Volume 2, Issue 7, July 2012
ISSN: 2277 128X
International Journal of Advanced Research in
Computer Science and Software Engineering
Research Paper
Available online at: www.ijarcsse.com
Protection of Transmission Lines Using Artificial
Neural Network
Isha Awasthi1, Aziz Ahmed2
M.Tech Scholar, Department of EEE,
2
HOD, Department of EEE,
AFSET ,Dhauj , Faridabad,
Affiliated to M.D.U., Rohtak, Haryana
1
Abstract-High voltage transmission lines are used to transmit electrical energy from source to substations. If any fault and
disturbance are generated in the transmission lines and not detected, located and eliminated quickly, it may cause
instability in the system. To find the exact location of fault occurring in transmission lines may be calculated by some
conventional methods. This paper presents the new approach to fault detection and clarification in power transmission
system by using ANN (Artificial Neural Network). The network is trained with Rosenblatt’s algorithm. The results have
been calculated by MATLAB. Several graphs have been shown in this paper. From the conventional methods, only one
fault may be detected between two substations, but in this paper there are several nodes and the distance between each
node has been calculated. The fault is generated in between these nodes.
Keywords- ANN (Artificial neural network), Rosenblatt’s algorithm, High voltage transmission lines, Fault, MATLAB
Introduction
Work on artificial neural networks commonly referred to
as “neural networks”, has been motivated right from its
inception by the recognition that the human brain
computers in an entirely different way from the
conventional digital computers. The brain is a highly
complex, nonlinear and parallel computer. It has the
capability to organize its structural constituents known as
neurons, so as to perform certain computations (e.g.
pattern recognition, perception and motor control) many
times faster than the fastest digital computer in existence
today.[1]
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. It resembles
the brain in two respects:
 Knowledge is acquired by the network from its
environment through the learning process.
 Inter-neuron connection strengths known as
synaptic weights are used to store the acquired
knowledge
(i) Artificial intelligence:
Designing of intelligence computer system from
characteristic associated with intelligence in human
behavior [2]
Example: - 1. Neural Network 2. Fuzzy Logic 3. Expert
system 4. Probabilistic reasoning.
Types: 1. Hard computing 2. Soft computing
© 2012, IJARCSSE All Rights Reserved
Characteristics: - 1. Cognition 2. Logical Interface 3.
Pattern Recognition
Human brain has two properties: Human brain
is getting experienced to adapt themselves to their
surrounding environments. So as a result the information
processing capability of the brain is rendered, when this
happen the brain becomes plastic.
1. Plastic: Capability to process information,
capability of adding. Must preserve the information it has
learn previously.
2. Stable: Remain stable when it is presented with
irrelevant information, useless information.
*. Synapses with large area are excitatory (+Ve weights)
& with small area are inhibitory (-Ve weights).
*. Synapses of the neuron are modulated as weights.
(Strength of the connection)
*. Biological neuron receive all inputs through dendrites,
sum them & produces an output. If the sum is greater then
the threshold value, then input signals are passed to the
cell body.
*. NN can mapped input patterns into output patterns.
*. NN’s are robust systems. They can recall full patterns
from incomplete pattern or noise channel.
*. Mc Culloh Pits neuron was centered on the idea that a
neuron will fire an impulse only if the threshold value is
exceed.
Total input (I) receive by the soma
Page | 70
Volume 2, Issue 7, July 2012
www.ijarcsse.com
𝑛
𝐼=
𝑤𝑖 . 𝑥𝑖
𝑗 =1
To generate the final output Y sum is passed through a
nonlinear filter called activation filter.
(ii) Activation function:
An activation function for limiting the amplitude of
the output of a neuron. The activation function
referred as squashing function in that it squashes the
permissible amplitude range of the output signal to
some finite value.
cognitive process. Training algo. of perceptron is
supervised learning.
It accept no. of inputs Xi (i=1,2,3…………..n) &
compute a weighted sum of these inputs. The sum is then
compared with a threshold θ of an output Y (which is
either 0 or 1)
y=1
if
y=0
if
𝑛
𝑖=1 𝑤𝑖 . 𝑥𝑖
≥ θ
𝑛
𝑖=1 𝑤𝑖 . 𝑥𝑖
≤ θ
The perceptron is the single transmission network consist
of sensor unit, association unit & output of response unit.
The sensor unit produce a binary output 0 or 1.
Association unit behave like a basic building block.
Response unit comprise pattern recognition. Thus the O/P
of the response unit could be such that is the weighted
sum of the input is less the nor equal to 0, then the output
is 0 else 1.
An externally applied bias bk has the effect of
increasing or lowering the net input of the activation
function depending upon whether it is positive or
negative respectively.
𝑚
𝑢𝑘 =
𝑤𝑘𝑗 . 𝑥𝑗
𝑗 =1
&
𝑦𝑘 = Ø. 𝑢𝑘 + 𝑏𝑘
where 𝑥1 , 𝑥2 , … … … 𝑥𝑚 are the input signals.
𝑤𝑘 1 , 𝑤𝑘 2 , … … … 𝑤𝑘 𝑚 are the respective synaptic
weights of neuron.𝑘, 𝑣𝑘 is the linear combiner output
due to the input signals, 𝑏𝑘 is the bias ∅(. ) is the
activation function and 𝑦𝑘 is the output signal of the
neuron[1]
𝑣𝑘 = 𝑢𝑘 + 𝑏𝑘
(v) DONALD HEBB’S LEARNING: - “When the axon
of cell A is near enough to excite the cell B & repeatedly
takes part of firing cell B. Some growth process takes
place in one or both cell & cell A efficiency in firing cell
B increases”.
“When neuron A repeatedly takes part of firing neuron B.
The synapses connection between them increases”.
Hebb’s learning rule can be used as an association
between 2 set of patterns.
Ex: (Pavlov’s ex.)
Suppose:- F: Sight of food (Unconditional stimulus)
B: Sound of bell (Conditional stimulus)
S: Salivation
(Response)
F→
S
→
(F ∩ B) → S
→
(B → S)
MATLAB TEST RESULTS
(i)
Identity Function
(iii) Recurrent Networks: - There could be neurons with
self feedback links.
Pattern Associator: -Here training is supervised. A set of
pairs of patterns are available. Ex: (Xi, Yi), i = 1, 2,
3……….n. NN learn to establish a relation between the
two sets of patterns. Relation (association) between the
two sets is stored in the network.
Auto Associator: Here training is unsupervised.
One set of pattern (Ex: (Xi), i = 1, 2, 3……….n.) is
repeatedly presented to the network. The NN learn &
remember these & then they are stored in the network.
(iv) Rosenblatt’s perceptron: Perceptron is the generic name given by the Frank
Rosenblatt because it is a model of EYE. Perceptron was
an attempt to understand human memory, learning &
© 2012, IJARCSSE All Rights Reserved
Page | 71
Volume 2, Issue 7, July 2012
(ii) Hyperbolic Tangent Function
www.ijarcsse.com
Distances between nodes are shown:
(iii) Logistic Function
(iv) Implementation of ANN :In this paper we consider nodes as the substations, the
several substations are connected from the transmission
lines.The fault occurred in transmission lines.In first table
the distance between two nodes is calculated.
Learning rate can not be 0 or negative
Enter the value of first node:1
Enter the value of second node:2
Table- 1
Nodes
Weights(Distance
between two nodes)
(4,1)
3.3621
(6,2)
3.5501
(2,3)
3.2068
(5,3)
2.4185
(6,3)
2.8447
(3,4)
3.4913
(5,4)
3.2214
(1,5)
3.0154
(2,5)
3.0860
(1,6)
2.8311
(4,6)
2.8565
© 2012, IJARCSSE All Rights Reserved
Now the fault is occurred in between nodes and the exact
location of fault would be calculated by Matlab.
Location of fault between nodes 2 & 3 by Matlab:
Learning rate can not be 0 or negative
Enter the value of first node:2
Enter the value of second node:3
weight = Columns 1 through 9
3.4147 3.5058 2.7270 3.5134 3.2324 2.6975
2.6785 2.9469 3.3575
Columns 10 through 11
3.3649 2.5576
value of shortest distance between specified nodes is
=2.73
Best path choosen is through the =2
Best path choosen is through the =3
Fault is generated at a alocation =3.48
Fault is located at = 3.479,distance from the node 2
Fault is located at = -0.752, distance from the node 3
Time taken by the routing:
Elapsed time is 8.688825 seconds.
Location of fault between nodes 4 & 5 by Matlab:
Learning rate can not be 0 or negative
Enter the value of first node:4
Enter the value of second node:5
weight = Columns 1 through 9
3.4147 3.5058 2.7270 3.5134 3.2324 2.6975
2.6785 2.9469 3.3575
Columns 10 through 11
3.3649 2.5576
value of shortest distance between specified nodes is
=5.26
Best path choosen is through the =4
Best path choosen is through the =1
Fault is generated at a alocation =5.48
Fault is located at = 5.479,distance from the node 4
Fault is located at = -0.223, distance from the node 5
Time taken by the routing:
Elapsed time is 24.250947 seconds.
Location of fault between nodes 5 & 6 by Matlab:
Learning rate can not be 0 or negative
Page | 72
Volume 2, Issue 7, July 2012
Enter the value of first node:5
Enter the value of second node:6
weight = Columns 1 through 9
3.3431 2.9922 3.2555 2.7712 3.3060 2.6318
2.6769 2.4462 2.4971
Columns 10 through 11
3.2235 3.0948
value of shortest distance between specified nodes is
=5.17
Best path choosen is through the =5
Best path choosen is through the =4
Best path choosen is through the =6
Fault is generated at a alocation =1.48
Fault is located at = 1.475,distance from the node 5
Fault is located at = 3.699, distance from the node 6
Time taken by the routing:
Elapsed time is 62.345801 seconds.
Location of fault between nodes 1 & 6 by Matlab:
Learning rate can not be 0 or negative
Enter the value of first node:5
Enter the value of second node:6
weight = Columns 1 through 9
2.6344 3.0387 2.9816 3.3655 3.3952 2.7869
2.8898 2.8456 3.0463
Columns 10 through 11
3.1094 3.1547
value of shortest distance between specified nodes is
=3.04
Best path choosen is through the =1
Best path choosen is through the =6
Fault is generated at a alocation =1.34
Fault is located at = 1.340,distance from the node 1
Fault is located at = 1.699, distance from the node 6
Time taken by the routing:
Elapsed time is 4.108720 seconds.
From these calculation the exact location of fault in
transmission lines can be find out
www.ijarcsse.com
damage to equipment and personnel.This saves cost of
replacement of the damaged parts or equipment, save time
and labour in tracing fault location by maintenance crew.
The paper therefore provides an economic saving to
power utility operatives.
REFERENCES
[1] “Neural Network” Tata Mc Graw Hill.
[2] “A Novel Fuzzy Neural Network Based Distance
Relaying Scheme” Ieee Transactions on Power
delivery
[3] M. Sanaye-Pasand and O. P. Malik, “High speed
transmission system directional protection using an
elman network,” IEEE Trans. on Power Delivery,
vol. 13, no. 4, pp. 1040–1045, 1998.
[4] C. T. Lin and C. S. G. Lee, “Neural-network-based
fuzzy logic control and design system,” IEEE Trans.
on Computers, vol. 40, no. 12, pp. 1321–1336, 1991.
[5] S. Horikawa, T. Furuhashi, and Y. Uchikawa, “On
fuzzy modeling using fuzzy neural networks with the
back-propagation algorithm,” IEEE Trans. on Neural
Networks, vol. 3, no. 5, pp. 801–806, 1992.
[6] Y. Lin and G. A. Cunningham, “A new approach to
fuzzy-neural system modeling,” IEEE Trans. on
Fuzzy Systems, vol. 3, no. 2, pp. 190–197, 1995.
[7] T. S. Sidhu, H. Singh, and M. S. Sachdev, “Design,
implementation and testing of an artificial neural
network based fault direction discrimination for
protecting transmission lines,” IEEE Trans. on Power
Delivery, vol. 10, no. 2, pp. 697–706, 1995.
[8] T. Dalstein and B. Kulicke, “Neural network approach
to fault classification for high speed protective
relaying,” IEEE Trans. on Power Delivery,
vol. 10, no. 2, pp. 1002–1011, 1995. : 77-84.
[9] “High speed protection of EHV transmission lines
using digital traveling waves”, International journal
of academic research
Conclusion and Recommendation
This paper reveals that protection of Transmission lines
can detect, calculate location of fault by using ANN.
When faults are detected and isolated faster, it minimizes
© 2012, IJARCSSE All Rights Reserved
Page | 73