A new hybrid differential evolution with simulated annealing and self-adaptive immune operation

摘要

Differential Evolution (DE) has become a very powerful tool for global continuous optimization. Many strategies have been proposed for the generation of new solutions and every strategy has its own pros and cons, so which one of them should be selected is critical for DE performance, besides being problem-dependent. In this paper, different new solution generation strategies are integrated together and the individual advantages of different generation strategies are utilized to enhance the exploring ability and/or to accelerate the convergence. Simulated annealing idea is introduced to escape from possible local optimum attraction. Clonal selection operation employs self-adaptive Gaussian hyper-mutation along each dimension to focus the exploitation on the promising areas and exerts different influences on different dimensions. Experiments show that the proposed ideas benefit the performance of the algorithm and the proposed algorithm performs comprehensively better than other DE variants in terms of convergence stability and solution accuracy.