基于遗传算法的新安江模型日模拟参数优选研究

在概念性水文模型的参数率定中,目前还没有一个传统优化方法能够提供保证足够高效和稳定性的算法。为了克服传统优化方法中局部收敛性的缺点,近年来利用遗传算法通过计算机准确稳定地进行概念性水文模型的参数优选的尝试得到越来越多的重视和发展。目前优选水文模型待定参数,大多是从次洪模型的方面去讨论,有关日模拟模型的遗传算法参数优选讨论的较少。本文系统分析了基于遗传算法的新安江模型日模拟参数的自动优选,同时针对遗传算法在模型参数众多的情况下时间效率低下问题,通过利用新安江模型参数分层原理与模型参数敏感性分析对优选结果影响,提出一套简化的日模型参数遗传算法优选方案。经过流域模拟检验,该优选方案可行,运行效率高,可以作为类似模型遗传算法参数率定快速、有效的方案。

Application of Genetic Algorithm for Model Parameter Calibration in Daily Rainfall-Runoff Simulations with the Xinanjiang Model

By now,there are still no traditional optimization methods being capable of providing enough steady and efficient algorithm to determine the prior-calibrated modeling parameters in conceptual model for rainfall- runoff simulations.The most knotty problem for this issue mainly attributes to the local convergence of the traditional optimization methods,the genetic algorithm capable of yielding accurate and stable optimization of the modeling parameters in conceptual models,therefore,has attract much attention and has rapidly developed with the computing advances.However,most of the efforts on genetic algorithm applications for modeling parameter calibration are concentrated on flooding simulations,few attempt has been made to modeling parameter optimizations with genetic algorithm in daily stream-flow simulations.Bearing this in mind,this study presents an approach to systematically automate-optimize the calibrated parameters in daily stream-flow simulations by using the Xinanjiang hydrological model.For resolving the frequently occurred problem of rather low computation efficiency in genetic algorithm applications,a set of simplified method in parameter optimization based on hierarchy principle of the model parameters and model sensitivity analyses is proposed.An experimental study on model parameter calibration by using the Xinanjiang hydrological model is conducted to a 2 341.6 km2 watershed located on the upper stream of the Hanjiang River Basin.The preliminary study shows that the proposed approach is feasible and of high computation efficiency,and can be transferred to model parameter calibrations for conceptual hydrological models in the similar categories.