您的位置:首页 > 论文页面
新安江产流模型与改进的BP汇流模型耦合应用
发表时间:2012-11-15 浏览量:1320 下载量:302
全部作者: | 阚光远,李致家,刘志雨,姚成,周赛 |
作者单位: | 河海大学水文水资源学院,水资源高效利用与工程安全国家工程研究中心;水利部水文局 |
摘 要: | 为提高新安江模型的汇流计算精度并减少经验因素对参数率定的影响,将新安江产流模型与改进的反向传播(back propagation,BP)汇流模型相耦合,建立XBK(XAJ-BP-KNN:XAJ,Xinanjiang,新安江;KNN,K-nearest neighbor,K-最近邻算法)模型。该模型以前期模拟流量和新安江产流模型计算的产流量作为BP网络的输入,出口断面流量作为网络输出,拟合汇流的非线性关系,代替新安江模型的分水源、线性水库及河道马法的汇流计算;采用相似原理和KNN算法,基于历史样本的模拟误差及相应影响要素对网络输出进行误差修正,实现了无前期实测流量的连续模拟;模型使用SCE-UA算法与遗传早停止(early stopping)LM(Levenberg-Marquardt)算法相结合的全局优化方法进行参数优选。在呈村流域的验证表明:XBK模型的模拟精度高于新安江模型,全局优化方法能找到最优参数,降低模型的使用难度。 |
关 键 词: | 水文学;新安江模型;人工神经网络;反向传播算法;K-最近邻算法;SCE-UA算法 |
Title: | Application of coupling of Xinanjiang runoff generation model and improved BP flow concentration model |
Author: | KAN Guangyuan, LI Zhijia, LIU Zhiyu, YAO Cheng, ZHOU Sai |
Organization: | College of Hydrology and Water Resources, Hohai University, National Engineering Research Center of Water Resources Efficient Utilization and Engineering Safety; Hydrology Bureau Ministry of Water Resources |
Abstract: | In order to improve the flow concentration accuracy of the Xinanjiang model meanwhile reduce the influences of experiential factors on calibration, a rainfall-runoff model named XBK (XAJ-BP-KNN) is proposed in this paper. In the XBK model, the runoff is calculated by the Xinanjiang runoff generation model, but the flow concentration module (runoff separation, linear reservoir and Muskingum routing method) of the Xinanjiang model is disused and the outlet flow is calculated by the improved BP flow concentration model. The BP flow concentration model uses BP neural network to simulate the nonlinear relationship of the flow concentration process, runoff calculated by the Xinanjiang runoff generation model and simulated antecedent outlet flow are the network inputs, outlet flow is the output. For the sake of simulating outlet flow continuously without using observed antecedent outlet flow and having a higher accuracy, the output of the network is corrected by adopting the similarity theory and the K-nearest neighbor algorithm, the historical samples’ simulation error and the corresponding impact factors are the inputs. A global optimization algorithm combined by the SCE-UA algorithm and the genetic early stopping LM algorithm is used to calibrate the model. The XBK model is applied to Chengcun watershed, the results indicate that the model has a higher accuracy than the Xinanjiang model, the global optimization algorithm can find the optimal parameter and the model is easy to use. |
Key words: | hydrology; Xinanjiang model; artificial neural network; back propagation algorithm; K-nearest neighbor algorithm; SCE-UA algorithm |
发表期数: | 2012年11月第21期 |
引用格式: | 阚光远,李致家,刘志雨,等. 新安江产流模型与改进的BP汇流模型耦合应用[J]. 中国科技论文在线精品论文,2012,5(21):2085-2093. |

请您登录
暂无评论