基于混沌变量优化的神经网络PID控制

Neural Network PID Control Based on Chaos Variables Optimization

对神经网络模型辨识器的输入量y(k),u(k)进行归一化处理,一种规范化PID控制方法作为控制器。采用Logistic映射构造多个不同的混沌变量,应用到神经网络PID参数域中,根据控制系统性能指标进行混沌寻优,获得近似最优解后,再通过时变因子Z(t)在近似最优解的附近继续混沌局部寻优。仿真实验表明该方案是有效的。

The inputs y（k） and u（k） of neural network model identifier（NNMI） are normalized, the controller is a normalized PID control method. Different chaos variables are constructed by Logistic mapping, which are applied to the NNM PID parameters range. Chaos optimizations are executed according of a performance index of control system. After the better value is found, the local chaos searching continue to execute near here by using the timevarying variable Z（t）. Simulation experiments show the method is very efficient.