摘要
目的针对结晶器液位系统在实际运行中会遇到突然干扰导致控制性能变差的问题,提出一种基于神经网络的智能PID和模型辨识相结合的控制方法。方法建立结晶器液位系统模型,采用对角回归型(DRNN)神经网络PID控制方法,并应用梯度下降法对结晶器控制器的权值进行调整和优化,结合传统PID控制方法对系统抗干扰效果进行分析。结果两种控制方法对比,DRNN控制算法的超调量更小,而在响应速度方面,与传统PID比较更具有调节周期短的明显优势。结论通过MATLAB仿真证明,采用DRNN神经网络PID的控制方法,可以使系统具有较强的自适应能力,当系统受到外来扰动时,不会出现过大的振荡,可以帮助系统及时恢复。
Objective For the problem of deterioration of control performance caused by sudden interference in actual operation,a combination method of intelligent PID and model identification of based on DRNN was proposed.Methods The mold liquid level system model was established,the diagonal regression(DRNN)neural network PID control method was adopted,and the gradient descent method was used to adjust and optimize the weight of the mold controller.Combined with the traditional PID control method,the anti-interference effect of the system was analyzed.Results Compared with the two control methods,the overshoot of DRNN control algorithm was smaller,and in terms of response speed,it had the obvious advantage of short regulation cycle compared with traditional PID.Conclusion MATLAB simulation shows that the DRNN neural network PID control method can make the system have strong adaptive ability.When the system is disturbed by external disturbance,it will not have too large oscillation,which can help the system recover in time.
作者
缸明义
夏兴国
潘小波
张奇
GANG Ming-yi;XIA Xing-guo;PAN Xiao-bo;ZHANG Qi(Department of Electrical Engineering,Maanshan Technical College,Maanshan,Anhui 243031,China;Department of Mechanical and Automatic Engineering,Xiamen City University,Xiamen,Fujian 361008,China)
出处
《河北北方学院学报（自然科学版）》
2023年第1期23-28,35,共7页
Journal of Hebei North University:Natural Science Edition
基金
2017年度安徽省高校学科(专业)拔尖人才学术资助项目(gxbjZD59)
安徽省高校自然科学研究重点项目(KJ2019A1245)
安徽省质量工程项目“工业机器人应用示范实验实训中心”(2020sxzx54)
院级自然科学研究目(MKJ2021010
2021ylm019
MKJ2021007)。
关键词
对角递归神经网络
连铸
结晶器
液位系统
梯度下降
diagonal recurrent neural network
continuous casting
mold
liquid level System
gradient descent
