Abstract
In this paper, a hybrid particle swarm optimization (PSO) and simplified swarm optimization (SSO) algorithm is developed for solving stochastic multimodal optimization problems. The most attractive features of PSO are its algorithmic easy implantation and fast convergence. However, most studies of PSO still focus on deterministic problem and mention that PSO has premature convergence in strongly multi-modal optimization problems. This paper proposes two methods (PSO and SSO) incorporating the optimal computing budget allocation (OCBA), hypothesis test (HT), and local search (LS) into the PSO or SSO algorithms, aimed at enhancing global search capability and searching efficiency in noisy environment. These two hybrid algorithms are called PSOOHL and SSOOHL, respectively. In these two hybrid algorithms, PSO and SSO are used to explore and exploit the solution space. The OCBA is applied to allocate computational budgets to provide reliable evaluation and identification in the swarm of PSO or SSO for updating next generation. Hypothesis test (HT) and local search are used for distinguishing the similarity of particles and generating the diversity particle to prevent premature convergence. Numerical experiments based on four benchmark functions are performed. The result shows that PSOOOHL is superior in terms of searching quality and efficiency.