基于BP网络的电动汽车用无刷直流电机转矩角控制技术研究

Study of Torque Angle Control Based on BP Neural Network for Brushless DC Motor Applied to Electric Vehicles

无刷直流电机低速下存在电枢反应,影响电机出力,并造成转矩脉动;而高速下又需要弱磁控制,以拓展恒功率范围。因此,转矩角控制是至关重要的因素。转矩角控制的目的是寻找最佳电流超前相角,由于电流超前相角与转速、转矩的非线性问题,传统的确定该角度的方法都是基于某种假设,因此与实际运行情况存在相当的差异,难以应用于工程实践当中。BP神经网络具有强大的非线性映射能力,可以解决转矩角控制中的非线性问题。针对全转速范围,提出了基于BP网络的无刷直流电机转矩角控制技术,将实验数据作为训练样本利用动态全参数自适应学习算法进行离线训练,网络收敛后用作在线控制。实验结果表明,该方法可以使无刷直流电机及控制系统在全转速范围内运行于高效区,满足电动汽车对驱动系统的要求。

Torque angle of brushless DC motor （BLDCM） represents the angle between the stator flux and rotor flux, and affects the output torque and operation efficiency. Armature reaction of BLDCM in low speed results in torque ripple. In addition, field-weakening in high speed is necessary to ensure constant power operation of BLDCM. So torque angle control is the most important factor to solve the above problems. The objective of torque angle control is to find the optimal advance phase angle. The traditional methods are based on the certain hypothesis because of the nonlinear relationship among advance phase angle, speed and torque. So it is hardly applied in the engineering practice. The powerful nonlinear mapping function of BP neural network can settle the nonlinear problem of torque angle control. Therefore, the torque angle control method based on neural network is presented, in which experimental data is applied to train BP neural network in off-line way using dynamic self-adaptive learning algorithm, and afterwards converged network to control on-line. The experimental results used in the electric vehicles show that the BLDCM and its control system based on the method run in high efficiency at full speed range and satisfy the electric vehicle.