Model-plant Mismatch Detection of a Nonlinear Industrial Circulation Fluidized Bed Boiler Using LPV Model

摘要

The performance of the model-based controller depends on the accuracy of the model it used. Since most industrial systems are nonlinear, the process model changes with the operating points, which increase the difficulty of the model-plant mismatch detection. A model-plant mismatch detection method for nonlinear systems based on multi-model linear parameter varying (LPV) model is proposed in this work. Firstly, the mismatch of the local linear models at each of the operating points is diagnosed by analyzing the partial correlation coefficients between the model residuals and the process inputs. Then the local linear models with significant mismatch are re-identified and the LPV model is obtained by re-estimating the model weightings. In order to verify the effectiveness of the proposed methodology, the mismatch of the LPV model of a nonlinear industrial circulating fluidized bed (CFB) boiler is diagnosed. The results show that the proposed method can locate the mismatch linear models correctly, and the accuracy of the re-identified LPV model is significantly improved with respect to the best fit percentage of the process outputs.