Fault Detection and Classification of Power Plant Using Neural Networks

This study focuses on the root causes of power station problems that result in system shutdown. A power plant may experience many types of faults, some of which include, for example, line-to-line, double-line-to-ground, and single-line-to-ground. Identifying faults in the 400kV high-voltage transmission line from the Samarra Thermal Power Plant in Iraq is the primary objective of this study using artificial neural networks (ANNs). Recognizing a wide range of electrical power plant faults is an innovative use of artificial neural networks (ANNs) to speed up system recovery. The ANN can perform a wide range of tasks, including pattern recognition, classification, matching, prediction, decision-making, and control; hence, it was selected for this work. This study utilizes two neural networks trained with an input of 8 variables: 3 phases of voltage and 3 phases of current, in addition to the absolute value of the zero sequence for voltage and current. The output for the first artificial neural network (ANN) designed for fault detection will possess a single output; it is for detecting faults only, whereas the second ANN will have five outputs: four of which go to the decoding circuit and the last goes to the fault location detection circuit. Additionally, the MATLAB/Simulink 2022a software is utilized to simulate the Samarra thermal power plant model. The model represents a three-phase power system network comprising two units. The power system has four transmission lines operating at a voltage of 400kV and a frequency of 50Hz. The transmission lines have lengths of 87.10km, 145.00km, 274.00km, and 306.00km.