Loss Prediction of Vehicle Permanent Magnet Synchronous Motor Based on Deep Learning

Based on the deep learning BP neural network algorithm, we establish the electromagnetic torque and loss prediction analysis model of permanent magnet synchronous motor to provide new design ideas and methods for optimizing motor structure design. In this paper, four-rotor structure parameters which are Rib, Air Gap, Magnet Thickness and Magnet Width, motor electromagnetic torque, and motor losses of the "V" type Interior Permanent Magnet Synchronous Motor are selected as the research object. The BP neural network structure prediction model with 2 visible layers and 2 hidden layers was built by 256 groups of sample data calculated by Maxwell transient electromagnetic simulation. 226 out of 256 randomly selected data samples were used to train the prediction model, and 30 groups were used to test the accuracy and generalization ability of the prediction model. and the prediction results data were compared with the deep learning prediction model through finite element simulation data. The results show that the BP neural network small-sample data prediction model has high prediction accuracy in the loss prediction of the vehicle permanent magnet synchronous motor, and verifies the feasibility of the motor torque and loss prediction model based on the deep learning algorithm.