Dynamic Voltage Restorer as a Solution to Voltage Problems in Power Systems: Focus on Sags, Swells and Steady Fluctuations

Abstract

In this study, the author presents the results of a survey on the utilisation of a dynamic voltage restorer (DVR) in power systems to alleviate voltage problems that result in sags, swells and fluctuations in voltage outside the required steady limits. A methodology based on the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement is adopted for conducting and reporting on the review, while the Scopus database is used to locate the relevant publications. A total of 68 publications qualify for inclusion in the survey. A bibliometric analysis covering the number of publications per annum, the top 10 most-cited journals and the top 10 most-cited publications is performed. The information from the selected publications is extracted, summarised and categorised into network scenarios for the use of DVRs, topologies and optimisation of DVRs; strategies for DVR controllers; and platforms that evaluate the feasibility of DVR topologies and controllers. Moreover, research trends and gaps are evaluated. Finally, potential areas for future research are proposed. This study provides an overview of the research on the use of a DVR to resolve voltage problems and is a resource for researchers generally interested in distributed flexible AC transmission systems (DFACTSs) and particularly interested in DVRs.

1. Introduction

Virtually all human activities rely on the availability of electrical energy and on electrically driven processes [1]. In fact, electricity is essential for meeting human needs for a comfortable living environment and propelling economic activities. There are several factors that need to be considered when thinking about the power systems that are built to generate, transmit and distribute electrical energy. Firstly, these systems must deliver power to the customer while meeting the minimum standards for supply of an acceptable quality. Furthermore, these systems must not have steady-state and stability constraints that can limit the transfer of power to ensure the expected load can be adequately provided.

Secondly, factors such as the increasing demand for electrical energy, the limited and dwindling nature of fossil fuels, the uncertainty associated with the prices of fuel and the imperatives of reducing greenhouse gas [2] have significantly spurred the use of renewable energy sources (RESs). Solar and wind energy are the leading RESs according to the amount of energy installed but are not problem-free. For example, if these energy sources are disconnected during severe contingencies or disturbances, there could be serious challenges to the supply of reactive power, leading to difficulties in maintaining bus voltages within acceptable ranges [3] and constraints to power transfer [4], which can lead to voltage instability [5]. These RESs are interfaced into the system using converters, which have a non-linear behaviour [6,7,8] and are sources of harmonics.

Thirdly, and lastly, these power systems are exposed to environmental factors such as pollution, weather and climatic phenomena, animals and human activity. Each of these factors is capable of initiating disturbances. This can result in a deterioration in the quality of supply [9] to customers. Additionally, the occurrence of faults means that currents that are much higher than normal operating conditions may flow, and this could lead to uncertainty regarding whether the protective equipment is adequately rated to clear such faults. Consequently, high-fault currents may [10] cause damage to equipment and harm human beings.

These challenges can be resolved by strengthening power systems to make them more robust to a variety of challenges. Strengthening options include adding new power system components, uprating the capacity of lines [11], utilising fixed compensation [12], installing variable compensation [13], and deploying flexible AC transmission system (FACTS) devices [14]. These solutions have different strengths and weaknesses related to cost, complexity, implementation time, environmental impact, and controllability, to mention just a few of the considerations.

At the heart of FACTS devices are power electronics, which enable these devices [12] to control system parameters smoothly, continuously and precisely. However, FACTS are costly and complicated, and concerns exist regarding their reliability, as the failure or outage of a single, costly device means that its capability is entirely lost. To overcome some of these challenges, a newer generation of FACTS, called distributed FACTS (DFACTS), was developed [15], with the distributed nature of these devices enabling [16] fine granularity in the rating. Thus, these systems can be scaled up as demand grows.

Dynamic voltage restorers (DVRs), series-connected devices, belong to the family of DFACTS, and the layout of a DVR, showing the components, is presented in Figure 1. The main transformer is used to inject a voltage into the line. The parallel switch changes the DVR to bypass this mode during fault conditions, protecting the inverter from the high currents that can flow. The passive filter converts the pulse-width-modulated (PWM) waveform into a sinusoidal output. The inverter converts the DC power from storage into AC power. The DC capacitor provides a stiff DC voltage input to the inverter. The battery is used to store DC energy, which can be utilised to provide the DVR with the active power required for compensation.

The issues stated in the introduction can impair voltages in at least three ways. Firstly, voltage sags, described in the IEEE Standard 1159–1995 [18] as a reduction in root mean square voltage magnitude to values ranging between 0.1 and 0.9 per unit for durations ranging from ½ cycle to 1 minute, can occur. Secondly, voltage swells, described as a type of voltage disturbance in which the magnitude of the root mean square supply voltage increases [19] to a value ranging between 1.1 and 1.8 per unit for a period that can range from ½ to 1 minute, can occur. Lastly, voltages can steadily fluctuate outside the acceptable operating ranges [20], resulting in a network performance that is deemed unacceptable.

An initial literature search was conducted, and it was established that there was a large volume of publications related to DVRs. No publications summarising the issues related to voltage sags, swells and steady fluctuations were found. Thus, in this study, the author aimed to present the results of a survey on the status of the research conducted on DVR use to solve the problems of voltage sags, swells and steady fluctuations in power systems.

The remaining parts of the manuscript comprise the following: In Section 2, the methodology for the survey is detailed. The results of the survey, including, firstly, a bibliometric analysis of the studies included in the survey; secondly, the extraction, summarisation and synthesis of key data from the selected publications; and thirdly, and lastly, an analysis of research trends and gaps in the literature, are presented and discussed in Section 3. Finally, in Section 4, the conclusions of the survey are drawn, and a future research direction is proposed.

2. Research Methodology

The approach based on a systematic literature review (SLR) was adopted in this study for the purposes of surveying works on the use of DVR for mitigating voltage sags, swells and variations. The SLR methodologies [21] have three distinguishable characteristics: the delineation of the research strategy, the identification of explicit criteria for inclusion and/or exclusion of publications, and, finally, the aim to seek, gather, evaluate and interpret as much of the available and relevant publications as is possible. For this SLR, the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement, as proposed by Moher et al. [22], was adopted for conducting and reporting on the review, and the process is summarised schematically in Figure 2. The use of the PRISMA method facilitates [23] the identification of publications, the screening of identified publications as a measure of assessing their quality, the choice of eligible publications following quality assessment, and, finally, the extraction and summarisation of information.

2.1. Identification

The process of identifying the relevant publications for the study involved a search in the Scopus database. In the beginning, the option of “Article, title, and abstract” in the “Search within” field was chosen, the “Search Document” option was selected and the string “dynamic voltage restorer” was inserted in the search field. A total of 1669 documents were returned by the search. After the above, additional selections were made for fields initially not considered in the search. The following choices were made: for “Subject area”, Engineering was opted for; for “Document type”, the choice was “Article”; for keywords, “dynamic voltage restorer” was chosen; and for “Source type”, journal was chosen. The outcome of this process was the identification of 240 publications.

2.2. Screening

At this stage, only publications in the English language were considered thenceforth. In addition, the author elected to focus on documents that were published in the period 2018 to 2022 so that the most recent developments could be reflected in the survey, with those for the year 2023 not included on the basis that the year is not yet complete. Thereafter, an assessment for the existence of duplication was made to eliminate them. Finally, the broad relevance of the publications to the subject of the review, based on the assessment of the abstract obtained from Scopus, was evaluated, and publications without sufficient relevance were excluded from consideration in the subsequent stages. After this stage of the review, a total of 92 publications remained in the process.

2.3. Eligibility

The author then began the process of downloading the full text of the 92 documents that remained after the screening process, as described in the previous step. For each downloaded document, if, based on the evaluation of the entire manuscript, the study was relevant to the topic of the survey, such a manuscript was considered eligible for the next stage of the process.

For some of the 92 documents, the full text was not accessible, and these references were not considered any further. Also, references that had reached the previous stage, but were rather reviews, and not primary research publications, were removed from further consideration. At the end of this stage of the review, 68 documents met the eligibility criteria and were considered in the next stage—data extraction and summarisation of the information. The key details of these publications are summarised in

Table 1. Final list of publications included in the survey.

Item No.	Ref. No.	Year	Authors	Item No.	Ref. No.	Year	Authors
3.1 Scenarios for the Use of DVRs				[24]	[25]	2022	Zhang et al.
[1]	[26]	2022	Chiranjivi and Swarnasri	[27]	[28]	2022	Ahmed et al.
[2]	[29]	2019	Farooqi et al.	[30]	[31]	2022	Inamdar and Iyswarya
[3]	[32]	2021	Farooqi et al.	[33]	[34]	2019	Kassarwani et al.
[4]	[35]	2020	Jeyaraj et al.	[36]	[37]	2020	Mohammed and Ariff
[5]	[38]	2018	Jian et al.	[39]	[40]	2020	Mohammed et al.
[6]	[41]	2021	Khergade	[42]	[43]	2022	Nasrollahi et al.
[7]	[44]	2022	Majumder et al.	[45]	[46]	2022	Shah et al.
[8]	[47]	2021	Neto et al.	[48]	[49]	2022	Zahra et al.
3.2 Topologies of DVRs				[50]	[51]	2022	Al-Gahtani et al.
[9]	[52]	2022	Ashraf et al.	[53]	[54]	2022	Al-Gahtani et al.
[10]	[55]	2022	Aydogmus et al.	[56]	[57]	2019	Danalakshmi et al.
[11]	[58]	2022	Bajaj	[59]	[60]	2018	Inci et al.
[12]	[24]	2018	Biricik et al.	[61]	[62]	2019	Singh et al.
[13]	[27]	2019	Jiang et al.	[63]	[64]	2018	Vo Tien et al.
[14]	[30]	2020	Karimi et al.	[65]	[66]	2021	Danalakshmi et al.
[15]	[33]	2022	Prasad and Dhanamjayulu	[67]	[68]	2020	Elkady et al.
[16]	[36]	2020	Prasada et al.	[69]	[70]	2022	Reddy et al.
[17]	[39]	2021	Rahman et al.	[71]	[72]	2019	Hung et al.
[18]	[42]	2018	Srikanth Babu et al.	[73]	[74]	2020	Merchan-Villalba et al.
[19]	[45]	2020	Toumi et al.	[75]	[76]	2018	Tien et al.
[20]	[48]	2019	Tousi et al.	[77]	[78]	2020	Kalyanasundaram et al.
[21]	[50]	2022	Chen et al.	3.4 Optimisation of DVR Controllers, Sizes and Locations
[22]	[53]	2018	Farhadi-Kangarlu	[79]	[80]	2019	Kassarwani et al.
[23]	[56]	2021	Biricik et al.	[25]	[81]	2018	Kassarwani et al.
[26]	[59]	2018	De Almeida Carlos et al.	[28]	[82]	2019	Tiacharoen and Chatchanayuenyong
[29]	[61]	2022	Karim et al.	[31]	[83]	2021	Zhong et al.
[32]	[63]	2022	Prasad and Dhanamjayulu	[34]	[84]	2022	Salman et al.
[35]	[65]	2020	Rajkumar et al.	[37]	[85]	2022	Salman et al.
3.3 Strategies for DVR Controllers				3.5 Platforms for Feasibility Evaluation of DVR Controllers
[38]	[67]	2022	Li et al.	[40]	[86]	2022	Mallajoshula and Naidu
[41]	[69]	2022	Ahmed et al.	[43]	[87]	2019	Torres et al.
[44]	[71]	2021	Boussaid et al.	[46]	[88]	2018	Roldán Pérez et al.
[47]	[73]	2020	Martins et al.	[49]	[89]	2018	Vu et al.
[52]	[75]	2021	Rajkumar et al.	[51]	[90]	2022	Li et al.
[55]	[77]	2018	SasiKiran and Manohar	[54]	[91]	2021	Pratap et al.
[58]	[79]	2022	Singh et al.

.

2.4. Extraction

As discussed after the last paragraph, a set of 68 documents qualified for the final stage of data extraction and the summarising of the key information. The data were summarised into the major themes shown in Figure 3, such as network uses of DVRs, topologies of DVRs, controller strategies for DVRs, platforms for investigating DVR performance and application of optimisation techniques. These results are presented in Section 3.2.

3. Results and Discussion

3.1. Bibliographical Analysis

3.1.1. Number of Publications per Year

The annual number of publications and the cumulative annual number of publications produced per year, from the set of 68 selected for the review, are plotted in Figure 4. The numbers produced between 2018 and 2021 seem to be steady and between 10 and 12 publications per annum. A big jump is observed in 2022, as shown first by the actual number of publications that reached 24 and, also, as demonstrated by an observable change in the slope of the cumulative number of publications.

3.1.2. Top 10 Most-Cited Journals

The ten most-cited journals, from the list of journals in which the articles selected for the survey are published, are presented in

Table 2. List of top 10 most-cited journals.

Journal Title	Total Citations
IEEE Access	54
International Journal of Power Electronics and Drive Systems	47
International Transactions on Electrical Energy Systems	35
Energies	30
International Journal of Electronics	28
IEEE Transactions on Industry Applications	23
Smart Science	22
Indonesian Journal of Electrical Engineering and Computer Science	21
CSEE Journal of Power and Energy Systems	17
International Journal of Electrical Power and Energy Systems	16

. Most citations come from publications in the journal IEEE Access, with 54 citations. The International Journal of Power Electronics and Drive Systems, with 47 citations, and International Transactions on Electrical Energy Systems, with 35 citations, are positioned second and third, respectively, according to the number of citations.

3.1.3. Top 10 Cited Studies

From the list of 68 publications selected for inclusion in the survey, the details of the top 10 most-cited publications are presented in

Table 3. List of top 10 most-cited publications.

Authors	Title	Year	Cited by
Tien et al. [76]	A multifunctional dynamic voltage restorer for power quality improvement	2018	28
De Almeida Carlos et al. [59]	Cascaded open-end winding transformer-based DVR	2018	23
Parreño Torres et al. [87]	A discrete-time control method for fast transient voltage-sag compensation in DVR	2019	23
Bajaj M. [58]	Design and simulation of hybrid dg system fed single-phase dynamic voltage restorer for smart grid application	2020	22
Farooqi et al. [29]	Mitigation of power quality problems using series active filter in a microgrid system	2019	18
Kassarwani et al. [80]	Performance analysis of dynamic voltage restorer using improved PSO technique	2019	18
Chen et al. [50]	Energy-saving superconducting magnetic energy storage (SMES)-based interline DC dynamic voltage restorer	2022	17
Jeyaraj et al. [35]	Development and performance analysis of PSO-optimised sliding mode controller-based dynamic voltage restorer for power quality enhancement	2020	15
Danalakshmi et al. [57]	A control strategy on power quality improvement in consumer side using custom power device	2019	14
Toumi et al. [45]	PV integrated single-phase dynamic voltage restorer for sag voltage, voltage fluctuations and harmonics compensation	2020	14

. The most cited study is titled “A multifunctional dynamic voltage restorer for power quality improvement”, authored by Tien et al., published in 2018 and has a total of 28 citations. The study “Cascaded open–end winding transformer–based DVR”, published in 2018 and authored by De Almeida Carlos et al., has a total of 23 citations, and another study, “A discrete–time control method for fast transient voltage–sag compensation in DVR”, authored by Parreño Torres et al. and published in 2019, has a total of 23 citations. These two studies completed the last two of the first three spots for the most-cited publications.

3.2. Some Scenarios for Uses of DVR to Mitigate Voltage Problems

The effectiveness of utilising a DVR to resolve voltage-related network problems has been assessed for a variety of scenarios that can be described based on the type of network, equipment connected and operating regime of the network, among other things. Some of these scenarios include a study [26] wherein Chiranjivi investigated the use of a DVR to compensate for voltage sags and swells in a distribution power system to which a photovoltaic plant was connected. For a distribution network to which a battery energy storage system and a mini-hydro turbine were connected, Farooqi et al. [29] evaluated a DVR as a solution for ameliorating balanced voltage sags and swells. To enhance the performance of modern pumping systems powered by solar energy, which, therefore, can only operate during solar hours, and reliability is affected by weather conditions, Farooqi et al. [32] proposed a photovoltaic (PV)–DVR solar water pumping (PV–DVRWP) system, with a controller used for a DVR, to mitigate voltage sags and swells to improve load voltage.

For a distribution network to which a 400 kW PV farm was connected, Jeyaraj at al. [35] proposed the use of a DVR that uses a sliding mode controller (SMC), the optimum parameters of which are based on the particle swarm optimisation (PSO) algorithm, to mitigate voltage sags and swells caused by faults and disturbances, thereby providing good-quality voltage to the load. In [38], Jian et al. proposed a scheme in which a DVR installed in a medium–voltage (MV) power grid to protect sensitive loads from voltage sags was provided with a control strategy that allows the DVR to act as a virtual impedance in series with sensitive loads, thereby allowing the DVR to additionally function as a reactor to limit fault currents or as a series capacitance to compensate for voltage losses during heavy loads.

To enhance the performance of adjustable speed drives (ASDs) during voltage sags, Khergade et al. [41] proposed the use of a DVR with a controller based on enhanced synchronous reference frame (ESRF) theory to obtain increased enhancement in the performance of ASD during voltage sags. To provide a common mitigation solution for voltage sags that ensures the optimal use of DVRs by exploiting economies of scale while ensuring partial excludability for customers who may not be willing to pay, Majumder et al. [44] proposed a graph-partitioning principle, which considers fault rates in the network serving the cluster and the customer’s willingness to pay, to derive multiple feasible contribution cluster sets where a single DVR can serve a particular cluster.

Although a conventional three-wire DVR topology is simple and costs less than others, and despite its proven capability to mitigate voltage disturbances, some researchers avoid this DVR based on the understanding that it cannot compensate for zero-sequence voltage disturbances. Despite this, Neto et al. proceeded to demonstrate evidence to the contrary [47], i.e., this DVR can compensate for zero-sequence voltage disturbances and can be used to compensate for unbalanced voltage sags/swells.

3.3. Topologies of DVRs

There has been a constant search for DVR topologies driven by, among other things, the need to enhance technical performance, improve efficiency and incorporate technological developments, as well as the need to better the economics of devices.

Several DVR topologies have been developed and assessed with a variety of improvement objectives. For a single–phase AC–AC converter with a direct power conversion approach and the capability to produce both regulated non-inverting and inverting forms of input voltage at the output, a feature that allows the converter to correct line voltage profiles once the converter is used in DVRs, Ashraf et al. [52] proposed a new circuit topology that was realised with only the use of one full bridge of four insulated gate bipolar transistors (IGBTs) and a full bridge of four diodes, the topology of which is, thus, simpler and has a reduced number of components, as well as reduced volume and size. In [55], to mitigate the impact of grid interruption and voltage sags, Aydogmus et al. presented a flywheel energy storage system (FESS) that was designed, optimised and analysed to operate as a DVR, using a flywheel designed with a natural resonance frequency outside the operating frequency range of the FESS and a matrix converter structure used for an FESS to bidirectionally convert the power between the grid and flywheel. To mitigate voltage sags and swells and reduce variation in the voltage supplied to commercial and domestic devices, Bajaj [58] proposed a hybrid distributed generation (DG) system model powered by a single-phase DVR in which the switching control strategy of the DG system enables the solution of voltage quality problems with minimum grid energy requirements.

To improve voltage stability in distributed-generation-integrated weak utility networks, Biricik [24] discussed a DVR topology in which a self-tuning filter (STF) was utilised with the $\mathrm{pq}$ control method to not only enhance the control performance of the DVR but also eliminate the need to have multiple filters as part of the control method, thereby reducing the complexity of the controller. To ensure good performance of DVRs used to compensate for voltage sags/swells in power distribution systems, which can be adversely affected by large-load circuit currents, Jiang et al. [27] proposed a dual-functional DVR (DFDVR) that can compensate for voltage fluctuation in an upstream grid.

To provide both active and reactive power compensation that can restore voltage values before faults to mitigate high-voltage sags, Karimi et al. [30] proposed a DVR–PV hybrid compensator that uses a fuzzy adaptive controller to improve the performance of the scheme in compensating for voltage interruption and different types of voltage sag. In grid-connected RES systems, the grid can suffer from voltage sags and swells, with adverse impact on consumers; thus, Prasad and Dhanamjayulu [33] presented a solar-PV-integrated DVR using a rotating $\mathrm{dq}$ reference frame controller to mitigate power quality problems, with enhanced incremental conductance maximum power point tracking (INC MPPT) used for the PV plant and proportional–integral (PI) controller for DVRs optimally tuned by using an adaptive neuro-fuzzy inference system (ANFIS).

In [36], Prasad et al. introduced a solar system and z-source inverter (ZSI)-based DVR topology, combining a PV system with a DVR that uses a ZSI voltage restorer to reduce voltage swell under the sudden addition of a balanced three-phase nonlinear load. The perturb and observe (P&O) algorithm was utilised to automatically determine the operating voltage of PV systems that would produce maximum power output, and the performance of this scheme was compared with that of conventional dynamic voltage restorers based on voltage and current source inverters. Rahman et al. [39] introduced DVRs based on direct converters to mitigate voltage sags, with the direct converter realised by using only two bidirectional switches.

Srikanth Babu et al. [42] analysed the performance of a DVR to resolve voltage sags and swells under two scenarios, i.e., scenario 1, in which the DC link was supported by battery energy storage, and scenario 2, in which the DC link was supported by the super-capacitor-supported DC link. To mitigate both voltage sags and swells, Toumi et al. [45] proposed a PV-based single-phase DVR system that requires a power source with a DVR to compensate for the sag/swell voltage, using the SMC method applied to the DVR. To compensate for voltage disturbances, such as voltage sags and voltage swells, Tousi et al. [48] proposed a DVR with a high-frequency link that enables reductions in the volume and weight of the DVR in comparison to conventional DVRs.

Several publications have reported on developments focusing on interline DVR schemes. Chen [50] investigated a new superconducting magnetic energy storage (SMES)-based interline DC dynamic voltage restorer (IDC–DVR) scheme (SMES IDC–DVR) with one SMES coil shared among multiple compensating circuits to exploit the fast-response feature from the SMES device, which predisposes it to suppress instantaneous voltage and power fluctuations, and to improve the energy utilisation rate and reduce the energy storage cost under multiple-line power distribution conditions, with the possibility of selectively controlling the SMES coil to compensate the preferable line according to the priority order of the line. To protect sensitive loads and automated industrial processes in industry from the consequences of poor power quality by mitigating balanced three-phase and unbalanced single-phase voltage sags, Farhadi-Kangarlu [53] proposed a multilevel-converter-based interline dynamic voltage restorer (MLC–IDVR) that uses multilevel converters, in which the IDVR comprises two DVRs that share a DC link and are installed in two independent distribution feeders.

In some works, there has been a strong focus on the use of multilevel converters as a key component of the DVR. In [56], Biricik et al. proposed a hysteresis-band-controller-based three-phase voltage source converter with twelve switches, where a common-mode DC voltage source three-level converter is used in DVR systems to mitigate voltage sags and swells, with the converters in this scheme having the advantages of being simple to control, having robust control and displaying rapid dynamic responses for distorted voltage, voltage dip and swell conditions. De Almeida Carlos et al. [59] proposed and generalised a class of multilevel-inverter-based DVRs, named DVR-cascaded open-end winding (COEW) (DVR–COEW), to compensate for voltage sags/swells experienced by high-power sensitive loads, where the proposed DVR is based on three-phase bridge converters series-connected that employ cascaded transformers and utilise the concept of open-end winding (OEW), and this topology offers the benefit of modularity when compared to other configurations.

Considering the advantages of high-power, high-voltage and low-harmonic levels associated with multilevel converters, as well as the modular structure of these converters, Karim et al. [61] proposed a design of a DVR-based multilevel converter topology to enhance voltage profiles and improve power quality in the network. To overcome the voltage swell and sag problems associated with grid-connected RESs, Prasad and Dhanamjayulu [63] proposed a solar-PV-integrated DVR that utilised a 23-level multilevel inverter, which also used an improved incremental conductance maximum power point tracking (INC–MPPT) technique for the boost converter to ascertain that maximum energy is extracted from solar PV modules. To mitigate voltage sags in medium-voltage and high-power applications, Rajkumar et al. [65] propose a T-type multilevel-inverter-based DVR, using a DVR controller based on the $\mathrm{abc}$ to $\mathrm{dq}$ controller, employing a scheme with a reduced switch count, which can be directly connected to a medium-voltage distribution system without injection transformer, offering several advantages.

3.4. Strategies for DVR Controllers

Controllers play a central role in the performance of DVR schemes. Voltage sag detection is a key component of the controller and is essential for capturing the instance of occurrence of a voltage sag and determining the depth of a sag, both to be further used in generating a reference voltage for the controller of a voltage-interactive device, such as a DVR. A variety of schemes have been proposed for enhancing the quality of voltage sag detection. To aid the acquisition of high-precision sag information for voltage sag detection under nonideal grid conditions (e.g., unbalance), Li et al. [67] proposed a selective harmonic extraction algorithm (SHEA) to enhance the immunity of sag detection.

For the generation of reference signals for a multifunctional operation of single-phase dynamic voltage restorers, Ahmed et al. [69] proposed an enhanced single-phase quasi type-1 phase-locked loop (QT1–PLL) that offers superior harmonic robustness over conventional QT1–PLL. Boussaid et al. [71] proposed an algorithm based on an instantaneous phase-locked loop (IPLL) using a multi-variable filter to synthesise unitary signals involved in compensation voltages required for compensation of voltage sags, thereby enhancing the control efficiency of the DVR in mitigating voltage sags. For voltage compensation using a DVR, Martins et al. [73] proposed a method that relies on the construction of a voltage reference that considers the load power factor and utilises a recursive least square (RLS) technique to estimate the grid voltage magnitude and phase, delivering control action that is robust to harmonic distortions, sags and swells.

To improve the estimate of the reference voltage for a transformerless T-type-converter-based DVR, Rajkumar et al. [75] proposed multiple delayed signal cancellation (MDSC) phase-locked loops (MDSC PLLs) with shorter delay times and smaller memory requirements for the accurate estimation of the grid phase angle in order to improve DVR performance for various grid conditions, such as voltage sag/swell, unbalanced voltage sag/swell, harmonics and DC offset. SasiKiran and Manohar [77] introduced a DVR aimed at mitigating voltage swell challenges that uses an estimation method, for the symmetrical components of the supply voltages, based on an unscented Kalman filter (UKF), to assist the DVR control algorithm in generating reference signals for the voltage source converter (VSC) of the DVR, and this estimation method ensures that voltage swells are accurately and speedily discovered to remove them quickly, thereby protecting sensitive loads connected to distribution systems.

To protect voltage-sensitive loads from balanced/unbalanced voltage sags/swells in a distribution system, Singh et al. [79] proposed a DVR controller using the least mean fourth (LMF) algorithm to calculate the reference load voltage for the VSC of a DVR; the algorithm did not require a PLL and, thus, had reduced complexity as well as increased robustness and easy implementation. The DVR exhibited superior dynamic response when compared to conventional adaptive least mean square (LMS) and SRF theory-controlled DVRs. To reduce the response time of a single-phase DVR aimed at mitigating the impact of voltage sags under a wide range of operating conditions, Zhang et al. [25] proposed a fast detection method for grid voltage sag based on comparisons of the multi-timescale variables of restructured voltages and a fast and reliable commuting strategy for bypass thyristors in the DVR based on a double-voltage-step method.

Researchers have also developed and evaluated various controller strategies for DVRs. Ahmed et al. [28] introduced a continuous terminal sliding mode controller (CTSMC) for single-phase DVRs that can significantly reduce chattering compared to the first-order SMC while achieving faster convergence speed compared to its super-twisting counterpart. For a DVR aimed at mitigating voltage sags, Inamdar and Iyswarya [31] introduced a terminal sliding mode control approach to design a DVR controller to improve the accuracy and robustness of a DVR.

To mitigate voltage variation due to voltage sag in the source voltage to which the critical load was connected, Kassarwani et al. [34] proposed a modified sliding mode control (SMC) method based on SRF theory. This method controls voltage injection using a DVR to regulate the terminal voltage of the critical load, restoring it to its pre-sag value. Notably, the SMC approach is robust and independent of system parameters. For a DVR intended to maintain voltage loads at constant values amidst voltage sags, Mohammed and Ariff [37] proposed a DVR model based on an approximate classical sliding mode differentiator (ACSMD) and a terminal sliding mode controller (TSMC) to improve load voltages after various types of faults (viz, three-phase, double-line-to-ground and single-line faults, as well as voltage imbalance conditions).

To maintain the magnitude of load voltage near 1 per unit, despite voltage sag/swell, Mohammed et al. [40] proposed the use of a robust differentiator called ACSMD in the DVR controller. This differentiator has low complexity, increases robustness and ensures the steady-state error stays within a small band. The most important problems of power quality are voltage sags and harmonics, which cause tripping or malfunctioning of equipment. Nasrollahi et al. [43] proposed a three-phase three-wire DVR to compensate for voltage sag, swell, undervoltage and overvoltage under balanced and unbalanced conditions, implementing a second-order generalised integrator (SOGI)-based phase-locked loop (PLL) in the DVR controller utilizing sliding mode control SMC to increase controller robustness against parameter variations, guarantee stability, simplify implementation and increase the speed of dynamic response.

To compensate for voltage sag/swell and reduce adverse impacts on customers and utilities, Shah et al. [46] proposed a DVR with an ultra-capacitor that uses a controller based on real twisting sliding mode control to successfully ameliorate the impacts of voltage sag/swell. To protect sensitive loads from voltage sags, Zahra et al. [49] proposed, modelled, analysed and simulated a DVR, with two controllers, namely, proportional integral controller and sliding–mode controller, considered for this DVR, and the effectiveness of these controllers was evaluated.

Controller strategies based on SRF theory have been widely studied in the context of solving voltage problems using a DVR. Al-Gahtani et al. [51] presented a DVR controller implementing a newly proposed version of the synchronous stationary frame ($\mathrm{dq}$) theory to control the DVR under severe transient voltage conditions. To enhance the dynamic and fast response of the DVR, Al-Gahtani et al. [54] proposed a modified instantaneous reactive power ($\mathrm{pq}$)-theory-based controller for a DVR operating under extreme transient voltage conditions.

To resolve voltage-related problems using a DVR, Danalakshmi et al. [57] proposed a DVR controller strategy based on the SRF theory due to the simplicity and suitability of this strategy. For simultaneous compensation of both symmetrical and asymmetrical voltage sags and swells, as well as harmonics, Inci et al. [60] introduced a controller method, called fast Fourier transform (FFT) with integrated improved synchronous reference frame (ISRF), for a multifunctional DVR a controller, in which an ISRF quickly and precisely detects the magnitudes of voltage sag/swell, and FFT very effectively extracts the selective components of voltage harmonics.

For mitigation of balanced and unbalanced voltage sag/swell as well as balanced and unbalanced voltage sag/swell with harmonic distortion, Singh et al. [62] introduced a self-supported DVR utilising a reduced-rule fuzzy logic controller based on SRF theory. Notably, this controller does not require a mathematical model of the plant and adequately provides control signals from the source side, resulting in a controller with superior performance compared to some existing techniques, simple implementation, lower complexity, high flexibility and robustness. Additionally, it requires less reactive power to compensate for various disturbances at the source side compared to a PI-based controller, resulting in reduced ratings of IGBTs and, as a result, lower overall costs.

Vo Tien et al. [64] present a DVR that utilises a controller based on a vector controller to minimise voltage sags in the grid, employing a structure built on a rotating dq coordinate system, which uses two cascaded loops to improve the properties of the DVR system (i.e., the system performs with higher accuracy, has a faster response and has lower distortion in voltage sag compensation).

The PI theory has also formed the basis of designing several controllers for DVRs. To mitigate voltage swell and sag, Danalakshmi et al. [66] studied the feasibilities of various PI and proportional resonant (PR) controllers for use in DVR control and assessed the performance of these controllers in detecting disturbances and injecting voltage to compensate for load voltages under these disturbances. Elkady et al. [68] presented an enhanced, optimised and less complex PI controller, which was implemented in a $\mathrm{dq}$ synchronous rotating reference frame, for a DVR that can enhance transient, steady-state and dynamic responses of the DVR scheme and eliminate inherent transient oscillations. For a DVR employed to compensate for continuous voltage sag in a distribution system network supplying linear as well as non-linear loads, Reddy et al. [70] proposed a control strategy using a PI-tuned fuzzy logic scheme that is cost-effective, validating the proposed control technique using MATLAB/Simulink software (https://www.mathworks.com/products/simulink.html last accessed 3 October 2023) to demonstrate its working performance and improved performance over a conventional DVR control in an efficient manner.

Some controller design strategies have relied on the PR theory for their development. To improve the performance of a DVR in maintaining the voltage of sensitive loads at nominal value amidst voltage sags and swells, Hung et al. [72] introduced a two-stage loop circuit control strategy for the DVR in which there is an external voltage control loop that uses a sequence-decoupled resonant (SDR) controller and an inner current control that utilises a PR controller implemented in a stationary frame ($\mathsf{\alpha }\mathsf{\beta }$). For a DVR that utilised a matrix converter (MC) as its core element and obtained compensation energy directly from the power system, which is intended to cope with balanced and unbalanced variations (sags and swells) as well as harmonic distortion, Merchan Villalba [74] presented a design of a decoupled linear control strategy for a DVR. For a multifunctional DVR, Tien et al. [76] proposed a new control method built in a stationary frame by combining PR controllers and sequence-decoupled resonant controllers, assessed the performance of the scheme under various conditions (e.g., voltage sag/swell due to symmetrical/unsymmetrical short circuits) and conducted simulations to assess the performance of the scheme.

Several works have reported on ANFIS-based controllers. To mitigate voltage sags, such as those caused by overload or the starting of electrical motors, and improve power quality in low-voltage distribution systems, Kalyanasundaram et al. [78] proposed the use of a DVR equipped with an ANFIS controller utilising a backpropagation-based algorithm.

3.5. Optimisation of DVR Controllers, Sizes and Locations

In some publications, robust optimisation algorithms have been considered for performing optimisation tasks. In comparing the control strategies of DVRs aimed at resolving voltage swell and sag in distribution systems so that an acceptable voltage profile can be maintained in case of disturbances, Danalakshmi et al. [66] compared the performance of PI and PR controllers, for which the optimal gain values to achieve the target of minimising global error and ensuring fast responses are determined using a robust optimisation algorithm called self-balanced differential evolution (SBDE).

Other research works have utilised metaheuristic algorithms to achieve optimisation objectives. To improve transient, steady-state and dynamic responses of a DVR, as well as to eliminate inherent transient oscillations in load voltage, Elkady et al. [68] proposed a DVR PI controller that is implemented in a $\mathrm{dq}$ synchronous rotating reference frame and has parameters selected using a population-based optimisation technique called the Harris hawks optimisation (HHO) technique, and the system response results are compared to the PSO and whale optimisation algorithm (WOA). To control a DVR intended for mitigating different voltage sags in the supply voltage of a distribution system supplying linear and non-linear loads, Kassarwani et al. [80] proposed a PI controller that utilises an SRF-theory-based control algorithm, an efficient improved particle swarm optimisation technique to optimise the gain parameters of the PI controller and integral squared error to assess the performance and suitability of the DVR, with extensive results showing that the proposed method outperforms conventional tuning methods, such as Ziegler–Nichols method and genetic algorithm (GA), for controller gain parameter optimisation.

To mitigate voltage sags affecting sensitive loads and to maintain acceptable voltages at the load terminal, Kassarwani et al. [81] proposed the use of a DVR, as a compensator, that uses PI controllers based on the SRF algorithm, with the values of the gains in the PI controllers optimised using the genetic algorithm. Tiacharoen and Chatchanayuenyong [82] considered both the use of a DVR in protecting critical or sensitive loads from voltage sags or swells and three potential controllers for the DVR, namely, a proportional–integral–derivative (PID) controller, an SMC controller, and a fuzzy sliding mode (FSM) controller; proposed the use of the bee algorithm (BA) to optimally to design the controller; and showed that the proposed strategy outperforms designs obtained through the trial-and-error method. Studying the problem of the optimal sizing and location of a DVR in a distribution network to meet customer demand for mitigating voltage sags, Zhong et al. [83] formulated an optimisation problem. The objective was to minimise the investment in DVRs, which was expressed as an objective function in terms of the unit price per kVA of DVR, with the constraint expressed as customer satisfaction and represented using a mitigation expectation index (MEI), indicating the proportion of sag events affecting customers that are expected to be mitigated; then, the optimisation problem was solved using the multiple population genetic algorithm (MPGA).

In some instances, trial-and-error approaches have been used as techniques for optimisation. To control a DVR aimed at mitigating voltage sags and swells, a PI controller was proposed by Salman et al. [84], with the parameters of this PI controller tuned using the trial-and-error method as well as the PSO method, and a comparison of the results was shown that indicated the better performance of the DVR using the PI controller tuned using PSO in terms of rise time, maximum overshoot and settling time, as well as total harmonic distortion (THD). For a DVR employed to tackle power quality problems, such as voltage sags and swells, spikes, distortions, etc., Salman et al. [85] proposed the use of PI and ANFIS for a DVR, with the settings of the PI controller firstly fine-tuned using the trial-and-error method and, secondly, the PSO, with latter being more effective based on settling time, overshoot, undershoot and disturbances around the final value. Thereafter, the ANFIS controller was employed to regulate the DVR’s responsiveness through the PI–PSO controller, and this scheme returned improved results.

Optimisation approaches based on ANFIS have also been used. To mitigate voltage sags and swells associated with the interaction between the grid and a grid-connected RES, Prasad and Dhanamjayulu [33] proposed a solar-PV-integrated DVR that uses a PI controller implemented in a rotating $\mathrm{dq}$ reference frame, with the optimal parameters of the controller tuned by using the ANFIS.

3.6. Platforms for Feasibility Evaluation of DVR Controllers

To evaluate the feasibility of DVR topologies and the effectiveness of proposed controller designs, various platforms have been used to perform the necessary studies. In many cases, one platform is used to validate the proposed approaches, and another platform is utilised to corroborate and verify the results from another platform. Some studies have relied on simulations in software to perform the evaluations. Simulations in this field of work are a very common and popular approach. One of the software packages used in this area of work is MATLAB/Simulink, which was utilised by Danalakshmi et al. [57] to study a specific DVR as a solution to shield loads from the distortion of utility voltages due to faults. This DVR has a typical three-phase inverter that transforms DC to AC and vice versa using a DC link capacitor, utilising a SRF-theory-based control strategy. The authors described its topology and implemented the whole DVR scheme in software to verify and validate the capability of the DVR.

Another software that is often used for simulations in this area of research is power systems computer-aided design/electromagnetic transients including DC (PSCAD/EMTDC). In their work, Farhadi-Kangarlu [53] proposed a new structure for IDVR that is based on multilevel converters, called multilevel-converter-based IDVR (MLC–IDVR), and verified its performance using simulation results carried out on PSCAD/EMTDC software (https://www.pscad.com/ last accessed 3 October 2023). A scheme proposed by Tousi et al. [48] comprising a DVR with a high-frequency link and intended for compensation of voltage sags and swells, which also enables reductions in the volume and weight of the DVR in comparison to conventional DVRs, was implemented and simulated in PSCAD/EMTDC software and examined under several disturbance conditions to verify its performance features.

Building laboratory-scale prototypes with the objective of experimental verifications is another popular approach in this area of work. Several prototypes have been based on a Texas Instrument C2000 Series micro-controller. In [69], Ahmed et al. proposed a single-phase QT1–PLL for generating reference signals in a multifunctional, single-phase DVR that is easy to implement, can provide rejection of any measurement offset and has better harmonic robustness compared with a conventional QT1-PLL, then built a laboratory-scale prototype in which a Texas Instrument C2000 series micro-controller was used to implement the proposed control and estimation algorithm in the experimental setup. The experiments conducted verified the method, demonstrating that the proposed method rapidly detects and compensates for any grid voltage anomalies, thereby maintaining constant load voltage despite voltage sag, swell and harmonic distortions.

Some laboratory DVR prototypes have relied on the use of dSPACE control boards for implementation. To assess the effectiveness of a DVR controller that was implemented using a newly proposed version of a synchronous stationary frame ($\mathrm{dq}$ theory), Al-Gahtani et al. [51] built a model of a complete system in MATLAB/Simulink and also constructed a laboratory-scale model of the proposed model, with the control circuit built in the dSPACE DS1104 control board, and then compared the obtained results with those of other algorithms and correlated the simulation and experimental results to evaluate the effectiveness of the proposed $\mathrm{dq}$ control scheme under severe transient voltage conditions. To enhance the dynamic and fast response of a DVR, Al-Gahtani et al. [54] proposed a modified instantaneous reactive power ($\mathrm{pq}$)-theory-based controller for a DVR operating under extreme transient voltage conditions to compensate for severe balanced, unbalanced (voltage sags and swells) and load changes. Simulations were conducted using MATLAB/Simulink to verify the mathematical models of the conventional $\mathrm{pq}$ and proposed $\mathrm{pq}$ control system for the DVR. The complete system was tested experimentally by building a laboratory system made of a three-phase AC voltage source, a transformer, a VSI powered by a DC power source and a 1 kVA load, with the control circuit based on the digital signal processing type dSPACE (DS1104), and the simulation and experimental results were correlated to demonstrate the efficacy of the modified $\mathrm{pq}$ control technique.

To demonstrate the performance of their proposed algorithm, which is based on IPLL using a multi-variable filter to synthesise unitary signals involved for compensation voltages required to mitigate voltage sags and enhance the efficiency of DVR control, Boussaid et al. [71] ran software simulations and conducted experimental validations based on a set-up comprising hardware and a controller, implemented using the dSPACE 1104 platform.

To avoid the abnormal operation of consumer electronics and domestic household appliances under weak grid conditions, Mallajoshula and Naidu [86] proposed an improved, robust SRF (IR–SRF) controller. The proposed scheme was implemented in a MATLAB/Simulink environment to analyse and evaluate its performance and a laboratory model of the system with a three-phase IGBT bridge of SEMIKRON used as a connected DVR. The embedded control board, dSPACE, was used to generate the required switching pulses for the IGBT switches of the DVR, and finally, the results of simulations and experimental results were correlated, and the performance of the IR–SRF was compared with a conventional control algorithm to validate its efficacy. Parreño Torres et al. [87] presented a discrete-time domain control scheme, implemented in an SRF, for balanced voltage sag compensation using a DVR. The time-domain performance of the entire system was analysed through detailed numerical simulations using a model implemented in PSCAD/EMTDC and experimental studies carried out with a 5 kW DVR, with the control algorithm for the DVR implemented using the DS1103 real-time platform. Finally, the discrete-time control method was compared with two control schemes previously proposed in the literature.

Roldán Pérez et al. [88] presented a systematic method for designing a high-performance proportional resonant controller (PRC) in discrete applications, and the effectiveness of this proposed control system was tested and analysed in a 5 kVA prototype of a three-phase series active conditioner (SAC) that protects a linear load against voltage sags and harmonics, which comprised a laboratory test rig representing the grid and the DVR, with the controller implemented on the dSPACE DS1103 platform.

Some laboratory-scale prototypes of the DVR have had control structures implemented via a Typhoon HIL402 real-time emulator kit to perform hardware-in-loop testing experiments. To validate their proposed, enhanced, optimised and less complex DVR PI controller, which was implemented in a $\mathrm{dq}$ synchronous rotating reference frame, that is capable of enhancing transient, steady-state and dynamic responses of the DVR scheme, and eliminating inherent transient oscillations, Elkady et al. [68] modelled the system and ran simulations using MATLAB/Simulink. The results were validated using hardware in-loop (HIL) testing experiments. The DVR control structure was implemented via a Typhoon HIL402 real-time emulator kit. The simulation and emulation results demonstrated the quick recovery of normal operation after voltage disturbances without overshoot and with near-zero steady-state error, exhibiting significant damping of inherent voltage oscillation that occurs upon DVR entry and/or exit. For their proposed proportional-resonant DVR controllers used to mitigate temporary unbalanced voltage affecting sensitive loads, e.g., industrial manufacturing or communication devices, Vu et al. [89] verified the feasibility of their approach via simulation through MATLAB/Simulink and verified the results by performing HIL real-time experiments using a Typhoon HIL402 device.

In some laboratory DVR prototypes, the control structure was implemented using OPAL–RT real-time digital simulator. Li et al. [90] introduced an optimal zero-sequence voltage injection strategy (OZVIS) for a three-phase DVR to improve the compensation capacity of the device in case of asymmetric sags in a three-phase, three-wire power system, and experimentally verified the feasibility and effectiveness of OZVIS, under asymmetric and symmetric situations, by implementing a three-phase, three-wire power grid and the three-phase DVR in RT–LAB, which is the software platform of OPAL–RT’s simulation systems for real-time simulations, with a step of 10 µs, and using an interface card to link the output to the I/O port of an external STM32F407 ARM controller.

For a solar-PV-integrated DVR that utilises a 23-level multilevel inverter (with advantages of reductions in the overall component count, cost and size of the inverter), to overcome voltage swell and sag problems associated with grid-connected RESs, Prasad and Dhanamjayulu [63] proposed an improved INC–MPPT technique and a DVR controller implemented in a rotating $\mathrm{dq}$ reference frame control, with the dynamic performance of the DVR evaluated using a balanced load. The proposed scheme was implemented using MATLAB/Simulink, and simulations were conducted for assessment, and experiments were run utilising the HIL testing platform, employing the Opal RT OP5600 simulator, and developed to verify the simulation results.

For a DVR aimed at protecting sensitive loads from balanced/unbalanced voltage sags/swells in a three-phase, three-wire distribution system, utilizing their proposed adaptive LMS controller, which employs energy-minimised voltage compensation, making this DVR self-supported in nature, i.e., there is no need for an extra energy storage device and only a capacitor is sufficient at the DC bus of the VSC, Pratap Singh et al. [91] verified the effectiveness of the algorithm by conducting simulations in MATLAB and verifying the results using simulation studies that performed a test of the algorithm through real-time experimentation using OPAL–RT (real-time digital simulator).

To improve voltage stability in distributed-generation-integrated weak utility networks, Biricik [24] discussed a DVR topology in which an STF is utilised with the pq control method to enhance the control performance of the DVR, but also eliminated the need to have multiple filters as part of the control method, thereby reducing the complexity of the controller. The proposed DVR control method was modelled using MATLAB/Simulink and also tested in real time; a software-in-the-loop (SIL) set-up was utilised to verify the correct operation of the environments using the OPAL RT real-time platform. The results were then presented as a verification of the proposed system.

3.7. Research Trends and Prospects for Future Research

3.7.1. Methodology

Based on the 68 eligible publications downloaded from the Scopus database and included in the survey, keywords from these publications were utilised to conduct an analysis of research trends and to identify potential areas for future research. These tasks were enabled using VOS viewer [92], a software tool for constructing and visualising bibliometric networks, with version 1.6.19 of the software used. For each keyword meeting this criterion, the total strengths of the co-occurrence links were calculated.

3.7.2. Results

The keywords that met the criterion of at least three occurrences are presented in

Table 4. List of keywords with at least 3 occurrences, with the number of occurrences and total link strengths for each of those keywords shown.

Cluster	Keyword	Occurrences	Total Link Strength
1 (red)	Control systems	3	25
	Dynamic voltage restorer	68	309
	Harmonic	3	22
	Harmonic analysis	6	53
	Power quality	41	231
	Pulse width modulation	4	25
	Resonance	3	15
	Total harmonic distortion	9	36
	Transients	3	20
	Voltage control	7	67
	Voltage fluctuations	7	49
	Voltage harmonics	3	23
	Voltage sag and swell	8	34
	Voltage swell	10	71
2 (green)	Electric inverters	4	34
	Electric power system control	4	45
	Electric power transmission networks	6	54
	Matlab	10	106
	Maximum power point trackers	3	29
	Multilevel converter	3	25
	Photovoltaic	5	44
	Renewable energy source	3	41
	Series compensation	3	9
	Solar power generation	3	35
	Topology	4	30
	Voltage regulators	33	230
3 (blue)	Converter control system	13	101
	Genetic algorithm	3	21
	Particle swarm optimisation	5	35
	Proportional and integral controllers	3	18
	Sliding mode control	9	50
	Sliding mode controller	4	25
	Synchronous reference frame	4	22
	Synchronous reference frame theory	4	25
	Voltage sag	31	193
	Voltage source converter	4	18
4 (yellow)	Adaptive control systems	4	38
	Electric fault currents	4	24
	Phase-locked loop	5	47
	Power distribution system	6	32
	Power system dynamics	3	33
	Quality control	10	99
	Reactive power	3	33
	Smart grid	3	24
	Transformer	3	30

, with the number of occurrences and the total link strength for each of these keywords provided. In total, 45 keywords met the criterion. In VOS viewer, these qualifying keywords were categorised into four clusters of relativity; thus, these clusters could be used to determine the focus areas of the eligible publications.

A network visualisation map showing the overall linkages between the keywords is presented in Figure 5. The four clusters discussed in the previous paragraph can be observed in this map, with each of the four colours (red, green, blue and yellow) associated with one of the clusters (1, 2, 3 and 4, respectively).

By observing Table 4 and Figure 5 together, the following can be stated. Fourteen keywords belong to cluster 1 (red), with the keyword “dynamic voltage restorer” having the highest number of occurrences (68) and the highest total link strength (309). There are 12 keywords in cluster 2 (green), with “voltage regulators” having the highest number of occurrences and the highest total link strength at 33 and 230, respectively. In cluster 3 (blue), there are 10 keywords, and “voltage sag” has the highest occurrences at 44 and the highest total link strength at 193. Finally, cluster 4 (yellow) has nine keywords, with “quality control” having both the highest number of occurrences and the highest total link strength at 10 and 99, respectively.

In Figure 6, an overlay visualisation map is presented, showing the linkages between keywords and also highlighting the years in which keywords were used over the period 2018–2022. From this map, trends of how the research foci have evolved over this time can be assessed. In the early phases of the period, research focused on the keywords in blue, then shifted to those in dark green, then to those in light green, and the latest phases focused on the keywords in yellow.

Looking at Figure 5 and Figure 6 simultaneously, many keywords in clusters 1 and 3 of Figure 5 correspond to the focus in years of blue and dark green in Figure 6, which is what was conducted in the early phases of the period of interest. On the other hand, the later and recent focus, represented by light green and yellow in Figure 6, seems to mainly encapsulate the keywords in clusters 2 and 4.

Some of the keywords that are at the core of this review were selected, and overlay maps for them were plotted. These words were “dynamic voltage restorer”, “voltage sag”, “voltage swell”, “voltage fluctuations” and “voltage fluctuations”. The overlay maps for each of these keywords are plotted in Figure 7, Figure 8, Figure 9 and Figure 10, with the same meaning of colours as described in the overlay visualisation map shown in Figure 6.

An analysis of these figures reveals the following:

• In Figure 7, the reference keyword is “dynamic voltage restorer”. In earlier years, the research focus was on converters, resonance, series compensation and optimisation of DVR using genetic algorithms. Later, the focus shifted to strategies for the design of controllers for DVR to mitigate voltage sags and swells, transients and reductions in fault currents. In addition, there was a focus on the application of DVRs to solve problems in smart grids. Recently, there has been a focus on the application of DVRs in voltage problems brought about by the integration of renewable energy sources.
• Patterns very similar to what was said about Figure 7 can be observed in Figure 8, where “voltage sag” is used as a reference keyword, and in Figure 9, where the reference keyword is “voltage swell”, except that in the latter there is also a focus on power system dynamics, which is highlighted as an area where attention was focused in the later phases of the period.
• Finally, some observations can be made by closely assessing Figure 7, Figure 8, Figure 9 and Figure 10, with Figure 10 being the overlay visualisation map in which the reference keyword is “voltage fluctuations”. In Figure 7, it is demonstrated that there is a weak linkage between “dynamic voltage restorer” and “electric power system control”. Again, looking at Figure 7, there is a weak link between “dynamic voltage restorer” and “power system dynamics”. Also, looking at the same figure, there is a strong link between “dynamic voltage restorer” and “genetic algorithm” for optimisation applications related to DVR. This does not extend to other available algorithms. These observations point to some gaps that can be closed by conducting related research in the future.

4. Conclusions and Proposed Directions for Future Research

In this study, the author discussed the results of a survey on the use of DVRs to alleviate voltage problems that lead to sags, swells and fluctuations in voltage beyond acceptable operating limits. To conduct and report on the review, the author adopted a research methodology based on the PRISMA statement, with the Scopus database used as the source for the publications to be included in the survey. The author compiled a bibliometric analysis addressing the number of publications per annum, the top 10 most-cited journals, and the top 10 most-cited publications.

From the set of publications included in this survey, information was extracted, summarised and categorised into network scenarios for the use of DVRs, topologies and optimisation of DVRs; strategies for DVR controllers; and platforms that evaluate the feasibility of DVR topologies and controllers. Moreover, research trends and gaps were evaluated. This study provided an overview of the research on the use of DVRs to solve voltage problems and is a resource for researchers generally interested in DFACTS and particularly interested in DVRs.

From the foregoing discussion, especially considering the potential gaps in the research that were identified earlier in the manuscript, the following are proposed directions for future research:

• Research should be extended to assess the feasibility of using DVRs to solve power flow problems, especially in networks with high stochasticity in loads and generation, addressing techno-economic issues, among others.
• The value of using DVRs should be addressed to enhance stability, i.e., voltage stability, small signal stability, and transient stability of power systems, including techno-economic comparisons with other solutions.
• There are an increasing number of situations in which embedded generation units are installed at medium network voltage levels, with series compensation, at times, required to control power flows. The use of DVR to mitigate sub-synchronous resonance (SSR) in such networks should be investigated.
• Research into the optimal design of DVRs should be strengthened, especially the aspect of optimisation of designs, using various optimisation techniques, such as mathematical programming and metaheuristic algorithms.
• It is proposed that DVR failure and outage data should be gathered and analysed to gain insights into the frequency and duration of outages. Such insights can assist in devising strategies to reduce the adverse impacts of failures and outages on the reliability and quality of supply.