带有改进自适应动量因子的四容水箱DRNN控制系统设计

Design of Quadruple-tank DRNN Control System with Improved Adaptive Momentum Factor

针对多变量耦合的四容水箱系统,提出一种带有自适应动量因子的对角递归神经网络(DRNN)自学习PID控制方法。首先依据四容水箱实验平台的结构,简要介绍了被控对象的数学模型,并给出了以PID为主控制器,DRNN用于辨识系统动态的总体控制方案;其次,考虑到耦合系统中控制输入与系统输出的之间耦合作用,为了减小整体跟踪误差,给出了调节PID控制器参数的全局性能指标;然后为了防止网络权值调整过程中出现收敛缓慢和震荡问题,设计了自适应动量因子;最后通过数值展示与仿真对比,验证了该文所提方法的有效性。

This paper proposes an improved PID self-learning control strategy based on diagonal recurrent neural network(DRNN)for quadruple-tank with the characteristic of multiple variables and coupling.Firstly,the mathematical model of the multi-variable coupling object is briefly introduced according to the structure of the quadruple-tanks experimental platform,simultaneously the overall control scheme is presented where PID is the controller and DRNN is utilized to identify the dynamics of system.Then,considering the coupling effect of the control input and output,a global performance index for controller parameter adjustment is given to reduce system errors in the process of adjusting controller parameters.An adaptive momentum factor is designed in order to prevent problems such as slow convergence and oscillation during network weight adjustment after that.Finally,the effectiveness of the proposed method is verified through numerical demonstration and comparative simulation.