You are on page 1of 17

Neural Network Application for

Fault Analysis
ECEN 679: Computer Relays
Spring 2014
Project 1

Anupam Thatte
UIN: 921000477

1
Outline
Fault Analysis
Pattern Recognition
Artificial Neural Network (ANN)
Learning
Classification
Advantages and Drawbacks of ANN
References

2
Fault Analysis
SCADA – Alarm analysis
Digital Fault Recorder (DFR) and
Sequence of Events Recorder (SER) file
analysis
Restoration strategies
Incipient fault detection on power
transformers
Monitoring circuit breaker operation

3
Pattern Recognition

Source: Google Images, http://ulcar.uml.edu/~iag/CS/Intro-to-ANN.html 4


Artificial Neural Network (ANN)

Source: Google Images 5


ANN for fault analysis

Source: [Dalstein 1995] 6


Learning
Determine weights w1, w2, w3,…
Need large dataset for training
EMTP simulation of different fault types,
locations etc. to generate
◦ Training data
◦ Testing data

7
Classification

Source: [Kezunovic Rikalo 1996] 8


Advantages
Samples of fault currents and voltages can
be directly used
Outcome in terms of classes
Robust with noisy data
Fast to process new input data

9
Drawbacks
May need large training data set
Training time may be long
Difficult to interpret – “black box”
Design is not straight forward
◦ Number of hidden layers
◦ Choice of threshold function
◦ Learning algorithm etc.
◦ No clear guidance on this in fault analysis
literature!
10
References
1. M. Kezunovic “A Survey of Neural Net Applications to Protective
Relaying and Fault Analysis,” Engineering Intelligent Systems, vol. 5, no.
4, Dec1997.
2. M. Kezunovic, I. Rikalo, and D. Sobajic, “Real-Time and Off-Line
Transmission Line Fault Classification Using Neural Networks”, Intl.
Journal of Engineering Intelligent Systems, vol. 4, no. 1, Mar.1996.
3. S. Haykin, Neural networks: a comprehensive foundation, Prentice Hall,
Upper Saddle River, NJ, USA, 1999.
4. A. K. Jain, J. Mao, and K. M. Mohiuddin, Artificial Neural Networks: A
Tutorial, Computer, pp. 31-44, Mar1996.
5. K. S. Swarup and H. S. Chandrasekharaiah, "Fault detection and diagnosis
of power system using artificial neural networks", Proc. 1st Int. Forum on
Applicat. Neural Networks to Power Syst., pp.102 -106, 1991.
6. K. R. Niazi, C. M. Arora, and S. L. Surana, Power system security
evaluation using ANN: feature selection using divergence, Electric Power
Systems Research, vol. 69, issues 2–3, pp. 161-167, May 2004.

11
References
7. 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 -1009, 1995.
8. E. E. Vazquez, H. Altuve, and O. Chacon, "Neural network approach to fault
detection in electric power systems", IEEE Trans. Power Delivery, vol. 11,
pp. 2090 -2095, 1996.
9. A.P. Alves da Silva, A. H. F. Insfran, P. M. Da Silveira, and G. L. Torres,
"Neural networks for fault location in substations," IEEE Trans. Power
Delivery, vol.11, no.1, pp. 234-239, Jan 1996.
10. M. Kezunovic and I. Rikalo, "Detect and classify faults using neural nets,"
IEEE Computer Applications in Power, vol.9, no. 4, pp. 42-47, Oct. 1996.
11. E. H. P. Chan, "Application of neural-network computing in intelligent alarm
processing (power systems)," Power Industry Computer Application
Conference (PICA), pp. 246-251, 1-5 May 1989.
12. ECEN 679 Reading – J.G. Webster, Wiley Encyclopedia of Electrical and
Electronics Engineering, Wiley & Sons, 1999. (Chapter: M. Kezunovic and H.
Drazenovic-Perunicic, Fault Location)

12
Backup
Learning Algorithms
FaultLocation requirements
Types of ANNs
Supervised vs. Unsupervised learning

13
Learning Algorithms
Error-correction learning: e.g.
backpropagation algorithm
Memory based learning: e.g. k nearest
neighbor
Hebbian learning
Competitive learning
Boltzmann learning: similar to error-
correction but considers probability
distribution
14
Fault Location Requirements
Determine fault type so proper auto-
reclosing option can be applied.
Accurate to within a span of two towers.
Typically 0.1% error is acceptable.
Accurate even if only few cycles (3-4) of
data is measured.
Accuracy should not deteriorate for
various types of faults.

15
Types of ANNs
Feed Forward Neural Network:
◦ Information moves in only one direction from
input through hidden layer to output
◦ No cycles or loops in the network
Recurrent Network:
◦ Include both feed-forward and feed-back
◦ Can do computations on an input of variable
length

16
Supervised vs Unsupervised
Supervised Models
◦ Neural Networks
◦ Multi-Layer Perceptron
◦ Decision Trees
Unsupervised Models
◦ Different Types of Clustering
◦ Distances and Normalization
◦ Kmeans
◦ Self Organizing Maps

17

You might also like