Font Size: a A A

Study Of Disturbance Rejection Performance And Improved Strategies For Model Predictive Control

Posted on:2013-01-26Degree:MasterType:Thesis
Country:ChinaCandidate:W B WangFull Text:PDF
GTID:2218330371457823Subject:Systems Engineering
Abstract/Summary:PDF Full Text Request
Along with the increasing complexity of industry systems, quantities of complex and time-varying disturbances affect the control system performance more and more seriously, and they will also affect the product quality and the safe operation of the system. As a result, the control algorithms are expected to be able to provide better disturbance rejection performance.Model predictive control has become an advanced control algorithm with rich theories and extensive practical applications. Dynamic matrix control algorithm (DMC) is widely used in industry process nowadays, however, DMC has sluggish rejection of disturbance and it can't deal with complex time-varying disturbances properly. Therefore, disturbance rejection strategies in model predictive control are researched in this dissertation. Their disturbance rejection abilities are analyzed in time domain and frequency domain, besides, a disturbance adaptation predictive control algorithm is improved. The main research works are as follows:1. The time domain performances of three different disturbance rejection strategies in MPC are researched. Principles of disturbance rejection strategies based on step disturbance model, Kalman filtering technique and adaptive disturbance model are analyzed separately. Effectiveness of the three rejection strategies for typical disturbances in industry process such as random disturbance, periodic disturbance and crude oil switch disturbance is demonstrated by simulations. Besides, disturbance rejection performances are compared by performance evaluation algorithm based on MVC criteria, and the results show that adaptive disturbance model-based rejection strategy has better performance for these typical disturbances.2. The closed-loop structure for unconstrained MPC is studied. Then frequency domain analysis methods of disturbance rejection performance are introduced. Frequency domain indices such as sensitivity function and bandwidth for controllers of the three disturbance rejection strategies are derived. By frequency domain method, closed-loop system's response speed, robustness and the rejection extent of disturbance with different frequencies can be investigated, thus this method compensates for the limits of time-domain analysis. Frequency domain simulations show that adaptive disturbance model-based controller has larger bandwidth and better rejection effect of low-frequency disturbance, but somewhat less robustness. 3. An improved disturbance adaptation MPC algorithm is proposed. The improved algorithm attenuates high-frequency disturbance by exponentially-weighted smoothing of the prediction error signal. Both time domain and frequency domain simulations illustrate that the improved algorithm can weaken the control action's high-frequency fluctuation effectively. As a result, ability of rejecting high-frequency disturbance is improved and robustness of the system is enhanced.4. Parameter design of disturbance adaptation MPC algorithm is researched. Relations between disturbance rejection performance and controller's parameters such as orders of the disturbance model, iteration times of the recursive identification algorithm, filter gain and prediction horizon for the disturbance are analyzed with simulation examples. Then practical rules for the controller's parameter design are given.
Keywords/Search Tags:model predictive control, disturbance rejection performance, frequency analysis, low pass filter, parameters design
PDF Full Text Request
Related items