Prediction of Partial Discharge Inception Voltage for Electric Vehicle Motor Insulation Using Deep Learning
Shakeel Akram, Peng Wang, Xi Zhu, Jialiang Huang, Feng Liu, Zhi Fang, H. Ahmed
Abstract
Monitoring the health of electric motor insulation by measuring the early breakdown phenomenon such as Partial discharge (PD) is required for the safer operation of electric vehicles (EVs). The prediction of partial discharge inception voltage (PDIV) is of great significance to evaluate the condition of motor insulation. This paper presents a method to predict the PDIV of electric motor magnetic wires winding. The data is driven by constructing a single point discharge model of turn-to-turn insulation of two magnetic wires. The model is simulated by varying the input parameters of magnetic wire configuration such as diameter, insulation thickness, winding temperature and insulation permittivity to calculate the PDIV based on Townsend discharge theory. The calculated PDIVs are used to train the convolution neural network (CNN)and extract the nonlinear relationship between characteristic variables and PDIV to test it for large data sets of magnetic wire. The predicted value is compared with the measured value, which proves the superiority and accuracy of the method in this paper. Based on the calculation results of the mean impact value algorithm, a method to improve the PDIV of the turn-to-turn insulation is given.