A Comparative Study on Condition Monitoring of HVDC Transmission System using Different Artificial Neural Network Techniques

Abstract

High Voltage Direct Current Technology has been a most attractive and beneficial
transmission technology not only when power has to be transmitted over long distances but
for Grid Interconnection & Synchronization. The healthy condition of HVDC system not only
secure uninterrupted supply of electricity but manages challenges of electrical power
network. faults on power system transmission lines are supposed to be first detected and then
be classified correctly and should be cleared in least possible time for that fulfillment ANN is
used and utilized by their different training algorithm. This paper present a comparative
study on a fault finding in HVDC transmission system using Artificial Neural Network (ANN)
with three different learning algorithms. ANN is used because it is a nonlinear data driven,
adaptive and very powerful tool for forecasting purposes. The various approaches available
for fault analysis of HVDC system ANN is one of them with accuracy and easiness as key
parameters. The performance of the proposed method is investigated using
MATLAB/Simulink environment. This paper present a comparative study of a fault finding
techniques in HVDC transmission system using Artificial Neural Network (ANN) with three
different learning algorithm. ANN is used because it is a non linear data driven, adaptive and
very powerful tool for forecasting purposes. Here an attempt is made to finding the fault
using ANN with Levenberg-Marquard (LM) training algorithm, Scaled Conjugte Gradient
(SCG) and Bayesian Regularization (BR) training algorithm and their results are compared
besed on their of Mean Square Error (MSE) and regressive curve. Here are the 4000 samples
per fault based on different value of current and voltages at ac and dc side which are the
input data, feed in ANN with target i.e. at normal condition ac fault value and dc fault value
is 0 and 0 respectively, at dc side fault value of dc fault is 1 and ac fault is 0 and all for the
rest of the two conditions and then results comparison occur individually based on MSE at
some number of hidden layers then after there will be comparison of all three training
algorithm for fault finding at high speed, because there should be fast switching in circuit
breaker during fault so that it can trip easily and can keep whole system healthy.