Modeling Based On Support Vector Machine And Application Of Predictive Control Technology

Model predictive control (MPC) refer to a class of control algorithms that optimize the current input by predicting the future behavior of system based on system dynamic model. Based on linear predictive model, linear predictive control has been matured theoretically, and many successful applications can be found in industrial process control. However, for the highly nonlinear processes, nonlinear model and the corresponding predictive control strategies have to be employed since linear models cannot accurately describe the system characteristics.This thesis focuses on the research of algorithm experimental and engineering application. On the simulation platform in the virtual plant lab, different processes are constructed and nonlinear predictive control strategies by minimizing the least square support vector machine are presented for a kind of nonlinear process objects. In industrial application, for the light hydrocarbon fractionating device that has complicated process and disturbances and long-ranged requirement, advanced control system mainly employing linear model predictive technique is designed, which produces good performance.The main contents are as follows:1. Several process objects and related model predictive control algorithms are constructed in the form of physical and digital simulation, in the virtual factory lab based on process industries.2. Nonlinear model predictive control algorithm based on support vector machine is presented, and nonlinear process control plant is designed on simulation platform of the virtual lab. Experiment shows the effectiveness of the presented nonlinear predictive control algorithm.3. After analyzing control state and requirement of the process of the light hydrocarbon fractionating device, advanced control system is designed according to the requirements, which gives good control performance.The last part is summary and perspective.