NEURAL NETWORK PREDICTION OF LEAKAGE CURRENT BASED ON THE THEORY OF TIME SERIES FORECASTING

Abstract

Among the modern electrotechnical means of monitoring the state insulation electric motors are devices for monitoring the magnitude the leakage current, which reflects the current state insulation electric motor. The use of more sophisticated devices that not only fix the dangerous value the leakage current by turning off the electric motor, but also predict the possibility of reaching a dangerous value the leakage current make it possible to inform the service personnel about the possible danger, reduce the time for simple equipment during the technological process due to the possibility of early maintenance, replacement or repair of electric motors during a technological pause, even before they fail. Neural networks, used for predicting the reliability electric motors, are mainly performed as mathematical models of parallel calculations, which have in their composition simple processing elements that interact with each other and are called artificial neurons. Non-linearity of neural networks allows establishing non-linear dependencies between future and actual values of processes. Other important advantages are scalability - the parallel structure of artificial neural networks accelerates calculations, which is extremely relevant on an industrial scale, when it is necessary to process terabytes of data.The neural network created on the basis the theory of time series forecasting is a technological suitability test for predicting the leakage current of an electric motor. The synthesized neural network can be the basis for creating a system for predicting the leakage currents of electric motors based on the theory of time series forecasting. The forecasting system includes a neural network based on the theory of time series, means of measuring the leakage current of electric motors and a database. The key decision for the developed system is made by a person.