Study Of Complex Systems For Multiple Modeling And Predictive Control

In the actual process of industrial production, the controlled object itself or by externalinfluence, there is often a lot of uncertainties, such as: the controlled system is a subsystem ordynamic changes faster parameter nonlinear systems, there is a big external disturbances orrandom system, the system parameters of linear systems or large, complex stochastic systems;systems, sensor or actuator failure. For such systems, when the conventional identificationalgorithms do not follow the changes in the real model parameters, therefore, the system isdifficult to fully describe the characteristics of conventional identification algorithm, makingrelative to the actual industrial process, model-based controller design It is difficult to meetthe accuracy requirements of control. To solve these problems, multi-model approach opensup a new direction, is an effective way to deal with complex systems. Methods of complexnonlinear system is decomposed into a number of simple linear systems, each solving a linearsystem decomposition and synthesis rules to get through some of the original complexsystems modeling and good control effect. Both in theoretical research and practical industrialapplications, since its been proposed, have achieved great results.Although the multi-model approach to the control of complex systems has a good effect,but in itself there is still room for improvement, including: how to select and optimize themodel set, how to ensure the stability of the model when switching on the uncertain existenceof random interference how the system control. For existing multi-model approach problemshave been studied in this paper, the main research work is as follows:(1) In the use of multi-model decomposition-synthetic strategies for complex nonlinearsystem modeling, the identification of system parameters based on fuzzy identification. Thedata from the input-output system starting by collecting data samples can fully exploit the useof inter-related information data clustering algorithm to be reasonably divided into multiplesub-space features a clear set of TS fuzzy model form set, then these TS fuzzy fuzzy modelidentification, get the model structure and parameters of each sub-space, the use ofsub-models to characterize the resulting fuzzy identification of nonlinear characteristics of the system.However, slow convergence and easy to fall into local optimal solution is the traditionalclustering algorithms biggest flaw, hindering the development of clustering algorithms. Catsswarm behavior by imitating the cat out of the evolution of intelligent algorithms, the cat’sbehavior into search mode and tracking mode, which greatly improves the clusteringconvergence speed and global search capability. By cat swarm of complex nonlinear system isdivided into sub-space, multi-form model set TS fuzzy. In the course of TS fuzzy modelidentification, the cats algorithms and applications to fuzzy identification process, theformation of cat identification method based on fuzzy clustering algorithm performed for eachTS fuzzy model identification, and finally out of the fuzzy multi-model identification set tocharacterize the nonlinear system. Simulation results for a nonlinear system modeling fuzzymulti indicating the practicality of the method.(2) A class of model parameters for the system over time mutations, through the use ofthe controlled system identification method based on fuzzy clustering class constructor cat,identified multiple model (model set) to represent, covering its uncertainty; modelrespectively corresponding concentration of each model predictive controller design. By fuzzyidentification model set by the fixed structure of a conventional and an adaptive model thatcan be reassigned parallel identification of the controlled system dynamics. At each samplingtime, the controller selects the best performance (switching) function to complete. Meanwhilestepped generalized predictive controller design to achieve global control system. Finally,through the design of the corresponding simulation results show that, compared with the othertwo control methods (single adaptive model control, multiple fixed model control) controlmethod described in this article better.