Predictive Functional Control Based On Characteristic Models And It's Application Research

In the past few decade, the research of model predictive control has gained more and more attention among many experts. With everyone's effort, many critical results are constantly spoken out. Model predictive control algorithm is obsessed of some advantages like good robustness, flexible ability of constraint solution, relatively higher comprehensive controlling quality, exclusively being adapted to industry process which has input-output constraint, time-delay and time-varying characteristic, reverse characteristic and variable target function, so it is widely accepted by industrial fields and practical results are emerging in an endless stream.As the 3rd generation of model predict control, predictive functional control maintains the advantage of model predict control, meanwhile, it strengths regularity of input controlled quantity, improved the speediness and accurateness of response by introducing the concept of base functions, which can effectively decrease calculation quantity of algorithm. However, same as other model predict control, once predict model happens not in line with actual controlled process, it would meet with degrade of controlling performance. Although a series of model mismatch research have been carried out, for example, introducing neural networks model and fuzzy model into predictive function control, this kind of operation can only make predictive model becoming more complicated, increasing online calculation quantity and depriving of all the original merits of predictive functional control like easy algorithm and fast calculation.Considering model mismatch problem of predictive functional control, this paper sets the assurance of predictive function control obsessing the strong points of easy algorithm and fast calculation as the starting line of research. Moreover, on the premise of decreasing model mismatch in order to assure maintaining the control characteristic of predictive functional control, this paper raises control scheme of predictive functional control on the basis of the characteristic model.The usual form of the characteristic model is a second-order slow time- varying linear model, which is built up under the combined requirements of controlled object dynamic features and control feature. The main features of characteristic model is that it has simple model form and easily practiced, besides, under same input control, object characteristic model and practical object are equivalent.The concept of characteristic model is abstracted by Hongxin Wu academician through his long years of theoretic research and project practice. This paper developed a further study on characteristic model theory, firstly, it concluded out several methods of acquiring characteristic models, especially the method on the basis of testing and emulation analysis with the help of Matlab. With these methods we can conveniently obtain characteristic model and realize model reduction. Secondly, in order to escape the calculation problem of time-varying parameters of characteristic model, this paper introduced in a concept of parameter zone and equivalent second-order slow time- varying to a cluster of second- order linear models thus it can assure when introducing characteristic model into predictive functional control, traditional predictive function control would not lose their advantages of easy and fast calculation. The result of simulation reflects that new algorithm is obviously better than traditional one. As the introduction of parameter zone leads characteristic polynomial parameter based on new algorithm having uncertainty parameter problem, in order to carrying out stability analysis, this paper utilized polynomial stability analysis theory to find out a method being good for stability analysis --- polynomial stability analysis theory. With the help of DCS platform, we did research under an experimental environment which fits to production control, experiment research further manifests that new algorithm has better controlling features over traditional one. The adopted DCS in experiment is from SUPCON JX-300X system, control program was made from SCX language provided by JX-300X, under the help of JX-300X platform we can quite easily plant control procedure into industrial production.The main contents of this paper are like following:1. Research on acquirement of characteristic model, introduction of some new methods of obtaining characteristic model, including the one of non-linear, especially taking advantage of testing method and simulation analysis method under the help of Matlab. The parameter zone is introduced to further simplify characteristic model.2. Introduced predictive function control algorithm based on characteristic model. New algorithm can conquer the effective control problem when model mismatch happened. With the help of Matlab, we did simulation research on many kinds of controlled objects by traditional algorithm and new one, the simulation results shows that new algorithm is obviously better than traditional one.3. Some methods analyzing on system stability of predictive function control algorithm , which model match and model mismatch, were given. Especially adopting polynomial stability analysis method to analyzing on system stability of predictive function control based on characteristic model algorithm.4. Carried on research in a production control oriented experimental environment with the help of DCS platform, experimental research further presents that new algorithm obsesses better control features then traditional one. The adopted DCS in experiment is from SUPCON JX-300X system, control procedure was made from SCX language provided by JX-300X, under the help of JX-300X platform we can quite easily plant control procedure into industrial production.