Performance Monitoring Of Model Predictive Controller Based On Subspace Identification

MPC has been increasely applied because of its advantages, so the performance monitoring of MPC is not only valuable to the theory research, but also to the engineering application of the industrial facilities, with the aspect of the control quality improvement, maintaining of efficient operation and increasing the application level of advanced process control strategies. Several performance monitoring methods for MPC controller are proposed afer reviewing some traditional technology. The main research works are as follows:Review current technology on both theory and application and propose the future needs for development. Introduce model predictive controller (MPC) based on subspace algorithm and simulate on Wood-Berry tower. Review several traditional technology and simulate them.The conventional LQG benchmark procedure requires an explicit parametric model to design a LQG benchmark. To design controller and construct performance assessment benchmark before the control action, we propose a MPC design and assessment method based on subspace identification. Just with the historic inputs and outputs, it designs the MPC and constructs the performance assessment benchmark only using the subspace matrix. As a result, the controller design and the performance assessment are realized in a unified framework. The simulation results on the Wood-Berry model validate the feasibility and effectiveness of proposed method.Because of no standard mehtod for selecting excellent data for historical-case inedx, and the traditional design-case index is stochastic with autocorrelation with the impect of stochastic noise. In order to overcome these shortcomings, an improvd design-case index is presented. Use the objective function of MPC as the benchmark, calculate the average index to assess the ensemble performance of a period and get the time series model of it, handle the residual with statistical process control to assess the system. Diagnose the fault source using the discriminant analysis based on distance. The simulation results on the Wood-Berry model validate the feasibility and effectiveness of proposed method.