The Research And Applications Of Hybrid Differential Evolution Algorithm

Differential evolution (DE) algorithm is a novel evolutionary algorithm. DE algorithm has been widely used in constrained optimization, optimal design of fuzzy controller, neural network optimization, filter design and so on for its simple concept, fast convergence and low domain knowledge required. Compared with other evolution algorithms, DE algorithm is more effective to solve optimization problems, but there are many problems to be improved, such as theroy and practice areas. So we must keep on researching it and extend its application aspect.This article concentrates on DE algorithm and its hybrid algorithm, by the analysis of DE algorithm's unified framework, the DE algorithm is used for solving 0-1 linear programming, mixed integer nonlinear programming, constrained optimization and multi-objective optimization problem and the simulation experiments are carried out. Main contents of this article can be summarized as follows:1,A kind of Differential evolution algorithm with adaptive mutation and index increased crossover operator and a dynamic differential evolution algorithm with random mutation are given. The results showed that the two algorithms have fast convergence, high precision solution and robustness.2,Combining DE algorithm and penalty function methods, adding 0-1 integer operation in the mutation operation, given a improved differential evolution algorithm of 0-1 non-linear programming problem; then the Hopfield neural network which has better local searching ability was combined with DE algorithm to solve a class of 0/1 knapsack problem. Numerical experiments show that these two algorithms have very good results.3,Based on dual populations co-evolutionary strategy of PSO and DE, given a modified dual populations hybrid algorithm based on PSO and DE. Numerical experiments show that The hybrid algorithm can solve efficiently the parameter estimation problem and portfolio problem.4,I construct an improved differential evolution algorithm of nonlinear mixed-integer programming problems and ACO/DE co-evolutionary algorithm for solving mixed-integer programming problems. Numerical results show that the two kinds of algorithms are efficient algorithms for solving mixed integer programming problems.5,By a new constraints handling mechanism of relax feasibility rules and a linear decreasing tolerance of constraints violations to guide the individual as much as possible to the feasible region, constructed revised selection DE algorithm for solving constrained optimization problems. Numerical results show that the new algorithm is effective, versatile and robust.6,Using different variants operators produce multiple trial vectors, and then follow the Pareto relations of domination and amendments crowding distance to selection operate. Multiple trial vectors in differential evolution algorithm for multi-objective optimization is given. Numerical experiment shows the new algorithm is an effective algorithm for solving multi-objective optimization problems.In general, the hybrid and application of DE algorithm are analyzed comprehensively. Finally, whole research contents are summarized, and further research directions are indicated.