Multimodel Nonlinear Predictive Control with Gaussian Process Model
In industrial practice, parameters of plants are often not fixed in production process. The variation of parameters gives rise to the variation of system model. With model-based predictive control, the plants will be out of control if a fixed predictive model is applied when the parameters of plants change frequently. This paper proposed a multimodel nonlinear predictive control based on Gaussian process models which can be applied to the nonlinear system control with varying parameters. On the basis of the management ability of Gaussian process in fitting nonlinear model, a feasible switch strategy based on deviation of the predictive output from the actual output was applied to identify change of parameters. A two order dynamical system with four models switch was demonstrated to illustrate this algorithm. According to simulation results, this switch strategy can identify the change of system model accurately and quickly.
KeywordsSupport Vector Machine Model Predictive Control Reference Trajectory Markov Chain Monte Carlo Method Switch Strategy
- 1.Qin SJ, Badgwell TA (1997) An overview of industrial model predictive control technology. American Institute of Chemical Engineers, New YorkGoogle Scholar
- 2.Maciejowski JM (2000) Predictive control with constraints. Prentice Hall, LondonGoogle Scholar
- 3.Qin SJ, Badgwell TA (2000) Nonlinear model predictive control: an overview of nonlinear model predictive control applications. Prog Syst Control Theory 26:369–392Google Scholar
- 7.Kocijan J, Murray-Smith R (2005) Nonlinear predictive control with a Gaussian process model. Springer Science+Business Media, BerlinGoogle Scholar
- 10.Bao ZJ, Sun YX (2008) Support vector machine-based multi-model predictive control. J Control Theory Appl 6:305–310Google Scholar