基于DRNN的非线性多变量系统控制算法

Research on Control Algorithm of Nonlinear Multivariable System Based on DRNN

针对一类多变量耦合的非线性系统,提出了一种基于对角递归神经网络(DRNN)优化的PID控制方法.首先根据系统模型机理给出了用于优化PID控制器参数的性能指标,然后利用神经网络辨识控制律所需的Jacobian信息,并针对加快神经网络权值收敛速度过程中的震荡问题,提出了一种新型的自适应动量因子.最后通过仿真验证了理论的可行性.

Aiming at a class of multivariable coupled nonlinear systems,this paper proposes an optimized PID control method based on Diagonal Recurrent Neural Network(DRNN).Firstly,according to the mechanism of the system model,the performance indicators for optimizing the parameters of the PID controller are given,and then the Jacobian information identified by the neural network is used to design the control law.Next,for the oscillation problem in the process of accelerating the convergence of the neural network weights,a new type of adaptive momentum factor is proposed.Finally,simulation verification proves the feasibility of the theoretical results.