Font Size: a A A

Model Predictive Control Based On SVR Optimized By Multi-agent Particle Swarm Optimization Algorithm

Posted on:2015-03-07Degree:MasterType:Thesis
Country:ChinaCandidate:Y LiFull Text:PDF
GTID:2298330422983073Subject:Control theory and control engineering
Abstract/Summary:PDF Full Text Request
With the progress of science technology and the development of industry, higherand higher requirements have been focused on model predictive control (MPC).Although Model predictive control theory has been more than30years, the traditionalcontrol strategies are very difficult to meet the demand of nonlinear systems. Thus MPCfor nonlinear systems has become a hotspot of research.For prediction model and parameter optimization of MPC, this thesis is devoted topredictive control strategies for nonlinear systems. The main research works are asfollows.(1) Multi-agent particle swarm optimization algorithm (MAPSO) is proposed bycombining particle swarm optimization (PSO) with the concepts f multi Agent system,and is applied to parameter optimization of model predictive control. On the basis of theiterative optimize principle of particle swarm algorithm, the MAPSO makes good use ofthe multi-agent system the concept of environment and competition and cooperationmechanism. The particle of MAPSO takes corresponding measures in accordance withthe environmental changes.Thus the MAPSO is quick and effective solution toparameter optimization.(2) As Support vector regression (SVR) algorithm is a very efficient technique fornonlinear function approximation, SVR is devoted to prediction model of MPC. Theprediction accuracy and generalization ability of the support vectorregression(SVR)model depends on a proper setting of its parameters to a great extent.,An optimal selection approach of SVR parameters is put forward based on MAPSOalgorithm. On this basis, a model predictive control method based on the MAPSO-SVRis proposed and applied to single-step nonlinear predictive control scheme to select theoptimal control inputs. For nonlinear system, the simulation results show that theproposed method is effective and has an excellent adaptive ability and robustness.Compared with the model predictive controllers based on SVR optimized by particleswarm optimization algorithm (PSO-SVR), SVR optimized genetic algorithm(GA-SVR), and RBF neural network algorithm, the proposed method is superior toother methods.(3) An optimal selection approach of support vector regression (SVR) parametersis proposed based on MAPSO. On the basis of single-step model predictive control, Amulti-step model predictive controller based on the support vector regression to predict nonlinear system is established; and the optimal parameters of which is searched byMAPSO. With objective function of rolling optimization, analytical solutions ofmulti-step predictive control laws are obtained by predictive control mechanism.Compared with the model predictive controllers based on SVR optimized by particleswarm optimization algorithm (PSO-SVR), SVR optimized genetic algorithm(GA-SVR), and RBF neural network algorithm optimized genetic algorithm (GA-RBF),the simulation results show that the proposed method has better prediction results thanothers, and is effective for nonlinear systems.
Keywords/Search Tags:model predictive control(MPC), support vector regression(SVR), multi-agent particle swarm optimization(MAPSO), nonlinear systems
PDF Full Text Request
Related items