A process transfer model-based optimal compensation control strategy for batch process using just-in-time learning and trust region method |
| |
| Authors: | Fei Chu Xiang Cheng Chuang Peng Runda Jia Tao Chen Qinglai Wei |
| |
| Institution: | 1. Underground Space Intelligent Control Engineering Research Center of the Ministry of Education, School of Information and Control Engineering, China University of Mining and Technology, Xuzhou 221116, China;2. State Key Laboratory of Process Automation in Mining & Metallurgy/Beijing Key Laboratory of Process Automation in Mining & Metallurgy, Beijing General Research Institute of Mining & Metallurgy, Beijing 100160, China;3. State Key Laboratory of Management and Control for Complex Systems, Institute of Automation, Chinese Academy of Sciences, Beijing 100190, China;4. School of Information Science & Engineering, Northeastern University, Shenyang 110004, China;5. Department of Chemical and Process Engineering, University of Surrey, Guildford, United Kingdom |
| |
| Abstract: | The advantages of maximally transferring similar process data for modeling make the process transfer model attract increasing attention in quality prediction and optimal control. Unfortunately, due to the difference between similar processes and the uncertainty of data-driven model, there are usually a more serious mismatch between the process transfer model and the actual process, which may result in the deterioration of process transfer model-based control strategies. In this research, a process transfer model based optimal compensation control strategy using just-in-time learning and trust region method is proposed to cope with this problem for batch processes. First, a novel JITL-JYKPLS (Just-in-time learning Joint-Y kernel partial least squares) model combining the JYKPLS (Joint-Y kernel partial least squares) process transfer model and just-in-time learning is proposed and employed to obtain the satisfactory approximation in a local region with the assistance of sufficient similar process data. Then, this paper integrates JITL-JYKPLS model with the trust region method to further compensate for the NCO (necessary condition of optimality) mismatch in the batch-to-batch optimization problem, and the problem of estimating experimental gradients is also avoided. Meanwhile, a more elaborate model update scheme is designed to supplement the lack of new data and gradually eliminate the adverse effects of partial differences between similar process production processes. Finally, the feasibility of the proposed optimal compensation control strategy is demonstrated through a simulated cobalt oxalate synthesis process. |
| |
| Keywords: | |
| 本文献已被 ScienceDirect 等数据库收录！ |
|
