A Study on the Optimal Design of PMa-SynRM for Electric Vehicles Combining Random Forest and Genetic Algorithm.

This paper proposes a method that combines random forest technology (RF) and a genetic algorithm (GA) for optimal design of traction motors for electric vehicles (EVs). The target motor is a permanent magnet assisted synchronous reluctance motor (PMa-SynRM), for which the design goal is to increase the average torque and efficiency while reducing the total harmonic distortion (THD) of the line-to-line back electromotive force (LtoL BEMF) and torque ripple. The superiority of the RF technology was verified by comparison with other machine learning methods through two mathematical test functions. Prediction accuracy was improved through hyperparameter tuning using the GridSearchCV technique. The design variables, selected for the optimal design of the target application, were proved to be an influential variable through sensitivity analysis techniques. Finally, stress analysis, thermal analysis, and irreversible demagnetization analysis were conducted to prove the validity of the selected optimal design. The applicability of the proposed method was verified by increasing the average torque by 12.92% and reducing torque ripple, cogging torque, and LtoL BEMF THD by 36.01%, 50.0%, and 4.58% respectively.