Time-current characteristic measurement of overcurrent relay in power system using multilayer perceptron network

Protection in power system is very important to ensure the systems are in a good condition without any failure. It is necessary that the protection system can operate at the shortest time to clear the fault as soon as possible. Overcurrent relay protection is depending on their time-current characteristic curve. In this particular characteristic curve, the required relay operating time can be determined in order to remove high fault current in the system. If fault current are not remove in correct possible time, it will cause damage of expensive equipment and loss of life as well. Therefore, until now, researchers are always come out with several methods to solve this problem. In this paper, artificial neural network (ANN) using multilayer perceptron network (MLP) is applied to determine an accurate operating time. The MLP network is trained by Levenberg-Marquardt algorithm as for its fastest convergence rate and with a minimum error. Experimental results proved that the proposed method produced acceptable results in determining an accurate operating time. Sum squared error (SSE) produced during training and testing is 1.04% and 0.73%, respectively. In addition, the result reveals that the relay operating time obtained by neural network is 1ms which is considerably fast time for the system to isolate fault.