Detection, Location, and Classification of Fault Applying Artificial Neural Networks in Power System Transmission Line

Abstract

Transmission lines transfer power from generating stations to substations and then to end customers in a power system. A reliable power system is a system with a robust fault location system that reduces downtime in case of any power failure. Therefore, today a technique is required which must be robust and accurate in finding the location of the fault. A technique for detecting the presence of a fault, and locating and classifying faults in transmission lines in the power system through an Artificial Neural Network (ANN) is presented in this paper. This technique through ANN is executed and tested in MATLAB on a 3-Bus power system. In this technique, a backpropagation algorithm is used in the feed-forward network. The Levenberg Marquardt optimization technique is selected. Linear regression and Mean Square Error (MSE) are the performance standard measures in MATLAB. Performance is the best when MSE is very low or close to zero and the Regression value (R) is 1. RMS values of currents, voltages of three phases, and zero-sequence voltage and current are taken as input in this method during fault and without fault. The Regression (R) value which measures the correlation between output and target is 1 and MSE is 1.462e-10 for the Neural Network created for detecting the presence of a fault in the line. ANN which locates and classifies the fault in the transmission line has the performance value R-0.9818 and MSE 0.16178. Thus, the performance parameters of both the ANN models have shown satisfactory and acceptable results.