多步预测误差方法在先控性能评估中的应用  

Application of Multi-step Prediction in Advanced Control Performance Assessment

作　　者：刘桂枝[1] 刘红梅[1] 陈爱军[1] LIU Gui-zhi LIU Hong-mei CHEN Ai-jun(School of Physics and Electronic Science, Shanxi Datong University, Datong Shanxi, 03700)

机构地区：[1]山西大同大学物理与电子科学学院,山西大同037009

出　　处：《山西大同大学学报（自然科学版）》2016年第4期28-32,共5页Journal of Shanxi Datong University(Natural Science Edition)

基　　金：国家青年自然科学基金[61307121];山西大同大学科研项目[2014K4];山西大同大学教改项目[XJY2013213]

摘　　要：本文在基于最优预报的基础上提出了一种新的MPC性能评估方法。通过将已有的基于k步预测误差的控制系统性能评估方法进行拓展,得到了一种新的基于多步最优预测误差的先进控制系统的性能评估方法。最后提出了一个新的基于多步预测误差的性能指标用来对先进控制系统的控制性能进行评估和分析。该方法不需要关于过程的任何先验信息,因此便于在实际工程中实施和应用。将所提出的方法应用在某石化企业的一套先进控制装置的性能评估当中,验证了该方法的有效性。This paper is concerned with the performance assessment of MPC systems with the prediction error approach. By extending the k-step prediction approach proposed in the literature, a multi-step prediction error approach is developed based on the multi-step optimal filter. Consequently, a novel closed-loop potential index is proposed based on the multi-step prediction error approach for the performance assessment of MPC systems. The proposed approach does not require any a priori information of process such as process model, controller structure etc., and is easily implemented in the practical industrial applications. The utility of the proposed approach is demonstrated by industrial application case study.

关 键 词：先进过程控制 控制性能评估 多步预报 模型预测控制 

分 类 号：X522[环境科学与工程—环境工程]