变尺度混沌-遗传算法在复杂河流水质模型参数优化中的应用

将变尺度混沌-遗传算法(MSCGA)应用于复杂河流水质模型参数优化.采用湘江衡阳段水质监测资料,以二维河流水质数学模型反演结果的均方误差为适应度函数,估计横向扩散系数Dx、纵向弥散系数Dy和污染物衰减系数κ.数值实验结果表明,MSCGA寻优过程具有明显的分级特征,级级收敛;在同样的条件下,MSCGA的收敛速度较快,为遗传算法(GA)的1.36倍;同时,MSCGA克服了GA早熟收敛的问题,其最优适应度函数值为7.6×10-4,而GA的最优适应度函数值9.6×10-4.将MSCGA应用于研究河段,求得Dx、Dy分别为0.1335、0.0011,BOD5、As、Cr的衰减系数κ分别为0.0229、0.0100、0.0107.

Application of mutative scale chaos genetic algorithm (MSCGA) to parameters estimation for river water quality model

Mutative scale chaos genetic algorithm (MSCGA) was applied to parameter estimation of water quality models for river with complicated hydrological characteristic. Fitness function was established according to mean square error of predicted data of two-dimensional water quality model and monitoring data from Xiangjiang, Hengyang Section. The estimated parameters included longitudinal diffusivity coefficient, horizontal diffusivity coefficient and pollutant decay constant. The results showed that the process of MSCGA optimization had the obvious character of step-by-step convergence. Under the same condition, the convergence of MSCGA was 1.36 times faster than genetic algorithms (GA). At the same time, MSCGA with fitness function value of 7.6×10-4 overcame problems of premature convergence by simple GA, while the value of GA was 9.6×10-4. By the application of MSCGA to the parameter estimation in water quality model, we obtained the optimal values of Dx and Dy as 0.1335 and 0.0011, and the values of pollutant decay constants as 0.0229, 0.0100, 0.0107 for BOD5, As,Cr respectively.