Robust Coordination Scheme for Microgrids Protection Based on the Rate of Change of Voltage

The wide application of microgrid concept leads to challenges for the traditional protection of distribution networks because of the changes in short circuit level and network topology during the two modes of microgrid operation. This paper proposes a promising solution for these problems by offering a new protection coordination scheme not affected by the variation of short circuit level or the changes in network topology. The proposed protection scheme is based on local measurements at relay location with low sampling frequency by computing the rate of change of fundamental voltage (ROCOV) to detect different fault types, identify the faulty zone accurately and guarantee robust coordination between primary and backup relays. The proposed coordination scheme can be achieved by optimizing either two settings for relay characteristic (time dial setting and pickup value) or four settings (time dial setting, pickup and the parameters that control the characteristic shape A & $B$ )). The proposed scheme is extensively tested using MATLAB simulations on the modified IEEE 14 bus meshed network embedded with synchronous/inverter-based distributed generation units under wide variations in operating conditions and short circuit levels for both grid-connected and islanded modes of operation. The achieved results confirm that the proposed coordination scheme can maintain the coordination between primary and backup relays for different fault locations, types and different topologies. It provides selective, reliable, and secured microgrid operation compared with conventional schemes, using fault current limiters and some other techniques discussed in the literature.


I. INTRODUCTION
A microgrid has become increasingly popular as an attractive solution for more sustainable and greener production of energy. It offers on-site power generation at the consumption point with improved reliability and reduced distribution losses [1].
Despite the numerous benefits of the microgrid with distributed generation (DG) integration, the protection challenges become serious concerns, where the performance of the traditional protection coordination schemes may be ineffective when applied to microgrids since they are susceptible to malfunctions and false tripping [2]. The protection relay The associate editor coordinating the review of this manuscript and approving it for publication was Bin Zhou . faces substantial difficulties as the fault current magnitude varies significantly depending upon the size, type and location of DG [3]. Furthermore, dynamically changing load, generation and network topology cause a significant change in fault currents which sometimes results in a miscoordination of one or more primary and backup traditional directional overcurrent relays (DOCRs) that are commonly used as main protection relays of microgrid networks. Such miscoordination results in unwanted false tripping for some healthy feeders and loads [4].
The available techniques in the literature for keeping the relays coordinated in both grid and islanded modes can be classified into local and communication based approaches. The schemes of the first category do not need communications [5]- [10]. In [5], due to the difference in fault VOLUME 9, 2021 This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/ magnitudes for grid-connected and islanded operation, fault current limiters (FCLs) are positioned and traditional DOCRs are optimally coordinated considering both microgrid modes of operation. By using FCLs, the infeasibility of conventional DOCRs to provide proper protection coordination was overcome; however, the coordination is violated with any change in the network operating conditions. A thyristor-based scheme is proposed in [6] to identify a distribution system operating condition and adapt the overcurrent protection of the grid. In that scheme, the system equivalent impedance is estimated, which differs for islanded or connected-operating conditions. Then, the pre-determined suitable setting is selected without any communications. However, that scheme is not effective with any variation in the system operating circumstances. A time-current-voltage characteristic (TCVC) is also proposed in both grid-connected and islanded modes of operation in [7]. The TCVC uses the faulted phase voltage and current magnitudes for determining the operating time of the relay. The TCVC does not require any communication system and achieves a notable reduction in total relay operating times. However, optimal settings of overcurrent relays are needed for every change in network topology. Moreover, the scheme is only tested with synchronous-based DG units and solid faults are only assumed. It may not be effective in a system that is dominated by inverter-based DG units due to their low fault current contribution. A voltage-based protection method for distribution systems with DG is proposed in [8]. In which, the relay characteristic is formulated from extensive analysis for voltage behavior during fault conditions. The method is communication-less and independent of mode of operation. However, it is very sensitive to any slight changes in voltage, thus leading to potential maloperation. Non-unit protection method is also investigated in [9] for fault discrimination within DC microgrid systems. It analytically studies the current and voltage signals, their rate of change (ROCOI, ROCOV ) and impedance profiles as measured at the generator converter terminals. In that study, there is no effective coordination method between primary and backup relays. The method is investigated for two fault locations only. When the fault conditions change, the protection relays cannot achieve the selectivity criteria since the fault location is based on constant threshold values. Another recent scheme for microgrid protection is introduced in [10] depending on dual protection settings for DOCRs. The first setting of TDS for primary protection is based on the very inverse curve while the second setting for backup protection is based on the normal inverse curve. The results show its superior performance over the conventional dual setting method by reducing the total operating time. Besides, the protection method does not require any communication links between relays. However, the method is still dependent on DOCRs and has not been evaluated under the wide variation in the short circuit levels under the changes in the operating conditions of the microgrid. Therefore, the pickup current and TDS need to be re-adjusted again under any changes in the operating conditions. Also, the method is not been tested with different types of DG. On the other hand, several microgrid protection schemes that rely on communications have been proposed in [11]- [21]. Adaptive settings of the relays have been applied in [11]- [13], where these schemes require a direct or indirect communication channel to reset the relays settings according to the prevailing conditions such as operational or topological changes. In [13], an adaptive protection coordination scheme has been discussed based on a centralized controller running the real time analysis of the data received from the intelligent electronic devices. The wide area monitoring system is implemented by the application of phasor measurements units and implemented for all nodes and branches of the AC grid in [14]. A differential protection strategy is developed using data mining techniques which relies on communicating measurements between two relays of the protected feeder [15], [16]. In [17], a travelling wave based protection scheme that utilizes a low bandwidth communication for meshed distribution systems with DG operating as a microgrid was proposed. In [18], [19], dual setting DOCRs have been applied for meshed distribution systems with DG. Relays are coordinated in such a way to reduce the overall relay operating time for grid-connected and islanded modes. As clearly shown, the communication system plays an important role in the adaptive protection methods. The cost, speed, redundancy, and reliability of the communication systems are vital factors that must be considered before implementing an adaptive protection method [20], [21]. Besides, the communication failure may lead to the inability of protection scheme [8].
Some other research studies were conducted towards applying the rate of change of the phasor voltage (ROCOV ) in islanding detection and distance protection. In [22]- [24], the rate of change of voltage and other parameters are used to detect the islanding condition at the point of common coupling between the distributed generation units and distribution networks. These algorithms are applied to detect DG operating modes correctly and quickly. Also, the conventional distance relay performance is enhanced when the ROCOV feature is added in [25]- [27]. Such algorithm can distinguish accurately the faulty cases and the stressed conditions. In fact, the above methods are not designed to handle the coordination of primary and backup relays to ensure secured feeder protection in microgrids. Figure 1 summarizes most of the microgrid protection approaches in literature with the general features and limitations of each category. Through the different research studies arranged schematically in Fig. 1, it can be deduced that there is an urgent need to propose a coordination approach which does not depend on the current as the current changes significantly with the changes occurring in microgrid topology. It is also necessary for the required proposed approach to avoid using communication systems to change relay setting frequently since it reduces system reliability and it has a high cost as well.
The main contribution of this paper is to propose a robust protective coordination scheme suitable for microgrids. The proposed coordination scheme is formulated here as an optimization problem for each mode of operation: grid-connected and islanded modes. The protection scheme is based on computing the rate of change of fundamental voltage (ROCOV ) to discriminate and locate the faulty section relying on local measurements only. The main feature of the new proposed coordination scheme is that it will not be affected by any variation of the network topology or short circuit level. It must be pointed out that ROCOV relay was developed by the authors in [28] and fully examined with simulation and practical implementations. As deduced in [28], ROCOV relay is stable during transient healthy conditions and provides a selective, reliable, and sensitive protection system in case of faults in distribution systems compared to conventional relays (overcurrent and under voltage relays).
The organization of this paper is presented as follows: description of the proposed protection scheme is offered in Section II. It briefly offers the basic idea of ROCOV relay, and then the problem formulation for ROCOV relays coordination is described. Test system description and the optimum settings of proposed protection scheme are presented in Section III. Simulation results for evaluating the proposed protection scheme on modified IEEE 14 bus system with wide variation in short circuit levels are discussed in Section IV. Comparison between the proposed scheme and other techniques is presented in Section V. Proposing user-defined characteristics for microgrid protection coordination using ROCOV relays is described in Section VI. Finally, the conclusions are drawn in Section VII.

II. DESCRIPTION OF THE PROPOSED PROTECTION SCHEME
In this section, the ROCOV relay for fault detection and the proposed coordination scheme formulation for grid-connected and islanded modes are presented.

A. DESCRIPTION OF THE ROCOV RELAY
The proposed scheme is based on the fact that the rate of change of the fundamental voltage is close to zero under normal operating conditions while it jumps to higher values under fault conditions at the fault instant. Fig. 2 shows the voltage (sinusoidal and RMS) and ROCOV waveforms at relay R1 during a fault at the mid-point of feeder 1 (F1) in Such ROCOV value can be calculated using the following equation: where: V A1(n) and V A1(n−1) are the calculated RMS values of fundamental voltage for phase A at present sample (n) and previous one (n − 1), respectively, while T is the sampling interval.
The ROCOV value at relay location is calculated using the pre and post RMS fault voltages during the sampling interval. The proposed ROCOV value is a function of fault distance. The ROCOV value is also a function of the sampling rate. The higher the sampling rate, the higher the accuracy. In fact, low sampling frequency in the range of 1-20 kHz can be applied to implement the proposed scheme. All results of studied cases in this paper are extracted with 20 kHz sampling frequency.
The ROCOV relay is supposed to be installed at the beginning and end of each protected main line section. In the event of a fault in any line section, the line should be disconnected from the beginning and the remote end of the line via installed ROCOV relays similar to traditional DOCRs used in the microgrid [5]. However, if the feeding is from one end only, then only one protection device is placed at the beginning of the line section as in radial distribution networks.
As discussed in [28], the relation between the measured ROCOV values and the relay operating time takes the same shape of the standard inverse-time characteristics of DOCR. Hence, a similar equation is used to describe this relation as given in (2).
where: t(op): is the operating time of the ROCOV relay.
-TDS : is the time dial setting of the ROCOV relay.
-ROCOV SC : is the maximum measured rate of change of fundamental voltage during the first cycle after fault occurrence. -ROCOV pick−up : is the setting point value of the ROCOV relay determined by normal load switching at relay location. -A and B: are constants that control the characteristic shape of the ROCOV relay. The setting of ROCOV relay in this paper will be designed based on the assumption of single fault occurrence in the relay protection zone, which can be considered a reasonable assumption when utilizing only local measurements. The constants A and B can be defined according to typical characteristics of Table 1 [29]. In this paper, A and B are chosen to be 0.14 and 0.02, respectively, to represent similar characteristics to the DOCR standard inverse characteristics. Besides, in Section VI, the coordination problem is also reformulated for optimizing both A and B using user-defined characteristics.
If the calculated ROCOV value from (1) exceeds a pickup value, the relay determines the required operation time, t from (2) and trips at the end of the estimated delay. It is worth mentioning that the ROCOV relay operates in conjunction with a directional feature (similar to traditional DOCR) in order to act only when the power flow is in the forward direction.
It should be noted that the methodology of the ROCOV relay is fully described by the authors in [28]. It is also worth clarifying that the ROCOV relay is extensively tested in [28] under different healthy and faulty conditions in distribution systems. The achieved results demonstrate the stability of the ROCOV relay under transient healthy conditions including dynamic load (starting transients of induction motors), static nonlinear load and capacitor switching. For all transient healthy conditions, calculated ROCOV value was less than the pickup value, and thus the relay does not generate any false tripping signal.

B. PROTECTION COORDINATION SCHEME FORMULATION BASED ON THE ROCOV RELAY
Two groups of settings are required to be stored in the ROCOV relay for grid-connected and islanded modes. The switching between the two modes can be easily achieved based on an effective islanding detection technique based on local measurements proposed by the authors in [30].
The protection coordination problem is typically formulated as an optimization problem (similar to typical DOCRs given for example in [18]), where the main objective is to minimize the overall relays operating time for each mode of operation: grid-connected and islanded modes.
For each mode of operation, the objective function T is the sum of the operating times of all ROCOV relays for all fault locations which needs to be minimized as follows: where: t pij and t bkj are the operating time of the primary ROCOV relay (i) and its all backup ROCOV relays (k) respectively for a fault location (j) calculated using (2). -N represents the total number of relays, while M denotes the total number of fault locations. The objective function should be achieved while fulfilling the following set of constraints for both modes of operation: where CTI is the coordination time interval that must be satisfied to achieve discrimination between the primary (i) and backup ROCOV relay (k) for a fault at j. The CTI usually takes a value between 0.2 and 0.5 s; it is set to be 0.2 s in this study. Other constraints include limits on the relays' are presented as follows: The minimum and maximum pick-up (ROCOV pick−upi ) depend on the maximum load switching condition at each VOLUME 9, 2021 relay location. As discussed in [31] for DOCRs, the pickup value of ROCOV relay can vary between 1.01 and 2 times the maximum value obtained under normal load switching [28]. The TDS min and TDS max are the minimum and maximum limits for relay i with values of 0.05 and 1.5 s respectively.
Various optimization methods, including heuristic and exact techniques can be applied to solve the optimization problem to achieve minimum operating time [32]- [33]. The problem is simply solved here using the MATLAB built-in fmincon optimization function. The protection coordination problem has been formulated as a non-linear programming (NLP) problem [34], by considering both the time dial and pickup settings to be continuous variables.

III. TEST SYSTEM AND THE OPTIMUM SETTINGS A. TEST SYSTEM
The modified IEEE 14 bus system with six added synchronous DG units is presented in Fig. 3. The synchronous based DGs are located at buses 1, 2, 3, 5, 7 and 10. The added DGs are rated at 2.4 MVA and 0.9 power factor. Other network parameters and data are presented in [5]. The modified IEEE 14 bus system is chosen in this paper as a test system for the purpose of comparison with the conventional DOCRs protection scheme which applies FCL on the same test system in [5].

B. THE OPTIMUM SETTINGS FOR ROCOV RELAYS
The microgrid operating philosophy is that in normal condition the microgrid is desired to operate in the grid-connected mode but in case of any disturbance, it would seamlessly disconnect from the utility at the point of common coupling (PCC) via relay R33 (shown in Fig. 3) and then continue to operate in the islanded mode [5]. Therefore, the optimum settings for the ROCOV relays are calculated for both grid-connected and islanded modes, according to the procedure mentioned in above sections, as presented in Table 2 (33 relays in grid-connected mode and 32 relays in islanded mode).

IV. EVALUATION THE PERFORMANCE OF THE PROPOSED COORDINATION SCHEME
The proposed protection scheme is extensively examined with different fault locations, types and different topologies in the next sections. Tables 3 and 4 list the relays operating time calculated by using (2) in case of different fault locations for both grid-connected and islanded modes of operation based on ROCOV . The results in Table 3 and 4 show that the required CTI is maintained between all primary and backup relays for all tested fault locations (F1 to F13 in Fig. 3) for both gridconnected mode (53 pairs of primary and backup relays) and islanded mode (51 pairs). As shown in the tables, the minimum and maximum recorded CTI for grid-connected mode was 0.2033 and 0.7645 s respectively, while the recorded CTI in islanding mode was in the range between 0.2 and 0.4582 s.

A. PERFORMANCE OF PROPOSED COORDINATION SCHEME WITH DIFFERENT FAULT LOCATIONS & TYPES
The results ensure the capability to get a complete coordinated protection system using ROCOV relays. As an example, a three-phase fault is applied at F3 in grid-connected mode, the calculated ROCOV value using (1) at the primary relay R5 is 5 × 10 6 Volt/s and at the backup relay R1 is 2.2 × 10 6 Volt/s. Based on the relays settings (TDS and pickup value) mentioned in Table 2, the relays operating time using (2) are calculated as 1.9359 s and 2.2534 s for primary and backup relay respectively, while the CTI is above 0.2 s.
The performance of the proposed coordination system is also examined for different fault types. For brevity, Table 5 shows a sample of the primary/backup relay operating times in case of different symmetrical/unsymmetrical fault types (2L-G, 3L, 1L-G, L-L) and locations in both grid connected and islanded modes of operation while using the proposed relay. The results ensure the capability to get a complete coordinated protection system using ROCOV relays (all the  CTI values are above 0.2 s) in case of different fault types and locations.

B. PERFORMANCE OF PROPOSED COORDINATION SCHEME WITH DIFFERENT TOPOLOGIES
The system operating conditions are likely to undergo frequent changes because of dynamically changing loads and generations. Further, topological changes can be caused by scheduled outage (for maintenance purpose) of any line or distributed generators from the live network. All these changes in the system severely affect any proper coordination using traditional DOCRs and clearly decrease the overall system reliability [35].
The performance of the proposed coordination scheme is examined with different scenarios in both grid-connected and islanded modes under different short circuit levels. The changes in network topology are simulated as described in Table 6.
For each scenario, three phase faults are applied on different feeders in the studied network (F1 to F13), and therefore a sum of sixty five fault cases are extensively investigated. The demonstrated results in Table 7 compare the calculated CTI using traditional DOCRs with the calculated CTI using proposed ROCOV relays for all tested fault locations in the aforementioned scenarios. According to the topological changes in tested scenarios, some relays are cancelled from the network in some fault cases and accordingly the corresponding CTI are excluded in Table 7, as indicated by the sign ''-'', e.g. CTI between R32 and R3 in Scenarios 1 and 2 for fault F2.
The results show that the traditional DOCRs failed to keep the protection system coordinated in many fault cases (CTI is less than 0.2 s) as illustrated in all shaded cells in Table 7. In some cases with traditional DOCRs, the operating time of backup relay was less than the operating time of primary relay and hence the backup DOCR acts before the primary DOCR.
The CTI in such cases got a negative value and was indicated in brackets in shaded cells, e.g. between R8 & R4 for F2 in Scenario 2.
As an example, where the line 11 is disconnected from the network in Scenario 3, the CTI for all 13  In Scenario 4, the total short circuit level is increased due to adding two DGs to the network in grid-connected mode, miscoordination cases are expected based on traditional DOCRs as between R4 & R1 for F1 (backup relay R4 acts before the primary relay R1) and between R6 & R2 for F2 (the primary relay R2 acts before the backup relay R6, but the CTI is less than 0.2 s), while the CTI for all fault locations based on ROCOV was above 0.2 s.
In Scenario 5, for islanded mode, two DGs are added to the network. Many miscoordination cases are noticed with DOCRs (conventional scheme) or with FCLs while the CTI was above 0.2 s for all fault locations based on ROCOV relays. Samples for this miscoordination cases are: -The miscoordination between R24 & R26 for F12 when using traditional DOCRs. The backup relay R26 operates before the primary relay R24. -The miscoordination between R14 & R16 for F7. The primary relay R14 acts before the backup relay R16 when using traditional DOCRs, and the CTI is less than 0.2 s.
For better illustration, the number of miscoordinated pairs for DOCRs and ROCOV relays in each tested scenario is summarized in the last row of Table 7. As clearly shown, the proposed scheme based on ROCOV relays is kept successfully coordinated for all the primary/backup relays, considering different operating scenarios. As shown, all CTI are equal or greater than 0.2 s for different fault locations, different short circuit levels and different network topologies. Thus, the achieved results have proved the selectivity feature of the proposed scheme.
It is noteworthy that the changes in the network topology or mode of operation affect the measured voltage at different buses slightly, unlike the effects on the fault current flowing in the network feeders. This adds a positive feature to the proposed scheme since it will not be affected by the variation in the network topology or changes in the mode of operation.

V. COMPARISON BETWEEN THE PROPOSED SCHEME AND OTHER TECHNIQUES
To further test and evaluate the performance of the proposed coordination scheme using ROCOV relays versus other existing techniques, a comprehensive comparison is carried out against the following schemes: -The conventional protection scheme using traditional DOCRs with two settings for grid-connected and islanded modes [6], -The conventional protection scheme using traditional DOCRs integrated with FCLs [5].
Such comparison will cover the effect of adding extra synchronous-based DGs, the effect of using inverter-based DGs and the effect of applying high fault resistance as will be presented in the following sections.

A. EFFECT OF ADDING EXTRA SYNCHRONOUS-BASED DGs
In this section, a comparison between the conventional protection schemes and the proposed scheme in terms of CTI is presented in Fig. 4    with the CTIs equal or greater than 0.2 s for all tested fault locations.

B. EFFECT OF USING INVERTER-BASED DGS
To further evaluation of the proposed scheme, the three synchronous-based DGs (each is 2.4 MVA) at buses 1, 3 & 5 in grid-connected are replaced by three inverter-based DGs, each is 0.1 MW. The inverter based DGs are selected with very low power rating of 0.1 MW DG size to simulate a very low fault current contribution case. The data for these added DGs is in [36]. Three phase faults are applied at all system feeders (F1 -F13) as indicated in Fig. 3

. The resulted
CTIs calculated by the proposed ROCOV relays are compared against traditional DOCRs in Fig. 5.
The miscoordination cases are recorded between primary and backup relays for both conventional DOCRs schemes without/with FCL. The results show that 39.6% and 62.2% of the total number of pairs are miscoordinated for conventional schemes without/with FCL respectively. Again, the proposed coordination scheme is kept successfully coordinated for all the primary/backup relays. It is concluded that, the proposed scheme has high performance with any type of DGs: inverter-based DGs or synchronous-based DGs. VOLUME 9, 2021

C. EFFECT OF HIGH FAULT RESISTANCE
An efficient protection scheme must be sensitive enough to detect high resistance faults, which may have current magnitudes close to normal magnitude values. The conventional overcurrent and under voltage relays may not be able to detect such faults [37]. Refer to Fig. 3, a three-phase fault is simulated at F1 with a fault resistance of 30 ohms at line 1. The conventional overcurrent and under voltage relays couldn't  detect such fault as illustrated in the first and second graphs of Fig. 6 since the measured fault current is less than the pickup current and the voltage almost does not change. On the other hand, the proposed primary ROCOV relay R2 was able to detect such fault. The ROCOV value has exceeded the pickup value as shown in the third graph of Fig. 6. The previous case ensured the effectiveness of the proposed protection scheme using ROCOV relay over the conventional techniques based on the current or voltage even. The results also proved the sensitivity of the proposed technique.

VI. PROPOSING USER-DEFINED CHARACTERISTICS FOR MICROGRID PROTECTION COORDINATION USING ROCOV RELAYS
Furthermore, the ROCOV relay parameters that define characteristics shape (A and B) can be optimized. In this section, the proposed coordination strategy will consider the four settings for relay characteristics (TDS, pickup, A and B) as continuous variable settings that can be adjusted to achieve coordination. The coordination problem is reformulated as a nonlinear programming problem where the main objective is to minimize the overall time of operation of relays during primary and backup operation considering faults at different locations. The proposed approach is also verified by MATLAB simulation on the same modified IEEE 14 bus system embedded with DGs previously shown in Fig. 3.
In addition to considering the constraints of Equations (5)-(6) for minimum and maximum pick-up and TDS settings (previously discussed in Section II), the following constraints VOLUME 9, 2021  are also considered for maximum and minimum values of A and B as follows: For the A and B constants; it has been chosen to have a minimum value of 0.01 and a maximum value of 80 and 2 respectively which represent the standardized values of the IEC 60255 standard for the very inverse time-current relay characteristics.
Accordingly, the achieved optimum four settings of relays using the proposed protection scheme for both grid-connected and islanded modes of operation are presented in Table 8 and Table 9 respectively.
For brevity, Table 10 and 11 list a sample of relays' operating time for different fault locations upon applying the optimized four settings for both grid-connected and islanded modes of operation respectively.
In the grid-connected mode of operation, the total operating time of all 33 relays was 97.1608 seconds, while in the island mode; the total operating time of all 32 relays for different faults was 77.3728 seconds. The results confirm that it is possible to obtain a shorter operating time for the protection relays through the optimized four settings while maintaining relays coordination. As an example, when a three-phase fault is applied at F3 in grid-connected mode, the calculated ROCOV value at the primary relay R5 is 5 × 10 6 Volt/s and at the backup relay R1 is 2.2×10 6 Volt/s. Based on the relays four settings (TDS, pickup, A and B) mentioned in Table 8, the relays operating time are estimated by 0.5442 s and 1.0599 s for primary and backup relay respectively, while the relays operating time based on only two optimized settings are estimated by larger operating time of 1.9359 s and 2.2534 s for primary and backup relay respectively.
Moreover, to evaluate the achieved settings, same scenarios in Table 6 (that simulate the change in short circuit level due to connection, disconnection of DGs) are applied to widely test the four settings in both grid-connected and islanded modes. Thus, sixty-five fault cases are simulated in the modified IEEE 14 system in both grid-connected and islanded modes of operation. For brevity, a sample for the performance of the proposed protection scheme based on ROCOV relays with four optimized settings in terms of CTI is examined as shown in Table 12. The results ensure proper coordination as the CTIs are equal or greater than 0.2 for different fault locations in all tested cases.

VII. CONCLUSION
This paper proposes a new microgrid protection scheme that is capable of operating in both grid-connected and islanded modes based only on local measurements. The protection coordination scheme depends on calculating the rate of change of fundamental voltage to detect different fault types and to estimate the proper operating time for all primary and back up relays in a meshed network.
The main contribution of the proposed coordination scheme can be summarized as follows: -The proposed scheme can identify the faulty zone accurately and guarantee robust coordination between primary and backup relays in both grid-connected and islanded modes. -It is robust against the change in short circuit level or change in network operating conditions. -The proposed scheme depends only upon locally available information which means it is more reliable and dependable than those that depends upon the information at the remote ends. -The proposed scheme does not require high sampling frequency. Actually, low sampling frequency in the range of 1-20 kHz can be applied to implement the proposed scheme. -The full coordination scheme can be achieved by optimizing two or four settings of relay characteristic. -Simulation results show the superiority of the proposed coordination scheme, in the presence and absence of DGs (with inverter-based DGs or with synchronousbased DGs), over conventional well-known DOCRs coordination scheme or using FCLs. -The proposed protection scheme can maintain the coordination between primary and backup relays for different fault locations, types and different topologies. -The results also prove the sensitivity of the proposed scheme compared to overcurrent and under voltage relays for high impedance faults.
Finally, as relays' manufactures can implement ROCOV relay as a new digital relay (or implement the idea as an additional feature to existing digital relays), the operators of microgrids and distribution networks can apply the proposed coordination scheme to estimate proper settings for gridconnected or islanded modes.