Optimized Fault Detector Based Pattern Recognition Technique to Classify and Localize Electrical Faults in Modern Distribution Systems

This research presents a method that integrates artificial neural networks (ANN) and discrete wavelet transform (DWT) to identify and classify faults in large power networks, as well as to pinpoint the zones where these faults occur. The objective is to enhance reliability and safety by accurately detecting and categorizing electrical faults. To manage the computational demands of processing the extensive and complex data from the power system, the network is divided into optimal zones, each made visible for fault detection. Niche Binary particle swarm optimization (NBPSO) is employed to place the fault detectors (FD) in each zone. This allows for precise measurement of fault voltage and current phasors without significant cost. The ANN module is tasked with identifying the fault area and locating the exact fault within that zone, as well as classifying the specific type of fault. Discrete Wavelet Transform is used for feature extraction, and a phase locked loop (PLL) is used for load angle computation. The proposed method's validity has been tested on the IEEE-33 bus distribution network.