1 KALU OBINNA OBUMA PG/M.ENG/14/68136 APPLICATION OF ARTIFICAIL NEURAL NETWORK FOR ENHANCED POWER SYSTEMS PROTECTION ON THE NIGERIAN 330kV NETWORK DEPARTMENT OF ELECTRICAL ENGINEERING FACULTY OF ENGINEERING Godwin Valentine Digitally Signed by: Content manager’s Name DN : CN = Webmaster’s name O= University of Nigeria, Nsukka OU = Innovation Centre
100
Embed
FACULTY OF ENGINEERING DEPARTMENT OF ELECTRICAL …€¦ · DEPARTMENT OF ELECTRICAL ENGINEERING FACULTY OF ENGINEERING Godwin Valentine Digitally Signed by: ... Chapter 1: INTRODUCTION
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
1
KALU OBINNA OBUMA
PG/M.ENG/14/68136
APPLICATION OF ARTIFICAIL NEURAL NETWORK FOR ENHANCED
POWER SYSTEMS PROTECTION ON THE NIGERIAN 330kV NETWORK
DEPARTMENT OF ELECTRICAL ENGINEERING
FACULTY OF ENGINEERING
Godwin Valentine
Digitally Signed by: Content manager’s Name
DN : CN = Webmaster’s name
O= University of Nigeria, Nsukka
OU = Innovation Centre
2
UNIVERSITY OF NIGERIA, NSUKKA
FACULTY OF ENGINEERING
DEPARTMENT OF ELECTRICAL ENGINEERING
APPLICATION OF ARTIFICIAL NEURAL NETWORK FOR
ENHANCED POWER SYSTEM PROTECTION IN THE NIGERIAN
330kV NETWORK
A THESIS SUBMITTED IN PARTIAL FULFILMENT FOR THE
REQUIREMENT OF THE AWARD OF M.ENG
(POWER SYSTEMS ENGINEERING)
BY
KALU OBINNA OBUMA
PG/M.ENG/14/68136
DECEMBER, 2015
3
Title Page
APPLICATION OF ARTIFICAIL NEURAL NETWORK FOR
ENHANCED POWER SYSTEMS PROTECTION ON THE NIGERIAN
330kV NETWORK
A THESIS SUBMITTED IN PARTIAL FULFILMENT FOR THE
REQUIREMENT OF THE AWARD OF M.ENG
(POWER SYSTEMS ENGINEERING)
BY
KALU OBINNA OBUMA
PG/M.ENG/14/68136
SUPERVISOR: PROF. T.C. MADUEME
DECEMBER 2015
4
APPROVAL PAGE
APPLICATION OF ARTIFICIAL NEURAL NETWORK FOR ENHANCED POWER
SYSTEM PROTECTION IN THE NIGERIAN 330kV NETWORK
BY
KALU OBINNA OBUMA
(PG/M.EMG/14/68136)
A THESIS SUBMITTED IN PARTIAL FULFILMENT OF THE REQUIREMENTS
FOR THE AWARD OF MASTER OF ENGINEERING DEGREE (M.ENG) IN
ELECTRICAL ENGINEERING
DECEMBER, 2015
Kalu Obinna Obuma Signature: Date:
(Student)
Engr. Prof. T.C. Madueme Signature: Date:
Engr. Prof. A.O. Ibe Signature: Date:
(External Supervisor)
Engr. Prof. E.C Ejiogu Signature: Date:
(Head of Department)
Engr. Prof. E.S. Obe Signature: Date:
(Fac. Of Engineering Rep. SPGS)
(Supervisor)
5
Certification
Kalu Obinna Obuma, a master’s degree student in the department of electrical engineering
with registration number, PG/M.ENG/14/68136 has satisfactorily completed the requirements
for the award of the degree of Masters of Engineering (M.Eng) in Electrical Engineering.
The work embodied in this project is original and has not been submitted in part or full for
any other diploma or degree of this or any other university to the best of my knowledge.
Engr. Prof. T.C. Madueme
(Supervisor)
Engr.Prof. E.C Ejiogu
(Head of Department)
Engr. Prof. E.S Obe
(Fac. of Engineering Rep. SPGS)
6
Dedication
I dedicate this work to all students of 2014 post graduate set of the department of Electrical
Engineering Department.
7
Acknowledgement
This work would not have been possible without the contribution of others, truly the saying
‘no man is an island’ adequately applies.
Many thanks to my supervisor, Prof T.C Madueme, whose consistency in lending assistance,
being there when you need him and giving needed assistance where necessary, truly a father.
Prof. Obe, your assistance with procuring much needed research materials is greatly
appreciated. Truly this thesis would not have taken much longer if not for your assistance in
procuring needed materials.
Prof A.U.Ekwue, many thanks to you as well, even from a distance your assistance be it
electronically, your tutelage as my lecturer made understanding this thesis easy. I will not fail
to mention Prof. Anih who answered important questions relating to my work. Dr Ogbuka,
your input with my Matlab Simulation saved me a lot of blunders, many thanks.
To the staff of the Protection, Control and Metering department of the New Haven & Onistha
Transmission Station, I can’t thank you enough for obliging me with needed data to make my
simulation more realistic, the head of department of both Departments for answering my
questions and granting me supervised access to the control room.
To my friends who supported me during my project work I will never forget; Onwaokangba
Anthony, Odoh Benjamin, Uzoeto Ifeanyi, Nwaogu Chijioke; am really grateful. To my
family, always a constant in my life, your prayers and support really kept me going. To my
siblings, thanks for looking up to me, truly responsibility, duty, and discipline. All these traits
I’ve developed thanks to you guys.
I thank almighty God, during the stressful moments trying to figure out my project, my
travels to transmission companies up to the moment it finally made sense. To you I give all
Table 4.3 Performance table for adaptive reclosure scheme network 66
Table 4.4 Performance table for fault locator using neural network 67
Table 4.5 Output of trained faulted phase detector network using current values only for zone 1 68
Table 4.6 Output of trained faulted phase detector network using current & voltage values 69
Table 4.7 Output of trained faulted phase detector network using current values for zone 2 70
Table 4.8 Output of trained faulted phase detector using voltage & current values for zone 2 71
Table 4.9 output of trained fault phase detector using current values for all zones 72
Table 4.10 output of trained faulted phase detector using voltage and current values for all zones 73
Table 4.11 Comparison of estimated and target output for fault locator zone 2 74
Table 4.12 Comparison of estimated and target output of fault locator zone 1 75
Table 4.13 Comparison of estimated and target output for fault locator zone 3 76
Table 4.14 Adaptive fault classifier for transient or permanent fault 78
16
Abstract
This work investigates an improved protection solution based on the use of artificial neural
network on the 330kV Nigerian Network modelled using Matlab R2014a. Measured fault
voltages and currents signals decomposed using the discrete Fourier transform implemented
via fast Fourier transform are fed as inputs to the neural network. The output plots of the
neural network shows its successful application to fault diagnosis (fault detection, fault
classification and fault location). The neural networks application to fault location shows a
mean square error of 3.5331 and regression value of 0.99976 which shows a very close
relationship between the output and target values fed to the neural network. Unlike
conventional protection schemes, the neural network can be adapted to distances which can
cover the entire length of the protected line. Numerical assessment carried out on the neural
network fault locator shows a reduced time of operation of 5.15miliseconds as compared to
the 0.350seconds with the use of ordinary numerical relays. This work also investigates the
adaptive auto reclosure scheme implemented using artificial neural network. The adaptive
reclosure scheme has been adapted for use in the Nigerian Network successfully to
distinguish transient and permanent faults. Simulation results prove that the adaptive
reclosure scheme was able to detect a line-to-ground transient fault and clear this fault in 0.1s
while the line-to-ground permanent fault is cleared after 0.14s. The auto reclosure scheme is
designed using two separate neural networks, one nework to distinguish the faults either as
transient or permanent fault, and using this fault distinguishing network as input to the second
network to classify decision, either as ‘safe to reclose’ represented by logic ‘1’ or ‘do not
reclose’ represented as logic ‘0’. The Fault diagnostic algorithm designed using artificial
neural network (A.N.N.) for the 330kV network was tested on a 132kV network. Results
show and prove that the algorithm is flexible and can be adopted to other networks.
17
CHAPTER ONE
INTRODUCTION
1.1 Background of the study
The demand for constant power supply in Nigeria is ever increasing; however the demand is
met with lots of constraint. One of them being system faults. Faults on transmission line in
particular is of great interest to the power holding company of Nigeria as more investment is
put into restructuring the current infrastructure and also expanding existing ones.
The power sector of Nigeria is subdivided into policy, regulations, customers, operations. The
operations division brings to light the activities of the transmission company of Nigeria that
controls the high voltage delivery of power from generating plants to the substations for
transmission to distribution stations. T.C.N handles a 330kv system capacity of 6870MW
over a total distance of 5650Km[1], their focus is to maintain power system stability,
reliability and sustainability.
The major protection schemes currently employed are distance protection, over current
protection, differential protection e.t.c. distance protection being the predominant suffers
from inaccuracy due to restraints of relays on protection schemes i.e. reach settings. The relay
cannot fully adapt to fluctuations in power system conditions especially in parallel lines as
well as distinguish between transient and permanent fault following a short circuit.
This work brings to view the application of artificial neural network for enhanced power
system protection in regards to fault detection, fault location, and application of the adaptive
auto reclosure schemes as opposed to conventional approach; travelling wave approach[2][3],
synchronous compensators[4] to name a few.
1.2 Statement of the Problem
Among several power system components, transmission line is one of the most important
components of the power system network and is mostly affected by several types of faults.
Generally, 80%-90% of the fault occurs on the transmission line and the rest of substation
equipment and bus bar combined[5]. The necessary requirement of all the power system is to
maintain reliability of operation which may be done by detecting, classifying and isolating
various faults occurring in the system. It is required that a corrective decision should be made
by the protective device to minimize the period of trouble and limit outage time, damage and
related problems. If any fault or disturbances occurred in the transmission is not detected,
located, and eliminated quickly, it may cause instability in the power system and causes
significant changes in system quantities like over-current, under or over voltage, power
factor, impedance, frequency and power. The appropriate percentage of occurrence of single
line to ground fault is about 70-80%, line to line to ground faults is 10-17%, line to line fault
is 8-10% and three phase is 3%[6]. The three faults occur rarely but if it exists in a system it
is quite expensive.
18
1.3 Significance of the Study
Distance protection is considered covering various effects like high fault impedance, non-
linear arc resistance and variable source impedance. Distance relaying principle, due to their
high speed fault clearance compared with over current relays is widely used protective
scheme for high voltage transmission lines in Nigeria. A distance relay estimates the
electrical distance to the fault and compares the result with a given threshold, which
determines the protection zone. There is need for measuring algorithms that have the ability
to adapt dynamically to the system operating conditions such as changes in the configuration.
Numerical relays acquire sequential samples of A.C. quantities in numeric (digital) data form
through the data acquisition system, process the data using the algorithm to calculate fault
discriminate and make trip decision. The reach accurateness of an electromechanical, static or
a microprocessor based distance relay is affected by different fault conditions and network
configuration settings. Artificial neural network makes use of samples of currents and
voltages directly as inputs without calculation of phasor and related symmetrical components.
The algorithm makes available automatic determination of fault direction and fault location
after one cycle from the initiation of fault. For protection of transmission line using artificial
neural network, it doesn’t necessitate any communication link to recover remote end data
from local end only i.e. voltages and currents are captured from the bus bar. Then, pre
processing of obtained signal can be done to pass it into A.N.N level making it the best tool
to solve under reach and overstretch problems which are very regular with conventional
distance relay design.
1.4 Aim/Objective of the study
The aim of this project is to demonstrate the application of artificial neural network to fault
diagnostic as well as implementation of adaptive single pole auto reclosure scheme in power
system protection. This work presents the outcomes of both the feed forward A.N.N and self-
organized neural network application to;
Fault detection of all types of faults
Fault classification (line to ground, line to line, line to line to ground, three phase
faults, three phase to ground faults)
Fault location on three zones presented in this work
Application to adaptive single pole auto-reclosure scheme
The objectives of the study are;
Fault pattern generation from transmission network modelled on Matlab Simulink
Environment.
Pre-processing of voltage and current signals using Fast Fourier transform.
Normalization of the extracted features in order to match A.N.N input level of ± 1.
19
Selection of appropriate neural network architecture for various protection problems;
fault detection, location e.t.c.
Training of appropriate neural network.
1.5 Scope of the Work
This work centres on the application of artificial neural networks on the Nigerian 330kv
network. For the purpose of this project, three transmission networks are considered; Onistha,
Benin and new-haven. Needed data’s like single line diagram of each network as well as line
and bus data of this networks are collated for the purpose of this work. The author did his
best to create a Simulink model of these networks taking Onistha network as the reference to
other networks represented as one entity. The fault breaker block is placed on each line
representing the three different protection zones to induce different fault types on each line.
Certain assumptions were made in the modelling of these networks; the generator as well as
step-up transformer data of a different generating plant but giving the same output voltage as
desired for this work, fault data also retrieved from Onistha T/S was used to compare
simulated to real time data
20
CHAPTER TWO
LITERATURE REVIEW
2.1 State of the Art Power System Protection
Power system protection is a branch of electrical power engineering that deals with the
protection of power systems from faults through the isolation of faulted parts from the rest of
the network. Protection systems in electricity delivery networks have a major role to play in
increasing of systems, and a broad understanding of their current and future application can
aid in better taking them into account for achieving future energy networks that adapt for the
incorporation of renewable energy generation sources. This chapter provides a survey of
faults generally, state of art of some protection techniques as well as protection schemes. The
unifying theme of this work is to highlight the potentials of artificial intelligence namely
artificial neural networks in overcoming the restraints of traditional protection techniques[7]
thus the enhanced protection scheme is introduced .
2.2 Faults in Power System
Fault is an unwanted short circuit condition that occurs either between two phases of wires or
between a phase of wire and ground. Short circuit is the most risky type of fault as flow of
heavy currents can cause overheating or create mechanical forces which may damage
equipment and other elements of power system. Faults can be classified into three types,
which are symmetrical faults, unsymmetrical faults, and open faults.
2.3 Symmetrical Faults
The fault that results in symmetrical fault current (equal currents with 120 displacement) is
known as symmetrical fault. Three phase faults is an example of symmetrical fault where all
three phases are short circuited with or without involving the ground.
2.3.1 Transient on a Transmission Line
To consider the short circuit transient on a transmission line, certain simplifying assumptions
made at this stage
The line is fed from a constant voltage source
Short circuit takes place when the line is unloaded
Line capacitance is negligible and the line is represented by a lumped RL series
circuit
With the above assumptions the line can be represented by the circuit model of fig 1 below.
The short circuit is assumed to take place at t=0. The parameter α controls the instant on the
voltage wave when the short circuit occurs. It is known from the circuit theory that the
current after short circuit is comprised of two parts i.e.
21
i = i + i (2.1)
where i = steady state current = √( )|| sin(ωt + α − ɵ)
Z = (R + ωL) < ɵ = tan! ω"# $ (2.2)
Where i = transient current'it is such that i(0) = i(0) + i (0) = 0) Being an inductive circuit, it decays correspondingly to the time constant L R . i = −i(0)e!*+$ = √|| sin(θ − α) e,*+$- (2.3)
Fig 4 shows a line to line fault at F in a power system on phases ‘b’ and ‘c’ through fault
impedance XY .
The currents and voltages at the fault can be expressed as
N7
a
b
c
F
XY N9 N:
Fig 2.4: Line to Line (L-L) fault through impedance XY
27
N_ = K N7 = 0N9N: = −N9L ; 69 − 6: = N9XY (2.30)
The symmetric components of the faults currents are
KNONONOAL = 13 K1 ∝ ∝1 ∝ ∝1 1 1 L K 0N9−N9
L
From which we get
NO = −NO (2.31)
NOA = 0 (2.32)
The symmetrical components of voltage at F under fault are
K6O6O6OAL = E K1 ∝ ∝1 ∝ ∝1 1 1 L a 676969 − XYN9
b (2.33)
The first two equations
36O = 67 + (∝ +∝)69 −∝ XYN9
36O = 67 + (∝ +∝)69 −∝ XYN9
From which we get
3(6O − 6O) = (∝ −∝)XYN9 = c√3 XYN9 (2.34)
Now,
N9 = (∝ ∝)NO (NO = NO; NOA = 0)
= d√3NO (2.35)
Substitute N9 from (3.31) and (3.35) parallel connection of positive and negative sequence
networks through a series impedance XY as shown in fig 5 since NOA = 0, the zero sequence
network is unconnected
28
In terms of the thevenin equivalents, we get
NO = \7X + X + XY
From (2.35) we get
N9 = N: = c√3 \7X + X + XY
2.4.3 DOUBLE LINE TO GROUND (LLG) FAULT
Fig 6 shows a double line to ground fault at F in a power system. The fault may be in general
having impedance XY as shown.
XY
NO NO
6O 6O
Fig 2. 5 sequence network for L-L fault
a
b
c
F
XY N9 N: 3NOA
NOA = 0
Fig 2.6: double line to ground (LLG) fault through impedance XY
29
The current and voltage (to ground) conditions at the fault are expressed as
N7 = 0
NO + NO + NOA = 0 (2.36)
69 = 6: = XY(N9 + N:) = 3XYNOA (2.37)
The symmetrical components of voltages are given by
K6O6O6OAL =
E K1 ∝ ∝1 ∝ ∝1 1 1 L K67696:
L (2.38)
From which it follows that
6O = 6O = E '67 + (∝ +∝)69) (2.39a)
6OA = E (67 + 269) (2.39b)
From (2.39a) and (2.39b)
6OA 6O = 13 (2∝ ∝)69 = 69 = 3XYNOA
Or
6OA = 6O + 3XYNOA
The sequence connections is shown in fig 2.7
3XY
F F F 6O 6O
6OA
Fig2. 7 Connection of sequence networks for a double line to ground (LLG) fault
30
In terms of the thevenin equivalents, the new equation translates from
NO = \7X + X (XA + 3XY)
= PQRSTRU.RVTERW/.RUTRVTERW/ (2.40)
2.5 Types of Protection
Protection of transmission or distribution network serves to protect the plant as well as the
personnel by disconnecting equipment which experiences an overload or a short to the earth.
Some forms of protection are;
Overload and backup for distance (over-current): overload protection requires a
current transformer which simply measures the current in a circuit. There are two
types of overload protection; instantaneous over current and time over current
(T.O.C). Instantaneous over current requires that the current exceeds a pre-determined
level for the circuit breakers to operate.
Earth-fault: earth fault protection again requires current transformers and series an
imbalance in a three-phase circuit. Normally the three phase currents are in balance,
which is roughly in magnitude. If one or two phases become connected to earth via
long impedance fault, their magnitudes will increase dramatically and cause
imbalance. If this imbalance exceeds a pre-determined value, a circuit breaker should
operate.
Distance (impedance relay): distance protection detects both the voltage and current.
A fault on a circuit will generally create a sag in the voltage level. If the ratio of
voltage to current measured at the relay terminals, which equates to impedance, leads
within a pre-determined level the circuit breaker will operate. This is useful for
reasonable length lines, lines longer than 10 miles because its operating characteristics
are based on the line characteristics. This means that when a fault appears on the line
the impedance setting relay is compared to the apparent impedance of the line from
the relay terminals to the fault. If the relay setting is determined to be below the
apparent impedance it is determined that the fault is within the zone of protection[8].
Back up: the objective of protection is to remove only the affected position of a plant
and nothing else. A circuit breaker or protection relay may fail to operate. In
important systems, a failure of primary protection will usually result in the operation
of back-up protection[9]. Remote back up protection will generally remove the
affected and unaffected items of plant to clear the fault. Local back-up protection will
remove the affected items of plant to clear the fault. Local back-up protection will
remove the affected items of the plant to clear the fault.
31
2.5.1 Distance Relays
The distance protection scheme is the dominant scheme used in the Nigerian 330Kv networks
thus a further review of this scheme and its implication as regards this work is paramount.
The distance protection is implemented in a transmission network by the protection
equipment known as distance relays. Distance relays respond to the voltage and current i.e.
impedance at the relay location. The impedance per mile is fairly constant so these relays
respond to the distance between the relay location and a fault location. As the power systems
becomes more complex and the fault current varies with changes in generation and system
configuration, directional over current relays are more difficult to apply and to set for all
contingencies, whereas the distance relay setting is constant for a wide variety of changes
external to the protection line. There are three general types; impedance relay, admittance
relay, reactance relay each is distinguished by its application and its operating characteristics.
In a three phase power system, 11 types of fault are possible; three single phase to ground,
three phase –phase to ground, three double phase to ground, and two three phase faults. It is
essential that the relays provided have the same setting regardless of the type of fault. This is
possible if the relays are connected to respond to delta voltages and currents. The delta
quantities are defined as the difference between any two phase currents, for example, \7 \9 is the delta quantity between phases ‘a & b’. In general, for multiphase-fault between
phases x and y,
PQ!PefQ!fe = X (2.41)
Where X is the positive sequence impedance between the relay location and the fault. For
ground distance relays, the faulted phase voltage, and a compensated faulted phase must be
used.
PQfQTgfV = X (2.42)
Where m is a constant depending on the line impedance, and NA is zero sequence current in
the transmission line. A full complement of relays consisting of three phase distance relays
and three ground relays. This is the preferred protective scheme for high and extra high
voltage systems[6].
32
2.5.2 Pilot Protection
Step distance protection does not offer instantaneous clearing of faults over 100% of line
segment. To cover the 10-20% of the line not covered by zone 1, the information regarding
the location of the fault is transmitted from each terminal to the other terminals. A
communication channel is used for this transmission. Pilot channels can be over power line
carrier, microwave, fibre optic, or wire pilot. Power line carrier uses the protected line itself
as a channel, superimposing a high frequency signal on-top of the 50Hz power frequency.
Since the line being protected is also the medium used to actuate the protective devices, a
blocking signal is used. This means a trip will occur at both ends of the line unless a signal is
received from the remote end. Pilot protection is not in use in the south-eastern Nigerian
transmission network due to some stations yet to be connected to the grid.The issues
associated with the distance relay, problems of under reach and over reach introduce a high
error in distance relaying. In the case of pilot protection, cost of implementing
communication channel presents a setback to its use regardless of its efficiency. The
constraints presented, informs the decision to research on a cost effective and robust
protection scheme; many research on improved protection, enhanced protection can be
characterized as ADAPTIVE PROTECTION; Adaptive protection is a protection philosophy
Fig 2.8 Three zone step distance relaying to protect 100% of a line and backup neighbouring line. (From
S.Horowitz,Transmission Line Protection, 2nd ed..,2007.CRC Press, Taylor % Francis Group)
33
which permits and seeks to make adjustments to various protection functions automatically in
order to make them more attuned to prevailing power system conditions. The ADAPTIVE
DISTANCE PROTECTION[10] is one of such research areas in power system protection;
this scheme seeks to keep the protected zone constant at a predetermined boundary by
adapting the tripping impedance under varying power system conditions. Another research
area is BOUNDARY PROTECTION[11][12], which happens to be a step beyond the
adaptive distance protection. A fast growing research interest is the single pole auto reclosure
technique further developed to the ADAPTIVE SINGLE POLE AUTO RECLOSURE
SCHEME[13][14][15].
2.6 Single Pole Auto Reclosure Technique
The most common faults on EHV transmission lines are single phase to ground types and for
such faults SPAR provides an improvement in the overall protection of transmission
system[13]. S.P.A.R is imperative in applications construction of additional circuits may not
be possible due to environmental pressure/costs, a practical example is the Nigerian
Transmission Network which does not utilize the scheme in the protection of its transmission
lines. The conventional S.P.A.R has cases of unsuccessful reclosures due to a fixed dead time
in the case of transient fault, or reclosure onto a permanent fault may aggravate the potential
damage to the system and equipment. Notwithstanding, the method of auto reclosure is
economical and effective technique for high capacity electric power systems to improve
reliability and stability if auto reclosure is successfully executed, it usually restores the
stability of the system and maintains the continuity of electric power transmission. In auto
reclosure techniques, it is very important to distinguish permanent fault from temporary faults
and to apply an adaptive algorithm in each case. In this respect, adaptive S.P.A.R offers many
advantages such as increased rate of successful reclosure, improved system stability and
reduction in system and equipment shock under a permanent fault.
2.6.1 Auto Reclosure Relaying System
A research work on auto reclosure scheme[16] proposes the utilization of adjustable dead
times by accurately identifying arc extinction times. It is inferred that if the dead time is too
short, is possible to reignite the arc, which leads to re-striking arc faults, so we must ensure a
long enough dead time to ensure improvement in power system stability and reduction in
system shock can be achieved easily by applying adaptive reclosing. In this study, the
adaptive S.PA.R is implemented using artificial neural network.
2.7 System Configuration
Although the fundamentals of transmission line protection apply in almost all system
configurations, there are different applications that are more or less dependent upon specific
situations.
Operational Voltages- transmission lines will be those lines operating at 138kV and
above, sub transmission lines are 34.5kV to 138kV, and distribution lines are below
34
34.5kV. These are not rigid definitions are only used to generically identify a
transmission system and connote the type of protection usually provided. The higher
voltage systems would normally be expected to have more complex, hence more
expensive, relay systems. This is so because higher voltages have more expensive
equipment associated with them and one would expect that this voltage class is more
important to the security of the power system. The higher relay costs, therefore, are
more easily justified.
Multi-Terminal Lines - occasionally, transmission lines may be tapped to provide
intermediate connections to additional sources without the expense of a circuit
breaker or other switching device. Such a configuration is known as a multi-terminal
line and, although it is an inexpensive, measure for strengthening the power system, it
presents special problems for the protection engineer. The difficulty arises from the
fact that a relay receives its input from the local transducers, i.e., the current and
voltage at the relay location. The total fault current is the sum of the local current plus
the contribution from the intermediate source, and voltage at the relay location is the
sum of the two voltage drops, one of which is the product of the unmonitored current
and the associated line impedance.
Line Length- the length of a line has a direct effect on the type of protection, the
relays applied, and the settings. It is helpful to categorize the line length as “short,”
“medium,” or “long” as this helps establish the general relaying applications although
the definition of “short,” “medium,” and “long” is not precise. A short line is one in
which the ratio of the source to the line impedance of a line varies more with the
nominal voltage of the line than with its physical length or impedance. So a “short”
line at one voltage level may be a “medium” or “long” line at another.
2.8 Transmission line protection
The study of transmission line protection presents many fundamental relaying considerations
that apply, in one degree or another, to the protection of other types of power system
protection. Each electrical element of course will have problems unique to itself, but the
concepts of reliability, selectivity, local and remote backup, zones of protection, coordination
and speed which may be present in the protection of one or more other electrical apparatus
are all present in the considerations surrounding transmission line protection.
Since transmission lines are also the links to adjacent lines or connected equipment,
transmission line protection must be compatible with the protection of all of these other
elements. This requires coordination of settings, operating times and characteristics[17]. The
purpose of power system protection is to detect faults or abnormal operating conditions and
to initiate corrective action. Relays must be able to evaluate a wide variety of parameters to
establish that corrective action is required. Obviously, a relay cannot prevent the fault. Its
primary purpose is to detect the fault and take the necessary action to minimize the damage to
the equipment or to the system. The most common parameters which reflect the presence of a
35
fault are the voltages and currents at the terminals of the protected apparatus or at the
appropriate zone boundaries. The fundamental problem in power system protection is to
define the quantities that can differentiate between normal and abnormal conditions. In this
study, Transmission line protection is carried out in the following categories; fault detection,
fault classification, fault location and adaptive auto reclosure technique
2.8.1 Fault Detection & Location
When a fault occurs on a transmission line, it is very important to quickly detect and locate it
in order to make necessary repairs and to restore power as soon as possible, the time needed
to determine the fault point along the line will affect the quality of power delivery. Fault
location has been a subject of interest for many years. Many fault locating algorithms have
been developed; the power frequency based approach[18], transient signals based
approach[19] and super imposed component based approach[20]. Currently the most widely
used method of overhead line fault location is to determine apparent reactance of the line
during the time the fault is flowing and to convert the Ohmic result into distance based on the
parameter of the line, however this method is subject to errors when the fault resistance is
varied and the line is fed from both ends.
Many successful applications of artificial neural networks to power systems have
demonstrated the use of artificial neural networks for direction estimation[21], faulted phase
selection[8], fault location under CT saturation. However these applications merely use the
A.N.N ability of classification i.e. the ANNs output of 1 or 0 and mainly work on singly fed
source without consideration to different zones of protection. In this study, three fault
detectors carried out on all zones of protection as well as fault locators are implemented using
A.N.N.
2.8.2 Fault Classification
Overhead transmission lines are vulnerable to faults since they extend over long distances
and are often exposed to severe climate conditions. Symmetrical faults and unsymmetrical
faults can easily be classified into transient and permanent faults. The fault classification
problem is mostly treated as a pattern recognition problem[22][23] which is implemented
using the back propagation learning algorithm, the usual case in many research works,
however this algorithm has generalization and convergence problems associated with it, an
algorithm with convergence issues has a high percentage error thus low effectiveness[24].
The paper[3] also utilizes feed forward ANN to fault classification current signals only as
input signals due to high economical cost of such devices. The use of only current signals
gives room for errors in fault detection and location for faulted phases to ground takes into
account zero sequence and voltage of faulted phases thus in this study, a different approach
that employs both voltage and current signals as well as zero sequence of both quantities is
used as input to the neural network designed. The approach implemented in most research
work treats fault classification as a pattern recognition problem carried out using the feed
forward back propagation algorithm[3][25]; however this algorithm is beset with
36
generalization problems if the input data set is not large enough which cause intolerable
percentage error. A different approach is treated in this work, the use of unsupervised
competitive layer self-organised map algorithm implemented as a clustering problem. Results
shown in chapter four proves the problems inherent in the previous algorithm are taken care
of with this approach.
2.8.3 Enhanced Power System Protection
The traditional line protection scheme based on fundamental frequency components of the
fault generated transient voltage and current signals can be classified into two categories;
non-unit protection and unit protection. The non-unit protection schemes use one end
transmission line data whilst the unit protection schemes use data from the two ends. The
non-unit protection such as distance relay cannot protect the entire length of the primary line
because it cannot differentiate the internal faults from external occurring around multi zone
boundaries. Back up protection may be introduced as a trade-off for protecting the entire
length of the transmission line. For unit protection such as pilot protection, it usually requires
a communication link to transmit the blocking or transfer tripping signals therefore the
reliability of the protection scheme highly relies on the reliability of the communication
link[5][6][11]. The cost of communication link also needs to be taken into account. Recently,
new techniques using high frequency components of the faulted generated signals were
studied and some useful solutions were obtained[11][12][15]. An approach known as
“adaptive single pole auto reclosure scheme” for solving the disadvantages of conventional
non-unit protection scheme was proposed. This approach introduces the possibility of
differentiating the permanent and transient fault using data from one end only; other proposed
solutions are boundary protection and adaptive distance protection[10].
Regarding the fault selection or classification, the traditional method is based on the
fundamental frequency phasors. The feature formed by a non-linear ratio between voltage
and current phasors is compared to the threshold to find out the faulted phase. This kind of
method is affected by different conditions such as fault resistance, mutual coupling of parallel
lines e.t.c. This study proves an alternative solution in the use of neural network based
algorithm based on fast Fourier transform and self-organized neural network and back
propagation neural network to realise fault classification as well as adaptive reclosure
scheme. A similar research[12][26] work utilized same self-organized neural network for
fault classification as well as adaptive reclosure for fault classification and boundary
protection. Although different, this study proves that both algorithms can be combined to
solve the generalization problems associated with fault classification problems associated
with only back propagation algorithm. To further illustrate the functionality of neural
network and its various algorithms, a summary of neural network description, back
propagation algorithm and self-organised neural network is presented in the next section.
37
2.9 Artificial Neural Network
A neural network is a massively parallel distributed processor made up of simple processing
unit that has a natural propensity for storing experimental knowledge and making it available
for use. Artificial neural network is inspired by biological neural network and is composed of
a number of interconnected units known as artificial neurons. Artificial neurons are used to
transmit signal from one layer to the other, its complex network of interconnected neurons is
analogous to firing of electrical pulses via its connections that leads to information
propagation. A.N.N. consists of three layers i.e. input layer, hidden layer and output layer
having number of neurons present in it[8].
Neural networks are primarily of three basic learning algorithms such as supervised learning,
unsupervised and reinforced learning. For the sake of this work only supervised and
unsupervised training algorithm is utilized. The supervised learning algorithm is the popular
error back propagation for diagnosis of faults in power systems. However due to slow
training speeds and generalization issues, the unsupervised was also adapted in this work. A
review of the multilayer perceptron, error back propagation as well as unsupervised training
algorithm is carried out for the purpose of better understanding of its functionality as applied
to this research.
2.9.1 Multilayer Perceptron
The cascaded layer perceptron is an example notation of the multilayer perceptron as
illustrated in fig 9 below. The output of the first network is the input to the second network,
and the output of the second network is the input to the third network. Each layer may have a
different number of neurons, and even a different transfer function. The weight matrix for
the first layer is written as h and the weight matrix for the second layer is h. to identify
the structure of a multilayer network, the following shorthand notations, where the number of
inputs is followed by the number of neurons in each layer[27].