A New Dynamic Matrix Control Algorithm Based On Feedback Correction By Curve Fitting

Predictive control was applied broadly in various engineering field, which was acquired from industry practicing in 70s. Internal model control (IMC) designed to analyze predictive control is developed quickly since 80s, and produced many design methods such as pole-zero cancellation, H2-optimal control, deadbeat control and so on. At the same time, the algorithms of predictive control were developed too. In these algorithms, dynamic matrix control (DMC) was researched and applied actively, which was based on non-parameter step response.For summarizing the merits and limitations of pole-zero cancellation, H2-optimal control and dynamic matrix control in application, this paper used a non-minimum phase second-order plant. In common premise by using a same feedback controller, the feed-forward controller was designed differently by the three design methods to compare the dynamic performance, robustness and disturbance-rejection.DMC is an important algorithm of predictive control. It's simple, strong robustness and broad using. But, it's very hard to compromise its dynamic and stable performances at the same time. This paper brought forward a new algorithm to realize feedback. That is, using the modelerrors of past time to predict the future model errors by curve fitting and weighting to the model errors to get the future predictive model outputs. The dynamic performance was proposed without affecting the system's stable performance,In the procedure of research, the declined performance of changed-rapidly system's robustness was found. The reason was using the fixed weight. So the system error and the non-matched model rate were introduced to tune the weight online piecewisely. As a result, the system performance was improved. Simulation results showed that the new algorithm is effective.