Improvement of Transient Performance in Microgrids: Comprehensive Review on Approaches and Methods for Converter Control and Route of Grid Stability

In conventional power systems, stability is ensured successfully because of large synchronous generators, where, during transient conditions, the inertia and dynamics of the synchronous generators help to retain stability. On the contrary, inverter-interfaced distributed energy resources suffer from maintaining stability because of their quick dynamics (opposite to the concept of inertia). Therefore, there exists a tradeoff between response and stability aspects and this is referred to as transient performance problem in microgrids. Conventionally, this problem is addressed from the controller side by increasing its bandwidth thereby leading to better disturbance rejection and immunity against parameter changes. Since the classification of the disturbances (as small or large) is not straightforward and uncertain, an increase in the controller's bandwidth beyond a certain limit is not possible. Therefore, it is understood that the transient performance improvement cannot be achieved from the controller alone. Hence, the chances of improving performance from other approaches in the case of grid-connected mode and advanced control strategies for the controller in the case of islanded mode need to be investigated. With this intent, a comprehensive investigation of state-of-the-art approaches and methods to improve transient performance in microgrids is carried out in this article. This work is carried out in three stages. Initially, it explores various technical challenges that are involved and possible approaches to address these challenges. Next, a critical review of each approach is carried out based on their philosophy to improve transient performance. Furthermore, a comparative analysis that projects the scope of each approach is elucidated. Finally, some future research directions are proposed to enhance the effectiveness of key approaches.

Transfer function of computational time-delay.G zoh (s) Transfer function of PWM time-delay.K PWM (s) Transfer function of PWM inverter.

I. INTRODUCTION
The microgrid is one of the recent greatest evolutions of power generation systems that can effectively augment the central utility grid by supplying the local loads, without depending on the utility grid [1].This can protect the utility grid from outages due to unplanned/sudden load changes.Generally, microgrids are formed with two kinds of DERs such as primary energy sources (e.g., wind and solar energy) and secondary energy resources (e.g., diesel generators, fuel cells, and batteries), which usually form a local dc-bus.Furthermore, to connect this dc-bus to the loads or distribution grid, the microgrid normally uses a VSI.The layout of a typical microgrid is shown in Fig. 1.The key advantages of these VSI-interfaced DERs are easy deployment, fast response (normal or transient), simple operation, and inherent closed-loop control to meet the desired operating point.These fast dynamics allow speedy disturbance rejection, thereby resulting in enhanced transient response in microgrids [2].Supplying required active and reactive power while keeping the frequency and magnitude of voltage within the tolerance limits is the fundamental objective of microgrids.However, the regulation of the microgrid's voltage and frequency is very much influenced by the type of control strategy that is employed [3].
Besides, the feature of fast dynamic response along with inherent no (or very small) reserve capacity (that affects the system inertia) of DERs is a significant setback in terms of transient stability margin.As a result, this introduces a large tradeoff between transient response and transient stability margin, which is referred to as a transient performance problem.Thus, a solution for this problem shall achieve a reduced tradeoff between these two indices (transient response and transient stability margin).
IEEE 2030.7 standard [4] defines two steady-state operating modes with four types of transitions between them for microgrids, as shown in Fig. 2. As per this standard, the transient performance problem will have a holistic solution only by considering the disturbances that occur during the steady state modes and the transients that occur during the transition between those modes.Besides, there are some key review works in the literature that discussed the transient performance issue in microgrids concerning a specific technical challenge or approach, given as follows.
An overview of islanding detecting methods and singlephase FLLs is provided in [5], [6].A comparative review of various aspects of inductors together with the LCL filter's   design procedure is discussed in [7].The evolution of control strategies based on multiloop feedback is presented in [8].In [7], an overview of active damping techniques based on state feedback is presented.Toward transient performance improvement, Singh et al. [9] have examined the potential benefits and drawbacks of droop control techniques for microgrids operating in SA mode, while Liu et al. [8] have made a comparison of controllers operating in GC mode.The VSI current control strategies for unbalanced faults based on how the current references are created are examined in [10].A virtual impedance loop was originated for the objective of providing disturbance rejection, improved stability, or as a reference generator for ancillary services.A review of this control strategy based on when applied to current source and voltage source converters is provided in [11].A survey on the most important VSG topologies is provided in [12].A review and comparison of VSG are provided in [13].Furthermore, based on the control techniques used, Khodadoost Arani et al. [14] explored the use of various ESS.In the same work, a comparative review of various ESS based on energy density, power density, response time, and so on.Each of these works focused on a different technical challenge or approach.In most of these works, only a subset of the modes is considered as depicted in Fig. 2. Furthermore, in many of these works, either the response or the stability issue is addressed, but not both.
Generally, better transient performance features such as maintaining voltage and frequency, low voltage ride-through, and more, are aimed to be achieved from inverter-interfaced DERs through the controller.Conventional classical controllers are mostly designed based on the assumption that the resiliency against disturbances is taken care of by the grid.Hence, the stability aspect is taken care of by the grid in the grid-connected mode.Concerning the improvement of the response aspect in this mode, increasing the bandwidth of the controller is a possible solution.However, increasing bandwidth beyond a certain limit is not possible due to noise and uncertainties in disturbances.Therefore, in gridconnected mode, the enhancement in transient performance is achieved by supporting the controller with improved grid synchronization approaches (e.g., PLL, virtual impedance, islanding detection, etc.) and an effective filter design approach.However, in the case of islanded mode, the support from the grid to ensure stability is lost and the classical controllers cannot handle this condition.In this situation, an advanced converter control approach (e.g.,: virtual inertia, power oscillation damping, etc.) is the only source of help to improve the performance.To the best of the authors' knowledge, a review work addressing this context is not comprehensively available in the literature.Thus, this present work systematically reviews state-of-the-art approaches and methods in a new context that leads to improved transient performance (response and stability) and reliability of microgrids during grid-connected, islanded, and other transitory modes.
Accordingly, a thorough literature review is performed to investigate various technical challenges and possible approaches/methods to improve the transient performance in microgrids.Based on these identifications, a comparative analysis of each approach and its scope in handling the identified challenges is performed.Furthermore, potential future directions are proposed.All these key constituents of this review paper are structured as shown in Fig. 3.

II. TECHNICAL CHALLENGES LEADING TO THE IDENTIFICATION OF APPROACHES
In this section, various factors that cause the deterioration of transient performance in microgrids are described.All the technical challenges are summarized and presented as shown in Fig. 4. The implications of these challenges are explained as follows.
1) Grid voltage distortions: The IEEE 929-1988 [15] standard recommends the allowable ranges for the current and voltage harmonic distortion introduced by the inverter, but they are based on a perfect sinusoidal source with a fixed source impedance.However, especially while connected to remote or weak networks, the inverter must work with the deformed grid voltage.2) Grid impedance variations: In the case of a stiff grid (small impedance), a disturbance may introduce a small-signal stability problem.However, the same disturbance will cause a large-disturbance stability problem when the grid is weak (large impedance) [16].Under grid impedance variation, a feedforward control loop, which is often employed to attenuate the grid voltage disturbances, can assist in maintaining the high current control bandwidth.But, when a feedforward loop is used, the changes in grid impedance are also brought inside the control system.Under the grid-impedance change, this could result in unstable open-loop poles, rendering the LCL filter design ineffective [17].3) Nonlinear loads: Lower order harmonic components emerge within the inverter output current due to the presence of nonlinear dynamic loads.While flowing through the filter-inductor, these components would cause a drop in the voltage.Consequently, the waveshape of voltage at the inverter's terminals gets corrupted.This issue becomes even more problematic when the output impedance is higher [18].4) Time delays: Due to the computing and PWM delays, damping performance will behave like a negative resistor which can certainly induce instability.The rise in time delay lowers bandwidth, thereby worsening the dynamics of the system [19].

5)
Changes in f res : Since the line inductance is unknown, which may be large, the LCL filter's actual f res is never accurately known.The current controller's bandwidth depends on the design of the grid filter.But, once the grid filter is designed, one cannot afford to redesign.Therefore, a grid filter designed for stiff grid conditions will show poor attenuation in weak grid conditions.This puts a limitation on the controller's bandwidth and the overall dynamic performance of the system.6) Coupling between active and reactive power controllers: Because of this coupling, any change in the input of APL results in an equivalent change in the output of RPL and vice-versa.This induces oscillations, which further increases the system's settling time.7) Planned islanding: When the grid connection is lost, the inverter should be switched from GC to SA state.PLL on the contrary will continue to operate till the control system notices an islanding event.As a result, the control system's response to a change in PLL behavior as a result of the outage is crucial.Huge swings in the frequency and the power angle are likely to take place during this transition.8) Unplanned islanding: Due to this, huge swings in frequency and power angle will occur.Also, transient power will flow from higher power set inverters to low power set inverters, potentially raising the inverters' dc link voltage and causing them to shut down [20].9) The transition from SA to GC mode: An improper synchronization could lead to a huge inrush of power from the grid to converters endangering the life of converters.IEEE std 1547-2003 prescribes ±10% in amplitude difference, 0.3 Hz in frequency difference, and 20°in phase difference as key ranges for synchronization [21].10) Black-start: Following a large outage, due to a lack of information about external-gird, the starting phaseangle of the input signal is uncertain to the PLL.Since the PLL links phase as well as frequency variables, this abrupt phase change is translated into major frequency transient.11) Large load changes: Due to its considerable inertia, a conventional synchronous generator retains a large quantity of kinetic energy.Under system uncertainties, this kinetic energy can adjust for the generator's mechanical and electrical power imbalance.However, with the increased penetration of DERs, there is a significant fall in the system inertia value [22].In view of all the above-said challenges and implications, various approaches and methods are proposed in the literature to address the issues.In this article, all those are summarized and through the general structure of the microgrid that is shown in Fig. 1, possible approaches aiming for transient performance improvement are shown.A comprehensive presentation of the identified approaches and methods is provided in Fig. 5.In this, four key approaches, namely: 1) grid synchronization control; 2) LCL filter tuning; 3) classical converter control; and 4) advanced converter control followed by the next level of approaches and methods are depicted.

III. APPROACH-1: GRID SYNCHRONIZATION CONTROL
Various approaches assisting distributed generation (DG) to stay in synchronism with the utility grid are presented in this section.It discusses the application of PLL, FLL, virtual impedance, islanding detection methods, and static transfer switches for achieving effective grid synchronization.

A. PHASE-LOCKED LOOP (PLL)
Estimating the grid phase is critical in the GC mode for processing the right amount of power.A general scheme in the application of the grid synchronization unit for the control of a 3-phase inverter operating in GC mode is shown in Fig. 6.In the literature, the technique of using PLL for grid synchronization is the chosen option.The PLL's estimated phase is utilized to synthesize the injected current's reference.The quality of power supplied to the utility system is harmed by errors in the estimated phase [23].
Working of SRF-PLL is very much sufficient for all purposes, i.e., fast and robust identification of phase and frequency under balanced conditions.However, in imbalanced and distorted grid conditions, this PLL has poor disturbance rejection.To fix this issue, the bandwidth of the SRF-PLL should be lowered at the sacrifice of transient response (i.e., slow detection speed), which cannot be affordable in most cases.This is the problem.Thus, many inventions took place in the development of new PLLs, which meet the following two necessary conditions.
1) Behaves the same as SRF under balanced conditions.
2) Under unbalanced as well as distorted grid conditions, the new PLL delivers effective disturbance rejection without sacrificing bandwidth or transient response.Accordingly, every developed PLL can be rated based on to what extent they meet the above two conditions.After justifying the above two conditions, a choice can be made among different PLLs, based on other indices such as the computational burden, design complexity, etc. Dynamic testing could be used to verify the proposed PLL's ability to track various forms of frequency fluctuations.These tests include the following: 1) magnitude and phase-angle jumps; 2) oscillatory variations of amplitude; 3) step variations of frequency; 4) linear changes in frequency; 5) changes in oscillations of frequency; 6) frequency tracking in a weak microgrid [24].A typical PLL system can be seen as a combination of a voltage-controlled oscillator, a loop filter, and a phase detector.The structure of a typical PLL and its small-signal model are shown in Figs.7 and 8.
Where u + αβ , u − αβ are positive and negative sequence components of u αβ , and û+ d,1 , û+ q,1 are the estimated dq components of the positive sequence component of PCC fundamental voltage.
When the input signal has some dc component (induced because of sensors and signal processing), the loop suffers from an error at the fundamental frequency.Mitigation of such a low-frequency error using in-loop LPF is undesirable because it extremely reduces the system bandwidth and degrades the transient response [25].Hence, prefiltering techniques started to compete with in-loop filtering techniques, thus avoiding the need to reduce bandwidth.Fundamentally, prefiltering techniques have the advantage of allowing the PLL to recognize sequences over in-loop filtering techniques.Prefiltering techniques are most commonly thought of as a group of two or more adaptive filters that work together to extract a certain sequential component from three-phase input signals [26].This prefilter is also an aspect of the phase detector.Based on this fact, it can be stated that in general, the implementation of a phase detector distinguishes the majority of PLL structures known from the literature.Table 1 shows a classification of the most popular PLL models based on the reference frame in which their prefiltering techniques were developed [26].
A comparison of different PLLs is provided in Table 2 based on how the transient performance was tried to be improved from the PLL's point of view.PLL development has been thoroughly researched and compared.
The estimated phase of PLL is used to create the reference for the injected current.However, if the respective influences of the current control loop and the PLL on one another are not  considered, the study of PLL would be incomplete.A detailed analysis of the performance of PLL including the interaction of the current control loop based on the development of a non-linear model for SRF-PLL is presented [27].The effect of grid impedance changes on the PLL's transient performance is discussed as follows.

1) INFLUENCE OF GRID IMPEDANCE VARIATIONS ON THE PERFORMANCE OF PLL
Based on grid-impedance value, the network can be categorized as a stiff grid and a weak grid as follows.
Under stiff-grid conditions, the grid synchronization loop adjusts the error signal (θ g − θ ) automatically to fix the disturbance d and secure u q = 0.However, when the disturbance d exceeds the maximum output of the phase detector, there is no steady-state equilibrium point to ensure u q = 0, causing

TABLE 1 PLL Model Groups Based on Prefiltering Techniques
PLL instability.From this, the requirements for large-signal stability can be defined.To put it another way, PLL operation is more stable if the grid-voltage level is higher, the current injected is lower, the grid input impedance is smaller, and the reactive component is smaller.
Under weak-grid conditions, the PLL instability can be either due to the increased deployment of DG units or due to the islanding of DG units.The first case falls under the category of GC mode.In this case, the grid impedance will be naturally high.Along with this, a higher injection current or a larger reactive component would result in a bigger disturbance (d) beyond the permissible limit, leading to instability of PLL.In the latter case, i.e., when islanding or outage occurs, the phase detector of the PLL will turn to absent leaving only the loop filter and voltage-controlled oscillator.At this stage, the output of the PLL should exhibit a step change in its output frequency value.This step change in frequency will be used by the control system to identify the islanding event and thereby shift the inverter operation mode to switch from GC mode to SA mode.The type of the load (inductive or capacitive) in islanded mode, determines the disturbance value.Based on this, the integrator in the loop filter drives the PLL output frequency to grow or fall.A special case is when the PLL exhibits no change in its output frequency value when an outage occurs.It is treated as the worst case for islanding detection.It happens when the load is of resonant type and its resonant frequency matches with the output frequency of the PLL at the time of the outage.

B. FREQUENCY-LOCKED LOOP (FLL)
There are two types of closed-loop three-phase grid synchronization techniques namely PLLs and FLLs [53].One advantage of FLL over PLL is that it is extremely resilient in the face of transient occurrences since grid frequency is far more stable than phase-angle [54].Moreover, PLLs are designed in SRF, while the design of FLL is carried out in a stationary reference frame.Because of the ease of designing in the SRF, improvements in PLL have been fully explored, leaving almost no room for further enhancement.However, since very few developments have taken place with FLL due to the difficulties of designing in the stationary reference frame [6], there is still potential for developing new FLLs with better performance.In view of these reasons, PLL and FLL are discussed separately in this work.
An alternative synchronization technique based on singlephase EPLL that provides independent frequency-adaptive synchronization without the use of synchronous reference frames is provided in [55].An EPLL can be treated as a PLL system providing a quadrature operator to an adaptive band pass filter.The fundamental requirement lies in the generation of direct and quadrature components of input grid voltage.However, the major disadvantages of EPLL, which later were addressed by SOGI-based FLL, are given as follows.
1) EPLL must be synched with the grid's sinusoidal reference signal, making it susceptible to phase changes in the input signal.
2) The assumption that α-β components are in quadrature may not always be true and is dependent on the grid frequency detection of the PLL. 3) When the dc term is present, the loop suffers from a low-frequency ripple, which also affects the amplitude estimation loop.For tracking the input signal, an SOGI-based quadratic signal generator (SOGI-QSG) was proposed in the literature, with the following characteristics and transfer functions: where (1) provides band-pass characteristics to ûα and (2) provides low-pass filter behavior to ûβ outputs.This is useful for reducing harmonics in the input u.A frequency adaptive loop (FAL) was used for the first time in [54] to adjust D-SOGI for positive and negative sequence component detection.The structure of a typical SOGI-QSG-based FLL is shown in Fig. 9.The resulting subsystem is simpler than a traditional PLL because neither any trigonometric functions nor phase-angle is employed to estimate frequency.To produce the estimated center grid frequency, an integrator with gain γ processes the product of − ûβ and (u − ûα ).The estimated frequency is calculated using (3) [56].A comparison of different FLLs based on improvements to prefilter, in-loop filter, and FAL is provided in Table 3 where γ = (k ω/u 2 max ) , in which is a positive constant, and u 2 max is the amplitude of the output voltage given by u

C. VIRTUAL IMPEDANCE
The voltage at PCC can be processed via a prefilter that identifies the fundamental positive sequence part to derive the phase angle properly under disrupted and imbalanced grid situations.A positive-sequence filter based on a generalized integrator [32], an adaptive filter based on the SOGI [33], the complex coefficient [34], and the delayed signal cancellation [47], [48], [49], [50], [51], [52] are examples of common prefilter.Besides, an in-loop filter-based revised phase-tracking method [43] was employed to improve SRF-PLL phase-tracking performance against grid voltage anomalies.Many filtered SRF-PLLs for GC converters had achieved improved estimation of frequency as well as phase under various grid-voltage anomalies.However, many of the PLLs of those schemes are asymmetrical [64], which probably causes frequency-coupled oscillations in weak microgrids [65].These oscillations increase the settling time, thus deteriorating the transient response.A subsynchronous oscillation will be noticed due to the PLL's frequency coupling effect.
For improved stability, a single input single output impedance shaping approach can be applied.The idea of this approach is to add virtual positive resistance that neutralizes the PLL's negative resistance [66].

D. ISLANDING DETECTION
Though the intention is to improve transient response, the case of transients due to faults and thereby UCB opening should be handled separately from the case of nonfault transients such as heavy load switching, capacitor switching, etc., for the sake of clarity.The former will be studied under this heading.
The main cause of unintended islanding is the presence of a problem in the upstream network, which causes UCB to open.The islanding operation of microgrids brings instability of frequency and voltage, synchronization difficulties, and power quality problems.Thus, the event of islanding should be detected accurately at the earliest [67].Before islanding, the power flow from DER to PCC can be expressed as P DG + jQ DG, and the power flow from the PCC to the local load as P L + jQ L [68].In such a case, the utility grid may be supplying or absorbing power, which can be described mathematically as follows: where u max and u min are the allowable highest and lowest voltages respectively, f max and f min are the allowable highest and lowest frequencies, respectively, u is the rated voltage, P is the rated active power, and Q f is the quality factor.

TABLE 3 Comparison of Different FLLs
When the condition P = Q = 0 has occurred, then a power mismatch is said to have occurred.This situation arises when the active power provided by DER matches the load's consumption or when the load-displacement power factor at resonant frequency converges to unity (i.e., ω 0 = ω res ).Under such conditions, due to a lack of significant voltage and frequency fluctuations during the islanding event, the voltage or frequency relays are unable to detect islanding, establishing a nondetection zone (NDZ).A grid-feeding inverter's NDZ characteristics can be mathematically stated as ( 6) and (7).
IDMs are divided into two types: local and remote.There are three types of local IDMs: passive, active, and hybrid.Along with NDZ, Li et al. [5] have made a mention of other performance indices namely error detection ratio, detection time, power quality, etc., based on which a comprehensive qualitative review of various IDMs is presented.

1) ACTIVE ISLANDING DETECTION (AID)
By introducing a small perturbance into the system at the inverter side, the active islanding detection (AID) approaches, identify the islanding.When compared with passive islanding detection (PID) approaches, these methods have a low NDZ.Injecting disturbance into the system is not an issue during grid connection but they pose serious power quality issues during the islanding mode of operation.Many improved versions of AID techniques were reported in the literature namely sandia frequency shift, impedance measurement, sliding-mode frequency shift, active frequency drift, and frequency jump.

2) PASSIVE ISLANDING DETECTION
To detect islanding, PID approaches will employ local metrics such as voltage, frequency, phase-angle, current, active power, reactive power, total harmonic distortion (THD), and so on at PCC.There can be no further deviations in these passive parameters when the DG is linked to the grid, but when the system is islanded, the changes in these values are beyond the norms and are used to detect islanding.To distinguish islanding from nonislanding events like switching transients and short-circuit faults, caution should be exercised while establishing the threshold values.The most preliminary passive islanding detection strategies depend on over/under frequency and over/under voltage.Frequency and voltage would be modified due to the disparity of real and reactive power at the onset of islanding.Islanding detection is accomplished by measuring the rate of change of frequency and voltage.Improved PIDs based on voltage and current harmonics distortion, phase jump identification, unbalance in voltage, rate of change of output power, and rate of change of frequency have been reported in the literature.Though passive approaches are easy to implement and have no power quality issues, they suffer greatly from large NDZ.

3) MODIFIED PASSIVE ISLANDING DETECTION TECHNIQUES
Signal processing techniques are used in the modified passive IDMs to improve detection performance, minimize detection time, and shorten NDZ.In Fig. 10, the process flowchart of this kind of IDMs is shown [69].Researchers were able to improve existing islanding detection strategies and establish new ways by using well-known frequency domain, time domain, and frequency-time domain signal processing methods such as the Fourier transform, wavelet transform, S-Transform, Kalman filter, etc.These technologies aid in the analysis and extraction of key elements from a measured signal, allowing for more effective power system operations.
A detailed review of these methods based on signal processing is presented in [70].These methods offer no power quality issues, low NDZ, low computational burden, and low detection time.However, the major issue in these lies in fixing the threshold value [71].Further intelligence-based classifiers under this category decision tree [72], artificial neural network [73], support vector machine [74], IoT and machine learning [75], etc., offer very low detection time but at the expense of a heavy computational effort.

4) HYBRID ISLANDING DETECTION (HID) METHODS
The HID procedures incorporate the benefits of both PID and AID methods.Large NDZ bothers PID methods, while perturbance bothers AID methods.As a result, HID approaches were created to address these issues.PID techniques suspect islanding, while AID techniques confirm nonislanding and islanding events.A hybrid automatic transfer switch that recognizes the microgrid's operation mode by inserting a suitable, variable impedance at the low voltage side of the grid is developed [76].A novel approach that involves the combination of voltage unbalance and voltage phase-angle to detect islanding in a variety of contexts is presented in [77].

5) REMOTE ISLANDING DETECTION
Classification of the remote techniques such as state monitoring, switch monitoring, and intertripping are done in [78].Communication links between the utility grid and DG sources are the basis for these strategies.These techniques are effective and have a low NDZ.Although remote IDMs are more reliable than local IDMs, they are more expensive.

E. STATE TRANSFER SWITCH
Utility interactive inverters ought to be able to supply electricity to local loads as well as the power grid in a highly adaptive grid.The grid can be unplugged under unusual circumstances, and the inverters should power the local loads.During this transition, there may be substantial transients in the frequency and magnitude of the voltage across the local dynamic loads.As a result, the inverter control system must include a mechanism that allows for smooth transfer among GC and SA states.Grid-connected inverters must be able to ensure power delivery while affecting the least amount of disruption to their local loads.An adjustable STS has been used in the majority of documented work on seamless transitions, with two pairs of voltage and current sensors used on opposite sides of the STS as shown in Fig. 11.
However, there are various ways that only require one set of sensors and do not require STS.A novel technique for smooth transfer that does not involve STS is presented in [79].The need for STS and its recommendation is found in [80].In the same work, a classification of the existing solutions from the literature related to mode transition based on employing the STS is presented in a novel way.

IV. APPROACH-2: LCL-FILTER TUNING
By replacing the L-filter with a properly designed LCL filter, the need for a large inductor can be overcome.However, the addition of a capacitor will create a new resonant peak and this peak if not addressed properly, may turn the loop unstable.Since the inductor resistance of the LCL filter does not provide sufficient damping, additional damping is required.Moreover, the region in which this resonant peak occurs determines the necessity of active damping [19].Since the line inductance is mostly unknown, the f res will also be mostly unknown which is a serious limitation.The two major approaches covered in this section are passive filter design and active damping for disturbance rejection.

A. PASSIVE FILTER DESIGN
An L-filter is used as a link between the converter and the grid to keep THD to a minimum of 5% and to limit the magnitude of various harmonics of current injected into the grid.The needed L-filter is large, bulky, inefficient, and costly.As a result, the goal of an LCL-filter design is to lower the analogous inductance L eq = L 1 + L 2 of a rather huge inductance value of simple L-filter inductance; thereby resulting in lesser losses.An LCL filter acting as an interlink between the inverter and grid is shown in Fig. 12.
The design of the LCL-filter should help the controller to address various disturbances such as grid voltage/frequency variation due to nonlinear or unbalanced loads, grid impedance variations, mode switching, etc., thereby contributing to the transient performance improvement.

1) FILTER DESIGN IN DIRECT CONTROL
In direct control, an LCL filter can alternatively be thought of as a CL filter that is added to an L filter.An L-type filter regulates the current injected into the distribution grid, and a CL-type filter attenuates the injection current ripple caused by inverter switching.When compared to the corresponding 20 dB/decade in the case of an L-filter, an LCL-filter has stronger attenuation and lower THD for frequencies around f sw of VSI.As a result, f sw is chosen to be higher than f res [81].
The control diagram of the conventional current-loop with grid-current-feedback (GCF) or inverter-current-feedback (ICF) is shown in Fig. 13.Where G c (s) is meant to attenuate steady-state tracking errors at the fundamental frequency.Also, the current controller is required to attenuate lower order harmonics and the LCL filter is intended to lower higher order harmonics.This critical frequency (f c ) which makes a fine demarcation between the two is found to be f c = f sw /6 [19].Accordingly, low-frequency and high-frequency regions are defined [82].In [83], it was proved that in the case of single loop GCF, if the value of the capacitor caused the resonant peak to fall after f c = f sw /6, then tuning the current controller's proportional gain constant to an appropriate value will be able to damp the resonance, thereby named it as the stable region.In the other case, if the resonant peak falls before f c , then the tuning of the current controller cannot suffice to dampen the resonance peak, so, this region is named unstable.
To bring stability in the unstable region, extra damping support either in the form of passive damping or active damping is required.The opposite happens in the case of single-loop ICF.As a result, to satisfy this requirement, the condition shown in ( 8) is established throughout the design of the LCL filter where f BW is the current controller's bandwidth frequency.
In addition, when using harmonic controllers to eliminate low-frequency dominant harmonics, harmonic compensators should be utilized up to a frequency that is lower than the system bandwidth.By any chance, if the frequency at which these harmonic controllers work is outside the bandwidth of the fundamental current controller, these harmonic controllers will fall into the LCL-filter dedicated region and can drive the system toward instability.This is referred to as "low-frequency stability" to distinguish it from the possible instability associated with the LCL-filter resonance.It may not be possible to raise the proportional gain in this situation since the LCL-filter poles may cause the system to become unstable.However, it has been shown in [82] that active damping can be used to improve the current controller's proportional gain without initiating any high-frequency resonance, allowing resonant controllers to be used even when the grid is very weak.One other way of avoiding overlap is by pushing the corner frequency, but it requires a redesign of the LCL filter.So, this option is not viable.Sometimes reducing the bandwidth of the primary current controller could be an option, but reducing the bandwidth reduces the dynamic response.Another way of controlling the bandwidth is by controlling the sampling frequency.In addition to the above restraints, the design of the LCL filter is very much conditioned by various design considerations such as the ratio of L 2 /L 1 and its impact on grid impedance variations, the ratio of f sw /f res , the ratio (f sw /f res ) and its significance in different damping regions, ratio (f sw /f res ) and its significance on different damping schemes and the connection between attenuation factors [81].The design of the LCL filter is complicated due to the constraints mentioned above.Fig. 14 presents a process flowchart for designing an LCL filter [84].Together with this, a comparative review of various aspects of inductors as presented in [7] will be very helpful for the physical realization of the LCL filters.To summarize, the LCL filter parameter selection is an iterative process that continues until all criteria are met.The finalized parameters of the LCL filter are calculated after completing the preceding steps.

2) ROBUST FILTER DESIGN AGAINST GRID IMPEDANCE VARIATIONS
The primary issue with the stability of the LCL filter would be its f res , which is not known precisely due to unidentified L 2 .While the real grid inductance is L 2 = L 2 + L g .The p.u. change of f res associated with the p.u. change of the LCL filter's inductance is influenced by the ratio of L 2 to L 1 .The system is fundamentally more resistant to grid inductance changes if the value of this ratio is greater than 1 [85].Generally, feedforward loops are used to compensate for grid voltage variations.Robust zones of LCL-filter that are resistant to grid impedance variation are presented in [17].Accordingly, ω res < ω sw /3 & ω sw /6< ω res < ω sw /4 is the zone for the single-loop GCF current control-based scheme and ω res < ω sw /6 is the zone for single-loop ICF current control scheme.Where ω res is the angular resonant frequency and ω sw is the angular switching frequency.In the same paper, an interesting finding was presented that under grid impedance variation, feedforward control can assist in maintaining a high bandwidth in current control.If numerous resonant controllers are used for harmonic compensation, this functionality is quite useful.Furthermore, many active damping-based approaches capable of compensating for grid impedance variations were discussed in previous approaches.

3) FILTER DESIGN IN INDIRECT CONTROL
In the direct current regulation method, the transient difficulty is expected during state switching from current to voltage control mode, i.e., when the transition occurs from GC to SA mode.However, in indirect control, the phase-angle and amplitude of capacitor voltage are controlled (voltage control mode) in both GC and SA modes, thereby a continuous voltage during total transition is ensured [86], [87], [88], [89].An outer current-control loop and an inner voltage-control loop are employed in this design to regulate the required amount of injected current and instantaneous capacitor voltage, respectively.Both control loops in GC mode are used to regulate the magnitude as well as phase-angle of capacitor voltage, ensuring that the right magnitude of the current is injected into the grid in phase with utility voltage.Whereas, in SA mode, only the voltage loop is engaged.
In direct current regulation, while designing an LCL filter, the inverter-side inductor is normally chosen initially to reduce the ripple current, and the CL section is added to fulfill the harmonic limitations.However, in this methodology, the LCL filter is structured in such a way that the LC filter is selected first to have desirable capacitor voltage ripple mitigation at the switching frequency, and a grid-side inductor is introduced to satisfy the harmonic requirement in indirect control.Keeping this basic distinction in mind, the design of the LCL filter has been presented in [87].However, the basic constraints, namely 1) to reduce the voltage drop across the inductors, the total inductance must be less than 0.1 p.u.; 2) the filter capacitance is constrained by the power factor; 3) f res must be between half of the f sw and f BW are maintained same as with direct control based LCL-filter design.
Table 4 presents a summary of various LCL-filter design techniques to mitigate the corresponding disturbances as proposed in the literature.

B. DISTURBANCE REJECTION AND ACTIVE DAMPING
Distorted grid voltage and changes in grid impedance are the causes of extrinsic effects.These effects cause a change in f res in the systems associated with LCL filters and hence require additional damping.The active damping techniques are popularly used for this purpose.The general classification of these techniques is shown in Fig. 15.
Filter-based active damping approaches as shown in Fig. 16 employ a controller of high order to dampen high-frequency transients as well as regulate low-frequency transients.The filter component of the controller is in charge of actively damping the highly unstable high-frequency dynamics.The filter is created as an analog filter, such as a notch filter, and thereafter transformed from analog to discrete, a process known as bilinear transformation [90].The comparison of transient performance improvement through the rejection of extrinsic disturbances using single and double-loop active damping approaches is carried out in Table 5.
Correspondingly, Fig. 17 shows the positioning of equivalent virtual-impedances emulated through the application of such state feedback-based active damping approaches [7].In the case of multiloop-based active damping techniques, additional system state variables that are measured/estimated are controlled to ensure stability.To gain a deeper insight into the evolution of control methods based on multiloop feedback, paper [8] can be referred to.

V. APPROACH-3: CLASSICAL CONVERTER CONTROL
This section discusses the traditional methods of controlling the converter to make it behave as a controlled current or voltage source by using single/multiple controllers.The structure with cascaded power and the current controller are investigated under current control.In voltage control, two structures are investigated, namely cascaded voltagecurrent (V-A) controllers and cascaded droop-voltage-current (D-V-A) controllers, as given follows.

A. CASCADED POWER-CURRENT CONTROLLERS (CURRENT CONTROL)
When microgrid inverters are connected to the grid, they function as current sources, and when they work independently, they function as voltage sources.The majority of microgrid inverters now use a VSI architecture.A current controller regulates the current delivered to the grid.Current-controlled inverters offer high-accuracy instantaneous current control, protection from peak current, overload rejection, and excellent dynamics.The quality of the employed current-control technique determines the VSI's performance.Inverters in microgrids must have excellent harmonic rejection capability to satisfy power quality criteria [110].Although there are various potential techniques for controlling GCIs, one common strategy is to attempt to drive a sinusoidal current into the grid independent of the grid voltage waveform.The block diagram of such a current control approach is shown in Fig. 18.
The current control's major objectives are reduced steadystate error, improved dynamic performance, and reduced harmonics in injected grid current [111].In this context, the following guidelines from the frequency domain help to develop a robust closed-loop system with good transient performance.1) The optimum gain crossover frequency is dependent on a tradeoff between the speed of response and switching harmonics rejection capacity.
2) The optimum phase margin is dependent on the degree of resonance damping and stable response.
3) The robustness of the closed-loop system against gridimpedance changes is dependent on the degree of flatness in its phase profile around the gain crossover frequency.4) A general 3 dB gain margin and 15°phase margin ensure robust stability and good transient response [104].PI controllers for current controllers are not capable of compensating for all poles of LCL-filter-based VSIs.Through active damping strategies, the effects of the uncompensated poles of the system are attenuated [101].These active damping methods shift the closed-loop poles to stable regions thereby providing a stable system [112].Therefore, active damping techniques help to improve transient stability.However, they do not improve the transient response in terms of the controller's speed, overshoot, and effort [113].This provides the necessity for controllers with complex control structures.Furthermore, a review of the current controllers was provided in [114].Besides these, a few other methods were also developed.These developments are presented in Table 6.

1) INFLUENCE OF DISTURBANCES ON STABILITY
Both the low-frequency and high-frequency ranges must be explored for the current loop's stability.Stability analysis is primarily concerned with the use of the current controller in a low-frequency range (frequencies up to the current controller's bandwidth).The damping of the LCL filter is connected to stability in the high-frequency region (frequencies higher than the current controller's bandwidth).Furthermore, both intrinsic and external influences can cause low-order harmonic distortion of grid current.The discussion on transient performance improvement through the rejection of extrinsic disturbances was carried out in Table 5.While the sources of intrinsic disturbances are given as follows.
1) Computational and PWM delays: Due to time delays, damping performance will behave like a negative resistor, which can easily induce instability [19].2) Overmodulation problems: When overlarge values of feed-forward gains are used in the process of mitigating harmonic effects through emulation of virtual impedance, will cause over-modulation problems [121].3) Inadequate positioning of f res within a suitable range: This results in a compromise between harmonic compensation and optimal resonance damping [122].4) Coupling between the active and reactive power of DG. 5) Measurement inaccuracies.6) Limited choice of loop gain: Increased time delay reduces bandwidth and, therefore, worsens the dynamics As a result, active damping isn't always beneficial, and it might harm stability if the gain isn't set carefully [83].However, large control gains are often necessary for good steady-state and transient responses.However, they decrease system stability.The system will have a sluggish reaction and high steady-state error if smaller gains are applied for a sufficient stability margin [91].
Hence, there exists a limited choice to fix the loop gain.

2) EFFECT OF TIME DELAYS ON DAMPING
Disturbances due to computational and PWM delays are considered crucial and received attention in the literature.A comparison of different methods to compensate for the errors that occur due to time delays (rejection of intrinsic disturbances) is shown in Table 7.If adequate damping is not used, then the closed-loop current regulation of the LCL-GC converter is not stable.The LCL-GC converter puts two resonant poles (at f res ) into the current control loop, which causes this.Adding a damper resistance (R D ) in series with a filter capacitor is the simplest approach to address this problem.It is worth noting that when R D rises, the system comes closer to the stability region.Measuring the capacitor current, multiplied by a constant K 1 , and deducting the resultant from the PI controller output produces active damping.The control diagram of the current control approach with capacitor current-based active damping while neglecting time delays is shown in Fig. 19.The constant K 1 works like a real damper resistance in this situation, with no additional power losses or burden.
To attain identical poles for the system with virtual resistorbased active damping and the system with a real damping resistor R D , ben Said-Romdhane et al. [123] have derived the  value of K 1 as (9).Now this will be compared with the case when the time delay is considered.One-sample computation delay as considered can be expressed as (10).The PWMbased reference is retained and compared with the triangle carrier to establish the duty cycle once it has been updated.A zero-order-hold can be used to model this behavior, which is expressed as (11).This means that a one-half-sampling period is introduced as a PWM delay.
Identification of the capacitor-current-feedback active damping as equivalent to virtual-impedance Z eq connected in parallel with filter capacitor is found in [124].The control diagram of such an approach considering the time delays is shown in Fig. 20, and Z eq is expressed as (12), where R D = L 1 /(CK 1 K PWM ) is the virtual damper resistance, equivalent to capacitor-current feedback active damping without any delay.Substituting s = jω yields ( 13) Z eq (ω) = R eq (ω)// jX eq (ω).
It means Z eq can be represented like a resistor R eq and a reactor X eq connected in parallel.The LCL filter's resonant peak is damped by component R eq , and the f res is changed by component X eq .As seen in ( 13), both R eq and X eq are frequency dependent.
The time delays transform the virtual resistance created by the proportional controller into a virtual impedance.For frequencies greater than one-sixth of the sampling frequency, this virtual impedance has a negative real portion.For frequencies greater than one-third of the sampling frequency, the imaginary part of this virtual impedance is negative, changing the actual LCL's f res [11].

B. CASCADED VOLTAGE-CURRENT CONTROLLERS (VOLTAGE CONTROL)
The block diagram of a general structure of cascaded voltage and current controllers is shown in Fig. 21 in which the current controller controls the inner loop, and the voltage controller controls the outer loop.The voltage controller is required to provide fast regulation of output voltage as well as improved stability.In SA mode, since the current controller has no role, it can be temporarily disconnected.However, it is verified in the literature that makes the current controller participate with the voltage controller in SA mode reduces instability.

1) SEAMLESS TRANSITION
The microgrid is said to operate in either GC or SA mode.In GC mode, the VSI is required to operate to control the current through a current controller.While in SA mode, the objective  is to control the voltage through a voltage controller.A transition between the modes involves switching between the two controllers.The control structures based on the deployment of STS for seamless transition can be broadly classified into two structures, namely, dual control structure algorithm and single control structure algorithm (unified control) [80].
In the first category, these DGs work in current-control mode when connected to the grid.These DGs cannot modify their output power in response to changes in grid frequency in GC operation, i.e., they are not designed to function as dispatchable DGs.They switch to voltage control mode in the absence of the grid and provide power to the linked loads.During GC operation, PLLs are used to achieve grid synchronization.DGs, on the contrary, create their reference for SA.Therefore, a dual control structure as shown in Fig. 22 is appropriate for this type of DG.
In the second category, these DGs operate with droop control in both GC and SA modes.They boost their active output power and get stabilized at a new steady-state operating point if grid frequency drops below the nominal value.In the same way, if grid frequency exceeds the nominal value, active power output is reduced, i.e., they are designed to function as dispatchable DGs.So, a single control structure as shown in Fig. 23 is appropriate for this type of DGs [80].

C. CASCADED DROOP-VOLTAGE-CURRENT CONTROLLERS (VOLTAGE CONTROL)
In this section, the focus will be on the application of droop controllers together with voltage and current controllers.This control scheme is shown in Fig. 24.Note that, in this scheme, the inverter is always regulated to operate as a voltage source in both the SA and the GC modes.

1) DYNAMIC RESPONSE OF DROOP CONTROL
The small-signal model of APL in a conventional droop control is shown in Fig. 25.This model helps to study the inherent limitation of the droop control toward improving the transient response.The closed-loop control's time constant can only be modified by adjusting m, as shown in (14).m, on the other hand, affects the DER frequency.As a result, there is a fundamental tradeoff between the control system's time constant and frequency regulation.This feature limits the dynamic response of droop control seriously 2) SEAMLESS TRANSITION Based on the state of the inverter's operation, Wang et al. [128] have presented a generalized control algorithm for the generation of the voltage reference.This is shown through ( 15)-( 22).If the voltage reference generation routines are set to run in parallel throughout three operating modes, the voltage reference generation does not have to start over after the switch.a) Grid forming mode (SA): b) Grid feeding mode (GC): where K p0 is the proportionality constant; K p,m and K i,m are the proportional and integral gain coefficients of P-δ droop; K p,n is the proportionality constant of Q-v droop.In Table 8, a comparison of different droop-based multiloop approaches for a seamless transition between different modes is presented.

3) VIRTUAL IMPEDANCE
The grid current is sent back to compute the voltage drop due to virtual impedance, simulating the impact of an impedance.The final reference for inverter voltage is obtained by subtracting this drop from the reference voltage (produced by the power loops).In SRF, the equations representing virtual impedance voltage drops are shown as follows: Table 9 compares different droop control methods.From this it can be noticed that droop control together with virtual impedance 1) is not affected by system parameter changes; 2) can handle both linear and non-linear loads; 3) can enhance decoupling between real and reactive power control.However, voltage regulation is not guaranteed.This limitation can be compensated using adaptive droop.Because of its simplicity to implement and above said advantages, the combination of adaptive droop control with virtual impedance is a favorite.

4) HANDLING OF NONLINEAR LOADS
Although output filters enhance the quality of voltage produced by the inverter, they can degrade the waveform quality while feeding nonlinear loads.Low-order harmonic components emerge within the inverter output current due to nonlinear loads.When these components pass through the filter inductor, a voltage drop occurs, which may deform the voltage waveform at the inverter's output terminals.In the case of low-power converters, where the output impedance is higher, the issue becomes severe [18].
Based on paper [143], the reduced model of a single-DG grid system with a voltage control approach, with the DG unit depicted as a regulated voltage source v(s) along with a series output impedance z inv is shown in Fig. 26(a).The voltage source e g as well as a grid impedance z g are used to represent the grid.The PCC nonlinear load is illustrated as a harmonic current source with passive loads in the middle.If a PCC harmonic voltage (u h ) is used to influence its DG harmonic voltage (v h ), this condition is expressed as follows: I 1,h and Z inv,h are the harmonic-current and harmonicimpedance of the DG, respectively.The expressions for I 1,h are shown as (26).The equivalent impedance at harmonic frequencies (Z inv,h,eq ) as illustrated in Fig. 26(b) is expressed as (27).DG's harmonic impedance will be reduced by a factor of (1 + K) if the DG harmonic voltage is properly regulated with only a positive feedback gain of K.As a result, the DG side's harmonic impedance can be significantly lower than the that of grid side.Thereby, the DG unit will absorb the majority of the nonlinear load current, resulting in better grid current as well as PCC voltage.Therefore, a greater K value reduces PCC voltage harmonics even more.If K equals 0, the system  becomes a typical voltage-controlled DG unit with no active compensation.In literature, many methods were formulated to mitigate the propagation of harmonics.The basic idea is to make the DG behave as an active power filter thereby absorbing the harmonics remaining the same.The following is a review of different methods provided in the literature, which include damping of harmonic propagation caused by grid voltage distortion at PCC as well as local nonlinear loads.a) Virtual resistance: A popular way of making the DG operate as a resistive-active power filter (RAPF); thereby only at harmonic frequencies the DG will behave like a virtual resistance [144].The two significant constraints of this technique are characteristic impedance getting affected by grid-side inductance and the need to extract line harmonic currents at all desired harmonic orders.b) Virtual inductance: An alternative way of shaping equivalent harmonic-impedance of DG unit is presented in [143].
An equivalent harmonic impedance is emulated as inductive as long as the physical feeder impedance is inductive, thereby providing better stability performance.c) Virtual impedance: In [145], a virtual impedance control scheme is presented by using a complex gain.In this, the need for fundamental and harmonic extraction is avoided leading to a major advantage.In [146], two separate harmonicimpedance loops were proposed one for fundamental and the other for harmonics.Positive resistance is used to achieve autonomous sharing of harmonic currents and harmonic resonance damping, while negative inductance is used to partially counterbalance the effect of grid-side inductance (one significant restriction of the RAPF system).Additionally, when harmonic resonance is present, the virtual harmonic impedance (VHI) loop can transfer the resonant spots to a higher frequency range, where the harmonic resonance can be more easily damped.In [147], a variable-VHI loop is developed in which the negative inductance emulated was below the effective inductance, which guaranteed the loop stability; and a larger positive impedance is used for better harmonic power sharing.It is established in [121] that higher values of emulated impedance cause overmodulation problems, and thus the impedance that can be realized should be limited.Therefore, the author has presented a new method where large values of virtual impedance can be realized without any overmodulation problems.In [148], a capacitive virtual-impedance loop is utilized to enhance the sharing of harmonic current and reduction of voltage-harmonics at PCC. d) Virtual admittance: Virtual impedance can be applied only to those DG units which operate in voltage control mode, i.e., SA mode; but cannot be applied to DG units that operate in current control mode, i.e., GC mode.This is verified in [149].To overcome this limitation, a virtual admittance loop as a substitute for virtual impedance is presented [148], [149], [150].

5) VIRTUAL IMPEDANCE FOR DECOUPLING OF P AND Q DROOP CONTROLLERS
Droop control is primarily introduced to ensure proper power sharing between the parallel DGs.As a result of implementing conventional P-ω and Q-v droop methods on a low-voltage microgrid, substantial coupling between active and reactive flows will be introduced, especially during transient conditions.Oscillations and lower transient stability will be caused as a result of this.To make up for this, an outer loop that emulates virtual impedance is inserted to dampen the oscillations of the system.However, although the system is properly damped by virtual impedance, the transient response is impaired, i.e., the outer virtual impedance loop improves transient stability but deteriorates transient response [151].

VI. APPROACH-4: ADVANCED CONVERTER CONTROL (VOLTAGE CONTROL)
This section discusses cutting-edge methods for controlling the converter to make it behave as a voltage source by employing advanced controllers.These methods control the VSI in such a way that the DG behaves like a synchronous generator.

A. VIRTUAL INERTIA EMULATION
In situations where the VSI supplies a considerable amount of power in a local grid, system inertia decreases, and fluctuations in generation or demand may cause severe frequency deviations, which may result in system instability.When the converter power output is equal to or more than the intended power, then the frequency of PCC voltage will stabilize at its nominal level.If this is not the case, the frequency of PCC voltage is altered proportionally to the variation of the actual power from the intended power.Thus, the VSI should be capable of providing frequency regulation as a primary control to the local grid as a result of this.When there is a sudden jump in power demand, the same shall be translated as a step change in mechanical input power P 0 (represented in the swing curve) and P is electrical output-power in terms of synchronous generator case.Whereas in terms of rotational inertia emulated through VSIs, P and P 0 are the actual and scheduled active power outputs from PCC to the grid, respectively.The only difference between virtual inertia emulation through droop and VSG is how the angular frequency reference ω * of the converter output voltage is derived.A general control scheme for the emulation of virtual inertia in VSI by mimicking a synchronous generator is shown in Fig. 27.

1) PHASE LOCKED LOOP BASED VIRTUAL INERTIA
During transients, estimating the system frequency is challenging due to time delay issues with PLL measurements [152].In PLL systems, including EPLL, both the phase as well as the frequency variables are integrated and estimated.In a power system, phase-angle can vary suddenly owing to system problems.Due to the mass of rotating machinery, the frequency cannot change rapidly.A PLL, by linking phase as well as frequency variables, converts abrupt phase changes into significant frequency transients.This might happen when a PLL is first started and the starting phase-angle of an input signal is uncertain to the PLL [153].In this case, if a time delay is introduced in the process of translation, the problems with abrupt and sudden phase changes can be resolved.This way of introducing time delay is similar to how a conventional synchronous generator emulates inertia.If VSI is phase-locked to the ac voltage V AC , it would not alter its output power when the load step occurs [154].This inverter has no electrical inertia.There are various ways to create required electrical inertia.The solution to emulate electrical inertia is to temporarily decouple the PLL phase-angle from the ac grid voltage V AC .However, this kind of delay is required to happen under grid-supporting mode.But under grid feeding mode a strict coupling should exist.
There are two ways in which a PLL can improve transient response: 1) by providing a similar response as how a synchronous generator behaves in case of disturbances by emulating its electro-mechanical characteristics [155]; and 2) by presynchronization for a seamless transition between the microgrid modes.The second way was already discussed in the PLL approach.In the first approach, the power controller for a said converter that emulates inertia is part of PLL.The phase detector generates a signal corresponding to the phase difference δ = (φ s − φ) of frequencies ω g and ω.In a true synchronous generator, the relation between ω g the EMF frequency and ω r the terminal voltage frequency is expressed as (28).In the case of a VSG, ω is the converter side frequency and ω g is the grid side frequency.Fixing the values of K I and K D , the open-loop transfer function of the PLL expressed as (29) will appear the same as the transfer function of a synchronous generator shown in (30), where K I is the gain with pure integrator of the loop filter

2) DROOP-BASED VIRTUAL INERTIA
The design of droop-controlled techniques based on virtual inertia injection can be considered an important evolution.Inertia emulation from droop control based on the swing equation is realized in [156] and [157].Voltage is measured at PCC and is used as feedback to fix set points for droop controllers.
In case of voltage tracking error at the converter output due to a small unbalance, oscillations of instantaneous active power will result.When these power oscillations are conveyed to the droop control the resulting output will also be oscillating.Therefore, a proper LPF is required which blocks the oscillatory part and provides the average value instead of instantaneous values [158].Replacing LPF with a real-time integration method to enhance the controller's response time is done in [159].First-order LPF's time constant to serve an equivalent function to virtual inertia is established in [160].This LPF is connected with active power measuring.This inertia, however, may be insufficient for transient conditions and would need to be strengthened to enhance transient performance.
The control diagram of the power frequency droop control strategy is shown in Fig. 28.The open-loop transfer function of the same is expressed as (31).If the delay time is 0, (32) represents the conventional droop control strategy; where m is the droop coefficient.In this method, the delay time is typically too short to affect both electro-mechanical transients Droop control-based techniques for inertia emulation are broadly classified into two categories namely fixed-inertia and variable-inertia.Literature has made much emphasis on strengthening the frequency response during transients but did not lay much stress on improving voltage response.To improve the dynamic response of both voltage and frequency, a delayed-droop control approach was described in [161].This method incorporates a first-order lead-lag controller into the classic droop laws.The systems with high inertia exhibit good transient responses but poorly damped oscillations of active and reactive powers.Hence, fixed inertia injection schemes are never a favorite [160].
Under variable inertia injection type, it is reported in [16] that droop coefficient variation in droop control can potentially add to the system's overall inertia.In the same paper, droop gains were modified based on the df/dt, i.e., virtual inertia is added as a function of df/dt which is observed by the inverter during the transition to enhance the transient response of frequency.This controller is tested under unintentional islanding case also.In [162], optimal tuning of inertia constant together with the parameters of load frequency controller and frequency droop coefficient was proposed.Inertia emulation realized for droop control based on the swing equation in [157] also comes under the category of variable inertia injection.None of the above methods has spoken about the aspect of issues with voltage response.It was demonstrated in [163] that a modern method of adaptive virtual inertia control is achieved by utilizing the VSG control idea in conjunction with the state of charge of an energy storage unit.This approach can simulate adaptive inertia.However, the control logic and storage control functionalities are difficult and limit their use.
An adaptive transient droop function is provided in [164].But this induces coupling between APL and RPL.Numerous articles have proposed a variety of techniques meant for decoupling so that APL can be tuned to respond quickly.However, most of them concentrated only on frequency response.A novel technique that addresses the above issues and provides improvement of both frequency and voltage responses is provided in [165].Table 10 shows the evolution of various techniques associated with droop control in the process of inertia emulation.

3) ROTARY INVERTER-BASED VIRTUAL INERTIA
In [173], a dc motor-synchronous generator set with proper control is used in place of the power electronic converter to emulate sufficient inertia.However, under normal conditions, slow transient response an inherent feature of rotary machines can be a drawback to the motive of the microgrid.

4) VIRTUAL SYNCHRONOUS GENERATOR-BASED VIRTUAL INERTIA
Apart from the virtual inertia injection through droop control laws, a novel approach named VSG meant to operate the VSIbased microgrids to mimic the synchronous generator-based conventional power plant's operation was presented in [174].The control diagram of APL of VSG is shown in Fig. 29.The corresponding open-loop transfer function can be written as (33), which is a first-order nonlinear differential equation, and the solution of it will yield ω * J • sω * = P 0 − P − k d ω * − ω g (33) This VSG emulates the moment of inertia and damping characteristics virtually into the microgrid's operation for short time intervals.A review of the fundamental topologies of VSGs that laid the foundation for later developments in this concept is provided in [12].The names of these models and their associated principles based on which they were developed are showcased in Table 11.
In IEPE's topology, there exist two important variants namely VISMA and Synchronverter [175].In terms of inertia and damping properties, the VSG presents the dc source to the utility grid as a synchronous generator.In fact, the system emulates virtual inertia by adjusting the real power in inverse proportion to the speed of rotation via VSI.In short, the virtual mass compensates for grid frequency dips and virtual damper damps grid oscillation; thus, these features are just as effective as synchronous generators.Table 12 shows the chronological development in the VSG technology through an emphasis on the type of the VSG model, APL, RPL, etc.

5) ENERGY STORAGE SYSTEMS BASED VIRTUAL INERTIA
With a decrease in the inertia of the system, load/generation fluctuations can result in substantial frequency variations, potentially resulting in instability of the system.When the scheduled power equals the power output of the converter, the PCC voltage's frequency will stabilize at the nominal value.Alternatively, the frequency will vary proportionally based on the difference between the power output of the converter and scheduled power.By utilizing selective ESS to act as a synchronous generator, the frequency can be controlled by injecting or absorbing the difference power.
A power converter with an appropriate control technique can be deployed to make ESS operate as an inertial unit.For this purpose of controlling frequency, Renjit et al. [201] used a basic swing equation (33) as a control technique for this purpose.Among the available ESS, superconducting magnetic energy storage (SMES) has the quickest response time with a high-power density next to the supercapacitor [14].Hence, these kinds of ESS should be chosen to provide inertial support.The dynamic model of a typical microgrid connected with a conventional power plant, a variety of DERs, loads, and selected ESS for inertial support is shown in Fig. 30.
The dynamic equation of the system can be expressed as (34).Here, H is the inertia of the system, D is the damping of the system, P M is the change in power generated by the conventional power plant, P DER is the sum of the change in power generated by aggregated DERs, P L is the change in power demand by the connected loads, and P VI is the change in power from selected ESS units to compensate for frequency deviation f.The ESS control equation based on the swing equation can be expressed as follows: 34) where J ESS is the inertia constant of the selected ESS, and T ESS is the time constant meant for modeling the selected ESS.In [202], fuzzy logic is utilized to adjust the value of the J ESS based on the change in frequency f and change in power generated by DERs P DER .

B. POWER OSCILLATION DAMPING
Though the design of VSGs by swing equation aims to inject inertia and suppress grid oscillations simultaneously, there always exists a tradeoff between the two.Heavy inertia presupposes that the VSG's active power output is oscillating; furthermore, transient power sharing amongst DG units will be slow as a result of virtual inertia, particularly under weak microgrid conditions [203].To reduce the settling time, the need to study power oscillations damping to improve the transient response under a separate heading is clear.During GC mode, if the power reference changes, the VSG-based inverter will exhibit large overshoot and oscillations in the output active power which is unacceptable.For the same conditions, the droop-based inverter will follow changes with small overshoots and oscillations.During SA mode, the output frequency of VSG based-inverter will change smoothly with the change in load power or disturbance.For the same conditions, the output frequency changes of the droop-based inverter are sharp which is unacceptable [204].Freedom of choosing virtual inertia and damping coefficient in VSG is highly constrained by the coupling between APL and RPL.In Table 13, a comparison of different approaches meant for damping power oscillations is presented.Optimization of various VSG parameters is presented in [223], [224].

VII. COMPARATIVE ANALYSIS AND FUTURE RESEARCH DIRECTIONS A. COMPARATIVE ANALYSIS
Various disturbances (or technical challenges) that cause deterioration of transient performance in microgrids that are described in Section II are combined in Table 14 as (a) to (k).
As discussed through Sections III-VI, the high-level four approaches and their 19 key subapproaches (numbered I to XIX in Table 14) are compared based on to what extent these approaches can improve the transient performance of the system when subjected to the disturbances.The summary of these comparisons is presented in Table 14.Where "NA" stands for not applicable to a specific approach, and "NF" stands for not found in the literature.Furthermore, all these subapproaches are ranked based on their capability to address a particular technical challenge as presented in Fig. 31(a)-(k).The capability of a specific approach is assumed in a range of 0 to 4, where the value 0 represents the lowest capability and value 4 represents the highest capability in relative comparison with other approaches.
In any of the plots shown in Fig. 31, it can be understood that highly ranked approaches concerning a particular issue  are highlighted in orange color.It is re-emphasized that out of the 19 subapproaches, I-V fall under the grid synchronization approach, VI-VII fall under the LCL-filter tuning approach, VIII-XIII fall under the classical converter control approach, and XIV-XIX fall under the advanced converter control approach.The importance of this comparison can be understood using the following cases.

1) CASE-1: IN CASE OF A SINGLE ISSUE
For example, consider a situation where grid-impedance change is the key issue.Verifying from Fig. 31(b), sub approach III from grid synchronization, VII from LCL-filter tuning, VIII-X, XIII from classical converter control, and XIV-XVII, XIX from advanced converter control are the suitable choices.

2) CASE-2: IN CASE OF MULTIPLE ISSUES WITH A COMMON SOLUTION
For example, consider a situation where grid-impedance changes, black start, and large load changes are the issues to be addressed.The corresponding plots for these issues are Fig.31(b), (j), and (k), respectively.Based on the comparative analysis from these plots, it can be justified that STS (V) from the grid synchronization approach for handling black start issues, active damping (VII) from the LCL filter tuning approach for handling grid impedance issues, and ESS (XIV) from advanced converter control approach to compensate for large load changes are a suitable choice.Accordingly, a configuration made up of combining these three approaches can yield STS in synchronization unit, active damping and ESS with virtual inertia-based converter control is an effective solution for this situation.With similar kinds of discretions, situations with multiple issues can be handled effectively.

3) CASE-3: MULTIPLE ISSUES WITH NO COMMON SOLUTION
For example, consider a situation where coupling between P and Q controls, and planned islanding are the issues to be addressed.The corresponding plots for these issues are Fig.31(f) and (g), respectively.Based on the comparative analysis from these plots, it can be justified that improved current controllers (VIII) to compensate for P and Q controls coupling and ESS (XIV) to compensate for planned islanding are suitable choices.
Thus, most of the technical works focus on handling only one technical challenge, which is not true in practical situations where multiple issues exist in parallel.The number of combinations of these multiple issues is very large, thereby providing many research gaps.In addition, in the process of addressing those research gaps, this article will be of great help to the researchers in terms of the time required to formulate novel effective configurations in the process of solving the considered issues.
Furthermore, the comparisons that are given in Table 14 and Fig. 31 help in formulating the future research directions that are discussed in Section VII-B.

B. FUTURE RESEARCH DIRECTIONS
As discussed above, it is identified that power quality is the pivotal issue in the grid-connected mode of operation, and stability/reliability is the pivotal issue in the islanded mode of operation.Hence, to address these issues, here, this article proposes some feasible future research directions corresponding to the approaches discussed in this article.

1) RESEARCH DIRECTIONS TO APPROACH-1: GRID SYNCHRONIZATION CONTROL
As discussed in Section III, this approach consists of various subapproaches and methods based on the concepts of PLL, FLL, virtual impedance, islanding detection, and STS.The future research possibilities in this direction are suggested as follows.
1) Most of the researchers considered SRF-PLL in their works.The limitations of this PLL have already been indicated earlier.Modified PLLs can avoid the reduction in bandwidth for maintaining a better transient response.This provides a vast scope to verify the system's transient performance with the remaining models.An increase in structural complexity is the only limitation.
2) The most favorable feature of an FLL is that it is more robust to disturbances than PLLs.Thus, the selection of a suitable FLL model can play an effective role than the use of a PLL.However, the studies on FLL models used in the microgrid context are limited when compared to the PLLs.Therefore, it is suggested to investigate the improvement in transient performance by applying various FLL models for microgrid application.
3) From Table 14, it is identified that the virtual impedance concept under the grid synchronization approach addresses the issues with grid impedance changes and nonlinear loads.Especially, if it is required to compensate for transient performance issues related to coupling between P and Q controllers, the virtual impedance approach is the only possible solution.4) From Fig. 31(h), it is inferred that the Islanding detection (IV) and STS (V) approaches have an excellent potential for handling transient performance issues raised with unplanned islanding.5) Furthermore, incorporating an STS helps in excellent soft transition between different modes (planned/unplanned islanding, SA-GC, and black start).It can be verified from Fig. 31(j) that, while handling transitions involving black start, STS is a suitable choice.Hence, the cost of deploying an STS can be justified by its excellent transition performance, especially, where the black start is a key issue.
However, this grid synchronization approach can't improve the situations involving a change in f res .
Scope: The above mentioned proposed research directions with respect to grid synchronization control help to improve the power quality in all three modes (grid-connected, islanding, and transition) of microgrid operation.

2) RESEARCH DIRECTIONS TO APPROACH-2: LCL-FILTER TUNING
From Table 14, it is noticed that passive filter tuning has a moderate effect on transient performance improvement.Elimination of high-order harmonics can be handled only through a passive filter.Even though, the redesign of a filter cannot be affordable, but, a proper design in the early stages helps to handle the changes in f res .Active damping can be treated as one of the best approaches for improving transient response during most disturbance conditions.Through multiple combinations, the effectiveness of this approach is also very large.However, the two major limitations of this approach are i) deterioration in performance due to improper active damping and ii) inability to handle issues relating to mode transitions and stability.
Scope: The abovementioned proposed research directions with respect to LCL-filter tuning help to improve the power quality in grid-connected mode of microgrid operation.

3) RESEARCH DIRECTIONS TO APPROACH-3: CLASSICAL CONVERTER CONTROL
As discussed in the above sections, proper design of the classical D-V-A controller can effectively improve the transient performance of the microgrids.Some potential future research directions related to classical control designs are suggested as follows.
1) Even though these approaches merely improved current controllers, low-order harmonic compensation by active damping, and control loop deficiency compensation, this group of things can be seen as an extension of remaining structures.Since this approach allows the converter to work in GC mode only, handling the transients due to mode transitions is beyond the scope of this approach.2) Complex current controllers offer very good robustness against intrinsic disturbances.It is important to remember that the dynamic response of the system mostly depends on the bandwidth of the current controller.Most of the remaining approaches extend their support to the current controller.
3) It can be known from the comparison that mitigating the effects due to time delays can be handled only by control loop deficiency compensation; thereby proving itself is yet another potential approach to adopt in the system.4) The inability of the current control approach toward handling mode transition transients can be handled through voltage control of the converter.The first of such schemes is a V-A controller.Multiple controller structures and compensations can be tried with both voltage controller and current controller to find the best.Many researchers are still finding a large scope in this approach.5) In addition to what was said with the V-A controller, a droop controller together with individual voltage and current controllers offers more combinations to try.Virtual impedance can be considered the most significant reason for placing the D-V-A structure on top of the list.Multiple ways of emulating virtual inertia can be tried to find the best among many.Scope: The abovementioned proposed research directions with respect to classical converter control help to improve the power quality in both grid-connected mode and during transition.Also, helps to improve the stability and reliability in the grid-connected mode of microgrid operation.

4) RESEARCH DIRECTIONS TO APPROACH-4: ADVANCED CONVERTER CONTROL
The virtual inertia mechanism is identified as a promising alternative to the classical control of microgrids.For this, the possible future research work is suggested as follows.
1) Virtual inertia through ESS is the best approach in terms of transient stability.However, the cost of its application has to be justified by the necessity.2) As found from the existing literature, presently both virtual inertia schemes, namely droop & VSG schemes are competing with each other in terms of both transient response and stability margins.This is presently a hot topic and has a wide scope to try with multiple combinations.However, damping of power oscillations is a major limitation associated with virtual inertia emulation.Hence, this combination offers large research opportunities.3) Power oscillation damping through virtual impedance is an effective approach next to ESS.Being a virtual entity, the application of virtual impedance has a clear merit over the cost factor of ESS.However, in ESS, the application of D-FACTS devices needs justification for its cost.Scope: The abovementioned proposed research directions with respect to advanced converter control help to improve the power quality in all three modes of microgrid operation and stability/reliability in the islanded mode of microgrid operation.

VIII. CONCLUSION
Due to the lack of a comprehensive and systematic analysis of all possible state-of-the-art approaches under one title, this article provides a comprehensive evaluation that helps to understand various issues and solutions about the transient performance of microgrids.The strategy involved in the development of this article is summarized as follows.
1) Based on the earlier research works, the technical challenges involved with the transient performance of microgrids are identified and highlighted.2) Different possible approaches meant for transient performance improvement are identified and showcased.3) A comprehensive literature review is conducted.The study is strictly constrained between the boundaries of available approaches and their role in mitigating the identified challenges.4) Based on the review, new strategies as available from the literature are added to the existing review works which makes the approach up to date.Analytical analysis and comparison of existing strategies wherever possible under the proposed approaches are investigated based on the critical issues considered and noticed from the first, second, and third stages.5) A comparative analysis that projects the scope of each approach in handling the said challenges is presented.Based on this, future trends under each approach are presented.It is identified that, among the different approaches, because of inherent physical limitations and cost justification of ESS, the converter control approach followed by the grid synchronization approach, offers more possibilities for improving the transient response.From this review, an effective solution can be sought through the optimal combination of various approaches.Thus, it is said that this review may provide good research gaps and can be a helpful reference for microgrid researchers.

FIGURE 1 .
FIGURE 1. Single line diagram of the general structure of microgrid.

FIGURE 3 .
FIGURE 3. Organization of content into various sections of the paper.

FIGURE 4 .
FIGURE 4. Technical challenges in transient performance improvement.

FIGURE 5 .
FIGURE 5. Approaches and methods for the improvement of transient performance in microgrids.

FIGURE 6 .
FIGURE 6. Grid synchronization unit for control of GC three-phase VSI.

FIGURE 7 .
FIGURE 7. Structure of PLL with prefilter included in phase detector and with in-loop filter included in loop filter.

FIGURE 8 .
FIGURE 8. Small-signal model of a generic SRF-PLL.

FIGURE 11 .
FIGURE 11.System with controllable STS and the VSI operating in GC mode.

FIGURE 13 .
FIGURE 13.Control system block diagram for single current-loop with GCF or ICF.

FIGURE 14 .
FIGURE 14. Flowchart for designing LCL filter in direct control.

FIGURE 16 .
FIGURE 16.Control system block diagram for filter-based active damping approach in single current-loop with GCF or ICF.

FIGURE 18 .
FIGURE 18. Cascaded power and current controllers approach.

FIGURE 19 .
FIGURE 19.Control system block diagram for current control approach with capacitor current-based active damping while neglecting time delays.

FIGURE 20 .
FIGURE 20.Control system block diagram for current control approach with capacitor current-based active damping while including time delays.

FIGURE 21 .
FIGURE 21.Cascaded voltage and current controllers approach.

FIGURE 22 .
FIGURE 22. Block diagram of dual control structure requiring STS for mode transition.

FIGURE 23 .
FIGURE 23.Block diagram of single control structure not requiring STS for mode transition.

FIGURE 26 .
FIGURE 26.Single-DG grid system with nonlinear loads at PCC.

FIGURE 27 .
FIGURE 27.Overall control scheme for emulation of virtual inertia in voltage-controlled inverters.

FIGURE 28 .
FIGURE 28.Control diagram of P-ω control in traditional droop control with measurement delays for generation of ω * .

FIGURE 31 .
FIGURE 31.Ranking of various approaches based on their capability to address a particular technical challenge.(a) With respect to Deformed Grid Voltage.(b) With respect to Grid Impedance Changes.(c) With respect to Non-Linear Loads.(d) With respect to Time-Delays.(e) With respect to Changes in f res .(f) With respect to Coupling Between P and Q Controls.(g) With respect to Planned Islanding.(h) With respect to Un-Planned Islanding.(i) With respect to SA-GC Transition.(j) With respect to Black Start.(k) With respect to Large Load Changes.

TABLE 7 Comparison of Methods for Error Compensation of
the system.As a result, by minimizing time delay, the stable region can be widened.But, such improvements are only achievable if the inner active damping gain is properly designed; otherwise, instability would return.

TABLE 8 Comparison of Different Droop-Supported Approaches for Seamless Transition 554
VOLUME 4, 2023Authorized licensed use limited to the terms of the applicable license agreement with IEEE.Restrictions apply.