Analysis of neural network vector control for IPM machine in electric vehicles

摘要

This paper investigates a neural network vector control for an interior permanent magnet (IPM) machine in an electric vehicle (EV). At the present, most electric vehicles are embedded with IPM machines because of the high torque capability and power density of the machine. Many literatures conducted research on how to control the machine most optimally. This paper proposes a neural network vector control method that implements the optimal control based on adaptive dynamic programming. As for the training, Levenberg-Marquardt and forward accumulation through time algorithm is used. The paper also develops an EV simulation system by utilizing off-the-shelf EV mechanical and electrical components and MATLAB SimDriveline and SimPowerSystems in order to evaluate the performance of the neural network controller in close to practical EV operating conditions. The design of the EV simulation system considers various electrical subsystems, vehicle dynamics, shaft and differential, and simple regenerative and hydraulic braking. The proposed neural network control is compared with traditional PI-based control. The result shows that the proposed controller can be a potential replacement of the existing control scheme, such as PID, fuzzy logic, or etc, and provides adequate traction control in EV application.