Parallelization and Performance Test to Multiple Objective Particle Swarm Optimization Algorithm

Abstract

In recent years, Model calibration and parameter estimation with high complexity is a common problem in many areas of researches, especially in environmental modeling. This paper proposes a comparatively simple technique on the parallel implement of Multi-objective Particle Swarm Optimization algorithm (MOPSO). The transformation of the sequential objective evaluation in the MOPSO is based on the Matlab parallel computing tool box. Two study cases of different complexity demonstrate that the parallel implementation resulted in a considerable time saving. The deviation of computational time indicates that MOPSO has the characteristic of randomness because of the crowding distance and the dominant ranking. The proposed parallel MOPSO therefore, provides an ideal means to solve global optimization problems that are comparatively with high complexity.