基于BP神经网络的开关磁阻电机直接转矩控制系统及实现

摘要

直接转矩控制可有效抑制开关磁阻电机（SRM）转矩脉动。由于开关磁阻电机双凸结构和磁路的严重饱和，造成其转矩是关于电流和转子位置的严重非线性函数，转矩计算非常困难。针对这一问题，本文提出一种采用基于BP神经网络建立开关磁阻电机转矩模型的方法。利用有限元仿真得到的转矩样本对BP神经网络经行进行离线训练，完成电流、位置角度到转矩的非线性映射，构造出基于BP神经网络的转矩观测器。再将构造好的转矩观测器应用于电机直接转矩控制系统中，对电机的转矩经行进行实时在线估算。最后，将估算转矩经行反馈，完成电机的直接转矩控制。该控制方法利用了BP神经网络泛化、逼近能力强的优点，同时控制过程简单，无需在线训练。实验结果表明，所提方法转矩计算速度快、计算精度高，可以满足实时控制的要求，有效地减小了电机的转矩脉动。%Direct torque control can effectively restrain the torque ripple of switched reluctance motor(SRM). With bi⁃convex structure and magnetic saturation of SRM,the torque is a nonlinear function with respect to current and rotor po⁃sition,which makes its calculation very difficult. To solve this problem,this paper proposes a method to establish the torque model of SRM based on BP neural network. First,BP neural network is off-line trained based on the torque sam⁃ple generated by finite element simulation,and BP neural network can achieve the nonlinear mapping from current and rotor position to torque,thus the torque observer is established. Then,the constructed torque observer is used in the di⁃rect torque control system to estimate the torque on-line. Finally,the feedback of torque is applied to the direct torque control of the motor. This method takes the advantages of BP neural network,such as its generalization and strong ap⁃proximation. Moreover,its control process is simple and does not need training on-line. Experimental results show that the proposed method has quick calculation speed and high precision,which can meet the requirement of real-time con⁃trol and effectively reduce the torque ripple of the motor.