WPD and RBFNN Based Fault Location Estimation on TCSC Based Series Compensated Transmission Line

Abstract

This paper presents a technique for estimation of fault location on the thyristor control series capacitor (TCSC) compensated transmission line for different types of faults. This technique is developed using wavelet packet decomposition (WPD) and radial basis function neural network (RBFNN) for fast and accurate estimation of fault location. The faulty current signal of 10 ms (half cycle) duration after fault inception is recorded. The fault signals recorded with 40 kHz sampling frequency are decomposed by WPD up to 3rd level with mother wavelet db6 to compute wavelet packet energy (WPE). The features of current signals are extracted by WPD to train the RBFNN for different fault locations. Training and Testing are done by varying fault resistance and position of TCSC. It is found that the combination of RBFNN and WPD coefficient’s energy of fault signals can locate the fault with high accuracy.