神经网络在线学习补偿自适应控制及其应用

Neural network based on-line learning indemnityadaptive control and its application

针对电液伺服系统的复杂非线性和不确定性特性,基于反馈误差学习法、小波分析理论并结合面向控制的辨识思想,提出了神经网络在线自学习自适应控制与"参征器"补偿控制相结合的控制方法.该方法将"过程辨识"和"参征器"引入反馈误差学习法的神经网络学习和控制中,控制参数的调整基于被控过程的小波变换结果信息,利用反馈误差学习法实现;"参征器"起监督和补偿控制作用,避免控制器的输出产生振荡或进入饱和状态.应用研究结果证明:该方法避免了采用直接反馈误差法可能造成的饱和和过调整问题;有效地提高了系统的稳定性、鲁棒性、控制精度和自适应能力.该方法能有效地处理工业系统中普遍存在的复杂非线性和时变不确定性特性,控制效果明显优于传统的反馈误差学习方法.为未知不确定非线性系统的智能控制提供了一条有效而可行的新途径.

With respect to the complex nonlinearities and uncertainties of electro_hydraulic servo system,a method of neural network on_line self_learning adaptive control and OSC(oscillatory surge controller) compensating control is introduced based on the feedback_error_learning and wavelets analytic theory combined with identification idea.By introducing process_identification and OSC in neural network learning and controlling with feedback error learning method, the control parameters can be regulated on_line by using the measured input/output data of wavelets and OSC for supervising and indemnity control, so as to prevent the oscillation and saturation of the controller output. The application results show that the method prevents the saturation and the overadjustment of using direct feedback error learning method, and effectively enhance stability, robustness, control precision and adaptive ability of the system. The method can effectively deal with the complex nonlinearities and uncertainties which are ubiquitous in industry system and its control effect is superior to that of the feedback error learning method. It provides a new way which is effective and feasible for intelligent control of nonlinear and uncertain systems.