Research On Modification Of Differential Evolution Algorithm Based On Control

As one kind of swarm intelligence optimization algorithms, differential evolution algorithm (DE) is always the hot research topic in the relevant field, which is aroused more attention with its simple structure, rapid convergence and strong robustness. Therefore, modifying and improving the optimization ability of DE algorithm has important theoretical and practical significance at global convergence and the speed of convergence by further analyzing and researching on the internal mechanism and evolution laws of DE.The aim of this paper is to modify and improve the optimization ability of DE based on the thoughts of control by using the feedback control and optimal control. The main control method is used with the information generated in the process of optimization. The specific research contents are as follows:Firstly, this paper introduces the basic principle and research status of differential evolution algorithm at home and abroad in detail, and analyses the process of optimization and convergence by descripting the algorithm as a control system. Secondly, DE algorithm based on the Model-free adaptive control is proposed to control the variation rate of average fitness in the feedback, which regards mutation factor as the control and the variation rate of average fitness as the set value. Dynamic population diversity is used as the index; mutation factor is also the control, then DE algorithm based on the Model-free adaptive control is proposed to control the population diversity in the feedback. Test functions are used to simulate the modified algorithm as proposed above. The simulation results show that the performance of the proposed algorithms is batter then the basic differential evolution algorithm, what’s more, each of the two kinds of proposed algorithm has its own advantage for different kinds test functions. Finally, another modified DE algorithm based on optimal control is proposed by using the thought of process optimization, which considers the factors of different characteristics of above methods and the set values of the optimization process. The new modified algorithm uses the modified rule base for track setting that adjusts the set values dynamically based on the information generated in the process of optimization. The final simulation researches show the results that the modified DE algorithm based on optimal control has further improvement in several indexes, such as the optimal value, robustness, rapid convergence and so on. All in all, the modified algorithm shows its feasibility and effectiveness...