Optimum Relaying Scheme of High Impedance

Fault Detection in Micro-Grid

Mr. Deepak Kumar Dash

School of Electrical Engineering Mrs. Tapaswini Biswal Dr. Sarat Chandra Swain
KIIT Deemed to be university School of Electrical Engineering School of Electrical Engineering
Bhubaneswar ,India KIIT Deemed to be university KIIT Deemed to be university
[email protected] Bhubaneswar, India Bhubaneswar, India
[email protected] [email protected]

Abstract—High impedance fault in micro-grid is not like asphalt, grass, concrete, sand etc. frequently then Hi-Z
encountered by typical over current protection, such as over fault occurs. As per the nature of Hi-Z fault, the fault current
current relays and fuses due to nonlinear, random, asymmetry constrains from mA to 75 A as explored in [3]. The
and low magnitude of fault current. Low impedance short conventional over current relay at the distribution side fails
circuit leads to enormous fault currents which are detectable to operate and unable to detect the low range current. The
by typical over current protection in the distribution feeder HIF current is arcing in nature that results fire hazard and
however Hi-Z fault current raise current magnitude in the 0 risk to the human life. Several researches have done to detect
ampere to 200 ampere range which triggers the over-current the HIF by using overcurrent relay. But the success rate is
relay to erroneous operation. One of the motivations of this
only below 20%. So, the detection and analysis of HIF and
research is the hazard produced when a conductor falls on the
ground or a tree contacts a line; or when a HIF occurs. In this
triggering of circuit breaker is very challenging task of every
paper, high impedance fault characteristics is analyzed and engineer in recent time.
various techniques are compared to develop a detection The detection and isolation of HIF occurring in multi
function and presented a solution and it is modeled using feeder radial distribution system have analyzed by
PSCAD Simulink tool for deploying proper protection scheme comparing residual voltage and -ve sequence current flowing
in Micro-Grid. in different feeder. The performance of the HIF detection
model is tested with different case study like balanced,
Keywords—Micro-grid, HIF, over current protection
unbalanced, and harmonic loads [4]. In paper [5]
I. INTRODUCTION Accumulated Absolute Difference method is used. In this
method, the cycle per cycle difference approach is used to
To make the unified electricity system, we are moving determine the non-harmonic content and randomness. The
towards the modern and small-scale version of electricity residual power is the key factor which is compared with the
system i.e. Micro-grid, which has excellent capability of reference voltage to detect the high impedance fault. HIF
generating, distributing and regulating the flow of electricity detection is analysed [6] by means of data mining Maximal
to consumers to a circumscribed level. Micro-grid needs an Overlapping Discrete Wavelet Transform (MODWT) and
integration of renewable resources in the cordiality level. Decision tree (DT). DT algorithm requisites mainly two
Integration of Distributed generation (DG) in Microgrid coefficients i.e., details and approximation which are
leads to the detrimental in voltage quality. In paper [1], obtained from the decomposition of the fault signal. The HIF
several drawbacks have made like voltage sag, voltage swell solution is addressed by Hilbert transform [7]. This method
leads to unhealthy voltage profile. The harmonic analysis is possesses the instantaneous current of each phase. The
constructed to know the fault current distortion. This paper detector ‘d’ is calculated and compared with the threshold
shows the optimum method which turns to structure and type value trigger the fault event from no fault event.
of DGs in Microgrid to reform the voltage quality. This study
focuses the installation of Distributed Generation in The use of smart meter and the calculation of Second
Microgrid which improves the system response to voltage order Harmonics Index (SHI) at the node voltage raise the
distortion. The micro-grid can operate in two modes i.e. grid high-Z fault event and detect the abnormal event in the
connected mode and islanded mode. The islanded mode is power distribution system [8]. The detection of high
independent from the micro-grid but in grid connected mode impedance is done by frequency change event. The sample
the micro-grid is connected with the main supply [2]. The data of fault event is analysed by normal Gaussian
integration of micro-grid raises the power quality issues such distribution curve (Bell Curve) as described in the paper [9].
as voltage fluctuation, voltage sag, voltage swell and impact The nominal healthy frequency is maintained by rejecting the
of large capacitor switching and harmonic to the bus voltage frequency of the HIF event. The rejection region is decided
profile. To attenuate voltage fluctuation and to improve by the frequency distribution of the HIF event. In paper [10]
voltage profile and to upgrade harmonic distortion, both Fuzzy Adaptive Network Interface System (ANFIS) method
rotating type DGs and inverting type DGs are scrutinized in is deployed to spot Hi-Z fault in a medium voltage network.
paper [1].
The above method has potential to detect some Hi-Z fault
The energized conductor in the distribution line generally in distribution system but restricted to all. The method that is
comes in contact with Hi-Z surface. The Hi-Z surface is used, have more computational complexity [10]. Discrete
categorized like trees, wood barricades, automobile etc. wavelet Transform and Decision Tree (Tree) based
When the conductor ruptures and contacts with Hi-Z surface algorithms are proposed in this paper to give optimum

XXX-X-XXXX-XXXX-X/XX/$XX.00 ©20XX IEEE

Authorized licensed use limited to: University of Canberra. Downloaded on June 25,2020 at 13:40:06 UTC from IEEE Xplore. Restrictions apply.
relaying scheme of relay. The new data mining technique i. DG1, DG2: 2.4kV,2kA,50Hz, Vbase=11kV
and classification learner are also used to separate the fault ii. DG3:575V,0.175kA,50Hz, Vbase=11kV
event from no fault event. ➢ Transformer (TRs):
i. TR1:15MVA,33/11 kV, 50Hz, Vbase=11kV
II. FEATURES OF HIGH IMPEDANCE FAULT CURRENT
ii. TR2, TR3:12MVA, 2.4/11 kV, 50Hz, Vbase=11kV
iii. TR4:2.5 MVA, 0.575/11 kV, 50Hz, Vbase=11kV
The surface contact is the key factor to decide the Hi-Z fault ➢ Distribution Lines:
current which may range from 10Ato 50A [11]. The DL1, DL2, DL3, DL4 are PI type
overcurrent relay faces difficulty to detect the lower range Length=2km, 50Hz, Vbase=11kV,
fault current. The probability of HIF on single phase to R1=0.3e-3[ohm/m], Xl1=0.25e-3[ohm/m],
ground is more than another fault event. If the comparison is Xc1=280.1[Mohm*m], Ro=0.75e-3[ohm/m],
made between HIF and another fault event, then arc Xlo=1.1740e-3[ohm/m],
production during HIF is smaller [12].
Xco=461.2546[Mohm*m]
The common prototype of high impedance fault is given ➢ Total Loading:
below. 10 MW/ph, 5 MVAR/ph (Inductive Loading)
⚫ Build-up –The build-up of fault current to its
optimum value with in ten second because the DG1 and DG2 are the distributed generation like, hydro
contact is made between the surface of the power plant or diesel generator whereas DG3 is a wind
conductor and ground. turbine. The proposed model tells about the Micro-grid used
in a grid connected mode. The impact of 3 DGs on the main
⚫ Non-linearity – V-I characteristics as shown in grid (Utility) current is studied on this paper. This paper
figure 1 represent the non-linear nature which will focuses the bidirectional power flow from supply side to
converge to instability. local load where local load is shared by all DGs. The
⚫ Asymmetry – Both +ve and -ve half of the fault parameters which are shown and discussed in this paper are
current magnitude is not symmetrical as shown in expressed in terms of Positive Sequence magnitude (PM)
figure 1. which is obtained from the decomposition of three phase
quantity.
⚫ Iintermittence – During fault event the contact is
disconnected up to few cycles however it is again
connected with ground in the coming cycle which
TR2
leads to ionization and arc formation during these 11/2.4 KV
cycles. PCC
B-2
B-1
⚫ Harmonic–Fault current consists of both harmonic DG-1
DL-1
and non-harmonic content as per the nature.
TR3
⚫ Frequencies of high value –The frequency TR1
B-3 L2 B-4
11/0.575
KV
component in the fault current ranges from 2 to 10 UTILITY

kHz which can be decompose for analysis. DL-2 DL-3

33 KV DG-3
B-5
33/11 KV L3
L1 DL-4 HIF

L4
L5
11/2.4 KV

TR-4

DG-2

Figure 2: Studied micro-grid with HIF

Figure 1: HIF current Characteristics
A. HIF MODEL:

III. SYSTEM STUDIED The HIF model used as shown in figure 2 is correspondent
to authentic fault in power distribution system. This model
represents a line to ground fault (L-G Fault) in the
Figure 2 represents the trial microgrid system for the
distribution feeder. The circuit visualizes a dc voltage
recommended HIF detection. The details of the test system source Vp, a diode Dp and Rp which is a variable resister
are given as follows: connected in one branch and in another branch the same
➢ Utility: f=50 Hz, rated kV=33, Vbase = 11 kV pattern repeats like a voltage source Vn, diode Dn and
➢ Distributed generation: variable resistance Rn are linked in series pattern. Rp and

Authorized licensed use limited to: University of Canberra. Downloaded on June 25,2020 at 13:40:06 UTC from IEEE Xplore. Restrictions apply.
Rn represent the resistance of the tree and/or earth
resistance. The magnitude of dc voltage is set as per the
phase voltage of the studied system. When the phase voltage
is greater than Vp, fault current starts to flow towards the
ground. In order to restrain the reverse current from negative
polarity of DC voltage, Vn must be less than Vph and Vph
must be less than Vp, i.e. Vn<Vph<Vp. This condition does
not allow the reverse flow of current. The value of the
resistors Rp and Rn during HI-Z fault are varied from 300
ohm to 400 ohms at every 2 ms interruption in such a way
that the fault current magnitude consistently less than the
5% of the rated load. The indiscriminate deviation possesses
asymmetry in the fault current. Figure 5: Fault voltage characteristics

HIF MODEL

CB 13

Rn
Rp
Figure 6: V-I characteristics of HIF
Dp Dn

Vp Vn

Figure 3.HIF Model

Figure 3 shows the HIF occurred between phase “R” and

ground at bus-5. Fault voltage and fault current are
measured by using specified measuring devices. The voltage
verses current attributes at the fault point is shown in figure-
6. The non-linearity behaviour can be seen in V-I
characteristic of HIF. Figure 4 and figure 5 show the fault
current and fault voltage characteristics where the Figure 7: Current at different buses during HIF
asymmetry and intermittence attribute are studied and
visualized.

Figure 8: Voltage at different buses during HIF

B. COMPARISON OF UTILITY CURRENT WITHOUT DG:

When micro-grid is serving the regional load with utility
Figure 4: Fault Current characteristics (grid connected load), all three distributed generation is

Authorized licensed use limited to: University of Canberra. Downloaded on June 25,2020 at 13:40:06 UTC from IEEE Xplore. Restrictions apply.
isolated from the Micro-grid by setting the circuit breaker Igr is the vector sum of all three-phase current as shown in
time. The local load at the different buses draws the current figure 10 and E1r is the vector sum of all three-phase as
from the main utility grid as shown in figure 14 (i.e. IgWDG shown in figure 11.
-kA). This figure shows the comparison of utility current in
case of with Distributed generation and without Distributed
generation. The figure indicates the clear difference between
the decreases in grid current which is accomplished by all
three DGs. The regional load is shared by the DGs which
results the decrease in utility current. As Hi-Z fault is
applied between 2 sec-2.20 sec the small change in grid
current can be visualized both the cases due fault event.

Figure 10: Residual current at Utility side

Figure 9: Comparison of Utility current

Figure 11: Residual voltage at Utility side
The sharing of load depends upon the type of DGs which
are connected with the Micro-grid. As described in figure 2,
The above graph visualizes a clear distinguish between the
Two DGs are small hydro generator and one DG is wind
fault event and no-fault event. In case of no-fault event the
turbine. All DGs are dependent on seasonal parameter like
residual current is zero and there is the transition of residual
water level, wind flow. So, the power delivered by the DGs
current in case of fault event (2 sec- 2.20 sec) as shown in
is not constant and we can’t connect on the distribution
figure 10.
generation on average load duty cycle basis.
B. CWT Analysis:
IV. ANALYSIS AND OPTIMUM RELAYING SCHEME FOR HIF
Continuous wavelet Transform analysis of the residual
voltage is done by using suitable wavelet at sampling
As fault occurs in bus-5 which is last bus of the Micro-grid, frequency 1 Hz as shown in figure 13. The figure 12 is the
the I5 and I4 are raised till the competition of fault. This can Time frequency analysis of the distorted residual voltage
be shown by figure 7. The asymmetry, non-linearity and where the fault event can be easily distinguished by
intermittence characteristics of fault current and fault observing the power frequency band. The white spot in the
voltage are shown in figure 4 and figure 5 respectively. figure 13 indicates the HIF event which affects the bus-1
Similarly, the voltage profile of each bus is distorted during voltage E1 at the utility side.
the fault event. Clear voltage sag is observed at E5 and E4
during HIF in figure 8. Approaching to the relaying scheme,
the relay is connected at the main utility grid. This relay
should response to the fault event however the transition of
utility current Ig during HIF is not feasible for relaying
application as shown in figure 7. We know that the fault
current of HIF is predominantly very less and leads to the
maloperation of relay at utility side. So, we are going for
residual current and residual voltage calculation for
approaching better relaying scheme.

A. Residual voltage and current analysis:

Residual voltage and residual current of utility side i.e. Ig,

E1 are done by summing all the three-phase quantity.
Igr = Ig_a + Ig_b + Ig_c
E1r = E1_a + E1_b + E1_c

Figure 12: Time frequency analysis of E1r using spectrogram

Authorized licensed use limited to: University of Canberra. Downloaded on June 25,2020 at 13:40:06 UTC from IEEE Xplore. Restrictions apply.
 The decomposition is done by choosing mother wavelet
doubesis. The decomposition tree and the level-6
approximation are shown by figure 15. The level-6
approximation coefficient a-6 is the major factor for HIF
detection.

Figure 13: Fine scale analysis of E1r by using CWT.

C. DWT analysis: Figure 15: Approximation at level 6.

D. DT Algorithm and classification learning:

CWT is only visualizing the fault event. To give proper
algorithm to detect the fault event we are going for DWT
Decision Tree (DT) technique is an excellent datamining
analysis. DWT (Discrete Wavelet Transform) is the
method for constituting classification systems based on
powerful tool which decompose the fault signal i.e. E1r.
numerous covariates or for establishing forecasting
Original signal is decompose to a level-6 with a suitable
algorithm for a destination variable. This procedure
mother wavelet as shown in figure 14.
organizes a population in to branch like portion that
construct a tree which is inverted type with a root node, leaf
nodes an internal node. The algorithm is non-parametric and
can accurately deal with dataset which is large and has more
complexity. For a greater number of sample size, analysis is
done by dividing the study data into training and validation
dataset. Here we are using the classification of residual
voltage E1r and the level-6 coefficient. The fault event can
be detected by DT algorithm as shown in figure 15.

Figure 14: Decompose signal by using discrete wavelet transform

Figure 15: Classification tree viewer with 4 pruning level.

Authorized licensed use limited to: University of Canberra. Downloaded on June 25,2020 at 13:40:06 UTC from IEEE Xplore. Restrictions apply.
The end node indicates the final conclusion of the fault V. CONCLUSION
event. Decision tree frames classification and regression
models in the form of a tree structure. This process
decompose dataset in to smaller subsets while during An optimum HIF detection methodology based on DWT
classification process the decision tree is equally developed. and DT are presented. Residual voltage waveform is used to
A complete tree consists of decision nodes and leaf nodes. detect HIF in the Micro-grid Architecture. The proposed
Final result is decided by the leaf nodes. Root node is the method can clearly eliminate the HIF from other transient
best predictor in a tree which is denoted by the topmost condition. The conclusion of the proposed method have
decision node. Leaf node represents the classification or excellent performance in quick detection, Secure and
decision node. Figure 16 shows the scatter plot of the dependable under diverse transient condition.
extracted data. Scatter plot shows the model prediction data
after training.

REFERENCES

[1] Wei Deng, Wei Pei and Zhiping Qi, "Impact and improvement of
Distributed Generation on voltage quality in Micro-grid," 2008 Third
International Conference on Electric Utility Deregulation and
Restructuring and Power Technologies, Nanjing, 2008, pp. 1737-
1741.
[2] Kar, S. and Samantaray, S. R.: ‘Time-frequency transform–based
differential scheme for microgrid protection’, IET Genr Transm &
Distrib., 2013, 8, (2), pp.310-320.
[3] J. Tengdin, R. Westfall, and K. Stephan, “Impedance fault detection
technology”, Mar. 1996, Report of PSRC Working Group D15.
Available:http://www.ppsrc.org/Reports/High_Impedance_Fault_Det
ection_Technology.pdf.
[4] K. Sarwagya, S. De and P. K. Nayak, "High-impedance fault
detection in electrical power distribution systems using moving sum
Figure 16: Scatter plot of extracted data approach," in IET Science, Measurement & Technology, vol. 12, no.
1, pp. 1-8, 1 2018.
The dark band indicates the predicted no fault event and the [5] A. P. Kujur and T. Biswal, "Detection of high impedance fault in
distribution system considering distributed generation," 2017
scattered dot indicates the fault event. The confusion matrix International Conference on Innovative Mechanisms for Industry
as shown in figure 17 which indicates the classification of Applications (ICIMIA), Bangalore, 2017, pp. 406-410.
data mining during fault event. The classification shows the [6] S. Kar and S. R. Samantaray, "High impedance fault detection in
clear distinction between true positive rate and false microgrid using maximal overlapping discrete wavelet transform and
negative rate. These is a confusion matrix of 2*2 which decision tree," 2016 International Conference on Electrical Power and
Energy Systems (ICEPES), Bhopal, 2016, pp. 258-263.
indicates that the difference between predicted class and true
[7] Premlata Jena and Ashok Kumar Pradhan, “Detection of High
class. The fault event we have classified and predicted and Impedance Fault”, IEEE- Inter. Conf., January 2012.
the actual fault event we get after training the data is 99% [8] A. Radhakrishnan and S. Das, "Location of High Impedance Faults
with false negative rate of 1%. using Smart Meters in Distribution Systems with DGs, Power
Electronic Loads and Electric Arc Furnaces," 2019 IEEE Milan
PowerTech, Milan, Italy, 2019, pp. 1-6.
[9] S. Kantra, H. A. Abdelsalam and E. B. Makram, "Application of
PMU to detect high impedance fault using statistical analysis," 2016
IEEE Power and Energy Society General Meeting (PESGM), Boston,
MA, 2016, pp. 1-5.
[10] A. Bouricha, T. Bouthiba, R. Boukhari and S. Seghir, "High
Impedance Faults Location in the Distribution Networks using
Adaptive Neuro-Fuzzy Inference System," 2018 International
Conference on Electrical Sciences and Technologies in Maghreb
(CISTEM), Algiers, 2018, pp. 1-5.
[11] B. Aucoin and R. Jones. "High impedance fault detection
implementation issues," IEEE Transactions on Power Delivery, vol.
11, no. 1, pp. 139 –148, jan 1996.
[12] D. I. Jeerings and J. R. Linders. "Ground Resistance - Revisited,"
Figure 17: Confusion matrix of extracted data. IEEE Transactions on Power Delivery, vol. 4, no. 2, pp. 949–956,
April 1989.

Authorized licensed use limited to: University of Canberra. Downloaded on June 25,2020 at 13:40:06 UTC from IEEE Xplore. Restrictions apply.