Received 21 November 2023, accepted 12 December 2023, date of publication 16 January 2024,

date of current version 25 January 2024.

Digital Object Identifier 10.1109/ACCESS.2024.3354790

Hardware Supported Fault Detection and

Localization Method for AC Microgrids
Using Mathematical Morphology
with State Observer Algorithm
FAISAL MUMTAZ 1 , KASHIF IMRAN 1, HABIBUR REHMAN 2, (Member, IEEE),
AND HAMMAD ALI QURESHI1
1 U.S. Pakistan Center for Advanced Studies in Energy (USPCAS-E), National University of Sciences and Technology (NUST), Islamabad 44000, Pakistan
2 Department of Electrical Engineering, American University of Sharjah, Sharjah, United Arab Emirates
Corresponding author: Kashif Imran ([email protected])
This work was supported in part by the Open Access Program from the American University of Sharjah.

ABSTRACT Microgrids are the future version of advanced distribution networks due to the fast growth of
renewable energy resources near consumers’ side. The microgrids are operated in on-grid mode (OGM) with
the utility grid, and isolation mode (IM) without the utility grid. This dual operational mode causes protection
and control challenges in the microgrids. This research paper suggests an advanced hardware-supported
fault detection, phase identification & localization method for AC microgrids. The scheme deploys a
Discrete Kalman Filter (DKF) for state estimation of voltage and current signals. Then, a Mathematical
Morphology (MM) is engaged for generating a novel fault detection/classification index named segregated
energy signature (SES) from estimated voltage and current signals. The system is considered to be faulty
if the SES is higher than a predefined threshold setting, while phase identification is achieved by default
because of the per-phase implementation of DKF&MM. Moreover, the directional features of the cumulative
energy signature (CES) are also computed from MM-based non-fundamental current and voltage to localize
the faulty section. The established scheme is tested on the CIGRE microgrid benchmark test bed on Matlab-
Simulink software. In addition, the suggested method is also examined on the dSPACE MicroLab testing
hardware setup in the Smart grid lab at USPCAS-E, NUST, Pakistan. The result illustrates that the proposed
scheme successfully detects, classifies, and localizes the low impedance fault (LIF) as well as high impedance
fault (HIF) in both operational modes and topological structures with 96.6% accuracy.

INDEX TERMS Distributed generation, discrete Kalman filter, fault detection, hardware in the loop,
microgrids protection, mathematical morphology.

I. INTRODUCTION several benefits [2]. However, microgrids face different

Microgrids are self-sufficient networks having various protection challenges notably i). Fault Detection and Local-
renewable energy-dominated distributed generations (DG), ization [3], ii). Coordination of protection devices [4],
loads, and battery energy storage systems, and are equipped iii). Islanding detection [5], [6], iv). High Impedance Faults
with self-sufficient protection infrastructure [1]. Microgrids detection [7], [8], v). Voltage and Frequency Control [9], and
are the future grids to provide clean and green energy due to vi). Cybersecurity Concerns [10]. Among all the microgrid
protection challenges our focus is only designing a protection
scheme for detection, classification, and localization of LIF
The associate editor coordinating the review of this manuscript and as well as HIF. These challenges are due to the following
approving it for publication was Arturo Conde . reasons.

2024 The Authors. This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License.
12446 For more information, see https://creativecommons.org/licenses/by-nc-nd/4.0/ VOLUME 12, 2024
F. Mumtaz et al.: Hardware Supported Fault Detection and Localization Method

1. Dynamic levels of current during different operational

conditions, the current in OGM of operation is very
high while the current during IM of operation is very
low.
2. The flow of electric current is bidirectional due to the
multiple DGs feeding the grid.
3. A different topological structure like radial, meshed,
or loop causes a challenging scenario [11], [12].

Therefore, under such circumstances, the traditional pro-

tection strategies exhibit maloperation during harsh system FIGURE 1. State-estimation through DKF.
conditions. Hence, we have to design a protection method that
is efficient to deal with these operating conditions without any
nuisance tripping. intelligence with a traveling wave approach was utilized
Previous literature addresses microgrid protection issues in [28] to identify, classify, and localize the faults in the
in many aspects. Different types of protection methods were microgrids. The authors in [29] utilized a Kalman filter-based
reported, such as over-current, differential, distance, intelli- approach for protecting the microgrids in both OGM &
gent, and some modified methods. A novel strategy based IM operational conditions under radial, meshed, and loop
on transient wavelet energy was proposed in [13] for the topologies.
protection of inverter-dominated AC microgrids. The pre- This work suggests a robust hardware-supported fault
sented scheme was very robust and capable of detecting localization and detection technique for AC microgrids.
and localizing different faults under various operating con- A DKF was used as a state observer in the proposed scheme
ditions. HIF was detected and classified in microgrids using to estimate the state of the voltage and current signals. Next,
Kalman filtering [14]. In [15] the authors utilized the wavelet the MM algorithm is engaged to create a novel index called
spectrum of energy to identify, classify & localize vari- SES, for fault identification and categorization. The system
ous faults in AC microgrids during the OGM of operation. is considered to be faulty if the SES is higher than a predeter-
Authors in [16] utilized the online current wavelet injection mined threshold value. To pinpoint the faulty section, parallel
method, traveling waves, and mathematical morphology to computations are also performed on the directional aspects
protect microgrids from different kinds of faults. Euclidean of the CES. On MATLAB/Simulink software, the established
distance was deployed in [17] to compute superimposed scheme is tested on the CIGRE microgrid benchmark. Also,
components with waveform similarity for the fault’s detec- the hardware arrangement based on the dSPACE MicroLab
tion/classification and localization. A Time-time-transform testing environment is used to test the presented method.
was proposed in [18] for microgrid protection in both the The results show that the suggested approach successfully
OGM and IM operations. Communication-based microgrid locates, classifies, and detects LIF and HIF in both operation
protection scheme was proposed in [7] and [19] through modes and topological structures. The value additions of the
the utilization of high-frequency and impedance respectively. suggested scheme are as follows.
The median filter was utilized as a state estimation algorithm
in [5] to extract harmonic features. These harmonic features 1. Application of both the DKF and MM in the time and
were then employed to identify, classify, and localize the fault frequency domain for AC microgrid protection.
events in the microgrids. 2. Computationally the proposed algorithms are less
Some intelligent microgrid protection strategies were also intensive.
reported in the previous literature. Convolution neural net- 3. Novel and very simple fault detection/classification and
work was deployed in [7], and [18] to identify, classify and localization indexes are developed.
localize the faulty events in the microgrids. Modal decom- 4. The proposed method deals with noisy measurements
position and combined data mining approach were proposed as well.
in [22] to identify, classify, and localize different types of 5. The proposed scheme is easy to implement on the
faults in AC microgrids. In [23] empirical mode decompo- hardware setup.
sition was deployed to identify, classify, and localize the 6. The suggested scheme detects both HIFs and LIFs.
faults in the microgrids. A support vector machine was uti- The remaining paper organization is as follows. The basic
lized in [24] for the protection of AC microgrids in OGM principles of the proposed technique are elaborated in
and IM. A novel protection method based on deep learning section II. A detailed methodology of the suggested scheme is
was proposed in [25] to protect the microgrid from fault mentioned in section III. The threshold setting and the utilized
incidents. State observer with the recurrent neural network CIGRE test benchmark are illustrated in section IV. Section V
was suggested in [26] to identify, classify, and localize vari- illustrates the simulation results comprehensively. The Hard-
ous faults. A hybrid classifier approach was utilized in [27] ware setup and results are explained in section VI. The
for fault location identification in AC microgrids. Artificial detailed performance analysis is mentioned in section VII.

VOLUME 12, 2024 12447

 F. Mumtaz et al.: Hardware Supported Fault Detection and Localization Method

TABLE 1. Pesudo code of MM algorithm. TABLE 2. Pseudo code of DKF algorithm.

At last, the concluding remarks along with future work are

provided in section VIII of this paper.

II. BASIC PRINCIPLE

A. DKF ALGORITHM
Fault detection in any scheme relies on correct parameter
estimation. Therefore, any scheme must choose a parameter
estimation algorithm that is capable of estimating electri-
cal parameters with high accuracy. The proposed scheme
deployed the DKF, a well-known state estimation algorithm
for that purpose. Rudolf E. Kalman, one of its principal theo-
rists, is honored by having the filter retain his name. Figure 1
depicts a detailed illustration of state estimation of per phase
voltage signal using the DKF algorithm while a similar esti-
mation is used for voltage as well. In comparison to estimates
based just on a single measurement, it generates estimates
of unknown variables that are often more accurate. A joint
probability distribution across the variables is calculated for
each timeframe by the Kalman filtering procedure, com-
monly known as a linear quadratic estimation. This algorithm
is utilized in control theory and statistics [25], [26], [27]. The
direction, navigation, and control of moving objects, notably
those on land, air, and space, are some technological uses
FIGURE 2. Mathematical morphological signature depiction.
of Kalman filtering. Moreover, Kalman filtering is a time
series analysis method that is commonly used in fields like
radar and signal processing. Because of these advantages,
we deployed DKF for the estimation of per-phase current and processing current & voltage signals. MM is a time-domain
voltage signals. The Pseudo code of the DKF algorithm is signal processing technique that, in contrast to STFT and WT,
demonstrated in Table 2. involves only simple computations (like maxima, and min-
ima) and whose correctness never depends on the periodicity
B. MM ALGORITHM of the signal. As opposed to the STFT and WT, which may
A non-linear signal processing tool called MM, created by deform the signal, a more significant feature of the MM is
Matheron and Serra may extract transitory features from its ability to extract the ascending or descending edges from

12448 VOLUME 12, 2024

F. Mumtaz et al.: Hardware Supported Fault Detection and Localization Method

FIGURE 3. Flow chart of proposed hardware supported scheme.

a processing signal more visibly [16]. Despite that structur- Here Ydil and Yero represent dilated and eroded signals respec-
ing elements (SE) are fundamental to all MM-based tasks, tively. Furthermore, f is a signal to be dilated or eroded,
including dilatation, erosion, opening, and closing as depicted whereas g is SE.
in Figure 2. The detailed computational steps of the MM

algorithm are shown in Table 1. When performing any type Yopen = fogn (3)
Yclose = f .gn


of MM-based operation, SE slides over the processing signal (4)
like a window to extract transitory features that resemble its
structure [32]. Hence, for applications in power systems, the Opening and closing operations of ‘‘f ′′ by ‘‘g′′ denoted by
best SE should be chosen. The SE for the electrical signal is ‘‘fog’’ and ‘‘f · g′′ respectively.
as follows.
III. PROPOSED METHOD
Ydil = max {f (n − m) + g(m)} (1) The proposed protection method designed for the micro-
0≤n−m≤n
Yero = min {f (n + m) + g(m)} (2) grid encompasses five operational units; readings acquisi-
0≤n+m≤n tion, DKF pre-processing unit, MM-based fault detection/

VOLUME 12, 2024 12449

 F. Mumtaz et al.: Hardware Supported Fault Detection and Localization Method

classification unit, MM-based fault localization unit, and at Where, Ie is the estimated current and Ve is the estimated
last the tripping unit. The flow diagram of the suggested voltage from the corresponding faulty bus. In is the magnitude
method is illustrated in Figure 3. of the estimated current while Vn is the magnitude of the esti-
mated voltage. However, ∅ depicts the random error or white
A. READINGS ACQUISITION UNIT Gaussian noise. These estimated voltage and current signals
The voltage/current readings acquired from the faulty are provided to a mathematical morphological algorithm for
bus were pre-processed through the Butterworth filter for the calculation of fault detection and classification index and
antialiasing purposes. Secondly, these measurements are also for fault localization indexing.
analog signatures, and our protection relay and the DKF
algorithm need discrete values. Therefore, the analog-to- C. MM-BASED FAULT DETECTION & CLASSIFICATION UNIT
digital conversion is done through the ADC module of the The MM is engaged in this unit for the calculation of fault
reading acquisition unit. The discrete version of both current detection & classification index. The estimated voltage &
and voltage signals is defined by eqs (5) and (6) respectively. current of each phase are provided at the input of MM in a
segregated manner to compute per-phase SES. The SES is
Ir = I sin (ωt + ϕ) (5)
computed as follows.
Vr = V sin (ωt + ϕ) (6)
SES a = b
Ie ∗ Vbe (14)
where Ir is the measured current and Vr is the measured
SES b = b
Ie ∗ Vbe (15)
voltage of the corresponding faulty bus. I is the current
magnitude while V is the voltage magnitude. However, ω and SES c = b
Ie ∗ Vbe (16)
ϕ are angular frequency and measurement noise respectively.
If the value of SES of any single/multiple phases is above
At the next step, the converted discrete current/voltage signals
the threshold of the relevant phase, the phase is regarded as
are sent to DKF for the exact state estimation because the
faulty. As the index is computed in phase segregated manner,
measured readings are noisy and are not suited for direct fault
no separate fault classification unit is needed. Moreover, this
index calculation.
unit sends two signals one of them is the OR-based fault
detected (OBFD) signal which is sent to the main protection
B. DKF PRE-PROCESSING UNIT
unit of the relay while the second is per phase fault detection
The discrete voltage & current signals received from the (PPFD) signal provided to the corresponding phase circuit
previous stage are provided to DKF for the estimation of breaker’s tripping unit.
electrical magnitudes. It is noted that the DKF is implemented
in a per-phase manner therefore, only a related mathematical D. MM-BASED FAULT LOCALIZATION UNIT
depiction is presented in the whole manuscript to avoid com-
This section mainly focuses on two parts: i. direction iden-
plexity. The remaining three phases are simply 120◦ degrees
tification using CES, and ii- communication-based main &
apart. The state-space model.
backup facility. Initially, the calculation of CES should be
Xn+1 = Bx n + W (7) done then a detailed description of communication-based
in = Yxn + un (8) main & backup be provided.
The faulty section is identified in the presented scheme
The tuned parameters of DKF utilized in this study are using the directional trends of the CES. The CES is sim-
defined by Equations 9. ply calculated by the mathematical sum of eqs (14), (15),
 −1  T and (16). The computed CES is as follows.
xn system states; X̂n = In In−1 ;b = 1 0 ;
2 cos ω0 −1
 
 T CES= SES a + SES a + SES a (17)
H = 10 ; A= (9)
1 0
It is noted that the CES is negative during forward faults
The Table 2 algorithm works using this state space model while positive during reverse faults. Therefore, the negative
and tuning parameters to generate desired estimates of cur- direction of CES in any section indicates the presence of
rent and voltage signals. The trigonometric derivatives of forward faults.
eqs (5) and (6) are taken for the acquisition of an iterative Secondly, as we can observe in the microgrid after the
conversion of the obtained signals as follows. occurrence of a fault the CES is negative on several buses.
Therefore, it is difficult to decide faulty section based on
e+1 = 2I n + ∅
Id (10)
just this direction signal. Based on this main reason the
Vde+1 = 2V n + ∅ (11) proposed scheme utilized the communication infrastructure
The estimated current and voltage signatures obtained of the microgrid to communicate with three adjacent relays
through DKF operation are given by equations (12) and (13). to take the correct decision. The detailed depiction of this
scenario is illustrated in Figure 4. It is the first three-bus
Ie = In + ∅
b (12) test system on which the whole proposed scheme is imple-
Vbe = Vn + ∅ (13) mented. It comprises two zones; i- Zone 1 and ii- Zone 2, the

12450 VOLUME 12, 2024

F. Mumtaz et al.: Hardware Supported Fault Detection and Localization Method

FIGURE 4. Initial 3-bus test bed for implementation of MM based fault localization logic.

FIGURE 5. CIGRE benchmark test bed for AC microgrid.

relay-1 (R-1) is considered a reference relay to understand Then, the directional trends on R-4 signals are cross-checked
the concept of a communication-assisted MM-based fault  
localization unit. if CES is negative at R1
Secondry unit operation (20)
The R-1 is communicated to R-2 and R-4 through the com- & CES is negative at R4
munication link, every relay sends and receives two signals Finally, the OR operation is performed at the output of the
from the other relay units: i- OBFD and, ii- CES direction main and backup protection units. However, it is noted that
signal. The following logic is followed for the main and there is an intentional delay of 3 ms to operate the main
backup unit’s operation. protection first. Then the trip signal (TS) is sent to the tripping
  unit.
if CES is negative at R1
Main unit operation (18)
& CES is negative at R2
  E. TRIPPING UNIT
if CES is negative at R1 The tripping unit receives two signals from previous stages
however (19)
& CES is positive at R2 one of them is PPFD and the second one is TS. They are

VOLUME 12, 2024 12451

 F. Mumtaz et al.: Hardware Supported Fault Detection and Localization Method

FIGURE 6. A single phase to ground fault in islanded operational mode when switch-1 is open.

processed through AND-gate. If both of them are triggered the suggested scheme was tested in different worst-case sce-
the corresponding CB gets operated on to protect the healthy narios and operating conditions for the exact calculation of
system. the threshold. It is observed that ideally, the SES is 0 during
  normal operating conditions while more than zero during the
if PPFD = 1 faulty situation. Therefore, to cover all kinds of faults in the
Tripping the CB (21)
& TS = 1 real-time practical environment the SES threshold is chosen
to be 0.05. If the SES is higher than this pre-defined threshold
IV. CIGRE BENCHMARK TEST BED & PRE-DEFINED value, the system is in a fault condition. Secondly, the CES is
THRESHOLD negative for forward faults, while positive for reverse faults.
This subsection mainly focuses on the AC microgrid network
that is engaged in the proposed scheme for fault analysis. V. SIMULATION RESULTS
Secondly, the threshold settings are explained in the scheme. The section described the fault analysis of the suggested
method in different operating modes, topology resistances,
A. CIGRE BENCHMARK TEST BED locations, and different DG’s intermittencies. Almost all
The proposed scheme utilized a well-renowned CIGRE kinds of phase-to-phase and phase-to-ground faults were
benchmark test system for extensive fault analysis. The tested to validate the performance of the suggested method.
single-line diagram of the CIGRE benchmark test network is The faults were also checked during the LIF and the HIF in
illustrated in Figure 5. The test system comprises 3 inverter- addition the single-phase tripping was also checked.
based DGs, 2 synchronous-based DGs, and one battery
energy storage system. It consists of 10 buses and the utility A. SINGLE-PHASE TRIPPING
grid is connected to the rest of the microgrid via a point Single-phase tripping is very important in any scheme to
of common coupling. The circuit breaker (CB), CB-S is detach only the faulty phase from the healthy network.
utilized in the proposed test system to switch the operation Therefore, numerous single-phase faults were tested in the
mode between OGM and IM. The test system had 9-line proposed scheme for validation, while only one of the cases is
sections, while switches S-1 and S-2 were provided for the presented here due to lack of space. A fault at a single-phase
topology changing purpose. The detailed parameterization of C occurred on the line section between bus 3 and bus 7 at
the CIGRE benchmark test system is acquired from [14]. a time stamp of 300 msec. It is seen from Figure 6 that
the presented method efficiently detects and classifies the
B. PRE-DEFINED THRESHOLD fault immediately. The negative direction of the Cumulative
Threshold settings are very important in any scheme for energy signature at the corresponding relay illustrates the
rapid fault detection/classification and localization. Hence, presence of a forward fault in this section. Therefore, the

12452 VOLUME 12, 2024

F. Mumtaz et al.: Hardware Supported Fault Detection and Localization Method

FIGURE 7. A double phase to ground HIF in OGM operational mode when switch-1 and switch-2 are open.

FIGURE 8. A three phase to ground fault in islanded operational mode when switch-2 is open.

main protection unit of the relevant relay of this line section performed to authenticate the capability of the suggested
successfully locates the fault. The tripping signals of some method. A double-phase BA-G fault occurred at the line
other single-phase fault cases are depicted in Figure 11(a). section between bus 4 and bus 10 of the CIGRE test system
at a time stamp of 200 msec. It is seen from Figure 7 that
B. HIGH IMPEDANCE RESULTS the presented method successfully detected and classified the
Detection of HIF is very challenging due to a very low fault immediately. The negative direction of the CES at the
fault current [10]. Many high-impedance fault cases are related relay illustrates the existence of a forward fault in

VOLUME 12, 2024 12453

 F. Mumtaz et al.: Hardware Supported Fault Detection and Localization Method

FIGURE 10. Hardware results of proposed scheme.

It indicates that the single-phase fault initiated from the dotted

FIGURE 9. Hardware supported experimental setup. green line and the voltage of the faulty phase is facing a sag.
The proposed fault detection method can be implemented
in dSPACE MicroLab hardware as follow.
this section. Therefore, the main protection unit of the related 1. By connecting the necessary sensors to the hardware
relay of this line section successfully locates the fault. The channels.
tripping signals of some double line-to-ground HIF cases are 2. Acquiring the measured voltage and current data.
depicted in Figure 11(b). 3. Preprocessing the data to remove noise.
4. Extracting the relevant features from the preprocessed
C. THREE PHASE FAULTS data.
Three-phase faults are the most threatening faults while less 5. Generation of fault detection indexes of proposed
occurring in power systems. Therefore, the effectiveness of algorithm using these features.
the presented method was also authenticated in three-phase 6. Validating and testing the algorithm with simulated or
faults under different operation conditions and microgrid real fault data.
topologies. A three-phase ABC-G fault occurred at the line 7. Integrating the algorithm into the control or monitoring
section between bus 9 and bus 10 of the CIGRE test system system, and continuously monitoring the system using
at the time stamp of 250 msec. It is seen from Figure 8 the hardware’s visualization tools.
that the suggested scheme effectively detected and classified 8. This implementation allows for real-time monitoring
the fault immediately. The negative direction of the CES at and detection of faults in the system.
the corresponding relay illustrates the existence of a forward
fault in this section. Therefore, the main protection unit of VII. PERFORMANCE DEPICTION
the relevant relay of this line section successfully located The performance of the proposed method was validated
the fault. The tripping signals of some 3-phase cases are through testing across various fault types, including phase-to-
illustrated in Figure 11(c). A case of a no-fault scenario is phase and phase-to-ground faults. Additionally, the method’s
also shown in Figure 11(d). effectiveness was evaluated in both LIF and HIF, with exam-
ination also conducted on single-phase tripping scenarios.
VI. HARDWARE IN THE LOOP RESULTS The presented microgrid protection scheme is compared
The presented microgrid protection scheme is also tested on with some previous methods to check the performance of the
hardware in the loop setup using the Dspace micro-lab box in proposed scheme as depicted in Table 3. The chosen methods
the smart grid lab at USPCAS-E, NUST, Pakistan. contain some signal processing-based schemes and some
A detailed illustration of the experimental setup developed intelligent methods. Three types of performance indicators
for hardware validation is shown in Figure 9. The experimen- under observation are as follows: i. computational burden, ii.
tal setup has been created by using a host PC that is used Accuracy, and iii. Internal operation time of the schemes.
as a Controlled desk, is attached to the Dspace micro-lab
box, and the analog outputs of the Dspace micro-lab box are A. COMPUTATIONAL BURDEN ANALYSIS
connected to the cathode ray’s oscilloscope. Different kinds Computational burden analysis is vital in evaluating the
of fault conditions are tested on hardware setup to validate the performance of microgrid protection schemes, especially
proposed scheme. concerning their implementation feasibility. Comparing a
A case study of single-phase tripping is depicted in proposed scheme’s computational burden against existing
Figure 10. In this, a real-time CRO signature of the voltage methods helps in determining its efficiency and practical-
signal was acquired at a faulty bus during a single-phase fault. ity. There are different parameters for computational burden

12454 VOLUME 12, 2024

F. Mumtaz et al.: Hardware Supported Fault Detection and Localization Method

FIGURE 11. (a) Tripping singnature of some single line to ground fault in different operational mode. (b) Tripping singnature of some Double
line to ground fault in different operational mode. (c) Tripping singnature of some Three phase faults in different operational mode. (d) No
fault condition.

TABLE 3. Performance analysis of proposed method with some existing counter parts.

analysis of any proposed scheme while we check the process- C. INTERNAL OPERATION TIME ANALYSIS
ing time of the proposed algorithm for computational burden The proposed scheme successfully detected the fault in less
analysis. The computational burden of the proposed scheme than 15 milliseconds in almost all cases. Therefore, by com-
is cross-checked with some existing methods as depicted in parison, it is observed that this detection time is much
Table 3. It is observed that the proposed approach is compu- less than most of the existing methods. Moreover, some
tationally less demanding than some of the existing methods. other parameters like operational modes, topologies, noise
consideration, ease of implementation, and hardware imple-
B. ACCURACY ANALYSIS mentation have also been compared to check the efficacy of
The proposed scheme is very accurate and reliable for the pro- the proposed scheme. The proposed scheme shows better per-
tection of microgrids as depicted in Table 3. Approximately formance as compared to some of the existing benchmarked
30 different types of faults are tested in various microgrid methods. However, the proposed scheme also lacks some
operational regimes, out of which 29 cases were successfully existing compared methods in a few aspects as mentioned in
detected. Hence, the proposed scheme is 96.6% accurate. The Table 3.
accuracy of the proposed scheme is computed as follows.
VIII. CONCLUSION
Accuracy = number of sucessfull
This paper proposed a robust hardware-based fault localiza-
operations/total number of tion method for AC microgrids. Initially, the state estimation
operations (22) of voltage and current signals was done through DKF. Then,

VOLUME 12, 2024 12455

 F. Mumtaz et al.: Hardware Supported Fault Detection and Localization Method

the MM was used to develop the SES, a new fault iden- [12] A. A. Memon and K. Kauhaniemi, ‘‘A critical review of AC microgrid pro-
tification and categorization index. The system is regarded tection issues and available solutions,’’ Electr. Power Syst. Res., vol. 129,
pp. 23–31, Dec. 03, 2015, doi: 10.1016/j.epsr.2015.07.006.
as faulty if the SES exceeds a certain threshold value, [13] D. Liu, A. Dysko, Q. Hong, D. Tzelepis, and C. D. Booth, ‘‘Transient
and phase-segregated DKF& MM implementation is done wavelet energy-based protection scheme for inverter-dominated micro-
for the autonomous classification of faults. To locate the grid,’’ IEEE Trans. Smart Grid, vol. 13, no. 4, pp. 2533–2546, Jul. 2022,
doi: 10.1109/TSG.2022.3163669.
faulty section, the directional properties of the CES were
[14] F. Mumtaz et al., ‘‘High impedance faults detection and classifica-
computed concurrently. On MATLAB/Simulink software, tion in renewable energy-based distribution networks using time-varying
the established scheme was tested in the CIGRE micro- Kalman filtering technique’’ Eng. Proc., vol. 20, no. 1, p. 34, 2022, doi:
grid benchmark. Also, the hardware setup based on the 10.3390/engproc2022020034.
[15] X. Zheng, Y. Zeng, M. Zhao, and B. Venkatesh, ‘‘Early identification
dSPACE MicroLab testing environment is used to validate and location of short-circuit fault in grid-connected AC microgrid,’’
the presented scheme. The results indicate that all types of IEEE Trans. Smart Grid, vol. 12, no. 4, pp. 2869–2878, Jul. 2021, doi:
faults are successfully detected by the proposed scheme in 10.1109/TSG.2021.3066803.
<15 msec with 96.6% accuracy and with a very low compu- [16] M. Dodangeh and N. Ghaffarzadeh, ‘‘Fault detection, location, and classi-
fication method on compressed air energy storages based inter-connected
tational burden. micro-grid clusters using traveling-waves, current injection method, on-
line wavelet, and mathematical morphology,’’ Int. Trans. Electr. Energy
Syst., vol. 31, no. 12, Dec. 2021, doi: 10.1002/2050-7038.13190.
ACKNOWLEDGMENT
[17] M. A. Jarrahi, H. Samet, and T. Ghanbari, ‘‘Protection framework for
This article represents the opinions of the author(s) and microgrids with inverter-based DGs: A superimposed component and
does not mean to represent the position or opinions of the waveform similarity-based fault detection and classification scheme,’’ IET
American University of Sharjah. This research utilized hard- Gener., Transmiss. Distrib., vol. 16, no. 11, pp. 2242–2264, Jun. 2022, doi:
10.1049/gtd2.12438.
ware in the Smart Grid Laboratory, USPCAS-E, National [18] O. A. Gashteroodkhani, M. Majidi, M. S. Fadali, M. Etezadi-Amoli,
University of Sciences and Technology. and E. Maali-Amiri, ‘‘A protection scheme for microgrids using time-
time matrix z-score vector,’’ Int. J. Electr. Power Energy Syst., vol. 110,
pp. 400–410, Sep. 2019, doi: 10.1016/j.ijepes.2019.03.040.
REFERENCES [19] R. Bhargav, C. P. Gupta, and B. R. Bhalja, ‘‘Unified impedance-
[1] A. Cagnano, E. De Tuglie, and P. Mancarella, ‘‘MicroGrids: based relaying scheme for the protection of hybrid AC/DC microgrid,’’
Overview and guidelines for practical implementations and IEEE Trans. Smart Grid, vol. 13, no. 2, pp. 913–927, Mar. 2022, doi:
operation,’’ Appl. Energy, vol. 258, Jan. 2020, Art. no. 114039, doi: 10.1109/TSG.2021.3129532.
10.1016/j.apenergy.2019.114039. [20] S. B. A. Bukhari, C. Kim, K. K. Mehmood, R. Haider, and M. Saeed
[2] N. Hussain, M. Nasir, J. C. Vasquez, and J. M. Guerrero, ‘‘Recent devel- Uz Zaman, ‘‘Convolutional neural network-based intelligent protection
opments and challenges on AC microgrids fault detection and protection strategy for microgrids,’’ IET Gener., Transmiss. Distrib., vol. 14, no. 7,
systems—A review,’’ Energies, vol. 13, no. 9, p. 2149, May 2020, doi: pp. 1177–1185, Apr. 2020, doi: 10.1049/iet-gtd.2018.7049.
10.3390/en13092149. [21] M. Manohar, E. Koley, S. Ghosh, D. K. Mohanta, and R. C. Bansal,
[3] H. R. Baghaee, D. Mlakic, S. Nikolovski, and T. Dragicevic, ‘‘Support ‘‘Spatio-temporal information based protection scheme for PV integrated
vector machine-based islanding and grid fault detection in active distribu- microgrid under solar irradiance intermittency using deep convolutional
tion networks,’’ IEEE J. Emerg. Sel. Topics Power Electron., vol. 8, no. 3, neural network,’’ Int. J. Electr. Power Energy Syst., vol. 116, Mar. 2020,
pp. 2385–2403, Sep. 2020, doi: 10.1109/JESTPE.2019.2916621. Art. no. 105576, doi: 10.1016/j.ijepes.2019.105576.
[4] R. Tiwari, R. K. Singh, and N. K. Choudhary, ‘‘Coordination of dual [22] S. Jamali and S. Ranjbar, ‘‘Phase selective protection in microgrids
setting overcurrent relays in microgrid with optimally determined relay using combined data mining and modal decomposition method,’’ Int.
characteristics for dual operating modes,’’ Protection Control Modern J. Electr. Power Energy Syst., vol. 128, Jun. 2021, Art. no. 106727, doi:
Power Syst., vol. 7, no. 1, pp. 1–12, Dec. 2022, doi: 10.1186/s41601-022- 10.1016/j.ijepes.2020.106727.
00226-1.
[23] M. Biswal, C. Durga Prasad, P. Ray, and N. Kishor, ‘‘Modified complete
[5] F. Mumtaz, K. Imran, A. Abusorrah, and S. B. A. Bukhari, ‘‘Harmonic
ensemble empirical mode decomposition based HIF detection approach for
content-based protection method for microgrids via 1-dimensional recur-
microgrid system,’’ Int. J. Electr. Power Energy Syst., vol. 141, Oct. 2022,
sive median filtering algorithm,’’ Sustainability, vol. 15, no. 1, p. 164,
Art. no. 108254, doi: 10.1016/j.ijepes.2022.108254.
Dec. 2022, doi: 10.3390/su15010164.
[24] A. Srivastava and S. K. Parida, ‘‘A robust fault detection and location
[6] F. Mumtaz, K. Imran, A. Abusorrah, and S. B. A. Bukhari, ‘‘An extensive
prediction module using support vector machine and Gaussian process
overview of islanding detection strategies of active distributed genera-
regression for AC microgrid,’’ IEEE Trans. Ind. Appl., vol. 58, no. 1,
tions in sustainable microgrids,’’ Sustainability, vol. 15, no. 5, p. 4456,
pp. 930–939, Jan. 2022, doi: 10.1109/TIA.2021.3129982.
Mar. 01, 2023, doi: 10.3390/su15054456.
[7] S. Beheshtaein, R. Cuzner, M. Savaghebi, S. Golestan, and J. M. Guerrero, [25] J. Hu, Z. Liu, J. Chen, W. Hu, Z. Zhang, and Z. Chen, ‘‘A novel
‘‘Fault location in microgrids: A communication-based high-frequency deep learning–based fault diagnosis algorithm for preventing protection
impedance approach,’’ IET Gener., Transmiss. Distrib., vol. 13, no. 8, malfunction,’’ Int. J. Electr. Power Energy Syst., vol. 144, Jan. 2023,
pp. 1229–1237, Apr. 2019. Art. no. 108622, doi: 10.1016/j.ijepes.2022.108622.
[8] M. Manohar, E. Koley, and S. Ghosh, ‘‘Microgrid protection against [26] F. Mumtaz, H. H. Khan, A. Zafar, M. U. Ali, and K. Imran, ‘‘A state-
high impedance faults with robustness to harmonic intrusion and weather observer-based protection scheme for AC microgrids with recurrent neural
intermittency,’’ IET Renew. Power Gener., vol. 15, no. 11, pp. 2325–2339, network assistance,’’ Energies, vol. 15, no. 22, p. 8512, Nov. 2022, doi:
Aug. 2021, doi: 10.1049/rpg2.12167. 10.3390/en15228512.
[9] S. M. Malik, X. Ai, Y. Sun, C. Zhengqi, and Z. Shupeng, ‘‘Voltage and [27] S. Jamali, A. Bahmanyar, and S. Ranjbar, ‘‘Hybrid classifier for fault
frequency control strategies of hybrid AC/DC microgrid: A review,’’ IET location in active distribution networks,’’ Protection Control Mod. Power
Gener., Transmiss. Distrib., vol. 11, no. 2, pp. 303–313, Jan. 26, 2017, doi: Syst., vol. 5, no. 1, Dec. 2020, doi: 10.1186/s41601-020-00162-y.
10.1049/iet-gtd.2016.0791. [28] D. A. Etingov, P. Zhang, Z. Tang, and Y. Zhou, ‘‘AI-enabled traveling wave
[10] F. Nejabatkhah, Y. W. Li, H. Liang, and R. Reza Ahrabi, ‘‘Cyber-security of protection for microgrids,’’ Electr. Power Syst. Res., vol. 210, Sep. 2022,
smart microgrids: A survey,’’ Energies, vol. 14, no. 1, p. 27, Dec. 01, 2020, Art. no. 108078, doi: 10.1016/j.epsr.2022.108078.
doi: 10.3390/en14010027. [29] F. Mumtaz, K. Imran, S. B. A. Bukhari, K. K. Mehmood, A. Abusorrah,
[11] B. Mahamedi and J. E. Fletcher, ‘‘Trends in the protection of inverter- M. A. Shah, and S. A. A. Kazmi, ‘‘A Kalman filter-based protection strat-
based microgrids,’’ IET Gener., Transmiss. Distrib., vol. 13, no. 20, egy for microgrids,’’ IEEE Access, vol. 10, pp. 73243–73256, 2022, doi:
pp. 4511–4522, Oct. 22, 2019, doi: 10.1049/iet-gtd.2019.0808. 10.1109/ACCESS.2022.3190078.

12456 VOLUME 12, 2024

F. Mumtaz et al.: Hardware Supported Fault Detection and Localization Method

[30] W. Zhou and J. Hou, ‘‘A new adaptive robust unscented Kalman filter HABIBUR REHMAN (Member, IEEE) received
for improving the accuracy of target tracking,’’ IEEE Access, vol. 7, the B.Sc. degree in electrical engineering from the
pp. 77476–77489, 2019, doi: 10.1109/ACCESS.2019.2921794. University of Engineering and Technology, Lahore
[31] J. Qi, K. Sun, J. Wang, and H. Liu, ‘‘Dynamic state estimation for (UET Lahore), Pakistan, in 1990, and the M.S. and
multi-machine power system by unscented Kalman filter with enhanced Ph.D. degrees in electrical engineering from The
numerical stability,’’ IEEE Trans. Smart Grid, vol. 9, no. 2, pp. 1184–1196, Ohio State University, Columbus, OH, USA, in
Mar. 2018, doi: 10.1109/TSG.2016.2580584. 1995 and 2001, respectively. He has wide expe-
[32] M. A. Shah, S. B. A. Bukhari, K. Imran, K. K. Mehmood, F. Mumtaz,
rience in the areas of power electronics, motor
A. Abusorrah, S. A. A. Kazmi, and A. Wadood, ‘‘High speed protection
drives, and power systems in both industry and
of medium voltage DC distribution system using modified mathematical
morphology,’’ IET Renew. Power Gener., vol. 16, no. 14, pp. 3134–3148,
academia. From 1998 to 1999, he was a Design
Oct. 2022, doi: 10.1049/rpg2.12564. Engineer with Ecostar Electric Drive Systems and Ford Research Laboratory,
where he was a member of the Electric, Hybrid, and Fuel Cell Vehicle
Development Programs. From 2001 to 2006, he was with the Department
of Electrical Engineering, United Arab Emirates University, Al Ain, United
Arab Emirates, as an Assistant Professor. In 2006, he joined the Department
of Electrical Engineering, American University of Sharjah, where he is
FAISAL MUMTAZ received the B.S. degree in
currently a Professor. His primary research interests include power elec-
electrical engineering power from the Univer-
tronics and their application to power systems, adjustable-speed drives, and
sity of Wah, Pakistan, in 2015, and the M.S.
alternative energy vehicles.
degree in electrical engineering power from the
U.S. Pakistan Center for Advanced Studies in
Energy, National University of Sciences and Tech-
nology (NUST), Islamabad, Pakistan, in 2021,
where he is currently pursuing the Ph.D. degree.
From 2017 to 2018, he was an Instructor with the
Department of Electrical Engineering, Swedish
Institute of Science and Technology. His research interests include machine
learning, microgrid protection, state estimation, distributed generation, dis-
crete signal processing, and system identification. He received the USAID
fully-funded scholarship for the master’s degree.

KASHIF IMRAN received the B.Sc. and M.Sc.

degrees in electrical engineering from the Uni-
versity of Engineering and Technology, Lahore
(UET Lahore), in 2006 and 2008, respectively,
and the Ph.D. degree in electrical engineering
from the University of Strathclyde, in 2015.
From 2006 to 2007, he was with the Transmis-
sion and Distribution Division, SIEMENS, and
the Power Distribution Design Section, NESPAK. HAMMAD ALI QURESHI received the B.S.
He was a Faculty Member of UET Lahore and degree in electrical engineering from Air Univer-
COMSATS University Lahore. From 2018 to 2021, he was the Inaugural sity, Islamabad, in 2019. He is currently pursuing
Head of the Department of Electrical Power Engineering, U.S. Pakistan the M.S. degree in electrical power engineering
Center for Advanced Studies in Energy (USPCSC-E), National University from the National University of Sciences and
of Sciences and Technology (NUST), Islamabad, where he is currently Technology (NUST), Islamabad, Pakistan. His
an Associate Professor. His primary research interests include energy pol- research interests include electric vehicle chargers
icy, optimization for power systems and electricity markets operations and and the integration of renewable energy into elec-
planning, and power system protection. He received the Commonwealth tric vehicle charging stations.
Scholarship for the Ph.D. degree.

VOLUME 12, 2024 12457