Research On Modified Predictive Control Methods Based On Real Coded Extremal Optimization And Their Applications

Predictive control methods have been widely applied to various complex industrial process control systems, but most of existing algorithms rely heavily on experiences of designers. Consequently, how to adopt evolutionary algorithms to modify the rolling optimization strategy of traditional predictive control methods in order to further improve the ability of solving the constrained optimization control problems efficiently is of great theoretical significance and engineering practice, so it has been one of research hotspots in academic and industrial fields. Moreover, the applications of predictive control in power converters and multi-area interconnected power systems have attracted increasing attentions by the researchers from electrical engineering field,but it is still its infancy. Therefore, this thesis focuses on the modified versions of predictive control and their applications in typical process control systems and multi-area interconnected power systems from the perspectives of real-coded population external optimization.The main research works and novel ideas of this thesis are summarized as follows:(1). A novel real-coded population extremal optimization (RCEO)method with multi-non-uniform mutation operation is presented. The simulation studies on a variety of continuous optimization well-known benchmark test problems and the optimal parameters design problem of fractional-order PID controller for automatic voltage control systems have demonstrated that the proposed RCEO method is superior to other evolutionary algorithms such as real-coded genetic algorithm (GA),particle swarm optimization (PSO) and other reported modified extremal optimization (EO) algorithms in terms of fewer adjustable parameters and better optimization performances.(2). Based on the above research work (1), this thesis presents two modified predictive control methods including RCEO-based constrained generalized predictive control method called CGPC-RCEO and RCEO-based constrained dynamic matrix control method called CDMC-RCEO. The respective simulation studies on the temperature control systems of recalculated water in heat exchanger and continuous stirred tank reactor have shown that the proposed CGPC-RCEO and CDMC-RCEO obtain better control performances such as system output response and control increment response than the traditional predictive control alogorithms, the corresponding GA and PSO based predictive control methods, respectively.(3). By extending the basic idea behind the proposed CDMC-RCEO to multi-area interconnected power systems on the basis of the research work (2), a novel CDMC-RCEO-based distributed load-frequency predictive control method is proposed for multi-area interconnected power systems. Its superiority to traditional integrator,proportional-integral controller, CGPC-RCEO, GA-based and PSO-based CDMC methods is demonstrated by the simulation results on two-area and three-area interconnected power systems in terms of better dynamic-state, steady-state and robust performances.