Elaborate Modeling And Exact Solution For LPV Model Nonlinear Predictive Control

In industrial process, the model predictive control (MPC) is one of the most successful advanced control strategies. For the system with wide setpoint ranges, strong nonlinearities and wide operating ranges, it is necessary to adopt the nonlinear model predictive control (NMPC) to meet the quality requirements. In order to improve the control performance, the design of controller always requires an accurate nonlinear model. The method of multiple models is a simple and effective modeling method in engineering. In this paper, linear varying parameters (LPV) model which is one of the methods of multiple models is selected as the predictive model. However, the traditional operating point setting for LPV model is basically experiential, artificial and lack of systematic approach guidance. Moreover, even though with the accurate nonlinear model, the design of NMPC controller is also very complex and the online computation is huge. From two aspects of model and optimization algorithm, an elaborate Modeling Method based on gap metric and NMPC based on the interior point method are respectively proposed for NMPC of LPV model. The contributions of this thesis include the following three aspects.1. In order to overcome the randomness of selecting the working point for LPV model. A method that building LPV model based on gap metric is proposed. This approach uses gap metric as the nonlinear measurement tool to analyze the nonlinear degree of system quantificationally, select the appropriate control strategy, and then decompose system in the operating space and set up reasonable working-points.2. In view of the online rolling optimization calculation is huge and the partition rule for operating area will directly influence the control performance and real-time. Comparison the LPV model by setting up reasonable working-points with the LPV model by selecting working-point randomly in open loop simulation and closed-loop response, verifies the effectiveness of building LPV model based on gap metric with the LV model of the distillation column proposed by Skogestad. The result show that the low order LPV model obtained based on gap metric can more accurately approximate real system and also speed up the solution.3. The optimization algorithm for solving the NMPC of LPV model is usually the multi-step linearization method, which needs to solve multiple QP optimization problems and the corresponding computational time is very long. The interior point algorithm is superior for solving large-scale optimization problems. NMPC based on the interior point algorithm is proposed. This method doesn’t need linearization and can solve optimization problems of LPV model directly in the frame of simultaneous approach. It has high precision. The example of continuous stirred tank reactor verifies that it can shorten the transition time, decrease the relative energy consumption. Furthermore, it’s advantages is more obvious under wide operating ranges.