小波神经网络PID控制在板球系统中的研究  

Research on Wavelet Neural Network PID Control in Ball and Plate System

作　　者：夏国锋 向凤红[1] 杨立炜 XIA Guofeng;XIANG Fenghong;YANG Liwei(Faculty of Information Engineering and Automation Kunming University of Science and Technology,Kunming 650000 China)

机构地区：[1]昆明理工大学信息工程与自动化学院,昆明650000

出　　处：《电光与控制》2022年第9期84-89,共6页Electronics Optics & Control

基　　金：国家自然科学基金(61163051);云南省重大科技专项计划项目(202002AC080001)。

摘　　要：针对板球系统PID控制震荡严重、人工整定PID繁琐以及动态品质差等缺点,研究了小波神经网络(WNN)辨识和WNN-PID自整定参数相结合的方案。首先,针对板球强耦合特性,通过拉格朗日方程建立了由两部分组成的板球系统模型;其次,为了克服PID人工整定繁琐和稳定性差等问题,构建了WNN-PID控制器;考虑到梯度下降法和固定学习速率容易陷入极值,利用动量梯度和AdaDec算法,加快了网络的训练速度;然后,利用Lyapunov稳定性理论验证了系统收敛性;最后,通过Matlab仿真实验可知,在板球系统中所提策略的稳定性和鲁棒性均优于常规PID和BP-PID策略。Aiming at the shortcomings of ball and plate system,such as severe PID control oscillation,tedious manual PID tuning and poor dynamic quality,a scheme combining Wavelet Neural Network(WNN) identification with WNN-PID self-tuning parameters is studied.Firstly,according to the strong coupling characteristics of ball and plate,a ball and plate system model consisting of two parts is established by Lagrange equation.Secondly,WNN-PID controller is constructed to overcome the problems of tedious manual tuning and poor stability of PID.Considering that gradient descent method and fixed learning rate are easy to fall into extreme value,the momentum gradient and AdaDec algorithm are used to accelerate the training speed of the network.Then,the convergence of the system is verified by Lyapunov stability theory.Finally,the Matlab simulation shows that the stability and robustness of the proposed stragegy in ball and plate system are better than those of conventional PID and BP-PID strategies.

关 键 词：板球系统 小波神经网络 LYAPUNOV稳定性理论 BP神经网络 

分 类 号：TP273[自动化与计算机技术—检测技术与自动化装置]