感应电机的神经网络逆系统线性化解耦控制

提出了一种新的感应电机的线性化解耦控制方法，其特点是不依赖于对象的精确数学模型与参数。通过用静态神经网络加积分器来构造感应电机的逆系统，将感应电机这一多变量、非线性、强耦合的复杂对象动态解耦成转速与转子磁链两个二阶线性子系统，然后运用线性系统理论进行综合。仿真与初步的实验结果表明系统具有优良的静态及动态解耦性能，且对电机参数的变化与负载扰动具有较强的鲁棒性。

DECOUPING CONTROL OF INDUCTION MOTOR BASED ON NEURAL NETWORKS INVERSE

In this paper, a novel linearization and decoupling method named as artificial neural network(ANN) inverse for induction motor control is proposed. It is characterized by that the construction of the ANN inverse is independent of the motor model and parameters. Cascading the ANN inverse which consists of a static ANN and four integrators with the motor, the multivariable, nonlinear and strongly coupled system is decoupled into two independent second-order linear subsystems, or rotor speed subsystem and rotor flux one, so as to be easy to design the closed-loop linear regulator to control each of the subsystems. Simulation and primary experiment results show that the good static and dynamic decoupling performance and the strong robustness to both variation of parameters and load torque disturbance can be achieved by using the proposed method.