Differential Evolution Algorithm Design Research

The differential evolution algorithm to search the global optimal solution of the stochastic optimization algorithm is an analog of the natural biological evolution. Because the differential evolution algorithm is simple and easy to implement less control parameters and search ability has been extensive research and application. And other population-based evolutionary algorithm, differential evolution algorithm there are also some shortcomings, mainly in the lack of premature convergence and local search ability. Because the differential evolution algorithm is simple and less control parameters reasonable parameters have a great impact on the solution performance of the algorithm. To solve these problems, this paper presents the improved method. And use some of the classic test functions of numerical experiments to study the application of differential evolution algorithm for optimization problems with equality constraints. The main results are as follows:For optimization problems with equality constraints this paper proposed a new differential evolution algorithm for equality constrained optimization problems. First of all, according to the dimension of the search space and equality constraints to determine the parameters and solving the parameter equation; equation with parameters in the initial population, crossover and mutation operation, update the individual to ensure that the individual meets the equality constraints for breach of inequality constraints individual penalty function for punishment. The numerical experiments verify that the new algorithm has a faster convergence speed and search for the optimal solution to the constrained optimization problems, but also to the greater probability.The optimization algorithm has the characteristics of strong local search ability, so the traditional optimization operator embedded in the DE algorithm is an effective way to improve the local search ability. Which will climb operator reached embedded DE algorithm to improve the local search ability of good results, but this method is more complex test function and algorithms later stage of evolution is still very difficult to play a role. Because the various stages of the algorithm in solving different test function and evolution need to choose a different control parameter values to ensure that the search capabilities of the algorithm. This paper presents a new adaptive hiking radius-based differential evolution algorithm for this problem. This algorithm is dynamically adjusted according to the individual search efficiency of the information obtained in the search process, mountain climbing radius, to improve the adaptability of the algorithm in the various stages of evolution. The numerical results show that the new algorithm improves the efficiency of climbing operator has a faster convergence speed and higher solution accuracy.