A novel decoupling control method based on neural network for EV's Driving PMSM

Permanent Magnet Synchronous Motor (PMSM) is often used in Electrified Vehicle (EV), while its I_d (direct-axis current) and I_q (quadrature-axis current) are coupled. Traditional FOC (Field Oriented Control) method can't get an accurate decoupling control of them when the motor speed changed with the vehicle's driving condition, especially at high speed. This coupling issue leads an unstable torque output. And further, it deteriorates the NVH (noise vibration and harshness) performance of EV. This paper focuses on the decoupling solution and puts forward a new control strategy which combines the neural network control idea with FOC method. The novel method gives a neural network controller based on four single neuron PID controllers, which its function is to realize the d-q axis interacted adjustment and decoupling. Four single neurons PID controllers achieve the negative feedback control of I_d to U_d (direct-axis voltage), I_q to U_d, I_q to U_q (quadrature-axis voltage), and I_d to U_q respectively. Creatively, it takes PMSM speed as one of the neuron inputs to adjust the feedback weight of I_d and I_q dynamically. A comparing simulation which is set up in the Simulink platform is given in this paper. Simulation results show that this method gives a good self-adaptiveness and decouples the influence of I_d and Iq, as well as improve the motor control quality at high speed.