2021 International Symposium of Asian Control Association on

Intelligent Robotics and Industrial Automation (IRIA)

Indian Institute of Technology, Goa, India. 20-22 September 2021
2021 International Symposium of Asian Control Association on Intelligent Robotics and Industrial Automation (IRIA) | 978-1-6654-3323-5/21/$31.00 ©2021 IEEE | DOI: 10.1109/IRIA53009.2021.9588709

Comparison of Phasor Estimation Techniques for

AC Microgrid Protection
Rudranarayan Pradhan Premalata Jena
Electrical Engineering Electrical Engineering Department
IIT Roorkee, India IIT Roorkee, India
r [email protected] [email protected]

Abstract—Microgrid(MG) provides numerous advantages such

as reduced transmission losses, low carbon emission, improved
system reliability. However, for the proper functioning of the
MG, it should be protected from different fault conditions.
Detection of fault conditions in MGs is challenging because the
fault current magnitude varies in various modes of microgrid
operations, types of DGs used, ratings of DGs, and bidirectional
fault current flow. This article presents a signal processing-
based technique to determine the faults in the microgrids and
state-of-the-art latest research on microgrid fault detection.
This paper analyzes different phasor estimation techniques, for
instance, Discrete Fourier Transform Technique (DFT), Kalman
Filtering Technique (KFT), and Least Square Technique(LST)
for microgrid fault detection. The pros and cons of DFT, KFT,
and LST are analyzed. The microgrid with various types of DGs
is simulated using RTDS/RSCAD (Real-time digital simulator), Fig. 1. The layout of an AC MG systems.
and the Least Square Technique can effectively detect different
fault conditions in the microgrid.
Index Terms—Microgrids, Distributed Generations(DGs),fault
on the distribution system’s type, MGs are primarily divided
detection, protection techniques, RTDS/RSCAD, DFT, Kalman
filtering,Least Square Technique,Faults. into three groups, i.e., AC MGs, DC MGs, and hybrid(AC/DC)
MGs. When a distribution system operates as a microgrid
I. I NTRODUCTION that has sufficient DERs and are placed close to the load
centers, it can provide power to the essential loads in case
With the ever-increasing need for electrical power and to
of faults in the system and reduces transmission losses. DERs
reduce the pollution level due to the conventional power plant,
are mainly classified into two types, i.e., electronically coupled
there is a need to increase the integration of distributed energy
and rotating machine coupled.
sources to traditional power plants, which are mostly thermal-
Even though the microgrid works at the distribution level,
based power plants. A microgrid is an interconnection of
traditional protection systems such as over current relays are
different distributed energy sources and loads which act as a
ineffective in protecting microgrids for the following reasons
single controllable unit concerning the grid and are contained
[1], [2].
within well defined electrical limits. Microgrids promote the
integration of different types of renewable energy sources and • The fault current magnitude varies significantly during

a variety of loads. Microgrid finds a wide range of applications grid-connected and islanded microgrid operation.
in enabling the integration of distributed energy sources, • The DERs size and configuration add challenges to

energy surety, reducing peak load demand and the availability the conventional protection schemes of the distribution
of other grid services like demand response management, an- network. For example, inverter-based DERs have a fault
cillary services,intentional is-landing for safety and reliability, current of (1.2∼2) times, while synchronous-based DG
add to local reliability and appreciate power quality, smoothing units have a fault current of (4∼10) times the standard
of intermittent and variable resources [1].Fig. 1 depicts the current.
layout of the AC MG system. The protection engineer in microgrids faces a difficult task
A microgrid can operate in grid-connected or island-mode due to the low fault current magnitude caused by high resis-
operations by connecting and disconnecting it from the pri- tance fault detection and inaccurate zone recognition. Most of
mary grid [2]. Microgrids can so operate either interconnected the relays in the distribution system are overcurrent relays, and
to the grid or islanded in the event of a grid failure. Depending to overcome the issues of bidirectional power flow, directional

978-1-6654-3323-5/21/$31.00 ©2021 IEEE 31

Authorized licensed use limited to: COLLEGE OF ENGINEERING AND TECHNOLOGY. Downloaded on April 29,2023 at 06:55:09 UTC from IEEE Xplore. Restrictions apply.
 overcurrent relays should be used. As fault current magnitude
in the islanded mode of operation is significantly less when
compared with grid-connected mode, the relay setting must be
updated on the way of operation.
In [3], the inverter output current is monitored by using
a transient monitoring function to detect the fault. The least
square technique calculates the fundamental component of
the VSC current waveform. In case of fault the actual signal
deviates from the reconstructed signal. The window’s length
determines the system’s accuracy, but the longer window
reduces the process speed and response time. By properly
designing the TMF value, fault condition and load switching
conditions can be determined.
In [4], a novel fault technique for the inverter-based islanded
microgrid is proposed. First, the output currents are limited
by the current limiting strategy by using the hysteresis control
approach and root mean square control technique to protect
Fig. 2. A Seven-bus microgrid test system
inverters against the short circuit. Then a fifth harmonic current
is injected into the fault current. Finally, the fault current
versus fault impedance droop is produced in the inverter to
provide considerable current across the output. II. DESCRIPTION OF THE MICROGRID SYSTEM

In [5],an active distribution network with impedance dif- RSCAD software is used to design real-time simulations
ferential protection is suggested.The amplitude difference be- of a medium voltage microgrid in [7]. The medium voltage
tween differential and restraint impedance determines the microgrid consists of a photovoltaic system, a synchronous
faulty site. Simultaneously, the present amplitude difference generator, a wind turbine with DFIG is shown in Fig. 2. During
is used to build an auxiliary criterion to handle the dead- the transition from grid-connected mode to islanded operation
zone effect. In particular, this research examines the impedance mode, a switch capacitor bank is utilized to restore power
differential protection’s protection performance in the presence balance and control voltage. The microgrid can function as a
of different influencing conditions such as CT saturation, redial network with switches S2 and S3 open and serve as a
transition resistances, and unsynchronized data. mesh network when both the switches are closed. When switch
In [6], Integrated Impedance Angel(IIA) is used for mi- S1 is turned on, the microgrid shifts into islanded mode. The
crogrid fault detection which is calculated using the positive PV in this microgrid is1.74 MW with two-level VSC to convert
sequence component of voltage and current information col- DC power from the solar PV to AC power to be injected
lected from the PMU. For internal fault conditions (near the to the grid. The PV system’s control system employs MPPT
DG, middle of the line, and away from the DG or excluding algorithms to maintain the DC voltage at its output power. The
DG), the IIA is negative. For external fault conditions, includ- wind power is of 2 MW rating is equipped with DFIG [8], the
ing DG or excluding DG, the IIA is positive. For security turbine and rotor blades transforms the kinetic energy of the
reasons, −50 to 50 is kept as no operating zone and validated wind to mechanical power which is given by
with several test cases. For internal fault, the IIA is −50 to
−900 , and with this setting, proper fault detection is achieved Pm = 0.5Cp ρAVw3 (1)
in microgrid.
where Cp represents the blade power coefficient, which varies
This paper presents the Least Square(LS) Technique to
from (0.2-0.5)
determine the faults in the AC microgrid. The seven bus AC
microgrid modeled in RSCAD software consists of solar PV, ρ denotes the air density (kg/m3 )
wind generation, DG. Different phasor estimation techniques A represents rotor swept area (m2 )
are studied for microgrid fault detection, and the LS techniques Vw express wind velocity (m/s).
prove the best results for fault detection. The content of this The diesel generator is of 3 MW rating, which is a diesel
report is as follows: Section II describes different phasor engine-driven asynchronous generator. The generator’s shaft
estimation techniques for microgrid fault detection. In section speed and frequency are maintained by the speed governor.To
III proposed technique is described. Section IV describes the meet the microgrid’s voltage magnitude and synchronization
results and discussion,and Section V brings the paper to a standards, a switchable capacitor bank with a rating of 0.5
conclusion. MVAR is fitted to bus B1.

32

Authorized licensed use limited to: COLLEGE OF ENGINEERING AND TECHNOLOGY. Downloaded on April 29,2023 at 06:55:09 UTC from IEEE Xplore. Restrictions apply.
 −Vimg
III. PROPOSED METHODOLOGY Vangle = tan−1 ( ) (7)
Vreal
Several fault detection techniques are explained in various
literature for microgrid fault detection. In this research, the The DFT technique is very simple and easy to calculate
phasor estimation technique detects the microgrid fault effec- the distortion of waveform and can have any integer num-
tively and accurately [9]. These techniques can be differenti- ber of samples in a data window however, the presence of
ated as follows decaying DC component, not suitable to analyze sub syn-
• Discrete Fourier Transform Technique
chronous frequencies,large memory requirement and extensive
N
• Half Cycle ( 2 ) DFT
computational time makes it unreliable for a broad range of
• Cosine Transform
applications.
• Recursive DFT
B. Kalman Filtering Technique
• Least Square Technique
• Wavelet transform Kalman filtering [11] is a methodology for estimating
• Kalman filtering Technique unknown variables based on measurements recorded over
This paper explains Discrete Fourier transform Tech- time.Kalman filters are generally simple and efficient and need
nique,Kalman filtering techniques and Least Square Technique minimal computing resources.
for microgrid fault detection. 1) Problem Statement: To determine states based on linear
dynamical systems in state space representation, Kalman filters
A. Discrete Fourier Transform Technique are utilised.
DFT presumes that any signal comprises its fundamental The state evolution from time t − 1 to t is described by the
frequency component as well as harmonics of that frequency process model as follows:
component. In DFT the signal is analyzed in a window manner
yt = F yt−1 + But−1 + St−1 (8)
which is defined for an interval of time. The DFT of an input
signal can be expressed as where F express the state transition matrix for the preceding
N −1 state vector yt−1 .
2 X 2πk
Vk = vn e−jn N (2) B denotes input for the vector control ut−1 ,st−1 is the process
N n=0 noise vector,and St−1 ∼ N (0, G) i.e.,assumed to zero-mean
v represents instantaneous voltage, Gaussian with the co variance G, which is supposed to be
n represents nth sample in specified data window, Gaussian with zero mean.
V represents phasor voltage, The process model defines the association between the mea-
k represents harmonic order, surement at present time step k and the state of a system as
N represents samples size in a particular data frame. At = Cyt + Nt (9)
In order to calculate the DC component k is taken as 0
and to calculate fundamental frequency of the waveform k is whereAt denotes the measurement vector
set to 1 and so forth. C marked as measurement matrix
The (2) can be expressed as Nt is the measurement noise
The objective of the Kalman filter is to determine yt for a
N −1 given time t, provided the first assessment of y0 , measuring
2 X 2πnk 2πnk
Vk = vn [cos( ) − jsin( )] (3) series, A1 , A2 , A3 , · · · , At and system data represented by F ,
N n=0 N N B, C, N and R. G and R are used to adjust tuning variables
to obtain the desired output.
The (3) can be further expanded to calculate the real and 2) Algorithm for Kalman filter: The two important steps of
imaginary part of the phasor as follows Kalman filter algorithm is prediction and update.
N −1 Prediction:
2 X 2πnk
Vk (real) = vn cos( ) (4) Predicted state estimate Ŷt− = F ŷt−1
+
+ But−1
N n=0 N − +
Predicted error covariance Dt = F Dt−1 FT + G
Update:
N −1
2 X 2πnk Measurement residual m̃t = At − H ŷt−
Vk (img) = vn sin( ) (5)
N n=0 N Kk = Dk− C T
Kalman Gain
×(R + CDk − C T )−1
The magnitude and phasor of the waveform can be calculated +
Predicted error covariance Dt − = F Dt−1 FT + G
q + −
Vmag = Vreal 2 2
+ Vimg (6) Update state estimate y˜t = y˜t + Kk m̃
Updated error covariance Dt+ = (I − Kk C)Dk−

33

Authorized licensed use limited to: COLLEGE OF ENGINEERING AND TECHNOLOGY. Downloaded on April 29,2023 at 06:55:09 UTC from IEEE Xplore. Restrictions apply.
 In the following expression, to represent updated and pre- Vn represents peak value of the nth harmonic component.
dicted estimates the superscripts ‘+’ and ‘-’ are used, re- θn represent nth harmonic component.
spectively. The operator ‘ˆ•’(hat) in expressions indicates an By excluding higher order components, the decaying DC
estimate of variable i.e., ŷ represents the estimate of y. The component can be represented in Taylor series expansion as
updated value of the preceding updated state estimate is used   N
to calculate the anticipated state estimate. D is the state error V0 X
v(t) = V0 − t+ Vn sin(nω0 t + θn ) (12)
covariance. The random variable y covariance is calculated as τ n=1

cov(y) = M[(y − ŷ)(y − ŷ)T ]T (10) With the assumption that the voltage is comprised of ex-
ponentially decaying dc component ,fundamental frequency
The above expression M represents the expected mean value component,with 2nd ,3rd ,4th and 5th harmonics component, the
of argument. (12) can be expressed for t = t1 as
During updated stage m̃t (measurement residual) is calcu-  
V0
lated, which can be get by differencing predicted measurement v(t1 ) = V0 − t
τ
(C yˆt − ) from the estimated measurement (C yˆt − ). In order
+ V1 sin(ω0 t1 + θ1 ) + V2 sin(2ω0 t1 + θ2 )
to evaluate the present measurement, the predicted state is
multiplied with the measurement matrix. After getting the + V3 sin(3ω0 t1 + θ3 ) + V4 sin(4ω0 t1 + θ4 )
updated state estimate the Kalman filter calculates the updated + V4 sin(4ω0 t1 + θ4 ) + V5 sin(5ω0 t1 + θ5 ) (13)
Dt+ (error covariance) which is used in the subsequent time
The above-mentioned (13) can alternatively be written as
step. An initialising stage is necessary for Kalman filter exe-  
cution. For first values, initial estimates are needed. For initial V0
v(t1 ) = V0 − t1
values, the initial assumption of ŷ0+ ( the state estimate first τ
assumption ) , D0+ (covariance matrix of the first assumption) + (V1 cos θ1 ) sin(ω0 t1 ) + (V1 sin θ1 ) cos(ω0 t1 )
are essential. The values of G, R, D0+ and plays a key + (V2 cos θ2 ) sin(2 ω0 t1 ) + V2 sin(θ2 ) cos(2 ω0 t1 )
role in achieving the necessary performance. Lastly the after
initializing all the estimates the Kalman filtering technique is + (V3 cos θ3 ) sin(3 ω0 t1 ) + (V3 sin(θ3 ) cos(3 ω0 t1 )
implemented by performing the prediction and update stage + (V4 cos θ4 ) sin(4 ω0 t1 ) + (V4 sin θ4 ) cos(4 ω0 t1 )
for every time step t = 1, 2, 3, · · · , n. + (V5 cos θ5 ) sin(5 ω0 t1 ) + (V5 sin θ5 ) cos(5 ω0 t1 )
KFT consider the variance of the initial state estimate and (14)
variance of the model order which is partially or entirely The (12) can be further simplified as
neglected by other filtering technique and provides information
about the quality of estimation and variance of the estimation y(t1 ) = c11 y1 + c12 y2 + c13 y3 + c14 y4
error however it can’t cope with the sudden and fast changes + c15 y5 + c16 y6 + c17 y7 + c18 y8
occurring in the power system and sluggish response and + c19 y9 + c110 y10 + c111 y11 + c112 y12 + .....
significant time constant in fault clearance.
.. + c1(2N +1) y(2N +1) + c1(2N +2) y(2N +2) (15)
C. Least Square Technique where y1 = V0 , y2 = −V0
y3 = V1 cos θ1 , y4 = V1 sin θ1 ,
τ ,
The least Error Squares (LES) technique [9] is mainly y5 = V2 cos θ2 , y6 = V2 sin θ2 , y7 = V3 cos θ3 ,
applied to determine the phasors of the fundamental and har- y8 = V3 sin θ3 , y9 = V4 cos θ4 , y10 = V4 sin θ4 ,
monics frequency component of voltage and current. It works y11 = V5 cos θ5 , y12 = V5 sin θ5 , · · · , y(2N +1) = VN cos θN ,
by reducing the mean square error between the actual and y(2N +2) = VN sin θN .
assumed wave forms.The fundamental frequency component,
an exponentially decaying dc component, and harmonics of c11 = 1, c12 = t1 , c13 = sin(ω0 t1 ), c14 = cos(ω0 t1 ),
defined orders are combined to mimic the voltage and/or c15 = sin(2ω0 t1 ), c16 = cos(2ω0 t1 ), c17 = sin(3ω0 t1 ),
current waveform. c18 = cos(3ω0 t1 ),c19 = sin(4ω0 t1 ), c111 = sin(5ω0 t1 ),
N c112 = cos(5ω0 t1 ),......c1(2N +1) = sin(N ω0 t1 ), c1(2N +2) =
v(t) = V0 e−( τ ) +
t X
Vn sin(nω0 t + θn ) (11) sin(N ω0 t1 )
n=1 If the signal is sampled every ∆t seconds, then (15) may
be expressed as by putting t1 = m∆t into the equation
where v(t) denotes instantaneous voltage at time t.
τ represents the decaying DC component’s time constant. y(m∆t) = cm1 y1 + cm2 y2 + cm3 y3 + cm4 y4
N denotes the highest level of harmonic component present + cm5 y5 + cm6 y6 + cm7 y7 + cm8 y8
in the waveform.
+ cm9 y9 + cm10 y10 + cm11 y11 + cm12 y12 + .....
ω0 is the fundamental frequency of the system.
V0 represent the magnitude of the DC offset when t = 0. .. + cm(2N +1) y(2N +1) + cm(2N +2) y(2N +2) (16)

34

Authorized licensed use limited to: COLLEGE OF ENGINEERING AND TECHNOLOGY. Downloaded on April 29,2023 at 06:55:09 UTC from IEEE Xplore. Restrictions apply.
 1
∆t = (17)
fs
where fs represents sampling frequency.m is sampling num-
ber. The coefficients c can be reformulate as
cm1 = 1, cm2 = m∆t,
cm(2i+1) = sin(iω0 m∆t), cm(2i+2) = cos(iω0 m∆t),
∀i = 1, 2, . . . , N.

By using (2N + 3) consecutive samples a total [(2N + 2) + 1]

equations can be formed formed similar to (17) and can be
represented as

[C](2N +3)×(2N +2) [X](2N +2)×1 = [v](2N +3)×1 (18)

In (18),N [= (P −2)
2 ] represent the highest order in the simu-
lated signal and P the sample size per cycle.
The (19) represents the least error squares estimate of [X].

[X] = [[C]T [C]]−1 [v] = [C]† [v] (19)

In (19) [C]† represents the left pseudo-inverse of [C].

Fig. 4. Magnitude and phasor for LG fault.
IV. RESULTS AND DISCUSSION
In RTDS/RSCAD environment, the medium voltage test
microgrid simulation is performed. The MATLAB software
is used for the analysis of simulation data.
Different types of fault is created in between bus no.1(B1)
and transformer (138/13.2kV ), and results are provided be-
low. It is also noted that different faults are also created at
various locations of the seven bus microgrid and the Least
Square Technique for fault detection is verified for the same.
1) L-G Fault: The line to Ground (L-G) fault is initiated
near Bus 1 of the microgrid. R phase- ground fault resistance
is 0.1Ω The pre Trigger time for fault is taken 20%.The LG
Fig. 5. LL fault

2) Line to Line Fault: The LL fault is created between R

and Y phase of the micro grid at the location specified the
fault current is shown in the figure below in Fig. 5 In Fig. 6
the magnitude and phasor value for LL to line fault is shown
between R and Y phase. When the current higher than the
value of the threshold, the trip signal will be initiated, and the
fault will be isolated.
3) Line- Line-line Fault(LLL fault): The LLL fault is cre-
ated among the three phases RYB of the LV microgrid. The
fault current is shown below in Fig. 7 The magnitude and
phasor is determined by using LS technique is presented in
Fig. 3. LG Fault Fig. 8. When the magnitude surpasses the specified threshold
value, the trip signal is said to be enabled.
fault current is as shown in Fig. 3 In Fig. 4, the magnitude,
phasor estimation of the fault signal and Trip signal is shown. V. CONCLUSION
When the fault current exceeds certain value (Threshold) of In this work, Discrete Fourier Transform Technique (DFT),
the magnitude, the trip signal is initiated. Kalman Filtering Technique (KFT), and Least Square Tech-

35

Authorized licensed use limited to: COLLEGE OF ENGINEERING AND TECHNOLOGY. Downloaded on April 29,2023 at 06:55:09 UTC from IEEE Xplore. Restrictions apply.
 Fig. 6. magnitude and phasor for LL fault
Fig. 8. magnitude and phase for LLL fault

for Inverter-Interfaced microgrids,” IEEE Trans. Smart Grid, vol. 9, no.

3, pp. 2097–2107, May 2018.
[4] Z. Chen, X. Pei, M. Yang, L. Peng, and P. Shi, “A Novel Protection
Scheme for Inverter-Interfaced microgrid (IIM) Operated in Islanded
Mode,”IEEE Trans. Power Electron., vol. 33, no. 9, pp. 7684–7697,
2018.
[5] G. Chen, Y. Liu and Q. Yang, ”Impedance Differential Protection for
Active Distribution Network,” in IEEE Transactions on Power Delivery,
vol. 35, no. 1, pp. 25-36, Feb. 2020.
[6] N. K. Sharma and S. R. Samantaray, “PMU Assisted Integrated
Impedance Angle-Based microgrid Protection Scheme,” IEEE Trans.
Fig. 7. LLL fault Power Deliv., vol. 35, no. 1, pp. 183–193, 2020.
[7] O. Nzimako and A. Rajapakse, “Real time simulation of a microgrid
with multiple distributed energy resources,” Proc. 2016 Int. Conf.
Cogener. Small Power Plants Dist. Energy, ICUE 2016, pp. 2–7, 2016.
nique (LST) for microgrid fault detection are discussed. The [8] T. R. Juttu, B. R. Bhalja, M. Andreev and O. P. Malik, ”Synchrophasor
Assisted Power Swing Detection Scheme for Wind Integrated Transmis-
pros and cons of the mentioned technique are discussed clearly sion Network,” in IEEE Transactions on Power Delivery,2021.
for fault detection in the microgrid.The LST can separate the [9] M. S. Sachdev and R. Das, “Undesrtanding Microprocessor-based Tech-
dc decaying component from the phasor value. Hence, at the nology applied to relaying,” Rep. Work. Gr. I16 Relaying Comm. IEEE,
p. 72, 2004.
inception of any fault, the magnitude change is free from any [10] A.G. Phadke and J. S. Thorp, Computer Relaying for Power Systems.
decaying component that can help better trip signal generation. John Wiley & Sons, Ltd, 2009.
Moreover, this technique can be used to estimate the sub- [11] Canyon Hydro et al., “Introduction to Kalman Filter and Its Applica-
tions,” Intech, vol. 32, no. July, pp. 137–144, 2013.
synchronous part of the fault current. In real-time operations of
the LV microgrid, the Least Square Technique technique works
effectively for all types of faults. The simulation results show
that for satisfactory protection of microgrid, the LST technique
can be suitably implemented.

R EFERENCES
[1] A. Hooshyar and R. Iravani, “microgrid Protection,” Proc. IEEE, vol.
105, no. 7, pp. 1332–1353, 2017.
[2] M. A. Zamani, A. Yazdani, and T. S. Sidhu, “A communication-assisted
protection strategy for inverter-based medium-voltage microgrids,” IEEE
Trans. Smart Grid, vol. 3, no. 4, pp. 2088–2099, 2012.
[3] I. Sadeghkhani, M. E. H. Golshan, A. Mehrizi-Sani, J. M. Guerrero,
and A. Ketabi, “Transient Monitoring Function-Based Fault Detection

36

Authorized licensed use limited to: COLLEGE OF ENGINEERING AND TECHNOLOGY. Downloaded on April 29,2023 at 06:55:09 UTC from IEEE Xplore. Restrictions apply.