神经网络自整定多变量PID控制方法研究

Study on Self-Tuning Multivariable PID Controller Based on Neural Network

关于污水优化处理系统控制问题。针对三相内循环生物流化床污水处理系统结构上的特殊性和生物反应过程的复杂性,以及污水处理过程的非线性、时滞、时变、多输入多输出等特点,提出一种对角递归神经网络(Diagonal Recurrent Neural Network,DRNN)的自整定多变量PID控制方法。通过DRNN对控制参数进行在线辨识,为PID控制器提供梯度信息;同时考虑到生物流化床污水处理系统输出变量之间的特殊耦合关系,将反馈偏差划分为不同等级,以调整PID参数修正速率的快慢。进行仿真验证,验证了该方法在生物流化床污水处理中的有效性和可行性。

Aiming at the particularity of structure and the complexity of biological reaction process of the three-phase inner circulation biological fluidized bed（BFB） sewage system,and considering the time delay,nonlinear,time-varying,multiple input and multiple output of sewage treatment process,an auto-turning and multi-variable PID control method based on the diagonal recurrent neural network（DRNN） was proposed in this paper.In order to provide the gradient information for the PID controller,we made use of the DRNN to identify control parameters on-line.In addition,considering particular coupling relationship between output variables,feedback was divided into different level to adjust the rate of modify PID parameters.Finally,the simulation experiment results show that the method is feasible and effective in the BFB sewage system.