A differential evolution algorithm considering multi-population based on birth & death process and opposition-based learning with condition

Abstract

Differential evolution algorithm with multi-population cooperation and multi-strategy integration (MPMSDE) has been proven to be a better efficient evolutionary algorithm for global optimization. In MPMSDE, dynamic resource allocation and multi-population cooperation are introduced to distribute computational resources rationally. However, there is no effective escape mechanism when MPMSDE falls into a locally optimal solution. Thus, to achieve automatic escape from the local optimum, a differential evolution algorithm considering multi-population based on B&D process and opposition-based learning with condition (MPNBDE) is proposed in this paper. Different from MPMSDE, MPNBDE develops a B&D process and an opposition-based learning mechanism with condition to automatically search high-efficiency for optimal solutions. Also, a Fermi rule in MPNBDE is utilized to control the extent of the maximum computational resource, which is affected by global information. In MPNBDE, a new mutation strategy, ”DE/pbad-to-pbest-gbest-Fermi/1”, is proposed. The new strategy can not only control the extent of information exchanged by the Fermi rule, However, it also can significantly accelerate the convergence of the algorithm. Meanwhile, compared with other differential evolution algorithms, e.g., MPMSDE and SMLDE, our MPNBDE shows better performance in searching optimum, especially in the calculation accuracy and convergence speed on 21 benchmark functions.