The Performance Assessment Research And Application Of Model Predictive Control System Based On LQG Benchmark

With the development of computer technology and putting forward the concept of industrial 4.0,industrial process tends to be more intelligent and automatic.This requires a more advanced control algorithm to control the industrial process to guarantee the stability and high efficiency of the system.Thus advanced control algorithm has developed rapidly in recent years,the Model Predictive Control is the most widely used and most efficient class of advanced control algorithms.Model Predictive Control is a combination of predictive model,roll optimization and feedback control and can make the system achieve the boundary,greatly improving the efficiency of the industry.But in the practical industrial application,due to insufficient controller setting or lack of maintenance,as well as the system factors such as equipment failures,Model Predictive Control has the phenomenon of control performance variation.If don't take timely measures,it can cause severe losses to the enterprise production.Therefore we need carry out the theory research of the MPC system control performance assessment to build the real-time performance monitoring system and achieve the goal of performance monitoring.This article is based on the Linear quadratic Gaussian benchmark as the MPC system performance assessment index.The main research content is as follows:At first,introducting briefly the purpose and significance of performance assessment of the Model Predictive Control system,respectively,the paper introduces the development and research status of the model predictive control and performance assessment of control system,and analyzes the development trend and problems of the current study.Then on account of the dynamic matrix control algorithm of Model Predictive Control,putting the predictive model,roll optimization and feedback correction as main aspects,respectively algorithm is derived in univariate and multivariate system,and choosing the single let water tank and rectified tower as the object of simulation.By analyzing the simulation results and comparing with the conventional PID control,fully verifies the superiority of the MPC control algorithm.And then introducing theory research of LQG benchmark,putting performance limit curve as the ultimate goal,respectively deducing the three different algorithms: based on state space model algorithm,based on the polynomial model algorithm and the algorithm based on subspace model is compared,analyzing their respective advantages and disadvantages and applicable scope.Finally based on the traditional assessment benchmark,applying the LQG benchmark scheme to the MPC system,it can take into account the limitation of input conditions and get more meaningful performance indicators,providing more realistic theoretical basis for the MPC system for performance assessment.To solve the problem that consistency estimates of the MPC system is difficult to obtain,on the basis of the subspace algorithm,using combinedinput/output closed-loop subspace identification algorithm as the theoretical basis,by minimizing the objective function to obtain LQG benchmark.Combining roll optimization of MPC with LQG benchmark theory to designed the LQG controller,and verify the effectiveness of the proposed method in the simulation experiment of the rectified tower.