Temperature estimation of vehicle electric drive system based on machine learning algorithm

The high temperatures of electric drive system (EDS) will affect the performance and reliability of the EDS, so it is integral to estimate the temperature of the EDS, avoiding too high temperature of the EDS. A prerequisite for the optimization of an electric drive thermal management system is that the temperature of the EDS can be accurately estimated. In this paper, an EDS temperature estimation method is proposed based on particle swarm optimization (PSO) and back propagation neural network (BPNN). And the temperature of key components of the EDS is estimated, including the temperature of motor winding, rotor, IGBT, and motor shaft gear. The results show that the PSO-BP estimation is more accurate than the BP estimation, and the R 2 values of PSO-BP for the temperature estimation of the four key components of the EDS are 0.994, 0.995, 0.990, 0.988. The mean absolute error (MAE) values are 0.731, 0.491, 0.489, 0.343, and the mean square error (MSE) values are 1.049, 0.479, 0.400, and 0.381.