H_∞ Robust Control Based On Recurrent Neural Network Of Permanent Magnet Linear Synchronous Motor

Permanent magnet linear synchronous motor (PMLSM) servo system is a dynamic complex nonlinear system with high precision and fast response characteristics. Its main performances are the tracking ability of the input command and rejection ability of the disturbances (including influence of parameter variations, model error and exterior destabilizations). An H_∞robust controller based on local feedback recurrent neural network (RNN) in this thesis is proposed for the position tracking control of PMLSM.First of all, using MATLAB/SIMULINK, a general position tracking simulation model of PMLSM controlled by space vector PWM is established based on the analysis and research of PMLSM and its vector control principle. Afterwards, the simple and practical RNN for real-time control is proposed to estimate the dynamic mapping of lumped element uncertainty. Based on the online parameter training result of RNN, the quadric form function of the H_∞robust controller is chosen and the performances of stabilizing the closed loop are obtained by the back propagation method, which not only equips the proposed controller with strong robustness under large perturbation of load and system parameters but also increases the learning capability of the RNN. The proposed method is capable of tracking both step and sinusoidal input command. The tracking performances are ensured in the cases of parameter variations, external disturbance and RNN estimation error once a H_∞performance requirement is achieved. Finally, contrasted with the H_∞robust controller integrated IP controller, simulated in three cases(no destabilizations, intestine destabilizations, exterior destabilizations), the simulated results show that the system has good robustness, fast response characteristic, and rejection ability of the large parameter change and large external disturbance.