A MODEL FOR DETECTING SHORT CIRCUITS IN STATOR WINDINGS OF AN ELECTROMECHANICAL ENERGY CONVERTER

Abstract

Electrical faults occurring in electromechanical energy converters mainly manifest in the operating characteristics of their windings. In this paper, a new approach to detecting and classifying phase short circuits in an electromechanical energy converter (EMC) based on an artificial neural network (ANN) is proposed. Characteristic dependencies of the EMC stator winding for single-phase, two-phase and two-phase short circuits on the neutral are obtained. The dynamics of change in the phase currents and voltages for various operating modes are obtained and studied. Based on the obtained data, the necessary database for testing and training is formed.

A block diagram of the algorithm for detecting and classifying the phase short circuits of in EMCs is developed. It consists of data evaluation and data training blocks. The data evaluation block collects, evaluates and creates a database of phase currents and voltages. Threshold values ​​​​of short circuits are used to detect and classify their occurrence. Phase short circuits are detected using an artificial neural network created in the MATLAB environment. The obtained results were tested on the example of a synchronous generator.

The developed artificial neural network has 4 layers. The number of input neurons is 48. They are obtained as a result of processing the variables Ia, Ib, Ic, Ua, Ub, Uc according to various criteria. The number appearing in the output neuron reveals the presence or absence of a short circuit. The high performance and accuracy of the constructed neural network are substantiated.

The results obtained can be successfully used to develop intelligent monitoring systems for electromechanical energy converters.