Multi-Output Deep Learning-Based Temperature Prediction for PMSM Motors in Electric Vehicles

Abstract

This paper highlights precise prediction of internal temperatures of the Permanent Magnet Synchronous Motor(PMSMs) which is occupied in many electric vehicle applications. Keeping track of motor temperature is very crucial for checking motor's efficiency, reliability, and safety. Insulation failure may occur, and the lifespan of the motor can also be reduced due to excessive heat. Physical measuring of internal temperature of the motor is difficult and also costly. To overcome these issues quantitative methodology is suggested using a multi output deep learning model. The study uses dataset which consist of mechanical, electrical and thermal parameters of the motor like ambient temperature, coolant temperature, voltages, currents in d-axis and in q-axis, motor speed and torque. For data preprocessing feature engineering and normalization is used and then dataset is divided into training and testing samples which refine model performance. The novelty of this work lies in the development of a multi-output deep learning framework capable of simultaneously predicting four critical PMSM temperature parameters using a single model. This approach reduces computational complexity compared to deploying separate models and supports real-time thermal monitoring in EV applications. Notice: We are not claiming a new algorithm. We are highlighting the multi-output framework.