基于NNs-MRAS无速度传感器双馈电机LQR控制

LQR Controller for Doubly Fed Induction Motor with Speed Sensorless Control Based on NNs-MRAS Method

针对双馈电机无速度传感器控制系统,提出了一种基于定子磁链的神经网络-模型参考自适应系统(NNs-MRAS)的速度观测法,采用差分算法设计了神经网络(NNs)模型,通过偏差反传算法对神经网络模型进行训练,使其具有良好的转速观测能力;设计了基于两相同步旋转坐标系下转子电流的线性二次型最优控制算法的控制器(LQR),并给出了状态反馈控制增益,实现了电流闭环参数的最优控制,改善了系统的动、静态性能。详尽地推导所述控制方案的实现过程,并通过基于DSP实现的样机试验,验证了控制方案的正确性和有效性。

Concerning the doubly fed induction motor(DFIM) control system without speed sensor, a speed estimation algorithm based on neural networks and model reference adaptive system(NNs-MRAS) is proposed. Differentiation method is adopted to design the neural networks model, and the good speed estimation of DFIM is obtained by training of the neural networks(NNs) through error back propagation. The linear quadratic regulator(LQR) optimal control algorithm is proposed to control the rotor currents in two-phase rotating coordinate system, and the state feedback gain is also designed in this paper. The proposed controller exhibits advanced features as: the parameters of current loop optimal control, improved dynamic and static performances of the control system. Detailed implementation of control strategies is deduced. The control schemes is verified correctly and validly via experimental results of prototype based upon DSP.