Non-Pilot Protection of the Inverter- Dominated Microgrid

Without utilizing costly communication systems, the existing protection strategies fail to reliably detect the occurrence and direction of faults in the inverter-dominated microgrid. To address this issue, this paper introduces a selective and reliable non-pilot protection strategy for the inverter-dominated microgrid. The proposed protection strategy (i) does not require communication signals, (ii) incorporates phase- and sequence-domain protective elements for reliable detection of symmetrical and asymmetrical faults, (iii) improves the existing sequence-domain directional elements and effectively utilizes them for accurate determination of the fault direction in the presence of inverter-interfaced distributed energy resources, (iv) selectively protects the inverter-dominated microgrid against internal and external faults, (v) is robust against the grid-connection mode of the microgrid, and (vi) enables fuse protection of laterals and non-critical circuits. The acceptable performance of the proposed protection strategy is verified through comprehensive fault studies conducted on a realistic study system simulated in the PSCAD/EMTDC software environment. This paper also demonstrates that the proposed protection strategy can be implemented using an off-the-shelf digital relay.


I. INTRODUCTION
The proliferation of the alternating current (AC) microgrid has been constrained by the lack of a cost-effective, selective, and reliable strategy for its protection against faults [1]- [4].The protection strategies used in traditional distribution networks are not generally applicable to microgrids [4], [5].The protection challenges are further complicated in the inverter-dominated microgrid to which the majority, if not all, of the distributed energy resources (DERs) are interfaced through inverters [2], [6], [7].The issue is threefold.First, the conventional over-current (OC) relays may fail to detect the limited fault currents contributed by the inverter-interfaced DERs (IIDERs) [2]- [4], [7], [8].Second, coordinating the OC relays in the inverter-dominated microgrid is challenging, due to the significantly different fault current levels under the grid-connected and islanded operation modes [2]- [4], [9].Third, the conventional The associate editor coordinating the review of this manuscript and approving it for publication was Bin Zhou .
phase-and sequence-domain directional elements fail to accurately determine the fault direction in the inverterdominated microgrid, under specific operating conditions, as demonstrated in this paper.
Different microgrid protection strategies have been proposed to address these issues.The differential protection strategy [6], [10] requires current measurement at all boundaries of its protection zone [3], [5], which may be costly and impractical in microgrids with multiple feeder sections and dispersed DERs [3].The adaptive protection schemes [11]- [14] resolve the fault detection and protection coordination issues but require prior knowledge of all possible configurations, operation modes, and the associated fault current levels [5], [14].The voltage-based protection strategy of [15] is intrinsically robust against variations in the fault current levels but suffers from lower sensitivity to faults in the grid-connected microgrid [5].A pilot protection scheme is used in [1] to enable coordination of directional OC relays using dual settings.A common disadvantage of the aforementioned protection strategies is that they all require remote communication which can be prohibitively expensive, especially in large-scale microgrids [3], [5].Relying on communication networks also increases vulnerability to communication failure and cyber-attacks [14].
The cost and reliability issues associated with the communication-assisted protection strategies can be avoided by utilizing non-pilot protection strategies, i.e., relays that do not require communication signals.However, the non-pilot protection strategies that have been introduced for microgrid applications [2], [6], [7], [16]- [18] do not reliably protect inverter-dominated microgrids under all possible operating conditions and fault scenarios.The harmonic measurement method of [2] requires special inverters for fifth harmonic injection and is only applicable to the islanded microgrid.A wavelet transform-based data mining strategy is proposed in [16] for fault detection and classification.This strategy requires building a data mining model based on the knowledge of all possible variations in the microgrid fault behavior, which is usually not practical.The symmetrical component-based protection method of [17] only detects asymmetrical faults [7], [18] and is only applicable to unigrounded microgrids [19].
The non-pilot directional protection strategies introduced in [6], [7], [18] effectively resolve the protection coordination issues by using the definite-time grading technique.However, these protection strategies have not been devised for inverterdominated microgrids.The study systems of [6], [7], and [18] contain sizable synchronous generators.As a result, the fault detection methods utilized in [7] and [18] suffer from low sensitivity to resistive ground faults in the islanded inverterdominated microgrid.The protection strategy of [6], which utilizes phase OC and negative-sequence OC elements, reliably detects asymmetrical faults but has low sensitivity to symmetrical faults in the islanded inverter-dominated microgrid.In addition, the protection strategies of [6], [7], and [18] would fail to accurately determine the asymmetrical fault direction under specific operating conditions in the inverterdominated microgrid, as explained in Section II.
This paper introduces a non-pilot protection strategy that effectively addresses the aforementioned issues related to fault detection, fault direction identification, and protection coordination, in the inverter-dominated microgrid.The proposed protection strategy incorporates (i) phase-and sequence-domain protective elements for reliable detection of symmetrical and asymmetrical faults, (ii) improved positive and negative sequence-domain directional elements for reliable detection of the fault direction, and (iii) a protection coordination strategy inspired by the definite-time grading technique of [6], [7], [18], for selective protection under both grid-connected and islanded operation modes.
Comprehensive fault studies are conducted on a detailed model of a realistic study system, in the PSCAD software environment.The results indicate that the proposed protection strategy operates reliably and selectively in an inverter-dominated microgrid, under both grid-connected and islanded operation modes.The practicality of the proposed protection strategy is also verified by implementing it in an off-the-shelf digital relay and demonstrating its desirable operation using an industrial relay testing platform.

II. DIRECTIONAL PROTECTION ELEMENTS
Directional protection elements have been widely used in traditional power systems, and their capabilities and shortcomings are well known to the research community.However, the consequences of utilizing these protective elements in the inverter-dominated microgrid are yet to be fully understood.This section briefly reviews the operating principles of the existing directional protection elements and highlights a misconception that leads to incorrect detection of the asymmetrical fault direction under specific conditions in the inverter-dominated microgrid.The information provided in this section constitutes the foundation of the microgrid protection strategy proposed in the next section.
The phase directional element discriminates between forward and reverse faults based on the phase-domain voltages and currents.This element may fail to accurately detect the fault direction if the current contains a large zero-sequence component [20], [21].The shortcomings of the phase directional element have led to the development of sequence directional elements, which usually assess the phase angle differences between the sequence-domain voltages and currents to identify the fault direction.Each of the positive-, negative-, and zero-sequence directional elements detects the directions of specific types of faults more accurately [3], [20].

A. SYMMETRICAL FAULTS
The positive-sequence directional element (PSDE) can be used to determine the symmetrical fault direction [3], [20], [22].Assume a symmetrical fault happens on a line connecting two areas of a microgrid, Fig. 1, where the per-unit distance to the fault is denoted by 0 ≤ m ≤ 1.Each of the areas A and B may contain sources, loads, lines, and the point of common coupling (PCC) with the utility grid.Representing these areas with the associated Thevenin equivalent circuits results in the simplified, yet accurate, circuit diagram of Fig. 2. In Fig. 2, Z L , Z A , and Z B respectively represent the line impedance and the Thevenin equivalent impedances of the areas A and B. Assume the relays at the two ends of the faulted line, i.e., R A and R B , utilize PSDEs with the reference directions shown in Fig. 2. The positive-sequence network corresponding to the symmetrical fault scenario is shown in Fig. 3, where the subscript 1 identifies the positive-sequence quantities and parameters.
The positive-sequence voltage phasors measured by the R A and the R B are defined by (1) and (2), respectively.
Neglecting the fault impedance (Z f ≈ 0), the approximate positive-sequence current seen by the R A and the R B are: The PSDEs of the R A and the R B measure the angles [22]: As indicated by ( 5) and ( 6), there is approximately 180 degrees difference between the angles of the positivesequence impedances seen by a relay, under forward and reverse symmetrical faults.This large difference is used by the PSDE to reliably determine the fault direction [22].The fault is assumed to be in the forward direction if the measured positive-sequence impedance angle falls in a half-plane of ±90 degrees around the element characteristics angle (ECA), which is typically set at Z 1L to maximize the security margin [20], [22].Fig. 4 shows the forward and reverse operating characteristics of the PSDE and the positive-sequence impedances measured by the relays R A and R B , when the fault of Fig. 1 is symmetrical [22], [23].
It is reported in [3], [21], [22], [24] that setting the ECA at Z 1L may cause false determination of the symmetrical fault direction in systems with high penetration of DERs, due to the associated reactive power injection.To prevent such issues, setting the ECA of the PSDE at values smaller than the Z 1L is proposed in [22] and [24].

B. ASYMMETRICAL FAULTS
The zero-sequence directional element (ZSDE) and the negative-sequence directional element (NSDE) have been widely used to determine asymmetrical fault direction in traditional power systems [3], [20], [21], [25].The ZSDE (i) does not identify the directions of phase-to-phase faults, (ii) has different design requirements depending on the system grounding strategy [3], [26], which drastically varies  between different microgrids [19], (iii) is prone to failure due to zero-sequence mutual coupling, and (iv) requires substation transformer neutral current measurement or brokendelta-connected voltage transformers [25], [26].The NSDE has been available in relays since 1990s and widely utilized by utilities, e.g., BC Hydro, [25], and does not suffer from the aforementioned issues [3], [20], [21], [25], [26].Thus, the rest of this section is focused on the operating principles and application challenges of the NSDE.
Assume the fault of Fig. 1 is asymmetrical and of single line-to-ground (SLG) type.The sequence network corresponding to this fault scenario is shown in Fig. 5, where the subscripts 0, 1, and 2 identify the zero-, positive-, and negative-sequence quantities and parameters [20], [26], [27].In Fig. 5, the areas A and B are represented by the corresponding Thevenin equivalent circuits in the sequence domain.Assume the relays at the two ends of the faulted line, i.e., R A and R B , utilize NSDEs with the reference directions shown in Fig. 2. The negative-sequence voltage phasors measured by the R A and the R B are defined by ( 7) and ( 8), respectively [20], [22], [27].The impedances Z 2A and Z 2B are the Thevenin equivalent negative-sequence impedances of the systems interconnected by the faulted line, hereafter referred to as the system negative-sequence impedances.By analyzing the sequence diagrams associated with line-to-line (LL) and line-to-lineto-ground (LLG) faults, it can be shown that ( 7) and ( 8) apply to all asymmetrical faults [6].
The negative-sequence currents measured by R A and R B are I R A = I 2A and I R B = −I 2B , respectively.Thus, if the fault of Fig. 1 is asymmetrical, the NSDEs of R A and R B measure the following impedance angles: The impedances Z 2A and Z 2B are not fixed and usually not accurately known.However, in traditional transmission systems and distribution networks that are not dominated by inverters, these are typically resistive-inductive impedances with angles close to the line impedance angle Z 1L .Therefore, the operating characteristic of the NSDE used in traditional power systems is similar to that of the PSDE, except the directional logic of the NSDE is inverted.Whenever the measured negative-sequence impedance is aligned with the Z 1L , a reverse direction is detected by the NSDE and vice versa [20], [27].Fig. 6 shows the forward and reverse operating characteristics of the NSDE and the negative-sequence impedances measured by the relays R A and R B when the fault of Fig. 1 is asymmetrical [22], [23].
The ECA of the NSDE is conventionally set at the line impedance angle Z 1L , based on a few assumptions that are only valid in traditional power systems.It has been reported that the NSDE may fail to correctly determine the asymmetrical fault direction in the presence of IIDERs [3], [21], [25], [28].The next part investigates this issue and proposes a simple but effective solution.

C. REVISITING THE NSDE
The practice of setting the ECA of the NSDE at Z 1L was initially adopted to maximize the operating torque produced in electromechanical relays.This is the reason that the ECA is also referred to as the maximum torque angle (MTA) [20], [22].The same strategy is still used in setting modern digital relays, assuming that the angles of the system negative-sequence impedances, i.e., Z 2A and Z 2B , are close to the line impedance angle Z 1L .This assumption is not necessarily valid in an inverter-dominated microgrid where the measured angles Z 2A and Z 2B depend on the control strategies of the IIDERs.This issue is theoretically explained below and also demonstrated using simulation results in Section IV.
Under asymmetrical faults, the IIDERs exchange different amounts of negative-sequence reactive current with their host systems, depending on their control and current limiting strategies [8], [9], [25], [29]- [33].It is shown in [29] that the IIDERs operating based on the voltage support control strategy, e.g., battery energy storage systems (BESS) in the islanded microgrid, can inject considerable amounts of negative-sequence reactive current into the faulted host system.Besides, the studies reported in [9] indicate that the IIDERs operating based on the power control strategy, e.g., photovoltaic (PV) generation system and wind turbines (WT), including those that aim to suppress the negative-sequence current, inject small amounts of negativesequence reactive current to the faulted host system, due to their harmonic filter capacitors.The reactive behavior of the IIDERs in the negative-sequence domain can change the impedance angles seen by the NSDEs and lead to their malfunction.
Fig. 7 shows the potential impact of the negative-sequence reactive current injected by the IIDERs in the Area B on the impedance measured by the NSDE of the R B , when the fault of Fig. 1 is asymmetrical.As shown in Fig. 7, depending on the sizes, types, and locations of the IIDERs, the resulting shift in the measured negative-sequence impedance can even cause the NSDE to see a reverse fault as a forward fault, if the ECA is set at Z 1L .The relay R A could experience similar issues during asymmetrical faults in the islanded microgrid where the strong grid no longer exists and the negative-sequence behavior of the IIDERs in the Area A become more impactful.Malfunction of the NSDEs due to the impacts of the IIDERs has been reported in [25] and [28], which confirms the above analysis.Application of the ZSDE instead of the NSDE is recommended in [25].Due to the wellknown adverse effects of the zero-sequence mutual coupling on the ZSDE, this is not an ideal solution.Using a smaller non-zero ECA is recommended in [28] to avoid the aforementioned issues.To maximize the reliabilities of the PSDE and the NSDE in the inverter-dominated microgrid, this paper proposes setting the associated ECAs at zero degrees.This results in complete desensitization of these sequence-domain directional elements to the reactive components of the positive-and negative-sequence currents.This strategy enables the PSDE and the NSDE to operate only based on the active power imbalance.The necessity and effectiveness of the proposed solution is verified in Section IV, using comprehensive fault studies performed on a realistic microgrid study system.

III. PROPOSED PROTECTION STRATEGY
The non-pilot protection strategies of [6], [7], [18] utilize an interface protection relay (IPR) at the PCC and multiple microgrid protection relays (MPR) along the microgrid feeder(s).In this section, improved IPR and MPR algorithms are proposed for protection of the inverter-dominated microgrid.Fig. 8 shows the reference directions assumed in this paper for the IPR and the MPRs as well as the locations of the circuit breakers associated with these relays in a typical microgrid.It should be noted that the proposed IPR and MPR are able to operate under faults in either directions, i.e., incorporate independently operating forward and reverse protection functions.The reference directions shown in Fig. 8 are only used to coordinate the relays.The proposed relays utilize a combination of phase-and sequence-domain elements to reliably detect the occurrence of symmetrical and asymmetrical faults.Besides, the fault direction is determined using the improved sequence-domain directional elements, i.e., the PSDE and NSDE with the zero ECA setting, as described in Section II.
The fault current magnitude in an islanded inverterdominated microgrid does not considerably depend on the fault location.The reason is that the fault current is mainly dictated by the current limits of the inverters [9] and not the fault loop impedance.Hence, the traditional time-current (inverse-time OC) grading strategy does not guarantee coordinated operation of non-pilot protective devices in the inverter-dominated microgrid.The proposed protection strategy is based on the definite-time grading technique of [6], [7], [18], with extra provisions that enable reliable and selective protection of the inverter-dominated microgrid.

A. INTERFACE PROTECTION RELAY
The IPR shown in Fig. 8 must discriminate between internal (forward) and external (reverse) faults and trip the PCC circuit breaker in a timely manner, whenever a fault is detected in either direction.The fault detection criteria and the tripping delay of the IPR depend on the fault type and direction, as described below.

1) PROTECTION AGAINST SYMMETRICAL FAULTS
Under internal symmetrical faults, the fault current contribution from the utility grid is expected to be relatively large for all three phases [7].Hence, the forward-direction fault timer of the IPR starts once a symmetrical fault is detected by three instantaneous phase OC elements, and a forward fault direction is detected by the improved PSDE.The pickup setting of the OC elements, I P−PU , is set at a value that is higher than (i) two times of the maximum balanced load current, and (ii) the maximum expected motor starting current, seen by the IPR.The IPR trips whenever the timer reaches the threshold TD fwd , i.e., the forward-direction time delay setting.
The response of the inverter-dominated microgrid to external symmetrical faults is drastically different.Due to the limited fault current contributions of the IIDERs, the IPR has to detect external symmetrical faults using Under-Voltage (UV) elements.Hence, the reverse fault timer of the IPR starts once a symmetrical fault is detected by three instantaneous phase UV elements, and a reverse fault direction is detected by the improved PSDE.The pickup setting of the UV element, V P−PU , is set at 50 percent of the rated voltage.The IPR trips whenever the timer reaches the threshold TD rev , i.e., the reverse-direction time delay setting.

2) PROTECTION AGAINST ASYMMETRICAL FAULTS
The IPR detects asymmetrical faults using a negativesequence OC element [6], [17], [34] which does not react to balanced load currents and thus can use a small pickup setting [34], [35].A negative-sequence over-voltage (OV) element is also used to enable the IPR to detect resistive asymmetrical faults.The IPR determines the asymmetrical fault direction using the improved NSDE.The forward (reverse) fault timer of the IPR starts once (i) an asymmetrical fault is detected, and (ii) a forward (reverse) fault is indicated by the NSDE.
The pickup setting of the OC element, I 2−PU , is set at twice the maximum negative-sequence current caused by unbalanced loads under normal condition.The pickup setting of the OV elements, V 2−PU , is set at a value higher than twice the maximum expected negative-sequence voltage caused by unbalanced loads under normal condition.The IPR trips whenever the forward (reverse) fault timer reaches the threshold TD fwd (TD rev ).Fig. 9 shows the logic diagram of the IPR, where D 1 and D 2 represent the output signals of the improved PSDE and NSDE and become high under forward symmetrical and asymmetrical faults, respectively.
The ''load current detection'' signal in Fig. 9 indicates whether the system is energized.This signal is used in the relay logic diagram to prevent the UV elements from unnecessarily tripping the circuit breakers when the system is de-energized.

B. MICROGRID PROTECTION RELAYS
Each MPR is expected to detect the occurrences and directions of internal and external faults and operate in coordination with other protective devices.Due to the reasons explained in the next part, the proposed protection strategy (the IPR) islands the microgrid under any fault, before the first MPR trips.Thus, the MPRs make protective decisions only in the islanded operating conditions, which simplifies the MPR algorithm.
Symmetrical faults in the islanded inverter-dominated microgrid cause the phase voltage magnitudes to drop significantly.Therefore, in the proposed protection strategy, each MPR detects symmetrical faults using instantaneous phase UV elements.The forward (reverse) fault timer of each MPR starts once (i) a symmetrical fault is detected, and (ii) a forward (reverse) fault direction is indicated by the improved PSDE.The MPR trips whenever the forward (reverse) fault timer reaches the threshold TD fwd (TD rev ).The pickup setting of the phase UV elements, V P−PU , is set at 50 percent of the rated voltage.For branch lines without DERs, phase OC elements are added to the fault detection logic to prevent false tripping under faults on adjacent feeders/lines [18].The asymmetrical fault detection algorithm of the MPR is identical to that of the IPR.these modules detects a fault and the corresponding timer exceeds a pre-determined delay.

C. PROTECTION COORDINATION
Selective protection of the inverter-dominated microgrid is achieved through coordinated operation of the IPR, the MPRs, and the lateral fuses.Under external faults, the IPR is expected to trip fast, without delaying its operation for coordination with other protective devices.This strategy prevents exposure of the microgrid components to prolonged voltage sags caused by utility grid faults, and also minimizes the risk of islanding a portion of the utility grid (energizing external faults) for an extended period of time.However, to prevent unwanted tripping of the IPR due to system transients, induced voltages, etc., the short delay of TD rev = 50 ms is used (Fig. 9).
Under internal faults, the IPR must trip the PCC circuit breaker adequately fast, in order to minimize exposure of the microgrid components to large fault currents.However, instantaneous tripping of the IPR may increase the operation times of the fuses within the microgrid under downstream faults, due to the subsequent drop in the fault current magnitude.When an internal fault takes place, the upstream fuse (if any) is the first protective element to react to the fault.Thereafter, the IPR causes forced islanding.To provide the fuses with sufficient fault clearing time, the IPR utilizes a fixed forward fault delay TD fwd for both symmetrical and asymmetrical faults.The delay is larger than the maximum total clearing time (TCT) of all fuses within the microgrid, in the grid-connected mode.It is determined by applying SLG faults to all fuse-protected laterals, with a reasonably large fault resistance which is assumed to be 40 [36].
After islanding, the MPRs operate in coordination with each other to selectively isolate the faulted feeder/line section, if the internal fault is not cleared by any fuse.The coordinating time interval (CTI) that is utilized to maintain sufficient margin between the operating times of distribution system relays is typically in the range of 0.2 to 0.5 s [1], [26].In this paper, all relays are coordinated with each other by a CTI of 0.2 s.

IV. STUDY RESULTS
This section presents the results of time-domain simulation studies performed on an inverter-dominated microgrid study system in the PSCAD/EMTDC software environment.The study system of Fig. 11 is developed by modifying the Canadian benchmark rural distribution system of [37] to enable its operation as a microgrid.The microgrid includes a PV generation system, a WT, two BESSs, and highly unbalanced loads that are distributed along a 24 km main feeder.All DERs are inverter-interfaced.The DERs and the associated controllers are represented by detailed models, as reported in [38].The microgrid is assumed to be low-reactance grounded and of the four-wire multi-grounded configuration [19].The parameters and specifications of the study system components are given in the Appendix.The relay R1 in Fig. 11 is the IPR, and the relays R2-R6 are the MPRs.
The study system is used to demonstrate the adverse effects of IIDERs on the operation of the conventional NSDE and also to verify acceptable performance of the proposed protection strategy that incorporates effective fault detection, fault direction determination, and protection coordination methods.Different types of symmetrical and asymmetrical faults, including LG, LL, LLG, and three-phase faults have been applied to six different locations in the microgrid, shown as F1-F6 in Fig. 11.The fault resistance R f is assumed to be zero, 10 , and 50 .The microgrid is in the grid-connected steady-state before the faults are applied.The reported relay settings and simulation results are either in per-unit or the primary values, i.e., correspond to the values at the primarysides of the instrument transformers.

A. NSDE IN THE INVERTER-DOMINATED MICROGRID
The malfunction of the conventional NSDE is demonstrated in this sub-section.Fig. 12 shows the negative-sequence impedances measured by the relays R1-R6, under 99 asymmetrical fault instances with the aforementioned fault types and fault resistances, at the locations F1-F6 shown in Fig. 11, under both grid-connected and islanded modes.To simplify the analysis, the measured impedances are divided into four categories in Fig. 12, depending on the grid-connection mode and the fault location with respect to the relay reference directions.The faults corresponding to the impedances in the shaded areas are seen by the conventional NSDEs as forward faults.Figs.12(c) and (d) show that the conventional NSDE fails to detect the correct fault direction under a large portion of reverse faults, which confirms the theoretical analysis presented in Section II-C and illustrated in Fig. 7. Fig. 13 shows that the proposed solution, i.e., setting the ECA of the NSDE at zero degrees, effectively resolves the issue.Fig. 13 also demonstrates the necessity of using an ECA setting of zero degrees (or the closest available setting to zero), since any ECA below -15 degrees would cause issues under forward faults, Fig. 13(a), and any ECA above 10 degrees would cause issues under reverse faults, Fig. 13(c).

B. PERFORMANCE OF THE PROPOSED PROTECTION STRATEGY
This part evaluates the performance of the proposed non-pilot protection strategy applied to the study system of Fig. 11.The relays R1-R6 are coordinated with each other and with the downstream fuses, according to the strategy introduced in part III-C.Coordination of the forward and reverse elements of the relays is performed separately, as follows: Forward-directional elements: R2→R3→(R5& R6)→R1 Reverse-directional elements: R5→R4→R3→R2→R1 The notation Rj →Rk means that, for any fault that is seen by the Rj and the Rk in the same direction, the relay Rj operates with a larger delay as compared with the Rk.It should be noted that the R1 operates faster than all relays in both directions, because it is designated as the IPR and   must always trip before all MPRs.The relay R6 does not include a reverse directional elements, because it is installed on a passive branch that would never feed reverse faults.The relay R4 does not include forward directional element due to the presence of the R6 and the R5 on the same bus.A forward element in the R4 would be redundant and would increase the forward operating times of the R3 and the R2 by 200 ms.The relays R1-R6 are coordinated with each other and with the lateral fuses, according to the coordination strategy described in Section III-C.The relay settings are shown in Table 1.
Faults of different types with various resistances are applied to different locations under both grid-connected and islanded operating modes of the microgrid.Table 2 shows the operating time delays of the relays and the TCTs of the fuses that protect the faulted laterals (if applicable), for internal faults.Table 3 shows the relay operating time delays under external faults.
The study results indicate that the proposed protection strategy selectively and reliably protects the microgrid under the grid-connected and islanded modes and various fault conditions.For the cases where the fault is applied to a lateral circuit protected by a fuse, i.e. the faults F2-F5 in Fig. 11, the corresponding lateral fuse is the first protective device that operates, as shown in Table 2.
Table 2 also shows that the reduced fault current level in the islanded microgrid increases the fuse TCTs.However, in both grid-connected and islanded modes, the minimum margin of about 200 ms is maintained between the operation times of all protective devices that see the fault in the same direction.The reason is that the relays are coordinated with the fuses, taking into account the fault current variations in both operation modes of the microgrid.
The proposed protection strategy enables fuse protection of the lateral circuits, provided that the current ratings of the fuses are below the fault current magnitude in the islanded inverter-dominated microgrid.However, if fuses with significantly higher current ratings are utilized (in case the loads protected by the fuses are considerably larger), the limited fault current magnitude in the islanded inverter-dominated microgrid may not be sufficiently large to enable fuse operation in a timely manner.Under such conditions, the fuses merely operate in the grid-connected mode, and the islanded microgrid is protected by the relays.It should be noted that this is not a shortcoming of the proposed protection strategy, because changing the relay operating times would not enable the fuses to operate in such conditions.To maximize the sensitivity of each fuse to faults in the islanded inverter-dominated microgrid, application of fuses with unnecessarily large current ratings should be avoided.
The maximum operating time of all protective devices under the investigated fault conditions is about 1205 ms.The aforementioned delay is due to the fact that six relays are coordinated with each other and with the lateral fuses in the study system of Fig. 11.These numerous protective devices are intentionally utilized in this study system to demonstrate the capabilities of the proposed protection strategy.However, fewer relays are typically utilized on a microgrid feeder, which reduces the maximum relay operation time.In addition, even in the study system of Fig. 11, none of the faults F1-F5 would remain uncleared for such long time periods unless multiple protective devices fail to operate.
It should be noted that the proposed protection strategy is not designed to achieve the highest speed among the existing microgrid protection strategies.The existing communication-based protection strategies typically operate faster.The main contribution of this paper is that the proposed protection strategy enables reliable and selective protection of the inverter-dominated microgrid, with reasonable speed, without utilizing communication systems.The proposed protection strategy is suitable for inverterdominated microgrids where application of communication systems is not feasible due to practical or economic reasons.In addition, the proposed protection strategy can be utilized as backup for the communication-assisted protection schemes.
The IPR and MPR algorithms of Figs. 9 and 10 are implemented in an off-the-shelf commercial relay [39] and evaluated using an industrial relay testing platform [40] The performance of the relay is evaluated under various fault scenarios by reading the event files recorded by the relay after the test signals are played back to the relay by the testing device.The investigation verifies that the proposed microgrid protection strategy can be implemented in an off-the-shelf digital relay, and the performance of the relay matches with the results obtained by PSCAD simulation.    2 and 3.

V. CONCLUSION
This paper introduces a non-pilot protection strategy for the inverter-dominated microgrid.The study results indicate that the proposed protection strategy: • enables reliable detection of symmetrical and asymmetrical faults, reliable identification of the fault direction, and selective protection under both grid-connected and islanded operation modes.
• enables fuse protection of lateral circuits.
• prevents microgrid exposure to prolonged fault currents from the utility grid, under internal faults.
• minimizes microgrid exposure to voltage sags during utility grid faults.• can be implemented using the existing commercial relays.Although the proposed protection strategy is devised to achieve improved performance in inverter-dominated microgrids, it is also applicable to microgrids that contain rotating machines instead of inverters.However, in systems that do not suffer from the issues associated with the inverter-dominated microgrids, higher protection speed can be achieved by using the traditional protection coordination method.

FIGURE 2 .
FIGURE 2. Simplified representation of the faulted system of Fig. 1.

FIGURE 3 .
FIGURE 3. Positive-sequence network corresponding to the faulted microgrid of Fig.1, when the fault is symmetrical.

FIGURE 4 .
FIGURE 4. Operating characteristics of the PSDE and the positive-sequence impedances seen by R A and R B when the fault of Fig. 1 is symmetrical.

FIGURE 5 .
FIGURE 5. Sequence network corresponding to the fault scenario of Fig.1, when the fault is of SLG type.

FIGURE 6 .
FIGURE 6.Operating characteristics of the NSDE and the negative-sequence impedances seen by R A and R B when the fault of Fig. 1 is asymmetrical.

FIGURE 7 .
FIGURE 7. Impact of the negative-sequence reactive power injection by IIDERs on the impedance measured by the NSDE during a reverse asymmetrical fault.

FIGURE 8 .
FIGURE 8. Locations and reference directions of the IPR and the MPRs.

FIGURE 9 .
FIGURE 9. Logic diagram of the proposed IPR.

Fig. 10
shows the logic diagram of the proposed MPR.Similar to the IPR logic diagram of Fig. 9, the MPR logic diagram consists of separate asymmetrical and symmetrical fault detection modules.The MPR trips whenever either of

FIGURE 10 .
FIGURE 10.Logic diagram of the proposed MPR.

FIGURE 11 .
FIGURE 11.Single-line diagram of the inverter-dominated AC microgrid study system.

FIGURE 12 .
FIGURE 12. Operating characteristics of the conventional NSDE, and the negative-sequence impedances measured by relays R1-R6, under (a) forward faults in the grid-connected microgrid, (b) forward faults in the islanded microgrid, (c) reverse faults in the grid-connected microgrid, and (d) reverse faults in the islanded microgrid.

FIGURE 13 .
FIGURE 13.Operating characteristics of the proposed NSDE, and the negative-sequence impedances measured by relays R1-R6, under (a) forward faults in the grid-connected microgrid, (b) forward faults in the islanded microgrid, (c) reverse faults in the grid-connected microgrid, and (d) reverse faults in the islanded microgrid.
. The voltage and current signals obtained by simulating various fault scenarios are loaded to the relay testing device which sends the recorded signals to the relay, as shown in Fig. 14.The MPR and IPR algorithms are implemented in the relay by appropriately setting the parameters of the available phaseand sequence-domain protective elements as well as defining new Logic Variables (LVs) and variable timers.The graphical logic diagram of Fig. 15 is created by the user interface software of the relay manufacturer and shows how the aforementioned LVs and timers are integrated with the internal protective elements of the relay to create the MPR algorithm.

FIGURE 14 .
FIGURE 14.Hardware setup for implementing the proposed protection strategy in an off-the-shelf digital relay and testing its performance under the fault scenarios of Tables2 and 3.

FIGURE 15 .
FIGURE 15.Graphical logic representation of the MPR algorithm implemented in the digital relay shown in Fig. 14, which is shown by the user interface software provided by the relay manufacturer.

TABLE 1 .
Relay types and settings.

TABLE 2 .
Relay and fuse operating time delays for internal faults.

TABLE 3 .
Relay operating time delays for external faults.

TABLE 4 .
Study system parameters.