Harmonic Content-Based Protection Method for Microgrids via 1-Dimensional Recursive Median Filtering Algorithm

Abstract

Microgrids (MGs) offers grid-connected (GC) and islanded (ID) operational modes. However, these dynamic modes of operation pose different microgrid protection challenges. This paper presents a new protection method for MGs based on a discrete one-dimensional recursive Median filter (1-DRMF). In the first step, the 1-DRMF is applied on a measured current signal on every single phase individually for targeted feature extraction. Then, the median filter deviation (MFD) and the selected harmonic distortion (SHD) are computed from the current signals of all phases independently. In the second step, the upsurges in the MFD and the SHD of all phases are cross-checked with the pre-established threshold value of 0.3 to identify and categorize fault incidents. Finally, the directional properties of three-phase (3-p) reactive energy are employed in order to pinpoint the faulty line section (LS). Many simulations were executed on MATLAB/Simulink to validate the sustainable performance of the established method. Results prove that the scheme can detect, classify, and locate the solid and high impedance faults (HIF) in the GC as well as the ID modes under radial and meshed scenarios.

1. Introduction

1.1. Problem Statement and Motivation

Distributed generation based on renewable energy resources (RERs) is an integral part of today’s modern power networks at the user end to form MGs with control-protection systems, loads, and distributed RER generations [1,2]. MGs are reliable and sustainable entities capable of overcoming today’s issues of reliability, energy scarcity, power quality, and climate change. Furthermore, under normal conditions, MGs operate in GC mode in parallel with the utility grid, but in the case of faulty conditions, they must operate in ID mode to reduce power quality and reliability concerns. However, their power flow is bidirectional, and the fault current is very high in GC mode while it is very low in ID mode [3]. Due to the rapid penetration of intermittent RERs, the dual-mode microgrid operation, and the low level of fault current in ID mode, the microgrids’ scheme designing becomes a great challenge [4].

Since conventional protection methods fail to provide safe and sustainable operation of microgrids, the purpose of this research is to design such a protection scheme that is capable of protecting a microgrid under all topologies and operational modes. Moreover, the scheme provides high reliability without any non-detection zone (NDZ) and is robust against false tripping [5,6].

1.2. Literature Review

Great efforts have been made in prior research to establish a reliable, sustainable, and secured protection approach for MGs. In ref. [7], over-current relays with distinctive relaying features were suggested to build a highly sensitive microgrid protection strategy. The current signal in ref. [8] was pre-processed using time-time transform (TTT) to calculate TT-matrixes. Moreover, a z-score vector was introduced as a fault index that was retrieved from current signals at both sending and receiving sides. This index, which represented an absolute value of the TT-matrix, was then used to detect and identify faults in the MGs. S. Baloch in [9], introduced a low-pass filter and/or square law technique for determining the current wave envelope. The variance in the autocorrelation function was generated for fault event detection. In addition, the directional characteristics of reactive energy were used for zone identification in an AC microgrid. S.B.A. Bukhari et al. [10] calculated the superimposed reactive energy (SRE) factors from the Hilbert transform. This SRE factor was then cross-matched with the system-dependent threshold level for fault detection and/or phase identification in the microgrids, while the directional characteristics of the SRE were utilized for fault localization in the microgrid. Wavelet transform (WT) for feature extraction was used in ref. [11]. Fault detection and/or classification was achieved through variations in the index for looped-microgrid topology. The authors in [12] also used WT for feature extraction, however, digital relaying and communication assistance were engaged in the established scheme for the protection of the MGs. In ref. [13] the authors suggested a communication-assisted, high-frequency-based approach for microgrid protection. A novel data-driven approach was proposed in ref. [14] for fault detection and location in microgrids.

Short-time Fourier transform (STFT) was implemented to extract features from the voltage signal and then the decision tree algorithm was applied for exact fault detection, classification, and location in ref. [15]. The authors of ref. [15] suggested a voltage-based protection scheme for MGs through active power calculation and sensitivity analysis that also depended on STFT. In ref. [16], the author presented a fault detection method through voltage-frequency controlled inverter-based distributed RER for inverter-dominated AC microgrids. A protection scheme for low voltage AC MGs based on active power flow, voltage sag, and current magnitude was suggested by J. Octavio in [17]. Many adaptive protection approaches for microgrid protection were reported in ref. [18,19,20,21]. In ref. [22], the authors suggested a novel protection strategy for inverter-based microgrids that depended upon injecting an off-nominal frequency. The authors of ref. [23,24] utilized a Kalman filtering-based signal processing tool for the protection of microgrids from different kinds of faults in GC and ID modes.

Some intelligent protection methods are also reported in the literature. In ref. [25] Gashteroodkhani proposed a TTT and deep belief network (DBN) based technique for microgrid protection. In ref. [26], it was proposed to use state observer-based fault detection and/or classification and localization of fault in MGs with recurrent neural network assistance. Refs. [27,28] used the CNN-based intelligent protection strategy for a microgrid. The author of ref. [29] proposed a type-2 fuzzy logic approach to recognize and/or classify, and then localize microgrid faults. The author introduced the support vector machine (SVM) model for detecting faults in microgrids [30]. These faults were then located based on harmonic injection that fed to the multi-class SVM. The ordinary overcurrent relays with intelligent electronic devices were used in ref. [31] to act as an agent within a multi-agent framework to protect the loop-based MG. A. Anand presented a differential protection scheme in ref. [32], which was based on ensemble empirical mode decomposition (EEMD) for AC microgrids.

The foregoing schemes solve some of the microgrid protection problems in many aspects. However, these microgrid protection schemes have many limitations as follows:

• The majority of given schemes cannot protect the microgrid in various scenarios;
• A large number of the protection strategies were only designed for a radial scenario of microgrids and do not cover looped or meshed scenarios;
• HIF identification is crucial in any reliable protection scheme. However, only a few schemes have the potential to deal with HIF faults;
• Some of the methods were very costly due to the deployment of high-cost protection devices;
• A few schemes were very complex and had a very high computational burden.

1.3. Significant Contributions

In this study, a well-known discrete signal processing tool named 1-DRMF was implemented for the protection of multiple distributed generation-based microgrids. In the first stage, a three-phase current signal was pre-processed through 1-DRMF to compute the MFD and SHD of each phase separately. The MFD and SHD fluctuations were then matched to pre-defined threshold levels to detect and classify fault occurrences in the microgrid. In the second stage, the 3-p reactive energy was calculated from the third, fifth, and seventh harmonic components of current-voltage. Furthermore, the directional characteristic of this 3-p reactive energy was used to recognize the faulty zone in the microgrid. The suggested protection strategy was inherently phase-segregated. The MATLAB/Simulink software package was used to perform extensive simulations to validating the performance of the proposed scheme in GC and ID modes under meshed and radial regimes. The proposed method makes the following contributions:

• Application of 1-DRMF as a time-frequency purview for harmonic analysis tool in the microgrid protection;
• Development of robust fault detection, classification, and location criteria that are effective for all types of faults;
• The presented scheme can protect microgrids against solid and HIF timely, as the scheme works satisfactorily in both main and protection;
• The developed approach applies to both GC and ID modes within radial and meshed microgrid layouts;
• The structure of the proposed strategy is very simple; therefore, the protective threshold setting is quite easy.

1.4. Paper Organization

The remaining paper is divided into the following sections: Section 2 describes the fundamental mathematical modelling of microgrid current and voltage signals, as well as the fundamental principle of the proposed 1-DRMF. Section 3 briefly explains the proposed protection scheme in detail. Section 4 illustrates the IEC microgrid test bed, threshold setting, and the extensive simulation results that are carried out to validate the proposed protection scheme. Finally, the paper is concluded in Section 5.

2. Underlying Principles of the Proposed Strategy

This section briefly explains the theoretical background and the mathematical concept of a 1-DRMF. Furthermore, this section outlines the mathematical model of current and voltage signals involved with the microgrid.

2.1. One-Dimensional Recursive Median Filter

1-DRMFs have several applications. They are predominantly utilized as a signal processing tool, as an image processing tool, and as a location estimator. Traditionally, they serve to reduce noise and generate smooth information at their output. Also, they have smaller computational complications compared with other existing signal-processing techniques. In addition, the 1-DRMF is a non-linear filter despite the fact; our microgrid system also poses a non-linear characteristic.

Table 1. Step-by-step description of the 1-DRMF operation algorithm.

Steps	1-DRMF Algo Description
Step 1; Initialize	3-phase voltage is measured at DG terminal at the nth sample with moving window (Non-tunable, Positive Integer) of Len-n.
Step 2; Median	The 1-DRMF is applied on 3-phase voltage signals to compute the median of the input signal for further third, fifth, and seventh harmonics feature extraction.
Step 3; Release	Allow property value and input characteristic changes, and release 1-DRMF resources.
Step 4; Clone	Create a median filter object with the same property values.
Step 5; Lock	Display locked status (logical).
Step 6; Reset	Reset the states of the median filter.
Step 7	Terminate.

depicts the detailed algorithm steps of a 1-DRMF.

In digital protective relaying, the finite impulse response filters with sliding data windows were used for band-pass filtering of currents and voltages measurement. Consequently, the 1-DRMF provided the most favorable feature extraction of electrical magnitudes from the set of their noisy, non-linear measurements in a very short time interval [33,34]. The 1-DRMF utilized the sliding window method (SWM) to compute the moving median.

In this method, a window of specified length moves over a specified channel sample by sample, and the MF computes the median of the data in this window. Therefore, in SWM, the output for each input sample was the median of the current sample and the (Len-1) previous samples where Len is the length of the window. To compute the first Len-1 outputs, when the window did not yet have enough data, the algorithm filled the window with zeros for dimension purposes. However, to avoid mathematical confusion statements in the workflow, the 1-DRMF pseudo-code with a window length of one is illustrated in the Appendix A [35,36].

2.2. Formulation of Microgrid Discrete Signals

During fault incidents or disturbances, a microgrid is more or less a non-linear dynamic system with balanced three-phase, non-sinusoidal, and noisy voltage- current signals. The proposed scheme employed a full-cycle sliding data window until a faulty incident was detected; however, the 1-DRMF strikes a tremendous balance between speed and accuracy.

In the established protection strategy, the 1-DRMF used a real-time measurement of the current-voltage signal as an input. However, simply the current signature was used for the detection and/or phase identification of a fault. Whereas, both current and voltage signals were used for the calculation of 3-p reactive energy based on the 1-DRMF. Implicitly, the model of the voltage signal was to be duplicated as current, which is why no separate description is obligatory. The 1-DRMF was implemented in each phase alone. Therefore, the single-phase representation of current-voltage signals of a microgrid is adequate. For measured current signal

$$i_k={\alpha }_k\cos \left({\omega }_{°}k+Ø\right)$$

(1)

Similarly, for a measured voltage signal

$$v_k={\alpha }_k\cos \left({\omega }_{°}k+Ø\right)$$

(2)

The $i_k$ and $v_k$ represents the measured current-voltage respectively at a k-th sample of a single phase. Other phases will be merely 120° from each other, while ${\alpha }_k$ and $Ø$ denotes the amplitudes of the recorded signals and noises. The ${\omega }_{°}$ is the angular frequency where ${\omega }_{°}=2\mathsf{\pi }\left(f_{°}/f_s\right)$ so, $f_{°}$ is the fundamental frequency. The trigonometric derivative of Equation (1) results in an iterative equation as follows.

$$i_{k+1}+i_{k-1}=2{\alpha }_k\cos \left({\omega }_0\right)i_k+e_k$$

(3)

where $e_k$ signifies an expected zero mean arbitrary error. The zero means represents a single-phase current signal in terms of measurement errors as well as other exogenous perturbations ($a_k$), which is given as follows:

$$I_k=i_k+a_k$$

(4)

Equation (4) was used as an input of the 1-DRMF. To keep away from ambiguous notations and for suitability of analysis, we applied the 1-DRMF in sequence to every sample of the signal, and we first interchanged each sample with the output of the 1-DRMF on that particular sample before shifting the window to the next spot. The function of the 1-D median filter for a current signal at the kth sample can be given as

$$Ⅎ{\left(t\right)}_k={\displaystyle \sum }_{k=1}^{M-1}f{\left(t\right)}_{i_k}$$

(5)

Specifically, the output of the 1-DRMF in expanded form for a current signal at the kth sample can be given as

$$Ⅎ{\left(t\right)}_k=median[ I{\left(t-l\right)}_n, I{\left(t-l+1\right)}_k\dots \dots I{\left(t\right)}_k\dots \dots I{\left(t+l-1\right)}_k,I{\left(t-l+1\right)}_k]$$

(6)

where

• $Ⅎ{\left(t\right)}_k$ = Output filtered current signal of 1-DRMF for k-th sample
• $I{\left(t\right)}_k$ = Input measured current signal of 1-DRMF for k-th sample

These 1-DRMF process equations were engaged further for detection, classification, and localization of faults in microgrids.

3. Proposed Protection Strategy

This section provides the stepwise detail of the proposed microgrid relay (MR), which was modelled in MATLAB/Simulink (2019b). The proposed MR is based on five functioning blocks, an input-signal-conditioning block, a feature-extraction block, a fault-detection-classification (FDC) block, a communication-supported zone identification block addition, with main and backup protection units, and the last tripping block. The step-by-step function of each block of the proposed scheme is clearly elaborated in Figure 1.

3.1. Input-Signal-Conditioning Block

The measured voltages and current at the corresponding buses had a lot of measurement and other random noise. Meanwhile, the signals were analog in nature. This block completed its work in two stages; initially, it acquired three-phase measured signals of current and voltage from the corresponding node. Then, these analog signals were provided to a 12-bit Analog Digital converter (ADC) having a 3.6 kHz sampling frequency. The digital signals were then provided to a Tschebyscheff second-order low-pass filter with a cut-off frequency of 1600 Hz, which was utilized for anti-aliasing purposes. These pre-processed discrete signals were then provided to the next block for both the fundamental and harmonic components-based feature extraction.

3.2. Feature-Extraction Block

The most relative and informative set of features must be employed to increase the scheme’s efficiency in terms of computational complexity, computation time, and accuracy. Therefore, the discrete current-voltage signals obtained in the previous stage were deployed as an input to the 1-DRMF for the fundamental-harmonics feature extraction, while 1-DRMF was applied to each individual phase. The extracted fundamental components of currents were further utilized in the computation of the MFD, and harmonic components of currents were used in the computation of the SHD in the fault detection-classification block. However, the extracted non-fundamental components of both current and voltage signals were used in the fault-direction block.

3.3. Fault-Detection-Classification Block

Precise and quick fault detection is necessitous in a protection scheme to keep away from false-tripping and protection-blinding problems. In the proposed scheme, the fundamental component of the input current signal and the filtered output signal of an individual 1-DRMF in each phase were used to compute the value of MFD through Equation (7). Moreover, the filtered non-fundamental third, fifth, and seventh current components were used to compute the SHD through Equation (8). Both the MFD and the SHD were used for the detection and/or classification of solids as well as HIF. To verify fault occurrence, the generated MFD and SHD were examined to a pre-defined threshold level of 0.3. If the MFD or SHD of any phase exceeded the pre-defined cut-off point, the system was regarded as faulty. The presented method for microgrid protection was phase-separated. The value of the MFD of the kth sample was computed by taking the difference of the filtered output signal $Ⅎ{\left(t\right)}_k$ from the measured input signal $I{\left(t\right)}_n$ of fundamental current components, which is formulated as

$$d{\left(t\right)}_k=I{\left(t\right)}_k-Ⅎ{\left(t\right)}_k$$

(7)

where;

$$\begin{gathered} d{\left(t\right)}_k=\mathrm{MFD} \mathrm{of} n\mathrm{th} \mathrm{sample} \\ Ⅎ{\left(t\right)}_k=\mathrm{filtered} \mathrm{output} \mathrm{current} \mathrm{signal} \\ I{\left(t\right)}_k=\mathrm{Measured\; input\; current\; signal} \end{gathered}$$

However, the SHD of the nth sample was computed from the filtered non-fundamental current component of orders third, fifth, and seventh of MF. The SHD can be expressed as follows

$$SH{D}_p=\frac{\sqrt{{{\displaystyle \sum }}_{n=3,5,7}^{\infty }{I}_{k\_rms}^2}}{I_1}$$

(8)

where

• $SH{D}_p$ = Selected harmonic distortion.
• $I_k$ = RMS value of the order of the harmonic.
• $I_f$ = RMS value of fundamental current.

The OR operation on the fault-detection-classification block output signals was applied to generate a fault detection (F_d) signal. The faulty situation was considered as a binary logic 1 and the normal situation is 0. The fault detection F_d signal was provided to the main protection unit for exact fault section identification through proper logical operation as shown in Figure 2, while the phase identification signals were provided to the fault isolation module to generate a trip signal to the concerned circuit breaker (CB) after receiving a trip (F_T) signal from either the main or the backup protection unit.

3.4. Communication-Supported Zone Identification Block

Precise fault location is crucial in shrinking the power outage time and to fast-track system recovery.

3.4.1. Direction Identification

In this study, the direction of the fault was decided based on computed 3-phase MFBRE.

$$3-{\mathrm{p} \mathrm{reactive} \mathrm{energy}}_k=\left(3v\ast i\cos Ø\right)dt$$

(9)

If the $3-{\mathrm{p} \mathrm{reactive} \mathrm{energy}}_k$ flowed in the negative direction at the fault point, the fault was considered a forward fault. However, if the $3-{\mathrm{phase} \mathrm{reactive} \mathrm{energy}}_k$ flowed in a positive direction, the fault was considered a reverse fault. In summary, $3-{\mathrm{p} \mathrm{reactive} \mathrm{energy}}_k$ was less than zero for the forward faults and more than zero for the reverse faults. The $3-{\mathrm{p} \mathrm{reactive} \mathrm{energy}}_k$ for the case of the forward fault is as follows:

$$3-{\mathrm{p} \mathrm{reactive} \mathrm{energy}}_k<0$$

(10)

However, the $3-{\mathrm{phase} \mathrm{reactive} \mathrm{energy}}_k$ for the case of reverse faults is as follows:

$$3-{\mathrm{p} \mathrm{reactive} \mathrm{energy}}_k>0$$

(11)

3.4.2. Main or Backup Protection Logic

A fault in one section can be a forward fault for many relays. So, the exact faulty zone was recognized by communicating information across different relays. In addition, communication improved selectivity and avoided nuisance tripping. A relay in a zone will identify through suitable logic that a particular section is healthy or faulty, which is explained in the next section. However, backup protection is very important in any communication-aided protection scheme. The efficiency and reliability of the proposed scheme can be enhanced using a backup protection unit. In the proposed protection scheme, each relay is designed with two protection zones as shown in Figure 2:

• Main protection zone
• Backup protection zone

The MR-1 is considered a reference to understand the concepts of protection zones of any relay in an adopted microgrid. Zone 1 is the main protection zone of MR-1; it covers the whole LS-1. Main protection operates rapidly when the fault occurs in LS-1. However, MR-1 is the backup protection of Zone-2. MR-1 operates after a pre-defined time delay (T_d) when the fault occurs in this zone. This time delay is provided to allow the main protection of zone 2 to operate first.

where

T_d > (operating time of main protection relay + operating time of its corresponding circuit breaker)

(12)

In the proposed protection scheme, faulty sections are identified by creating a communication link between three adjacent relays. Each relay not only sends but at the same time receives the fault detection (F_d) and fault direction (F_dir) signals from the other two adjacent relays. This confirms whether a fault is detected in this relay region or not. It also confirms the direction of the fault for the relay position. Furthermore, a proper logical pattern is designed that determines the faulty section of the microgrid.

The fault zone-identification logic of MR-1 is shown in Figure 2. This shows that the MR-1 was not only receiving the f_D and the f_Dir signals from MR-2 for the identification of fault in the primary protection zone; it also received both signals from MR-4 to operate as backup protection for LS-2. If the relays at both ends of the LS-1 detected the fault as a forward fault, then the primary protection zone of LS-1 can be considered faulty and the E_mz(trip) signal will be issued. On the other hand, if the MR-1 detected the fault as a forward fault but the MR-2 detected the fault as a reverse fault, then the fault zone identification unit of the MR-1 checked the status of the MR-4 from its received signal. If MR-4 identified the fault as a forward fault, then the backup protection zone was considered faulty and the E_dz(trip) signal will be issued after a predefined time delay. Finally, OR-gate was implemented on both E_mz(trip) signal and E_dz(trip) signals, resulting in a trip f_FF signal to the isolation module. In our proposed scheme, we take a 3 s time delay to simulate the most severe scenario.

3.5. Fault-Tripping Block

The proposed scheme can inherently classify faults. Finally, an AND-operation is applied on an identified faulty phase from the FDC module and a forward fault f_FF signal from the FL module. This generates a decision of tripping by issuing a trip signal to the relevant CB of a faulty zone to clear the fault.

4. Simulations and Results

4.1. IEC Benchmark MG Test System

An IEC test bed was implemented in MATLAB/Simulink software for faults data generation in various case studies. The IEC test system has five LSs from LS-1 to LS-5. The single-line depiction of the IEC test system is illustrated in Figure 3. The considered test system consisted of six buses from Bus-1 to Bus-6. However, the main CB connected the microgrid to the main grid at the point of common coupling (PCC). Four RERs-based distribution generations were connected at different busses through four transformers named T/F-1 to T/F-4. Three of them were inverter-based RERs and one was a synchronous generator-based RER. Main Circuit Breakers at PCC were used for the operation of microgrid test models in grid-tied or islanded mode. CB loop-1 and CB loop-2, the remaining two circuit breakers, were employed to operate the microgrid in radial, looped, and meshed topologies. The test system and load parameters were taken from the reference [23,30]. The operating voltage of the MG test system was 25 KV. Six loads from L-1 to L-6 were connected to each bus.

4.2. Relay Pickup Threshold Setting

The relay pickup threshold settings prevented protection blinding and false tripping issues. The SHD and the MFD are ideally 0 in operating normal conditions. However, due to the noise in measurement input signals, the MFD and the SHD were not 0. So, a constant relay pickup value of 0.3 for the SHD and the MFD was easily calculated by running the simulations under healthy but worst system conditions. A fault was indicated by the relay if the disturbance was above the pre-specified pickup values. All types of faults were detected, located, and classified in GC and ID modes in a precise and timely manner under radial and meshed scenarios.

4.3. Tested Cases

The proposed protection scheme has been evaluated in the MG test model in this section. Several fault types were simulated in various operation modes and topological structures. For all cases in our study, a window length of 1 was used for a good estimate. From the results, it is unambiguous that the proposed scheme can protect the microgrid in different operation modes under numerous faulty conditions. The following are different case studies for the validation of the proposed protection scheme. All types of faults were tested in different scenarios, with different fault locations and along with different buses; few generalized cases are represented for clear validation of the proposed scheme.

4.3.1. Solid Faults in GC Mode

The fault current in the GC mode of microgrids was very high due to the main grid contribution. This case was considered for the validation of the presented strategy during solid 3-p faults in the GC mode. Moreover, the meshed scenario was also created by simultaneously closing switches S-1, and S-2.

Figure 4 depicts the result of a 3-p solid fault that occurred in LS-2 at 0.16 s between the PCC and Bus-3. The corresponding fault was located on 40% of LS-2′s total length. Figure 4a depicts the signature of the fault current before and after the initiation of the 3-ph fault. Figure 4b,c show the SHD and MFD for each singular phase. It is observed from the results that the value of SHD and the MFD at 0.16 s were larger than the pre-specified threshold value, therefore the proposed schemes detected and classified the fault within half a cycle successfully. Furthermore, the 3-p reactive energy at MR-56 was recorded to be less than zero as depicted in Figure 4d, which initiated the E_mz(trip) signal of the presence of forward faults in the faulty section.

4.3.2. Solid Faults in ID Mode

This case validated the proposed scheme in ID mode for solid 3-phase to ground faults. Due to the absence of the main grid, the fault current in the ID mode of microgrids was very low. As a result, conventional schemes had great difficulty in detecting fault incidents. Even so, the radial topology was taken into account by simultaneously opening switches S-1, and S-2.

The result for the corresponding 3-p solid fault is depicted in Figure 5; the fault occurred at the time instant of 0.22 s at 50% of the total length of LS-5. Figure 5a shows the signature of the fault current before and after the inception of the ABC-g fault. Figure 5b,c represent the SHD and MFD for each phase. The MFD failed during ID conditions, while the SHD values of faulty phases A, B, and C exceeded the pre-defined thresholds. As a result, the results demonstrate that the proposed schemes successfully detect and classify the fault within half a cycle. The existence of forward faults in LS-5 was revealed by the negative 3-p reactive energy at MR-10 depicted in Figure 5d.

Hence, it is summarized that in the first two cases, the proposed scheme was validated for solid faults during both GC and ID microgrid operational modes under radial and meshed scenarios.

4.3.3. HIF Faults in GC Mode

Due to the low fault current level, HIFs endanger human safety instead of electrical equipment. However, it is very difficult for conventional overcurrent protection to detect the HIFs [37].

A double phase to ground HIF with the Z_f of 50 ohm hit LS-3 at 20% of the total line length. However, the corresponding AC-g HIF was initiated at 0.25 s as depicted in Figure 6a. Hence, as it is depicted in Figure 6b,c, the SHD of the faulty phases A and C was more than the specified threshold value, while MFD failed to detect the HIF. Consequently, the reliable dual fault detection criterion efficiently detected and classified the HIF in less than a half cycle, even though the MFD was not enough to cross the threshold value. In addition, the MR-62 indicated the presence of forward faults in LS-3 due to the negative direction of 3-p reactive energy, as shown in Figure 6d.

4.3.4. HIF Faults in ID Mode

HIF detection is a challenging issue in microgrids due to the small magnitude of fault current. Similarly, the ID mode microgrids have a low level of current as well. Therefore, a separate case study is also discussed here to validate the proposed scheme when both HIF and ID situations occur simultaneously [38].

A 3-phase to-ground HI fault with the Z_f of 100 ohms arose at the LS-4 at 80% of the total line length. The corresponding ABC-g HI fault began at 0.15 s, as shown in the current signature of Figure 7a. Figure 7b shows that the SHD of faulty phases A, B, and C were greater than the specified threshold value. Hence, the scheme dual fault detection criterion detected and classified the HIF successfully in less than a half cycle, even though the MFD was not adequate to cross the threshold. Furthermore, the MR-78 indicated the presence of forward faults in LS-4 due to the negative direction of 3-p reactive energy as shown in Figure 7d.

Hence, it is summarized that in the above two cases, the proposed scheme was validated for HIF faults during both GC and ID microgrid operational modes under radial and meshed scenarios.

4.3.5. Line-to-Line Faults

Unsymmetrical line-to-line short circuit faults are also very important to address in microgrids. Therefore, the suggested microgrids protection strategy was also tested for various line-to-line fault scenarios. A double-phase line-to-line fault during the ID mode of operation is presented in Figure 8a.

Figure 8 depicts the results for the BC fault; the current signature of the corresponding figure indicates that the fault is initiated at 0.14 s. However, the SHD and MFD of phases B, and C were greater than the threshold values, as depicted in Figure 8b,c.

Accordingly, the proposed scheme efficiently detects and classifies the line-to-line fault in less than a cycle. However, the negative value of 3-p reactive energy indicates the forward fault in LS 2, as depicted in Figure 8d.

4.3.6. Single-Phase Tripping

Even though single-phase faults are the most common in power systems, the proposed technique was tested in a variety of single-phase fault scenarios. Several cases were simulated, but due to lack of space only one such case is presented in Figure 9 for the validation of the suggested scheme.

The current signature in Figure 9a indicates that a single-phase B-g fault occurred at 0.122 s. The corresponding fault was initiated at 90% of the total length in LS-1. In this case, the meshed topological structure was under consideration in the GC mode of microgrids. Hence, the results of both the faults indicate that the SHD and the MFD were greater than threshold levels as shown in Figure 9b,c. Therefore, the scheme detected the single-phase fault in less than half a cycle. However, the fault classification was by default due to the phase-segregation feature of the proposed scheme. In addition, the 3-p reactive energy in Figure 9d at relay-32 was negative, thus confirming the incidence of a forward fault in LS-1.

4.3.7. Backup Protection

Failure of the main protection unit of any scheme causes huge damage to the power system. Therefore, in the purposed scheme a backup protection unit was provided to avoid such loss. Several protection scenarios were performed in both the ID- and the GC modes to validate the proposed scheme operating during primary protection failure, although due to space limitation only two are discussed here.

The results as depicted in Figure 10 show that a single-phase A-g fault occurred in 30% of the total length of LS-3 at 0.1 s. MR-62 acted as the primary protection relay of LS-3, which might have failed to detect the fault in the corresponding section. The protection unit of MR-62 operated after a pre-specified time delay of 0.3 s. Hence Figure 10 demonstrates that the A-g fault was successfully and promptly cleared by the protection unit of MR-62.

In Figure 11, the results show that a three-phase ABC-g fault occurred in 80% of the total length of LS-5 at 0.2 s. MR-10 acted as the primary protection relay of LS-1, which might have failed to detect the fault in the corresponding section. The protection unit of MR-10 operated after a pre-specified time delay of 0.3 s. Hence, Figure 10 depicts that the ABC-g fault was successfully cleared by the backup protection unit of MR-10 on time.

4.3.8. Scheme Failure Consideration

In summary, the proposed microgrid protection method tried to solve different microgrid protection challenges in GC and ID operational modes during radial and meshed topologies including scheme failure during some 3-phase fault conditions, despite these faults being very rare in power systems. In Figure 12a–c, a scheme failure case study is considered on a 3-phase ABC-g fault at 50% of the total length of LS-3 at 0.15 s. As Figure 12 shows, the scheme failed to protect the microgrid in such conditions. Nevertheless, the scheme only failed in the rare case of a very-high impedance three-phase fault under islanded condition.

4.4. Performance Analysis

To validate the performance of the proposed work, the scheme was compared with some exciting benchmark techniques in both grid-tied and island operation conditions. Four benchmark methods –KF-based method [23], State Observer (SO)-based method [26], CNN-based method [28], and Fuzzy Logic (FL)-based method [29]—were chosen for a comparative study based on two most critical parameters, namely accuracy and speed of action, as depicted in Figure 13a,b.

5. Conclusions

Microgrids are modern power networks with RER-based distributed generation, loads, distribution networks, and their own protection and control systems. MGs are operated in DC and ID mode and provide three topological structures: radial, loop, and meshed. However, designing microgrids protection schemes is challenging in such harsh and dynamic operational conditions. Therefore, we critically need a protection scheme to protect microgrids under such harsh dynamics. In this paper, a new 1-DRMF-based microgrid protection strategy was introduced. 1-DRMFs are simple, non-tuneable, and have a very low computational burden, which makes the scheme fast. Initially, the current signal measured at any considered bus and the current signal was pre-processed through 1-DRMF. Then, dual fault detection and classification indices were computed and named as SHD and MFD. The fault in the microgrid was detected through dual check, which included MFD and SHD. MFD was computed from fundamental components of current, while the SHD was calculated through the third, fifth, and seventh harmonic components of the current signal. Because the suggested technique was phase-separated, fault classification took place by default. Secondly, the directional properties of three-phase $3-{\mathrm{p} \mathrm{reactive} \mathrm{energy}}_k$ were utilized for location identification of a fault. Additionally, the scheme provided a protection unit, making it more robust during main protection failure. Hence, it was proved from the extensive simulation results that the proposed method can protect microgrids from solid as well as HIF in both GC and ID operating modes under meshed and radial topologies. The proposed microgrid has been validated on model-based simulation and analysis on MATLAB/Simulink software, while hardware in the loop is considered for future work.