基于动态递归神经网络的自适应PID控制

提出一种基于动态递归神经网络的自适应PID控制方案，该控制系统由神经网络辨识器和神经网络控制器组成。辨识器采用单隐层的动态递归神经网络，网络结构为2-4-1；辨识算法为动态BP算法；控制器采用两层线性结构的神经网络，输入为系统偏差及其一阶、二阶微分，因此具有增量型PID控制结构。应用该控制系统对一非线性时变系统进行仿真研究，仿真结果表明该控制方案不仅具有良好的跟踪特性，而且对系统参数变化具有较强的鲁棒性。

Adaptive PID Control Based on Dynamic Recurrent Neural Network

An adaptive PID control scheme based on dynamic recurrent neural network is presented. The control system is consisted of the neural network identifier and the neural network controller. The identifier is a dynamic recurrent neural network with one hidden layer, in which the structure of the network is 2-4-1. The dynamic BP algorithm is used to identify the plant on line. The controller is a neural network with two-layer linear structure. The inputs of the controller are the system deviation, the first and second differentiation, so it has a mechanism similar to the PID controller with incremental form. The control system is used to control a nonlinear time-variant system. Simulation results show that it not only has good traceability, but also has good robustness to the variation of parameters.