Improved particle swarm optimization for parameter inversion of Muskingum model

In this study, the parameter inversion problem of the Muskingum model is considered. To overcome the premature phenomenon of particle swarm optimization, a new niche particle swarm optimizaiton (NPSO) is presented. NPSO combines traditonal particle swarm optimization with the fitness-sharing principle. By applying four test functions and parameter inversion based on the Muskingum model and comparing these with traditional particle swarm optimization, the efficiency in convergent speed and precision of this method are verified. However, inversion results are not good because of stochastic noise. To improve the antinoise capability of NPSO, a multiscale NPSO is constructed by combining the multiscale strategy with NPSO and by applying the parameter inversion of the Muskingum model with 5% stochastic noise. Inversion resultsverify the effectiveness of the improved algorithm; the antinoise performance of the NPSO has been increased, and the precision of the parameter inversion result is significantly improved. © 2016, Editorial Board of Journal of Harbin Engineering. All right reserved.