A model mismatch assessment method of MPC by decussation. (November 2020)

• Record Type: Journal Article
• Authors: Li, Lijuan; Lu, Luheng; Huang, Zhen; Chen, Xingyu; Yang, Shipin
• Is Part Of: ISA transactions. Volume 106(2020)
• Page Start: 51
• Page End: 60
• Publication Date: 2020-11
• Subjects: Model predictive control -- Model mismatch -- Correlation analysis -- Model quality index -- Decussation; Engineering instruments -- Periodicals; Engineering instruments; Periodicals; Electronic journals; 629.805
• DOI: 10.1016/j.isatra.2020.06.021
• Languages: English
• ISSNs: 0019-0578

Abstract

Abstract: For the model plant mismatch (MPM) assessment of MPC systems, a correlation analysis method between the input and the disturbance (CAID) is proposed and is combined with model quality index (MQI) for its application in multivariable system. By deducing the input–output correlation coefficient variation law of adding independent variables in a function, the paper defines the CAID index of assessing the MPM. To realize the exact location of sub-model mismatch in multivariable system, the synthetical disturbance variable is constructed through principle component analysis (PCA), then it is used to define the synthetical CAID index of multivariable MPC system, and finally the decussation method combining synthetical CAID and MQI is presented. The proposed method is validated in Wood-Berry distillation process and an industrial air separate process. Highlights: The input–output correlation coefficient variation law when adding independent variables in a function is deduced. The model assessment index denied by the correlation analysis between the input and the disturbance (CAID) is proposed. The synthetical disturbance variable is constructed for a MIMO system. The synthetical model assessment index is denied for a MIMO system. The discussion of CAID and MQI (model quality index) method is proposed for the sub-model assessment for a MIMO system.