Neural Network-Driven Sensorless Speed Control of EV Drive Using PMSM

To reduce pollution and energy consumption, particularly in the automotive industry, energy saving is the main concern, and hence, Electric vehicles (EVs) are getting significantly more attention than vehicles with internal combustion engines (IC engines). Electric motors used in Electric Vehicles (EVs) must have high efficiency for maximum utilization of the energy from the batteries or fuel cells. Also, these motors must be compact, lightweight, less expensive and very easily recycled. Further, to obtain better dynamic performance, various motor control strategies are used to control the speed of the motor. And to have increased reliability, sensorless speed control techniques that offer sufficiently high performance are used. The sensorless speed control techniques are largely divided into three groups: state observer methods, indirect measurement methods and saliency-based methods. Generally, the state observer uses back emf or flux linkage to estimate the speed of the motor. Since the back emf is directly proportional to the rotor speed, at low-speed back emf based method will give poor performance. The current-based Model Reference Adaptive System (MRAS) model is also popular for estimating low speed; however, assessments deteriorate during high performance applications such as EV. This paper presents an artificial neural network (ANN)-deployed sensorless speed control of permanent magnet synchronous motor (PMSM) drive used in EVs. In this paper, the estimation of speed using the current-based MRAS model is discussed and compared with the proposed ANN-based controller, which shows significant improvement in the performance of EV motor drives. The MATLAB simulation and experimental results are presented to validate the proposed algorithm.