Improved Differential Evolution Algorithm Based on Mutation Strategy of Tournament Selection for Function Optimization

The traditional mutation strategy of differential evolution algorithm can not reach a good tradeoff between robustness in global convergence and the search efficiency.The operators are constants,and the differential evolution algorithm leads to many problems,such as the low search efficiency and the premature convergence.Based on analysis of performance of the mutation strategies,a new mutation strategy with tournament selection rule taking the best individual vector from the random individual vectors as the base vector was proposed in this paper.Meanwhile,selecting the direction for the difference vector beneficial to search and making strengthen on the difference vectors is to improve the convergence rate and maintain the diversity of population.The random normal scaling factor F and the time-varying crossover probability factor CR are used synchronously to advance the local search and global search.Finally,the improved differential evolution algorithm was tested on four benchmark functions.The simulation results show that the improved algorithm can effectively avoid the premature convergence,as well as improve the global convergence ability and the search efficiency remarkably.