基于BP网络的反馈控制器设计

针对常规反馈控制器参数在对象时变情况下难以获得最优的问题，利用BP神经网络构成系统反馈控制器，通过自适应学习速率在线调整网络权值以逼近对象的逆动态模型，并利用Lyapunov方法给出了该算法的收敛的条件。将算法应用于循环水温度控制系统表明：该控制器对模型参数不依赖，能有效地适应控制对象参数的变化，系统具有较强的鲁棒性。

Design of feedback controller based on BP network

The parameters of general feedback controllers difficult to be optimized according to time-varying object are tackled by using alternatively back-propagation network which acts as feedback controller, approximating invert dynamic model by on-line adjusting weights value with adaptive learning rate and the convergence condition of this algorithm due to the use of Lyapunov principle. The application of the algorithm to a temperature control of cycling water shows that the controller exhibits no dependence on the mathematic model, effective control of variation of object parameters and strong robustness.