Advanced Frequency Control Technique Using GTO with Balloon Effect for Microgrids with Photovoltaic Source to Lower Harmful Emissions and Protect Environment

Abstract

Renewable energy (RE) resources such as wind and PV solar power are crucial for transitioning to carbon-free and sustainable energy systems, especially for agricultural and domestic applications in the desert and rural areas. However, implementing RE resources may lead to frequency penetrations, especially in isolated microgrids (µGs). This study proposes an adaptive load frequency control (LFC) technique for power systems. An integral controller can be tuned online using an artificial gorilla troops optimization algorithm (GTO), which is supported using a balloon effect (BE) identifier. Adaptive control is used to control the system frequency in case of variable loads and fluctuation due to 6 MW photovoltaic (PV). Three other optimization methods have been compared with the GTO + BE technique, namely the Grey Wolf Optimization method (GWO), the standard artificial gorilla troops optimization (GTO) and the Jaya technique. Digital simulation tests approved the efficiency of (GTO + BE) during system difficulties such as load disturbance and system parameter variations. In addition, the same test conditions have been repeated using a real-time simulation platform. The real-time simulation results supported the digital outcomes.

1. Introduction

1.1. Literature Review

Recently, conventional methods for generating power, especially from fossil fuels, are the primary problems concerning climatic conditions [1,2]. There is a strong interest from scientists and researchers to focus on the conservation and generation of energy through green energy resources (RE) such as solar (e.g., PV, CSP), hydrogen [3] and wind [4]. REs have advantages, including that they are free and economical, and play a crucial role in promoting environmental sustainability, reducing greenhouse gas emissions and ensuring a reliable and clean energy supply for several agricultural and industrial activities. According to [5], solar PV succeeded in reducing the emission of 270 million tons of CO₂ in 2021. In addition, in some areas, solar PV can be used to reduce air pollution by up to 30%, as concluded by [6]. Compared with conventional power sources, no water is needed in the PV operation and this acts as a water-saving factor [7]. RE resources are not feasible to be used as a single source of power due to their intermittent nature. Because of this, controllable energy storage technologies (ESTs) are commonly deployed with renewable energy sources [2,4]. The use of thermal energy, which previously predominated, is likely to decline in the future while the use of renewable energy is expected to increase. Additionally, a study into the microgrid’s frequency control is important because of its high absorption capacity and possible effect [8]. For the purpose of integrating RESs into the grid at the distribution level, MGs would offer an appropriate infrastructure. MGs may function in either islanded or grid-connected modes. They might face difficulties controlling frequency and voltage in islanded mode, though. Due to the low grid inertia, these issues would be more severe in MGs with a high percentage of power-electronically interfaced DERs. Actually, power imbalances in low-inertia MGs cause abrupt frequency fluctuations that might risk system stability. Therefore, in an isolated μG, it is crucial to use a controller that is robust and performs well under a variety of conditions [9]. Load frequency control maintains the balance of the power system within specified parameters within which it deviates from its nominal value and in accordance with a practically acceptable dynamic performance [10,11]. The μG operates in two modes (on grid and off grid). In off-grid mode, energy sources, synthetic virtual inertia, and energy storage terminals (ESTs) are often employed techniques to sustain frequency [12,13]. However, the capacity of one μG is constrained and susceptible to a variety of nonlinear random fluctuations impacts [13,14].

RESs in power systems lead to instability and complexity in this system. The primary μG input variables in power systems are the rise in environmental and economic concerns as well as the dependability of conventional power systems [15]. The gain of an LFC device can be adjusted offline with integral controller I, which is commonly used in LFC applications. The system gives poor dynamic performance when loading changes and systems are changed. Fixed parameters proportional integral PI controllers have been used to solve this problem [16,17,18].

Problem formulation: Policymakers try to integrate as much variable renewable energy (VRE) into electricity networks as they can to lower CO₂ emissions. This has resulted in goals for VRE as a percentage of overall power output and guidelines intended to make the best use of the electricity produced by VRE sources. RESs, ESTs, fuel generators, and other equipment make up the majority of μGs. These sources are connected to electricity, which may significantly improve security. But compared to a standard μG, its topology is more complex, which makes managing energy sources, planning and regulating the structure, and synthesizing the system more challenging [19,20]. It goes without saying that RESs alter the distribution system’s voltage and frequency [21]. Furthermore, power systems are adversely affected if these sources are not managed properly. For these reasons, effective solutions should be required to maintain the characteristics of frequency variations and voltage changes to remain within specified values [22,23,24]. A comparison of the cost, reliability, and consistency of the models has been carried out based on the information in [25,26], where the description of the models is based on the communication system and model design.

1.2. Contributions

An artificial gorilla troop optimization algorithm (GTO) has been used in many industrial applications to adaptively tune the gains of conventional controllers [27]. GTO with a cascaded proportional integral–fractional order proportional–integral derivative (PI-FOPID) controller can improve the hybrid microgrid’s frequency response [28]. The results of GTO are also compared to those of Grey Wolf Optimization (GWO), whale optimization algorithm (WOA), and sine cosine algorithm (SCA) using a multi-layer perceptron (MLP) [29].

This research addressed this issue and suggested a balloon effect (BE) adjustment to increase the sensitivity of the optimization technique to disturbances and parameter changes [30,31,32].

This study offers to perform adaptive frequency control for changeable loads and parameters in smart μG by using an artificial gorilla troop optimization method with balloon effect (GTO + BE). The electrical load, PV, and diesel generators make up the evaluated μG. We investigate the impact of frequency variations arising from random demand loads and RESs on the (GTO + BE) method. Additionally, gray wolf optimization (GWO) and Jaya approaches are compared in order to show their accuracy and resilience. Furthermore, the MATLAB simulation findings are verified by the implementation of a real-time simulation. A lab version of the suggested controller integrated with the system under study is presented. This stage involves applying the system’s GTO + BE, GWO, and Jaya Algorithms to a real-time simulator utilizing a desktop PC with (Win 10 operating system, Core i5, 4.2 GHz processor, and 16 GB of RAM) supported by QUARC pid_e data acquisition card and + QUARC 2.3.

Here are the key distinguishing characteristics of this work:

• An online adaptive LFC is presented using the GTO + BE optimization method;
• The effectiveness of a GTO + BE optimizer-adjusted integral controller in regulating frequency is discussed;
• A system with the proposed adaptive technique gives performance better than those systems with conventional GTO, GWO and Jaya Algorithms.

1.3. The Organization of the Paper

The paper is organized as follows:

Section 2 describes the mathematical model of the isolated single area μG. Section 3 explains the artificial gorilla troops optimization algorithm (GTO). Section 4 discusses the balloon effect identifier (BE). Section 5 discusses the modified (GTO + BE) algorithm. Section 6.1 offers the simulation results and discussions of the proposed controlled systems. Additionally, Section 6.2 presents a real-time implementation for the studied system. At the end, Section 7 shows the conclusions of the work.

2. Power System Dynamic Model

The block diagram for a microgrid power system is displayed in Figure 1. The following equations explain the proposed microgrid power system’s dynamic model [30].

The frequency deviation (∆f) can be expressed as

$$∆f=\left(\frac{1}{M}\right).∆Pd-\left(\frac{1}{M}\right).-∆PL-\left(\frac{D}{M}\right).∆f$$

(1)

The diesel generator’ dynamic can be expressed as

$$∆Pd=\left(\frac{1}{Td}\right).∆Pg-\left(\frac{1}{Td}\right).∆Pd$$

(2)

The governor’s dynamic can be expressed as

$$∆Pg=\left(\frac{1}{Tg}\right).∆Pc-\left(\frac{1}{R.Td}\right).∆f-\left(\frac{1}{Tg}\right).∆Pg$$

(3)

3. Artificial Gorilla Troops Optimization Algorithm (GTO)

This section presents a modern metaheuristic algorithm based on gorillas’ group behavior, which is called GTO, which illustrates five distinct operators that explain the two phases of the GTO Algorithm (exploration and exploitation), as indicated in Figure 2. There are three operators in the exploration phase while there are two operators in the exploitation phase.

The three operators of the exploration phase are the following:

• GTO exploration can be increased by migration to an unknown location;
• Moving on to the other gorillas will increase the balance between exploitation and exploration;
• GTO can search for different optimization spaces in more detail when it migrates towards a known location.

The two operators in the exploitation phase are:

• Staying with the silverback, and
• A struggle associated with female adulthood.

The optimization space flowchart of the GTO method is shown in Figure 3. The GX is the gorilla candidate position vector, which is formed in each phase and operates in the event that it outperforms the current one. The X is known as the gorilla position vector. The silverback, the best answer, is presented on each iteration. There is just one silverback in the population as a result of the quantity of search agents used for optimization activities. Gorillas live in social groupings in the wild. There are three types of solutions based on X, GX, and silverback. Providing better food or establishing a fair and robust group can increase a gorilla’s power. In the GTO algorithm, solutions are generated in each iteration called GX. New solutions (GX) replace existing ones (X) if they are new. In any other case, it will remain in memory (GX). It is impossible for gorillas to live on their own because of their tendency to live in a communal manner. The algorithms can be widely applied because GTO’s unique features make it effective in several optimization issues given the basic concepts of gorilla group life.

3.1. Exploration Phase

The silverback gorilla provides the best candidate solution in each iteration of the GTO algorithm, which considers all gorillas to be candidates. The exploration phase involved three different strategies: migration to unknown positions, migration to known positions, and migration to another gorilla group. A general process is followed in choosing each of these methods. Figure 4 and Equation (4) show the three methods to obtain the solution of the next iteration. Iteration (t + 1) of the gorilla candidate position vector is represented by GX(t + 1). A gorilla’s current position is represented by X(t). Furthermore, $r_1$, $r_2$, $r_3$ and rand represent random values between 0 and 1. To determine the migration method to an unknown place, an optimization procedure uses a parameter p with a range of [0, 1]. An unknown place is selected as a potential migration mechanism when rand < p. On the other hand, the moving mechanism towards other gorillas is selected if rand > 0.5. It is noteworthy, however, that when rand < 0.5 is taken into account, migration to a known place is selected. Every one of these methods benefits the GTO algorithm. The algorithm can monitor the whole challenge space in the first mechanism, perform better during exploration in the second mechanism, and escape local optimum portions of the task space in the third mechanism. The upper and lower boundaries of each variable are represented by the variables UB and LB. Xr and GXr were chosen at random from the total gorilla population to be part of the group. Positions that are modified throughout each phase are added to a vector of randomly chosen gorilla candidate positions.

$$GX\left(t+1\right)=\left\{\begin{array}{c}\left(UB-LB\right)\times r_1+LB,\\ \left(r_2-C\right)\times X_r\left(t\right)+L\times H,\\ X\left(i\right)-L\times \left(L\times \left(X\left(t\right)-G{X}_r\left(t\right)\right)+r_3\times \left(X\left(t\right)-G{X}_r\left(t\right)\right)\right),\end{array} \right.$$

(4)

$$C=F\times \left(1-\frac{It}{MaxIt}\right),$$

(5)

$$F=\cos \left(2\times r_4\right)+1,$$

(6)

$$L=C\times l$$

(7)

$$H=Z\times X\left(t\right),$$

(8)

$$Z[-C,C]$$

(9)

Currently, it represents the number of optimization iterations being used, while MaxIt represents the total number of optimization iterations. A cosine function is represented by cos, while $r_4$ shows a random number that is updated with each repetition, ranging from 0 to 1. When l is a random value between −1 and 1, it denotes the leadership of the silverback. There is a group formation operation conducted following the exploratory phase. When determining the cost of a solution after the exploration phase, we use the GX(t) solution rather than the X(t) solution if the GX(t) solution is less expensive. Silverbacks are therefore sometimes referred to as the best solution produced during this phase.

3.2. Exploitation Phase

This phase involves two behaviors: competing for adult females and following the silverback. In Equation (5), C determines which behavior to choose. If C ≥ W, the silverback mechanism is chosen; however, if C < W, the competition for adult females is chosen. The optimization process requires a parameter called W.

3.2.1. Follow the Silverback

Gorilla males follow the silverback’s instructions to search for food supplies in a variety of locations. The movement of the group may also be influenced by each member. This behavior is simulated using Equation (10). Figure 5 shows this process.

$$X\left(t+1\right)=L\times M\times \left(X\left(T\right)-X_{silverback}\right)+X\left(t\right)$$

(10)

$$M={\left({\left|\frac{1}{N}{\sum }_{i=1}^{N}G{X}_i\left(t\right)\right|}^g\right)}^{\frac{1}{g}}$$

(11)

$$g=2^L$$

(12)

In this instance, the gorilla’s position vector is indicated by X(t), while the silverback gorilla’s position vector is represented by Xsilverback (optimal solution). Moreover, at iteration t, GXi(t) finds the vector position of each potential gorilla. The total number of gorillas is represented by the number N.

3.2.2. Competition for Adult Females

Young gorilla males compete violently with each other in their group for the choice of adult females when they reach puberty. Members of the group may be involved in these conflicts for days at a time. We can simulate this behavior using Equation (13):

$${GX\left(i\right)=X}_{silverback}-(X_{silverback}\times Q-X(t)\times Q)\times A$$

(13)

$$Q=2\times r_5-1,$$

(14)

$$A=\beta \times E,$$

(15)

$$E=\left\{\begin{array}{c}{N}_1, rand\ge 0.5\\ N_{2, } rand<0.5 \end{array}\right.$$

(16)

A vector defining violence in conflict is coefficient vector A. In order for the optimization operation to take place, the parameter β must have a value. Simulation of violence and its effect on dimensions is carried out using E. In the case where rand ≥ 0.5, the dimensions of the problem and E are identical to random values received from the normal distribution, but in the case where rand < 0.5, E is also a random value received from the normal distribution. The Rand value ranges between zero and one. Figure 6 indicates how the solutions can be changed to arrive at the best solution.

Silverbacks are considered to be the best solutions found across the entire population. A cost estimate for all GX solutions is performed at the end of the exploitation phase. The GX(t) solution is used as X(t) if the costs of GX(t) and X(t) are equal.

4. Balloon Effect (BE) Identifier

The balloon effect (BE) is described as the effect of air on balloon size. With the balloon effect, system challenges like disturbances and parameter uncertainty can have a significant impact on Gi(s). Figure 7 shows how the BE identifier affects an optimization strategy objective function at any iteration. The algorithm process is therefore enhanced using this technique [30,31,32,34].

The representation of the transfer function of the studied microgrid at any moment (i) will be as follows:

$$G_i\left(\mathrm{s}\right)=\frac{Y_i\left(s\right)}{U_i\left(s\right)}$$

(17)

Furthermore, Gi(s) is a function of its previous value $G_{i-1}\left(s\right)$. ${AL}_i$ stands for a gain and $G_0\left(s\right)$ represents the nominal process transfer function.

$${\mathrm{G}}_{\mathrm{i}}\left(\mathrm{s}\right)={AL}_i{G}_{i-1}\left(s\right)$$

(18)

$$G_{i-1}\left(s\right)={\rho }_i{G}_0\left(s\right)$$

(19)

where

$${\rho }_i={\prod }_{n=1}^{i-1}{AL}_n$$

(20)

$$G_i\left(s\right)={{AL}_i\rho }_i{G}_0\left(s\right)$$

(21)

In brief, BE acts as an online identifier that can sense any system difficulties, such as load disturbances or system parameter variations, and this will positively affect the job of the coupled optimization technique.

5. GTO-Based Balloon Effect Identifier

Figure 8 illustrates a simplified microgrid model; it is utilized to calculate the parameters of a second-order closed-loop system for the controlled area:

$$\mathrm{T}.\mathrm{F}=\frac{{wn}^2}{S^2+2\eta Wn+{Wn}^2}=\frac{\frac{Ki}{Mo}}{S^2+\left(\frac{\left(Do+\frac{1}{Ro}\right)}{Mo}\right)S+\frac{Ki}{Mo}}$$

(22)

where Do, Ro and Mo are the nominal values of D, R and M, respectively.

$${\omega }_n=\sqrt{\mathrm{K}\mathrm{i}/\mathrm{M}\mathrm{o}}, \eta =\frac{\frac{(Do+\frac{1}{Ro})}{Mo}}{2{\omega }_n}$$

(23)

$$T_r=\frac{\pi -\sqrt{(1-{\eta }^2)}}{{\omega }_n\sqrt{(1-{\eta }^2)}}, T_s=\frac{4}{{\eta \omega }_n}, M_P=e^{\frac{-\pi \eta }{\sqrt{(1-{\eta }^2)}}}$$

(24)

This is the objective function of the GTO-based BE identifier.

$$J=min\sum {(T}_r+T_s+M_P)$$

(25)

This means that the objective function $J$ is a function of ${AL}_i$ and $k_i$ to address the system challenges.

6. Results and Discussion

The LFC controller of a small isolated power system is adjusted using the proposed improved approach (GTO + BE). The first segment uses a MATLAB/Simulink environment for the simulation testing, whereas the second section uses a real-time simulator. Figure 9 shows the 20 MW diesel generator that is part of the proposed microgrid. The nominal parameters of the system and the GTO parameters are provided in

Table 1. Parameters of the studied microgrid.

D	H = (M/2)	R	Tg	Td
(Pu/Hz)	(Pu·sec)	(Hz/Pu)	(sec)	(sec)
0.015	0.08335	3	0.08	0.4

and

Table 2. Suggested parameters of GTO optimizer.

Population Size (K)	5
Maximum Iteration (IT max)	50
Parameter to be set before optimization operation (P)	0.03
Beta (β)	3
Paramter select mechanism of migration (W)	0.8
The initial values of the design variables (Ki)	[0.04, 0.025, 0.017, 0.08, 0.030]

, respectively.

All simulations are executed on a high-performance desktop PC with (Win 10 operating system, Core i5, 4.2 GHz processor, and 16 GB of RAM) supported by QUARC pid_e data acquisition card and MATLAB R2014 software with QUARC sub-program.

6.1. MATLAB Simulation Results

6.1.1. First Case

In this instance, testing was carried out on the system under nominal conditions with a step load variation. In this instance, at t = 3 s, the load changes from 0 pu to 0.02 pu. Governor dead-band and turbine GRC are taken into consideration. Turbine GRC is 10% per minute, while governor dead-band is 0.05 pu. [35,36]. This control scheme was evaluated by comparing it with GTO described in [31], adaptive Grey Wolf Optimization (GWO) described in [37] and the Jaya optimization method presented in [38]. Gray wolves also exhibit a group hunting behavior in addition to their social hierarchy. The following are the main phases of grey wolf hunting according to Mirjalili et al. [37]: first, tracking, chasing, and approaching the prey; then, pursuing, encircling, and harassing the prey until it stops moving; and then attacking the prey.

Table 2 contains a list of the employed GTO’s stated specifications.

Table 3. Data for Grey Wolf Optimization (GWO).

Maximum Iteration (IT max)	50
Population Size (k)	5
Convergence Constant (a)	2

contains a list of the employed GWO’s defined parameters. The parameters of the Jaya optimization employed are listed in

Table 4. Data for Jaya Algorithm.

Population Size (k)	5
Number of Generations (i)	50
Number of Design Variables (m)	2

. The system frequency deviation change is depicted in Figure 10 for four different scenarios: the adaptive integral controller utilizing adaptive Grey Wolf Optimization (GWO), the adaptive integral controller using GTO and the adaptive integral controller using conventional the Jaya optimization approach provided in [38]. The system frequency deviation changes after using GTO and utilizing the suggested GTO + BE is depicted in Figure 11. It can be shown that, when employing the standard GTO, Mp has been reduced to around 20% in comparison to the value produced when utilizing the Jaya method and Grey Wolf Optimization (GWO), while there is not quite an overshoot with the suggested GTO + BE.

6.1.2. Second Case

In this case, the system with the suggested (GTO + BE) adaptive controller has been tested in the face of system parameter variation. The power system’s time constant is increased by a factor of 200%, D is decreased to 0.08 pu MW/Hz, and the time constant is increased by a factor of 200%, as well. Figure 12 illustrates the system response obtained using four different controllers (I, I adjusted by Jaya Algorithm [38], I adjusted by GWO [37], and I adjusted by GTO). Comparing the situation of the system responses determined by (I tuned by GTO and I tuned by (GTO + BE)) is shown in Figure 13. These results show that, whereas systems with I adjusted by Jaya, GWO, and GTO can successfully manage these issues, the frequency response of the fixed parameter I controller is unacceptable. Furthermore, the system utilizing the suggested GTO + BE provides optimal time responsiveness and high performance.

6.1.3. Third Case

Multiple operating conditions have been tested on the system. Testing was performed with a 12 MW variable load, considering the nominal parameters of the system and adding a 6 MW PV to the microgrid as an additional generation source where Figure 14 depicts a model of a variable solar power system [39,40] provides information on how temperature and radiation affect the PV module system’s output power. The considered PV system output power is calculated by

$$P_{pv considered}=-S\phi \eta \{1-0.005\left(T_a+25\right)\}$$

(26)

The PV first order transfer function of PV can be given as

$$G_{pv}\left(S\right)=\frac{K_{pv}}{1+T_{pv\left(S\right)}}$$

(27)

For PV array:

(S) is the measured zone, which is 4084 m², ($\phi$) is the diversion efficiency, which ranges from 9 to 12%, (η) is the sun radiation, which is 1 kW/m², and (Tampient) is the ambient temperature in degrees Celsius. In this work, Ppv is only linearly varied with η, and Tampient is maintained at 25 °C. Additionally,

$$∆P_{pv}=P_{pv rated}-P_{pv considered}$$

(28)

According to Figure 15, the system with the suggested controller performs best in terms of frequency response when employing the suggested (I + GTO + BE). When compared to normal GTO, GWO and Jaya approaches, this figure demonstrates the superiority of the proposed online tuned controller employing the (GTO + BE) methodology.

6.2. Real-Time Simulation Results

The block diagram for the suggested system’s implementation utilizing a real-time simulator is shown in Figure 16. A computer with a QUARC pid_e data acquisition card has the studied islanded microgrid installed. Using a storage oscilloscope (DSO-X 2014A, Keysight, Santa Rosa, CA, US), the output frequency signals were captured. Figure 17 shows how the suggested real-time simulation system is physically configured. The initial scenario involves doing a real-time simulation test with nominal system characteristics and a step load adjustment. Line charts of the maximum overshoots and settling times of the system frequency deviation change (in the case of a step load change) are displayed in Figure 18 for the applications of integral controllers with fixed parameters and adaptive controllers that use the standard Jaya optimization approach described in [38]. adaptive integral controller using GTO, adaptive Grey Wolf Optimization (GWO) and using proposed GTO + BE. In the second case, the same variation of the system parameters used in case 2 of simulation results is applied, and a system response in this case is shown in Figure 19.

In the case of a previous random load and PV source, Figure 20 compares the system response produced with four different controllers (fixed I, adjusted by Jaya Algorithm, tuned by GWO, and tuned by GTO + BE). According to Figure 20, the system with the proposed controller performs best in terms of frequency response when employing the suggested (I + GTO + BE). These numbers demonstrate the superiority of the proposed online tuned (GTO + BE) controller over traditional I-controller, GWO, Jaya, and regular GTO techniques. In addition, when the proposed adaptive controller depends mainly on a strong computer with a data acquisition card and signal sensors, it can be applied to large-scale power systems by adding a Phasor Measurement Unit PMU. In addition, the tuning of MG’s parameters such as H and D will be an effective extension of the proposed adaptive controller that can improve the system frequency response in case of large MGs.

7. Conclusions

PV is an important source of RES where the presence of PV in power system PV is an important source of RES where the presence of PV in power system reduces CO2 and air pollution while also contributing to water-saving goals. The system under study in this work is a microgrid with a 30% PV source. This work proposes an adaptive LFC mechanism based on GTO and (GTO + BE). Different from the standard controller, which is not guaranteed to offer sufficient performance throughout a wide range of operating conditions, the GTO + BE control strategy allows for frequency regulation in situations of load disturbances and parameter uncertainties. A digital simulation has been run to examine the suggested control strategy in the presence of random demand loads and variable PV power generation. The obtained responses of the system with the proposed controller have been compared with the responses of the systems with other controllers (conventional I controller and adaptive controller using JAYA, GWO, and GTO). The comparisons explained that the proposed controller has superiors over the other controllers in terms of the overshoot and the settling time, which are decreased by less than 0.05 and 10 s, respectively. In general, the suggested adaptive control strategy with GTO + BE may do so with reduced system complexity and computation time while effectively handling large system issues (such as disturbances and parameter fluctuations). With the suggested method, oscillations may be efficiently suppressed and a notable enhancement can be obtained within the suggested μG by online tuning of the controller gains. Therefore, it is recommended to use a controller whose gains may be tuned using GTO + BE in order to handle LFC concerns and reduce system oscillations. Lastly, a PC via QUARC pid_e data collection card has been used to demonstrate a laboratory implementation of the required controller with the investigated system, confirming the efficacy and robustness of the suggested adaptive controller for variable load and parameter modification on islanded μG. In addition, the suggested algorithm can be extended to large-scale isolated MGs by tuning MG’s parameters such as H and D to improve the system frequency.