MOVPSO算法在TOPMODEL参数优化中的应用

摘要

为了对水文模型中难以直接测算的参数进行调试和优化,将多目标涡流粒子种群优化算法(Multi-Ob-jective Vortex Particle Swarm Optimization,MOVPSO)应用于水文模型参数优化计算中,并以逼近性(Generation-al Distance,GD)及超体积值(Hyper-Volume,HV)作为算法性能评价指标.将MOVPSO算法与NSGA-Ⅱ算法及多目标粒子种群优化算法(Multi-Objective Particle Swarm Optimization,MOPSO)独立运行50次所得Pa-reto前沿的GD值及HV值进行统计分析,并结合方差分析比较3种算法的性能.将MOVPSO算法迭代过程中的粒子种群速率、种群半径的估计值与测量值进行对比分析并判别两者的拟合程度.用尼泊尔巴格玛蒂河流域2005~2011年期间实测洪水日径流过程资料作为TOPMODEL模型参数率定系列,运用MOVPSO算法对模型参数进行优化,得出 Pareto 最优解,并利用2013年5场洪水日径流过程进行模型检验.结果表明:MOVPSO算法所得Pareto解集性能优于NSGA-Ⅱ及MOPSO算法,拟合历史洪水平均确定性系数达到0.85,模型预报精度高,表明MOVPSO优化算法在解决多参数多目标优化问题中具有优势.%In order to adjust and optimize the parameters of hydrological model( TOPMODEL) that can not be measured and calculated directly,MOVPSO algorithm (Multi-Objective Vortex Particle Swarm Optimization) was applied to the optimization of TOPMODEL parameters. In efficiency evaluation of algorithms, we use generational distance (GD) and Hyper - Volume (HV) as the performance index. We compared algorithm performances through independent 50 runs of MOVPSO,NSGA-II and MOPSO (Multi-Objective Particle Swarm Optimization) by analyzing the statistic results of GD and HV. And the estimation values of particle swarms'velocity and radius in MOVPSO iteration process were compared with the measured ones to judge the fit-ting degree. Taking the daily flood discharge process data from 2005 to 2011 in Bagmati River Basin(in Nepal) as the parameter determination series of TOPMODEL model,MOVPSO algorithm was used to optimize the model parameters and the Pareto optimal solutions were obtained. Finally,the model was tested by using daily runoff process of 5 floods in 2013. The results showed that the MOVPSO algorithm performs better than NSGA-II algorithm and MOPSO algorithm,the fitted average uncertainty coefficient of historical flood reaches 0.85 and the forecast accuracy is high,indicating that the MOVPSO optimization algorithm has advan-tages in solving the multi-objective optimization issue of multi-parameters.