Research On Improved Dynamic Matrix Control

Model predictive control is a type of advanced computer optimization control algorithms. It is based on model prediction, rolling optimization with feedback correction and generated directly from industrial process control. For the mathematic model of control system is easy to acquire in industrial field, it do not need complicated system identification and precise modeling. Rolling optimization based on feedback correction is used instead of conventional optimization control in the algorithm, so the model is of low demand and it has strong robustness on the influence of model mismatch, non-minimum phase systems and uncertainty interference. Since online calculation of the algorithm is relatively simple, it is applicable to digital computer control. The basic principle of the algorithm is to expand single step prediction of traditional self-tuning technology to multi steps prediction, so the sensitivity of algorithm to the change of model parameters is retained effectively. The typical algorithms of model predictive control are model algorithmic control (MAC), dynamic matrix control (DMC), generalized predictive control (GPC). They all based on model predictive, rolling optimization and feedback correction.Dynamic Matrix Control algorithm is one of the important representative model predictive control algorithms. Traditional autocorrection of single step prediction is expanded to multiple steps prediction. Based on the practical feedback information, repeating optimization of the algorithm restrains effectively the algorithm sensitivity to parameter change of the model. It has strong adaptability to uncertainty of model error and environmental disturbance. DMC algorithm is designed on a low order linear approximate model, but complicated industrial processes have many uncertainties on the aspects of model order, nonlinear, environmental disturbance and time delay. So, the research of improved DMC algorithm and the discussion of parameter design to the influence of robust performance have important theoretical significance and practical value of application. In this paper, two improved dynamic matrix control algorithms are proposed.Traditional feedback control algorithm—PID control, which is combined proportional, integral and derivative of feedback system deviation and is simple in principle, easy to understand and implement in engineering, is still one of the most extensive application of conventional control algorithm in industrial process control. But conventional PID control algorithm can not achieve ideal control effect in some practical production process with nonlinear and time varying uncertainty. Based on the analysis of PID control algorithm and dynamic matrix control algorithm, the derivation of proportional integral derivative dynamic matrix control (PIDDMC) algorithm is given. Combined the virtues of the two algorithms, this kind of algorithm is of the advantages of PID algorithm, which is simple in structure, reliable in operation and robust in performance, and of predictive function. The influence of PIDDMC controller parameter tuning on system performance is investigated in time domain. The parameter election range is researched. Compared simulation results, PIDDMC algorithm has better performance than conventional DMC algorithm to make the system have satisfactory stability and transition characteristics.Most industrial processes have characteristic of time delay. The communication delay introduced by the networks can lead to performance degradation and even instability. To address network-induced delays, a time-stamped dynamic matrix control (TSDMC) algorithm which uses a communication delay model to improve reliability over network control systems is proposed. The network-induced delays are measured by a time-stamp method, based on which the step response coefficients and control coefficients ate corrected in each sampling period and the algorithm derivation is given. Simulative validation of this new algorithm resulted in improved performance and stability over traditional dynamic matrix control.