时变过程在线辨识的即时递推核学习方法的研究.pdfVIP

时变过程在线辨识的即时递推核学习方法研究 1 2 1 刘毅 ，金福江 ，高增梁 1. 浙江工业大学过程装备及其再制造教育部工程研究中心化工机械设计研究所, 杭州310032 E-mail: yliuzju@ 2. 华侨大学信息科学与工程学院, 厦门361021 E-mail: jinfujiang@163.com 摘 要：为了及时跟踪非线性化工过程的时变特性，提出即时递推核学习的在线辨识方法。针对待预测的新样 本点，采用即时学习策略，通过构造累积相似度因子，选择与其相似的样本集建立核学习辨识模型。为避免传 统即时学习对每个待预测点都重新建模的繁琐，利用两个临近时刻相似样本集的异同点，采用递推方法有效添 加新样本，并删减旧样本，以快速建立新即时模型。通过一时变连续搅拌釜式反应过程的在线辨识，表明了所 提出方法在保证计算效率的同时，较传统递推核学习方法提高了辨识的准确程度，能更好的辨识时变过程。 关键词：过程辨识，即时学习，核学习，最小二乘支持向量回归，递推辨识 Online identification of time-varying processes using just-in-time recursive kernel learning approach 1 2 1 Yi Liu , Fujiang Jin , Zengliang Gao 1. Engineering Research Center of Process Equipment and its Re-Manufacturing, Ministry of Education, Institute of Process Equipment and Control Engineering, Zhejiang University of Technology, Hangzhou, 310032 E-mail: yliuzju@ 2. School of Information Science and Engineering, Huaqiao University, Xiamen 361021 E-mail: jinfujiang@163.com Abstract: An online identification method using just-in-time recursive kernel learning is proposed to trace the time-varying characteristics of nonlinear chemical processes. For each query sample, a just-in-time kernel learning model is established using the similar set constructed by a presented cumulative similarity factor. Different from traditional just -in-time learning approaches discarding their models at eac