**Conceptual Framework of Antecedents to Trends on Permanent Magnet Synchronous Generators for Wind Energy Conversion Systems**

**K. Padmanathan 1,\*, N. Kamalakannan 1, P. Sanjeevikumar 2,\*, F. Blaabjerg 3, J. B. Holm-Nielsen 2, G. Uma 4, R. Arul 5, R. Rajesh 6, A. Srinivasan <sup>7</sup> and J. Baskaran <sup>8</sup>**


Received: 1 April 2019; Accepted: 3 July 2019; Published: 8 July 2019

**Abstract:** Wind Energy Conversion System (WECS) plays an inevitable role across the world. WECS consist of many components and equipment's such as turbines, hub assembly, yaw mechanism, electrical machines; power electronics based power conditioning units, protection devices, rotor, blades, main shaft, gear-box, mainframe, transmission systems and etc. These machinery and devices technologies have been developed on gradually and steadily. The electrical machine used to convert mechanical rotational energy into electrical energy is the core of any WECS. Many electrical machines (generator) has been used in WECS, among the generators the Permanent Magnet Synchronous Generators (PMSGs) have gained special focus, been connected with wind farms to become the most desirable due to its enhanced efficiency in power conversion from wind energy turbine. This article provides a review of literatures and highlights the updates, progresses, and revolutionary trends observed in WECS-based PMSGs. The study also compares the geared and direct-driven conversion systems. Further, the classifications of electrical machines that are utilized in WECS are also discussed. The literature review covers the analysis of design aspects by taking various topologies of PMSGs into consideration. In the final sections, the PMSGs are reviewed and compared for further investigations. This review article predominantly emphasizes the conceptual framework that shed insights on the research challenges present in conducting the proposed works such as analysis, suitability, design, and control of PMSGs for WECS.

**Keywords:** permanent magnet synchronous generators; wind energy conversion system; finite element analysis; soft computing techniques.

### **1. Introduction**

Energy is predominantly the driving factor of human life and the economy of global countries. Henceforth, the research investigation in this area is highly critical and the need lot of time to invest

for in-depth study [1,2]. Due to the fast depletion of the natural conventional resources, sustainable alternative energy sources, for instance tidal wave, solar, wind, biogas/biomass and hydro energy, must be tap together for developmental activities. Therefore, there is currently a tremendous increase in the lookout for sustainable and alternative energy sources to generate electricity. Wind energy seems to be a promising and potential alternative renewable energy source with its enhanced sustainability and eco-friendly nature. According to 'Global Energy Outlook and the Increasing Role of India', in the year 2040, the electricity generation capacity of India will be equivalent to what is produced by today's European Union [3]. Figure 1 shows a summary of electricity generation by selected region and its electricity generation by 2040. The Global Wind Report (GWP, 2018) mentioned that the wind energy is one of the cheapest forms of electricity in a number of markets. Has it is a cost-effective option for countries which have ever-growing power demands and distribution challenges with centralized grid system [3].

**Figure 1.** Electricity generation by selected region up to 2040. Source: International Energy Agency [3].

The Global Wind Energy Council (GWEC) suggested that wind energy sector (both the on-shore and off-shore) supplies 300 GW of wind power capacity to come online by 2024 for global consumption. The global wind energy capacity increased with 51.3 GW in 2018. In spite of the fact, it is less than 2017 in about 4.0%; it is still a good achievement in wind energy capacity addition. From the year 2014, there is a 50 GW capacity addition occurring for every year though some markets behave differently. Thus, wind energy may contribute to electricity generation in India about 34,046 MW, which was 49.3) compared with all other renewable energy mix in the end of year 2018. By the year 2030, the wind power capacity is expected to generate 2300 GW power, fulfilling 22% of the global electricity demands. The report published by Global Wind Energy Outlook 2018 [4] predicted the future of the wind energy industry until 2050. In 2018, 50,100 MW was added, which was lesser than that of the 2017's capacity addition 52,552 M). It is viewed in 2018 as the consecutive year with increased new installations accounting to 9.1%, but this is lesser than the previous year's data i.e., 10.8% growth in 2017. The global electricity demand met by 6% of the wind turbines installed in 2018. In Figure 2, the cumulative production based on wind sources for the year 2018 shown along with the newly added capacity for the year 2018 [4].

**Figure 2.** Cumulative installed capacity of wind energy in the world end-of-year by 2018 and newly added capacity by different country in 2018 [4].

Figure 3 presents the overall baseline information of various settings, such as the new polices, moderate, and advanced scenarios. A global status report, published at the end of 2018, reported that global installed wind capacity was approximately 590 GW, which meant that Asia topped the regional market scale for the 9th consecutive year. It accounts for a whopping 48% of the added capacity (a total that exceeds 235 GW by the end of the year 2019) followed by Europe (over 30%), North America (14%), and Latin America and the Caribbean (almost 6%). In case of new installations, China retained the top position, though there was a contraction for two years. This was followed by US, Germany, UK, and India in respective positions.

**Figure 3.** Global total breakdown of cumulative capacity up to 2030. Source: Global Wind Energy Outlook [4].

Globally, the energy demands were 282.5 GW and 318.105 GW in the years 2012 and 2013, respectively. This denotes that there was a strong market growth of more than 19% and 12.5% in the years 2012 and 2013, respectively. However, this seems to be the lowest growth rate i.e., 22% and 21% of global electricity, when compared with annual average growth rate in the past decade. This is predicted to increase in the range of 8%–12% by the year 2020. The wind penetration level increased up to 10% in the year 2016, in alignment with the guidelines for international agreements on environmental commitment. By the years 2030 to 2035, the predicted saturation level is about 1.9 <sup>×</sup> 10<sup>9</sup> kW. The work by International Renewable Energy Agency (IRENA) titled 'Global energy transformation: A roadmap to 2050 (2019 edition)' inferred that by the year 2050, electricity would be the central energy carrier with growth up to 50% share from its current 20% share on final consumption. This would make the consumption of gross electricity double. The power demand across the globe (accounting to 86%) will be met by renewable resources-based power. Overall, the final energy will have two-thirds of contribution from renewable energy [5]. According to the literature [6], the current study focuses on the hypothesis subjects such as Wind Energy Conversion System (WECS) history, transformation of Permanent Magnet Synchronous Generators (PMSG), Finite Element Method (FEM) leveraging, Soft Computing (SC) applications, and the upgradation of Computer Aided Design (CAD) which looks to be a novel perspective as the first step. Generally, the wind turbine is moved by the wind pressure as in step-like method, though its design is different. In wind energy production, low (cut-in) and abundant (cut-out) wind speeds are labelled as risk potentials. On the basis of size and design parameters, the risk potential of every turbine is decided. Generally, the electricity yield of a wind turbine ranges from 3 to 25 m/s whereas high generation is examined once it crosses 10–15 m/s values. Each turbine has cut-in as well as cut-out values that are contingent on size as well as design parameters [7]. Therefore, the wind turbine design plays an important role in energy production. Dai et al. (2019) stressed that, in recent years, the incorporation of wind turbine generators, such as Permanent Magnet Synchronous Generator (PMSG), and Doubly Fed Induction Generator (DFIG), in which the former is predominantly utilized in wind energy conversion system's has been commonly seen, since it is cost-effective, highly reliable, and has flexibility in control [7]. This paper aims to address the technical issues and fitness of WECS components and integration with electrical grid. Furthermore, it will explore the study of PMSG comprehensive comparisons with other topologies of generator. In addition, this paper will also shed insights on the gaps in research and areas to further enhance research, in the context of WECS.

### **2. A Brief Review of WECS**

In 2004 article discussed wind engineering in general and wind power meteorology with special reference to turbine and generator technology. Further, they discussed the economics, which are involved in this regard [1]. In a study conducted in 2007, the researchers stressed that the conversion of wind electricity is currently a green technology factor due to (1) structural design improvements, (2) design and manufacturing of blades, and (3) efficient power processing techniques, on the bases of power-electronics followed by new generator design, to achieve variable-speed operation [8]. In 2013, [9] discussed a list of possible changes in the methodology towards the implementation of utility-scale wind energy into the power grid and follow up in accordance to the updated research with their obtainable alleviation techniques. Figure 4 disseminates the growth in size of wind turbines since 1980 and for predicted future prospects. The scaling up of turbines to lower cost has been effective so far, but it is not clear that the trend can continue forever [10].

**Figure 4.** Growth in size of wind turbines since 1980 and future prospects [10].

In 2012, [11] developed a 5 MW baseline design in deep wind concept with more than 150 deep Darrieus-type floating wind turbine systems. In this research article, the technology used in previous works employing various generator types and manufacturers of large power direct drive wind turbines were detailed. In Figure 4, the developments that occurred in the tower, blades, rotor diameter, power rating, and wind turbine hubs heights are illustrated. Amongst the available turbines, the 7.5 MW turbine seems to be the most powerful one with a 126 m rotor diameter. The global wind report published in 2012 cited the new Alston Haliade 6 MW turbine to be the world's large turbine with a 150.8 m rotor diameter [12]. In the future, the next-generation wind turbines are predicted to hold 20,000 kW capacity with a 250 m rotor diameter.

In 2010, [13] investigated the power output density functions of different WECS for a variety of operating wind regimes with the help of a probabilistic approach. In 2007, [14] conducted a review of information regarding global wind energy scenarios, performance, and stability of wind turbines, sizes of wind turbine, wake effects, evaluation of wind resourced, site selection, wind turbine aerodynamics, and challenges faced in wind turbines followed by wind turbine technology. Which is inclusive of control system, design, loads, blade behavior, generators, transformers, and grid connection. In 2014, a review of notable technical as well as environmental impacts of wind farms, wind power resource assessment techniques, control strategies, and grid integration techniques, were conducted [15]. A comparative investigation was conducted using a Maximum Power Point Tracking (MPPT) control

device in 2009 [16] between the optimized configurations of passive wind turbine generators with that of the active ones that operate at optimal wind power.

### **3. Wind Turbine, Types, and Generator Technologies**

In the past decade, there has been a tremendous growth observed in wind turbine technologies and that have resulted in the development of new-age wind turbine concepts. With developments in wind generator systems, cost-effectiveness of the systems has become the new mandate. In a wind power generator system, there is a tower which supports rotating as well as the stationary parts. The nacelle that has the generator in it, power converter, grid side step-up transformer, monitoring and control equipment are present in the stationary part. In 2014, [17] developed a summary about compact and lightweight wind turbines along with the technical hindrances with special reference to Horizontal Axis Wind Turbines (HAWT). There are two broad categories of wind turbine technology at present; such has the HAWT and the Vertical Axis Wind Turbines (VAWT). The HAWT main rotor shaft rotates in alignment with the wind direction, whereas it is perpendicular to the ground, generator, transformer, converters, and other equipment in the case of the VAWT rotor shaft.

In HAWT, the nacelle is placed at the top position in the tower. The HAWT showcase better aerodynamic performance when compared to VAWT, due to which the former is largely deployed in large-sized offshore wind farms [17]. According to [18], there are approximately 8000 different components present in a typical wind turbine. This information is based on a RE power MM92 turbine with the blades' lengths being 45.3 m and the tower height being 100 m.

Figure 5 shows the major components in a wind turbine and the share of the overall wind energy system parts cost. A direct-drive radial flux permanent magnet generator was checked for its suitability [19] to act as a drive-train runner. FEM software was used to test the generator fitness, based on structural design (or in other terms the stability of the air-gap present between the rotor and the stator) as per PMSG. So as to deduce the differences in flux density and force along the periphery of the rotor. In this study, the researchers used a simple analytical model. Further, 2D magneto-static simulations where also used to check the validity of the analytical model by making use of FEM software carried out [19].

**Figure 5.** Main components of a wind turbine and their share of the overall cost [9].

According to the literature [20], the induction and synchronous generator models are general candidates used to convert wind energy to electrical energy. In 2009, [20] listed Danish wind power status and various topologies of other wind farm configurations. A classification was done by [21] to differentiate the wind turbine technology schemes. To be specific, the different categories are Full Rate Converter Wind Turbine (FRCWT), PMSG, Fixed Speed Wind Turbine-Squirrel Cage Induction Generator (FSWT-SCIG), Variable Speed Wind Turbine-Direct Drive Synchronous Generator (VSWT-DDSG), Squirrel Cage Induction Generator-Wind Turbine (SCIG-WT), Full Rate Converter Induction Generator (FRCIG), Direct Drive Synchronous Generator (DDSG), Variable Speed Wind Turbine-Doubly Fed Induction Generator (VSWT-DFIG), Squirrel Cage Induction Generator (SCIG), Fixed Speed Wind Turbine-Permanent Magnet Synchronous Generator (FSWT-PMSG), Fixed Speed Wind Turbine (FSWT), Doubly Fed Induction Generator (DFIG) and Variable Speed Wind Turbine-Full Rate Converter Induction Generator (VSWT-FRCIG) [21].

This segregation is done on the basis of power level, working principle, application type, and the usage in a number of commercial applications. The research and development in this area is still happening, and various novel configurations and advanced applications are in the testing stage. In 2006 compared different classification types and explained them in detail [22]. In general, based on the working principle, three electric generators are considered as main types: induction, synchronous machines. Parametric which are associated with magnetic anisotropy and permanent magnets. The study further mentioned that the parametric generators in most cases be called as doubly salient electric generators [22]. Since they are mostly equipped with doubly salient magnetic circuit structures. When classified according to the magnetic flux penetration, there are three types of permanent magnet generators present: transversal-flux, axial flux, and radial-flux machines [22].

Since the efficiency provided is better, most of the high-power direct-driven wind power applications prefer low-speed and high-torque PMSGs [23]. These are generally applied in a wide range of applications due to cost-effective Permanent Magnets (PM). According to the literature [23], Permanent Magnets can provide high-power densities, higher efficiency, and chances of compactness which eventually results in the reduction of turbine size. The advantages of Permanent Magnet generators are when it excludes the exciter field winding, slip rings, and brushes in association with the capability to self-excite making option, so as to achieve good efficiency as well as the high power factor. In a standalone system, the PMSG has overloading and full torque capability, a highly competitive feature, due to which it is unique when compared to other traditional electrical machines. The PMSG is capable of self-excitation, another exciting feature which makes it the best option for operating at higher power factors and efficiencies. Further, PM machines possess the ability of overloading and full torque at zero speed, as well as at lower speeds [24]. To be specific, the standalone power systems are utilized in the isolated areas. When compared with the traditional electrical machines, this is inevitably effective.

In 2009, [25] studied the prospective site matching of direct-drive wind turbine models on the basis of electromagnetic design optimization of PM generator systems. In this study, a three-phase radial-flux PM generator was developed with a back-to-back power convertor. The study had a total of 45 PM generator systems which were designed, optimized, and grouped as a collage of five-rated rotor speeds in the 10–30 rpm range and nine-power ratings in the range of 100 kW to 10 MW, respectively. Following this, the study also determined the rotor diameter and the rated wind speed of a direct-drive wind turbine under optimum PM generator on the basis of the maximum wind energy capture design principle. This study also calculated the Annual Energy Output (AEO) with the help of the Weibull density function. At last, at eight potential sites, the maximum AEO Per Cost (AEOPC) of the optimized wind generator systems was calculated along with yearly mean wind speeds ranging between 3 and 10 m/s [25].

In 2008, [26] developed a concept of Permanent Magnet Generators Design. In this study, the researcher discussed the geared as well as direct-driven PM generators. Further, they also classified the direct-driven PM generators and the researchers dealt with various topologies of design aspects and unique nature in PM generators [26]. In 2012, [27] conducted a techno-economic evaluation of the basic assembly and magnetic topographies of the Salient Pole Synchronous Machine and Permanent Magnet Synchronous Machine. The study also provided the economic analyses of the machines that accompanied wind turbines.

### **4. Various Aspects of Comparison for PMSG's**

The design of electrical machines is important for any kind of applications. The basic design of an electrical machine involves certain procedures and analytical strategies. For calculation of magnetic circuit, electrical circuit, efficiency, insulation type, number of slots/poles combinations, winding dimensions, cogging torque analysis, control strategies, usage of materials, cost of products, thermal and structural design of electric machines, and manufacturing techniques etc. Finite Element Analysis (FEA) software can provide support for design and optimization tools to determine the best performance parameters. In 2008, [28] elaborately briefed and further used a deterministic global mathematical optimization which became a vital tool in the processes of design. Several mathematical models and optimization techniques could handle such problems associated with multi-faceted design. Figure 6 describes a complex range of ideas and significances of parameters for electrical machine design, analysis and characteristics studies, it has been simplified with partial adoption [28]. The studies conducted so far in this research areas, and various viewpoints have been established [28].

**Figure 6.** Electrical machine design parameters for analysis and characteristics studies.

In 2012, [29] conducted a general, as well as magnetic, analysis of various parameters, such as size, topology, voltage, magnetic field air-gap flux, weight, torque, losses, and efficiency between Permanent Magnet Synchronous Machines (PMSMs) and Conventional Salient Pole Synchronous Machine (CSPSMs) with the help of FEM. Figure 7, the weights of active material and costs are compared, and analyzed. Based on the comparison, it is observed that the total weight of the active material in the PMSM is reduced by 6.55% more than the conventional salient pole machine. In Figure 8, the losses at full load are presented [27].

With the same output power generated by the Permanent Magnet used in the machine, there will be reduction in machine weight which eventually becomes lighter to produce and so it increases the efficiency. Once the investigation was complete, it was observed that the CSPSM expressed less efficiency when compared to PMSM's. Further, when it comes to enhancement of magnet and semi-conductor expertise, the PMSMs reaped a cost-based benefit. Therefore, at the time of designing electrical machines, it is advised to follow their strategy in terms of machine efficiency and efficient use of energy [29].

**Figure 7.** Active material weights and Cost comparison of PMSM and conventional machines [29].

**Figure 8.** Losses comparison for PMSM and conventional machines at full load conditions [27].

A seven type of systems such as variable-speed constant frequency (VSCF) wind generator system, PMSGDD, PMSG1G, PMSG3G, DFIG3G, DFIG1G, EESG\_DD (Electricity-Excited Synchronous Generator with direct-driven), and SCIG\_3G (Squirrel Cage Induction Generator with three-stage gearbox) has been compared. In this comparative study, the researcher made optimization designs for different wind generator systems in the range of 0.75, 1.5, 3.0, 5.0, and 10 MW [30,31]. The results inferred that the PMSG\_DD was cost-effective when compared to EESG\_DD systems due to the cost incurred in lower generator system and enhanced Annual Energy Production (AEP) per cost. When there is an increase in wind turbine, the cost spent on direct-drive wind generator seems to be reduced. However, when there is an increase in the rated power, there is an enhanced performance exhibited by the PMSG\_DD system when compared to the EESG\_DD system.

Following is the description for a single-stage gearbox drive train concept. Due to the low-cost generator system and high AEP per cost, the focus shifted to the DFIG\_1G system which seems to be the best alternative. Further, when viewed from AEP per cost perspective, the DFIG\_1G system seems to be the most cost-effective and is close to 1.5 MW. Following is the concept behind three-stage behavior drive-train. Due to the least cost generator system and high AEP per cost, the DFIG\_3G system was considered as the best solution among other three wind generator systems. Additionally, in terms of AEP per cost aspect, more emphasis is given to the PMSG\_3G system compared to the SCIG\_3G system [31]. Figure 9 compares all five various wind generator systems of respective manufacturers in a wide range of aspects.


**Figure 9.** Comparison of five different wind generator systems [31].

When compared in terms of cost between a multi-hybrid PM wind generator system loaded in single-stage behavior and the direct-drive concept, the former seems to be cost-effective. When there is an increase in the size of the wind turbine, then the adoption of gear ratios may also widely vary. Based on the rated power levels, the optimum gear ratio may vary from 4:1 to 10:1. In the case of larger power ratings, the literature [17] suggests making use of higher gear ratios would be better performance. In 2014 mentioned that PMSGs are predominantly employed by giants such as the manufactures as follows GE energy, Vestas, Siemens, Gamesa and Goldwind. The stator of the PMSG is wound where the rotor is present with the PM pole system and may possess salient cylindrical poles. At most of the time, the low-speed synchronous machines project the salient-poly type with predominantly numerous poles. One can develop a direct drive system based on a synchronous generator with an ideal number of poles (a multi-pole PMSG). Some common types are transversal flux machine, axial flux machine, and the radial flux machine. The PMSG machine expressed highest the efficiency in an induction machine since the excitation was supplied excluding any energy flow. However, it is difficult to manufacture the PMs, whereas its inventory is cost-consuming too [17].

The long-term-unaddressed issue comes with the mandate to maintain the rotor temperature less than the magnet's threshold temperature. This may further be influenced by the magnetic material's Curie point and the binding material's thermal criterion in the case of power metallurgy composites. In turn, the synchronous process generates the issue according to the start-up, synchronization, and voltage regulation [32]. In 2011, Sandra Eriksson et al. performed an excellent comparison of direct-driven PMSGs. A total of six different-range generators were compared among each other [33]. Figure 10 clearly depicts the considerations of various factors with respect to the fixed and variable parameters for different ranges of generators.

**Figure 10.** Characteristics of rated speed and power for stationary simulations [33].

In 2013, [34] compared three configurations such as gearless-drive Permanent magnet induction generator PMIG-WECS, gearless-drive PMSG, geared-drive squirrel-cage induction generator (SCIG), and in the index every system was allocated with a number such as 1, 2, and 3 to position itself in the rank in accordance to other two systems. According to Table 1, the geared-SCIG system seems to be prominent in 61.5% of the indexes, while at the same time 38.5% of indices where dominated by the gearless-PMSG system. There was a 60% similarity in advantages between gearless-PMIG and gearless-PMSG. Therefore, the geared-SCIG system exists in alignment with the number of indexes. However, there is a domination of gearless-PMSG in the three top priority indexes such as generation efficiency, Operation & Maintenance (O&M) cost, and the duration of failure behavior. Further, there was a domination of geared-SCIG in the four top priority indexes such as kWh production at low speed, frequency of failure, generator O&M cost, and capital cost. In order to achieve the results with best accuracy, the weight of an index should be considered as per the order. Among the different configurations considered for the study, the results concluded that the gearless-drive PMSG-based and geared-drive SCIG-based systems seem to be the most desirable solutions. From Table 1, it is identified that the gear less PMSG is the only machine, which has the best option in efficiency, as there is no gearbox and copper loss [34].



In 2010, [35] with the field-circuit method for rapid calculation of load characteristics for stand-alone PM synchronous generators (PMSGs) that were developed with various rotor structures. The study results were compared with load characteristic calculations and results. The field-circuit method was defined, and utilized to determine the load characteristics of PMSGs with surface-mounted, inset or interior mounted permanent magnets and with inner or outer rotors [35]. In a comparative study conducted 2013, two PM generator types such as radial flux PM (RFPMG) and axial flux PM (AFPMG) generators were compared. To compare the generator performance during mechanical energy storage, the study measured the output powers of both RFPMG and AFPMG [36]. Results shown in Figures 11 and 12, concludes that there was a better performance exhibited by RFPMG when the machine's electrical parameters were in very similar condition and in relatively small power applications. It was inferring that the RFPMG has fewer copper, core, and rotor losses with respect to the varying generator and wind speed when compared to AFPMG [36].

**Figure 11.** Measured generator efficiency comparison [34].

**Figure 12.** Electromagnetic loss according to generator speed [33]. R: Radial Flux PM Generator, A: Axial Flux PM Generator.

### **5. Di**ff**erent Design Perspectives**

Designing PMSG has several challenges, which make it complicated when compared to conventional machine design procedure. The combination called 'Slot and Pole' poses various other challenges, which include reducing eddy current losses and cogging on permanent magnets. In 2013, [37] a technique was proposed to improve the air-power gap apparently transferred under the constraint of tangential stress using analytical optimization algorithms. The processes of optimization have been optimized for expressions that are relevant for the design of main variables, external derivations, and operational restrictions for the formulation of mathematical derivations.

In general, terms, during PMSG design, the optimum design includes various mandatory requisites, which are to improve profitability, and mitigate utilization of material to reduce cost and weight [38]. In addition, the design considerations should also take into account availability, high reliability, and low serviceability and maintainability for TC Ia that is wind class [38]. Furthermore, the utilization of gearless or semi-geared drive machines improves efficiency and reliability of wind power generators. Additionally, such requisites are associated characterization of compactness in terms of weight and dimensions. In addition, during the design of PMSG, the mechanical forces and voltage waveforms are quite imperative in several applications [38].

The design of machines is generally concerned with the electric and magnetic circuits; however, there are several losses which are measured using empirical equations [39]. In 2011, [39] explored the various design aspects concerned with the radial and axial field of synchronous machines with permanent magnets. In addition, the analysis of three fractional-slot and concentrated winding permanent magnet synchronous machine topologies are suited especially for specific applications [39]. According to a study [40], which explored the performance of wind power generators fitted with external permanent magnet rotors. The authors analyzed the FEM and electromagnetic results that examined the turbine characteristics and variations of the nominal wind speeds; various systematic methods were employed in previous research. For the calculation of the electrical characteristics, such as synchronous inductance, Electromotive force (EMF) constant, and phase resistance, an electromagnetic analytical and magnetic field distribution method was applied. In this study, a d-q model coordinate transformation theorem was employed for the analysis of performances. In addition, FEMs and curve fitting are used for the analysis of core losses [40]. Furthermore, a dissertation [41] presented a transformation theorem that developed a technique for the optimization and design of machines mounted with Surface Mount Permanent Magnet (SMPM), as impacted by mechanical loads, energy source, thermal effects, and state-of-the-art developments in manufacturing and material capabilities. A method was proposed for the design and development of cage rotor induction machines that can be optimized for better performance. Both genetic algorithm GA and particle swarm optimization (PSO) were used for optimization of the machines. Different integrated methods were applied and the Electromagnetic-Thermo-Mechanical method was used for the fabrication of Surface Mount Permanent Magnet (SMPM) machines [41]. An iron-less brushless permanent magnet machine was proposed and designed in 2013 [42] for the design and optimization of generator applications. The proposed approach constituted a dimensioning technique that involves comprehensive geometric techniques; both electrical and magnetic methods were used followed by the use of a detailed 3-D finite element (FE). In addition, the machine configurations used were both circular and rectangular designs, and were compared against each other. Furthermore, the performance of ironless stator designs configurations and the effectiveness of materials used were compared [42]. Tangential magnetic flux and stator concentric windings were incorporated in wind power generators in 2009 [43] with the rotation frequencies of 75–300 rpm. The parameters associated with the developed generators were depicted in the research. The intention of the previous research was to analyze the working of synchronous generators fitted with permanent magnets, which is in line with the concept of mitigating the problem of magnetic field distribution that was studied separately using FEM. During the development of such models, as given below in Tables 2 and 3 the following parameters to acquire synchronous machines should be considered and varied:

In addition, the following parameters should be considered for mathematical simulation.


**Table 2.** Parameters need to be considered for acquiring synchronous machines [43].

**Table 3.** Parameters need to be considered for mathematical simulation.


In 2012, [44] examined and designed PMSG using FEM simulation software that involves low speed three phase generators associated with external rotors. The aim of the research paper was to obtain sinusoidal voltages that are induced in stator windings which are espoused magnetization and arrangement path of permanent magnets within the rotor structure [44]. Again in 2012 [45], used the multi-physics approach for the design and development of a 10-MW doubly fed induction generator (DFIG). The optimal design and analyses were considered for the operation of direct drive of wind turbines with a conversion that has reduced size. In 2005, [45] performed a study that comprised of PMSGs that were used in wind power generation systems that are small. The output voltage was examined using FEM wherein both no-load and load conditions were considered. The influence of shapes and magnetic dimensions was examined. The previous research is a novel study wherein the outcomes of FEM were analyzed that revealed the PMSG's cogging torque frequency was influenced by number of poles and stator slots. However, the performance was influenced by factors such as magnet dimension, air-gap length, and cogging torque magnitude [46]. Research conducted by [47] (2008) depicted the design, prototyping, and analysis of relatively small and cheap axial-flux three-phase coreless permanent magnet generator. In the previous research, the FEM approach was used for the measurement of equivalent circuit inductances. In addition, the end winding inductance calculation and equivalent resistance of eddy-current loss where calculated using traditional methods. In 2002, [48] proposed a method for performance improvement using soft magnetic composite inter poles in drive permanent magnet machines. Several factors such as suitable pole arc shapes, magnet dimensions' influence, material usage efficiency, and labor costs where considered. In 2011, [49] examined the design considerations of double rotor radial flux permanent-magnet wind generators in terms of the mechanical and electromagnetic non-overlap air-cored (ironless) stator windings. The developed model was examined using finite-element analysis. The results of the analysis revealed that the electromagnetic design determines the mass, cost, rotor yoke dimensions, and leakage flux paths. In 2012, [50] examined the axial flux PM generator performances using wind turbine characteristics and electromagnetic field. The analytical approach could mitigate the analysis time required when compared with the FEM that is three-dimensional, which could use for the calculation of performances in the preliminary design phase. In 2010, [51] proposed and developed an optimal design high-speed DC generation system that uses a slot-less PM machine. In the previous research, the researcher used

soft magnetic composite (SMC) stator yoke and a controlled rectified fitted to the stator winding [51]. In [52] (1997) further examined the multi pole PMSG with the radial field. PMSG machines have been used as direct-coupled grid-connected generators with ratings between 100 KW and 1 MW. However, the previous research revealed that the poles that are between 100 and 300 are found to render better performance in terms of efficiency and reactance. The stator and rotor section design present the suitable pole and power number. Standard ferrite magnet wedges are used in the rotor sections. The stator sections however are made up of E cores with a single rectangular coil in each core. The researcher also developed a lumped-parameter magnetic model that permits the calculation of machine parameters in a rapid manner [52]. In [53] (2007) examined the direct-coupled an Axial Flux PMSG (AFPMSG) that is appropriate for a wind turbine system. Furthermore, the researcher used horizontal-axis and vertical-axis wind turbine generator systems. FEM analysis was undertaken for the analysis of the AFPMSG magnetic flux density distribution. The results analyzed were compared with the proposed machine configuration wherein the voltage from the output line was found to be of sinusoidal pattern. AFPMSG design feasibility was confirmed using a prototype generator [53]. In 2010, [54] further displayed an Axial-Flux Permanent-Magnet Generator for Induction Heating Gensets whereas ([55], 1997) and ([56], 1994) proposed a straightforward approach for the design of brushless permanent-magnet machines; the results are supported by several analytical results. The main difference between sine wave and square wave motors are detailed and described in terms of EMF, self-inductance, flux density and so on. A stage by stage method is involved with the design of computer-aided systems which are elaborated in detail. The previous research detailed the information such as torque, shape, magnet poles and phases, slots, poles, teeth, energy and co-energy, magnetic circuit concepts, yokes, basic relationships, magnetic materials, flux linkage and inductance, influence of stator slots, tooth flux, back-EMF, need for the field analysis based design FEM, cogging torque, series and parallel connections, and loss modeling [56]. Though machines achieve infinitely, the core of the machine that operates under unsaturated conditions and deep rectangular slots are not appropriate and not suitable for the design of today electrical machines with non-linear materials. The machine's performance should be predicted with great accuracy to solve non-linear equations which is expressed in terms of the Magnetic vector potential. The irregular machine geometry confirmation makes the analytic method configuration challenging. Hence, there is a need to use appropriate field computation, and modeling techniques utilizing electromagnetic fields such as the energy minimization. Includes, differential/integral functions, variational method, discretization, shape functions, stiffness matrix, 1D and 2D planar and axial symmetry problem and computation of electric and magnetic field intensities, capacitance and inductance, force, torque, and energy for basic configurations of electrical machines [57].

In Figure 13, various electromagnetic analytical methods are illustrated. Every method contains a set of advantages as well as disadvantages. In this scenario, the finite elements were found to be robust in nature to conduct general electromagnetic analyses [57].

*Energies* **2019**, *12*, 2616

**Figure 13.** Different methods of electromagnetic analysis [57].

### **6. Consideration of Losses Calculation for PMSGs**

One of the important design factors discussed in this study is the determination of losses in PMSGs. In 2010, published a model with an elaborate loss computation and calculation method with updated analytical loss calculation. In this model, conventional losses, for instance, stator core iron losses, ventilation losses, I2R losses, and other detailed losses like stator end region losses, were discussed. However, being a separate engine, the cooling caused by the bearing friction and the losses incurred via excitation system were not considered. The components which were lost are discussed here in detail [58].

### *(a). Iron Losses*

Excluding the stator and rotor windings, there seemed to be losses in eddy current as well as some more losses in entire metallic parts, which can be segregated as follows.


### *(b). Winding Losses*

Various types of losses in stator, rotor, and damper windings are inclusive in winding losses

1. Stator winding copper I2R losses.


### *(c). Ventilation Losses*

Ventilation losses segmented further into following parts


In 2011, a study [59] of extreme excellence was conducted by on electromagnetic losses which were incurred in direct-driven PMSGs. Using electromagnetic model, the solutions were obtained from FEM. By utilizing a MATLAB-driven model, the researchers performed the simulations. The results obtained inferred that the iron and copper losses were completely based on the rated voltage and rated current. In terms of a fixed output power, the experiment achieved larger machine volume with an increase in rated voltage. Further, higher frequency and increased iron loss were observed in parallel to decrease rated current and reduce copper losses. At the time of simulations, the generator losses were determined for various wind speeds, using which the loss distribution was calculated. Furthermore, they tested an analytical model to predict the eddy current losses in PMSG rotor magnets by feeding a rectifier load. The eddy current loss achieved during time stepping resulted in the coherence of 2-D FEM and coupled-circuit when performing the investigation. In 1997, conducted an experiment and designed losses for the model of a 1 MW machine design prepared in alignment with the parasitic losses. These were stator back-iron reluctance, rotor and stator slotting, rotor reluctance, stator back-iron reluctance, stator module weld loss, rotor eddy-current loss, stator beam loss, the polygon effect, and stator structure cage loss [60]. In 2014, [61] experimented on eddy current losses in PMs of surface-mounted magnet synchronous machines. This study introduced a true analytical method on the basis of magneto-dynamic problem of a conductive ring. The results were obtained and compared with the information retrieved from 3-D FEM analysis. In the analytical model, the effect of the width on magnet loss was considered. The axial effect was considered via a correction coefficient. In the comparison executed, the researchers included impact of the circumferential segmentation, instantaneous losses, effect of the frequency on magnet losses, and induced current density. Through stressing the criticality of the skin effect and magnetic reaction due to magnet currents, this analytical model yielded an accurate measurement of magnet eddy current losses [61].

### **7. Faults and Protection**

At the time of designing PMSGs, researchers must be considered for the chance of fault occurrences and protection schemes methods. In 2013, [62] listed the influence of asymmetrical magnet faults upon PMSG rotors. Mechanical looseness, eccentricity, and damage in any one magnet are the most commonly found attributes that result in rotor faults. Further, the rotor eccentricity is caused by unequal distribution of static, dynamic, or mixed air-gap. In the presence of static eccentricity, the air-gap seems to be the least and positioned as per the stator. On the contrary, in the case of dynamic eccentricity, there seems to be no coincidence between the rotor's centers and the center of rotation. Therefore, the minimum air-gap position rotates in line with the rotor. There are notable reasons behind the cause of eccentricities such as looseness, incorrect assembly, load unbalances, misalignment, and sometimes the bending of the rotor. At the time of analysis, the study conducted series as well

as parallel-connected windings. In order to quantify the demagnetization in a single magnet, the study defined the faulty severity factor. As per the study investigations, one can conclude that, for a generator where all windings are series-connected, the induced EMF value gets decreased due to the demagnetization of a single magnet. Likewise, if the load is a resistive type, the current also may decrease. Therefore, one may not be able to identify the frequency components which are in association with the fault whereas one can observe only the decreased total flux linked to the windings [62]. In 2012, Rodrigues et al expressed his ideas on direct or indirect lightning strokes after thoroughly reviewing the over-voltages and electromagnetic transients [63]. The transient behavior can easily be explained via the lightning protection of the wind turbines accurately, for which the modified version of EMTP (Electro-Magnetic Transient Programme) was utilized. In this study, the researcher adopted a case study model in which two interconnected wind turbines were used so as to study the direct lightning stroke to the blade or the lightning strikes which happens in the soil near a building. Further, this study also conducted a holistic computer simulation in addition to EMTP-RV [63]. Investigation in 2011 [64] which evaluated the fault conditions and identified efficient fault ride-through and protections schemes in electrical systems of both small-scale (land) and large-scale (offshore) wind farms. In their study, the researcher considered two variable-speed generation systems such as PMSGs and DFIGs. After discussing the protection issues associated with DFIGs, the research proposed a new protection scheme as well. Following this proposal, the protection scheme options for fully rated converter and direct-driven PMSGs were analyzed and simulation results were compared.

The development in magnetic materials and its impact on the electric machine design investigated (2007) [65]. In addition to that, few potential faults were also selected using a fault-tolerant system design. Two fault types may occur in the system, of which the electromagnetic faults are as follows:


The power converter faults are listed herewith


One should focus on development of a fault-tolerant system, if the operation needs to be continued even in the presence of faults, if any. In this design, every phase should have a stand-alone single-phase PWM inverter that has a modular system in which the modules are isolated by every phase fault. When a module has less thermal interaction or electrical/magnetic interactions, then the system is likely to proceed with the operations excluding the faulty phase [65].

By 2012, inducted a rotor core design and FEA simulation, to diminish the mechanical stress put upon the core bridge. After considering rotor speed variations, the researcher performed the mechanical transient analysis. The experimental result was presented for the S-N curve (S-N curve is deduced from material test data) of rotor care material so as to assure the validity of the model against fatigue failure [66].

### **8. Damping and Oscillation**

In order to handle damping and oscillation, the PMSG-based stability issues in WECS should be taken into consideration. In 2011, a torque compensation strategy was devised [67] based on DC-link current determination of the converters, after the stability challenges faced in PMSG-WECS were studied. In general, the instability issue is caused at the time when generators are in direct connection with the wind turbine during which the speed oscillations occur because of the lack of damper in design, and torsional vibration. With the purpose of reducing the oscillation amplitude and enhancement of the

system stability, one can make use of generator torque controller. However, due to limited ability, it may impact the WECS' power response. The torque compensation strategy, when deployed with the sole purpose of enabling positive damping of the oscillations, may lead to enhancements in small signal and transient stabilities of the WECS [67]. In 2018, [68] opined the influence on system oscillations because of the grid-connected wind farms. The study focused on the contributions made by the damping of power system oscillations and the assistance rendered by the inner wind turbine oscillations upon the changes in several aspects of power system behavior, which is inclusive of stability. The stability of the power systems gets connected with electro-mechanical interactions and the generator's behavior which is already in connection with the grid. Therefore, the influence of wind power penetration over the power system becomes a key challenge to tackle. In the literature [68], an elaborate investigation was conducted about the oscillations in power systems and its influence and control schemes in the wind farms for various wind turbine technologies [68]. The growing technologies that focus on magnetic gears were the primary theme of the study conducted [69] (2011). The concept of magnetic gears has an advantage of dealing with inherent overload capability surpassing the mechanical gears. However, there is a less amount of torsional stiffness found in magnetic gears than their counterparts i.e., mechanical gears. This leads to oscillations at the time of transient changes in load and speed alike, and the damper windings utilized in synchronous generators to alleviate the oscillations that occur due to transients [69]. In a study conducted in 1996, the researchers display the damping of PMSG power-angle oscillations in terms of wind turbine applications [70]. The small pole pitch present in the generator allows it to work in every low speed and this is conjoined with the wind turbine, thus a direct electrical grid connection is maintained. In this research paper, an alternative damping system was proposed in which the stator is allowed to confine the rotational movement through a connection with the wind turbine that is located near a spring and mechanical damper. This proposed method enables high damping of power-angle oscillations when compared to conventional damper windings. The design's efficiency can be illustrated via the generator's response to initiate the changes in driving torque. In order to showcase the new design's vibrant nature and viability, the generator's behavior on synchronization as well as on operation front, where there is a difference in wind occurs, is described [70].

### *Short Circuit*

In 2011, [68] stressed the occurrence of sudden short circuit when applied in large PMG machines thus denoting the differences in short-circuit behavior amongst the would-field generator and the PMG. With the help of FEM analytics, the researcher calculated the sub-transient reactance and time constants of the PMG and utilized it at the typical circuit theory simulation in short-circuit fault. The FEM was then used in the risk evaluation of magnetization loss in magnets. Being complex, the transient magnetic field looks for transient non-linear circuit-coupled FEA in 3-D in association with voltage-source excitation. Various calculation methods where summarized in this research paper with further discussions on implications of futuristic design and PMG application after considering the attributes that are relevant to application of standard tests and specifications [71].

### **9. Several Aspects of Cost Factor**

In the study conducted by Salem Alshibani et al. (2014) [72], the high CAPEX (Capital Expenditure) issue was taken into account since, at the beginning of a project, it is always a hindrance for such techniques, especially in case of PMSGs. The study proposed a method, which utilized to assess typical PMSGs designed and reported in this article. The results of the proposed method were compared with the results of the traditional methods. The results inferred that the lifetime cycle assessment (LCA) seemed to favor the gearless PMSGs that incur high CAPEX. Further, in the case of inclusion of lifetime cost in the design optimization, the scenario develops machines which can yield significantly higher lifetime revenues than the extra CAPEX required [72]. Figures 14 and 15 compare the CAPEX values of geared as well as gearless PMSGs in a range of power ratings with percentage.

**Figure 14.** Depicted CAPEX (Capital Expenditure) comparison of geared and gearless PMSGs at a range of power ratings with percentage difference in cost shown at each power level [72].

**Figure 15.** Cost comparisons of the machines with lifetime losses cost added and gear cost calculated twice. The percentage difference in cost is shown at each power level [72].

To conclude, it can be inferred that higher power rating-based wind turbines are the most preferred ones in reducing the development and maintenance time and eventually increasing the energy yield [72].

### **10. Soft Computing Technique Based Optimization Used for PMSGs**

There are two critical issues that influence an electrical machine's optimal design considering the usage of FEM, the computation time from FEM simulation, and the different parameters concerned with the electrical machine. In the present day scenario, the use of soft computing techniques-based optimization has gained momentum owing to the use of the statistical analysis with multiple correlation coefficients and moving least squares (MLS) approximation as proposed (2007) which are compatible with the electrical machines [73]. In general parlance, the process of optimization includes several computations which are all dependent on parameters; the effort of computation is very minimal when compared to the time that is saved. Such a method is assessed by the same application to synchronous machine's optimal design. The results of such analysis reveal the increase in the torque per weight ratio by 13% when compared with the results that are acquired from traditional optimization techniques [73]. In 2010, [74] used the Fuzzy and FEM method for the analysis of the comparison that includes leakage field analysis witnessed in the electrical generator. The process of leakage field analysis is performed by developing a fuzzy model of the generator with the technology called adaptive neuro-fuzzy inference system (ANFIS). In this regard, the researcher performed a comparative evaluation on fuzzy model and FEM model wherein a good correlation was found to be present between them [74]. Furthermore, a study (2008) [75] revealed the new and novel approaches towards automating optimization processes that are manual, and examined the implementation obstacles that are witnessed by the engineering community. Based on the effort for design evaluation and the degrees of freedom viewpoints, engineering design optimization was subjected to classification. In the previous research, the researchers presented a holistic view on the various design optimization approaches. Furthermore, the major challenge witnessed was scalability for the techniques of design optimization considered in the study. Large-scale optimization requires effective algorithms such as swarm intelligence and a considerable computing power [75]. However, 2001, [76] proposed the use of a neural network in comparison with the Finite Element Technique (FET) based sensitivity analysis for the optimization of permanent magnet generators. In 2012, [77] further identified the challenges that were witnessed during design optimization for minimizing or maximizing the fitness function which positively influence the design purpose. Genetic algorithm which is incorporated in the optimization technique that is population based does not consider certain inferences, such as the magnet, copper, and magnetic laminations, and raw active materials. The main intention towards the reduction of the fitness function is based on the cost of energy that is generated by the system which further accounts for the variables that are uncertain in nature [77]. In 2009, [25] further explored the use of direct-drive PM wind generation system optimum design models wherein the PM was designed and developed using enhanced genetic algorithm with a PM generator fitted with 500 kW direct drive wherein the minimization of the active material cost tends to improve the design optimization effectiveness.

In 2009 used the concept of direct-drive PM wind generation system optimum design models in which the PM is developed using an improvised genetic algorithm along with a 500 kW direct drive PM generator; this actually reduces the cost of generator active material which further illustrates design optimization effectiveness [25]. Furthermore, [78] (2007) proposed a novel approach for the design of electrical rotating machines wherein a rational solution of predesign was done by integrating exact global optimization algorithms and analytical model. However, prior to developing an extensive prototype, validation of previous solutions should be performed using FEM. The purpose of the previous research was to extend the accurate global optimization algorithm through the introduction of an automatic numeric tool. Such a novel technique is used in resolving rationally the design problems. Furthermore, several examples were evaluated to examine the effectiveness of the novel technique [78]. In 2008, [79] further established a new hybrid machine with 36/24 pole outer rotor permanent magnet (PM) that is directly coupled with a wind power generator. For effective control of the flux control, two excitation (PMs and DC field windings) hybridization in the double-layer stator is utilized. This result in constant output with wide range of speeds and a load varying where examined. In 2001, [80] further used genetic algorithms wherein a new algorithm called orthogonal genetic algorithm along with quantization/quantification for global numerical optimization was used with continuous variables. Furthermore, a quantization technique and orthogonal design were used for the development of a new crossover operator; this crossover operator generates representative sample points which are small, however are a potential offspring. Such a proposed algorithm solves 15 benchmark problems with 30–100 dimensions belonging to the local minima [80]. It was 2005 arrived at new dimensions in this research area of evolutionary computation and structural design [81]. Furthermore, [82] (2008) examined soft computing (SC) techniques associated with the design of engineering concepts. Through the inspection of soft computing methods, techniques, and their competence, to further address the high complexity issues and design tasks, the researcher reviewed Fuzzy logic (FL), artificial neural networks (ANN), and Genetic Algorithms (GA) [82]. In 2012, [83] further made an overview to compare research that was conducted to optimize the parametrization of machining process of modern and conventional machining. Following are the most important techniques used: genetic algorithm (GA), particle swarm optimization (PSO), simulated annealing (SA), artificial bee colony (ABC) algorithm, and ant colony optimization (ACO). Amongst the aforementioned algorithms, GA is widely applied in the literature [83]. In 2004, [84] proposed a new solution called the multi-agent genetic algorithm (MAGA) which is an integration of the genetic algorithms and multi-agent systems to solve the problem

of global numerical optimization. In 2008, [85] further proposed a disagreement versus randomness in the various SC techniques. In 2011, [86] further reviewed the state-of-the-art research developments associated with the use of soft computing techniques used for the optimization of problems associated with design, planning, and control in the field of sustainable and renewable energy. Furthermore, several soft computing methods were reviewed and presented regarding the current state of the art in computational optimization methods applied to renewable and sustainable energy, wherein a vibrant visualization of the state-of-the-art research progresses was proposed [86]. It is important to generate random numbers using soft computing methods, as random numbers are used during the beginning of the estimation or during the processes of learning and searching. When compared between simultaneous randomness consideration and opposition and pure randomness, it was revealed that the former is better than the recent results acquired from evolutionary algorithms, neural networks, and reinforcement learning. To further increase the performance of soft computing algorithms, it was revealed that opposition-based learning provides an inclining effect. This was experimentally and mathematically proven that SC has better merits when applied to improve the differential evolution (DE) [86]. In 2010, [87] also presented the Genetic algorithm (GA) with memetic algorithm and MADS (Mesh Adaptive Direct Search) for the optimal design of an electric machine. To acquire an effective optimal design of an electric machine with longer computation time and many local optima, the previous research proposed a hybrid algorithm to acquire global optimum. To maximize further Annual Energy Production (AEP), the prospective algorithm was referred. By 2006, [88] classified the modelling and optimization techniques for process problems shows in Figure 16, which displays the conventional and non-conventional optimization techniques and tools used in this regard.

**Figure 16.** Conventional and non-conventional optimization tools and techniques [88].

To further conclude, it was deemed that MADS is combined with GA as an effective computation time reduction method for optimal PM wind generator design and is considered over other parallel computing methods [89]. Further offered a type of multidisciplinary design and optimization (MDO) of a diffuser for an incompressible and steady magneto-hydrodynamic (MHD) method. The design problem can be resolved using GA-based programme that is optimized with the FEM based MHD simulation technique for which least-square FEM was used and developed in later research [89]. In 2017, [90] presented about Multiple Criteria Decision Making (MCDM) concepts and has been used for economics analysis. Similarly, this concept can be used for lifecycle cost analysis of machine design [90]. In (2002) [91] presented the non-dominated sorting GA to mitigate the performance related problems wherein the performance was analyzed through the comparison of the results from the

other four algorithms. Further discussion was performed on the multi-objective optimization process solution using evolutionary algorithms, wherein the findings of the research revealed effectiveness against analytical and electro-magnetic problems [91]. Furthermore, [92] (2001) displayed an approach that was used to design PM for wind power applications wherein the approach was made up of two phases: preliminary design stage and optimization stage. In 2008, [93] further examined the use of Differential Evolution (DE) and Particle Swarm Optimization (POS) Algorithms with technical analysis. It was ascertained that the Artificial Bee Colony (ABC) algorithm could be used as an innovative swarm optimization algorithm with fine results of numerical optimization.

Furthermore, [94] (2011) proposed an enhanced algorithm called the fast mutation artificial bee colony algorithm (FMABC). In 2012 proposed an improved ABC algorithm, which was used to solve numerical optimization issues, which further improved the capability of the ABC algorithm's exploitation feature. An alternate search mechanism and a varying probability function were proposed by the previous researcher. Seven numerical optimization problems were tested on the enhanced ABC algorithm [95].

In 2012, further utilized genetic algorithm (GA) for the achievement of an optimal design for an axial-flux PMSG (AFPMSG) [96]. In 2009, [97] proposed an approach based on a numerical optimization algorithm wherein a generalized receding horizon control of fuzzy systems was proposed. To further resolve generic fuzzy dynamic systems' optimal control problem, a numerical method was developed. Fine optimization was developed in the previous research.

The researcher made a thoughtful understanding of soft computation techniques in the electrical engineering field applications, with integrated pseudo-code operational summaries [98]. In 2010, [99] considered population-based algorithm and its application to solve numerical optimization problems. In certain cases, there are complexities in computing search problems which is associated with high dimensionality of search spaces. Until there is an employment of appropriate approaches, a search process could reduce effectiveness and increase cost. The use of nature inspired algorithms could tackle such difficulties. For example, fish schools tend to increase the mutual survivability since a large number of constituent individuals are deployed.

In 2008, first to introduce a method that searches high dimensional spaces that consider account behaviors that are obtained from fish schools. The derived algorithm—Fish-School Search (FSS) was made up of three operators: feeding, breeding, and swimming. In a cumulative scale, these operators tend to afford the evoked computation: (i) wide-ranging search abilities, (ii) automatic capability to switch between exploitation and exploration, and (iii) self-adaptable search process for global guidance [100].

S.L. Ho et al. (2006) examined the use of particle swarm optimization (PSO) methods wherein the previous research considered several variables such as age; new strategies were figured out to examine the optimum particle solutions, the original formula for velocity updating, and intensified search phase integration with enhanced PSO method. The findings of the previous research revealed that the proposed method contains a refined ability to perform a pinpointing search and the overall global ability improved when compared to traditional PSOs [101].

It was [102] offered the use of support vector machine (SVM) classifier for the detection of broken electrical induction machines. Furthermore, the previous researchers also considered the analysis of Gaussian, linear and quadratic kernel function as opposed to the error rate and the support vector numbers. The findings of the previous researchers revealed the successful detection of broken bars in different situations wherein there also evidences fast, precise, and robust load changes which tend to qualify for the right use of such techniques in real-time online applications in industrial drives.

Furthermore, in 2002 proposed a tabu-search algorithm to identify multi objective optimal design problems' pareto solutions from which there is a utilization of the contact algorithm to assess the previous aspects. During the initiation of the iteration cycle, identification of the new current points, fitness sharing function, and ranking selection approaches are introduced. A more detailed explanation of the numerical results is displayed in the previous research to highlight the power of the proposed

algorithm to ensure that there is uniform sampling performed which yields Pareto optimal front of the multi-objective design problems. Furthermore, effective execution strategies for the proposed algorithms were also displayed [103].

In 2011, further displayed the Improved Discrete Particle Swarm Optimization (IDPSO) searching technique, which is applied on the head of an electromagnet and for the optimization of the magnetic field gradient. For the previous research, COMSOL software was used for the measurement of the magnetic forces and field. The aim of the optimization algorithm is the search of optimal pole shape geometry in a refined manner, which results in the distribution of the homogeneous magnetic field with the desired holding force in the specific area of interest [104].

Furthermore, [105] (2007) displayed an innovative recursive fuzzy logic categorizing (R-FL-C) strategy for the PM generators design approach that is utilized to mitigate search space and for expelling the local minima in due course of the process of optimization. In the previous research, finite element state space models are used to examine the space database with the knowledge that is acquired from off-line.

In 2012, [106] assessed the numerical functional optimization wherein the use of artificial bee colony optimization led the researcher to derive the use of the same bee swarm foraging behavior in their approaches. Furthermore, the ABS's efficacy was found to be high when compared with the genetic algorithm (GA), ant colony optimization (ACO), and the Particle swarm optimization (PSO). Though the ABC technique is found to be pretty important and efficient during exploration, the capacities associated with exploitation are found to be poor with issues regarding convergence speed in several instances. To mitigate this, the researcher further introduced the improved ABC algorithm or the I-ABC, which during the process of search with refining used the acceleration and inertia weight as the fitness functions.

In addition, [107] (2012) provided a heuristic structural optimization for the Surface Mounted PMSG. The use of structural optimization is the process of identifying the material distribution in an optimal way in every machine part; this technique is very prevalent in the field of mechanical engineering. Similarly, the use of structural optimization can also be witnessed in the field of electrical engineering. When compared to the other methods reported to deploy the continuous models for the elaboration of the material properties with Heuristic Search Algorithm [107], it gives a solution to the structural optimization issues.

By 2019, [108] proposed an identification method on K-means-singular value decomposition and least squares support vector machine which the simulations were proposed for voltage sags based upon an annealing algorithm for multi-objective optimization. To gain the pareto solutions in a significant manner. This is completely dependent on Pareto and can successfully be introduced in addition to parameter and objective space strings. The novel method proposed in this study questions the stop criterion, new rank formula, fitness sharing functions, and other such enhancements. For the purpose of validating, the proposed method's robustness, the study validated two numerical examples [108].

In 2001, [109] proposed an enhanced tabu-search algorithm to practically applied, it when finding optimal designs for electromagnetic devices. In parallel, the study also conducted team workshops and mathematical test functions. Based on the numerical results, it was inferred that there is a less significant iteration number achieved for the proposed method when compared with simulated annealing and other such algorithms.

In 2008, proposed a novel methodology with reference to PSO in order to find out the parametrically non-linear model structure. In this study, an existing method used in PMSM's dq-model to identify the parameters. Both the disturbed load torque as well as the motor stator resistance was established for PMSM variable-frequency drive system application. In order to question the efficiency of the identification method, the study conducted a simulation and the experimental results were provided. The results inferred excellent precision in terms of time-varying parameters when the PSO algorithm was used [110].

In the study conducted 2000), an auto-learning simulated annealing algorithm was proposed. This algorithm was developed by collaborating simulation annealing as well as the characteristics of the domain elimination method. This study utilized the standard mathematical function to assess the algorithm in addition to optimization of the power transformer practical end region [111].

In 2005, [112] demonstrated single as well as multi-objective optimizations by experimenting with a PMSM with rotor feedback with the help of a Genetic Algorithm. This artefact's extensions are nothing but the implementation of core losses cited with the help of the Steinmetz approach. A few other up-front changes are the modifications in tooth shape (especially the base), addition of voltage drawbacks, and changes in the volume expression for addition of end turns.

In the study conducted by [113] (2005), an improved Ant Colony Optimization Algorithm was proposed to be used in Electro-Magnetic Device Designs. The experiment deployed the algorithm in an inverse problem along with a mathematical function where its performance was contrasted with other better-designed methods.

A comparative study was conducted (2007) between the performance of ABCs upon the optimization of numerical function with swarm intelligence and population-based algorithms such as PSO, GA, and Particle-Swarm Inspired Evolutionary Algorithm (PS-EA) [114]. In order to explore the performance of the ABC, a total of five high dimensional benchmark functions that consisted of multi-modality were deployed. From the simulation results, the authors made a strong recommendation that the proposed algorithm is capable of expelling local minimum and can be used well in multi-variable multi-modal function optimization. The scope for future researchers in this study was the investigation of influence exerted by the control parameters in the convergence speed and performance of ABC [114].

In 2009, a comparative study conducted to assess the performance of ABC algorithm with Evolution Strategy (ES), DE, GA, and PSO using a large set of unconstrained test functions. From the results, it was concluded that there was an excellent performance exhibited by the ABC algorithm when compared to other algorithms, though the study made use of only less-control parameters thereby efficiently solving multi-dimensional as well as multi-model optimization problems [115]. The results further inferred that the performance of the ABC algorithm is superior compared to other such algorithms.

A beneficial design procedure was proposed (2012) for the controller utilized in the frequency converter of a variable speed wind turbine (VSWT)-driven PMSG with GA and RSM [116]. A mess-less technique was recommend by the study conducted in 2004, which focused on connecting the radial basis functions (RBFs) as well as wavelets. This new method proposed in this study leveraged the advantages of RBFs as well as the wavelets. In order to maintain the linear independence as well as consistency, the bridging scales were utilized so as to safeguard the mathematical properties. With the purpose of validating the proposed method, a numerical example was utilized [117].

A hybrid Genetic Algorithm (GA) was proposed in 2003 [118], in order to optimize the electromagnetic topology. After taking a 2-D encoding technique into account, the geometrical topology was at first applied to electromagnetic topology. In the later stages for the crossover operator, the study utilized a 2-D geographic crossover. In order to enhance the convergence features, the study used a novel local optimization algorithm, otherwise called an on/off sensitivity method, which is hybridized with 2-D encoded GA. Once the algorithm was verified with different case studies, the results were published [118].

### **11. Novel Topology Development in PMSGs**

In 2012 stated the assessment of low maintenance slip-synchronous, PM wind generator, which was developed using the concept of PM induction generator [119]. In 1926 introduced the PMIG (Permanent Magnet Induction Generator) concept upon which the slip-synchronous permanent magnet generator (SS-PMG) was constructed. In generator design, there exists an induction machine cage-rotor, traditional stator winding along with an add-on of second free-rotating PM-rotor. The second PM-rotor runs synchronous speed while the cage-rotor operates at a relative slip speed in accordance to the PM

rotor and rotating synchronous stator field. This is a gearless wind turbine generator that is connected with the grid directly i.e., no power electronic convertor or such behavior is required in the drive train. In the summary developed by [17] (2014), a comparison was performed between large-sized wind turbines which can produce more electricity at less cost with small-sized turbines. This comparison was executed since the costs involved in experimental set-up and maintenance do not impact the size of the machine. Therefore, more than 7 MW output power is being achieved from today's wind generators. For instance, from 2011, Enercon manufacturing an E-126/7500 wind turbine with 7.5 MW power capacity. At present, Sway Turbine and Windtec Solutions are in the process of developing 10 MW wind turbine generators which might hit the commercial markets in 2015 [17]. Figure 17 shows the voltage ratings of seven various models of common wind turbine generators with respect to the turbine power which clearly depicts the model performance.

**Figure 17.** Summary of the voltage ratings of a few common wind turbine generators [15].

An innovative model with a Surface-Inserted Permanent Magnets Synchronous Generator was proposed in 2011, with air slots in the rotor that can be adjusted. This model removes the disadvantage present in PMSGs i.e. fluctuation of regulating voltage. When a comparison was performed between conventional machines and superconducting machines, it was found that the latter exhibited novel advantages such as efficiency, compactness, lightweight and significant stable operation in power systems [120].

In 2007 proposed an eccentricity topology with a promise to enhance the power density and made use of it in the design, development, and testing on an eight-pole superconducting rotating machine. Further, the study discussed the results retrieved from the magnetic scalar potential from a Coulomb formulation by Markov Chain Monte Carlo (MCMC) method. Additionally, the flux density was calculated using derivation from the regularization method. With the purpose of reducing the computation time, the MCMC method was deployed which in turn perform the magnetic scalar calculations in specific regions of discrete geometry. By using YBaCuO high-temperature superconducting (HTS) bulk plates and low temperature superconducting NbTi wires, a high magnetic field was generated. In order to increase the cooling operation, there is a stationary superconducting inductor and a rotating armature coiled with copper wires present in the superconducting machine [121].

A detailed differentiation study was conducted [122] (2012) on the differences in development and settlement of active materials for transversal-flux machines from radial and axial ones. Lower stator copper losses were gained by increased windings space in the absence of any impact from the available space for flux in the transversal flux. As the electromagnetic structure is sophisticated, the transversal-flux machines seemed to be costly [122].

A novel low-fare methodology was proposed in 2012 to develop wind turbine electric generators from the generator from the burnt-out squirrel cage induction motors. The author first detailed the list of properties generally required for a wind turbine generator following which the methodology described the PMG, workability, multi-pole, and low-speed. The study conducted a cost comparative analysis and performance comparative analysis based on the test results achieved from a 500 W generator run at 900 RPM and a 1500 W generator at 650 RPM [123].

The efficiency of an air-cored PMSG was estimated in the study conducted [124] (2011) using finite elements and equivalent circuit modelling. The emerging trends showcase that the air-cored machines are predominantly used in wind energy systems. Instead of iron, the magnets which are captivated between the mild-steel-based rotors are present. At zero-load, the two-sided, axial-flux, air-cored machine's flux path can be seen as a stable magnetic flux that crosses axially from a magnet on one rotor to the opposite rotor which is a facing magnet. Further, the study stated that the coil is held by the stator on a plane in the middle of two magnetic sets [124].

In 2012 [125], an alternative viable solution was proposed the traditional PMSGs at MW level in direct-drive wind turbine applications via a Halbach array. It is a must to optimize the machine dimensions in order to achieve the maximum benefit of the Halbach array. This research article provides an overview of calculating the Halbach array application using analytical equations which are prevalent in the studies published earlier. The study recorded extraordinary performance by making few modifications in the existing PMSG design in which a constant magnet volume is maintained. When compared, the conventional array seemed to be more valued than the Halbach array at the time of considering the critical rotor radius. When the number of poles were increased, the critical radius got shifted to larger sizes and thus it allowed a positive leverage of the Halbach array at MW level. The analytical equation findings were verified using FEA simulation [125].

In 2008 [126], Halbach magnet array with the help of the numerical optimization method, which in turn relied upon finite element analysis. The magnetization direction of every element was designated as the design variable. In order to enhance the repulsive, attractive, and tangential magnetic forces present between the magnetic layers, the researcher investigated the optimal magnet arrays composed of two and three linear magnet layers. Two and three magnet rings altogether are present in a torsional spring and it receives the tangential force maximized by the magnet array. In this study, the researcher employed few optimization techniques such as adjoint variable methods and sequential linear programming in 2-D finite element analysis [126].

In 2005 developed a theoretical study about the magnetic circuit for a longitudinal flux PM synchronous linear generator. In order to assess the machine performance, the researcher used a coupled field and circuit model which was solved using the time-stepping finite-element technique [127]. In 2008 [128], and 2010 [129], conducted a comparison of different configurations in an axial-flux nine-phase concentrated-winding PMSG for a direct-drive wind turbine.

Various prototypes where investigated by [130] (2012) in which one of the prototypes demonstrated that the active mass of a PMG unit in a SS-PMG curtailed in a considerable fashion. For different slip-PMG concepts, the evaluation was also performed. To be specific, it is feasible to have a notable amount of minimization in active and PM mass for the new brushless-DC winding slip-PMG in comparison to existing non-overlap winding configurations. Further, it can be projected that the copper can be replaced by aluminum and there is no need to increase the mass of slip-PMG without changing the machine cost performance [130].

A low-speed three-phase generator was considered in 2014 with high induced voltage, low harmonic distortion as well as high generator efficiency, optimal generator parameters such as pole-arc to pole-pitch ratio and stator-slot-shoes dimension topology for investigation. For the purpose of obtaining sinusoidal induced voltages in stator windings, the researcher arranged the PMs in rotor structure and adopted the magnetization direction in an appropriate manner [131]. An insight was

published (2006) about the basis behind the development of PMSG, a novel hybrid in Hybrid Excitation Permanent Magnet Synchronous Generator (HEPMSG). It was developed through the insertion of an exciting winding in rotor or stator [132].

In 2008, [133] developed the Flux Reversal Machine (FRM) coupled with a doubly salient stator permanent magnet machine in addition to flux linkage reversal present in the stator concentrated winding. The study conducted a comparative analysis on Full Pitch Winding Flux Reversal Machine (FPFRM) and Conventional Concentrated Stator Pole Winding FRM (CSPFRM) on the design. The results revealed that FPFRM exhibited high power density than CSPFRM [133].

In order to shuffle the standard claw pole alternator in the place of automobile application, a single-phase FRM was introduced. It has few advantages, such as it has a simple construction process, expresses high-power density, low in inertia etc. Reference [134] (2010) investigated and proposed a distributed winding for FRM (Flux Removal Machine). A high-power density is provided by FPFRM and it enhances the efficiency as well. Being a doubly-salient Permanent Magnet machine with concentrated windings, FRM has advantages of both Switched Reluctance Machine and Permanent Magnet (PM) machines. FEM analysis was carried out in order to achieve the induced EMF, winding inductances, and flux linkages. The winding function strategy received the inductance of both the machines and it was compared with FEM results. On the basis of fabricated 'electrical gear', the power densities of both CSPFRM and FPFRM with PMSM were compared. The gear ratios were provided from various FRM configurations. As the design of CSPFRM, FPFRM, and PMSM are similar with respect to outer dimensions, the volume of the magnet, and the rotor speed, the comparison of those three machines were graphically represented in Figure 18. From the graphical representation, it is observed that the machine FPFRM requires very high compensating kVAr, when compared to CSPFRM and PMSM. However, as far as active weight/kVA is concerned the FPFRM is less when compared with other two [134].

**Figure 18.** Comparison of Conventional Concentrated Stator Pole Winding Flux Reversal Machine (CSPFRM), Full Pitch Winding Flux Reversal Machine (FPFRM), and PMSM [134].

In 2007, [135] developed low-revolution magneto-electric generators which are custom designed for wind power engineering applications. The best and efficient way to diminish the own drag torque is to incorporate a magnetic rake so that the EMF do not exhibit a significant decrease and the adaptability of the magneto electric machine design is preserved as per the manufacturing. Among the available ones, the best alternative is the one, which is found in the magnetic rakes situated outside and inside of the rotor inductors, which is equivalent to the width of the spline way slots that were found inside and outside of the stator.

An optimal design method was proposed in 2009 [136], in which a double-layer permanent magnet (PM) Dual Mechanical Port (DMP) machine was present for wind power application with random low-wind turbine speed input and stable steep synchronous speed output. The torque was compared between the outer-rotor and inner-rotor. Further, they also compared the THD variations with a pole arc co-efficient for inner-rotor and stator winding [136].

With the purpose of overcoming the potential barriers of dimension, cost, and reliability, in 2011 [137], a multi-generator architecture was recommended. They suggested that a total of two PMSGs should be shared with one turbine-driven shaft. The outputs need to be recorded from the two PMSGs, and then rectified in order to be connected in series with an intermediate DC chopper, whereas the back-end inverter is provided with similar option [137].

In 2012, [138] developed an investigation about the novel form of transvers and axial-flux magnetic fields of the PMSG. With novel machine configuration such as rotation, the flow of the main flux would be in the transverse direction. A novel Outer Rotor-Permanent Magnet (PM) Vernier (OR-PMV) machine was introduced [139] (2010) for direct-driven wind power generation that comes packed with low speed. This is because the wind power can be easily capture, and it triggers the high-speed rotating field design in order to enhance the power density [139].

In 2003, [140] developed an operating principle called Consequent-Pole Permanent-Magnet Machine. In addition to the finite element analysis and sizing analysis, the experimental results were achieved for the prototype machine. There are many advantages associated with CPPM machine, one of which is the control on air-gap flux level excluding the demagnetization risk from magnetic pieces. In terms of low-reluctance iron poles, it is possible to execute the control action. In addition to the low field AT requirement, a wide range of air-gap flux control was also yield and this could be leveraged to either increase or diminish the air-gap flux.

In 2012, [141] sought the winding functional theory as the basis and detailed the inductance of a multi-phase synchronous machine supported with a PM or a wound field rotor. Due to the three magneto-static simulation results produced from simple machine-based geometric models, it is easy to determine the permeance function. For the inductance of stator phase, inductance of phase-to-field, and PM flux linkage calculations, the existing method was used. In order to have an accurate incorporation in numerical machine models that pertain to dynamic simulations, the study proposed the machine inductances, which are relevant to Fourier-series expansion.

A novel interpolating strategy was proposed in 2011 [142] for air-gaps by antiperiodic boundary condition when applied to AFPMSG. With the help of coupled-circuit, element analysis, and time-stepping, the performance of AFPMSG in case of isolated load was investigated. The investigation was also conducted upon the performance of short circuit. In order to produce the results of the accurate analysis, the researcher used second-order serendipity quadrilateral elements [142].

A novel three-phase 12/8-pole doubly salient permanent magnet (DSPM) machine was investigated in 2006 [143] so that it can be used in wind power generation. This in turn can be utilized to design and study the recommended DSPM generator i.e., a novel machine structure design which yields high efficiency, high power density, and high robustness in the device of system operation. In order to obtain static characteristics of the proposed generated, the researcher used FEM [143].

In 2017, [144] executed a complete model, design, and development of a novel slip-synchronous Permanent Magnet (PM) wind generator for direct-drive direct-grid connection. There is variation present in the proposed generator with that of the traditional PM induction generators mentioned in the literature. The non-overlap winding was used in the proposed model for the very first time. For the generator design to be effective, a mixture of analytical, finite element calculation and optimized design methods were employed by the researcher. The researcher minimized the critical design parameters such as load torque ripple and no-load cogging torque to the best possible minimum level in the design optimization stage. The model was verified, and the design was completed with measurements of wind generator system prototype.

#### *Energies* **2019**, *12*, 2616

Under linear condition (2007) [145] compared the predictions of the two methods listed herewith in the calculation of electro-magnetic torque with inductance of a synchronous reluctance machine.


In the methods mentioned in the literature, the stator winding connections, stator slot effects, as well as the rotor geometry, were considered. The WFA simulation results were contrasted with that of the 2-D FEA whereas the results were the same. In case of magnetic linear condition, the winding function method seems to be quick, incur less computational costs, and quite simple.

In 2009, [146] analyzed the Hybrid Exciting Synchronous Generator (HESG) in addition to unique operating principles and structure. The upkeep required for the main output is generally taken out by PM generator and, however, the terminal voltage is regulated by a homo-planar inductor alternator. In order to execute the computation of EMF and analyze the performance of HESG, 3-D FEM was utilized.

### **12. Control Mechanism for WECSs**

Numerous control mechanisms are employed in WECS, when designing a generator, it is important to consider the vital parameters such as aerodynamic efficacy, statistical wind distribution, and control system, since these decisive factors can be used in performance evaluation. A study considered high overload capacity generators for this specific application. It was concluded that the optimization of generator is a must to diminish the losses and achieve the highest overload capability. The wind power generation systems act to achieve the sole aim of harnessing the maximum amount of wind energy and consequently converting it to electrical energy. One can achieve this easily through the help of a control structure which allows the operation range as well as the ideal algorithm of stable system with MPPT (Maximum Power Point Tracking). The objective of MPPT presents the harnessing of maximum energy by making few changes in operating point of the system in order to tap the full energy from wind. A control structure was proposed [147] (2013) with specific reference of wind energy systems on the bases of PMSG. With the purpose of enhancing the reliability and the robustness, the study determined the optimum structure using speed and torque control. This retrieved from the analysis of conventional control structures, which used variable speed, and fixed pitch wind energy generation systems [147].

The time-stepping FEA was offered (2006) [148] pertaining to a variable speed synchronous generator in addition to the rectifier. The bi-directional alternator speeds are maintained by this model, the application is a linear generator in terms of the ocean wave energy conversion [148].

In the study conducted in 2014 [149], the authors described the output power fluctuations faced in a wind farm and the relevant problems created in the power system. The study compared the fluctuations that occurred in the output power of conventional schemes with the proposed methods. However, this study's proposed scheme had the tracking of optimal rotational speed in such a manner that the output power was smoothened. A fuzzy PID controller used instead of traditional vector control, which resulted in the tracking of the turbine's optimal rotational speed and the smoothening of wind farm's output power [149].

In 2009, [150] presented a system design model and its control approaches for a 2 MW direct-driven PMSG fed through parallel-connected full power back-to-back PWM converters. Both the electromagnetic FE analysis as well as the optimal generator design was executed in this application in terms of wind generation [150].

An elaborate design and an experimental approach proposed in 2010 [151] for a completely passive wind turbine system without the active electronic part (power and control). The efficiency of the devices predominantly relied upon the condition where the system's design parameters were reciprocally adapted using an integrated optimal design methodology. This methodology ensures simultaneous optimization of wind power extraction and losses because of the global system in a specific wind speed

profile. This way, the weight of the wind turbine generator was decreased. Based on the approaches discussed in earlier studies, optimal PMSG was obtain for critical features of passive wind turbines like geometric and energetic features [151]. Figure 19 illustrates the general representation of a typical variable-speed direct-driven PMSG wind turbine connected to the grid distribution.

The study conducted by [152] developed a holistic modelling of direct-driven PMSG-based grid-connected wind turbines in addition to control schemes for the interface converters. There were two distinguishable control schemes designed in this configuration for generator and grid-side converters.

**Figure 19.** Typical variable-speed direct-driven PMSG wind turbine. Source: GE [153]

### **13. Conclusions**

This review article developed a conceptual framework with an overview of research challenges, using which the proposed work of analysis, suitability, design, and control of PMSGs for Wind Energy Conversion Systems (WECS) to be carried out in future needs and development in the wind sector. The predicted influence and the preliminary results will further the progress beyond the threshold level set by the state-of-the-art envisioned research. In the literature, the WECSs and its classification as per wind turbine and generator schemes, the types of PMSG technologies where discussed elaborately. In addition, the study also suggested the solution found for optimization problems using the field computation technique. This study related to WECS development provided an advanced inter-disciplinary approach on technical parts, and compared with the pros cons of as previous studies. Information provided in this article can also be helpful in improving the WECS. It also reviewed the soft computing (SC) techniques that where applied for the optimal design methodologies of the PMSGs. When exploring the literature, unraveled mysteries and unexplored areas from the developmental perspective to take forward in future.

**Author Contributions:** All authors involved equally, and intensively investigated to produce the comprehensive research survey article in current form for renewable energy system for Wind Energy and its machine dynamics.

**Funding:** This research received no external funding.

**Conflicts of Interest:** The authors declare no conflict of interest.

### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### *Article* **Multi-Channel-Based Microgrid for Reliable Operation and Load Sharing**

### **Ali Elrayyah and Sertac Bayhan \***

Qatar Environment and Energy Research Institute, Hamad Bin Khalifa University, Doha 34110, Qatar; aelrayyah@hbku.edu.qa

**\*** Correspondence: sbayhan@hbku.edu.qa; Tel.: +974-4454-7188

Received: 6 May 2019; Accepted: 28 May 2019; Published: 30 May 2019

**Abstract:** This paper presents a novel approach to distribute available power among critical and non-critical loads in microgrids. The approach is based on supplying power over a number of channels with distinguishable frequencies where loads could be served by these channels according to their level of importance. The multi-channel scheme not only offers flexibility to supply loads but also to share power among adjacent microgrids. The control system, which can deal with multi-channel scheme, is presented and different applications that can be offered whereby are discussed. The number of channels that can be supplied by any inverter is determined based on the parameters of the used filter. Moreover, the power exchange efficiencies over the active channels at various power levels are determined and approximated formulas for quick evaluation are presented. To verify the proposed solution performance, simulation and experimental studies were performed. The obtained results demonstrate the effectiveness of using multi-channel scheme for power exchange in microgrid and also confirm the accuracy of the provided formula related to power exchange efficiencies.

**Keywords:** multi-channel power exchange; microgrid; power electronics-based source/loads; distributed generators; critical loads; uninterruptable power supply; power exchange efficiency

### **1. Introduction**

The development of the microgrid concept has become an important element in the of the future utility grids and a priority in many countries due to its considerable environmental, economic, and social benefits. An increase of microgrid deployment rate will play an important role to meet future electricity demands without significant investment in new power plants. The main advantage of the microgrid is that it operates either in islanded mode or grid-connected mode. On the other hand, the power generation capacity of the distributed generators (DGs) is limited and it is often not sufficient to meet load demand in the islanded mode of operation. Therefore, a control system should be designed to ensure the continuity of the energy at critical loads (hospitals, data centers, etc.) and special attention should be paid for such loads in the islanded operation [1].

There has been extensive research and development in technologies, methods, and systems to secure a reliable power supply for various electric loads (critical and non-critical loads) [2]. In some aspects of their operation, microgrid systems still rely on the same operating principles that were established over a century ago [3]. Considering the advancement in power electronics technologies, loads and sources which are based on power electronics systems are expected to have significant share within power systems in near future [4–6]. Power electronics-based sources and loads can thus provide new means to revolutionize power exchange in microgrids.

One principle that remains rooted in microgrids is the non-differentiable sharing of power among loads. In all power systems, sources set up a single channel voltage and loads draw power from that channel. This indicates that critical as well as non-critical loads extract their need using the same approach. It would be very useful if there was an ability to enable sources to limit the supplied power to serve critical loads when there is a lack of sufficient power supply. However, sources cannot discriminate among the loads in these cases. There are solutions to address this problem which are based on establishing centralized energy management systems (EMSs) that monitor the whole system and adjust the loads accordingly [7,8]. However, EMSs could be challenging to establish especially when addressing large systems [9]. Moreover, the dynamics in the future inverter-based system could be too fast to be accommodated by such centralized-based systems [10]. The power packet presented in [11] proposes a scheme for power switching from one source to a specific load similar to the concept of circuit switching used in old telephone networks. This approach could be very effective but it requires a number of parallel lines and it needs high synchronization between switches.

In this paper, multi-channel-based microgrid is proposed to enhance system reliability and power sharing flexibility. The concept of using more than one frequency channel in microgrid is used in [12]. However, that concept is limited to superimposing a small AC signal driven by droop relation on a DC microgrid to ensure proper sharing among sources. The idea is further extended in [13] to manage power sharing among interlinking converters that transfer power between AC and DC sides of hybrid power systems. The approach proposed in this paper aims to use multi-channel not as communication mean, but rather to achieve an intelligent power exchange within/among microgrids.

The paper is organized as follows. The structure, operation, and control of multi-channel-based microgrid are discussed in Section 2 while the control of the sources and loads within that system are covered in Section 3. Section 4 provides theoretical analysis for the number of channels that can be used within a microgrid and estimate for the associated losses. The effectiveness of the proposed concept is demonstrated through simulation and experimental studies covered in Section 5. Finally, the paper is concluded in Section 6.

### **2. The Proposed Multi-Channel-Based Microgrid Structure and Operation**

Figure 1 illustrates a conceptual diagram of the proposed multi-channel-based microgrid. Power sources—renewable-based or energy storage—can generate power at different frequency channels (50, 100, and 200 Hz in this case) through power electronics converters. The loads, either non-critical or critical, can then tune their power reception to one of more of these channels. The power sources can limit their supply to channels devoted to critical loads such that their power supply service is maintained uninterrupted when overloading occurs. The ability to exchange power over number of channels is not limited to individual sources and loads, but rather it can take place between subsystems such as microgrids.

**Figure 1.** The conceptual diagram of the proposed multi-channel-based power system.

Fortunately, power electronics converters can support multi-channel-based power exchange with merely changes in their control algorithms and no hardware modification. Although the generated voltage signals contain harmonics, these are injected deliberately for power distribution in this application. To block the harmonic components, serial compensators (SCs), which are inverters connected in series with line to apply compensating voltages, are employed. The SCs can also be used in exchanging power among adjacent power systems. The operation of SCs is similar to the electric springs proposed in [14]. Electric springs regulate voltage and perform demand response by adding voltage components in series with load branches [15]. However, the operational scope of SCs is wider than that of electric springs as it deals with several channels.

The single-line diagram of the multi-channel microgrids is depicted in Figure 2. The system contains a number of multi-channel sources/loads as well as conventional ones. The power system is divided into subsystems. Grid and conventional loads are operated by a single frequency (50 Hz) while subsystems A and B can exchange power in multi-channel mode. In Figure 2, subsystem A can supply power in three channels; grid, local critical loads, and power exchange with subsystem B. It is well known that sharing power among interconnected subsystems in coordinated manner can provide several benefits for efficient and reliable system operation [16]. As shown in Figure 2, subsystem A can secure sufficient supply to its critical loads and also designate some power to subsystem B that does not get interfered by other loads within the system. The connections between the subsystems are done through SCs which are operated to block the current flow of certain channels. A communication system can be used to set the various frequencies and to manage the power flow, but it is not critical for the system reliability.

**Figure 2.** Single-line diagram of the multi-channel-based microgrids.

The power sources and loads in this system are connected to the network through power electronic converters. The line voltage in this power system composes of a number of frequency channels (harmonics). Each source can contribute power in any of these channels and likewise loads can receive power from any of these channels. Each source acts on parts of these channels as a voltage source where droop control can be applied. For the remaining channels, the source acts as a current source to feed some or zero current on each one of them.

Multi-channel-based microgrids shown in Figures 1 and 2 require changes in power system structure and the associated cost with these changes needs to be considered to evaluate the effectiveness of such approach. Fortunately, power electronics-based loads and sources are expected to have a larger share within microgrids in the future. As the following section shows, only changes in the inverter control logic are required to apply a multi-channel power exchange which does not impose any extra hardware cost. The need for SCs, on the other hand, might require an investment in additional hardware. However, the deployment of series compensating devices is expected to increase within distribution systems to accommodate for renewable energy sources integration [17]. Accordingly, the hardware components needed to apply the multi-power exchange are expected to be available within microgrids in the near future which opens the doors for this scheme to be utilized once the appropriate control logic is developed.

### **3. Control of Multi-Channel Sources and Loads**

The control system of a multi-channel source is shown in Figure 3. Central controllers manage the various sources by informing them about the active channels and the amount of power that needs to be supplied in each channel. The phase locked loop block (PLL) in each source monitors the magnitude and angle of each channel to:


**Figure 3.** The block diagram of the control system of a source in a multi-channel-based power system.

Based on the applied operation scheme, the channels angles detected by the PLL are sent to current and/or voltage controllers to supply power through those channels. The outputs of these controllers are then combined to drive the inverter switches through the power width modulation (PWM) generator.

The multi-channel loads on the other hand operate as current sinks. In general, some channels could be defined for critical services and the other are for non-critical ones. Any load can then consume the power it needs for its essential functions from the critical channels, while it consumes the remaining demand from the other ones. The SC control structure is shown in Figure 4. This system works merely as a current source inverter where it adjusts the current at all channels that need to be blocked to zero. However, it does not contribute any voltage at the channel allowed to pass.

**Figure 4.** The block diagram of serial compensator control system.

The flowchart of multi-channel operation is given in Figure 5. The cycle starts by updating the list of active channels based on commands sent from central controller. The controller then identifies active channels in the line through the PLL block. The importance of this step is of two folds. First, it is needed to identify the frequency and phase of each channel. Second, the PLL can detect channels' activation/deactivation immediately as it does not have to wait for command from the central controller. The control system then updates the power flow in various channels. Generally, every source needs to maintain a certain power reserve to accommodate for demand increase. As the sources supply more power, this reserve might not be maintained. To recover the required reserve level, sources can perform the following steps:


**Figure 5.** The flowchart of multi-channel operation scheme.

The SC operates similar to the sources without the part related to maintaining the reserve. After detecting the active channels, the currents at blocked channels are set to zero through their current controller operation.

### **4. Determination of Number and Frequencies of Usable Channels**

Activated frequencies in multi-channel systems cannot be arbitrarily selected. For effective operation, there must be clear separations between their frequencies, otherwise the PLL may not be able to identify them accurately. Moreover, even if the channel frequencies are selected with reasonable difference among them, their higher-order harmonics might not be distinguishable from one another. In this paper frequencies that meet these requirements are selected as:

$$f\_k = 2^{k-1} \ f\_{\mathbf{L}\prime} \ k = 1 \ \mathbf{2} \ \dots \ \mathbf{2} \ \dots \ \mathbf{k}\_{\text{mx}} \tag{1}$$

where *fL* is the frequency of the fundamental channel, *fk* is the frequency of the *kth* channel, and *kmx* is the number of channels that can be activated in the power zone. Since merely odd harmonics are experienced in power systems [18], there is at least 50 Hz separation between the harmonics components of the channels indicated in Equation (1).

Different types of filters could be used for multi-channel inverters. However, LCL is considered in this paper since in requires relatively low inductors. Having a low inductor is very important to reduce its associated voltage drop. As indicated in [19], the resonance frequency of LCL filter needs to satisfy

$$10f\_L < f\_{res} < \frac{1}{2}f\_{sw}, f\text{res} = \frac{1}{2\pi}\sqrt{\frac{L\_1 + L\_2}{L\_1 L\_2 \mathbb{C}}}\tag{2}$$

where *fres* is the resonance frequency of the LCL filter and *fsw* is the switching frequency. In the case of a multi-channel system, Equation (2) implies

$$110 \times 2^{k\_{\text{max}} - 1} \text{ fL} \prec f\_{\text{res}} < \frac{1}{2} f\_{\text{sur}} \tag{3}$$

Equation (3) leads to the following relation

$$2^{k\_{\rm{mx}}-1} < \frac{f\_{\rm{sw}}}{20f\_{\rm{L}}} \to k\_{\rm{mx}} < 3.32 \left( \log \left( \frac{f\_{\rm{sw}}}{f\_{\rm{L}}} \right) - 1 \right) \tag{4}$$

Accordingly, for a 50 Hz system with a 20 kHz switching frequency, *kmx* ≤ 5.

To investigate the other constraints for acceptable value of *kmx*, the following information needs to be provided for each channel: frequency (*fk*), rated power (*Pmx*,*k*), and voltage magnitude (*Vmx*,*k*). Consider a multi-channel inverter system shown in Figure 6 where power is transferred over the *kth* channels from a source to a load. As multi-channel systems are based on power electronics sources and loads, it is more effective to transfer only real power as the reactive power needed by loads could be provided by their local inverters. Accordingly, the line voltage *Vo*,*<sup>k</sup>* and output current *io*,*<sup>k</sup>* are assumed to be in phase. The voltage at the capacitor *VC*,*<sup>k</sup>* of the source inverter is given by

$$V\_{\mathbb{C},k} = V\_{o,k} + j\omega\_k L\_2 i\_{o,k} \to \left| V\_{\mathbb{C},k} \right| = \sqrt{V\_{o,k}^2 + \left( \omega\_k L\_2 i\_{o,k} \right)^2} \approx V\_{o,k} \tag{5}$$

The approximation in Equation (5) is done to simplify the analysis and since the voltage drop ω*kL*2*io*,*<sup>k</sup>* should to be very small in comparison with *Vo*,*<sup>k</sup>* for reasonable power transfer. The reactive power supplied to the capacitor *QC*,*<sup>k</sup>* can then be approximated by

$$Q\_{\mathbb{C},k} \approx -\omega\_k \mathbb{C} V\_{o,k}^2 \tag{6}$$

The current that flows over the inductor *L*<sup>1</sup> (*ii*,*k*) accordingly becomes:

$$
\dot{\iota}\_{i,k} \approx \dot{\iota}\_{o,k} + j\omega\_k \mathcal{C}V\_{o,k} \tag{7}
$$

By summing the reactive power supplied to *L*1, *L*2, and *C* total reactive power *Qk* supplied by the inverter can then be written as:

$$\begin{split} Q\_k &\approx -\omega\_k \text{CV}\_{o,k}^2 + \omega\_k \left(\omega\_k \text{CV}\_{o,k}\right)^2 L\_1 + \omega\_k l\_{o,k}^2 (L\_1 + L\_2) \\ &= \omega\_k \left(-\text{CV}\_{o,k}^2 \{1 - \omega\_k^2 L\_1 \mathbb{C}\} + P\_{o,k}^2 / V\_{o,k}^2 (L\_1 + L\_2)\right) \end{split} \tag{8}$$

Assuming *RL*1, *RL*2, and *RC* are the resistances of *L*1, *L*2, and *C* the power transmission efficiency (η*k*) from source to load could be estimated using the relation:

$$\eta\_{k} = \frac{P\_{o,k}}{P\_{o,k} + 2\left(\frac{P\_{o,k}}{P\_{o,k}}\right)^2 (R\_{L1} + R\_{L2}) + 2\left(Q\_{C,k}/V\_{o,k}\right)^2 (R\_{L1} + R\_C)}\tag{9}$$

Equations (8) and (9) can be used to determine whether certain power could be transferred over a specific power depending on the associated reactive power and losses.

Another constraint for multi-channel inverter is related to its DC bus voltage (*VDC*). The DC bus voltage needs to be high enough to generate the required output voltage at all channels. The appropriate value for the DC bus voltage depends on the phase angle of the various channels. When the phase angles of the channels are allowed to be arbitrary set, the DC bus voltage needs to satisfy the following relation

$$\sqrt{2}\sum\_{k=1}^{k\_{\text{max}}} V\_{m x,k} < V\_{DC} \tag{10}$$

Allowing the channels to have any phase is very important for certain applications. For example, when droop control is implemented over the various channels, the exact channel frequency varies slightly around the nominal channel frequency and so does the phase angle. In this case Equation (10) needs to be satisfied. However, in another application the master source could be responsible to set the line voltage (*VL*). In this case, the channel frequencies and phases can be maintained at fixed values set by the master source as:

$$V\_L = \sqrt{2} \sum\_{k=1}^{k\_{\text{max}}} V\_{\text{max},k} \sin \left( 2^{k-1} \pi f\_L t + \phi\_k \right) \tag{11}$$

If φ*<sup>k</sup>* of the various channels do not have the same value, the positive and negative half cycles will have different shapes. It is therefore preferred that all channels have the same value of φ*<sup>k</sup>* which can be taken as zero. To determine the peak voltage of *VL* in this case, let π*fLt* be defined as θ, then the following equation needs to be solved for θ in the range [0, π]

$$\frac{dV\_L}{d\theta} = \sqrt{2} \sum\_{k=1}^{k\_{\text{mx}}} 2^{k-1} V\_{\text{mx},k} \cos 2^{k-1} \theta = 0 \tag{12}$$

The values of θ that solves Equation (12) can then be used to determine the peak voltage. For example, consider the case of three channels with RMS values of *Vmx*,1, *Vmx*,2, and *Vmx*,3 given by *Vm*, 0.5*Vm*, and 0.5*Vm*, respectively. Solving Equation (12) yields a peak voltage for *VL* as 1.37 <sup>√</sup> 2*Vm*. The DC bus voltage in this case can be taken as 1.37 <sup>√</sup> 2*Vm* representing 68% of the value set by Equation (10). This indicates that, depending on the intended application, selecting the right setting for the channel frequencies and phases can have significant impact in components' sizing and design.

Besides the constraints related to switching/resonant frequencies, reactive power supply, and DC bus voltage, ripple in inverter current and voltage drop over the inductors are usually considered while designing LCL filters. The ripple in the inverter output current Δ*iL*<sup>1</sup> in unipolar PWM switching is given by [20]:

$$
\Delta i\_{L1} = \frac{V\_{DC}L\_1}{8f\_{sw}}\tag{13}
$$

*L*<sup>1</sup> must then be selected to maintain Δ*iL*<sup>1</sup> below the required limit. This value of *L*<sup>1</sup> can be used to estimate the voltage drop of the LCL filter. The voltage drop over the filter inductors is usually needed to be kept below a certain value for proper voltage regulation. For the *kth* channel, the voltage drop over the filter inductors can be written as:

$$
\Delta V\_{o,k} = V\_{o,k} - m\_{a,k} V\_{DC} \approx j\omega\iota\_k L\_1 (\mathbf{i}\_{\mathbb{C},k} + \mathbf{i}\_{o,k}) + j\omega\iota\_k L\_2 \mathbf{i}\_{o,k} \\
= \omega\_k^2 L\_1 C V\_{o,k} + j\omega\iota\_k (L\_2 + L\_1)\mathbf{i}\_{o,k} \tag{14}
$$

*ma*,*<sup>k</sup>* is the inverter modulation index. The term *Vo*,*k*ω<sup>2</sup> *k L*1*C* has a 180 degree phase shift from *Vo*,*<sup>k</sup>* and it is added to the inverter output voltage while the term *j*ω*kio*,*k*(*L*<sup>1</sup> + *L*2) has a 90 degree phase shift from *Vo*,*k*. Based on any requirement for the line voltage drop, the relation in Equation (14) can be used to check the possibility of activating a certain channel with the power system under consideration.

**Figure 6.** LCL filter for multi-channel inverter.

### **5. Simulation and Experimental Studies**

To verify the effectiveness of the proposed multi-channel-based power systems, simulation and experimental studies have been conducted. In the experimental setup, a Chroma regenerative grid simulator is used to set up the three channels. Since the grid simulator can absorb power it is used to set the line voltage and to emulate loads in the system. Two sources are then used to supply power over the various channels. Each of the sources is supplied by Magna-Power power supply and it uses IAP inverter and controlled by TMS320F28335 DSP. The system is also simulated using Matlab to analyze its efficiency at different power levels. The considered diagram for the microgrid is shown in Figure 6 where the source and load inverters have identical parameters which are listed in Table 1. The resistors *RL*1, *RL*2, and *RC* are assumed to be connected in series with *L*1, *L*2, and *C*, respectively.


**Table 1.** Inverter parameter of the considered system.

From Table 1, *fres* = 2.1 *kHz* indicating that *kmx* can at most be 3. For source inverter, Figure 7a compares the simulated per unit reactive power and the one provide by Equation (8). For the small value of *Po*,*<sup>k</sup>* the term <sup>−</sup>ω*kCV*<sup>2</sup> *<sup>o</sup>*,*<sup>k</sup>* is dominant, especially for higher-order channels, and the inverter supplies more reactive power. However, as *Po*,*<sup>k</sup>* increases, the value of ω*kP*<sup>2</sup> *o*,*k* /*V*<sup>2</sup> *o*,*k* (*L*<sup>1</sup> + *L*2) causes the supplied reactive power to decrease. Note that the power ranges that correspond to high amounts of reactive power need to be avoided to minimize operating losses. This can be observed clearly in Figure 7b, since when channel 3 supplies less than 0.5 kW, the transmission efficiency is relatively low. When the transmitted amount of power increases, the losses due to the term *Qk*/*Vo*,*<sup>k</sup>* 2 (*RL*1) decrease, while that of *Po*,*k*/*Vo*,*<sup>k</sup>* 2 (*RL*<sup>1</sup> + *RL*2) becomes dominant and hence the efficiencies of all channels become comparable and they all decrease by the same rate.

**Figure 7.** Reactive power supplied by multi-channel inverter to LCL components over the various channels: (**a**) Reactive power supply using Equation (8) and simulation study; (**b**) power exchange efficiency over various channels.

The microgrid system in Figure 8 is considered for experimental studies. Two sources (Src1 and Src2) feed power as current sources to loads in three channels ch1, ch2, and ch3 which have the frequencies 50, 100, and 200 Hz, respectively. The line voltage in the system is given by:

$$
\omega\_L = V\_1 \sin \omega\_1 t + V\_2 \sin \omega\_2 t + V\_3 \sin \omega\_3 t \tag{15}
$$

where *V*<sup>1</sup> = *V*<sup>2</sup> = *V*<sup>3</sup> = 83. Though all channels are generated by Chroma power simulator, ch1 is taken to represent the grid. Two case studies are considered where the sources feed power over the channels differently and the event of grid disconnection are analyzed.

**Figure 8.** Considered microgrid for experimental study.

### *5.1. Case Study I*

In the first case study, Src1 supplied 7 A at ch1 only while Src2 supplied 7 A in each of the three channels. This can be used in applications where the supply of Src1 is unreliable and thus is used to supply uncritical loads over ch1, while Src2 supplies critical loads that are served by ch2 and ch3 as well as uncritical ones in ch1. Figure 9a shows the line voltage and current supplied by the two sources.

**Figure 9.** Experimental study sources' behavior in multi-channel power system after grid disconnection where one source supplies power to the grid channel only while the other feeds power to all channels; (**a**) Src1 feeds power to ch1 and Src2 feeds power over all channels; (**b**) grid disconnection causing elimination of ch1; (**c**) Src1 and Src2 output currents after grid disconnection.

At t = 4.5 s, the grid disconnection was imposed to analyze the system performance. The controllers of Src1 and Src2 could detect this event locally and eliminate any power supply over ch1. The line voltage and sources currents waveforms reflected that as shown in Figure 9b. Eventually, the line voltage and sources current were adjusted to feed the loads that were served by ch2 and ch3 as shown in Figure 9c.

The power and energy fed by the two sources for case study I are demonstrated in Figure 10. Before t = 4.5 s, Src2 fed three times the power of Src1 as it was active over three channels. However, after the grid disconnection, the power supplied over ch1 was dropped, which represented all the power supplied by Src1 and one third of Src2 supplied power. This behavior could be very useful to enable sources to play different roles in supplying load-demand with fast and reliable responses to system changes and without explicit inter-controllers communication.

### *5.2. Case Study II*

Case study II represents an application where certain critical loads need not to experience any disruption in their power supply. In this case, sources were configured such that one source (Src2) was exclusively used to serve critical loads in ch2 and ch3 while Src1 participated a guaranteed amount of power to critical loads, with any extra power being fed to ch1. Originally, as Figure 11a shows, Src1 supplied 7 A over the three channels and Src2 supplied 7 A in each of ch2 and ch3. At t = 4.5 s, grid disconnection was imposed which eliminated ch1 from the line voltage. As shown in Figure 11b, Src1 could respond to this event by adjusting their current at ch1 to zero without any major disturbance in the sources current supply over ch2 and ch3. The system then maintained the operation under the new condition highlighted in Figure 11c.

**Figure 10.** Power and energy supplied by Src1 and Src2 in case study I.

**Figure 11.** Experimental study sources behavior in multi-channel power system after grid disconnection where one sources supply power to all channels while the other supplies power merely to critical loads on non-grid supplied channels; (**a**) Src1 feeding power in ch2 and ch3 and Src2 feeding power over all three channels; (**b**) Grid disconnection casuing elimination of ch1, (**c**) Src1 and Src2 output currents after grid disconnection.

The power and energy fed by the two sources are demonstrated in Figure 12. The power supply of Src2 was maintained at the same value throughout the experiment duration as it goes in its entirety to critical loads. On the other hand, one third of the Src1 supplied power was withdrawn after t = 4.5 since that power was used to feed uncritical loads or to be fed back to the grid.

**Figure 12.** Power and energy supplied by Src1 and Src2 in case study II.

### **6. Conclusions**

The use of sources that supply power over a number of frequencies in multi-channel microgrids provides many benefits to power systems. A multi-channel-based power supply helps maintain an uninterruptable power supply to critical loads, enhances system reliability, and allows more liberalized power trading among adjacent microgrids. The paper shows the viability of supplying power over more than one channel through the proper configuration of voltage and current controllers. The parameters of an inverter filter determine the maximum number of channels that can be supplied by that inverter. Moreover, supplying low power at higher-order channels has poor efficiency as a significant amount of reactive power is supplied in these cases to the reactive components of the filter. The formulas provided in the paper for the allowed number of channels and power exchange efficiencies assist in determining the effectiveness of applying the concept of a multi-channel scheme in any microgrid. The paper focuses on introducing the concept of a multi-channel power exchange, its implementation scheme, some of the related advantages, and discussion of its limitations. In future publications, we plan to cover and analyze in detail subjects such as system stability, controller design, and interactions among various channels.

**Author Contributions:** The research presented in this paper was a collaborative effort among both authors. A.E. and S.B. conceived, implemented, and got the results along with the paper write-up. A.E. and S.B. wrote the paper and discussed the results and revised the manuscript critically.

**Funding:** The publication of this article was funded by the Qatar National Library.

**Acknowledgments:** The publication of this article was funded by the Qatar National Library.

**Conflicts of Interest:** The authors declare no conflict of interest.

### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### *Article* **A Novel Integrated Topology to Interface Electric Vehicles and Renewable Energies with the Grid**

### **Alfredo Alvarez-Diazcomas 1,†, Héctor López 1,†, Roberto V. Carrillo-Serrano 2,†, Juvenal Rodríguez-Reséndiz 2,†,***∗***, Nimrod Vázquez 1,† and Gilberto Herrera-Ruiz 2,†**


Received: 19 September 2019; Accepted: 23 October 2019; Published: 26 October 2019

**Abstract:** Electric Vehicles (EVs) are an alternative to internal combustion engine cars to reduce the environmental impact of transportation. It is common to use several power sources to achieve the requirements of the electric motor. A proper power converter and an accurate control strategy need to be utilized to take advantage of the characteristics of every source. In this paper is presented a novel topology of a multiple-input bidirectional DC-DC power converter to interface two or more sources of energy with different voltage levels. Furthermore, it can be used as a buck or a boost in any of the possible conversion of energy. It is also possible to independently control the extracted power in each source and any combination of the elements of the system can be used as source and destiny for a transfer. Finally, the interaction with the grid is possible. The operation, analysis and design of the converter are presented with different modes of power transfer. Simulation results are shown where the theoretical analysis of the converter is validated.

**Keywords:** DC-DC/DC-AC power converter; electric vehicle; multi-input converter; sliding mode control; photovoltaic module; grid; renewable energies

### **1. Introduction**

Global warming is one of the greatest challenges today for humankind. The transportation sector is one of the largest contributors to the emissions of greenhouse gases. It represents 27% in the European Union in 2016 and 28% in the United States for the same year which represents the major contribution. Moreover, it is responsible for the greatest growth in emissions currently due to the growth of tourism, the globalized economy and the increase in living standards. A viable alternative to reduce emissions due to transportation is using electric vehicles (EVs), which practically behave like zero-emission cars [1].

In this type of cars, it is common to use several power sources to achieve the requirements of the electric motor such as fuel cells, batteries, ultracapacitors (UCs), and so forth. The aim of a hybrid energy storage system is to make use of the strong features of each source while eliminating their weaknesses [2]. Researchers have hybridized batteries and UCs in References [3–6] to create an ideal energy storage unit with high energy/power density, low cost/weight per unit capacity and a long cycle life. The battery can be used when the vehicle maintains a relatively constant velocity to take advantage of its high energy density characteristic. Also, the peak power transients during acceleration and regenerative braking can be avoided by the inclusion of a higher power density element such as an UC. The ability of the UC to handle higher power for a higher number of charge/discharge cycles not only increases the life span of the battery but also improves the overall system efficiency [7–9]. The active hybridization of the aforementioned energy storage system, in which the power/current in its output can be fully controlled, is only possible by means of utilizing power converters.

These converters can be isolated or non-isolated. In Reference [10] a novel multiport isolated bidirectional DC-DC converter for hybrid battery and supercapacitor applications is presented, which can achieve zero voltage switching for all switches in the whole load range. Moreover, the current ripples are greatly decreased by interleaved control, which is good for battery and supercapacitor. In Reference [11] it is proposed architecture eliminates two boost switches which are present in the two-stage counterpart. Moreover, the input inductors are operated in discontinuous conduction mode; thus, power can be shared between input sources through proper selection of input inductors. In Reference [12] a new modified LCLC series resonant circuit based dual-input single-output isolated converter is proposed for hybrid energy systems. With this novel converter topology, two different voltage sources can be decoupled completely and transfer the power from two separate dc sources to dc load simultaneously. Moreover, it consists of only two controllable switches for integrating two separate voltage sources; it can provide good voltage regulation and soft switching over a wide load range. Nevertheless, these converters use a transformer to achieve galvanic isolation between sources and output; therefore, are much more complex in terms of designing and control when compared to the non-isolated ones.

In References [13,14] is presented a simple way to build a non-isolated hybrid energy storage system, connecting one of the sources directly while linking the other utilizing a DC-DC converter; yet this method does not permit the adjustment of the DC bus voltage. Another technique is to link each of the sources with the DC bus with an individual converter as presented in References [15–18]. In this way, it is possible to manage the DC bus voltage but it is an expensive solution due to the utilization of multiple converters. In order to decrease the cost, multiple-input converters have been proposed to achieve the goals of EVs. In Reference [19] is demonstrated that the multiple-input converters are cost effective, reliable, simple and easy to control. In Reference [20], energy flow between *N* different sources and the DC link are discussed. In this topology, it is not possible to transfer energy directly between DC sources. In Reference [21] a Z-source converter for EV application is presented, although this topology is suitable for optimal devices and components, the number of voltage sources is limited to two and it is not possible to extend this topology for multiple-input sources. In Reference [22] a modular multipleinput converter is presented, whose input ports are connected to the DC bus via half-bridges. However, with this topology it is not possible to transfer the energy directly between the sources. In Reference [23] is presented a converter with the same characteristics as that presented in Reference [22] but with a reduced element count. In Reference [24] a flexible topology is presented that can be used as a boost or a buck in any transfer of the energy and allows the direct exchange of energy between the sources. A non-desirable characteristic of this converter is the presence of an inductor per input, due to the intrinsic weight of these elements. In Reference [25] a topology is introduced that presents a greater gain compared to other existing ones, for its use on fuel cell-based EVs. In detriment of this converter, it can be said that it is not bidirectional. In References [26] and [27] are proposed converters with multiple-input and multiple-output, utilizing only one inductor. These types of converters are very useful for its use with multilevel inverters. Nevertheless, are not bidirectional converters and that can be a limitation in this type of application. Moreover, in Reference [28] is presented an inverter for the injection of energy generated in a panel into the grid. As stated in Reference [29] it is very important to take advantage of the renewable energies and for that reason in this work the topology presented in Reference [28] was the base of the DC-DC converter proposed in order to achieve the interaction with the grid.

There are several solutions for converters to harvest the energy generated in a Photovoltaic module (PV) and store it in the battery.Reference [30] presents an isolated multiport bidirectional DC-DC converter capable of parallel power management of various renewable energy sources. The advantage of this converter is it utilizes less number of controllable switches and provides soft switching for converter primary switch. Reference [31] proposes a battery charger for an EV based on a Zeta converter. This converter has the advantage of an output current without ripple and it is possible to buck or boost the input voltage. On the other hand, in Reference [32] it is utilized a cascaded buck-boost converter for the application. Such converters are typical of PV-battery systems for its simplicity, its bidirectional capability and the versatility to buck or boost the input voltage. Another solution is presented in Reference [33]—a non-isolated three-port switching boost converter. By controlling three degrees of freedom, the ports have boost, buck and buck-boost characteristics. Nevertheless, as stated in Reference [34], the boost converters provide the lower cost and higher efficiency of the non-isolated converters for PV systems.

In order to minimize the size, weight and cost of the traditional on-board chargers, integrated chargers have been proposed, some topologies and techniques are reviewed in Reference [35]. One concept of integration was proposed by Rippel and Cocconi in References [36,37], consisting in the use of the existent inverter and motor windings for the charging operation. Since the traction operation and charging the battery are not simultaneous, using the drivetrain components could reduce the size and cost of the on-board chargers. This modified structure supports the charging/discharging process of the battery. Some examples based on an induction machine are presented in References [38–40], based on PMSMs in References [41,42], based on windings rearrangement in Reference [43] and finally based on multiphase machines in Reference [44]. Another technique consists of combining a modified DC-DC converter with the AC-DC bidirectional rectifier. In Reference [45] the integrated converter is able to function as an AC-DC battery charger and to transfer electrical energy between the battery pack and the high-voltage bus of the electric traction system. The converter has a reduced number of high-current inductors and current transducers and presents fault-current tolerance in PHEV conversion. In Reference [46] a single-stage integrated converter is proposed based on direct AC-DC conversion theory. The proposed converter eliminates the full-bridge rectifier, reduces the number of semiconductor switches and high current inductors and improves the conversion efficiency. In Reference [47] is presented a bidirectional converter that not only enables beneficial vehicle-to-grid (V2G) interactions but also ensures that all power delivered to and from the grid has good power factor and near zero current harmonics. To accomplish this task, a multi-level bidirectional AC-DC converter is combined with an integrated bidirectional DC-DC converter. The proposed converter has four different modes of operation that allows it to supply power to or from the battery to either the grid or the high voltage bus of the EV. As stated in Reference [48], the increase of the renewable energies can affect the power system efficacy, the power quality, the security, among other problems; therefore, these converters need to have the possibility of power factor correction (PFC) when charging the battery and low total harmonic distortion (THD) when injecting current into the grid to have an interaction with the grid without disturbing the quality of the energy.

The aim of this work is to propose an integrated topology that combines a multiple-input DC-DC converter and a bidirectional rectifier for the interaction of the storage devices with the grid. The proposed converter topology has all the considered advantages from the architectures presented in the literature, such as allowing the bidirectional power flow, the possibility of directly transferring the energy between sources, that for every transfer it is possible to boost or buck the input voltage, that any storage element of the system can be the source or destination for a transfer and has only one inductor, which means less weight, and finally, that it permits the interaction of the storage devices with the grid.

For this purpose, in the present investigation three main sources are considered—a PV, an UC and a battery. Furthermore, a pedal was thought to be mechanically connected to the engine to generate energy, through the electric machine when the vehicle is parked. In addition, when the automobile is in motion, it can help the engine with its mechanical load. All these elements can be seen in Figure 1, which shows a diagram of the considered EV. The focus of this paper is the DC-DC converter.

The operation and steady-state analysis of the proposed converter topology, with all its cases, is explained in Section 2. The size of the elements of the circuit is presented in Section 3. The control

strategy used to regulate the converter is shown in Section 4. Finally, the simulation results for all cases are presented in Section 5.

**Figure 1.** Diagram of the electric vehicle (EV) considered in this investigation.

### **2. Operation and Analysis of the Converter**

The proposed power converter topology is displayed in Figure 2. It can be built by connecting each bidirectional source/output through two switches to the inductor *L*<sup>2</sup> while the unidirectional sources/outputs only need one switch. For a better understanding and analysis of the converter, this is divided into several stages in the present work. This can be done because they work independently. One of these stages is dedicated to the maximum power point tracking (MPPT) of the PV, shown in Figure 3. The other stage is the multiple-input converter shown in Figure 4. In case that an interaction with the grid is required to charge the UC or the battery, the diagram is shown in Figure 5. Furthermore, the injection of current into the grid was considered starting from the energy generated in the PV. In this way, when the vehicle is parked in daylight the energy can be harvested. Figure 6 indicates the circuit for this case. This article presents a detailed analysis of the operation of these converters, as well as simulation results for each of the possible cases.

**Figure 2.** Proposed topology of the power converter.

**Figure 4.** Multiple-input converter.

**Figure 5.** Circuit of the converter when it is needed the charge of the ultracapacitor (UC) or the battery.

**Figure 6.** Circuit of the converter when the vehicle is parked in the sunlight.

In Table 1, various topologies are compared, attending several parameters of interest. Compared to the converters presented in the literature, the proposed topology presents more switches to its operation. Nevertheless, this is justified because it gathers all the advantages identified in the other converters.


**Table 1.** Comparison of the proposed converter with the existing ones.

Now, the converter associated with the PV will be analyzed. When it is dedicated a converter to achieve the MPPT it is easiest to control the system. Some modifications were done to the boost converter such as the addition of the elements *Sprot* and *Dprot* for the protection of the system. It allows to isolate the PV in case no more energy is required without damaging the inductor *L*1. Finally, the *SUC* switch is added to provide greater versatility to the circuit.

In order to make better use of the energy the UC voltage is monitored to define the switch on which to act. In case that the *VUC* > *V*<sup>1</sup> the energy of the PV is transferred to the UC and the capacitor *Cpv*. Switch *S*<sup>1</sup> is first turned on and the voltage *V*<sup>1</sup> appears across *L*1, resulting in the increase of the current in this device with a slope of *V*1/*L*1. In the next interval, this switch is turned off and switch *SUC* and diode *D*<sup>1</sup> are turned on. In this way, the energy stored in the inductor is transferred to the UC and the capacitor *Cpv*. It was decided to not include a switch in the position of the diode *D*<sup>1</sup> because it is acceptable for the application that the voltage in capacitor *Cpv* matches the voltage in the UC. Figure 7 shows the two stages needed to achieve the transfer of the energy.

**Figure 7.** Operating modes of the converter associated with the PV when *VUC* > *V*1, (**a**) charge of the inductor, (**b**) discharge of the inductor.

In Table 2 are represented the operating modes of the converter associated with the PV. If *V*<sup>1</sup> > *VUC*, the inductor is charged through the *SUC* switch in order to emulate the behavior of the boost converter. In both cases, the point of operation in steady-state is giving by Equation (1). It is necessary to point out that in case that *VUC* is greater than *V*1, then *Vpv* = *VUC*.

$$V\_{pv} = \frac{1}{1 - D} V\_1 \tag{1}$$

where *D* is the duty cycle giving by Equation (2).

$$D = \frac{t\_{1A}}{t\_{1A} + t\_{1B}} \tag{2}$$

where *t*1*<sup>A</sup>* is the length of time of the mode 1*A* of Table 2 and *t*1*<sup>B</sup>* is the length of time of the mode 1*B* of Table 2.


**Table 2.** Operating modes of the converter associated with the PV.

*<sup>a</sup>* Increase the state of charge; *<sup>b</sup>* The same state of charge was kept; *<sup>c</sup>* Decrease the state of charge.

Now, the multiple-input converter will be analyzed. The operating modes used in the multiple-input converter emulates the behavior of a buck converter or a boost one. Figure 8 shows the operating modes for a transfer from the battery to the capacitor *Cbus*. Switches *S*<sup>4</sup> and *S*<sup>8</sup> are first turned on and the voltage *Vbat* appear across inductor *L*2, resulting in the increase of the current in this device with a slope of *Vbat*/*L*2. In the next interval, switch *S*<sup>8</sup> is turned off and switch *S*<sup>7</sup> is turned on to complete the transfer of energy to the bus. Therefore, *S*<sup>4</sup> stay on throughout the transfer and switches *S*<sup>7</sup> and *S*<sup>8</sup> work in a complementary manner. As mentioned before, the proposed converter is very flexible, allowing that any element of the system can be source or destiny for a transfer and any transfer can be performed by boosting or reducing the input voltage. Hence, the aforementioned transfer can be realized reducing the input voltage with the correct combination of switches. Nevertheless, for this application, only the transfers present in Table 3 were considered.

One of the main concerns when using the number of switches utilized is the efficiency of the system. However, this is not a serious problem since only three switches are used to perform a transfer of energy; one that stays on and two that work in a complementary way, as can be appreciated in Table 3.

**Figure 8.** Operating modes of the multiple-input converter, (**a**) charge of the inductor, (**b**) discharge of the inductor.


**Table 3.** Operating modes of the multiple-input converter.

The losses analysis of the proposed topology can be done with the circuit presented in Figure 9. The power dissipated by the devices *S*4, *D*<sup>4</sup> and *L*<sup>2</sup> is mainly by means of conduction due to the current is flowing in these devices at all time and is given by the Equation (3). The conduction power loss of the switches *S*<sup>7</sup> and *S*<sup>8</sup> are given by the Equation (4), while the switching power loss is given by Equation (5). Finally, total power dissipation due to all the elements is given by the Equation (6). Where *I*<sup>2</sup> is the current through the inductor *L*2, *D* is the duty cycle, *fsw* is the switching frequency, *tr* is the rise time of the Metal-Oxide-Semiconductor Field-effect Transistors (MOSFETs), *tf* is the fall time of the MOSFETs and *RSL*, *RDS*(*on*)*S*4, *RDS*(*on*)*D*4, *RDS*(*on*)*S*8, *RDS*(*on*)*D*8, *RDS*(*on*)*S*7, *RDS*(*on*)*D*<sup>7</sup> are the series resistance of the inductor *L*2, switch *S*4, diode *D*4, switch *S*8, diode *D*8, switch *S*<sup>7</sup> and diode *D*<sup>7</sup> respectively. From this analysis, it can be appreciated that it is not a major issue the power loss compared to topologies that exist in the literature and the use of all the switches can be justified by the advantages gathered in the proposed topology.

$$P\_{L\_1} = I\_2^2 (R\_{SL} + R\_{DS(on)S4} + R\_{DS(on)D4}) \tag{3}$$

$$P\_{L\_2} = l\_2^2 D \left( R\_{DS(on)\,S8} + R\_{DS(on)\,D8} \right) + l\_2^2 (1 - D) \left( R\_{DS(on)\,S7} + R\_{DS(on)\,D7} \right) \tag{4}$$

$$P\_{L\_3} = V\_{in} I\_2 f\_{sw}(t\_r + t\_f) \tag{5}$$

$$P\_{L\_{-}Total} = P\_{L\_{1}} + P\_{L\_{2}} + P\_{L\_{3}} \tag{6}$$

**Figure 9.** Circuit considered for the analysis of the losses.

For the efficiency calculation it can be assumed that the on-resistance of the MOSFETs and diodes is of 0.5 Ω, a rise time (*tr*) of 28 *ns* and a decrease time (*tf*) of 44 *ns*. Finally, the series resistance of the inductor is estimated to 1 Ω. The transfer that causes a greater loss is from UC to the bus and for that reason this conversion was considered for the power loss calculation. In this conversion the current through *I*<sup>2</sup> is of 10 A and the voltage in the UC is of 125 V. Substituting the corresponding values in Equations (3)–(6) it is obtained that the total power loss is of 309 V. The efficiency is given by Equation (7), which for an output power of 1.8 kW the efficiency is of 85.35%.

$$\eta = \frac{P\_O}{P\_O + P\_{L\_{\text{Total}}}} 100 \tag{7}$$

### **3. Design of the Converter**

In this section, the analysis for the sizing of the inductors *L*<sup>1</sup> and *L*<sup>2</sup> will be explained, as well as the capacitors *Cpv* and *Cbus*. First, the calculation of the components associated with the PV converter were performed. For the calculation of these elements, the methodology described in Reference [49] was followed. In this way, the capacitor was calculated as described in Equation (8), while the inductor was obtained as set out in Equation (9).

$$\mathcal{C}\_{pv} = \frac{DP\_O}{V\_O \Delta V\_O f} \tag{8}$$

$$L\_1 = \frac{D(1-D)^2 V\_O^2}{2f P\_O} \tag{9}$$

where *D* is the duty cycle, *PO* is the output power, *VO* is the output voltage and *f* is the switching frequency.

To make the calculation was necessary to take into account some design considerations with respect to the PV and the desired operation for the converter. In this application was considered a PV that has an open circuit voltage of 50 V, short circuit current of 10 A and with a maximum power point of energy transfer around 25 V and 4–5 A (100–125 W). As proposed in [31], a *D* = 0.5 is used, the switching frequency was set in 50 kHz, while the typical output power was fixed in 150 W. Finally, the variation in the output voltage was defined as 10 V. For this data, the capacitor value needs to be greater than 3 μF. Finally, the inductor value was calculated and determined as 21 μH.

*Energies* **2019**, *12*, 4091

Next, the values of the inductor and the capacitor of the multiple-input converter were calculated. As explained in the previous section, this converter always behaves as a buck or as a boost. For this reason, the strategy followed to obtain these parameters was to calculate the necessary values for each case and take the critical values. The equations used to dimension the power elements in a buck converter are described in Reference [50] and are presented in (10) and (11). While the equations used for the sizing of the power elements in a boost converter are described in Reference [51] and are presented in (12) and (13).

$$C = \frac{\Delta I\_L}{8f \Delta V\_O} \tag{10}$$

$$L = \frac{V\_O(V\_{in} - V\_O)}{\Delta I\_L f V\_{in}} \tag{11}$$

$$\mathcal{C} = \frac{I\_{\mathcal{O}} D}{f \Delta V\_{\mathcal{O}}} \tag{12}$$

$$L = \frac{V\_{in}(V\_O - V\_{in})}{\Delta I\_L f V\_O} \tag{13}$$

where *IL* is the current through the inductor, *f* is the switching frequency, *VO* is the output voltage, *Vin* is the input voltage, *IO* is the output current and *D* is the duty cycle.

Taking into account these equations, and the behavior that was highlighted in the previous section for each state, Table 4 was constructed. In this table are presented the nominal values of the parameters needed for the sizing of the power elements. In this way, was possible to calculate the minimum inductance and capacitance allowable for the proper operation of the converter. It is important to highlight that the transfers concerning the PV were not included because this source of energy is considered as auxiliary. It can be concluded after analyzing this table that the value needed for the inductor *L*<sup>2</sup> must be higher than 3.86 mH and for the capacitor *Cbus* it must be greater than 144 μF.

**Table 4.** Sizing of the power elements of the multiple-input converter.


### **4. Control Strategy**

The control diagram of the system is shown in Figure 10.

It can be appreciated that there are several levels of hierarchy. The purpose of the lower level is to achieve the regulation of the state variables of the converter. Moreover, are present the blocks of MPPT, which are responsible for the extraction of the maximum power. Finally, there is an energy management system (EMS) that dictates the proper transfer at every moment. There are several strategies to control the state variables in a switching power converter as the ones presented in Reference [52]. Nevertheless, in this work, the current through *L*<sup>2</sup> was regulated utilizing a sliding mode control (SMC) because this technique is naturally suited for the regulation of switched controlled systems such as DC-DC power converters [53] and for its robustness, as stated in References [54–56]. Moreover, the control parameter can be the same to regulate the current in a boost and a buck converter as demonstrated in Table 5. In this table, *Vin*, *vC* and *iL* are the variables of the respective converter, while *u* is the input signal (for this case is the state of the switch), *σ* is the sliding surface and S is the control parameter of the SMC.

**Figure 10.** Block diagram of the controller of the system.

**Table 5.** Demonstration of the homogeneity of the control parameter for the regulation of the current with SMC in a buck and a boost converter.


The other state variable is the bus voltage (*Vbus*). This voltage was regulated with an on-off controller, because it is not necessarily an exact value in the input of the inverter and because of its simplicity. The range of the voltage was defined from 400 V to 500 V. This controller enable/disable the input current to the capacitor *Cbus*.

In order to achieve the MPPT of the PMSG there are several techniques like that presented in Reference [57] and is used in this work due to its simplicity. This algorithm is based on the principle that if the voltage in the capacitor *Cbus* is kept constant, then the maximum possible energy is being extracted. On the other hand, the MPPT applied for the PV was based on a Perturb and Observe (P&O) algorithm. This algorithm was implemented with the PV as the active input in the multiple-input converter and manipulating the current through the inductor *L*2. Nevertheless, this is not always possible as the multiple-input converter can be needed for other functions. If this were the case, then the algorithm manipulates the state of the switches *SUC* and *S*1.

Finally, the EMS is the element that establishes the proper transfer in every moment. It is based on if-else rules to achieve its purpose. Figure 11 shows the main cases treated in this controller. If the vehicle is parked, then the source generating energy transfer is the same as the storage device that needs it. In this case, the UC is prioritized because it is desirable that the demanded energy at the start of the motion was supplied by this device. If the vehicle is in motion with positive acceleration and there is energy in both power sources, the flowchart depicted in Figure 12 is followed. If, on the contrary, there is only one power source with energy, then this device supplies the power during the driving cycle. Lastly, when the vehicle is in regenerative braking the energy generated is stored in the capacitor *Cbus*. For each case, the EMS establishes a code and the output decoder with this information and the control signal from the SMC sends the correct signal for each switch.

**Figure 11.** Cases taking into account for the implementation of the EMS.

**Figure 12.** Flowchart followed by the EMS when the vehicle is in motion and there is energy in both power sources.

The control strategy previously presented requires a complement when it is required to interact with the grid. For this purpose, there are two well differentiated cases; charging the battery or the UC from the grid or injecting the energy generated in the PV into the grid. Figure 13 shows the block diagram for the charge of the storage elements. The objective is to regulate the current in the inductor *L*<sup>2</sup> and send it to the device that requires it. In Reference [58], it was proved that the SMC technique is adequate for achieving the control of an inverter interacting with the grid. For that reason this strategy was used and, in addition, it was necessary to know if the voltage grid is in the positive or negative half cycle to act on the necessary switch. Table 6 shows the behavior of the charge decoder. The switches that are not considered in Table 6 are kept off all the time.

**Figure 13.** Block diagram of the control strategy of the charge from the grid.

**Table 6.** Behavior of the charge decoder.


On the other hand, Figure 14 shows the block diagram for the injection of current into the grid. One of the objectives is to create a hysteresis behavior in the voltage *Vpv* acting on the current *I*<sup>2</sup> reference. If the voltage increases above a set value, the reference is augmented in order to decrease the voltage below a permissible value where is increased the current reference again. It is necessary to maintain the voltage *Vpv* above 170 V to guaranteed a satisfactory injection of the current into the grid. Moreover, in order to achieve an interaction with low THD at the beginning and at the end of the injection are executed in the zero-crossing. Finally, two separated SMC are implemented for each inductor (*I*<sup>2</sup> and *Ig*). With this information and the half-cycle in which the grid is operating, the injection decoder establishes the corresponding signal in each inductor. Table 7 shows the behavior of the injection decoder. The switches that are not considered in Table 7 are kept off all the time.

**Figure 14.** Block diagram of the control strategy of the injection of current into grid.


**Table 7.** Behavior of the injection decoder.

### **5. Simulation Results**

The validation of the proposed converter and the control strategy was realized by the simulation of the system in the PSIM software (version 9.0.3). The main considerations for the simulation are presented in Table 8.

**Table 8.** Considerations for the simulation of the multiple-input converter.


The validation of the blocks of MPPT was possible in a separate way. Figure 15a shows the MPPT achieved for the PV utilizing the multiple-input converter for this purpose. It can be appreciated that the system has a good response with this method. The settling time is less than 0.1 *s* for a power variation of 100 *W*. Moreover, in steady-state, the ripple current is less than 0.25 *A* and the ripple power is less than 1 *W*. Otherwise, the second method is manipulating the switches *SUC* and *S*1, obtaining the response shown in Figure 15b. It is realized that this is a response of lesser quality. The settling time is of 0.02 *s* and in steady-state, there is a current ripple of 15 *A* and a power ripple of 8 *W*. Despite the lesser quality response obtained with this method, it has the advantage that the multiple-input converter can be dedicated to other tasks.

**Figure 15.** Methods implemented to achieve the MPPT of the PV, (**a**) utilizing the multiple-input converter and manipulating the reference of the current *I*2, (**b**) manipulating the switches *SUC* and *S*1.

The PMSG also needs an MPPT algorithm for the harvesting of the energy. Figure 16 shows the simulation results for the implemented algorithm. It can be appreciated that by maintaining the voltage *Vbus* constant the MPPT is achieved. With a previous characterization of the electric machine it can be concluded that 92% and 97% of the possible total were extracted, respectively, for the conditions of the test.

**Figure 16.** MPPT achieved in the PMSG by keeping the voltage in the capacitor *Cbus* constant.

For the validation of the EMS and the performance of the converter, the driving cycle ECE-15, which was legislated in the European Union in 1970, was utilized. Figure 17 shows the behavior of the state variables of the converter during the cycle. It can be concluded that the regulation of these variables was successful, with a ripple current of 0.8 A in steady-state and the desired hysteresis of 100 V for the voltage *Vbus*. Figure 18 shows the performance of the power sources. It is important to point out that the UC supplies the energy during the velocity transitions and when the velocity remains relatively constant the demanded energy is supplied by the battery as desired. Finally, the behavior of the PV and the PMSG is shown in Figure 19. The PV is generating energy during the entire duration of the cycle and verifies the proper switch between the two methods implemented. The generation of energy from the pedals was simulated in the beginning of the cycle, obtaining the desired behavior of the system.

**Figure 17.** Behavior of the state variables through the ECE-15 cycle. The speed of the motor following the ECE-15 cycle; the current through inductor *L*2, *I*<sup>2</sup> and the voltage in the capacitor *Cbus*, *Vbus*.

**Figure 18.** Behavior of the energy storage devices through the ECE-15 cycle. The speed of the motor following the ECE-15 cycle; the voltage in the ultracapacitor *VUC* and the voltage in the battery *Vbat*.

**Figure 19.** Behavior of the energy sources through the ECE-15 cycle. The speed of the motor following the ECE-15 cycle; the power harvested from the solar panel *Pout*\_*PV* and the power harvested from the PMSG *Pout*\_*PMSG*.

Finally, the interaction with the grid was validated. Figure 20 shows the charge of the battery. It can be appreciated that the battery has been charged and the current *I*<sup>2</sup> is regulated in 10 A with a ripple of 1 A. Despite the main objective being achieved, the system presents a deficient power quality. In order to achieve a better response, a PFC strategy need to be implemented.

**Figure 20.** Charge of the battery from the grid: Current through the inductor *L*2, *I*2; voltage in the battery *Vbat*; current in the grid *Igrid* and the voltage in the grid *Vgrid*.

Figure 21 demonstrates the injection of current into the grid. It can be appreciated that the MPPT is working properly as is the hysteresis of the voltage *Vpv*. This behavior of the voltage is permissible because the only restriction is that needs to be greater than 170 V and it is achieved by modifying the current amplitude injected into the grid. Figure 22 shows a zoom of the injected current where it can be appreciated that it is in phase with the normalized voltage grid and it is also notable that the start of the injection coincides with the zero-crossing.

**Figure 21.** Injection of current into the grid—the power harvested from the solar panel *Pout*\_*PV*; the voltage in the capacitor *Cpv*, *Vpv*; the current in the inductor *L*2, *I*<sup>2</sup> and the voltage and current of the grid *Vgrid* and *Igrid* respectively.

**Figure 22.** Zoom of the currents in the interval 0.5 *s* to 1 *s*—the current in the inductor *L*2, *I*<sup>2</sup> and the voltage and current of the grid *Vgrid* and *Igrid* respectively.

### **6. Conclusions**

In this work, a bidirectional multiple-input DC-DC converter topology is proposed for electric vehicles. A very flexible topology has been achieved, which gathers all the advantages identified in the converters presented in the literature. This converter allows the interaction of two or more energy sources with a wide range of voltage in its inputs with a direct current bus. It can be operated in buck or boost mode for each possible transfer. In addition, each element of the system can be a source or destination in a transfer. Moreover, it permits the direct transfer of energy between sources and it only presents a single inductor which will impact the weight of the vehicle. Finally, due to the arrangement of the switches, it can be said that it allows the interaction with the grid to charge the battery and UC and to deliver energy from these devices.

Likewise, a detailed analysis of the operation of the converter was presented in the operating modes of interest for the application. A control strategy based on different levels of hierarchy was implemented to achieve the proper performance of the system. The regulation of the current of the converter was achieved by an SMC, while the voltage was regulated with an on-off controller. The MPPT was realized in the PV using two different converters, based on the P&O algorithm. By using

two converters it simplifies the control system. Meanwhile, the MPPT in the PMSG was accomplished by controlling the voltage on the capacitor of the direct current bus. In addition, an EMS was developed to decide the corresponding transfer at each time.

Because of its versatility, this converter is not only limited to its application in electric or hybrid vehicles. It can also be used in smart grids, microgrids, as a resource in distributed energy systems, battery charging management systems and so forth, in any application where the interaction of two or more sources is needed with the possibility of bidirectional power flow.

For future works the overall control system will be redesigned; an EMS based on fuzzy logic, which will increase the efficiency of the system and it will be more robust in situations that were not considered and a stability analysis will be developed. Moreover, a PFC strategy needs to be implemented for the charging of the power sources, and a control strategy that permit the boosting of the input voltage needs to be designed for the injection of current into the grid in order to be able to use the battery. Finally, experimental results that validate the system need to be performed, where it is assumed that the main constraints will be associated with the switching frequency in the current control loop of the *L*<sup>2</sup> inductor and in the transitions between the EMS states.

**Author Contributions:** Conceptualization, A.A.-D. and H.L.; methodology, A.A.-D., H.L. and R.V.C.-S.; software, A.A.-D., H.L. and N.V.; validation, A.A.-D., H.L., R.V.C.-S., J.R.-R. and N.V.; formal analysis, H.L., R.V.C.-S., J.R.-R. and N.V.; writing—original draft preparation, A.A.-D., H.L., R.V.C.-S. and J.R.-R.; writing—review and editing, G.H.-R.; supervision, R.V.C.-S., J.R.-R. and N.V.

**Funding:** This research was funded by the "Consejo Nacional de Ciencia y Tecnología (CONACYT)".

**Conflicts of Interest:** The authors declare no conflict of interest.

### **References**


c 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### *Article* **Large Scale Renewable Energy Integration: Issues and Solutions**

### **G. V. Brahmendra Kumar 1, Ratnam Kamala Sarojini 1, K. Palanisamy 1,\*, Sanjeevikumar Padmanaban 2,\* and Jens Bo Holm-Nielsen <sup>2</sup>**


Received: 16 April 2019; Accepted: 21 May 2019; Published: 24 May 2019

**Abstract:** In recent years, many applications have been developed for the integration of renewable energy sources (RES) into the grid in order to satisfy the demand requirement of a clean and reliable electricity generation. Increasing the number of RES creates uncertainty in load and power supply generation, which also presents an additional strain on the system. These uncertainties will affect the voltage and frequency variation, stability, protection, and safety issues at fault levels. RES present non-linear characteristics, which requires effective coordination control methods. This paper presents the stability issues and solutions associated with the integration of RES within the grid.

**Keywords:** renewable energy sources; energy storage system; voltage; frequency; grid integration

### **1. Introduction**

The majority of countries across the world are concentrating on RES, due to an increase of environmental concerns and depletion of fossil fuels [1]. For example, China is aiming for RES to represent somewhere around 35 percent of consumption by 2030. India has set an ambitious RES target of 175 GW [2]. The European Union and the United States also have targets regarding RES [3]. Some countries successfully integrated large shares of RES in the power grid in 2017, and most countries have set targets to receive their power generation from RES by 80% in 2050, as is shown in Figure 1.

**Figure 1.** Renewable energy share and targets by countries in 2017 and 2050 [4].

Due to the irregularities and fluctuating features of RES, new challenges to the grid over a unified anticipated generation have been created. The RES outputs are influenced by meteorological conditions. These characteristics affect the power system efficacy, power quality, system reliability, load management, security, and safety in various ways and are also very significant factors to be viewed in the integration of RES within the grid [6]. This influence is commonly observed with solar and wind energy, but geothermal, hydropower and biomass energy resources are more anticipated and have irrelevant difficulties in their association with the grid [7]. The comparison of RES with a synchronous generator (SG) is shown in Table 1.


**Table 1.** The comparison of renewable energy sources (RES) with synchronous generator (SG) [5].

The RES integrated with the grid is shown in Figure 2. Figure 2 is incorporated with RES, grid and electronic components. The power generation from RES to the grid is unidirectional. The RES requires the converters to be connected to the grid. These converters will achieve efficient operation and obtain the energy quality requirements related to the harmonics level. Also, this allows the integration of RES in a power inverter with high energy features and safety.

**Figure 2.** RES integration with grid.

The remainder of this paper is organized as follows: Section 2 considers the grid integration issues with a high share of RES, and solutions for variable RES is discussed in Section 3. The Energy Storage System (ESS) support is explained in Section 4, and followed by smart grid features in voltage control with RES in Section 5. Section 6 is the conclusion.

### **2. Grid Integration Issues with a High Share of RES**

As the power generation from RES increases, the installed capacity of power converters increases. By utilizing the large scale of solar energy, the RES system is able to supply power to the grid. Thus, by increasing the capacity of the RES, the grid-connected RES will create a negative impact on the grid when fault or disturbance occurs. Hence, the Point of Common Coupling (PCC) voltage drop is triggered by a fault in grid power and the RES will become off-grid across a wide area range. Moreover, these fault impacts cause the grid voltage and frequency collapse, also affecting the safe, stable and reliable operation of the grid and even triggering large economic losses [8].

### *2.1. Impact of Large-Scale Integration of RES on Frequency*

The frequency of a power system must be preserved near to its nominal value (either 50 Hz or 60 Hz based on the grid). The frequency deviations will only arise when there is a mismatch between generation and load. A stiff power system preserves the frequency subsequent to a contingency event [9].

The frequency of the power system is maintained at the nominal value only when the active power of generation and demand is balanced as indicated in Figure 3. If the demand is more than the generation, then the frequency decreases from the nominal value. In the case of surplus generation, the system frequency increases. The kinetic energy (KE) stored in the rotor of the rotating machines present in the power system contribute to inertia. The inertia in the power system regulates the frequency in demand generation imbalances. If the inertia is more in the power system, then it is less sensitive to small power imbalances. The inertia in the power system provides energy for some time to reduce the frequency deviations.

**Figure 3.** Power balance in a power system network.

In order to balance the power generation and demand, several control techniques are employed in a power system in multiple time frames as shown in Figure 4. The frequency response of a power system is comprised of the following steps [10]:


When the power imbalance occurs, the frequency starts to fall. The Rate of Change of Frequency (ROCOF) is more at the initial stage. ROCOF is associated with how much inertia is present in the system; the inertia slows down the initial frequency deviation, and this is called the inertial frequency response. The governor response is the primary frequency technique which acts within the first few seconds (typically 10–30 s) after a frequency event and seeks to reduce the frequency deviation. As the frequency reaches a minimum value (the frequency nadir); then the governor control is activated. Hence, the active power output increases and the frequency settles at a point slightly below the nominal value [11].

As the penetration level of RES increases, the frequency deviations are more frequent. As these RES are connected to the grid through a power electronic inverter, substituting the conventional SG with power electronic inverters will decrease the inertia of the power system. To handle the frequency stability issues raised from the low inertia and reserve power, new frequency control techniques need to be employed for RES to participate in a frequency regulation process.

**Figure 4.** Frequency response in a power system network.

### *2.2. Voltage Rise and Fluctuation*

It is considered that electricity is distributed at the consumer's terminal within the tolerable limit. The normal allowable voltage range is ±6% of the nominal value [12]. When large RES is integrated with lightly loaded feeders, the impact is unbearable. Whenever there is any change in load, the voltage fluctuation occurs at PCC. The voltage fluctuations and dips occur due to earth leakage faults and earth short-circuits located in the electrical power system (EPS). These faults weaken the voltage quality at PCC, based on fault situations. This is a significant element, particularly for solar and wind energy sources which possess irregular characteristics due to the wind speed disparity and solar irradiance that changes with time. The sensitivity of both electrical and electronic appliances that contributes to the life span deficiency of maximum devices is due to the voltage deviation [13].

### Impact of RES on Voltage Drops in the Grid

Voltage is a significant factor in EPS. Hence, voltage control requirements are essential in EPS for both transmission and distribution levels. The source voltage and voltage drop at the feeder is determined at the end of the feeder [14]. The voltage drop in the feeder is due to the conductor impedance, current flow, and load. Also, the drop should not be lower when in a peak load situation, and it should not be more than the maximum voltage in a light load condition.

$$
\Delta V = V\_1 - V\_2 = \frac{R\_{LN}(P\_L - P\_G) + X\_{LN}(Q\_L - Q\_G)}{V\_2} \tag{1}
$$

Equation (1) shows the injection of reactive power from RES within the grid; the RES will continually diminish the drop at the feeder. Where Δ*V* is the voltage between bus 1 & 2, *P* and *Q* are active and reactive power, and *RLN* & *XLN* are the line resistance and reactance, respectively. The grid is incorporated with RES and the load is shown in Figure 5. If the supply power is lower than the demand, the RES will inject power into the grid. Hence, this process is subject to RES active/reactive power that is relevant to the load active/reactive power and line X/R ratio [15].

**Figure 5.** Grid connected system with RES and load.

### **3. Solutions for Variable RES**

Various control techniques need to be employed to increase the power generation from RES and to decrease the negative impact of the RES on the power grid. This section gives a brief overview of the frequency and voltage control techniques for RES.

### *3.1. Frequency Control Techniques for RES*

The frequency control techniques employed for RES are shown in Figure 6. The frequency control techniques used for both wind and solar are discussed in this section. For RES-based power plants, the frequency can be controlled by reserving the active power by using the de-load operation or by using the energy storage system (ESS). Generally, Inertia and frequency control methods for RES are categorized in two ways: control techniques to de-load the RES, and control practices for RES with ESS.

**Figure 6.** Frequency control techniques for RES.

### 3.1.1. Control techniques Used for Wind Turbines

The frequency control techniques for the wind turbine can be classified in three ways, i.e., 1. Inertia control, 2. Droop control, 3. De-loading control.

### 3.1.2. Inertia Control

The inertia of the wind turbines needs to be emulated to maintain the frequency stability under the high penetration of wind power generators. The inertia of the wind turbines can be emulated by either "hidden" inertia emulation or by the fast power reserve emulation.

### (a) Hidden inertia emulation

The emulation of hidden inertia present in the wind turbines is used to reduce the frequency deviation and to maintain the frequency stability. The suitable control algorithm for the power electronic converter for the wind turbines allows the wind turbine to release the KE stored in the rotating blades. The KE stored in the blades of the wind turbine helps to regulate the frequency under unbalance condition through its inertial response [16]. There are two ways to emulate the inertial

response, either by considering the ROCOF alone or by incorporating both ROCOF and frequency deviation. Figure 7 shows the control diagram for the inertia emulation, considering only ROCOF to release the KE stored in the wind turbine. The inertia constant (*H*) is used to express the inertial characteristics. The "hidden" inertia of wind turbines can be known as,

$$H = \frac{J\omega\_{nom}^2}{2S} \tag{2}$$

where *J* the inertia of the wind turbine is, *S* is the VA rating of the machine, ω*nom* is the nominal angular frequency.

**Figure 7.** Hidden inertia emulation control [17].

The inertial active power control signal *Pinertia* of the hidden inertial emulation control is known as,

$$P\_{inerti} = 2H \* \omega\_{sys} \* \frac{d\omega\_{sys}}{dt} \tag{3}$$

where ω*sys* the angular frequency of the system is, ω*<sup>r</sup>* is the reference angular frequency.

The inertia control algorithm with both ROCOF and frequency deviation is shown in Figure 8. In the power balanced condition, the active set power is monitored by the MPPT controller. Under power imbalance/frequency disturbance conditions the control algorithm acts and produces the extra inertial power signal. In this controller, the inertial power is calculated from both the ROCOF loop and frequency deviation loop. The inertial power calculated from the Figure 8 is known as,

$$P\_{\text{inertia}} = K\_I \frac{df}{dt} + K\_{droop} \Delta f \tag{4}$$

where *KI* and *Kdroop* are the inertia and drooping gains respectively.

**Figure 8.** Inertia emulation control with droop control [18].

The wind generators can store and release KE instantly compared to the SG, due to the power electronic converter controller. The variable speed wind turbines can participate more in the frequency regulation by releasing more KE than fixed speed wind turbines and SG [19].

### (b) Fast Power Reserve

Usually, the emulated inertia for the wind turbines can be determined based on the frequency deviation or ROCOF, as illustrated in the above section. Whereas the fast power reserve is defined as the constant active power support regardless of the wind speed [16]. The fast power characteristics are shown in Figure 9. The fast power reserve is the temporary power, released from the KE stored in the rotating blades of the wind turbine.

**Figure 9.** Fast power reserve characteristics.

The control diagram of fast power reverse is shown in Figure 10. This fast power reserve can be realized by regulating the set value of the rotor speed. This is given by,

$$P\_{\text{Const},t} = \frac{1}{2}J\omega\_{r,0}^2 - \frac{1}{2}J\omega\_{r,t}^2\tag{5}$$

where *t* (*t* < *t*max) is the lasting time of the fast power reserve since the beginning of the frequency event, ω*r*,0 is the rotor speed at the start and ω*r*,*<sup>t</sup>* is the rotor speed at *t*, *PConst*,*<sup>t</sup>* is the constant active power available at *t*, hence the refence angular speed in this method can be calculated as ω*re f* .

$$
\omega\_{ref} = \sqrt{\omega\_{r,0}^2 - 2\frac{P\_{\text{Const,t}}}{J}} \tag{6}
$$

The fast power reserve control operates when the frequency deviation is more than the threshold value. This control delivers the extra power in the frequency event through the KE stored in the rotor and can be known as "over-production". To recover the KE after the event is known as "under-production". The shifting of an over-production state to an under-production state must be in a sloped manner, to avoid a sudden dip in the active power [20].

**Figure 10.** Fast power reserve control.

### 3.1.3. Droop Control for Frequency Regulation

The droop control of a wind turbine adjusts the active power response based on the frequency deviation. The droop controller shown in Figure 11 decreases the frequency nadir. Regulated active power and frequency is linearly related and it is shown in Figure 12.

**Figure 11.** Droop control for wind turbine.

When the frequency drops from *fref* to *fcal*, the wind generator increases the output of power from *P*<sup>0</sup> to *P*<sup>1</sup> to regulate the frequency deviation. Hence, the active power regulated by the droop control can be given as,

$$
\Delta P = P\_1 - P\_0 = -\frac{f\_{\text{mens}} - f\_{\text{nom}}}{R} \tag{7}
$$

where *R* is the droop coefficient, *fmeas* and *P*<sup>1</sup> are the measured frequency and wind turbine output power, respectively, while *fnom* and *P*<sup>0</sup> are the initial operating points.

The analytical method for estimating the influence of inertia and droop responses from wind for frequency control was presented [21]. The droop control for wind turbine is shown in Figure 12. In order to use the wind turbine in frequency regulation under higher penetration, the adaptive gains for both the inertia, and droop controller was proposed in [22] by Yuan-Kang Wu. The inertia and droop gains are altered from time to time depending on the frequency imbalance.

**Figure 12.** Droop characteristics.

### 3.1.4. De-Loaded Operation in Wind Turbines

The control techniques presented in this section would help to eliminate the adverse impact of the higher RES penetration level on the frequency stability. This section addresses the inertia and frequency control techniques with the de-loaded operation of RES. The reserve power in RES-based generators from de-loaded operation can be utilized for the inertial response, and primary frequency support. Generally, the wind turbines are operated with a maximum power point tracking technique and this does not contribute to frequency regulation. To maintain the stability of the power system, the wind turbines need to participate in the frequency regulation with the increasing penetration level of wind in the power system. The de-loaded control techniques enable the wind turbines in frequency regulation. The de-loaded operation of wind generators for the fast frequency reserve was initially proposed in [23]. Some of the authors [9] proposed maintaining the active power reserve for high wind speeds, and the reserve being unavailable in lower wind speeds. Another way to maintain the reserve power for the wind generators is by de-loading them in low wind speeds. Although, the reserve is not available above the wind-rated speeds [24]. The author of [25–28] proposed the de-loaded operation of a wind generator over the entire speed range, either in lower wind speeds or in higher wind speeds. The power and rotor characteristics for the de-loaded operation of wind turbine are shown in Figure 13.

**Figure 13.** De-loaded operation for wind turbine.

### (a) Speed Control

The de-loading operation of the wind turbine can be realized by changing the operating point from the maximum power point (*Pmpp*) to a sub-optimal power point (*PSuboptimal*). The power can be varied from *Psub* to *Pmpp* by regulating the rotor speed. The speed control method controls the tip speed ratio (λ) and it is known as,

$$
\lambda = \frac{\omega\_r R}{v} \tag{8}
$$

where *R* the rotor radius and ν is is the velocity of wind.

By controlling the speed ratio (λ) the operating point of the wind turbine can be altered and is shown in Figure 14.

**Figure 14.** Speed control for wind turbine.

Therefore, the reference power given for the wind turbine can be known as [29],

$$P\_{ref} = P\_{del} + \left(P\_{\text{max}} - P\_{del}\right) \ast \left[\frac{\omega\_{r,ld} - \omega\_{r,\text{max}}}{\omega\_{r,dd} - \omega\_{r,\text{max}}}\right] \tag{9}$$

where *P*max is maximum power, *Pdel* is de-loaded power, ω*r*,max is rotor speed at *Pmax*, ω*r*,*del* is de-loaded rotor speed at *Pdel*.

In the rotor speed control of wind turbines, there are two different possibilities to regulate the active power output: over-speeding and under-speeding of the turbine. In case of over-speed mode, the wind turbine delivers the KE until the operating point has reached the maximum power point (*PMPP*). Whereas, in under-speed mode the wind turbine absorbs the KE until the operating point has reached the maximum power point [30]. The under-speeding of the rotor control results in stability problems. Hence, over-speeding of the rotor control is used.

The active power reference from Equation (9) uses a de-loaded power extraction curve shown in Figure 14. This technique regulates output of the active power from the wind turbine under a frequency event. Generally, the speed control technique is appropriate for low wind speeds.

The rotor speed control of the wind turbine is presented in [23] and regulates the active power output under a frequency event. In this paper, the author utilized the optimum power extraction curve to regulate the active power of a wind turbine. The power transferred to the grid is dependent on the slip for a doubly-fed induction generator. If the slip is increased, then the power transferred from the wind generator to the grid is increased.

(b) Pitch Angle control

Another type of de-loading control for the wind turbines is created by changing the blade angle, known as "pitch angle control". The pitch angle control can be applied to both variable speed and fixed speed wind turbines. Figure 15 illustrates the power-speed characteristics of a wind turbine for different pitch angles.

**Figure 15.** Pitch angle controller.

To de-load the wind turbine in pitch angle control, the pitch angle (beta) needs be increased to receive the active power reserve which is shown in Figure 16. Whenever the frequency event occurs, the pitch angle controller increases the value of beta and the operating point without changing the rotor speed. In pitch angle control, the pitch angle of the wind turbine is set at a sub-optimal point and reaches optimum value under frequency events to deliver/absorb active power. Pitch angle control is suited to high wind speed conditions [31].

Generally, the de-loading control can be selected depending upon the wind speeds. In low wind speeds, the de-loading is realized by the rotor speed control. The coordinated control of both pitch angle and rotor speed control is needed in medium wind speed conditions. The pitch angle control alone is used in high wind speed conditions.

**Figure 16.** De-loaded curves with pitch angle control for wind turbine.

3.1.5. Solar Photovoltaic Array in Frequency Regulation

The installed photovoltaic (PV) generators do not contribute to the active reserve power for frequency regulation. All the PV systems are operated at the maximum power point using MPPT algorithms and do not alter their active power output when load variation occurs. The PV systems also need to participate in the frequency regulation to maintain the power system stability, with an increasing penetration level of PV system in the power system. The frequency regulation is employed in two ways for the PV generators:


In solar photovoltaic arrays, both the inertial and primary frequency response can be implemented by using de-loaded operation of PV and ESS with proper control algorithms.

### 3.1.6. De-Loaded Operation of a PV Generator [32]

In the de-loaded operation of PV, the power electronic converters need to inject the required amount of power under imbalance conditions depending on the control signal generated in Figure 17.

**Figure 17.** De-loaded operation control for solar photovoltaic (PV).

The solar power plants need to be operated under a sub-optimal power point below the maximum power point to maintain the active power reserve for the PV plants, as is shown in Figure 18. When the frequency event occurs, the operating voltage of the PV is decreased from the *Vdel* to *VMPP*, to increase the active power output. This reserve power will not be released until the frequency is deviated. When the frequency deviation occurs, the DC-link voltage is changed based on the frequency. The active power reserve available under de-loaded operation is known as,

$$P\_{reserve} = P\_{MPP} - P\_{del} \tag{10}$$

The voltage reference calculated in the de-loaded operation is,

$$V\_{dc,ref} = V\_{MPP} + V\_{del} - V\_{dc} \Delta f \tag{11}$$

where, *Vdc* is the DC-link voltage.

**Figure 18.** Frequency response in power system network.

### *3.2. Voltage Control Techniques*

The control methods for voltage are discussed in two ways: (1) Low-voltage ride through (LVRT) and (2) High-voltage ride through (HVRT). The power disturbance in the grid is higher because of the output power from RES is irregular. Hence, LVRT should be attributed to the large capacity of grid-connected RES. The main purpose of LVRT is restricting the PCC current of the RES inverter, and power elements of RES inverter should not be damaged and tripped. The RES inverter is operated at 1.1 times the rated current in a short time, and it can supply a reactive current to support the grid voltage recovery. Thus, the control strategy of LVRT for RES is particularly dependent upon controlling the current output of the inverter during a grid fault [33]. Hence, to improve the power quality and stable operation of the grid, the RES is associated with the power system to serve the grid recovery from grid fault and to sustain the regulation of grid voltage and frequency.

Developing the penetration of distributed generation (DG) systems into the grid, one of the major problems is sudden tripping of DG from the grid during faults occurring in the system such as power outages and voltage flickers. Hence, DG units need to support the grid during fault conditions. The reactive power support diagram is shown in Figure 19. According to the theory of instantaneous reactive power, the active and reactive currents are controlled by varying the amplitude and phase of the output voltage of the inverter. In daytime situations, the active power output and the reactive power compensation (RPC) of the system are obtained concurrently [34,35]. When the PV power is not available, the RPC characteristic of the RES can be employed to enhance the utilization factor of the system.

The main factors for RPC in a system are:


During the LVRT, the RES has the ability to be associated with the grid when the voltage at PCC drops to a prescribed cut-out point. When the drop is below the prescribed point, the RES can switch from the grid. Hence, a certain amount of reactive power is needed to provide effective voltage support to overcome this fault situation [36].

**Figure 19.** Reactive power support diagram.

There are various causes for the PCC voltage to be above/below rated values, such as the load has been suddenly disconnected, earth faults, and the irrational control strategies. During HVRT conditions, The RES should consume a required reactive power in order to make the PCC voltage become lower. The amount of absorbing power is determined by the capacity of the inverter and voltage at the above level of the prescribed cut-off point. Hence, the active power supplied from RES should not be changed, so the RES inverter can utilize its capacity to consume reactive power [37].

*Energies* **2019**, *12*, 1996

The voltage requirements during the VRT at PCC are shown in Figure 20. The HVRT handles the grid faults, which depends on voltage rise. The HVRT functionality is used for each of the three voltage phases. Hence, the functionality is actuated when the voltage rises higher than the voltage level determined by the utility grid. The control requirements for LVRT and HVRT equations are presented in Equations (12)–(14).

**Figure 20.** Voltage requisites during the low-voltage ride through (LVRT) and high-voltage ride through HVRT at PCC [38].

The equation is to satisfy the demand requirement as follows,

$$I\_d^2 + I\_q^2 \le I\_{\text{max}}^{\prime 2} \tag{12}$$

where *Imax* is maximum reference current from RES system, and *Id* and *Iq* are the active and reactive currents. During a fault situation, the reactive support should meet the demand of Equation (12), and the active and reactive currents are set as follows,

$$I\_d = \left\{ \begin{array}{c} I\_{d, \text{set}} \\ \text{or } \min\left(I\_{d0,} \sqrt{I\_{\text{max}}^2 - I\_q^2}\right) \end{array} \right. \tag{13}$$

$$I\_{q,refH} = \frac{\sqrt{I\_{\text{max}}^2 - \left(K I\_{d,refH}\right)^2}}{K} \tag{14}$$

where *Id*<sup>0</sup> is the pre-fault current and *IdrefH* is the reactive reference current. The active power control activities are presented in Table 2.


**Table 2.** Active power control activities [38].

The control strategy of LVRT and HVRT is shown in Figure 21. In this [38], the control strategy is decided by the voltage at PCC. The *Idref* and *Iqref* are generated various working positions, and in that way to understand the LVRT and HVRT.

**Figure 21.** Control diagram of the LVRT and HVRT [38].

### **4. ESS Support**

The above section considers the various controllers for both solar and wind and discusses the maintenance of the power system stability without ESS. RES is intermittent in nature and may lead to stability issues if we completely rely on these sources for frequency stability. Hence, ESS needs to be used along with the RES to maintain the stability of power system under high penetration of RES. The frequency of the power system can be balanced by using an energy storage device. Whenever there is an imbalance between the demand and generation, the control algorithm acts at the ESS to deliver the required amount of the active power. The ESS support for RES issues diagram is shown in Figure 22.

The performance of ESS in the smart grid is an effort towards balancing generation, consumption, reduction of variations in RES, and also delivers high quality supply and reliability of the supply. There has been an evolution of the enhanced intelligent electricity network called "smart grid", and it has a prominent contest of supporting all the sources connected with effective load control, powered from large penetration of non-dispatchable RES. These irregular energy sources are given higher levels of grid perception. Hence, ESS is the most significant smart grid advanced element to produce stable power at the grid level. ESS produces vital benefactions in defeating the complexity of irregular variation created by RES. It compensates the variation and lessens the variance among supply and demand [39]. ESS consists of pumped hydro-storage, compressed air energy storage (CAES), flywheel, superconducting magnetic energy storage (SMES), battery, and capacitors that are supposed to be widespread in RES integrated to the grid. ESS employs a power transformation system to integrate into the system network; they can add or receive both active and reactive power to balance for voltage changes in a short or medium duration. The ESS integration with the grid will increase the power quality, reliability, voltage support, backup power and decrease losses. When the irregular energy source reaches higher levels of grid penetration, ESS is the solution for providing a reliable energy supply [40,41]. Hence, proper coordination is needed for micro-grid (MG) and ESS operations. The main control methods among MG and ESS are the voltage, frequency, and active and reactive

power controllers. Thus, ESS can sustain a balance within the generation and load at the instant of operation, and can also produce a firming potential of RES [42].

**Figure 22.** Energy Storage System (ESS) support for RES issues.

### **5. Smart Grid Features in Voltage Control with RES**

Smart grid is being increasingly used in practice by electric utilities and involves the use of information, communication and control abilities to enhance the performance of the grid. The fundamental concept of a smart grid involves developing analysis, control, monitoring abilities of the traditional grid system, and reducing energy consumption. These objects are feasible over a system that can yield accurate and precise monitoring of the grid. A promising selection to minimize the challenges created by variations in RES is to add ESS into the grid. Another approach is to achieve flexible operation in energy consumption by employing demand side integration (DSI) or the usage of the micro-grid system. Moreover, the RES, ESS, and DSI are listed as one of the DERs. Hence, combining various aspects of certain sources is collectively essential in enhancing the utility of RES in the energy market [43].

### **6. Conclusions**

The large-scale integration of RES into the power grid will have an impact on the stability of the power system. Both the frequency and voltage stability are highly affected by integrating large scale RES into the grid. This paper presented the issues related to high integration of RES in detail and reviewed their possible solutions available in the literature. To minimize frequency and voltage related issues, the several control techniques that are applied to wind and solar-based generators are summarized. Furthermore, ESS support could be used to maintain the stability and reduce the negative impacts related to grid integration as discussed.

**Author Contributions:** Initial idea purpose, data collection, formal analysis, original draft writing and editing: G.V.B.K. and R.K.S.; resources, supervision, review and editing: K.P., P.S., and J.B.H.-N.

**Funding:** No source of funding was attained for this research activity.

**Acknowledgments:** The authors would like to acknowledge the support and technical expertise received from the center for Bioenergy and Green Engineering, Department of Energy Technology, Aalborg University, Esbjerg, Denmark which made this publication possible.

**Conflicts of Interest:** The authors declare no conflict of interest.

### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### *Article* **A Hybrid PV-Battery System for ON-Grid and OFF-Grid Applications—Controller-In-Loop Simulation Validation**

### **Umashankar Subramaniam 1, Sridhar Vavilapalli 2, Sanjeevikumar Padmanaban 3,\*, Frede Blaabjerg 4, Jens Bo Holm-Nielsen <sup>3</sup> and Dhafer Almakhles <sup>1</sup>**


Received: 24 December 2019; Accepted: 7 February 2020; Published: 9 February 2020

**Abstract:** In remote locations such as villages, islands and hilly areas, there is a possibility of frequent power failures, voltage drops or power fluctuations due to grid-side faults. Grid-connected renewable energy systems or micro-grid systems are preferable for such remote locations to meet the local critical load requirements during grid-side failures. In renewable energy systems, solar photovoltaic (PV) power systems are accessible and hybrid PV-battery systems or energy storage systems (ESS) are more capable of providing uninterruptible power to the local critical loads during grid-side faults. This energy storage system also improves the system dynamics during power fluctuations. In present work, a PV-battery hybrid system with DC-side coupling is considered, and a power balancing control (PBC) is proposed to transfer the power to grid/load and the battery. In this system, a solar power conditioning system (PCS) acts as an interface across PV source, battery and the load/central grid. With the proposed PBC technique, the system can operate in following operational modes: (a) PCS can be able to work in grid-connected mode during regular operation; (b) PCS can be able to charge the batteries and (c) PCS can be able to operate in standalone mode during grid side faults and deliver power to the local loads. The proposed controls are explained, and the system response during transient and steady-state conditions is described. With the help of controller-in-loop simulation results, the proposed power balancing controls are validated, for both off-grid and on-grid conditions.

**Keywords:** battery; cascaded H-Bridge; chopper; energy storage; multi-level; PV inverter

### **1. Introduction**

A low voltage (LV)-rated solar PCS containing a conventional inverter with two-level topologies is preferable for PV systems rated for lower power. For higher power solar power stations, it is better to opt for the system with medium voltage (MV) rating. Multi-level inverters (MLI) are more suitable for MV applications. The cascaded H-bridge (CHB) inverter is a popular MLI configuration which requires isolated DC sources/DC links. Hence CHB configuration is highly suitable for static compensator (STATCOM) and solar applications due to the availability of isolated DC links [1]. CHB inverter needs multiple H-bridge modules, and to control the multiple H-bridge modules, the required input-output channels in the processor are more when compared to other MLI configurations, but in MV high power systems, the CHB inverter enables the independent maximum power point (MPP) controls to attain enhanced efficiency [2]. In high power applications, the number of H-bridge modules to be used in CHB is more which results in better power quality [3,4]. The transformer on the AC side can also be eliminated with this configuration.

Solar PV stations can reduce carbon emissions and provide clean energy but may not be able to supply the load requirements due to sudden changes in weather conditions and when the solar irradiation is weak. The system remains in an idle state during nighttime, which affects the utilization factor of the system drastically. Hence there is much attention to battery energy storage systems along with PV to reduce power disturbances in the system, to improve the stability, for providing continuous power to the load and for improving utilization factor of the system. In such hybrid PV-battery stations, power is transferred from PV array to battery and load during daytime and batteries transfer power to the load during night time. The battery storage system also improves the system dynamics for sudden weather changes [5,6]. Importance of hybrid PV-battery stations and operation strategy is discussed in [7]. A review of energy storage for large-scale PV systems, grid integration issues, stability concerns and the selection of batteries is available in [8]. With the battery storage system in PV applications, the utilization factor of the PCS can also be improved since the system can be made operational for all the time. To keep the grid power non-negative always, an approach called 'Solar Plus's which is the combination of energy storage, PV and load controls are presented in [9].

Cost of the battery storage systems is also decreasing over time due to the advancements in the areas of different battery technologies, battery charging and discharging methods. In [10], various battery types such as lithium-ion, lead-acid, aluminium–ion, sodium-sulphur (NaS), flow batteries, etc. suitable for large-scale PV ESS systems are explained and compared. Various discharge strategies are presented for grid-connected hybrid PV-battery applications in [11] and different battery storage technologies suitable for residential applications are also elaborated.

In a conventional grid-connected PV power conditioning system, the anti-islanding feature will be incorporated as a feature of protection. But as mentioned earlier, it may be required to operate the PCS in stand-alone mode also when there are frequent disconnections from the grid. In such scenarios, a power management system at a higher level of control architecture needs to isolate the grid from the critical loads.

Hybrid PV-battery stations can also be operated in standalone mode; hence it is possible to provide power to local critical loads during grid-side faults or during maintenance on the grid side. Such systems help in catering continuous power supply in remote locations which are not connected to the central grid or which are facing problems of regular power supply failures from the grid. Optimal design for a PV + Diesel + Battery storage hybrid system is presented in [12].

From the above, the following research gaps are identified:


Present work mainly focuses on medium voltage CHB-based power condition system to meet the large-scale PV system requirements. For the uninterrupted power supply and to improve the utilization factor of the system, battery energy storage is incorporated in to the PV power system. The following are the contributions of present work to address the research gaps in earlier systems:


This paper is organized, as explained below: In Section 2, ESS configurations suitable for CHB inverter and advantages of chopper-based ESS are explained. Operation of buck chopper and bi-directional chopper are also elaborated in this section. Section 3 covers the procedure for selection of components such as PV array, batteries, filter components and other accessories. Controls adapted for inverter and battery charger are discussed thoroughly in Sections 4 and 5, respectively. The setup for the controller-in-loop simulation validation is explained in Section 6 and results are presented in Section 7. Contributions and the conclusions of the present work are presented in Sections 8 and 9, respectively.

### **2. Chopper-Based ESS for CHB Inverter**

An AC coupled ESS configuration suitable for CHB inverters based on a voltage regulator is presented in [13], but AC-side coupled configurations are not suitable for higher rated energy storage systems as the power quality may get affected due to the multiple interfaces connected on the grid side. DC-coupled systems are more suitable for large-scale energy storage systems and for obtaining better power quality. A dual active bridge (DAB)-based ESS configuration is presented in [14], it is it has been observed that independent power control through PCS and battery charger is possible with this configuration but the control is complex due to a greater number of power modules. Since each DAB needs eight gate pulses, the controller hardware requirements are more. To reduce the cost, controller hardware requirements and the control complexity, it is preferable to select chopper-based systems. In this work, two types of choppers namely buck-chopper and bi-directional choppers, are discussed to operate as a battery charger.

Figure 1 shows the circuit diagram for a buck chopper-based battery charger. In this system, PV voltage shall be more than battery voltage during charging mode. By regulating the duty cycle of IGBT-S1, the charging current is regulated during charging mode. The output filter is used to limit the ripple content of output current and voltage. During the nighttime, PV voltage tends to be less than the battery voltage. Then battery starts giving power to DC link through the third diode 'D3 . The changeover is instantaneous; hence the system response for transient conditions is also instantaneous. After the transition, the DC voltage is clamped to battery voltage. Since the above charger is operational in charging mode alone, the battery charger and filter inductor need to be selected for the charging current, i.e., usually less than the battery rated current. This reduces the cost of the system. Since the discharging of the battery is through D3, this diode is selected for full current to handle current during discharging mode.

**Figure 1.** Battery charger based on a buck-chopper.

The main drawback with this type of charger is that the battery starts discharging only when the irradiation is weak. In grid-connected mode, the battery cannot transfer power to DC link even when the local load demand is higher than the power available at PV array since the PV array is operating at its MPP voltage which is more than the battery voltage. In this case, the power required for the local load needs to be taken from the grid. In standalone mode, the battery needs to provide power along with the PV source for the local critical load requirement due to the absence of the grid. In this case, the PV array cannot be operated at its MPP voltage but operates at battery voltage which reduces the efficiency of PV [15]. So, this charger is more suitable for grid applications. For standalone PV systems, it is required to have control during the discharging mode of operation also which is possible through bi-directional chopper based ESS.

Figure 2 shows a battery charger based on the bidirectional chopper. Unlike the buck chopper, the battery voltage can also be higher than the DC link in this configuration. Battery terminals need to be connected to output terminals of the charger in case the battery voltage is less than DC link. Similarly, battery terminals need to be connected to input terminals of the charger in a case when the battery voltage is more than DC link. This DC-DC converter needs two gate pulses, whereas buck-chopper based system requires only one gate pulse. Cost of this system is less than the DAB based system but slightly more than the buck-chopper based system since the battery charger is selected to the full capacity of the battery. With this charger, battery current in discharging mode can also be controlled hence the PV source can always be operated at MPP voltage in both on and off-grid operations.

**Figure 2.** Bi-directional chopper-based battery charger.

### **3. Design Calculations for the System**

In present work, a high-power PV system with a solar inverter based on a CHB MLI and a bi-directional chopper-based battery charger is studied. To discuss the controls proposed at various operating points and in different modes such as grid-connected and standalone modes, a single-phase system with ratings given in Figure 3 is chosen. In this work, a single-phase system is considered as a case study due to limitations in the controller hardware. However, the system controls can be upgraded to the three-phase system, which is suitable for high power applications.

**Figure 3.** DC-coupled ESS for CHB-based PCS.

Design calculations for CHB inverters, selection of devices, device loss calculations etc. are discussed in [16,17]. The procedure for the selection of CHB inverter and AC side filter components is shown in Table 1.


**Table 1.** Design calculations for CHB Inverter and L-C-L Filter.

To cater to the power needed for local load and power to be transferred to the grid, the required power rating of CHB inverter is 350 kW. In this system, 4 H-bridges are selected hence the levels achieved on the output PWM voltage is nine. The voltage and the power ratings of each H-bridge are 1/4th of CHB ratings and the current rating is equal to the rated inverter output current i.e., 280 amperes. Minimum DC input required for each H-bridge is calculated based on the AC output. In this work, level-shifted PWM is adapted with a carrier frequency of 1 kHz. A filter is used at AC terminals of CHB MLI for obtaining better voltage and current THDs.

In this system, the minimum battery backup power is selected as 100 kW so that the critical load requirement can be met during the standalone mode of operation also. In this work, the battery is selected in such a way that the maximum voltage of the battery is less than the nominal voltage of PV array which is suitable for both the types of chopper based ESS configurations. The lithium-ion battery is selected with a minimum battery voltage depending on the minimum DC input required for each H-Bridge. Procedure for obtaining battery ampere-hour value is also shown in Table 2. After selecting the battery, the PV array is selected by considering that the nominal voltage of the PV array is always higher than the maximum battery voltage. Short circuit current and open-circuit voltage of the PV module is obtained from the datasheet. Minimum MPP voltage of PV array is at the maximum operating temperature range i.e., at 75 ◦C. Hence ten PV modules in the series are selected to meet the system requirements at 75 ◦C also. Procedure for selection of PV array is also explained in Table 2.


**Table 2.** Selection of battery and PV sources.

Input source to the battery charger is a PV array and the battery is connected to the output terminals. In the case of a buck chopper-based system, the battery charger rating is decided based on charging current which is obtained from the charging time. Ratings of bi-directional chopper-based battery chargers are obtained by a maximum of battery charging and discharging currents. In this work, battery charging and discharging times are selected equal, to maintain equal ratings for Buck-Chopper and bi-directional chopper-based battery chargers. The switching frequency of 5 kHz is selected for the IGBT based battery chargers. An L-C filter is connected to chopper output terminals. Design calculations for filter inductance and capacitance are also shown in Table 3.



### **4. Inverter Controls in Di**ff**erent Modes of Operation**

Figure 4 shows the building blocks of the CHB inverter which consists of PV switch, IGBT based H-Bridge, bypass module, and transducers. Each building block is fed by independent PV arrays and also connected to independent batteries through chopper-based DC-DC converters. Transducers in the H-bridge module measure PV array voltage and currents for independent MPPT controls. A bypass switch module consisting of antiparallel thyristors and a bypass contactor is used along with H-bridge. Bypass module bypasses the H-bridge during fault so that the system can continue to be operated at reduced power rating.

**Figure 4.** The basic building block of a CHB-based PCS.

In [18–20], regulation of power flow through CHB MLI based PCS are discussed. The control diagram for the power regulation through the inverter is as shown in Figure 5. In grid-connected mode, closed-loop current regulation is implemented and closed-loop voltage regulation is adapted during the standalone mode.

**Figure 5.** Control diagram for the CHB inverter in hybrid PV-ESS.

In the grid-connected mode of operation:


In standalone mode, closed-loop voltage control is adapted to provide the rated voltage to the load. For smooth changeover in the mode of operation, it is required to match the phase angle of inverter voltage with the earlier grid voltage. Following activities are carried out in the inverter controls when the system is in standalone mode:


### **5. Battery Charger Controls in Di**ff**erent Modes of Operation**

With a bi-directional chopper-based battery charger, regulation of battery current in charging and discharging modes is possible which allows the PV array to operate at its MPP in all operating conditions. In grid-connected mode, the power available at PV array is used for the charging of the battery and the balance power is transferred to the grid/load as long as irradiance is available. In standalone mode, the power available at PV array power is transferred to the grid/load and the balance power if any is used for charging the battery as long as irradiance is available. In both these modes, if the battery is fully charged or if the power demand on the grid side is more, then the battery can also feed the additional power demand. In this work, battery voltage less than PV voltage is selected, hence the converter acts like a buck-chopper during charging mode and acts like a boost-chopper during discharging mode. Battery charging is always through current control whereas discharging of the battery can be carried out either through the current controller or uncontrolled. Figure 6 shows the control diagram of the bi-directional charger.

**Figure 6.** Control diagram for the bi-directional chopper in PV-ESS.

In grid-connected mode:


In the standalone mode of operation:


In this work, a bi-directional chopper-based battery charger is considered for ease of understanding and due to limitations in the controller hardware. For high power applications interleaved buck-boost converters presented in [21] may be adapted with minor modifications in the control logic.

### **6. Validation of Control Algorithm for Chopper Based ESS Configurations**

To validate the system, controller-in-loop simulation validation is adapted. With the controller-in-loop simulations, the control software can be tested prior to the site trials and can be tested at various operating points which are difficult to test with a real plant [22]. The following activities are carried out as part of controller-in-loop simulation validation:


**Figure 7.** Block diagram of the controller-in-loop simulation setup.

Controller-in-loop simulation setup is shown in Figure 8, Table 4 lists the particulars of the real-time simulator hardware.

**Figure 8.** Controller-in-loop simulation setup.

**Table 4.** Hardware details of real-time simulator setup.


### **7. Results and Discussion**

In grid-connected mode, the advantage with the bi-directional chopper-based system is the possibility of discharging even when the PV array is active. The response of the inverter current is observed by operating the battery in charging mode and discharging modes intermittently. Irradiance is continuously maintained at 1000 watt per square meter hence the PV array current is constant. The battery reference current is varied from +50 Ampere to −50 Ampere and vice versa with an equal interval time of 1 s. When the battery current is negative, additional current is flowing through the inverter as shown in Figure 9. The dynamic response of the system in this condition is also found to be satisfactory.

**Figure 9.** Change in the battery, PV array and actual inverter currents with a change in reference battery current at constant irradiance of 1000 watt per square meter in a bi-directional chopper-based ESS in grid-connected mode.

Irradiance on the PV arrays is varied from zero watts per square meter to 1000 watt per square meter in steps of 200 watts per square meter and the currents of inverter output, battery and PV array are observed.

As discussed earlier, when the irradiance is zero, the battery current is negative as it is in discharging mode and the inverter power is equal to the rated battery power and PV current is zero as displayed in Figure 10. MPP tracking is carried out in inverter controls hence the inverter current and PV currents are varied in proportion to the irradiance inputs whereas the battery current is almost constant.

**Figure 10.** Inverter, battery and PV currents with varying irradiance value in grid-connected mode.

Irradiance on the PV arrays is varied from 350 watts/sq.m to 1000 watt per square meter in steps as shown in Figure 11 and inverter current is observed. When the grid side is healthy, the current required for charging the battery is provided by PV array and the remaining power is supplied to the grid/load. So, the current through inverter decreases with the reduction in irradiance. PV power is less than the power required for the battery charging when irradiance is 350 watts/sq.m; hence the inverter current is zero. As the irradiance value increases, the power transferred to the grid also increases and shows good dynamic during a sudden change in irradiance input.

**Figure 11.** Change in inverter current with varying irradiance value in grid-connected Mode.

A fault on the H-Bridge module is created to verify system performance. The system continues to operate with reduced power as shown in Figure 12. The proposed algorithm enables the fault-tolerant operation instantaneously and the obtained results are in line with the simulation results presented in [1].

**Figure 12.** Inverter PWM voltage and RMS current during a fault in one H-bridge module

Mode of operation is changed from grid-connected to the standalone operation and the transients in the inverter AC voltage i.e., the input voltage to the load are observed. Smooth transition in the inverter voltage is observed during mode transfer as shown in Figure 13. The operation mode is changed from grid-connected to an off-grid mode for a bi-directional chopper-based system. Operation of the system for a fixed load is studied by varying irradiance value from zero to 1000 watt per square meter in steps of 200 Watts/sq.m. Since the load is fixed, inverter current is constant for all the values of irradiance inputs. MPP tracking is carried out in battery charger controls; hence the battery current varies with the irradiance value. At the instant when the PV power is not sufficient to meet the load demand then discharge current of the battery is regulated through current control. In this mode, DC link voltage has maintained the value of the MPP voltage of PV array. As shown in Figure 14, when irradiance value is zero, PV array voltage tends to zero hence the DC link is clamped to the battery voltage level. From the presented results it is observed that the dynamic response of the battery charger system is good as the settling time is in the range of 50 milliseconds, whereas the settling time is in the range of 200 milliseconds in the battery charger systems based on voltage regulator and dual active bridge configurations proposed in [13,14], respectively.

**Figure 13.** Grid voltage, internal oscillator voltage and inverter output voltage during a changeover in the mode of operation.

**Figure 14.** Change in battery current, inverter current and DC link voltages with a change in irradiance in standalone mode operation.

It is observed that the dynamic response of the solar PV inverter for sudden changes in irradiance input is found satisfactory as the settling time of inverter current is in the range of 50 milliseconds whereas the settling time is in the range of 300 milliseconds in PV inverters with improved perturb and observe method presented in [23].

### **8. Contributions and Future Scope**

In this work, various energy storage system configurations suitable for cascaded H-bridge-based PV inverters are discussed. Design calculations for the chopper-based ESS for PV applications are presented in detail. Controls for the bi-directional chopper-based energy storage systems are studied and a control algorithm is developed and validated through real-time simulations. As future work, the proposed control algorithm can be validated on prototype models of the chopper-based systems or the AC/DC buck-boost converters proposed in [24]. The controls can be extended further for interleaved buck-boost converters to improve the power rating further. In this work, the operation of these systems is explained considering equal irradiances on each PV arrays and the state of charge of each battery bank is also considered equal. System operation for unequal irradiances can be studied as future work in line with the controls proposed in [25]. In present work, a single-phase system is considered; the controls can be extended to the three-phase system in future work. An additional feature of reactive power compensation can also be planned. The present system provides the solution for AC micro-grids for rural areas with frequent disruptions in grid supply. From earlier studies, it found that DC micro-grids are more feasible as PV sources, wind power generator and other renewable sources provide DC current and DC micro-grids are more stable compared to an AC micro-grid [26]. Feasibility of the energy storage system and its controls can be studied for DC grid systems proposed in [27].

### **9. Conclusions**

In this paper, a chopper-based ESS suitable for CHB-based PCS is presented. With the proposed system, the power to the grid/load can be supplied without any interruption. Cost, control complexity and controller hardware requirements for the chopper-based system are less compared to other configurations such as voltage regulator-based ESS and DAB-based ESS configurations. Buck-chopper based ESS configuration identified to be more suitable for grid-connected system and the bi-directional chopper based ESS configuration is suitable for standalone operation as well. The controls proposed for the bi-directional chopper-based ESS is analyzed with the help of controller-in-loop simulations by using a real-time simulator. Good dynamic response of the system for sudden changes in irradiance is observed from the presented results. Smooth transition in the system controls is also achieved during mode change over.

**Author Contributions:** All authors were involved to articulate the research work for its final depiction as the full research paper. "Conceptualization, U.S, S.V, S.P.; Methodology, U.S., S.P., D.A.; Software, S.V.; Validation, S.P, F.B. and J.B.H.-N.; Formal Analysis, U.S.; Investigation, S.V.; Resources, J.H.N and F.B.; Data Curation, S.P., D.A.; Writing-Original Draft Preparation, U.S., S.P., S.V., D.A.; Writing-Review & Editing, F.B., and J.B.H.-N.; Project Administration, F.B., S.P., J.B.H.-N.; Funding Acquisition, S.P." All authors have read and agreed to the published version of the manuscript.

**Funding:** This research received no external funding.

**Acknowledgments:** The authors like to express sincere gratitude to Renewable Energy lab, Prince Sultan University, Saudi Arabia for making this research work for execution and technical implementation in real-time, and Center for Bioenergy and Green Engineering/Center of Reliable Power Electronics (CORPE), Aalborg University, Esbjerg/Aalborg, Denmark.

**Conflicts of Interest:** The authors declare no conflict of interest.

### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).
