Recursive Identification Algorithm And Its Application On Model Predictive Control

With the application of model predictive control in industry more and more widely, how to improve the performance of controller in the background of complex industry system gets also more and more attention. There are plenty of theories and practical experiences in predictive control and system identification, but in actual production, the on-line control system all have their own problems and space to be improved. In this paper, we study the recursive identification and a disturbance adaptive predictive control algorithm, and put forward some new ideas. The specific work is as follows:1. Recursive identification of ARMAX based on multi-iteration. A recursive identification algorithm using the Gauss Newton method is derived. After analyzing the difference of the parameter estimation accuracy between the off-line algorithm and the recursive algorithm, the idea of data multi iteration is put forward. Simulation results show that the algorithm based on multi-iteration is superior to the original recursive algorithm in both convergence speed and precision of the parameters.2. Modeling and predicting ability of two time series models. This article begins from the derivation of common AR and ARMA model recursive estimation, and idea of multi-iteration is applied to ARMA model identification to improve the tracking capacity. Multi-step prediction studies of the two models found that the parameters of the ARMA model are more accurate than the AR model, but the robustness of algorithm is not as good as the latter.3. Analysis, improvement and parameter design of disturbance adaptive predictive controller. A predictive control algorithm based on disturbance model DMCA is introduced, which makes the controller more powerful. We analyzed the reason why the ARMA model based on the DMCA algorithm is not as stable as original, and put forward two improved methods:a DMCA algorithm based on AR model and adding a small perturbation to disturbance model estimation data. Finally, the parameter tuning method of the disturbance adaptive controller is studied, which provided some guidelines and reference for the controller parameters optimization.