A distribution-knowledge-guided assessment strategy for multiobjective particle swarm optimization

The selection of global best (gBest) is an important and challenging issue for multiobjective particle swarm optimization (MOPSO) algorithms. In this paper, a distribution-knowledge-guided assessment strategy (KS) is proposed to obtain the suitable gBest in MOPSO. The novelties of KS-MOPSO include the following three aspects. First, the distribution knowledge, including both the current and historical distributions of nondominated solutions, is designed to describe the distribution information of the optimal solutions. Second, an adaptive assessment mechanism using this knowledge is designed to select the appropriate gBest to improve the search performance. Third, an optimal technique is developed to update the archive to improve the computational efficiency. Finally, the performance of KS-MOPSO is compared with that of other algorithms on benchmark functions and a zinc electrolysis optimization problem. The experimental results show significant improvement over these state-of-the-art algorithms.