将单纯形-粒子群混合算法应用于分析二维河流横向扩散情况下的水团示踪试验数据，估计河流的横向扩散系数、断面平均流速和污水排放位置。数值试验结果表明：①加速因子c1,c2和参数初值取值范围综合影响粒子的搜索能力，当加速因子c1=c2=1.72时，有利于保持粒子的搜索能力；②在同样的条件下，混合算法的时间性能指标值小于单一的粒子群优化算法；③参数初值的取值范围对混合算法收敛性几乎没有影响；④混合算法可以有效地应用于河流水质数学模型参数识别问题。混合算法能改善粒子群算法在迭代后期出现的收敛速度慢、早熟的不足，是分析河流水质模型参数的一种有效方法。

Simplex-particle swarm hybrid algorithm (SM-PSO) was applied to analyze the experimental data of water quality of river in two-dimensional transverse dispersion, and to estimate the transverse dispersion coefficient, mean velocity of river, and location of continuous pollutant discharge. The results of numerical experiment show that: 1) SM-PSO algorithm can be effectively employed to analyze the experimental data of water quality and estimate water quality parameters. 2) Under the same condition, the time performance indicator of SM-PSO is less than that of PSO algorithm. 3) The range of initial guess value of water quality parameters has little influence on the convergence speed . 4) c1, c2 and the range of initial guess value have synthetic influences on the search capability in operation. When c1=c2=1.72, the search capability can be kept properly. SM-PSO algorithm can overcome the problem of PSO algorithm where it easily drops into local convergence and premature convergence. The hybrid algorithm was proved to be an effective way to estimate parameters for river water quality models.

国家自然科学基金（11171043）