Testing of Current-Only Directional Relay Algorithm in a Realistic Distribution Network

Testbed

Francisco Javier Vivent Barahona

Thesis submitted to the faculty of the Virginia Polytechnic Institute and State University in
partial fulfillment of the requirements for the degree of

Master of Science
In
Electrical Engineering

Virgilio A. Centeno, Chair

Jaime De La Reelopez
Vasileios Kekatos

08/11/2023

Blacksburg, Virginia

Keywords: Hardware testbed, overcurrent directional relay, DER integration, phasor estimation
 Testing of Current-Only Directional Relay Algorithm in a Realistic Distribution Network
Testbed

Francisco Javier Vivent Barahona

ABSTRACT

The growth of Distributed Energy Resources is accelerating, causing significant changes in utility
distribution systems due to the bidirectionality of power flow. As a result, protection systems will
require upgrades to operate with these new requirements. The current-only directional relay
(CODR) is a novel proposal that detects the direction of fault currents without the need for new
devices, but instead upgrades the software logic of existing devices, making it a cost-effective
solution. This work provides a complete description of how to implement a hardware testbed to
review the CODR method. Findings show that CODR performs successfully in a real environment,
but its algorithm needs to be upgraded if used in distribution systems where lines have a non-
negligible resistive component.
 Testing of Current-Only Directional Relay Algorithm in a Realistic Distribution Network
Testbed

Francisco Javier Vivent Barahona

GENERAL AUDIENCE ABSTRACT

The use of renewable energy sources, such as solar panels and wind turbines, is growing rapidly.
This is causing changes in the way electricity is distributed, as power can now flow in both
directions. To keep up with these changes, the systems that protect the electricity grid need to be
updated. One cost-effective solution is to use a new type of relay, called the current-only directional
relay (CODR), which can detect the direction of fault currents without the need for new devices.
This work describes how to test the CODR method using a hardware testbed. The results show
that the CODR performs well in a real environment, but its algorithm needs to be updated for use
in certain types of distribution systems.
Acknowledgements

I would like to express my deepest indebtedness to my thesis advisor, Dr. Virgilio Centeno, for his
expert knowledge, patience, and unwavering commitment to my academic growth. His guidance
and feedback have been instrumental in shaping this research. I am also deeply grateful to Dr.
Jaime de la Ree, whose valuable insights and constructive feedback significantly improved the
quality of this work. His mentorship and encouragement were critical in keeping me focused and
motivated. I am also thankful to Dr. Vassilis Kekatos for sharing his knowledge and expertise with
me throughout my academic journey.
I would also like to extend my appreciation to Dr. Jason Lie and Dr. Eric Hsieh for their assistance
with the components needed to build the testbed, and to Biqui Wang for sharing her knowledge on
inverter models.
My sincere thanks go to my friends and colleagues, Alok, Suiksha, Vishal, Imtiaj, Arian,
Mohammad, Reza, Ashkan, and colleagues from PEC, for their endless support and camaraderie.
Their encouragement and understanding made the challenges of completing this thesis more
achievable.

iv
Dedication

I would like to dedicate this work to Natalia, who encouraged me to pursue this academic
achievement, to my family who always supported me from distance, and to my friends in
Blacksburg who made the town, and this experience, a beloved place.

v
Table of Contents

Chapter 1 Introduction .................................................................................................................... 1

1.1 Motivation ............................................................................................................................. 1

1.2 Testbed Description............................................................................................................... 2

1.3 Thesis Outline ....................................................................................................................... 4

Chapter 2 Literature Review ........................................................................................................... 6

2.1 Fault detections ..................................................................................................................... 6

2.2 Classic directional relay ........................................................................................................ 6

2.3 Current-only directional relay ............................................................................................... 7

2.4 Phasor Estimation ................................................................................................................. 9

Chapter 3 Testbed Implementation ............................................................................................... 14

3.1 Hardware ............................................................................................................................. 14

3.2 Software .............................................................................................................................. 17

Chapter 4 Testing and Results ....................................................................................................... 22

4.1 Introduction ......................................................................................................................... 22

4.2 Forward flow – forward fault testing .................................................................................. 23

4.3 Forward flow – reverse fault testing ................................................................................... 26

4.4 Discussion ........................................................................................................................... 29

4.5 Recommendations ............................................................................................................... 30

Chapter 5 Conclusion.................................................................................................................... 32

5.1 Conclusion .......................................................................................................................... 32

5.2 Main Contributions ............................................................................................................. 33

5.3 Future works ....................................................................................................................... 33

vi
References ..................................................................................................................................... 34

Appendices .................................................................................................................................... 37

Appendix A - Phasor Estimation Code ..................................................................................... 37

vii
List of Figures

Figure 1.1: Bidirectional Power flow due to DER integration. ...................................................... 2

Figure 1.2: Distribution Grid with DER Testbed. ........................................................................... 3
Figure 2.1: Comparison of phase angle Voltage and forward-reverse load-fault current. .............. 7
Figure 2.2: Pre-fault, forward, and reverse fault current in a distribution system with DER. ........ 7
Figure 2.3: Phasor diagram for currents sensed by the relay ......................................................... 8
Figure 3.1: Single-phase distribution system ................................................................................ 14
Figure 3.2: Testbed hardware components ................................................................................... 18
Figure 3.3: Testbed equivalent circuit ........................................................................................... 18
Figure 3.4: CCS Expression window ............................................................................................ 19
Figure 3.5: Waveforms oscilloscope sample rate configuration ................................................... 20
Figure 3.6: Distribution hardware testbed for CODR testing ....................................................... 21
Figure 4.1: Ideal circuit for testing ............................................................................................... 22
Figure 4.2: Forward power flow – forward fault current testing .................................................. 23
Figure 4.3: Forward power flow – forward fault 0.02 InvIoRef injection .................................... 24
Figure 4.4: Forward power flow – forward fault 0.05 InvIoRef injection .................................... 24
Figure 4.5: Forward power flow – forward fault 0.07 InvIoRef injection .................................... 25
Figure 4.6: Forward power flow – forward fault 0.10 InvIoRef injection .................................... 25
Figure 4.7: Forward power flow – reverse fault current testing ................................................... 26
Figure 4.8: Forward power flow – reverse fault 0.12 InvIoRef injection ..................................... 27
Figure 4.9: Forward power flow – reverse fault 0.15 InvIoRef injection ..................................... 27
Figure 4.10: Forward power flow – reverse fault 0.18 InvIoRef injection ................................... 28
Figure 4.11: Forward power flow – reverse fault 0.20 InvIoRef injection.................................... 28
Figure 4.12: Phasor diagram for currents sensed by the relay considering an R+jX distribution line
....................................................................................................................................................... 30

viii
List of Tables
Table 1.1: Testbed Hardware elements ........................................................................................... 4
Table 3.1: Single-phase inverter TIEVM-HV-1PH-DCAC features............................................. 15
Table 3.2: Grid-tied inverter main features on powerSuite page .................................................. 19
Table 4.1: Forward fault - summary results .................................................................................. 29
Table 4.2: Reverse fault – summary results .................................................................................. 29

ix
Chapter 1

Introduction

1.1 Motivation
In recent decades, the electric grid has faced a range of challenges, including environmental
concerns, more frequent and severe weather events, changing consumer needs, and advances in
related technologies. One notable technological advancement is the increased attractiveness of
renewable energy (RE) generation due to its improved efficiency, reliability, and cost-
effectiveness. It is expected that RE will become the primary source of energy generation in power
systems in the United States[1], but this transition is not without challenges [2][3].
The power systems are composed of three main segments: generation, transmission, and
distribution. Traditionally, the distribution segment has been designed as a radial-passive structure,
where the power flow is unidirectional, from the source end to the customer load. However, the
integration of Distributed Energy Resources (DERs) allows for the injection of power from the
load end into the distribution network, creating new operational scenarios, see Figure 1.1. This
integration offers various benefits, including increased resilience, reduced emissions, and
improved energy efficiency. However, it also creates operational difficulties for Distribution
System Operators (DSOs), such as system stability, voltage variations, energy quality issues, and
difficulties with conventional protection devices [4].
Particularly for distribution protection elements, the integration of DERs has resulted in power
flows being able to reverse direction, leading to changes in the amplitude and direction of the
system’s currents. This, in turn, can cause several issues, including difficulties with the settings of
distance and differential protections, recloser-fuse discoordination, and the non-functionality of
conventional overcurrent relay (OCR). These challenges become more pronounced as the
integration of DERs increases [5] and the large number of distribution protection devices is
considered. Although future technologies such as robust communication infrastructures, new

1
 Figure 1.1: Bidirectional Power flow due to DER integration.

sensor devices, and advanced information technologies are expected to address these issues [6],
existing control and protection devices were not designed to handle these new scenarios. Replacing
these large number of devices can be costly, and DSOs may not be able to afford these costs in the
short term [7]. As such, DSOs must find cost-effective and feasible solutions to ensure the stability
and reliability of distribution networks in the face of increasing DER integration.
The author in [8] aims to solve the OCRs bidirectionality using the current-only directional relay
(CODR), sensing the directionality of the current avoiding the angle reference given by the
voltage, as normally bidirectional OCR does. This solution could be implemented in existing
OCRs saving the investment in new directional overcurrent relays and voltage instrument
transformers. While the results in [8] seem promising, conducting a simulation with components
that better resemble real-world conditions would be appropriate to assess the effectiveness of this
solution in environments that closely resemble a distribution system.
Given the potential impact of Reverse Power Flow (RPF) on the safety and reliable operation of
distribution grids, the objective of this study is to evaluate RPF and enhance the solution proposed
by the CODR. To achieve this, a distribution grid testbed is set up and utilized to assess the
effectiveness of the proposed technique under various power flow scenarios.
1.2 Testbed Description
The objective of the testbed is to simulate a distribution grid with DER integration where the
direction of power flow can be controlled. The experiment consists of an inverter-based DER
connected to resistive loads where a switch controls the connection of one of the loads. These two
elements are connected to one end of a shunt resistor, the other end is connected to the distribution
line model, which emulates a feeder with one lateral load in its path. The feeder also connects with
an alternate voltage source that models the distribution system connection to the grid. Furthermore,
a fault to ground in the side of the inverter is performed with a switch. The voltage drop in the

2
shunt resistor is sensed by the acquisition unit consisting of a USB oscilloscope and a personal
computer. The USB oscilloscope detects the voltage drop and captures the resulting data. The
personal computer is equipped with a custom-built algorithm created in MathWorks MATLAB,
which is used to analyze the collected data and determine the current direction.

Figure 1.2: Distribution Grid with DER Testbed.

1.2.1 Inverter-based DER

The North American Electric Reliability Corporation (NERC) defines DERs as any element in the
distribution system able to generate electricity and is not comprised of the Bulk Electric System
defined by NERC [9]. Some resources, such as microturbines, combined heat and power systems,
and small synchronous generators, can directly connect to the grid. However, other generator units,
such as Photo Voltaic (PV), wind turbines, and fuel cells, require an inverter to convert their
generated DC electricity to AC electricity that meets the grid's rated operational values. These
inverter-based DERs are typically represented by PV solar panels, wind turbines, and fuel cells. In
recent decades, the former group has seen an increase in participation, and a further rise in their
adoption is expected in the near future [7] and [10], and they are known as inverter-based resource
(IBR).
For this study IBR is manufactured by Texas Instruments, with a model number of TIEVM-HV-
1PH-DCAC. This is a single-phase device designed to operate as a grid-tied inverter with an output
LCL filter. It can be controlled using Code Composer Studio software via a USB link. A
controllable DC power source is chosen to emulate the DC electricity generated by the PV, which
operates at 150 Vdc.
1.2.2 Acquisition Station
The acquisition station comprises a shunt resistor, a digital oscilloscope, and a personal computer.
The shunt resistor, manufactured by Macrodyne, has a resistance of 0.05 Ω. The digital

3
oscilloscope is the Analog Discovery 2 model from Digilent, with two differential analog channels
that digitize the analog signal with a 14-bit resolution, a sample rate of 10 MS/s, and a bandwidth
of 10 MHz. The personal computer (PC) has an Intel Core i7-9750H at 2.6 GHz, 16 GB RAM, and
it runs on 64-bit Windows 10. In addition, the PC has the Waveforms software installed, developed
by Digilent, which serves as the interface between the oscilloscope and the PC.
1.2.3 Lines and Grid
According to [11][12], distribution line impedances can have an R/X ratio between 0.6 and 3.0.
This project model one feeder composed of an inductor of 32.5 mH and a resistor of 8 Ω, having
a R/X ratio equal to 0.64.
The distribution grid is simulated with a variable AC power source 0-140Vac 60Hz operating at
50 Vac.
1.2.4 Loads
In this experiment, two resistive loads were implemented, named Load 1, Load 2, in Figure 1.2
Both loads have a resistance value of 50 Ω. Load 1 is continuously connected to the grid, while
Load 2 is controlled by switch 2, allowing control over the power flow.
The testbed components are summarized in Table 1.1.

Table 1.1: Testbed Hardware elements

Testbed component Manufacturer Observations

Inverter Texas Instruments TIEVM-HV-1PH-DCAC
USB Oscilloscope Digilent Analog Discovery 2
Shunt resistor Microdyne 0.050 Ω
Grid Staco Energy Products 50Vac, 60Hz
Feeder Power and Energy Center ≅ 14.8 Ω
Loads Ohmite 50 Ω

1.3 Thesis Outline

The main objective of this work is to test the CODR’s technique for sensing current directionality
for RPF detection in a distribution network that includes DERs, while accounting for the
limitations of existing protection equipment. The following chapters are included: Chapter 2
presents the literature review for sensing directionality currents in relays and signal digitalization.

4
Chapter 3 explains the design and implementation of the testbed. The results of the experiments
conducted are presented and discussed in Chapter 4. Finally, Chapter 5 provides the conclusions
drawn from the study and offers recommendations for future work.

5
Chapter 2

Literature Review

2.1 Fault detections

In the event of a fault within power systems, it is typical for the magnitude of the current to increase
while the voltage is likely to decrease. Furthermore, system variations may occur in frequency,
harmonic components, power demand, and phase angles between current and voltages. Directional
relays utilize the latter phenomenon to determine the direction of the current by measuring the
angular difference between voltage and current [13].
2.2 Classic directional relay
Directional overcurrent relaying is essential in multiple source circuits to limit relay tripping for
faults occurring in only one direction. This type of relay compares the relative phase angle between
two AC quantities to determine the direction of the current flow. Directional relays are used to
coordinate all relays that detect a specific fault, requiring two inputs: operating current and a
constant reference quantity, known as the polarizing quantity, traditionally the measured voltage
where the relay is installed [13].
2.2.1 Phase Angle Comparison
For regular power flow in a determined direction, the angular difference between voltage and
current is determined by the power factor angle 𝜙𝜙 ; for the reverse of the power flow, the difference
will be (180° + 𝜙𝜙). The same behavior can be considered for the forward and reverse direction of
a fault current, considering as 𝜑𝜑 and (180° + 𝜑𝜑), been 𝜑𝜑 the angular difference between voltage
and current in a fault. In transmission systems, the phase angle (𝜑𝜑) is approximately -90° due to
the predominantly inductive nature of transmission lines. However, for distribution systems the
phase angle (𝜑𝜑) may differ from 90° because the resistive components of its lines cannot be
neglected, especially in medium and low voltage systems.

6
Directional relays are designed to take advantage of the difference in phase relationships caused
by a fault, responding to the phase angle differences between two input quantities, such as fault
voltage and fault current [1].

Figure 2.1: Comparison of phase angle Voltage and forward-reverse load-fault current.

2.3 Current-only directional relay

The directional relay requires two quantities to be compared to establish the direction of the power
flow expecting that the polarizing quantity’s phase angle should remain relatively constant, even
when the other quantity experiences significant changes in its phase angle. Traditionally, both
voltage and current have been used to determine direction in directional relays. However, recent
research suggests that current only can be used to determine its path [14]-[20]. Implementing this
solution can result in significant economic benefits because the non-directional OCR would be
updated rather than replaced, eliminating the need for voltage transformers installation increasing
reliability in faults close to where the current relay is installed.
2.3.1 Principle
The authors in [14]-[16] establish the principle in the OCR considering the analysis of Figure 2.2.

Figure 2.2: Pre-fault, forward, and reverse fault current in a distribution system with DER [15].

7
Assuming the power flow forward direction going from the grid to the DER if the fault occurs at
A, the current 𝐼𝐼𝑅𝑅 will be determined by the voltage 𝑉𝑉𝑆𝑆 and the value of the line impedance between
the DER and location A. This can be expressed as:
𝑉𝑉𝑆𝑆
𝐼𝐼𝑅𝑅 = (2.1)
𝑍𝑍𝑆𝑆𝑆𝑆
Similarly, if a fault occurs at location B, the current 𝐼𝐼𝐹𝐹 will be defined by the voltage 𝑉𝑉𝐺𝐺 and the
line impedance between the grid source and location B. The relationship can also be expressed as:
𝑉𝑉𝐺𝐺
𝐼𝐼𝐹𝐹 = (2.2)
𝑍𝑍𝐺𝐺𝐺𝐺
Therefore, using Kirchhoff ‘s current law, the reverse and forward currents sensed by the relay
when a fault occurs are determined by:
𝑉𝑉𝑆𝑆
𝐼𝐼𝑟𝑟𝑟𝑟𝑟𝑟 = 𝐼𝐼𝑝𝑝𝑝𝑝𝑝𝑝 − (2.3)
𝑍𝑍𝑆𝑆𝑆𝑆
𝑉𝑉𝐺𝐺
𝐼𝐼𝑓𝑓𝑓𝑓𝑓𝑓 = 𝐼𝐼𝑝𝑝𝑝𝑝𝑝𝑝 + (2.4)
𝑍𝑍𝑆𝑆𝑆𝑆
For transmission systems, impedances 𝑍𝑍𝑆𝑆𝑆𝑆 and 𝑍𝑍𝐺𝐺𝐺𝐺 their values are unknown, but their nature
can be assumed as a reactive inductance, resulting in the phasor diagram depicted in Figure 2.3.

Figure 2.3: Phasor diagram for currents sensed by the relay [15].

According to the above figure, a forward fault will result in a change in the phase angle of the
measured current between 0° and less than -90°. On the other hand, a reverse fault will cause a
change in the phase angle of the sensed current between 0° and less than 90°.

8
2.3.2 Techniques for detection
Different authors have proposed various techniques for varying the current’s phase angle. Eissa in
[14] uses pre-recorded data from faults on 138 kV and 240 kV systems to address that for a forward
fault, the absolute sum of the fault current signal and the polarized current signal is higher than the
absolute value of the fault current signal. In contrast, for a reverse fault, the absolute sum of these
two signals is lower than the absolute value of the fault current signal. The author in [15] proposes
a phasor estimation of the positive sequence current using the Kalman Filter technique. Phasors
are calculated at each data cycle, allowing their phase angles to be computed and used to detect
the angular difference between two consecutive phasors. In [16] a phasor is estimated for the
sensed current, employing the discrete Fourier transform (DFT) to calculate the phase angle
between two consecutive phasors, arguing that this method is faster than Kalman Filtering. Authors
in [17] remark that the detection offered in [16] is not able to detect an accurate current direction
if the power flow previous to the fault is in the inverse course; to address this issue, they proposed
to use the signal present in the cable shield to determine the accurate direction of the fault current
independently of the pre-fault current direction. Another solution to the problem mentioned above
is presented in [8], which notes that a 180° difference in the phase angle of the pre-fault current
indicates a change in power flow. As a result, the logic used to detect the direction of the fault
current should also be changed. The 180° changes of the pre-fault current are stored and used as a
condition to determine the logic in the algorithm of the fault direction detection.
Other methods for detecting the direction of a fault current have been explored. One approach,
described in [18], uses the rise in the current slope and Lissajous curves to determine the direction
of the faulted system. Another method, presented in [19], calculates the reference quantity using
the square of the first-order derivative of a fault current, which changes polarity when the fault
current crosses zero. The author in [20] developed a detection system that uses only the post-fault
current as the polarizing reference. This work improves on the technique presented in [8] to
implement the detection algorithm.
2.4 Phasor Estimation
By the end of the decade of 1970s, the relaying technique based on symmetrical components
analysis of line voltage and current increased the efficiency for calculating the fault current in a
wide-area power system. This encouraged the development of modern phasor measurement
systems, which with the advance of the Global Positioning System (GPS), was the most effective

9
way for synchronizing power systems over large distances [21]. The use of time synchronization
allowed the use of remote angles as references leading to the development of algorithms that
computed the whole phasor (magnitude and angle).
2.4.1 Phasor representation
A phasor is a way to represent a sinusoidal function to facilitate the analysis of the operation of
power systems. Considering that a sine function can be as:

𝑥𝑥(𝑡𝑡) = 𝑋𝑋𝑚𝑚 cos(𝜔𝜔𝜔𝜔 + 𝜙𝜙) = Re�𝑋𝑋𝑚𝑚 e𝑗𝑗(𝜔𝜔𝜔𝜔+𝜙𝜙) � = 𝑅𝑅𝑅𝑅[�e𝑗𝑗𝑗𝑗𝑗𝑗 �𝑋𝑋𝑚𝑚 e𝑗𝑗𝑗𝑗 ] (2.5)

Knowing that e𝑗𝑗𝑗𝑗𝑗𝑗 is the system frequency, the sinusoid can be re-expressed as a complex function
called phasor.
𝑋𝑋𝑚𝑚 𝑋𝑋𝑚𝑚
𝑋𝑋 = � � e𝑗𝑗𝑗𝑗 = (𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐 + 𝑗𝑗𝑗𝑗𝑗𝑗𝑗𝑗𝑗𝑗) (2.6)
√2 √2
The magnitude of the phasor and its angle 𝜙𝜙 represent the RMS magnitude value and the phase
shift of the original sine function, respectively.
2.4.2 Signal sampling
Phasor estimation requires that the analog signal be sampled and digitalized to be processed. To
efficiently digitize signals, it is necessary to take data samples at uniform periods 𝑘𝑘∆𝑇𝑇.
Considering the sinusoidal signal 𝑥𝑥(𝑡𝑡) being sampled at intervals of 𝑘𝑘∆𝑇𝑇, in which each magnitude
is 𝑥𝑥(𝑘𝑘∆𝑇𝑇)
∞

𝑥𝑥′(𝑡𝑡) = � 𝑥𝑥(𝑘𝑘∆𝑇𝑇)𝛿𝛿(𝑡𝑡 − 𝑘𝑘∆𝑇𝑇) (2.7)

𝑘𝑘=−∞

The sampled signal, 𝑥𝑥′(𝑡𝑡), is obtained by multiplying the magnitude of the original signal, x(t),
with a unit impulse train function, 𝛿𝛿(𝑡𝑡), at regular intervals of 𝑘𝑘∆𝑇𝑇. To perform frequency analysis,
it is essential to obtain the Fourier transform, 𝑋𝑋′(𝑡𝑡), of the sampled signal 𝑥𝑥′(𝑡𝑡). The following
expression can represent this:
∞
1 𝑘𝑘
𝑋𝑋′(𝑓𝑓) = � 𝑋𝑋 �𝑓𝑓 − � (2.8)
∆𝑇𝑇 ∆𝑇𝑇
𝑘𝑘=−∞

The Fourier transform of the sampled data comprises an infinite sequence of Fourier transforms of
𝑥𝑥(𝑡𝑡), this is 𝑋𝑋, centered at frequency intervals of 𝑘𝑘∆𝑇𝑇 for all 𝑘𝑘. Noting the sampling interval is
1
denoted by ∆𝑇𝑇, resulting in a sampling frequency 𝑓𝑓𝑠𝑠 = ∆𝑇𝑇.

10
The Fourier transforms are limited to a cutoff frequency 𝑓𝑓𝑐𝑐 , which must be less than half of the
sampling frequency to avoid the phenomenon of aliasing, which is an overlapping of two
consecutive sampled Fourier transformations, resulting in an error in frequency estimation. This
constraint is known as the Nyquist criterion. A detailed analysis of the expressions indicated in this
section can be consulted in [21].
2.4.3 Discrete Fourier Transform
The Discrete Fourier Transform (DFT) is the representation of the frequency domain of an input
signal 𝑥𝑥(𝑡𝑡) for a determined number of samples, 𝑁𝑁, of 𝑥𝑥(𝑡𝑡). The DFT is calculated at discrete
frequency steps, just as the input signal is sampled at discrete time instants.
To obtain the desired 𝑁𝑁 samples, a window function, 𝑤𝑤(𝑡𝑡), can multiply the train of samples of
𝑥𝑥(𝑡𝑡), considering 𝑤𝑤(𝑡𝑡) as a rectangular function with a unity magnitude and a span of 𝑁𝑁∆𝑇𝑇.
Therefore, the samples selected from 𝑥𝑥(𝑡𝑡) can be addressed as:
𝑁𝑁−1

𝑦𝑦(𝑡𝑡) = 𝑥𝑥(𝑡𝑡)𝛿𝛿(𝑡𝑡)𝑤𝑤(𝑡𝑡) = � 𝑥𝑥(𝑘𝑘∆𝑇𝑇)𝛿𝛿(𝑡𝑡 − 𝑘𝑘∆𝑇𝑇) (2.9)

𝑘𝑘=0

Hence, the DFT will be the Fourier transform of the windowed samples of 𝑥𝑥(𝑡𝑡) which can be
expressed as:
𝑁𝑁−1
j2πkn
𝑋𝑋′(𝑓𝑓) = � 𝑥𝑥(𝑘𝑘∆𝑇𝑇)e− N (2.10)
𝑘𝑘=0

Where the DFT 𝑋𝑋′ of an input signal 𝑥𝑥(𝑡𝑡) can be obtained by adding the product of the amplitude
2πkn
values 𝑥𝑥(𝑡𝑡), given in regular intervals of ∆𝑇𝑇, and a periodic function with phase equal N
, noting
2π
that the step angle is N
. A detailed analysis of the expressions indicated in this section can be

consulted in [21].
2.4.4 Representing phasors with DFT
Considering that the phasor representation of a sinusoid 𝑥𝑥(𝑡𝑡) regarding its Fourier series
coefficient 𝑎𝑎𝑘𝑘 and 𝑏𝑏𝑘𝑘 can be expressed as:
1 𝑏𝑏
𝑋𝑋𝑘𝑘 = ��𝑎𝑎𝑘𝑘2 + 𝑏𝑏𝑘𝑘2 � 𝑒𝑒 𝑗𝑗𝑗𝑗 with 𝜙𝜙 = arctan �− 𝑘𝑘 � (2.11)
√2 𝑎𝑎 𝑘𝑘

And knowing that the Fourier series coefficients of a sinusoidal function are equal to 2/N times its
DFT sampled data, expression 2.10 can be written as:

11
 𝑁𝑁−1
1 2 𝑗𝑗2𝜋𝜋𝜋𝜋𝜋𝜋
𝑋𝑋𝑘𝑘 = � 𝑥𝑥(𝑛𝑛∆𝑇𝑇)𝑒𝑒 − 𝑁𝑁
√2 𝑁𝑁 𝑛𝑛=0
𝑁𝑁−1
√2 2𝜋𝜋𝜋𝜋𝜋𝜋 2𝜋𝜋𝜋𝜋𝜋𝜋
= � 𝑥𝑥(𝑛𝑛∆𝑇𝑇) �cos � � − 𝑗𝑗 sin � �� (2.12)
𝑁𝑁 𝑁𝑁 𝑁𝑁
𝑛𝑛=0

Considering that expression 2.12 represents the 𝑘𝑘𝑘𝑘ℎ harmonic of the periodic signal and stating
𝑥𝑥(𝑛𝑛∆𝑇𝑇) = 𝑥𝑥𝑛𝑛 , and the sampling angle 𝜃𝜃 = 2𝜋𝜋/𝑁𝑁 dependent on the period of the fundamental
frequency component, the DFT of the phasor is:
𝑁𝑁−1
√2
𝑋𝑋𝑘𝑘 = � 𝑥𝑥𝑛𝑛 {cos(𝑘𝑘𝑘𝑘𝑘𝑘) − 𝑗𝑗 sin(𝑘𝑘𝑘𝑘𝑘𝑘)} = 𝑋𝑋𝑘𝑘𝑘𝑘 − 𝑗𝑗𝑗𝑗𝑘𝑘𝑘𝑘 (2.13)
𝑁𝑁
𝑛𝑛=0

As 𝑥𝑥𝑛𝑛 = 𝑋𝑋𝑚𝑚 cos (𝑛𝑛𝑛𝑛 + 𝜙𝜙), and phasor measurement cares for the fundamental frequency, which
means 𝑘𝑘 = 1, the cosine summation results in
𝑁𝑁−1
√2 𝑋𝑋𝑚𝑚
𝑋𝑋𝑐𝑐𝑁𝑁−1 = � 𝑋𝑋𝑚𝑚 cos(𝑛𝑛𝑛𝑛 + 𝜙𝜙) cos(𝑛𝑛𝑛𝑛) = cos 𝜙𝜙 (2.14)
𝑁𝑁 √2
𝑛𝑛=0

Applying the procedure for the sine summation of expression 2.13

𝑁𝑁−1
√2 𝑋𝑋𝑚𝑚
𝑋𝑋𝑠𝑠𝑁𝑁−1 = � 𝑋𝑋𝑚𝑚 cos(𝑛𝑛𝑛𝑛 + 𝜙𝜙) sin(𝑛𝑛𝑛𝑛) = − sin 𝜙𝜙 (2.15)
𝑁𝑁 √2
𝑛𝑛=0

Noting superscript N-1 in 𝑋𝑋𝑐𝑐 and 𝑋𝑋𝑠𝑠 identify the phasor according to the last N-1 sample for its
estimation. Joining formulas 2.14 and 2.15 gives the phasor 𝑋𝑋 𝑁𝑁−1
𝑋𝑋𝑚𝑚 𝑋𝑋𝑚𝑚
𝑋𝑋 𝑁𝑁−1 = [cos 𝜙𝜙 + 𝑗𝑗sin 𝜙𝜙] = 𝑒𝑒 𝑗𝑗𝑗𝑗 (2.16)
√2 √2
A detailed analysis of the expressions indicated in this section can be consulted in [21].
2.4.5 Phasor updating
Given that the calculation of phasors in a relay is an ongoing process, it is essential to use
algorithms that update the phasor estimate as new data samples are acquired. A non-recursive
algorithm consists of repeating the calculations from equation (2.16) for every new sample data
that is coming, having the following expression:

12
 𝑁𝑁−1
√2
𝑋𝑋 𝑁𝑁−1 = � 𝑥𝑥𝑛𝑛 [cos(𝑛𝑛𝑛𝑛) − 𝑗𝑗sin(𝑛𝑛𝑛𝑛)]
𝑁𝑁
𝑛𝑛=0
𝑁𝑁−1
√2
𝑋𝑋 𝑁𝑁 = � 𝑥𝑥𝑛𝑛+1 [cos(𝑛𝑛𝑛𝑛) − 𝑗𝑗sin(𝑛𝑛𝑛𝑛)] (2.17)
𝑁𝑁
𝑛𝑛=0

For this algorithm, if the frequency of the time domain signal is constant, the phasor estimate will
have constant magnitude and will rotate in the counterclockwise direction by an angle 𝜃𝜃 at every
new sample data. This algorithm is stable but lacks computing efficiency compared to the
following algorithm.
If two consecutive phasors are generated with the non-recursive algorithm, they will share almost
the same data samples, except phasor 𝑋𝑋 𝑁𝑁−1 will not have the last sample 𝑋𝑋 𝑁𝑁 , and the phasor 𝑋𝑋 𝑁𝑁
will not include the first sample of 𝑋𝑋 𝑁𝑁−1. This can be addressed as:
𝑁𝑁−1
𝑁𝑁 √2
𝑋𝑋 = � 𝑥𝑥𝑛𝑛+1 𝑒𝑒 −𝑗𝑗(𝑛𝑛+1)𝜃𝜃
𝑁𝑁
𝑛𝑛=0

√2 (2.18)
𝑋𝑋 𝑁𝑁 = 𝑋𝑋 𝑁𝑁−1 + (𝑥𝑥𝑁𝑁 − 𝑥𝑥0 )𝑒𝑒 −𝑗𝑗(0)𝜃𝜃
𝑁𝑁
As this algorithm employs the calculation of the previous phasor 𝑋𝑋 𝑁𝑁 , is called the recursive
algorithm. This method is more efficient in estimating phasors, but if not properly implemented
the sampling error propagates in more phasors when compared with the non-recursive algorithm.
A detailed analysis of the expressions indicated in this section can be consulted in [21].

13
Chapter 3

Testbed Implementation

In [14], [15], and [16], the CODR’s performance is evaluated using clean environment pre-
recorded data, and Simulink/MATLAB simulations. This work aims to test the OCR in a hardware
testbed that more closely emulates a distribution system, as the primary objective of this relay is
to be applied in distribution systems with DER. Therefore, more accurate results can be obtained
if the OCR is assessed in an environment with actual distribution system elements. Considering
that this testbed could be implemented in Power Lab classes, for safety concerns, the nominal
value for the experiment will be 50 Vac.
Because this thesis aims to test the detection of fault current direction, a simple testbed is
implemented considering a single-phase system with two sources: a conventional and a non-
conventional source representing the grid and a DER, respectively. These sources are connected
through a feeder, including inductive and resistive components to describe the distribution line. At
each end of the feeder, there are two resistive loads. The CODR is suited at the side of the DER.
Figure 3.1 shows the system diagram.

Figure 3.1: Single-phase distribution system

The testbed had two main components: the hardware representing the distribution system and the
software that processes, computes, and controls the information obtained from the hardware
system.
3.1 Hardware
The hardware consists of loads, sources, a feeder, the current measurement, the oscilloscope,
switches, and two computers.

14
3.1.1 Loads
To simplify the system, the two loads are pure resistors, each drawing 1 A of current. This means
that their resistance value is 50 Ω. Ohmite manufactures the resistors employed in the testbed, its
resistance can be adjusted from 0-50 Ω, and its power rating is up to 280W.
The resistors are named Load 1 and Load 2, and their terminals are connected in nodes 1 and 3,
respectively, and the system's reference.
3.1.2 Sources
3.1.2.1 IBR
As mentioned in section 1.2.1, IBRs are expected to become more prevalent in distribution
systems, therefore is a valid and beneficial choice to include in this experiment. Texas Instrument
offers the model TIEVM-HV-1PH-DCAC, a single-phase inverter development kit that supports
voltage source and grid-connected modes, being the last employed in this experiment. The control
for the single-phase inverter is implemented using the C2000™ F2837xD microcontroller
connected to a PC manufactured by Dell, model OptiPlex.
When viewed as a black box, the single-phase inverter has two power connections. The first is the
DC input that feeds the DER. In this case, the input is provided by a DC source manufactured by
Sorensen, model DCR 150-12B. The second connection to the IBR is the AC output attached to
node 1 of the system and the ground.
Table 3.1: Single-phase inverter TIEVM-HV-1PH-DCAC features [22].

According to the IBR manual [22], the DC voltage must be greater than the grid voltage to ensure
power flows from the inverter to the grid. Knowing that the system voltage is 50 Vac and based on
the relationship between input and output voltages indicated in Table 3.1, the input voltage for the

15
DC source is set at 150 Vdc. Additionally, since each load draws 1 A, they will consume 100 W.
Therefore, the DC power source and the IBR must meet this power requirement.
3.1.2.2 Grid
The grid is emulated with an AC power source 0-240 Vac, 3300 W, 60 Hz. This power source can
select between three-phase and single-phase output, with the latter chosen for node 1 and ground
connection. The voltage level output is adjusted to 50 Vac using a dial on the power source and
monitored with a Fluke model 45 multimeter. The power requirement for this circuit is 100 W, as
this source will feed two loads.
3.1.2.3 Feeder
Distribution lines cover a shorter distance than transmission lines, the short model is adequate for
this test. The main characteristic of a distribution line is that is modeled as a series branch
composed of resistive and inductive elements, and its length is restricted to 50 miles.
The resistive component is more prominent than in transmission lines, hence resistance should not
be avoided when modeling distribution lines, especially when they conform to medium and low-
voltage systems. According to [12], the impedance value of a feeder of 12 kV and 35 kV are 0.122
+ j0.175 Ω/mile and 0.119 + j0.190 Ω/mile, respectively. Taking the former impedance value and
assuming that a feeder can have 15 MVA triphasic capacity, the scale-down model of the feeder
can be obtained as indicated [23]:
𝑆𝑆𝑠𝑠𝑠𝑠𝑠𝑠(1−𝜙𝜙) = 5 MVA
12
𝑉𝑉𝑠𝑠𝑠𝑠𝑠𝑠(𝐿𝐿−𝑁𝑁) = kV = 6.93 kV
√3
2
𝑉𝑉𝑠𝑠𝑠𝑠𝑠𝑠(𝐿𝐿−𝑁𝑁)
𝑍𝑍𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 = = 9.6 Ω (3.1)
𝑆𝑆𝑠𝑠𝑠𝑠𝑠𝑠(1−𝜙𝜙)
𝑅𝑅𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙(Ω/mile) 0.122
𝑅𝑅𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠(𝑝𝑝.𝑢𝑢.) = = = 0.0127 𝑝𝑝. 𝑢𝑢/𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚 (3.2)
𝑍𝑍𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠(Ω) 9.6
𝑋𝑋𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙(Ω/mile) 0.175
𝑋𝑋𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠(𝑝𝑝.𝑢𝑢.) = = = 0.0182 𝑝𝑝. 𝑢𝑢/𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚 (3.3)
𝑍𝑍𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠(Ω) 9.6
Performing the change to the base of the testbed:
𝑆𝑆𝑡𝑡𝑡𝑡𝑡𝑡(1−𝜙𝜙) = 100 𝑉𝑉𝑉𝑉
𝑉𝑉𝑡𝑡𝑡𝑡𝑡𝑡(𝐿𝐿−𝑁𝑁) = 50 V

16
 2
𝑉𝑉𝑡𝑡𝑡𝑡𝑡𝑡(𝐿𝐿−𝑁𝑁)
𝑍𝑍𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡 = = 25 Ω (3.4)
𝑆𝑆𝑡𝑡𝑡𝑡𝑡𝑡(1−𝜙𝜙)
𝑅𝑅𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡(Ω/𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚) = 𝑅𝑅𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠(𝑝𝑝.𝑢𝑢.) ∗ 𝑍𝑍𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡 = 0.3177 Ω/𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚 (3.5)

𝑋𝑋𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡(Ω/𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚) = 𝑋𝑋𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠(𝑝𝑝.𝑢𝑢.) ∗ 𝑍𝑍𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡 = 0.4557 Ω/𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚 (3.6)

Assuming the feeder emulated is 25 miles long, the testbed line impedance is 7.94 + j11.4 Ω. The
elements selected are an 8 Ω resistor, manufactured by Ohmite with 500 W capability, and an
inductor of 33.5 mH which gives a resistive inductance of 12.6 Ω, considering 60 Hz.
The resistors are connected in series to form the feeder, whose terminals are connected to node 2
and node 3.
3.1.2.4 Current measurement
The current is measured using a 0.050 Ω shunt resistor manufactured by Macrodyne Inc.. The
power terminals of the shunt resistor are connected to nodes 1 and 2. At the same time, the
measurement terminals are joined to a motherboard where the USB oscilloscope probe is also
connected to the motherboard to measure the voltage across the shunt resistor.
3.1.2.5 USB Oscilloscope
The USB oscilloscope manufactured by Digilent model Analog Discover 2, is employed to acquire
and digitalize the current signal coming from the shunt resistor. Its analog terminal is connected to
the motherboard, where the terminals of the shunt resistor are connected. The digital output is
connected to the laptop Lenovo Thinkpad P53, where the digital data can be observed and stored
with dedicated software.
3.1.2.6 Switch
A double pole double throw (DPDT) switch is fixed across node 1 and ground to perform the fault-
to-ground of the experiment.
Figure 3.2 and the equivalent circuit in Figure 3.3 show the physical elements utilized in the
testbed.
3.2 Software
The software components are to control the inverter, read, and store the data samples coming from
the current measurement, and for data processes implementing the detection algorithm.

17
 Figure 3.2: Testbed hardware components

Figure 3.3: Testbed equivalent circuit

3.2.1 Code Composer Studio

Code Composer Studio (CCS) is an integrated development environment (IDE) for developing
and debugging embedded applications on Texas Instruments microcontrollers and processors. The
grid-tied inverter mode implementation, and its control, is operated with the software v12.2.0.
The powerSuite Page section of the software [22] allows to set different features of the grid-tied
inverter, Table 3.2 summarize them.
It is worth to mention the fault current will be generated connecting one tie to earth, having a value
of 𝐼𝐼𝑓𝑓 = 𝑉𝑉𝑛𝑛𝑛𝑛𝑛𝑛 /|𝑍𝑍𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙 | = 50 /14.92 = 3.35 A. Therefore, the fault will demand 167 W. To have a
slack, a power rating of 200 Watts.

18
 Table 3.2: Grid-tied inverter main features on powerSuite page
Feature Value/configuration
Type of program CLOSED Current Loop & Grid Sync
Output AC
Frequency [Hz] 60
Power rating [W] 200
Power operating [W] 100
Vrms output [Vac] 50
Vbus (input) [Vdc] 150

The control scheme of this inverter is Current control. The grid voltage is sensed, and a phase lock
loop (PLL) algorithm is used to compute the angle and phase of the grid. This value is multiplied
with the reference current command to generate the instantaneous current command reference,
which is compared with the sensed output current. The error is fed into the current compensator,
which uses proportional resonant (PR) controllers to zero the error at the AC frequency [22].
The amount of power injected into the grid is controlled with the reference current command,
which can be manipulated during the experiment with the invIoRef expression shown in the
Expressions window. This window is beneficial to the experiment since it shows the measurements
that the inverter is taken from the system. Hence, parameters such as voltage input, voltage output,
power injected, and frequency can be observed during the experiment. See Figure 3.4.

Figure 3.4: CCS Expression window [22]

19
3.2.2 Waveforms
Waveforms is a software developed by Digilent that displays the data obtained by the USB portable
oscilloscope used in this testbed. The oscilloscope software provides data acquisition, triggering,
and viewing functionality like a benchtop scope. It is the former feature that is employed by this
work, the signal obtained from the shunt resistor is viewed and recorded with Waveforms
oscilloscope. Once the experiment is finished, the data is exported in a CSV format file for further
analysis.
One crucial aspect to be considered is the selection of the oscilloscope sample frequency because
this value is one of the inputs for the phasor estimation algorithm. This project uses a 3 kHz as a
sample rate as used by precision based on relays [24].

Figure 3.5: Waveforms oscilloscope sample rate configuration

3.2.3 MATLAB
MATLAB is a proprietary programming language and computing environment developed by
MathWorks. It allows users to perform matrix manipulations, plot data, implement algorithms,
create user interfaces, and interact with other languages.
The phasor estimation algorithm is based on Equation (2.17) and is implemented as follows: the
data samples recorded on the CVS file from each simulation are read in windows lengths

20
determined by the sampling frequency and the nominal frequency, indicating the samples per cycle
that will compose a phasor. That means, 𝑁𝑁 = 𝑓𝑓𝑠𝑠 / 𝑓𝑓𝑛𝑛 = 3000 / 60 = 50 𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠. Then, the
angular step 𝜃𝜃 is calculated dividing 2𝜋𝜋/𝑁𝑁 and each of these steps are multiplied by the
corresponding sample value of the window. When the multiplication of the data sample and the
angular step reach the length of the windows, all those products are added and stored as a phasor.
This procedure is repeated with a new window that will start immediately after the last sample of
the previous windows, establishing the ratio of one phasor per cycle for this system.
With the vector of the phasor is easy to calculate its magnitude and angle with the MATLAB
functions abs and angle, respectively. A loop runs the phasor vector obtaining their angle and
subtracting two consecutive angles; if the angular difference is between 0° and -90° is due to a
forward fault, it is between 0° and 90° is a reverse fault, as Figure 2.3 indicates.
As every simulation starts with a pre-fault current, the angle of the first phasor is taken as reference
and subtracted for the rest of the phasor angles. This generates that the pre-fault angle will be 0°
facilitating the reading of the results.
One important aspect to consider when analyzing phasors is the presence of transient samples.
These samples do not provide an accurate depiction of the system’s state and should be discarded
[21]. To address this issue, the implemented algorithm searches for and compares the standard
deviation of two consecutive groups of samples. If the standard deviation exceeds a certain
threshold, indicating the presence of transient values, the phasor is discarded. This ensures that
only phasors that accurately represent the system’s state are used in the analysis.

Figure 3.6: Distribution hardware testbed for CODR testing

21
Chapter 4

Testing and Results

4.1 Introduction
This chapter presents the testing result obtained from the testbed described in the previous section.
The main goal is to test if the CODR is able to detect the current fault direction in a distribution
grid.
Ideally, the test should be directed by closing switch 1 for the forward fault direction and switch 2
for the reverse fault direction as determined by the pre-fault current and the polarity marks of the
shunt resistor. Figure 1.1: Bidirectional Power flow due to DER integration. Figure 4.1 depicts this
scenario.

Figure 4.1: Ideal circuit for testing

The inverter current control does not respond to changes in system impedances, it responds to the
setting of the invIoRef set manually at the beginning of each test, therefore, the reverse fault
detection cannot be obtained with the circuit of Figure 1.1: Bidirectional Power flow due to DER
integration. Figure 4.1. To solve this constraint, the inverter injects power into the grid at the same
time that the grid injects the pre-fault current, and the shunt resistor measurement polarity will be
inverted. In this case, the shunt resistor will have pre-fault and fault currents in the opposite
direction, and the polarity of the measurement instrument will agree with the pre-fault current
entering into the positive terminal of the shunt resistor.

22
4.2 Forward flow – forward fault testing
This test examines the angular difference between pre-fault and post-fault currents when both flow
in the same direction. The test is performed for different power injections from the IBR setting,
addressing different levels of penetration from the DER. The invIoRef expression is set at values
of 0.02, 0.05, 0.07, and 0.1, corresponding to current injections of 0.24, 0.62, 0.78, and 1.11 A,
respectively. The voltage of the AC source is increased from 50 Vac to compensate for the voltage
drop in the line. This adjustment is stopped when the node where the inverter is connected reaches
50 Vac, as measured by the inverter using the guiVrms expression. To generate the post-fault
current, switch 1 is closed. See Figure 4.2 for details.

Figure 4.2: Forward power flow – forward fault current testing

Results are shown in Figure 4.3 to Figure 4.6. As a general observation, the upper graph of each
figure depicts the phase angle of each phasor, noting the difference between two consecutive
phasors. The bottom plot of each figure displays the amplitude of the currents during the
experiment, pre-fault, and post-fault. Additionally, the phasor angles have a slope because the
phasor estimation algorithm assumes a nominal frequency of 60 Hz. In contrast, the actual
frequency of the system rarely is fixed at that value.
At the moment of the fault, Figure 4.3 shows an angular difference detection of -41.47° and the
ratio between the amplitude of the fault and pre-fault current is the lowest for the forward fault
tests with a value of 4.98.

23
 Phasor angles - 0.02 InvIoRef
50

40

30

20

10

Angle [°]
-10

-20

-30

-40

-50
222 223 224 225 226 227 228 229 230

Phasor number

Pre fault/Post fault - 0.02 InvIoRef

0.4

0.2
Amplitude [mV]

-0.2

-0.4

-0.6
1.11 1.115 1.12 1.125 1.13 1.135 1.14 1.145 1.15

samples 10
4

Figure 4.3: Forward power flow – forward fault 0.02 InvIoRef injection
Phasor angles - 0.05 InvIoRef
50

40

30

20

10

0
Angle [°]

-10

-20

-30

-40

-50
190 191 192 193 194 195 196 197 198

Phasor number

Pre fault/Post fault - 0.05 InvIoRef

0.4

0.2
Amplitude [mV]

-0.2

-0.4

-0.6
9500 9550 9600 9650 9700 9750 9800 9850 9900

samples

Figure 4.4: Forward power flow – forward fault 0.05 InvIoRef injection

The figure above shows the fault current's angular difference is -44.15° and an amplitude ratio of
7.0.

24
 Phasor angles - 0.07 InvIoRef
50

40

30

20

10

Angle [°]
-10

-20

-30

-40

-50
211 212 213 214 215 216 217 218 219

Phasor number

Pre fault/Post fault - 0.07 InvIoRef

0.4

0.2
Amplitude [mV]

-0.2

-0.4

-0.6
1.055 1.06 1.065 1.07 1.075 1.08 1.085 1.09 1.095

samples 10
4

Figure 4.5: Forward power flow – forward fault 0.07 InvIoRef injection

The 0.07 InvIoRef shows an angular difference is -45.14° and the amplitude ratio is 10.4. As in the
previous figure, the waveform of the current lacks harmonics due to a low or IBR power injection.
Phasor angles - 0.10 InvIoRef
50

40

30

20

10

0
Angle [°]

-10

-20

-30

-40

-50
194 195 196 197 198 199 200 201 202

Phasor number

Pre fault/Post fault - 0.10 InvIoRef

0.4

0.2
Amplitude [mV]

-0.2

-0.4

-0.6
9700 9750 9800 9850 9900 9950 10000 10050 10100

samples

Figure 4.6: Forward power flow – forward fault 0.10 InvIoRef injection

25
Figure 4.6 shows an angular difference of -34.88° and the amplitude ratio is 29.12. The ratio
post/pre-fault current is the greatest for this fault. The current waveform presents more harmonics
than previous tests.
4.3 Forward flow – reverse fault testing
The test examines the angular difference between pre-fault and post-fault currents, which flow in
opposite directions. The test is conducted with various power injections from the IBR setting,
representing different levels of DER penetration. The invIoRef expression is set at fixed values of
0.12, 0.15, 0.18, and 0.20, corresponding to current injections of 1.29, 1.67, 2.00, and 2.21 A,
respectively. Throughout the test, the voltage of the AC source is maintained at 50 Vac, and the
oscilloscope probe polarity is inverted to ensure consistency with previous sections. To generate
the post-fault current, switch 1 is closed. Further details can be found in Figure 4.7.

Figure 4.7: Forward power flow – reverse fault current testing

The general observations of the forward fault current apply to the reverse fault testing.
Figure 4.8 shows an angular difference of 122.72°and post/pre fault current ratio of 16.06. The
current waveform presents more harmonic components than all the forward fault current tests.

26
 Phasor angles - 0.12 InvIoRef
160

140

120

100

80

Angle [°]
60

40

20

0
231 232 233 234 235 236 237 238 239

Phasor number

Pre fault/Post fault - 0.12 InvIoRef

0.4

0.2
Amplitude [mV]

-0.2

-0.4

-0.6
1.155 1.16 1.165 1.17 1.175 1.18 1.185 1.19 1.195

samples 10
4

Figure 4.8: Forward power flow – reverse fault 0.12 InvIoRef injection
Phasor angles - 0.15 InvIoRef

120

100

80

60
Angle [°]

40

20

-20

171 172 173 174 175 176 177 178 179

Phasor number

Pre fault/Post fault - 0.15 InvIoRef

0.4

0.2
Amplitude [mV]

-0.2

-0.4

-0.6
8550 8600 8650 8700 8750 8800 8850 8900 8950

samples

Figure 4.9: Forward power flow – reverse fault 0.15 InvIoRef injection

The above figure indicates an angular difference of 119.55° and an aomplitude ratio of 14.88.

27
 Phasor angles - 0.18 InvIoRef

120

100

80

60

Angle [°]
40

20

-20

194 195 196 197 198 199 200 201 202

Phasor number

Pre fault/Post fault - 0.18 InvIoRef

0.4

0.2
Amplitude [mV]

-0.2

-0.4

-0.6
9700 9750 9800 9850 9900 9950 10000 10050 10100

samples

Figure 4.10: Forward power flow – reverse fault 0.18 InvIoRef injection

Figure 4.10 shows an angular difference of 126.97°. The pre-fault current amplitude has been
decreasing according to InvIoRef increases, in this test its value is 8.3.
Phasor angles - 0.20 InvIoRef

120

100

80

60
Angle [°]

40

20

-20

201 202 203 204 205 206 207 208 209

Phasor number

Pre fault/Post fault - 0.20 InvIoRef

0.4

0.2
Amplitude [mV]

-0.2

-0.4

-0.6
1.005 1.01 1.015 1.02 1.025 1.03 1.035 1.04 1.045

samples 10
4

Figure 4.11: Forward power flow – reverse fault 0.20 InvIoRef injection

28
Figure 4.11 indicates a 116.44° angular difference and the smallest post-pre fault amplitude ratio
with a value of 5.12.
Table 4.1 and Table 4.2 summarize the results of all the tests conducted in this experiment.
Table 4.1: Forward fault - summary results

Injection 0.02 0.05 0.07 0.1

Fault
Angular difference [°] -41.47° -44.15° -45.14° -34.88°
Amplitude ratio [p.u.] 4.98 7.00 10.4 29.12
Table 4.2: Reverse fault – summary results

Injection 0.12 0.15 0.18 0.2

Fault
Angular difference [°] 122.72° 119.55° 126.97° 116.44°
Amplitude ratio [p.u.] 16.06 14.88 8.30 5.12

4.4 Discussion
Regarding Figure 2.3 and Figure 4.1, the results for the forward fault current meet expectations.
The angle difference falls in the fourth quadrant of a Cartesian plane, allowing for accurate
detection of a forward fault current in a distribution system. The angle difference remains stable
around -30° and -40°, unaffected by DER penetration.
However, the reverse fault current results do not perfectly match the analysis in [8]. The angular
difference falls in the second quadrant instead of the first. As the penetration of a source increases
(meaning the inverter injection decreases), the angle decreases from 122.72° to 116.44°. However,
it remains in the same quadrant regardless of other sources’ penetration in the system. One
explanation for this discrepancy is that this work considers a distribution line with a resistance
component that affects the fault angle of the system. If the line model includes resistance, the
assumption that the fault impedance is close to 90° is not valid. The angle could be assumed to be
at least 60° or even closer to zero degrees if the resistive component is more prominent in the line
impedance. Figure 4.12 shows a phasor diagram where a lower angle is assumed for the fault angle
impedance. As can be observed, the forward and reverse faults obtained in this experiment fall in
the quadrants indicated by this diagram.

29
 Figure 4.12: Phasor diagram for currents sensed by the relay considering an R+jX distribution line

Although the angle showed a slope due to the off-nominal frequency of the testbed, that did not
impede having a good reading in the angular difference. The angular difference between the pre-
and post-fault current is more significant than the difference presented by the slope. However, that
problem has been fixed in [8] using a DAC with a switching frequency that responds to the system
frequency.
This experiment utilized an IBR to simulate the penetration of DER. Inverters are known for
injecting harmonics into the grid due to the nature of their power electronics components. As
observed, harmonics were present in the recorded currents in the results. However, this setting did
not impede the phasor estimation algorithm because it intrinsically employs a filtering technique
that uses the natural frequency for phasor calculation.
The relationship between the amplitude of the pre-fault and post-fault current is greater than the
ratio typically used by overcurrent (OC) relays to discriminate a fault. This scenario is more
pronounced when two sources are feeding their respective areas and there is only a small amount
of power interchange through the feeder.
4.5 Recommendations
The CODR is a cost-effective solution for addressing the changes that distribution networks are
experiencing with DER. It is a suitable solution for transmission and some distribution systems
which operate at high voltages. However, as the voltage decreases in distribution systems, the
transmission lines usually have a higher resistive component. Suppose the resistive component is
larger than the inductive component, as in lateral feeders or low distribution networks. In that case,

30
the line angle approaches zero, resulting in a power factor similar to loads. In the event of a failure,
the angles of the pre-fault and post-fault conditions are similar, meaning there is less resolution to
discriminate the direction of the fault. This makes CODR unsuitable for these types of networks.
.

31
Chapter 5

Conclusion

5.1 Conclusion
The growth of Distributed Energy Resources (DER) is accelerating, and power system engineers
anticipate significant changes in utility distribution systems in the near future. One of these
changes is the bidirectionality of the power flow in distribution systems, which demands upgrades
in the distribution protection system to address this phenomenon. The CODR is a cost-effective
solution due to can be implemented at the software level of actual OC relays, avoiding replacing
existing equipment in distribution grids. This thesis provides a complete description of how to
implement a hardware testbed that reviews the CODR. The testbed comprises a grid-tied inverter,
an AC source, a feeder, two resistive loads, and a digital oscilloscope that have been interconnected
to form a distribution network to review the CODR fault detection direction algorithm. A code
developed in MATLAB was employed to implement the DFT non-recursive algorithm for a phasor
estimation vector, which was used to detect the difference between the angle of two consecutive
phasors, determining the path of the fault current. The result for the detection of forward fault
agrees with previous investigations, while the angle for reverse fault detection needs to be
upgraded. The reason for the disagreement resides in the assumptions made by previous works in
the components of the line. This result is not going against what was developed by previous
researchers but is a complement to updating the algorithm that discriminates the reverse fault
current when the relay is designed to operate in distribution systems where the line resistance
component cannot be neglected.
Even though the estimated angle presents some deviation due to the system’s non-nominal
frequency, it did not impede testing appropriately. The angular difference between pre and post
current is large enough to discriminate between the two phenomena.
Additionally, The DFT technique is suitable for implementation in CODR in environments with
harmonic components due to its filtering nature.

32
5.2 Main Contributions
This work presents a significant advancement in CODR testing, offering a more realistic and
reliable approach for detecting the fault direction of the relay in real distribution environments.
The proposed testbed utilizes simple elements commonly found in undergraduate Power
Laboratories, making it easily implementable as an experiment for senior or graduate students. The
combination of practicality and accuracy makes it an ideal tool for enhancing students'
understanding of relay operations in real-world scenarios.
5.3 Future works
The developed testbed can be used to provide guidance on how the CODR discriminates the
direction of the fault current in the face of different types of loads. For example, test capacitive
loads, which can be the connection of capacitor banks in the system; inductive loads, such as motor
starts, or lines with an R/X ratio greater than one. The results may lead to adaptive algorithms that
change settings based on detected changes in the grid.
Additionally, the inverter control algorithm could be updated to respond according to changes in
the system load. In this way, it would be possible to see if its transient response affects the detection
of fault direction and avoid changing the polarity of the oscilloscope.
Finally, experiments could be conducted in real-time to determine the latency between the
occurrence of a fault, its detection, and the trip signal. In this case, various phasor estimation
algorithms can be tested to find the most effective one for detecting the direction of the fault.

33
References

[1] J. R. Aguero, E. Takayesu, D. Novosel and R. Masiello, "Modernizing the Grid: Challenges
and Opportunities for a Sustainable Future," in IEEE Power and Energy Magazine, vol. 15,
no. 3, pp. 74-83, May-June 2017, doi: 10.1109/MPE.2017.2660819.

[2] P. Denholm, D. J. Arent, S. F. Baldwin, D. E. Bilello, G. L. Brinkman, J. M. Cochran, et

al., "The challenges of achieving a 100% renewable electricity system in the United States,"
Joule, vol. 5, no. 6, pp. 1331-1352, Jun. 2021, doi: 10.1016/j.joule.2021.03.028.

[3] P. Denholm, P. P. Brown, W. Cole, et al., "Examining Supply-Side Options to Achieve

100% Clean Electricity by 2035," National Renewable Energy Laboratory, Golden, CO,
USA, Tech. Rep. NREL/TP-6A40-81644, Sep. 2021, doi: 10.2172/1847089.
[4] A. H. Javed, P. H. Nguyen, J. Morren and J. G. H. Slootweg, "Review of Operational
Challenges and Solutions for DER Integration with Distribution Networks," 2021 56th
International Universities Power Engineering Conference (UPEC), Middlesbrough, United
Kingdom, 2021, pp. 1-6, doi: 10.1109/UPEC50034.2021.9548198.

[5] U. Shahzad, S. Kahrobaee, and S. Asgarpoor, "Protection of Distributed Generation:

Challenges and Solutions," Energy and Power Engineering, vol. 9, no. 10, pp. 497-508,
Oct. 2017, doi: 10.4236/epe.2017.910042

[6] J. G. Kassakian, R. Schmalensee, G. Desgroseilliers, T. D. Heidel, K. Afridi, A. M. Farid

et al., "The Future of the Electric Grid: An Interdisciplinary MIT Study," Massachusetts
Institute of Technology, MIT Energy Initiative, December 2011. [Online]. Available:
http://energy.mit.edu/wp-content/uploads/2011/12/MITEI-The-Future-of-the-Electric-
Grid.pdf.

[7] K. Horowitz, Z. Peterson, M. Coddington, F. Ding, B. Sigrin, D. Saleem, et al., "An

Overview of Distributed Energy Resource (DER) Interconnection: Current Practices and
Emerging Solutions," National Renewable Energy Laboratory, Golden, CO, Technical
Report NREL/TP-6A20-72102, December 2019. [Online]. Available:
https://www.nrel.gov/docs/fy20osti/72102.pdf.

[8] A. Ukil, "Detection of direction change in prefault current in current-only directional

overcurrent protection," IECON 2016 - 42nd Annual Conference of the IEEE Industrial
Electronics Society, Florence, Italy, 2016, pp. 3829-3833, doi:
10.1109/IECON.2016.7793704

[9] North American Electric Reliability Corporation, “Distributed Energy Resources:

Connection Modeling and Reliability Considerations,” February 2017. [Online]. Available:
https://www.nerc.com/comm/Other/essntlrlbltysrvcstskfrcDL/Distributed_Energy_Resour
ces_Report.pdf

34
[10] J. E. O’Brien et al., "Solar Futures Study," U.S. Department of Energy, Washington, DC,
USA, Sep. 2021. [Online]. Available: https://www.energy.gov/sites/default/files/2021-
09/Solar%20Futures%20Study.pdf.

[11] T. A. Short, "Electric Power Distribution Handbook," Boca Raton, FL: CRC Press, 2014.

[12] J. Burke, “Hard to Find Information About Distribution Systems,” 5th ed., ABB Inc.,
Raleigh, NC, 2002. [Online].Available: https://library.e.abb.com/public/
91ad3a29a50978bf85256c550053db0d/Hard.To.Find.6th.pdf

[13] S. H. Horowitz, A. G. Phadke, and J. K. Niemira, "Power System Relaying" 4th ed., John
Wiley & Sons, 2013.

[14] M. M. Eissa, "Evaluation of a new current Directional Protection technique using field
data," in IEEE Transactions on Power Delivery, vol. 20, no. 2, pp. 566-572, April 2005,
doi: 10.1109/TPWRD.2005.844356.

[15] A. K. Pradhan, A. Routray and S. Madhan Gudipalli, "Fault Direction Estimation in Radial
Distribution System Using Phase Change in Sequence Current," in IEEE Transactions on
Power Delivery, vol. 22, no. 4, pp. 2065-2071, Oct. 2007, doi:
10.1109/TPWRD.2007.905340.

[16] A. Ukil, B. Deck and V. H. Shah, "Current-Only Directional Overcurrent Relay," in IEEE
Sensors Journal, vol. 11, no. 6, pp. 1403-1404, June 2011, doi:
10.1109/JSEN.2010.2094186.

[17] T. Hua, X. Yi, W. Chen, Y. Zhuo and K. Arzig, "A novel method of current-only directional
overcurrent protection," 2016 IEEE International Instrumentation and Measurement
Technology Conference Proceedings, Taipei, Taiwan, 2016, pp. 1-5, doi:
10.1109/I2MTC.2016.7520341.

[18] J. Faiz and A. Mahmoudi, "An efficient algorithm for fault detection by directional
overcurrent relay based on current waveform in distribution systems," 2017 International
Conference on Optimization of Electrical and Electronic Equipment (OPTIM) & 2017 Intl
Aegean Conference on Electrical Machines and Power Electronics (ACEMP), Brasov,
Romania, 2017, pp. 43-47, doi: 10.1109/OPTIM.2017.7974945.

[19] P. Mishra, A. K. Pradhan and P. Bajpai, "Time-Domain Directional Relaying Using Only
Fault Current for Distribution System With PV Plant," in IEEE Transactions on Power
Delivery, vol. 37, no. 4, pp. 2867-2874, Aug. 2022, doi: 10.1109/TPWRD.2021.3118712.

[20] I. Kiaei, S. Lotfifard and M. Ghanaatian, "Current-only Directional Overcurrent Protection

Using Postfault Current," 2019 IEEE Texas Power and Energy Conference (TPEC),
College Station, TX, USA, 2019, pp. 1-5, doi: 10.1109/TPEC.2019.8662199.

35
[21] A. G. Phadke and J. S. Thorp, Synchronized Phasor Measurements and Their Applications,
Cham, Switzerland: Springer International Publishing, 2017

[22] Texas Instruments, “Grid Connected Inverter Reference Design,” Texas Instruments,
[Online]. Available: https://www.ti.com/lit/ug/tidub21d/tidub21d.pdf.

[23] S. V. Chavan, R. V. Patil, and S. S. Kamble, “Hardware Implementation of Transmission

Line and Its Study Using MATLAB,” International Journal of Engineering Research &
Technology (IJERT), vol. 3, no. 1, Jan. 2014.

[24] C. Venkatesh and K. S. Swarup, "Investigating performance of numerical distance relay

with higher sampling rate," 2012 North American Power Symposium (NAPS), Champaign,
IL, USA, 2012, pp. 1-6, doi: 10.1109/NAPS.2012.6336316.

36
Appendices

Appendix A - Phasor Estimation Code

%% Fault detection algorithm

%Variables, constants, and sample vectors definition

clear;
clc;

Fs = 3000; %sample rate in Hz

f0 = 60; % fundamental frequency in Hz
Nr = round(Fs/f0); % Samples per one fundamental(f0) cycle in units
count = 1;
aux = 1;

Dsam = readtable(['\Path of record data\']);

Xn = Dsam{750:15.75e3,2}; %it starts at 750 sample (1/4 sec after recording) and
stops 5 sec after ,

%% sampling Angular steps vector

% Vector that contains all the angular steps of one cycle regarding the
% samples per cycle Nr

for n = 0:Nr-1

Mang(n+1)= cos((2*pi/Nr)*(n)) - j*sin((2*pi/Nr)*(n));

end

%% Phasor Estimation
%It will be counting the number of samples of the sampling vector. When the
%division remaining between the count and the samples per cycle by design (Nr) is
zero,
%an auxiliary vector is created containing all the sample values and multiplied by
the angular vector, getting a new phasor estimation.
%which is stored in a Phasor vector.

for b = (Nr):length(Xn)
a = b-Nr+1;

if rem(count,(Nr)) == 0 % denominator controls at each sample the phasor will

be created. Nr means every cycle
Xn_wind = Xn(a:b); % window samples
Std(aux) = std(Xn_wind);
Phas_nc(aux) = (2/Nr)*Mang*Xn_wind; %Every phasor will be moving n*2*pi/Nr
angle. In this case, n is equal to Nr at each phasor estimation.
aux=aux+1;
end
count = count+1;

37
end

%% Eliminating transient
%Transient gives error phasor estimation. Standard deviation comparison
%between two consecutive Nr samples groups identifies transient to be
%removed from the phasor and samples data vectors

er = 0.001;

for br = 2:length(Std)
a=br-1;
if Std(br)-Std(a) > er
Phas_nc = [Phas_nc(1:br-1), Phas_nc(br+1:end)];
Xn_c = [Xn(1:Nr*(br-1))', Xn(Nr*(br+1):end)'];
break
end
end

%% Correcting to zero angle

% To get angles starting at zero, the first angle is considered as
% reference and subtracted to all the other angles of the Phasor vector.

Phas = Phas_nc*(conj(Phas_nc(1))/abs(Phas_nc(1))); % normalizing with the angle

reference, amplitude = 1.
angleref = angle(Phas_nc(1))*180/pi;

%% Fault detection
%%The angular difference between two consecutive phasors is calculated and
%%stored.

err=5;

for b=2:length(Phas)
a=b-1;
diff_ang(a)= (angle(Phas(b))-angle(Phas(a)))*180/pi;

if diff_ang(a) < -err

sprintf('Forward fault at %d phasor',a)
elseif diff_ang(a) > err
sprintf('Reverse fault at %d phasor',a)
end
end

%% Plotting results
a=br-5;
b=a+8;

figure(1)
subplot(2,1,1)
plot(angle(Phas)*180/pi, 'Linewidth', 2)
xlabel('Phasor number')
ylabel('Angle [°]')
xlim([a b])
ylim([-30 130])

38
title('Phasor angles - 0.18 {\itInvIoRef}')
subplot(2,1,2)
plot(Xn_c,'Linewidth', 2)
xlabel('samples')
ylabel('Amplitude [mV]')
xlim([a*50 b*50])
ylim([-0.6 0.6])
title('Pre fault/Post fault - 0.18 {\itInvIoRef} ')

%% Difference and Amplitude ratio

diff = diff_ang(br-1)
rtamp = max(Xn_c(Nr*(br):end)/max(Xn_c(1:Nr*(br-1))))

39