Font Size: a A A

Research On Performance Assessment Of Model Predictive Control System

Posted on:2016-06-21Degree:MasterType:Thesis
Country:ChinaCandidate:Y H WenFull Text:PDF
GTID:2308330461983621Subject:Control Science and Engineering
Abstract/Summary:PDF Full Text Request
Model Predictive Control is in the late 1970 s, after the Control algorithm development and produce a new algorithm based on computer Control.Can be understood as the basic principle of the MPC in the limited time domain, by minimizing the objective function method, to calculate the control action of the next step.The three elements of model predictive control for the model prediction, rolling optimization and feedback control, and be able to handle the constraints and the coupling.Be Advanced Control of the main Control strategy.MPC controller is widely applied to industrial process, but in the real industrial site MPC controller design.Based on the field running, performance indicators and not the people.Disturbance by external factors, the change of unpredictable factors, there is no regular maintenance, will cause the loss of control performance, therefore to carry out the control of the MPC controller performance evaluation is of profound significance.Background in this paper, based on the above topic for model predictive control system research of performance evaluation, the concrete research content includes the following aspects:Model predictive control and the controller performance assessment and the background of the topic, the basic principle and development process based on. To carry out the work, according to the dynamic matrix model predictive control thinking, in-depth analysis of single variable and multi-variable control system, and the simulation experiment was made using a single variable single tank model, DMC algorithm of model predictive control based on advanced verification. Control process for single variable system, firstly the traditional minimum variance performance evaluation algorithm based on single variable model, and integration of the specific steps in the evaluation of the minimum variance. The introduction of multi-variable system, introduces the correlation matrix for prepare knowledge, for multi-variable system two kinds of evaluation methods are proposed: the MIMO system of traditional feedback control method, the multi-variable minimum variance benchmark evaluation method based on.The thought and the minimum variance performance evaluation combined with the principle of rolling time-domain model predictive controller, designed for minimum variance controller of rolling time-domain thought, and carries on the simulation experiment for the tank model of single variable and multivariable distillation column model, the simulation verifies the performance evaluation method than the traditional minimum variance evaluation more practical significance, and validated. Theeffectiveness of this evaluation method. Proposed generalized based on the idea of minimum variance(GMVC) on the performance of MPC controller, control method, detailed and specific steps of evaluation method, comparative analysis of the advantages and disadvantages of the GMVC benchmark and the MVC benchmark data curve.According to the model predictive control, evaluation of the use of historical data in statistics based methods,combined with the historical performance index and design of performance index of MPC controller online monitoring data acquisition and analysis. According to the model mismatch, the unmeasurable noise interference model, using the method of residual CUSUM control chart to monitor variables, self related phenomenon is serious,smaller, and makes a classification of the reasons for the decline because of its performance through statistical control chart. Through laboratory methanol refined semi physical simulation model, the distillation column model for multi-variable system simulation, through the analysis of the data and the simulation curve, verify the feasibility and effectiveness of the proposed method.
Keywords/Search Tags:Model Predictive Control, Performance Assessment, Minimum Variance, Historical performance benchmark, CUSUM chart
PDF Full Text Request
Related items