摘要
基于相关性即时学习法(Co-JITL)在对化工过程建立模型时,面对不同工况需要人为对多个参数进行调节,这导致模型在使用上缺乏便利性,也容易使模型陷入过拟合情况,不利于模型及时在线更新。针对上述情况,对相关性即时学习法的阈值参数进行简化,并使用仿真案例验证简化阈值参数后模型的预测精度及其应用在控制器设计中的控制效果。结果表明,简化阈值参数后的Co-JITL方法虽然会损失很少的模型精度,但是也能达到较为满意的预测和控制效果,并且与原方法相比,面对新工况时无需使用试错法去调节阈值参数,这将节约大量的离线优化时间,使该方法更加简单实用。
When correlation-based just-in-time learning(Co-JITL)method is used to establish a model for a chemical process,multiple parameters need to be adjusted manually in different operating conditions,which makes the model inconvenient to use and easily fall into the over-fitting.It is not conductive to the timely online update of the model.In order to solve the above problems,the threshold parameters of the Co-JITL method are simplified.The prediction accuracy of the model after simplifying the parameters and the control effect of its application in the controller design are verified by simulation.The results show that the model established by two Co-JITL methods with simplified threshold parameters can achieve satisfactory prediction and control performance although it loses little accuracy.Compared with the original Co-JITL method,the proposed method does not need to use the trial-and-error method to adjust the threshold parameters in new operating conditions,which saves a lot of offline optimization time and makes the Co-JITL method more simple and practical.
作者
曹向军
杨鑫
CAOXiang-jun;YANGXin(College of Chemistry and Chemical Engineering,Chongqing University of Technology,Chongqing 400054,China)
出处
《控制工程》
CSCD
北大核心
2022年第10期1850-1856,共7页
Control Engineering of China
基金
国家自然科学基金青年科学基金资助项目(21306234)。
关键词
基于相关性即时学习法
阈值参数简化
统计分析上线
Correlation-based just-in-time learning
threshold parameter simplification
upper limit of statistical analysis
