Hybrid technique for analysis of the faults in an underground cable

The suggested fault classification and location technique uses single cycle post fault voltage and current signal from the sending end of the underground cable (UC). Then signal disintegration is done using different signal processing method to collect set of statistical features. Thereafter by using feature extraction method (forward feature selection (FFS)) is implemented to collect the optimum features and the redundant features are discarded from the total feature set. The chosen features are fed to artificial neural network (ANN)/extreme learning machine (ELM) for fault classification and location. From the simulation of the test system it is noticed that wavelet packet transform-ELM with FFS based hybrid method provides least fault locality error (0.2%) and highest classification accuracy (99.5%) as associated with other method implemented by other researcher