A differential amplitude variation based pilot relaying scheme for microgrid integrated distribution system

1 Introduction

Recently, the power-producing industries have boosted their investments in distributed generation (DGs) using renewable energy sources (RES) at the primary and secondary distribution levels. Through power electronics-based converters, the RESs are incorporated into the current network. The microgrid architecture, which shares many traits with the smart grid, is defined by the linkage of DGs, loads, and control frameworks with a communication system. Microgrid uses a communication network to exchange data and get accurate data for a variety of activities, including administration, control, and protection. Data acquisition uses sophisticated monitoring tools like smart meters, intelligent electronic devices (IEDs), and micro-phasor measuring units (µPMUs) [1], 2]. Reductions in the cost of generation and transmission, power loss, carbon emissions, increased dependability, and the ability to operate in both grid-connected (GC) and islanded modes (IM), etc. are some of the primary advantages of microgrids [3], 4]. However, in terms of network protection, microgrids encounter considerable difficulties.

However, high DG penetration in active distribution networks drastically alters existing network topologies, resulting in a bilateral flow of fault current and power. The typical overcurrent precautions in this situation have significant hurdles regarding both their selectivity and sensitivity. Therefore, a proper protective operation cannot be ensured. Earlier literature [5] provides a comprehensive overview of microgrid protection challenges. Due to the power-electronic converter-based DGs, low fault current contribution, the conventional over-current-based protection approaches are insufficient [6]. Therefore, an effective protection approach is required to quickly identify and disconnect faulty zones while ensuring power supply continuity to boost network dependability, resilience, and security.

Numerous protective strategies are suggested in the literature for distribution system protection. Among them, the majority of research works are focused on the development of high-impedance fault (HIF) detection schemes [7], [8], [9]. Some studies investigated single-line-to-ground (SLG) fault detection and faulty phase identification, considering the various adverse effects [10], [11], [12]. In this context, the use of differential protection methods for feeder protection is growing in popularity as a result of global advances in technology and communication infrastructure. The literature discusses differential protection strategies based on various power system parameters.

To identify the faulty branches in distribution networks, Reference [13] proposes a technique based on phase jump. The plan, though, also fails when it comes to high-impedance flaws (HIFs). Reference [14] uses a data mining methodology with a unit protection strategy based on a sequence component. The procedure is for determining the relay settings and their characteristics during fault conditions in an isolated microgrid. In [15], a data mining-based intelligent differential relaying technique is proposed for fault detection that employs both the voltage signal and the fault current to determine the decision-making parameter. However, because the earlier stated techniques rely on an optimization technique, it is nearly impossible to model them [14], 15]. Time-frequency transforms, such as the S-transform [16], are another method for fault identification. With these transforms, the difference in spectral energy of the relaying signals is acquired as the detection index. A methodology based on the Hilbert-Huang transform (HHT) for fault identification is detailed in [17]. This paper demonstrates the superiority of HHT. However, the methods have not been tested on any real-world standard system, and there is still room for improvement in the speed of defect disturbance detection. In [18], a different plan is put up that makes use of HHT and machine learning methods. The method uses the decomposition of fault currents. However, the proposed technique’s practical use is rather difficult and useless as a result. A differential technique based on positive-sequence components of current signals is suggested [19]. All of the techniques outlined in [13], [14], [15], [16], [17], [18], [19] are precise for locating low-impedance faults (LIFs), but they have drawbacks for locating high-impedance faults (HIFs). The feed-forward neural network strategy is covered in [20] for identifying and pinpointing the defect. However, the online use of this method necessitates more data set storage, and the method is also network configuration dependent.

A few methods for obtaining the logical trip signals include positive sequence impedance [21], differential positive sequence power angle (DPSPA) [22], integrated impedance [23], 24], and differential frequency [25]. In distribution grids, voltage transformers may not be accessible at every bus, which restricts the use of the aforementioned methods that call for voltage measurement [26]. In [27], a modified current differential algorithm using self-synchronisation is proposed. However, the detection of HIFs is very challenging using this method. In [28], a current differential approach is proposed with self-data synchronization. This method doesn’t require any additional synchronization instrument. In [29], a synchrophasor-assisted current differential scheme is proposed. However, methods mentioned in [28], 29] are much more sensitive to high-resistance faults beyond 500 Ω and towards HIFs. In [30], a superimposed differential energy based method is illustrated for discriminating low and high impedance faults. In [31], a current amplitude difference based protection technique is presented for active distribution networks considering the impact of DGs. A quasi theorem based pilot relaying method for the protection of active distribution networks is described in [32].

PMU or µPMU applications in distribution network protection integrated with and without microgrids have recently been proposed [33]. A high reporting rate (100 or 120 Hz), time-stamped data, the fast phasor estimation method, and the availability of hardware set-up are properties that make it easier to employ for microgrid protection [34], 35]. Using synchronized phasor data from PMUs, some experts have suggested differential protection strategies in this context.

2 Proposed method

In general, numerical relays initiate their operation after fault detection. In this regard, as seen in (1), the phase current based criterion is typically employed in numerical relays fault detection. In this study, the proposed relaying logic will initiate its operation only after fault detection criterion (1) is met. Thus, the fault starting criterion is given by,

where N is the number of samples per cycle, and i(k) is the phase current sample at time k; I _RMS stands for the typical load current’s root mean square (RMS) value. The beginning coefficient in this study is called κ, and its value is fixed at 0.1.

The proposed method suggests computing DAVI by utilizing the fault current signals accumulated at the control center by µPMUs. The fault current from each measuring point is collected and the DAVIs for each line segment are computed to find out the faulty segment hence, corresponding relays will be tripped. The single-line diagram of the two-section feeder model with relays is shown in Figure 1. The µPMUs are placed in bus-M, N & P. Line-PN is taken for further analysis. Two faults (internal and external) are simulated as shown in Figure 1. The detailed analysis and derivation of the current variation indexes for internal (F _int) and external (F _ext) faults are presented in the following sub-sections.

Figure 1:

Equivalent network single line diagram.

The sequence networks for internal and external faults are shown in Figures 2 and 3.

Figure 2:

Equivalent network single line diagram for in-zone fault.

Figure 3:

Equivalent network single line diagram for out-zone fault.

The positive sequence pre-fault voltage and current signal at the relay point (Figure 1) during steady-state operation is represented as I ‾ pre and V ‾ pre . For any internal fault can be expressed by the post-fault values of relaying quantities I ‾ int , post and V ‾ int , post at the relay location. Similarly, for the external side, these post-fault values of relaying quantities are represented by I ‾ ext , post and V ‾ ext , post at the relay location. Finally, the expression for the superimposed components of V and I for an internal and external fault are given as follows:

where Δi _int/Δi _ext and Δv _int/Δv _ext are the superimposed components of current and voltage for the internal and external sides fault respectively.

The pre and post-fault variations in the V and I values utilizing the data at buses M and N are given as,

where I ‾ pre   &   I ‾ post are the pre-fault and post-fault current variations and V ‾ pre   &   V ‾ post are the pre-fault and post-fault voltage variations, respectively.

Finally, the DAVI utilizing superimposed components computed by utilizing both ends relaying signals for internal and external faults can be depicted as,

2.1 Proposed zone discrimination method

Equations (7) and (8) show the DAVI at relay location PN (on analyzed line segment PN) for internal and external fault conditions. The DAVI, on the other hand, is used to separate internal and external faults. The condition for identifying the internal and external faults is stated as:

(10) { DAV I int = + Ve for int e rnal fault DAV I ext = 0  for external fault

To boost the precision and enhance the security of the protection scheme against non-fault disturbances, computation of threshold is desired. Further, the threshold should be adaptive and independent of system topology. This protection strategy also adopts the computation of an adaptive threshold whose steps are discussed below.

1. According to Equations (7) and (8), the DAVI (internal and external faults) for each phase at the relay location are computed.

2. Then, the mean of the DAVI for each phase is evaluated.

3. Finally, the threshold value is calculated based on Equation (11)

(11) T H j ( ξ ) = ζ ⋅ Max [ ∑ ( [ DAV I PN j ] − mean [ DAVI ‾ PN j ] ) 2 S ]

where ζ indicates the coefficient of margin. The length of the signal is denoted as S, ‘j’ represents the corresponding phase, PN represents the line segment and TH is the computed threshold value.

In the power system, the dynamic fault nature of DGs, measurement noises, and errors may cause variation of current at the relay point which causes differences in ( [ DAV I PN j ] − mean [ DAVI ‾ PN j ] ) 2 and hence, the variation increases and the set value reduces. This leads to the maloperation of the proposed scheme. Thus, based on numerous tests considering the 20 % more overloading and pick-up setting of overcurrent elements, a value of ζ = 1.2 is computed.

Therefore, the final criterion for discriminating internal and external faults immune to non-fault disturbance and system topology is set as,

(12) DAVI > TH }  for internal faults  ( i ) DAVI < 0 & DAIV < TH }  for external faults / for non- fault event  &  other feeder faults  ( ii ) DAVI = 0 }  No Fault  ( iii )

The steps of the proposed relaying scheme to differentiate the internal and external faults are illustrated in Figure 4. The proposed scheme computes the current variation at the relay location utilizing the current signals from different CTs locations using µPMUs. Then these signals are processed through the second-order low pass antialiasing filter to die out the noise components. Next, the DAVIs are computed using Equations (8) and (9). The phasors are obtained using full cycle recursive discrete Fourier transform (DFT) with 20 samples in each window. The sampling rates of 1 kHz and 1.2 kHz are used for 50 Hz and 60 Hz test systems respectively.

Figure 4:

Flow chart of proposed zone discrimination logic.

2.2 Observations

For a better understanding of the proposed method, let us discuss a fault case. First, a three-phase internal fault (R _F = 1 Ω) is simulated at 0.484 s. The line segment between buses 4 and 5 of feeder-1 (in Figure 6) is taken under investigation. As shown in Equation (8) following a fault, the relay at a given location will evaluate DAVI which is named an internal component, whose value should be greater than the threshold value. As shown in Figure 5 the DAVI for the internal fault (for all phases) satisfies the detection criterion thus, a fault has occurred in lines 4–5 which is an internal fault.

Figure 5:

Results for internal and external fault cases.

Secondly, a three-phase fault in line 3–4 is simulated and the relay of line 4–5 is considered for the analysis for which it is an external fault. The DAVIs computed for all phases using Equation (12) can be depicted in Figure 5. As seen from the figure, the DAVI for all phases meets the condition presented in Equation (12)(ii) and the index is lying below the threshold value, indicating an external fault scenario.

3 Results and discussion

In this section, the proposed differential logic is tested for various worst disturbance cases (comprising of fault and non-fault events). For validation, a modified two-feeder distribution test system integrated with IIDGs is taken into consideration as shown in Figure 6 whose parameters are borrowed from [25]. It is a 20 kV, two feeder radial distribution test network, each having a rating of 2.3 MVA. Three different IIDGs are considered in this work. First, a PV of 1 MVA is connected to the feeder through a 20 kV/480 V interfacing transformer. Next, a diesel generator of 6 MVA and a wind turbine of 5.5 MVA are connected through 20 kV/12.47 kV and 20 kV/0.69 V/0.69 V transformers, respectively. The fault inception time of 0.484 s is taken throughout the further analysis as it is a voltage zero crossing time which produces more transient. Numerous tests, as pointed out below, have been carried out to validate the efficacy of the proposed relaying scheme:

1. Different internal and external faults

2. High resistance faults

3. Variation in DG penetration level

4. Faults in the different feeder

5. During capacitor and load switching transients

Figure 6:

Modified two feeder test system [25].

3.1 Results for two feeder test system

3.1.1 Different internal faults

Different internal faults (11 types) having symmetrical and unsymmetrical types were created. The fault resistance is kept constant at 1 Ω and the fault location is in the middle of line 4–5. The µPMUs at bus-4 and bus-5 are monitoring the relaying quantities and sending them to the control unit for the decision-making process. Figure 7(i)–(iv) show the DAVIs at the relay location computed using Equation (11) for four different types of fault. These DAVIs of corresponding faulty phases cross the adaptive threshold value which depicts that faults occurred inside the zone.

Figure 7:

Results for internal faults (i) AG, (ii) AB, (iii) ABG, (iv) ABC.

3.1.2 Different external faults

Numerous test scenarios have been conducted for different external faults keeping fault resistance (R _F) constant to 1 Ω. The line section between bus 4 and bus 5 is taken into consideration. AG, AB, ABG, and ABC faults are simulated in the middle of the line 3–4. The DAVIs for different phases at the relay location are computed according to Equation (8). The outcomes are illustrated in Figure 8(i)–(iv). It is observed from the figure that the DAVIs of the faulty phases satisfy the external fault criterion and hence these faults are considered as out-zone faults for which the relay will restrain its operation. On the other hand, the internal component DAVIs will be within the set thresholds.

Figure 8:

Results for external faults (i) AG, (ii) AB, (iii) ABG, (iv) ABC.

3.1.3 Different fault resistance

In this segment, the sensitivity of the proposed protection strategy is examined for variation in R _F (low to high). The AG (SLG) fault in the mid-point of line 4–5 is simulated for R _F starting from 0.01 Ω to 1000 Ω. However, Figure 9 depicts the results for a few R _F cases such as 5 Ω, 50 Ω, 100 Ω, 200 Ω, 500 Ω, and 1000 Ω which indicate the correctness of the proposed scheme. Hence, the functioning of the proposed relaying logic is insensitive up to a R _F of 1000 Ω.

Figure 9:

Results for different fault resistance.

3.1.4 Faults in different feeders

This case is performed to validate the efficacy of the suggested method for faults in different feeders. As shown in Figure 6, the studied test system has two feeders in which line Sections 4 and 5 of feeder-1 is considered throughout the paper for investigation. But, now the fault is simulated in feeder-2 keeping the investigated line constant. An LG fault (AG type) is created in line 8–9 of feeder-2. During this fault, the relay should restrict the operation. The outcome obtained using the proposed technique is shown in Figure 10, from which it can be clearly observed that the DAVIs for all phases at the relay location satisfy the relaying criterion mentioned in Equation (12). Thus, the relay will restrict its operation.

Figure 10:

Results for AG fault in different feeder.

3.1.5 Fault during variation in DG penetration

To analyze the variation of DGs power output on the proposed relaying method, the penetration of DGs within the microgrid is varied from 60 %–100 % in a step of 10 %. For every scenario, AG (LG) fault at the mid-point of line 4–5 is simulated keeping R _F = 1 Ω. Various combinations of the power produced by the DGs are considered. Firstly, a total power output of 1.770 MW from DGs is considered (PV = wind = 135 KW, and diesel generator = 1.5 MW). Similarly, other possible variations in the penetration level is considered and the result obtained for all these variations is depicted in Figure 11. From the figure, it can be observed that fault and faulty phases can be effectively detected irrespective of the variation of the penetration levels of the renewable sources.

Figure 11:

Results for AG fault during variation of penetration level.

3.2 Results for modified IEEE-13 node test system

By simulating the redesigned IEEE 13-bus standard distribution network using EMTDC/PSCAD shown in Figure 12, the proposed approach is assessed. By including solar PV at bus 634 and a WT of DFIG type at bus 680, respectively.

Figure 12:

Studied test system-2 [24].

The studied test system is a 4.16 kV radial and unbalanced in nature as distribution feeders are not transposed. Capacitor banks, generating units, transformers, voltage regulators, and unbalanced loads are all included in the system. It consists of unbalanced laterals having one phase or two phases. Bus numbers 645, 646, and 684 only have two phases (A and C), while bus 611 only has one phase. The rest buses are balanced in nature. Reference [24] provides all the parameters of the studied model.

3.2.1 During internal and external faults

As stated in the aforementioned sections, an internal fault must be quickly detected and isolated from the healthy section. For validation, different fault types, including LG, LL, LLG, and LLL faults are simulated. As seen in Figure 12, these faults are incepted in the middle of the line connecting buses 632–633. The simulation is run for a range of fault resistance values between 1 Ω and 1000 Ω.

First, an LG (AG) fault is created in line section 632–633. The fault resistance of 1 Ω is considered. The outcome produced by applying the proposed logic is depicted in Figure 13(a), where the DAVI of phase-A crosses the set threshold level while the remaining phases are within the limit set for the threshold. Hence, the relay will detect it as an in-zone phase-A fault. Similar observations can be pursued for LL, LLG, and LLL faults as depicted in Figure 13(a)–(d).

Figure 13:

Results for internal faults.

Similarly, various external faults (different types with fault resistance variation within 1–1000 Ω) are simulated. The fault locations are presented in Figure 11. Figure 14 shows the obtained results for AG fault cases using the proposed logic. It can be seen from the figure that following a fault, the DAVIs of the respective phases are lying below the set value, and its magnitude is almost close to zero, indicating an out-zone (external) event.

Figure 14:

Results for external fault.

3.2.2 During switching events (CS and LS)

Switching transients generated other than faults may trigger the false operation of any protection scheme. Thus, the relays should clearly discriminate these non-fault transients and restrict the relay operations. CS and LS are the most common events in distribution systems which produce transients at the point of connection and disconnection. A well-designed protection scheme should correctly discriminate between the fault and non-fault events. To verify the performance of the proposed logic during such events, both CS and LS scenarios are considered and described below.

First, a CS event is initiated at 0.7 s, and a capacitor bank of 0.350 MVAR is switched on bus-632. Second, an LS event is simulated at bus-632 with real and reactive power ratings of 0.17 MW and 0.125 MVAR. Figure 15(a) and (b) demonstrate the results for both the events using proposed logic. In both cases, DAVIs of each phase are lying within the set thresholds. Hence, the relay will not issue the trip signal and restrain its operation. This condition is classified as a CS or LS or other feeder fault event. It should be remembered that the proposed method is not capable of classifying the event type rather than indicating the disturbance zone.

Figure 15:

Results during (i) load switching, (ii) capacitor switching.

3.2.3 During high impedance fault (HIF)

High impedance faults (HIFs) introduce significantly more nonlinearities in the current signal as compared to non-HIFs. Apart from that, it also provides high non-linear impedance to the fault currents resulting significant reduction of fault currents at the relay location. Thus, it is desired to test the performance of the proposed scheme in such a scenario. To corroborate, a total of four HIF cases (phase-A and phase-B) are simulated. Two HIFs each are simulated at the middle of the analyzed line section (632–633), which is internal and external HIFs between nodes 633–634. Results as shown in Figure 16(a) and (b) reveal that the proposed approach successfully detects the in-zone HIFs.

Figure 16:

Results for HIF case.

3.2.4 Results during unavailability of data

Performing the DAIV computation inside each data window is necessary for the suggested technique to detect the fault. To establish the appropriate threshold, this windowed DAIV is also utilized. Since the relaying data may be absent or unavailable within the data window, the performance must be examined during such scenario. LG (AG) faults after the middle of the investigated lines are analyzed with one cycle of missing data following the fault initiation. Figure 17 displays the results both with and without missing data. The figure shows that while the threshold and DAIV have not changed significantly, however, the relay decision is postponed by 18 ms. As a result, it may impact the operational speed and lead to instability problems.

Figure 17:

DAIVs during AG fault with and without missing data.

4 Comparative assessment with other techniques

Despite the potential theoretical benefits of the suggested method, this section compares its accuracy to various recently published advanced relaying techniques and commercial relaying algorithms. These methods are divided into two subcategories and are discussed in the next section.

4.1 Comparison with commercial relay algorithms

The 87L line differential scheme [39], the alpha-plane scheme [40], and the IQ-based method [41] are very powerful techniques for line protection. Because of their immunity to external imperfections, these strategies are extremely successful and frequently used. However, evaluating the performance of such strategies in the presence of RESs is very appealing and desired.

In the two-feeder test system presented in Figure 1, a high resistance (R _F = 1000) AG fault is simulated in order to test these methods. A high resistance fault is a very serious scenario since it represents a very small quantity of current at the relay site. However, these methods malfunction and are unable to effectively identify the fault as a result of the high resistance characteristic imposed by the fault, as shown in Figure 18(a) and (b). Indeed, the defective phase-“a” is still present inside the alpha plane’s non-trip zone, and the Δv _res value is higher than the Δv _op, indicating that the significant fault resistance of such commercially existing relaying techniques is intended to be high fault resistance. The outcome is depicted in Figure 18(c) and also includes consideration for “87A”-line current differential relay with a single slope characteristic. The differential current is seen to be in the restraining zone in the figure, which means that the 87A relay is unable to detect the fault due to the influence of high fault resistance.

Figure 18:

In-zone high resistance fault (1000 Ω) (i) IQ based method, (ii) 87-L alpha based method, (iii) 87 percentage based method.

4.2 Comparison with various signal processing based algorithms

Recently, various signal-processing-based approaches have attracted attention in the area of power system protection. These algorithms are classified as time-frequency and signal decomposition-based techniques. For evaluation, Hilbert-Huang transform based Differential Discrete Teager Energy Operator (DDTEO) [16], and fast discrete Stockwell transform based Discrete Spectral Energy (DSE) [17], and Differential Energy (DE) using Intrinsic Time Decomposition (ITD) [41] based techniques are considered.

First, a capacitor switching case is taken into consideration, whose result has already been presented in sub-section 3.2.2. The results obtained using DDTEO and DSE based techniques are shown in Figure 19(a) and (b), which indicate that the relay utilizing these techniques will maloperate as the corresponding indexes of each phase surpass the set threshold value after the switching event. Hence, these methods are sensitive to switching events.

Figure 19:

Result for capacitor switching using (i) DDTEO, (ii) DSE based methods.

Secondly, a high fault resistance (1000 Ω) case as mentioned in Section 4.1 is considered. The results utilizing the DSE and DE-based approaches are depicted in Figure 20(a) and (b). The indexes for each phase computed using each method are lying within the set value indicating non-fault events that restrain the relay operation. Thus, these methods are prone to high fault resistance. However, the proposed method is very efficient in function for fault cases with high fault resistance as presented in Figure 19.

Figure 20:

Result for in-zone high resistance fault (1000 Ω) using (i) DDTEO, (ii) DSE based methods.

Furthermore, as indicated in Table 1, a significant number of test scenarios, covering a range of situations, are considered. The findings of the aforementioned approaches are utilized to create a correlation matrix. Next, the accuracy of several fault detection methods is calculated for different scenarios including faults and non-faults, as seen in Figure 21. The figure indicates that while the accuracy of the proposed method is good, that of the previously described approaches.

Table 1:

Simulated number of test cases.

Events fault	Cases	Category
	50-internal 16-external	Various types of fault (LG, LL, LLG, LLL), different fault resistances (0.01 Ω–1000 Ω)
	30	Different fault locations (0–95 % of the line)
	20	HIF cases
	10	Variation in DGs penetration level
Non-fault	8	Load switching events
	10	Capacitor switching scenarios 0.125 MVAR–0.350 MVAR

Figure 21:

Accuracy comparison of various methods with proposed method.

5 Conclusions

This paper presents a current variation index at the relay location, utilizing the information of the other feeders and line sections using µPMUs in the presence of IIDGSs. The benefit of this research is that such a problem has received little attention in the literature. The modeling of the various IIDGSs is the first step in this endeavor. Then, the internal and external fault components of the respective phases are computed using the proposed method. In order to distinguish the fault zone, the internal and exterior fault components of the DAVI are calculated. For different active distribution systems with varying IIDGSs, the suggested relaying logic is evaluated. Case studies of various disturbances (faults and switching events) are simulated, including fault situations. The outcomes indicate that the suggested method’s operation time to distinguish the fault zones is 10 ms. Lastly, comparison analysis demonstrates the superiority of the relaying reasoning that has been given.