Computationally Efficient Estimation of PWM-Induced Iron Loss of PMSM Using Deep Transfer Learning

As the demand for increasing the efficiency of traction motors for increasing the mileage of electric vehicles, it is necessary to accurately estimate the efficiency of traction motors at the early design stage. Since the iron loss of the traction motor is highly affected by the pulse width modulation (PWM) frequency, the PWM current should be considered when designing the motor. However, it is difficult in considering the PWM current at the early design stage because of its high computation cost due to the small time step for representing the high-frequency harmonics. Therefore, we propose a method to reduce the computation cost for the calculation of PWM-induced iron loss using deep transfer learning (DTL) even with a small amount of data. The proposed method can be achieved by training a deep neural network (DNN) that can predict PWM-induced iron loss accurately using a large amount of sinusoidal current-based iron loss and a small amount of PWM-induced iron loss. As a result, the PWM current can be practically considered in the design stage of the traction motor because the computation cost can be decreased by using the proposed method.