| 分布式水循环模型的参数优化算法比较及应用 |
| |
| 引用本文： | 孙波扬,张永勇,门宝辉,张士锋.分布式水循环模型的参数优化算法比较及应用[J].资源科学,2013,35(11):2217-2223. |
| |
| 作者姓名： | 孙波扬 张永勇 门宝辉 张士锋 |
| |
| 作者单位： | 华北电力大学可再生能源学院, 北京 102206;中国科学院地理科学与资源研究所陆地水循环及地表过程重点实验室, 北京 100101;中国科学院地理科学与资源研究所陆地水循环及地表过程重点实验室, 北京 100101;华北电力大学可再生能源学院, 北京 102206;中国科学院地理科学与资源研究所陆地水循环及地表过程重点实验室, 北京 100101 |
| |
| 基金项目： | 国家973计划项目(编号:2012CB955304);国家自然科学基金面上项目(编号:41271005);国家973计划项目(编号:2009CB421403);武汉大学水资源与水电工程科学国家重点实验室开放基金(编号:2011B078)。 |
| |
| 摘 要： | 分布式水文模型的优势在于还原水文过程的时空变异性,可以很好地模拟和反映各种水文要素和下垫面因素的时空分布不均匀性。由此也导致模型参数过多,在子流域过多的情况下,人工调节参数繁琐复杂,应用优化算法实现参数自动调节成为首选。本文选取石羊河流域九条岭站1988-2005年实测径流资料,分别应用SCE-UA算法、遗传算法(GA)和粒子群算法(PSO)对分布式水循环模型(时变增益模型)进行参数率定,对比3种算法的收敛速度、所需迭代次数和算法稳定性。结果表明:通过SCE-UA、GA和PSO的优化,模型水平衡系数都控制在0.0左右,而相关系数和效率系数分别能达到0.90和0.84以上,模拟精度较好。但粒子群算法的全局搜索能力和收敛速度优于SCE-UA和遗传算法,所需迭代次数最少,初值敏感性小,更适合时变增益模型的参数寻优,有很高的扩展性和改进潜力。
|
| 关 键 词： | SCE-UA算法 GA算法 PSO算法 分布式水循环模型 |
Optimization Algorithm Comparison and Application to Parameter Calibration in a Distributed Hydrological Model |
| |
| SUN Boyang,ZHANG Yongyong,MEN Baohui and ZHANG Shifeng.Optimization Algorithm Comparison and Application to Parameter Calibration in a Distributed Hydrological Model[J].Resources Science,2013,35(11):2217-2223. |
| |
| Authors: | SUN Boyang ZHANG Yongyong MEN Baohui and ZHANG Shifeng |
| |
| Institution: | School of Renewable Energy, North China Electric Power University, Beijing 102206, China;Key Laboratory of Water Cycle and Related Land Surface Processes, Institute of Geographic Sciences and Natural Resources Research, Chinese Academy of Sciences, Beijing 100101, China;Key Laboratory of Water Cycle and Related Land Surface Processes, Institute of Geographic Sciences and Natural Resources Research, Chinese Academy of Sciences, Beijing 100101, China;School of Renewable Energy, North China Electric Power University, Beijing 102206, China;Key Laboratory of Water Cycle and Related Land Surface Processes, Institute of Geographic Sciences and Natural Resources Research, Chinese Academy of Sciences, Beijing 100101, China |
| |
| Abstract: | The distributed hydrological model is so advantageous that the spatial and temporal variability of hydrological processes can be restored and hydrological elements and inhomogeneities of spatial and temporal distribution can be simulated. However, there are many parameters in model simulation, especially in cases with many sub-basins, and manually adjusting parameters is cumbersome and complex. Optimization algorithms, which can automatically adjust parameters, are a preferred approach to counter these issues. After selecting and analyzing runoff data from 1988 to 2005 for the Jiutiaoling Station, Shiyang River, distributed time-varying gain model parameters (Distribute Time Variant Gain Model, DTVGM) were calibrated by applying the SCE-UA algorithm, genetic algorithm (GA) and particle swarm optimization (PSO) algorithm. We compared performance by looking at convergence rates, iteration number and the stability of the three algorithms. We found that after optimization using the three algorithms, the water balance coefficients were well controlled at around 0.0; the correlation coefficient reached 0.90, and the efficiency factor was more than 0.84. All model simulation results had much better precision and accuracy; however, the global search ability and convergence rates of PSO are superior to the SCE-UA and genetic algorithms. The PSO algorithm had the best performance in the minimum number of iteration processes and showed little sensitivity to initial values. This algorithm is more suitable for parameter optimization of the time-varying gain model. |
| |
| Keywords: | SCE-UA GA PSO Distributed hydrological model |
|
| 点击此处可从《资源科学》浏览原始摘要信息 |
| 点击此处可从《资源科学》下载免费的PDF全文 |
|
