化工学报 ›› 2019, Vol. 70 ›› Issue (12): 4749-4759.DOI: 10.11949/0438-1157.20190716

• 过程系统工程 • 上一篇    下一篇

基于改进回声状态网络的游离氧化钙预测控制

李德健1(),刘浩然2(),刘彬1,刘泽仁1,王卫涛1,闻岩3   

  1. 1. 燕山大学电气工程学院,河北 秦皇岛 066004
    2. 燕山大学信息科学与工程学院,河北 秦皇岛 066004
    3. 燕山大学机械工程学院,河北 秦皇岛 066004
  • 收稿日期:2019-06-25 修回日期:2019-07-29 出版日期:2019-12-05 发布日期:2019-12-05
  • 通讯作者: 刘浩然
  • 作者简介:李德健(1994—),男,硕士研究生,501807366@qq.com
  • 基金资助:
    河北省自然科学基金项目(F2019203320);国家自然科学基金项目(51641609)

Predictive control of free calcium oxide based on improved echo state network

Dejian LI1(),Haoran LIU2(),Bin LIU1,Zeren LIU1,Weitao WANG1,Yan WEN3   

  1. 1. College of Electrical Engineering, Yanshan University, Qinhuangdao 066004, Hebei, China
    2. College of Information Science and Engineering, Yanshan University, Qinhuangdao 066004, Hebei, China
    3. College of Mechanical Engineering, Yanshan University, Qinhuangdao 066004, Hebei, China
  • Received:2019-06-25 Revised:2019-07-29 Online:2019-12-05 Published:2019-12-05
  • Contact: Haoran LIU

摘要:

在非线性时延水泥烧成系统中,针对传统预测控制方法调节时间长、控制精度不高的问题,提出一种改进的在线型回声状态网络预测控制模型。首先将带有L1范数约束项的递归最小二乘法与回声状态网络相结合构建在线型预测模型,解决传统预测控制模型辨识精度较低、无法进行实时预测的问题;然后基于改进的回声状态网络预测模型,构建预测控制模型结构,并采用具有全局优化能力的粒子群算法进行滚动优化,保证实际输出量快速、准确、平稳地跟随被控量的设定值;最后利用改进的预测控制模型对水泥烧成系统中的游离氧化钙含量进行预测控制仿真实验,结果表明改进的预测控制模型具有良好的性能和应用前景。

关键词: 模型预测控制, 神经网络, 回声状态网络, L1正则化, 优化, 烧成系统

Abstract:

In the nonlinear time-delay cement burning system, an improved on-line echo state network predictive control model is proposed for the problem that the traditional predictive control method has long adjustment time and low control precision. Therefore, we firstly combine the L1-norm constrained recursive least squares method with the echo state network to construct an on-line prediction model to address these issues. Then, the structure of predictive control model is constructed based on the improved prediction model of echo state network. And particle swarm optimization (PSO) algorithm with global optimization capability is utilized for rolling to ensure that the actual output follows the setting value of the controlled variable quickly, accurately and smoothly. Finally, the simulation prediction experiments of the free calcium oxide content in the cement burning system are conducted using the improved predictive control model. The results show that the improved predictive control model has good performance and application prospects.

Key words: model predictive control, neural network, echo state network, L1 regularization, optimization, burning system

中图分类号: