Study Of Optimal Control Based On Genetic Algorithms

This paper studys the development of GA . the improvement of GA and the application of GA in the control system. The topics discussed include GA design GA improvement, optimal design of parameters of classical controller, simultaneous optimal design of the structure and the parameters of digital controller and emulation of on-line design. Chaper I summarizes the current study state of optimal control based on Genetic Algorithm, brings forward the science idea of the paper, and points out the significance and application foreground of the study. Chaper 2 introduces and analyses canonical Genetic Algorithm and its main influence components. Since being proposed by Professor Holland, the improvement to GA emerge in endlessly, the content involves coding strategy operation procedure~ selection mechanism~ self-adaptive crossover and mutation rate parallelism and implementing in the combination with other methods, etc., and most of the work on self-adaptive crossover and mutation rate aim at classical binary coding. Though there are so many study, unfortunately, when dealing with a practical application problem, there is no general feasible algorithm, and one has to design a algorithm which is well connected to the specific characteristics of the task at hand. Optimal design of the parameters of the classical controller belongs to the optimal problem of the continuous parameters, thus, using real number coding will get a better performance. In chapter 3, the GA used in our study is discussed. In chapter 4, a new adaptive crossover and mutation rate designed for real number coding GA is presented, and is validated in the numerical example. Chapter 4 studys the optimization of the construct and the parameters of controller comprehensively when the model of the controlled object is given. Generally, there are more than one performance target in a control system, thus, the design belongs to multi-target optimal task and the route of computing integrate performance value turn into the key problem which determines effect-, efficiency and the applicability of the design method. Solving state equation to obtain the value of time region performance index of the control system, then construct fuzzy integrate performance function is one of the methods, this needs to workout program to solve the differential equations and other assistant program of the system, the work is somewhat boring and the method has no currency. Another route to obtain the value of time region performance index is to use SIMULINK emulation software, the advantage of this route is breaking away from the fussy task of solving differential equation, but the integrate performance function remains having to be designed. In the latter part of the thesis, a simple convenient and implemental method is established, which insert the M document directly into the SIMULINK emulation block figure, thus the integrate performance value of the system can get directly form the emulation block figure, the soft interface of GA and emulation program is successfully developed, and the final design effect is proved to be excellent. Another main content of Chapter 4 is the study of simultaneous optimal design of the structure and the parameters of digital controller. As the problem comes down to the structure and the parameters of controller simultaneously, the coding method is different from the ordinary one. Thus, a mix coding of binary and real number is used, the binary and the real number represents the structure and the parameter value of the controller respectively. The operation of HGA is designed next. Some techniques are used to deal with t...