Development of a Fault Detection Technique in

Transmission Line using Extended Kalman Filter

Project Report
Submitted in partial fulfillment of
the requirements for the award of
Bachelor of Technology Degree in
Electrical and Electronics Engineering
of the APJ Abdul Kalam Technological University

Submitted by
Aswin Raj (TVE18EE041)
Bensen Bernard (TVE18EE044)
Devika M. K. (TVE18EE049)
Shivani H. Nair (TVE18EE109)

Guided by
Dr. Vinod V.

DEPARTMENT OF ELECTRICAL ENGINEERING

COLLEGE OF ENGINEERING TRIVANDRUM
KERALA
2022
DECLARATION

We undersigned hereby declare that the project report DEVELOPMENT OF A

FAULT DETECTION TECHNIQUE IN TRANSMISSION LINE USING EX-
TENDED KALMAN FILTER, submitted for partial fullfillment of the requirements
for the award of degree of Bachelor of Technology of the APJ Abdul Kalam
Technological University, Kerala is a bonafide work done by us under the supervision
of Dr. Vinod V. This submission represents our ideas in our own words and where ideas
or words of others have been included, we have adequately and accurately cited and
referenced the original sources. We also declare that we adhered to ethics of academic
honesty and integrity and have not misrepresented or fabricated any data or idea or fact
or source in our submission. We understand that any violation of the above will be a
cause for disciplinary action by the institute and/or the University and can also evoke
penal action from the sources which have thus not been properly cited or from whom
proper permission has not been obtained. This report has not been previously formed
as the basis for the award of any degree, diploma or similar title of any other University.

Trivandrum Aswin Raj

27-06-2022 Bensen Bernard

Devika M. K.

Shivani H. Nair

1
 DEPT. OF ELECTRICAL ENGINEERING
COLLEGE OF ENGINEERING TRIVANDRUM
2021 - 22

CERTIFICATE

This is to certify that this project report entitled “Development of a Fault

Detection Technique in Transmission Line using Extended Kalman Filter” is a
bonafide record of the project work done by Aswin Raj, Bensen Bernard, Devika
M. K. and Shivani H. Nair of 8th semester B.Tech under our guidance towards the
partial fulfillment of the requirements for the award of B.Tech Degree in Electrical
and Electronics Engineering of the APJ Abdul Kalam Technological University of
Kerala during the year 2022.

Dr. Vinod V. Dr. Jisha V. R.

Assistant Professor (Guide) Project Coordinator
Department of Electrical Engineering, Department of Electrical Engineering,
College of Engineering Trivandrum. College of Engineering Trivandrum.

Dr. Jisha V. R. Dr. Savier J. S.

U.G. Co-ordinator & Professor Professor & Head
Department of Electrical Engineering, Department of Electrical Engineering,
College of Engineering Trivandrum. College of Engineering Trivandrum.
Acknowledgement

First of all, we would like to express our deepest appreciation and gratitude to our
guide Dr. Vinod V., Assistant Professor (Guide),Department of Electrical Engineering,
College of Engineering Trivandrum, for all his guidance and support. We could not
have completed this project if it wasn’t for his constant guidance and advice.

We gratefully acknowledge, Dr. Savier J. S., Head of the Department, Department

of Electrical Engineering, for his wholehearted cooperation and encouragement.

We express our gratitude to Dr. Jisha V. R., U.G. Co-ordinator & Professor,
Department of Electrical Engineering, College of Engineering, Trivandrum, for all
necessary help extended to us during the evaluation process in doing the project work.

We also acknowledge our gratitude to all other members of faculty in the

Department of Electrical Engineering, our family members and friends for their
wholehearted cooperation and encouragement. Above all, we thank GOD Almighty,
without whose help we wouldn’t have reached this far and completed the work.

3
Abstract

Distance relay plays a pivotal role in the detection of faults and subsequent isolation
in transmission lines. However, severe distortion in the measured voltage and
current waveforms (such as due to Current Transformer (CT) saturation), affects
their performance. This project report focuses on a compensation method for signal
reconstruction using Extended Kalman Filter (EKF) algorithm to improve relay
functioning, enhancing the safety of transmission lines. The proposed fault detection
scheme was tested extensively by running simulations using PSCAD and MATLAB for
different types of faults at various distances and CT burden. The CT model in PSCAD
was developed based on the Jiles-Atherton phenomenological theory in order to closely
resemble a real-time CT with its saturation effects. The simulation results show that the
proposed method is highly efficient in determining the exact fault point with improved
zone approaching times even with measured signals having high noise and harmonic
content. In cases where the conventional relay fails to detect the fault in the correct
zone, the compensated data from the proposed method accurately detects the zone.
Hardware experimental tests were carried out to further verify the effectiveness of the
scheme in real-time, where the EKF based algorithm was implemented using a DSP-
based microcontroller.

4
Contents

1 Introduction 1
1.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Problem Statement . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.3 Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.4 Objectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.5 Outline of the Report . . . . . . . . . . . . . . . . . . . . . . . . . . 3

2 Literature Review 4

3 Proposed System Model 7

4 Methodology 9
4.1 Extended Kalman Filter (EKF) Algorithm . . . . . . . . . . . . . . . 9
4.1.1 Integration of EKF Algorithm in Distance Relay Protection
Scheme . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
4.1.2 Implementation of EKF Algorithm . . . . . . . . . . . . . . . 11
4.2 Selection of q and r values . . . . . . . . . . . . . . . . . . . . . . . 13
4.3 Modelling of Current Transformer in Real Time . . . . . . . . . . . . 14

5 Hardware Design 17
5.1 Developed Prototype . . . . . . . . . . . . . . . . . . . . . . . . . . 17
5.2 Transmission Line Model . . . . . . . . . . . . . . . . . . . . . . . . 18
5.3 Current and Voltage Sensor . . . . . . . . . . . . . . . . . . . . . . . 18
5.4 Controller . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19

6 Software Design 21
6.1 System Modelling in PSCAD . . . . . . . . . . . . . . . . . . . . . . 21
6.2 Prediction of Fault Location in MATLAB . . . . . . . . . . . . . . . 24

i
 6.3 Performance under Noisy Conditions . . . . . . . . . . . . . . . . . . 26
6.4 Performance under Harmonic Contents . . . . . . . . . . . . . . . . . 28
6.5 Controller Software in Code Composer Studio . . . . . . . . . . . . . 29

7 Results and Discussions 31

7.1 Simulation Results . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
7.1.1 Simulation Results during Current Transformer Saturation . . 31
7.1.2 Simulation Results with Noisy Signals . . . . . . . . . . . . . 35
7.1.3 Simulation Results with harmonic components . . . . . . . . 37
7.2 Hardware Verification and Results . . . . . . . . . . . . . . . . . . . 38
7.2.1 Verification of Zone Approaching Time . . . . . . . . . . . . 38
7.2.2 Calculation of R and X values . . . . . . . . . . . . . . . . . 40

8 Conclusion 41

References 42

ii
List of Figures

3.1 Block Diagram . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7

4.1 Block Diagram of EKF Algorithm . . . . . . . . . . . . . . . . . . . 12

4.2 B-H curve . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16

5.1 Circuit diagram of the experimental set up . . . . . . . . . . . . . . . 17

5.2 Developed Prototype . . . . . . . . . . . . . . . . . . . . . . . . . . 19

6.1 Single line diagram of 220kV test system with frequency dependent
transmission line . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
6.2 Overall dimensions of transmission tower employed in the test set up . 22
6.3 PSCAD Simulation Test Set Up . . . . . . . . . . . . . . . . . . . . 23
6.4 Variation of secondary voltage waveform on application of fault: (a)
without the addition of White Gaussian Noise (b) with the addition of
White Gaussian Noise of Signal to Noise ratio of 10dB . . . . . . . . 27
6.5 Generation of Harmonics in PSCAD . . . . . . . . . . . . . . . . . . 28
6.6 Variation of secondary current waveform on application of fault: (a)
with the addition of harmonic content (10% of 2nd harmonics, 10% of
3rd harmonics and 10% of 5th harmonics) (b) without the addition of
harmonic content . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
6.7 Workspace in Code Composer Studio . . . . . . . . . . . . . . . . . 30

7.1 Secondary voltage, uncompensated current and compensated current

waveform for three phase fault at 20km with a CT burden of 5Ω . . . 31

iii
7.2 Zone-1 and Zone-2 approaching time of the uncompensated and
compensated Z-locus . . . . . . . . . . . . . . . . . . . . . . . . . . 32
7.3 Zone approaching time for three phase fault at various fault distances
and CT burden . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
7.4 Secondary uncompensated voltage, compensated voltage, uncompen-
S
sated current and compensated current waveform for a N
of 20 dB . . 35
7.5 Zone approaching time for three phase faults at 20km with different
S
CT burden and N
ratios . . . . . . . . . . . . . . . . . . . . . . . . . 37
7.6 Measured secondary voltage, compensated and uncompensated cur-
rents for three phase fault at zone-2 with a CT burden of 10Ω . . . . . 39
7.7 Variation of R and X values during fault condition . . . . . . . . . . . 40

iv
List of Tables

2.1 Merits and demerits of different compensation schemes . . . . . . . . 6

4.1 RMSE values for Saturated Condition of CT . . . . . . . . . . . . . . 14

4.2 RMSE values for Unsaturated Condition of CT . . . . . . . . . . . . 14
4.3 Current Transformer Parameters . . . . . . . . . . . . . . . . . . . . 15

5.1 Hardware equipment details . . . . . . . . . . . . . . . . . . . . . . 20

7.1 Comparison of zone approaching times under various types of faults

and fault distances . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
7.2 Comparison of zone approaching times for different S/N ratios and
fault distances . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
7.3 Comparison of zone approaching times in the presence of different
harmonic components . . . . . . . . . . . . . . . . . . . . . . . . . . 38
7.4 Hardware results: Comparison of zone approaching times for different
CT burden and fault points . . . . . . . . . . . . . . . . . . . . . . . 39

v
Chapter 1

Introduction

1.1 Background
The increase in electrical power demand cannot be met with a proportionate increase
in electrical power generation alone; an efficient power transmission system plays
an integral role in realising this. However, power transmission networks have an
associated risk of fault occurrence which causes severe economic losses and poor
reliability. Rapid fault detection and subsequent isolation is therefore crucial for
uninterrupted flow of electric power. Distance relay protection is one of the many
methods employed for this purpose.

Real-time measurement of power system parameters for continuously monitoring the

system for inception of faults is carried out through the use of current and potential
transformers. These equipment scale down the system current and voltage levels to
be further analysed. A major drawback of current transformers is the possibility of
erroneous measurement resulting from a saturated core. This can have detrimental
effects on the detection of a fault and its location by the protection system. Therefore,
compensation of saturation of current transformers is necessary for improving the fault
detection capability of the distance relay.

1
1.2 Problem Statement
The aim of the proposed method is to design and implement a fault detection
scheme to overcome malfunction in distance relay operation in transmission lines.
The proposed method will accurately predict the fault location even in the case of
severe disturbances/noises such as current transformer saturation. This is achieved
through the execution of Extended Kalman Filter (EKF) algorithm, which is capable
of estimating the impedance and consequently the distance to the fault point during
disturbance for different types of faults.

1.3 Motivation
Power system disturbances may cause stable or unstable power swings in the trans-
mission system. As a result, apparent impedance trajectories may or may not enter
into distance characteristics of the relay during a power swing causing distance relay
maloperation. This is a direct consequence of distortion in the fault current wave
caused by Current Transformer saturation combined with the presence of DC offset
resulting in overreach condition. This leads to serious safety concerns, as the relay will
not operate as desired. The proposed model using EKF algorithm accurately predicts
the fault point, ensuring the tripping of circuit breaker within prescribed seconds.

1.4 Objectives
General Objective :
To enhance the safety and protection of power transmission line in case of severe
fault current, that would otherwise go undetected due to saturation effect in current
transformer and resulting DC offset.

2
Specific Objective :

• To utilise EKF algorithm to estimate the accurate current waveform, while

compensating for CT saturation.

• Reduce the time delay in the operation of Distance Relays.

• Verify the proposed method through hardware implementation.

1.5 Outline of the Report

Chapter 1 gives a brief introduction to the project. In chapter 2 a literature survey is
conducted. Chapter 3 gives an idea about the proposed system used. Chapter 4 deals
with the methodology of the system. The hardware design is given in chapter 5.The
software design is discussed in chapter 6. Results and discussions are conducted in
chapter 7.Finally, the report is concluded in chapter 8.

3
Chapter 2

Literature Review

The main purpose of Current Transformers (CTs) in a high voltage power system is to
step down the current into one that can be handled by delicate electronic devices. They
are hence essential in the use of distance relays for protection. However, saturation
effects in the core of a CT causes miscalculation of impedance in the distance element.
This impact of CT saturation in distance protection is studied extensively in [1]. Due
to CT saturation, there is a change in the calculated phase shift [2] of the measured
current with its magnitude falling short of the actual value. The resulting error causes
the distance element to have slower operation time leading to under reach condition.
According to IEC 60255-151 standard [3], the major downside of CT saturation is the
loss of dependability and security. This poses a major concern of safety in transmission
and distribution systems and hence needs to be compensated.

The algorithm proposed in [4] , consists of two consecutive compensation filters that
aims to reconstruct both the fault and inrush currents distorted by the saturated CT. In
[5] and [6] , novel CT saturation detection and compensation methods based on Least
Error Squares (LES) technique is introduced. Additionally, a simple offline lookup
table is also utilised in [6]. The Multistage Least-Squares (MLS) and Partial Sum
(PS)-based method used in the hybrid algorithm [7] is able to successfully estimate the
current signal phasor, regardless of dc offset and CT saturation. In the event of severe
noise as well as harmonics, the waveform reconstruction methods [4–7], prove to be
insufficient in estimating the fault current accurately.

4
 The detrimental effects of CT saturation can be counterbalanced through the use of
Artificial Neural Networks(ANN). In order to estimate the true current of a saturated
transformer, the ANN is trained using data generated from actual CTs as well as
from EMTP simulations [8]. In this paper, two different sets of network coefficients
are employed to handle large dynamic ranges of fault current. The operation of the
algorithm for running the network in real-time is demonstrated on a DSP processor
simulator. An alternative algorithm for the correction of saturated CT secondary
current using an ANN based scheme is proposed in [9]. In the multi-layer feed-
forward network (MFNN) method presented in [10], the distorted CT current is
detected using wavelet transform and all possible currents are partitioned into nine
clusters. These nine MFNNs are trained individually and finally integrated to obtain
corrected secondary current. The compensation algorithm presented in [11], introduces
an adaptive network-based fuzzy interference system that converts a sampled, distorted
current waveform to a compensated one. The operating performance of the algorithm
under various faults and CT conditions are also reported. Bayesian Deep Neural
Network (BDNN), is another technique [12] to compensate for saturated current
waveform, comprising of Deep Neural Network (DNN) and Bayesian Optimization
(BO). This method proves to be superior to ANN and LES techniques through a
comparative analysis. However, these methods require a considerable amount of
training data for varying system and CT parameters to provide reliable output.

Several hardware addition-based methods are also proposed to overcome CT saturation

effects such as the use of adjustable switched resistor [13], negative voltage feedback
[14] and negative resistance [15]. In [13], prevention of CT core saturation is achieved
with the help of an adjustable resistor and an auxiliary transformer. The method
discussed in [14], involves the application of negative voltage feedback to the CT
secondary side such that it maintains constant magnetic flux under any fault condition.
In [15], compensation is attained through the use of a negative resistance to control the
applied voltage to the secondary terminal of CT.

5
 Table 2.1: Merits and demerits of different compensation schemes
Compensation Scheme Merits De-merits
Reconstruction depends on depth of saturation
More reliable because
Waveform reconstruction Estimation is inaccurate, if the fault current
of no additional equipment
is contaminated with noise and harmonics
Hardware addition Hardware failure affects reliability
Exact reconstruction of saturated current
in CT secondary Increased cost
Large number of training data required
Artificial Neural
Captures the non-linear behaviour of CT accurately Complex real-time implementation
Network
Intense computational burden
Optical Current Better linearity and greater accuracy Massive replacement of existing CT
transformer Extensive operating current range Increased cost

Alternatively, Optical Current Transformers (OCTs) can be employed in the place

of magnetic/conventional CTs. As studied experimentally in [16], an OCT is not
saturated by the dc offset current under simulated short circuit conditions and also
gives acceptable performance with white noise. Rogowski Coil (RC) is yet another
viable substitute for CTs with better linearity, greater accuracy and extensive operating
current range [17]. The merits and demerits of different compensation schemes for
saturated CTs are outlined in TABLE 2.1.

The aim of the proposed work detailed in this report is to develop an efficient
and fast-converging alternative scheme with minimal hardware modifications. A
mathematical/algorithmic approach for compensation would meet these requirements,
accurately predicting the current waveform even in noisy and harmonically polluted
environments. The scope of the Extended Kalman Filter (EKF) algorithm for this
application is extensively analysed in [18]. Additionally, the work presented in this
report obtains precise values for both process and measurement covariance matrices.
This will, in turn, enhance the accuracy of the proposed method by predicting the
heavily distorted voltage and current waveform. Furthermore, the implementation
of hardware circuitry and real-time compensation using a high speed microcontroller
augments the credibility of the algorithm put forth.

6
Chapter 3

Proposed System Model

The proposed system consists of a scaled down model of a three phase transmission
line, with adequate supply and load provided at both its extremities. The system also
includes current transformer to measure line current. The measured line current as well
as the line voltage is fed into a DSP-based microcontroller through carefully calibrated
voltage and Hall-effect current sensors. Fault creation is achieved through a contactor.

Figure 3.1: Block Diagram

7
The EKF algorithm is implemented through the advanced DSP-based microcontroller,
TMS28377s. Various kinds of faults such as three phase and single line to ground
faults are created using three phase and single phase contactors which are activated
upon receiving signals from the microcontroller itself. Function of the sensors in this
circuit is to scale down the high value of current and voltage which would otherwise
damage the microcontroller. During the fault condition, EKF will predict the fault
point with minimum error consequently isolating the fault point instantaneously.

8
Chapter 4

Methodology

4.1 Extended Kalman Filter (EKF) Algorithm

The distance relay is designed for fast action in case of a fault in its Zone-1 operating
region. However, distortion in the voltage and current waveforms and hence the
impedance calculated deviates considerably from the actual value, impeding the
response of the relay. As a result, the zone approaching time of the Z-locus (time
at which the Z-locus intersects with the zonal characteristics) tends to be higher. This
delay in isolating the fault can have severe safety implications. One of the major
causes of current waveform distortion is the saturation of CT used for measurement.
This section concentrates on the prediction of the actual current waveform using EKF
algorithm.

4.1.1 Integration of EKF Algorithm in Distance Relay Protection

Scheme

The state-space representation of the non-linear simulation model is essential for its
implementation.
During fault, the current waveform during kth state can be defined as follows:

Xn
Ik = F j cos( jωk + θ j ) + Geλk (4.1)
j=1

9
where n = 5, i.e., considering harmonics up to fifth order. F j is the amplitude of the jth
harmonic, θ j is the corresponding phase shift, and ω = 2 f / f s , in which f and f s are
the current signal frequency and sampling frequency, respectively. The second term
accounts for the dc offset that occurs at the instant of fault. Here G is a constant and
λ is the time constant of the exponential term, which depends on the X/R ratio of the
power system. The system state equations are as defined in (4.2) where j is varied from
1 to 5.

S ( j)(k) = F j cos( jωk + θ j )

S ( j+1)(k) = F j sin( jωk + θ j )

(4.2)
λk
S (2n+1)(k) = Ge

S (2n+2)(k) = λ

State equations at (k + 1)th state can be derived and represented as in (4.3)

S ( j)(k+1) = F j cos( jω(k + 1) + θ j )

= F j cos( jωk + jω + θ j )

= F j cos( jωk + θ j ) cos( jω)

(4.3)
| {z }
S ( j)(k)

− F j sin( jωk + θ j ) sin( jω)

| {z }
S ( j+1)(k)

= S ( j)(k) cos( jω) − S ( j+1)(k) sin( jω)

Similarly, S ( j+1) , S (2n+1) and S (2n+2) at the (k + 1)th state can be obtained as shown in
(4.4)-(4.6)

10
 S ( j+1)(k+1) = F j sin( jω(k + 1) + θ j )

= F j sin( jωk + jω + θ j )

= F j cos( jωk + θ j ) sin( jω)

(4.4)
| {z }
S ( j)(k)

+ F j sin( jωk + θ j ) cos( jω)

| {z }
S ( j+1)(k)

= S ( j)(k) sin( jω) + S ( j+1)(k) cos( jω)

S (2n+1)(k+1) = Geλ(k+1) = Geλ eλk

(4.5)
= eλ (Geλk ) = eλ S (2n+1)(k)
|{z}
S (2n+1)(k)

S (2n+2)(k+1) = S (2n+2)(k) (4.6)

The matrix form representation of all the (2n+2) state equations at the kth state is:

 
S k = S (1)(k) ; ...; S ( j)(k) ; ...; S (2n+1)(k) ; S (2n+2)(k) (4.7)

4.1.2 Implementation of EKF Algorithm

The EKF Algorithm, an optimal state estimator for a non-linear system, can be
sectioned into three processes, viz. Prediction, Gain Calculation and Correction. The
estimation of the state, before and after the measurement of current is termed as priori
and posteriori state estimate, respectively. The flow of calculation of posteriori state
estimate from (k − 1)th state to kth state is explained with the help of Fig. 4.1. The filter
can be initialised at the prediction stage with the initial state S o and covariance matrix
Po . The priori state estimate S k and the priori current value I k is calculated in the first
stage using (4.8) and (4.9).
S k = g(Ŝ k−1 ) + w(k) (4.8)

11
 Figure 4.1: Block Diagram of EKF Algorithm

Ik = h(S k ) + v(k) (4.9)

where g(Ŝ (k−1) ) is the state transition function formed using (k − 1)th posteriori state
estimate and w(k) is the associated process noise; h(S k ) is the measurement function
formed using S k and v(k) is the corresponding measurement noise. The first step is to
linearise Ŝ k−1 and S k , achieved by forming corresponding Jacobian matrices Gk and
Hk .
∂g(Ŝ k−1 ) ∂h(S k )
Gk = , Hk = (4.10)
∂Ŝ k−1 ∂S k

This is followed by the estimation of priori error covariance matrix at kth state
(4.11)

Pk = Gk P̂k−1Gk T + Q (4.11)

where Q = q2 ×[I]12×12 is the diagonal covariance matrix for process noise. The cross
covariance matrix P12 denotes the relationship between process and measurement
errors. The Kalman Gain is calculated in the gain calculation stage using (4.13).

P12 = Pk · Hk T (4.12)

12
 Kk = P12 (C)−1 (4.13)

where matrix C is the Cholesky factorization (4.14), introduced to reduce the

computation time and improve the numerical stability of the algorithm.

C = ClCu = Hk P12 + R (4.14)

where R = r2 ×[I]1×1 represents the intentionally added measurement noise. Cl and

Cu are the upper and lower triangular matrices of same eigen vectors obtained during
Cholesky factorization. In the corrector stage, the Kalman gain Kk determines how
heavily the measurement contributes to the posteriori state estimate Ŝ k (4.15) in
addition to the priori state estimate S k . The corrector stage also calculates the kth
state posteriori covariance matrix P̂k and the current Iˆk using (4.16) and (4.17).

Ŝ k = S k + Kk (Ik − I k ) (4.15)

P̂k = Pk − Kk Hk Pk (4.16)

Iˆk = h(Ŝ k ) (4.17)

4.2 Selection of q and r values

The decision criteria to detect saturation, utilized in this work, is the same as that
defined in [18]. However, there is a difference in the selection of q and r values during
unsaturated and saturated region of the current waveform. Hence, this method gives
more importance to CT measurements in the unsaturated region while depending more
on the predicted value in the saturated region. The values q and r, are obtained through
trial and error so as to minimise the Root Mean Square Error. The selected values for
q in the event of saturation and unsaturation are 0.01 and 1 respectively. Similarly, the
values for r are 100 and 0.1.

13
 q 0.01 0.1 1 10 100
r
0.01 10.7106 10.7125 10.7124 10.7124 10.7124
0.1 10.8431 10.7272 10.7120 10.7124 10.7124
1 11.6728 10.6040 10.6823 10.7120 10.7124
10 5.0638 5.6524 9.9048 10.6815 10.7120
100 2.6134 3.1253 4.6112 9.8922 10.6815

Table 4.1: RMSE values for Saturated Condition of CT

q 0.01 0.1 1 10 100

r
0.01 0.3700 0.3674 0.3676 0.3676 0.3676
0.1 1.6883 0.3700 0.3513 0.3676 0.3676
1 12.3511 1.6883 0.3700 0.3674 0.3670
10 21.1325 12.3504 1.6883 0.3700 0.3674
100 21.8293 21.4453 12.3507 1.6883 0.3700

Table 4.2: RMSE values for Unsaturated Condition of CT

4.3 Modelling of Current Transformer in Real Time

The Jiles-Atherton phenomenological model [19], [20] of a ferromagnetic core is
adopted in order to closely resemble the saturation characteristics of a CT in electric
power transmission. A modified Langevin function is used to model the saturation
characteristics of a soft magnetic material in the Jiles-Atherton theory, which describes
the relationship between magnetic moment (M), and magnetic field intensity (H) [21].
The anhysteretic magnetization curve between He and Man (4.18) obtained from the
Langevin function f (He ), is used to replace the magnetization relationship between
B and H. Here, He = H + aM (a represents inter-domain coupling) and M s is
the saturation magnetization. However, the major drawback of Langevin function is
its inability to accurately match the slope of the B–H loop in the shoulder region,
necessitating the need to replace it. The novel function with additional parameters

14
 Table 4.3: Current Transformer Parameters

Parameter to adjust K and M 0.96

Interdomain Coupling , a 1.325e-5
Sat. Anhysteretic Magnetization , M s 1.72e6 A/m
Coefficent 1 of Anhys. Curve, α1 2730
Coefficent 2 of Anhys. Curve, α2 3209
Coefficent 3 of Anhys. Curve, α3 20294
Coefficent 4 of Anhys. Curve, β 2
Primary turns, N p 1
Secondary turns, N s 200
Secondary Resistance, R s 0.5 Ω
Secondary Inductance, L s 0.8e-3 H
Area, A 2.601e-3 m2
Path Length 0.6377 m

(4.19) provides improved flexibility to trace a B-H loop with a better overall shape.
The relevant parameters used to simulate a CT are tabulated in TABLE 4.3.

Man = M s f (He ) (4.18)

α1 He + H β
Man = M s (4.19)
α3 + α2 He + H β

Man = M s f (He ) (4.20)

The anhysteretic magnetization curve between He and M (Fig. 4.2) obtained from
the Langevin function, f (He ), is used to replace the magnetization relationship between
B and H. Here, He = H + aM (a represents inter-domain coupling) and M s is the
saturation magnetization.

c dM
dHe
an
+ Man −M
δK α(Man −M)
dM µ0 − 1−c
= (4.21)
dH 1 − αc dM
dMe
an

15
 The alternate function should adhere to the following constraints:

lim Man = 0, lim Man = M s ,

He →+∞ He →+∞
(4.22)
dMan
≥ 0 ∀ He
dMe

On taking the derivative of (19) with respect to He , the resulting function satisfies
the constraints provided α1 > 0, α2 ≥ α1 , α3 > 0, and β ≥ 1.

Figure 4.2: B-H curve

16
Chapter 5

Hardware Design

5.1 Developed Prototype

Burden
3-ϕ Is
CT
Supply L L Load

Va F1 F2 F3

F1 F2 F3

F1 F2 F3

Relay
From
GPIO1
Contactor
TMS320F28377S
GPIO1 To Relay
Distorted Voltage
Voltage ADCIN1
Sensor
Va DAC1 Compensated
Voltage
Saturated
Current Current
ADCIN2 Compensated
Is Sensor DAC2 Current

Figure 5.1: Circuit diagram of the experimental set up

The circuit diagram for the hardware experimental setup is shown in Fig 5.1. The
scaled down transmission line model is excited from one end using a three phase
autotransformer and the other end is connected to a balanced three phase star connected

17
load. A CT (with variable burden) is connected in R phase to measure the phase
current. Three phase faults are generated using a three phase contactor. Three tappings
are provided in each phase to create these faults at various distances that fall under the
two zones of protection. The first two tappings in each phase lies under Zone-1 and
third tapping comes under Zone-2. The details of each equipment in the set up are
summarized in Table 5.1.

5.2 Transmission Line Model

The scaled down transmission line model is equivalent to two inductances of value
25mH each connected in series. An inductive model is adopted since the effect of
capacitive reactance can be neglected when scaled down. The core of the inductor is
constructed using CRNGO (with maximum flux density of 1.5T) laminations of E-I
type with double layer copper windings wound on top of it. Centre-taps are provided
along the length to change fault application points. The model is designed to safely
carry a maximum current of 5A.

5.3 Current and Voltage Sensor

The current sensor is used for sensing the primary and secondary currents of the
Current Transformer. This is achieved by converting those currents into voltages
suitable for DSP-based microcontroller. The Hall-effect sensor used is HO-10-
P(LEM), with a sensitivity of 50mV/A. The voltage sensor used for stepping down the
voltage level is designed for an operating range upto 350V (rms). The sensitivity of
the sensor is 3.2 mV/V. Appropriate signal conditioning circuits are also incorporated
in the PCB design for the sensors to limit the voltage below the safe level of the
microcontroller. The PCB design software utilised to develop the schematic and
generate the corresponding gerber file is KiCAD.

18
5.4 Controller
The DSP-based high speed microcontroller used is TMS320F28377S. It is a 32-
bit, floating point microcontroller unit which provides 200MHz of signal processing
performance, with 4 independent 16-bit ADCs. The contactor for creating faults is
activated upon receiving appropriate signal from the GPIO1 pin of the microcontroller,
through a relay. Inputs to the microcontroller are given from appropriately placed
current and voltage sensors in the transmission line. Under faulty conditions, distorted
voltage and saturated current waveforms are fed into the microcontroller, through
ADCIN1 and ADCIN2 respectively. The conversion is initiated by timer interrupt after
every 200µs (sampling frequency=5kHz). Also, the ADC interrupt is generated at the
end of conversion to acquire the input sample data from the ADC register. The EKF
algorithm is then implemented on this erroneous data, giving compensated voltage and
current waveform as the output through DAC1 and DAC2 pins.

Figure 5.2: Developed Prototype

19
Item Description
Autotransformer Output Voltage - (0-415)V
CT Rated Voltage - 660V
Rated VA - 5VA
Turns Ratio - 5:10
WPL type
Burden (5-15)Ω
Inductor(L) 25mH
Contactor 3 pole, 3 phase, 18A, 415V, Automatic type
Current Sensor LEM HO-10P, Sensitivity-50mV/A
Voltage Sensor Sensitivity-3.2mV/V
Microcontroller TMS28377s, 32-bit, 200MHz
Load 5kW, 3 Phase Load

Table 5.1: Hardware equipment details

20
Chapter 6

Software Design

6.1 System Modelling in PSCAD

1 2 3
CB CT

PT

Section 1 Section 2
200kM 200kM

Figure 6.1: Single line diagram of 220kV test system with frequency dependent
transmission line

The single line diagram of the test system is shown in Fig. 6.1. It consists of a
3 phase voltage source with the base specifications: 100 MVA, 220 kV (L-L rms),
50 Hz. The 400km long transmission line has been sectioned into two of 200km
each supported by flat towers (Fig. 6.2). Each section is represented by a frequency
dependent (phase) model having 2 bundled Chukar ACSR conductors per phase and
ground wires made of ½ high strength steel wire. The zero and positive sequence
impedances of the transmission line in Ω /m are 0.3048 ×10−3 + j 0.1069 ×10−2 and
0.1747 ×10−4 + j 0.2953 ×10−3 respectively. The fault is inserted at various locations
in each section using the timed fault logic block.

21
 7.5[m]
7.5[m]

5[m]

0.4572[m]
30[m]

Figure 6.2: Overall dimensions of transmission tower employed in the test set up

The line currents and phase voltages are measured using current and potential
transformer blocks. The entire simulation test set up is as shown in Fig.6.3. The
Zone-1 operating region with relay tripping time of 100ms (5 cycles) corresponds to
distances up to 160km and Zone-2 operating region with relay tripping time of 300ms
(15 cycles) corresponds to distances from 160km to 400km as per IEEE Std C37.113-
2015.

22
Figure 6.3: PSCAD Simulation Test Set Up

23
6.2 Prediction of Fault Location in MATLAB
The current and voltage data is imported from PSCAD into MATLAB as numeric
matrix. The entire array is segmented into windows containing 200 samples each. The
resistance (R) and reactance (X) for each window is computed using Fourier Full Cycle
Algorithm (Digital Full Cycle Algorithm). The impedance diagram of the distance
relay (mho circle) calculated using the positive sequence impedance and resistance of
the entire transmission line is plotted for different operating zones. The Z-locus (R-X
trajectory for each window) is also plotted alongside. On the inception of fault, the R
and X values drop significantly, wherein the Z-locus enters into the operating zones of
the distance relay, causing it to trip with appropriate delay.

f u n c t i o n [ x , P]= e k f ( f s t a t e , x , P , hmeas , z , Q, R)
[ x1 , A]= j a c c s d ( f s t a t e , x ) ;
P=A* P *A’+Q;
[ z1 , H]= j a c c s d ( hmeas , x1 ) ;
P12=P *H ’ ;
R= c h o l (H* P12+R ) ;
U=P12 / R ;
x=x1+U * (R ’ \ ( z−z1 ) ) ;
P=P−U*U ’ ;
end

With compensation using EKF algorithm, the erroneous values of voltage and current
is predicted after inception of fault using the reliable data before fault condition alone.
This is achieved by suitably adjusting the Kalman Gain to rely more on the a priori
estimate value than the measured value. Once the measured data has been compensated
for in this manner, it is once again segmented into windows and the corrected values
for R and X are computed in the same way.

24
for k = s i n d e x : s number
[ Vs Vc ] = FOURIER FULL ( v ( k− s r a t e : k ) , s r a t e ) ;
[ I s I c ] = FOURIER FULL ( i ( k− s r a t e : k ) , s r a t e ) ;
Vm( k− s r a t e ) = s q r t ( Vs ˆ 2 + Vc ˆ 2 ) ;
Im ( k− s r a t e ) = s q r t ( I s ˆ 2 + I c ˆ 2 ) ;
R( k− s r a t e ) = ( Vc * I c + Vs * I s ) / ( I c ˆ 2 + I s ˆ 2 ) ;
X( k− s r a t e ) = ( Vc * I s − Vs * I c ) / ( I c ˆ 2 + I s ˆ 2 ) ;
end

f u n c t i o n [ Ys Yc ] = FOURIER FULL ( y , K)
Ys = 0 ;
Yc = 0 ;
f o r k = 1 :K
Ys = Ys + y ( k ) * s i n ( k * 2 * p i / K ) ;
Yc = Yc + y ( k ) * c o s ( k * 2 * p i / K ) ;
end
Ys = 2 * Ys / K;
Yc = 2 * Yc / K;
end

25
6.3 Performance under Noisy Conditions
The performance of the proposed fault protection scheme is analysed for noisy
conditions in both voltage and current waveforms. Consider the voltage waveform
at kth state during fault as :

X2
Vk = F j cos( jωk + θ j ) (6.1)
j=1

The corresponding state equations can be written as (without considering harmonics

and dc offset) :
S (1)(k) = F1 cos(ωk + θ1 ) (6.2)

S (2)(k) = F2 sin(2ωk + θ2 ) (6.3)

In order to verify the effectiveness of the algorithm in such worse conditions, the
addition of White Gaussian noise is made to these waveforms in MATLAB. In the
secondary voltage and current waveforms, Signal to Noise ratio is varied from 10dB to
30dB and the results were analysed.

26
Figure 6.4: Variation of secondary voltage waveform on application of fault:
(a) without the addition of White Gaussian Noise
(b) with the addition of White Gaussian Noise of Signal to Noise ratio of 10dB

27
6.4 Performance under Harmonic Contents
The functioning of the proposed algorithm in the presence of harmonics is also
analysed. Harmonic content is introduced in the CT secondary current data (as
a percentage of the fundamental) through PSCAD. The resultant distorted current
waveform is given as input to the MATLAB program.

Figure 6.5: Generation of Harmonics in PSCAD

28
Figure 6.6: Variation of secondary current waveform on application of fault:
(a) with the addition of harmonic content (10% of 2nd harmonics, 10% of 3rd harmonics
and 10% of 5th harmonics)
(b) without the addition of harmonic content

6.5 Controller Software in Code Composer Studio

The microcontroller in use has an ADC of 12-bit resolution. The signal conditioning
circuits of the sensors were designed to scale down the data measured to a range
between 0-3V, suitable for the microcontroller. The two inputs, corresponding to

29
 Figure 6.7: Workspace in Code Composer Studio

current and voltage data, are interfaced with the ADC A and ADC B pins of the
microcontroller, respectively. The analog data is then converted to their 12 bit
digital equivalent values. The actual values, obtained from their digital equivalent
through appropriate conversion equations, are then utilized to find the resistance
(R) and reactance (X) values using Fourier Full Cycle Algorithm. The R and X
values (their digital equivalents) are then given as output from DAC A and DAC B
pins. The results can be viewed using DSO connected to these pins. Additionally,
EKF can also be implemented on these values thus compensating for the erroneous
(saturated/noisy/harmonically distorted) data in real-time. The code for the controller
software was implemented using Code Composer Studio.

30
Chapter 7

Results and Discussions

7.1 Simulation Results

The efficacy of the proposed method is evaluated in terms of desired zone approaching
time. Different cases such as CT saturation during fault, noisy voltage and current
data etc. are analysed. The results with a comparison of both uncompensated and
compensated data (using EKF algorithm) corresponding to each case are discussed in
the following sub sections.

7.1.1 Simulation Results during Current Transformer Saturation

Voltage (V)
Current (A)

Time (sec)

Figure 7.1: Secondary voltage, uncompensated current and compensated current

waveform for three phase fault at 20km with a CT burden of 5Ω

31
 Unc
ompens
ate
dZ-loc
us
Co
mp e
nsa
tedZ-l
ocu
s

m
Zt2=160.
2ms

noh
Zt1=178.
1ms
Xi Zt1=109.
4ms

Zt2=108.
8ms

Ri
nohm

Figure 7.2: Zone-1 and Zone-2 approaching time of the uncompensated and
compensated Z-locus

A three phase fault was simulated at 20km in the system depicted in Fig. 6.1.
The secondary waveform (referred to as the uncompensated data) is shown in Fig.
7.1. After the inception of fault at 0.1s, it is evident that the CT of burden 5Ω goes
into saturation, as a result, the waveform is distorted. The EKF algorithm comes
into action at the instant of saturation and the predicted current waveform (referred
to as the compensated data) is obtained. Fundamental phasors of voltage and current
waveforms are estimated using Digital Full Cycle algorithm. Further analysis is carried
out by plotting the Z-locus in the R-X diagram (Fig. 7.2). The Zone-1 and Zone-2
approaching time is represented as Zt1 and Zt2 respectively. Due to saturation at fault,
the uncompensated Z-locus initially drifts away from Zone-2 and slowly approaches
Zone-2 at Zt2 =160.2ms and Zone-1 at Zt1 =178.1ms. As can be inferred from the graph,
Z-locus corresponding to compensated data enters Zone-2 at Zt2 =108.8ms and Zone-1
at Zt1 =109.4ms, proving the advantage of the proposed algorithm.

The results of the simulation for various types of faults and fault distances, falling
under both relay operation zones, is summarized in TABLE 7.1. A general trend
that can be observed is that as the fault distance increases, Zt1 and Zt2 increases as
well. For three phase fault at a distance of 160km (end of Zone-1), the inaccuracy
in the uncompensated data is evident as Zone-1 operation of relay fails (no crossing
of uncompensated Z-locus into Zone-1 mho circle). The distance relay also fails

32
 Uncompensated
Compensated

Figure 7.3: Zone approaching time for three phase fault at various fault distances and
CT burden

after 120km in the case of single line to ground fault. However, the compensated
data obtained using EKF algorithm successfully overcomes these issues and the relay
operates within the desired time during all types of faults.
The zone approaching time of both uncompensated and compensated data for a
three phase fault simulated at varying distances and CT burden is plotted in Fig. 7.3.
It is found that the time required for uncompensated data worsens with increasing
fault distances and CT burden, whereas, compensated data is independent of these
parameters. Therefore, its zone approaching time is well within the desired limit.

33
Table 7.1: Comparison of zone approaching times under various types of faults and
fault distances

Uncompensated Compensated
Type of Fault Distance
Zt1 (ms) Zt2 (ms) Zt1 (ms) Zt2 (ms)
20 178.1 160.2 109.4 108.8
40 181.9 162.9 110 109.1
60 188 167 110.6 109.5
80 201.9 178.3 111.4 109.9
Three phase fault 100 210.5 183.9 112.2 110.3
120 231.3 189 113.6 110.7
140 271.6 200.1 117 111.2
160 Fails 210.2 119.8 111.8
180 - 224.7 - 112.6
200 - 243.1 - 113.9
20 162.7 142.4 109 108.5
40 219.8 183.9 109.6 108.9
60 247.8 201.8 110.3 109.4
80 280.4 220.7 111.2 109.9
Single line
100 329.9 238.9 112.5 110.5
to
120 419.2 260.5 117.3 111.1
ground
140 Fails 290.4 123.9 111.8
160 Fails 331.3 124.2 112.9
180 - 401.3 - 117.5
200 - 680.1 - 121.7
20 128.5 125.3 121.9 114.3
40 183.7 124.5 114.8 114
60 128.8 123.4 115.4 114.4
80 127.4 122 116.1 114.8
Double line
100 126.3 116.3 116.9 115.2
to
120 125.7 116.1 118.1 115.7
ground
140 125.8 116.3 121.6 116.2
160 126.7 116.8 124.9 116.8
180 - 117.4 - - 117.6
200 - 118.2 - - 118.9

34
7.1.2 Simulation Results with Noisy Signals

Voltage (V)
Current (A)

Time (sec)

Figure 7.4: Secondary uncompensated voltage, compensated voltage, uncompensated

current and compensated current waveform for a NS of 20 dB

The performance of the proposed fault protection scheme is analysed for noisy
conditions in both voltage and current waveforms. Inorder to verify the effectiveness
of the algorithm in such worse conditions, the addition of White Gaussian noise is
made to these waveforms in MATLAB. The secondary waveforms with Signal to Noise
ratio of 20dB (uncompensated data) and the EKF compensated waveforms during a
three phase fault is shown in Fig.7.4. The simulation results for different Signal to
Noise ratios are summarized in TABLE 7.2, for a three phase fault with a CT burden
of 8Ω. As in the previous case, the fault locations are varied from 20km to 200km,
and for each distance the EKF algorithm returned improved values as opposed to the
uncompensated secondary data. For instance, the zone approaching time for S/N ratio
of 30dB and fault at 140km, has improved significantly from 271.7ms (uncompensated
data) to 119.4ms (compensated data) and from 209.1ms to 111.7ms for Zt1 and Zt2
respectively. Fig. 7.5 better illustrates the results for different values of CT burden and
S/N ratio for a fixed fault location of 20km.

35
Table 7.2: Comparison of zone approaching times for different S/N ratios and fault
distances

Uncompensated Compensated
S/N Ratio Fault km
Zt1 (ms) Zt2 (ms) Zt1 (ms) Zt2 (ms)
20 179.1 161 110.6 109.4
40 186.4 165.8 110.9 109.5
60 188.1 167.2 111.7 109.9
80 208.6 180.2 113.8 110.3
100 219.2 188 114.4 110.5
30 dB
120 239.7 189.3 117.1 111.1
140 271.7 209.1 119.4 111.7
160 385.9 210.4 120.1 112.4
180 - 230.7 - 113.8
200 - 251.3 - 115.9
20 179.3 160.8 110.4 109.3
40 183.6 166.1 111.5 109.8
60 188 167.2 111.6 109.8
80 208.6 180.2 110.6 109.4
100 221 188.1 112.4 110.2
20 dB
120 239.6 196.6 115.7 110.8
140 272.1 208.9 113.1 110.6
160 416.6 212.4 119.3 111.8
180 - 230.9 - 111.8
200 - 251.6 - 112.4
20 180.1 161 109.3 108.7
40 182.5 165.9 109 108.3
60 198.1 167.3 109.4 108.7
80 208.6 178.6 109.7 108.9
100 216.7 188.4 110.2 109.2
10 dB
120 244.1 189 110.5 109.5
140 261.5 207.3 110.7 109.7
160 336.1 209.6 112.5 110.8
180 - 217.9 - 111.1
200 - 251.4 - 111.5

36
 Uncompensated
Compensated

Figure 7.5: Zone approaching time for three phase faults at 20km with different CT
burden and NS ratios

7.1.3 Simulation Results with harmonic components

The functioning of the proposed algorithm in the presence of harmonics is also

analysed. Harmonic content is introduced in the CT secondary current data (as
a percentage of the fundamental) through PSCAD. The resultant distorted current
waveform is given as input to the MATLAB program. The zone approaching
time obtained for different percentages of harmonic content, tabulated in TABLE
7.3, further highlights the reliability of the proposed compensation method. For
a current waveform distorted with 20% 2nd , 25% 3rd and 15% 5th harmonics, Zt1
for uncompensated data is 208.6ms, whereas for compensated data Zt1 is 116.1ms,
ensuring the fast action of relay.

37
Table 7.3: Comparison of zone approaching times in the presence of different harmonic
components

Harmonic Content
Uncompensated Compensated
(as % of fundamental)
2nd 3rd 5th Zt1 (ms) Zt2 (ms) Zt1 (ms) Zt2 (ms)
5 20 15 208.7 180.2 115.3 110.5
15 20 15 208.7 180.1 115.3 110.5
25 20 15 208.7 179.4 116.2 110.6
20 25 15 208.6 179.9 116.1 110.6
25 25 15 208.7 178.8 115.7 110.5
5 20 20 208.7 180.3 115.3 110.5
15 5 20 208.7 179.6 114.5 110.4
15 15 20 208.7 179.6 114.5 110.4
15 25 20 208.7 180.2 115.3 110.5
25 15 20 208.7 178.9 115.7 110.5

7.2 Hardware Verification and Results

7.2.1 Verification of Zone Approaching Time

The effectiveness of the EKF based compensation can be noted from Fig. 7.6,
containing both voltage and current waveforms for a three phase fault at Zone-2 and a
CT burden of 10Ω. The saturation in the CT core and hence, distortion in the measured
secondary current is shown as the uncompensated current data. The corresponding
compensated current data obtained as the output from the microcontroller has little to
no trace of saturation. The zone approaching time during three phase fault at F2 (Zone-
1) and F3 (Zone-2) are tabulated in TABLE 7.4. It is observed that the uncompensated
relay fails to operate at fault points F2 and F3 for a burden of 15Ω. Even in this
worse situation, the proposed scheme works satisfactorily with zone approaching time
of 77.8ms and 83.4ms respectively.

38
Table 7.4: Hardware results: Comparison of zone approaching times for different CT
burden and fault points

Uncompensated Compensated
Fault Point Burden
Zt1 (ms) Zt2 (ms) Zt1 (ms) Zt2 (ms)
5 77.7 76.6 77.7 76.5
F1 (Zone-1) 10 86 78.6 79.6 78.3
15 96.9 84.3 85.1 76.3
5 100.1 82.8 73.9 71.2
F2 (Zone-1) 10 97.8 76.6 77.9 75.1
15 Fails 86.2 77.8 74.5
5 - 82.5 74.7 73.1
F3 (Zone-2) 10 - 134.5 75.5 72.9
15 - Fails 76.5 83.4

Figure 7.6: Measured secondary voltage, compensated and uncompensated currents

for three phase fault at zone-2 with a CT burden of 10Ω

39
7.2.2 Calculation of R and X values

During fault conditions, there is a marked change in resistance and reactance values of
the system. The resistance (R) reduces from a constant positive value to a value close
to zero at the onset of fault. A similar variation is observed in the reactance (X) values
as well. Suitable scaling factors must be applied to obtain the actual values for R and
X from their digital equivalent.

Figure 7.7: Variation of R and X values during fault condition

40
Chapter 8

Conclusion

In this project report, an improved mathematical fault detection scheme is proposed for
distance relays using the Extended Kalman Filter algorithm. The suggested method
is evaluated through simulating a test set up in PSCAD and MATLAB for different
types of faults at various distances and CT burden. For a three phase fault simulated
in a 400km long transmission line model, the zone approaching time of the Z-locus is
considerably reduced. This reduction in time remains even with added White Gaussian
Noise of different Signal-to-Noise ratios. Furthermore, the performance of this method
is found to be superior for a range of harmonic content in the measured signals. This
indicates that the EKF based approach is highly efficient in reconstructing the signal
with severe distortion resulting from CT saturation, noise and presence of harmonics.
To verify the effectiveness of the scheme in real-time, hardware experimental tests
are carried out where the EKF based algorithm is implemented using a DSP-based
microcontroller. The experimental results are in line with the simulation results, giving
a significant improvement in the zone approaching time.

41
References

[1] K. Wannous and P. Toman, “The impact of current transformer saturation on the
distance protection,” in 2016 17th International Scientific Conference on Electric
Power Engineering (EPE). IEEE, 2016, pp. 1–5.

[2] Y. Zheng, T. Wu, F. Hong, G. Yao, J. Chai, and Z. Wei, “Transmission line
distance protection under current transformer saturation,” Journal of Modern
Power Systems and Clean Energy, vol. 9, no. 1, pp. 68–76, 2020.

[3] S. IEC, “Measuring relays and protection equipment-part 151: Functional

requirements of over/under current protection,” IEC 60255-151, 2009.

[4] E. Hajipour, M. Vakilian, and M. Sanaye-Pasand, “Current-transformer saturation

compensation for transformer differential relays,” IEEE Transactions on Power
Delivery, vol. 30, no. 5, pp. 2293–2302, 2015.

[5] F. B. Ajaei, M. Sanaye-Pasand, M. Davarpanah, A. Rezaei-Zare, and R. Iravani,

“Compensation of the current-transformer saturation effects for digital relays,”
IEEE Transactions on Power Delivery, vol. 26, no. 4, pp. 2531–2540, 2011.

[6] A. Hooshyar and M. Sanaye-Pasand, “Accurate measurement of fault currents

contaminated with decaying dc offset and ct saturation,” IEEE transactions on
power delivery, vol. 27, no. 2, pp. 773–783, 2012.

[7] S.-R. Nam, J.-Y. Park, S.-H. Kang, and M. Kezunovic, “Phasor estimation in the
presence of dc offset and ct saturation,” IEEE transactions on power delivery,
vol. 24, no. 4, pp. 1842–1849, 2009.

42
 [8] D. C. Yu, J. C. Cummins, Z. Wang, H.-J. Yoon, and L. A. Kojovic, “Correction of
current transformer distorted secondary currents due to saturation using artificial
neural networks,” IEEE Transactions on Power Delivery, vol. 16, no. 2, pp. 189–
194, 2001.

[9] H. Khorashadi-Zadeh and M. Sanaye-Pasand, “Correction of saturated current

transformers secondary current using anns,” IEEE Transactions on Power
Delivery, vol. 21, no. 1, pp. 73–79, 2005.

[10] Y.-Y. Hong and P.-C. Chang-Chian, “Detection and correction of distorted current
transformer current using wavelet transform and artificial intelligence,” IET
generation, transmission & distribution, vol. 2, no. 4, pp. 566–575, 2008.

[11] K. Erenturk, “Anfis-based compensation algorithm for current-transformer

saturation effects,” IEEE Transactions on Power Delivery, vol. 24, no. 1, pp.
195–201, 2008.

[12] S. Key, S.-H. Kang, N.-H. Lee, and S.-R. Nam, “Bayesian deep neural network to
compensate for current transformer saturation,” IEEE Access, vol. 9, pp. 154 731–
154 739, 2021.

[13] S. Sanati and Y. Alinejad-Beromi, “Avoid current transformer saturation using

adjustable switched resistor demagnetization method,” IEEE Transactions on
Power Delivery, vol. 36, no. 1, pp. 92–101, 2020.

[14] M. Delzendeh and H. K. Karegar, “Current transformer saturation compensator

by using negative voltage feedback,” IEEE Transactions on Power Delivery,
vol. 35, no. 3, pp. 1200–1208, 2019.

[15] S. Sanati and Y. Alinejad-Beromi, “Prevention of the current transformer

saturation by using negative resistance,” IET Generation, Transmission &
Distribution, vol. 15, no. 3, pp. 508–517, 2021.

43
[16] S. Kucuksari and G. G. Karady, “Experimental comparison of conventional and
optical current transformers,” IEEE Transactions on power delivery, vol. 25,
no. 4, pp. 2455–2463, 2010.

[17] A. Sarwade, P. Katti, and J. Ghodekar, “Enhancement of digital distance

protection scheme with novel transducer,” in International Conference on
Intelligent Computing and Applications. Springer, 2018, pp. 287–297.

[18] F. Naseri, Z. Kazemi, E. Farjah, and T. Ghanbari, “Fast detection and

compensation of current transformer saturation using extended kalman filter,”
IEEE Transactions on Power Delivery, vol. 34, no. 3, pp. 1087–1097, 2019.

[19] D. C. Jiles, J. Thoelke, and M. Devine, “Numerical determination of hysteresis

parameters for the modeling of magnetic properties using the theory of
ferromagnetic hysteresis,” IEEE Transactions on magnetics, vol. 28, no. 1, pp.
27–35, 1992.

[20] R. Jajasinghe and P. McLaren, “Transformer core models based on the jiles-
atherton algorithm,” in IEEE WESCANEX 97 Communications, Power and
Computing. Conference Proceedings. IEEE, 1997, pp. 121–125.

[21] U. Annakkage, P. McLaren, E. Dirks, R. Jayasinghe, and A. Parker, “A current

transformer model based on the jiles-atherton theory of ferromagnetic hysteresis,”
IEEE transactions on power delivery, vol. 15, no. 1, pp. 57–61, 2000.

[22] J. Pan, K. Vu, and Y. Hu, “An efficient compensation algorithm for current
transformer saturation effects,” IEEE Transactions on Power delivery, vol. 19,
no. 4, pp. 1623–1628, 2004.

44