摘  要：针对永磁同步电机这一多变量、非线性、强耦合的控制对象，提出了一种基于神经网络在线辨识的永磁同步电机逆系统解耦控制新方法。通过静态神经网络加积分器来构造永磁同步电机的逆系统，并在实际运行中不断地修正神经网络权值，使其更精确地逼近逆系统。将逆系统与永磁同步电机原系统复合成两个伪线性子系统，使永磁同步电机解耦成二阶线性转速子系统和一阶线性磁链子系统，在此基础上，运用线性系统理论进行综合。仿真试验表明这种控制策略能够实现永磁同步电机转速和定子磁链之间的动态解耦控制，并且系统具有良好的动静态性能。
关键词：永磁同步电机；神经网络；逆系统；在线辨识；解耦控制

Abstract: A novel approach of inverse system decoupling control based on artificial neural network(ANN)online identification was proposed for permanent magnet synchronous motor(PMSM),which was multivariable,nonlinear and strong coupled system.A static ANN and three integrators were used to construct the ANN inverse system of the PMSM,and the connection value of ANN could be amended on-line in order to approach the inversion exactly.Consequently,two pseudo-linear subsystems were completed by combining the ANN inverse system with the PMSM.The PMSM was decoupled into a second-order linear speed subsystem and a first-order linear flux subsystem.Then the linear system theory was used to design the closed-loop linear regulators to control each of the subsystems.Simulation results show this kind of control strategy can realize dynamic decoupling control between speed and flux of the PMSM,and the control system has fine dynamic and static performance.
Key Words: PMSM;ANN;Inverse system;On-line identification;Decoupling control