Development of a Simulation Model for Detecting Phase Short Circuits of a Three-Phase Synchronous Machine Based on an Artificial Neural Network

Abstract

This paper presents a new approach to the detection and classification of phase short circuits of a three-phase synchronous machine based on an artificial neural network. A simulation model of phase short circuits occurring in an electric machine is constructed in the MATLAB software environment. With its help, it became possible to study the dynamics of changes in phase currents and voltages and obtain the database necessary for testing and training. A two-stage algorithm for detecting and classifying phase short circuits of a three-phase synchronous machine is proposed. At the first stage, the collection, evaluation and creation of a database of phase currents and voltages obtained using a simulation model is carried out. To assess the malfunction using phase currents and voltages, their threshold values were obtained. At the second stage, the artificial neural network was trained using the back-error-propagation algorithm. The results obtained were tested on a 2.85 kW generator.The root mean square error of training a four-layer artificial neural network developed in the MATLAB software environment was 0.0198, the process took 17 minutes. Observation of the linear regression plot showed that the performance of the constructed network is high. The results obtained can be successfully used for the development of intelligent monitoring and diagnostics systems for synchronous machines.