**1. Introduction**

According to the CIGRE C6.22 working group definition, microgrids are electrical distribution systems containing loads and distributed energy resources (DERs) like distributed generators (DGs) (renewable/nonrenewable), energy storage devices or controlled loads that can be operated in a controlled and coordinated way either while connected to the main power network or while islanded [1]. Microgrids can be classified as either AC microgrids, DC microgrids or AC/DC hybrid microgrids, each having their own advantages, limitations and challenges, as described in [2]. The technical challenges of AC microgrids can be broadly divided into two main categories: control challenges and protection challenges. The protection challenges can be further divided into two categories according to operational modes of the AC microgrid: grid-connected mode and islanded mode protection challenges.

When the AC microgrid is operated in the grid-connected mode, a large magnitude of fault current (ten times the full-load current or more) is available from the main grid in order to activate the overcurrent protection devices within the AC microgrid. When the AC microgrid is operated in the islanded mode, a very low magnitude of fault current is available from DGs within the AC microgrid, and hence, overcurrent devices with a single setting become insensitive. The consequences

are the miscoordination of overcurrent devices, longer tripping delays and even no trips at all during di fferent fault situations. The magnitude and duration of the fault current is mainly limited by the control of the converter-based DGs within the AC microgrid, which can be overcome by an additional fault-current source (FCS), like an energy storage device with high short-circuit capacity, and thus, single-setting overcurrent devices will become e ffective. However, the connection of an additional FCS will make the protection scheme unreliable due to dependence upon the single energy storage device. Moreover, the connection of many such FCSs will make the scheme quite expensive [3]. Another alternative approach for using only single-setting overcurrent devices can be the limitation of the fault current from the main grid or directly coupled DGs using fault-current limiters (FCLs) in grid-connected mode and using lower fault-current trip settings, which can also work in islanded mode with reduced short-circuit currents. This approach causes overcurrent devices to be more sensitive in grid-connected mode and prone to nuisance tripping during transient events [4]. The huge di fference of the fault-current magnitude and duration in grid-connected and islanded mode calls for adaptive protection schemes for the AC microgrid.

The adaptive protection schemes can be only overcurrent-based [5] or a combination of overcurrent-based and unit type (current di fferential) or based on other new protection methods like traveling waves-based [3]. The adaptive overcurrent protection necessarily requires such overcurrent devices that provide the flexibility for changing the tripping settings like numerical overcurrent (OC) relays with several setting groups [5]. The overcurrent schemes can be used more e ffectively in AC microgrids with the majority of directly coupled DGs (synchronous generators) compared with only the converter-based DGs, since the latter provide very limited fault currents for a very limited duration of time. Another reasonable adaptive approach is to use only the overcurrent protection scheme in the grid-connected mode and other protection schemes like directional overcurrent, harmonic content-based, voltage-based, symmetrical component-based, etc. in islanded mode for the AC microgrid with the converter-based DGs, with all functions included in a single protection device called the IED (intelligent electronic device). However, the protection schemes proposed for the islanded mode are not e ffective in every fault situation, and the majority of them need high-speed communication to remain e ffective [4]. Finding a suitable and cost-e ffective combination of di fferent effective protection schemes for the islanded mode with the converter-based DGs to work as primary and backup protection in a coordinated manner is still a huge challenge. An adaptive protection can be implemented either in a centralized manner by using a microgrid central controller to change the active-group settings [5] or in a decentralized manner in which IEDs in the microgrid change their own active-settings groups by receiving a trip-signal/breaker status from another IED or circuit breaker. The centralized adaptive protection scheme necessarily requires a redundant microgrid central controller to maintain a certain level of reliability. For a decentralized adaptive protection scheme, the IEDs must be equipped with the required intelligent agents and logics in order to perform various functions in an autonomous manner using the available information (data/measurements/signals) both locally and remotely.

Previously, the adaptive protection for the AC microgrid using centralized protection and communication architecture was proposed in [5–7]. An adaptive overcurrent protection for microgrids using inverse-time directional overcurrent relays (DOCRs) was presented in [8]. In this paper, artificial neural networks (ANNs) at the central human machine interface (HMI) or data concentrator are implemented for the detection and location of the faults. The protection coordination of OC relays using the linear programming approach is presented for the radial and looped configurations of microgrids in both the grid-connected and islanded modes. An adaptive protection combined with machine learning for medium voltage (MV) microgrids was reported in [9]. The proposed methodology requires a database available beforehand, which has been obtained through simulation in this research. Then, using the data mining methodology, the meaningful information is extracted quantitatively from the database. The ANN is used for fault detection and support vector machine (SVM) for fault location. The proposed method also requires relay settings calculations and recordings in the control center or relays beforehand. Moreover, the proposed scheme may generate inaccuracies in the case of data corruption, and therefore, additional countermeasures will be required. A new adaptive protection coordination scheme based on the Kohonen map or self-organizing map (SOM) clustering algorithm was proposed recently in [10] for the inverse-time OC relays. In this paper, the protection coordination is improved gradually in the three phases of the proposed algorithm, namely conventional, clustering and sub-clustering phases. The proposed method uses digital OC relays with four setting groups. The performance of the method was presented in terms of the total miscoordination time (TMT) index using a modified IEEE 33-bus network with two synchronous generators and two electric vehicle (EV) charging stations. A decentralized adaptive protection scheme using DOCRs, teleprotection and a fuzzy system in real time was proposed in [11] for the transmission system. In this paper, the transient stability constraint satisfying the maximum operating time of DOCRs was considered. Due to the dynamic adaption of the fuzzy system to the changing system conditions, the actuation time of relays was decreased, keeping the stability and coordination intact. An optimal overcurrent relay coordination in the presence of inverter-based wind farms and electrical energy storage devices was presented in [12]. In this paper, the optimal protection coordination of inverse-time DOCRs with varying load demands and changing unit commitments of DGs is presented using mixed integer nonlinear programming. A hybrid particle swarm optimization-integer linear programming (PSO-ILP) algorithm was suggested recently in [13] for the proper coordination of OC relays by suggesting proper settings groups for the changing network states. The adaptive di fferential protections for wind farm-integrated networks were reported in [14,15]. However, the di fferential protection inherently cannot provide the backup protection, and usually, the time-coordinated overcurrent protection is used as the backup protection.

The modeling of the inter-substation communication based on the IEC 61850 standard was presented in [16] for the di fferential protection (Sampled Values (SV) messages) and in [17] for the distance protection (generic object-oriented substation event (GOOSE) messages). In both [16] and [17], the virtual simulated communication networks were used based on a non-real-time tool called the riverbed modeler network software. In both references [16] and [17], the tunneling communication mechanism between substations was used for the di fferential and the distance protection functions, respectively. In [16], it was evaluated that the dedicated fiber optic network link had better performance in terms of the end-to-end delay of SV and GOOSE messages compared with an asynchronous transfer mode (ATM) link and synchronous optical networking (SONET) links. It was concluded in [17] that the links with lower bandwidths were not suitable for long distances; however, a more accelerated distance protection can be implemented, even with lower bandwidth links, compared with the conventional distance protection scheme. An adaptive protection system based on the IEC 61850 for MV smart grids was presented in [18]. In this paper, the dynamic publisher/subscriber reconfiguration of the protection devices for the implementation of the advanced fault location, isolation and service restoration (FLISR) was suggested. Since, the remote changes of the IED settings are not supported by the current versions of the IEC 61850 standard, therefore, the change of the operational settings after the network reconfiguration was suggested using the exchange of MMS (manufacturing message specification) messages with IEDs. Additionally, the logic selectivity was proposed to support remote changes of GOOSE communication schema without interrupting the FLISR operation. A mixed-layer 2/3 approach was also suggested in the paper to support both the MMS and the GOOSE implementations for the field demo of an Italian MV network. A detailed survey of di fferent adaptive protections of microgrids was presented recently in [19]. For a further detailed review of di fferent microgrid protection schemes, their challenges and developments, the recent review articles [20–23] are suggested, in addition to the previous review article [4] by the authors. For further information related to IEC 61850-based substation automation systems and related issues, the recent literature survey done in [24] is also recommended.

Based on the recent literature review presented above, it was found that less literature is available for the role of IEC 61850 standard-based communication in the protection coordination of the AC microgrids with decentralized protection and communication architecture. Moreover, a low-voltage ride through (LVRT) capability with reactive power support from the converter-based DGs in the case of AC microgrid faults has rarely been used for adaptive protection. The high risks of communication link failures and unacceptable and unpredictable communication delays are still the limiting factors to use communication links for high-speed protection functions. However, the use of a communication link is inevitable for protection functions like transfer trips from the breaker/IED at the point of common coupling (PCC) to another breaker/IED within the AC microgrid for loss of mains detection and changing preplanned active-groups settings/functions during the transition from the grid-connected to island mode for deactivating sensitive anti-islanding protection during faults and for reverse interlocking schemes. In this paper, the main focus is to discuss how an IEC 61850 communication can be applied for a decentralized preplanned adaptive overcurrent protection in a radial AC microgrid. Additionally, the DGs with LVRT capability and reactive power support in islanded mode are considered in order to implement the adaptive overcurrent protection.

The rest of the paper is organized in a manner that Section 2 presents adaptive protection based on the IEC 61850 communication standard by explaining a generalized architecture of the adaptive AC microgrid. Section 3 gives a case study background of the adaptive protection of a radial AC microgrid, explaining GOOSE (generic object-oriented substation event) message delays (transfer time) for IED to IED communication for di fferent functions, the schematic diagram of radial the AC microgrid and adaptive protection settings of di fferent IEDs. Section 4 explains the details of the proposed adaptive protection methods and results for both the grid-connected and islanded modes of operation. Additionally, the control of DGs and the LVRT capability of DGs are also explained in this section. Section 5 gives a brief discussion about the previous methods, the contribution of the research presented in this paper and what is needed for the practical implementation of the proposed method in the future. Section 6 provides the conclusion of the paper.

#### **2. Adaptive Protection Based on IEC 61850 Communication Standard**

An adaptive protection is necessarily required for AC microgrids due to changing operational modes (grid-connected and islanded), due to the formation of controlled islands due to faults within the AC microgrid, due to intermittent DGs and periodic load variations and due to the economical operations of the AC microgrid [4,25]. An adaptive protection is defined as an online activity that changes to the preferable protection device response for modified system conditions or requirements. An adaptive protection is normally automated, but some timely human interventions may also be included. Adaptive relay is a relay that includes various settings, characteristics or logic functions capable of speedy online modifications by means of externally generated signals or control actions [26]. The modern intelligent electronic devices (IEDs) not only provide various protection functions (overcurrent, over/under voltage, etc.) integrated in a single physical device but, also, o ffer various setting groups for each of the available protection functions. The various setting groups of the protection functions can be modified or altered in an adaptive manner using the communication link between IEDs and IEDs and circuit breakers (CBs). Recently, the popularity of the IEC 61850 communication standard for application in electric power substation automation has increased considerably due to its promise of interoperations among IEDs from di fferent manufacturers. The initial focus of the standard is on communication between IEDs within a single substation, but its extension for communication between several substations in the future is possible. The IEC 61850 standard explains the standardized structures for the data model and definitions of rules for the exchange of these data. IEDs from di fferent manufacturers that comply with these standard data model definitions can then communicate, understand and interact with each other [26]. The IEC 61850 standard as a common protocol enables the integration of all protection, control, measurement and monitoring functions [27].

The generalized architecture for adaptive AC microgrid protection based on the IEC 61850 communication standard is depicted in Figure 1. The IEC 61850 communication architecture for adaptive AC microgrid protection can be subdivided into three levels: process level, bay level and substation level. At the process level, the electrical parameters measurement data (MMXU) from the voltage and current sensors (VTs and CTs) and status of the circuit breakers (XCBR) inside the AC

microgrid will be collected and digitized by merging units (MUs). At the bay level, the IEDs for lines, DGs and loads of the AC microgrid will collect the digitized measurement data (MMXU) and circuit breaker status signals (XCBR) from the process bus. Each MU will publish data to process the bus, and each IED will subscribe to the respective published data from the process bus. Each of the line, DG and load IED will use measurement data (MMXU) from their respective MU for performing the active protection function like overcurrent protection in the case of faults. The status signal of the circuit breakers inside the AC microgrid (XCBR) will be used by each adaptive IED to change the active setting groups of the protection function in the case of a fault inside the AC microgrid in islanded mode. Moreover, a XCBR signal can also be used for the transfer trip of nonadaptive IEDs that are unable to detect the faults within the islanded AC microgrid. All IEDs at the bay level will also receive the status signal (XCBR\_pcc) from the circuit breaker at the point of common coupling through the station bus at the substation level. The status signal from the PCC breaker (XCBR\_pcc) will be used by each adaptive IED within the AC microgrid to change the active setting group of the protection function from grid-connected mode settings to islanded-mode settings and vice versa. The signal (XCBR\_pcc) can also be used for the detection of the loss-of-mains event by DG IEDs and to deactivate the sensitive loss-of-mains protection functions in order to maintain stability and reliability of supply within the AC microgrid during the transition from the grid-connected to islanded mode. The station bus at the substation level will provide primary communication between the various logical nodes of IEDs. In other words, all IEDs at the bay level will communicate and share data/information (MMXU, XCBR, and XCBR\_pcc) with each other using the station bus. At the station bus, a remote access point will also exist to share data with remote clients (for wide-area measurement and protection, etc.) through a wide-area network (WAN). This remote access point will implement security functions like data encryption and authentication for all data transfers and, thus, will unburden the individual IEDs to perform these tasks.

For an adaptive OC protection, the coordination between the control and protection of the AC microgrid will also be required, and control action will be required first, followed by protection action. In the grid-connected mode, a high fault current from the grid will be available, so depending on the protection settings of IEDs, it may be required to limit the magnitude of the fault current by the activation of FCLs, and in the islanded-mode with converter-based DGs, the enhancement of the fault current magnitude may be required by the activation of additional FCSs. The numerical results presented in [28] indicate that a majority of the photovoltaic (PV) inverters contribute a fault current of 200% or less for a duration of only an initial half-cycle and 110% of the rated current for an additional duration of 10 cycles or less. It is mentioned in [29] that the grid-connected converters can feed fault currents of 1.1–1.5 p.u. of their nominal currents. It should here be noted that extra FCSs like batteries, flywheels or supercapacitors with quick response times (≤10 ms) [30] will either be necessarily required to support some type of DGs like photovoltaic DGs for providing standard LVRT capability or extending the LVRT duration of other types of DGs like wind turbine generators (WTGs) for proper protection coordination if the WTG is not capable of providing LVRT. The results presented in [31] show that a wind turbine of 1 MW can provide a fault current of magnitude equal to 120% of the rated current for seven cycles of supply frequency. This duration of seven cycles with a 50-Hz supply frequency is approximately equal to the initial duration of 150 ms after fault in the LVRT characteristic of the German BDEW-2008 standard [32]. Although the duration of 150 ms looks su fficient for the maintenance of proper protection coordination between two successive IEDs within the AC microgrid, assuming high speed communication with 3–10-ms one-way fast trip message transfer as per the IEC 61850 standard and high-speed circuit breakers (one-cycle operation). However, in some cases like data loss in the transmission channel, the retransmission of the message is required, which will result in an additional delay. Moreover, the coordination between various IEDs for breaker failure protection may be required. In such situations, the extension of the initial duration after the fault in the LVRT curve beyond 150 ms will be required, and hence, additional FCSs (flywheels or supercapacitors) will be required. In addition to that, a redundant communication and redundant synchronization clock architecture will be required to cover the communication link and synchronization clock failures as recommended in [33].

**Figure 1.** Adaptive AC microgrid protection based on the IEC 61850 communication standard. HV: High Voltage, WT: Wind Turbine, ESS: Energy Storage System, BES: Battery Energy Storage, FES: Flywheel Energy Storage, MMXU: Measurement, XCBR: Circuit Breaker, CT: Current Transformer, PT: Potential Transformer (VT: Voltage Transformer), CLK: Clock, PIOC: Instantaneous Overcurrent Protection, PTOC: AC Time Overcurrent Protection, PTOV: (Time) Overvoltage Protection, PTUV: (Time) Undervoltage Protection, PTOF: Overfrequency Protection, PTUF: Underfrequency Protection.

In this paper, the main focus was given to the adaptive OC protection using fault contributions from DGs with LVRT capability, particularly in the islanded mode of the AC microgrid. Hence, the control of DGs is not discussed in detail, except a few control actions for maintaining the voltage and frequency at the islanded sections, as explained in Section 4. Moreover, the loads and generation are considered balanced in islanded mode of the AC microgrid. The same is true even for the islanded MV and LV (low voltage) sections of the AC microgrid. The paper is limited to single fault events (three-phase short-circuit faults only) during the grid-connected and islanded modes with smooth transitions to islands. However, the method presented can be extended to other types of faults. In this paper, it is not considered how the islanded AC microgrid is reconnected back to the main grid after the removal of the fault events, which is mainly related to resynchronization procedures and not directly related to AC microgrid protection. Considering the previous research, the fault contributions from DGs (both PV and WTG) are taken as 1.2 p.u. or 120% of their rated nominal currents for a duration of 150 ms after the fault. During the initial fault duration of 150 ms, the active, passive and other islanding detection and protection schemes are considered normally interlocked and can be activated quickly after the loss of communication. This means that the anti-islanding protections like under-voltage protection at DG locations can be set by default to only detect fault conditions but not trip, and DGs start providing fault currents instantaneously according to the LVRT characteristics. The trip-blocking signal to anti-islanding protection can be sent additionally from an IED at PCC after a fault is detected on the main grid side; it should be done as fast as possible and within 3 ms after fault detection, as per the IEC 61850 standard. In this paper, the term "adaptive IED" mainly refers to the communication-assisted definite time overcurrent (DTOC) relay with two preplanned setting groups: higher setting group for the grid-connected mode and lower setting group for the islanded mode of operation. The case study of a typical radial AC microgrid equipped with adaptive DTOC relays and DGs with LVRT capability is presented in the next section.
