南水北调与水利科技
南水北調與水利科技
남수북조여수리과기
SOUTH-TO-NORTH WATER
2014年
5期
109-112,123
,共5页
BP 人工神经网络%径流模拟%水文模型%日径流%模拟精度
BP 人工神經網絡%徑流模擬%水文模型%日徑流%模擬精度
BP 인공신경망락%경류모의%수문모형%일경류%모의정도
BP artificial neural netw ork%runoff modeling%hydrological model%daily runoff%simulation accuracy
以湖北省宜昌市某一小流域为例，详细介绍了研究区径流模拟的 BP 神经网络模型建模方法，并利用模型对研究区的日径流进行了模拟预测研究。首先，根据研究区降雨2径流的时间分布特征，确定了对丰水期、枯水期分别建模的建模方案；接着分析了流域产流的主要影响因素，确定了将前五日径流量、前三日降雨量、当前降雨量和蒸散发量为作为模型的输入变量；并在反复试验的基础上，选取了合适的模型结构和学习效率参数；最后，利用确定性系数对模型的径流预测精度进行了评定。结果表明，针对丰、枯水期分类建立的 BP 神经网络模型克服了以往模型对极值事件模拟精度较差的不足，对高流量和低流量的模拟精度高。
以湖北省宜昌市某一小流域為例，詳細介紹瞭研究區徑流模擬的 BP 神經網絡模型建模方法，併利用模型對研究區的日徑流進行瞭模擬預測研究。首先，根據研究區降雨2徑流的時間分佈特徵，確定瞭對豐水期、枯水期分彆建模的建模方案；接著分析瞭流域產流的主要影響因素，確定瞭將前五日徑流量、前三日降雨量、噹前降雨量和蒸散髮量為作為模型的輸入變量；併在反複試驗的基礎上，選取瞭閤適的模型結構和學習效率參數；最後，利用確定性繫數對模型的徑流預測精度進行瞭評定。結果錶明，針對豐、枯水期分類建立的 BP 神經網絡模型剋服瞭以往模型對極值事件模擬精度較差的不足，對高流量和低流量的模擬精度高。
이호북성의창시모일소류역위례，상세개소료연구구경류모의적 BP 신경망락모형건모방법，병이용모형대연구구적일경류진행료모의예측연구。수선，근거연구구강우2경류적시간분포특정，학정료대봉수기、고수기분별건모적건모방안；접착분석료류역산류적주요영향인소，학정료장전오일경류량、전삼일강우량、당전강우량화증산발량위작위모형적수입변량；병재반복시험적기출상，선취료합괄적모형결구화학습효솔삼수；최후，이용학정성계수대모형적경류예측정도진행료평정。결과표명，침대봉、고수기분류건립적 BP 신경망락모형극복료이왕모형대겁치사건모의정도교차적불족，대고류량화저류량적모의정도고。
The BP neural net work model method for runoff modeling was int roduced and the model was applied to simulat e the daily runoff in a watershed of Yichang in Hubei Province. First, based on the temporal distribution of rainfall-runoff in the study area, the modeling approaches for the wet and dry seasons w ere developed separately. Second, the main factors affect ing the run-off were analyzed, and t he input variables of the model included the runoff offive previous days, rainfall of three previous days, current rainfall, and current evapotranspiration. Third, the appropriat e model structure and learning eff iciency parameters were determ ined through trial-and-error tests. Finally, det erminacy coefficient was used to assess the accuracy of simulation results. The results showed that the BP neural netw ork models of the wet and dry seasons overcome the disadvantages of low accuracy in previous models w hen simulat ing the extreme events, and the BP neural network model can simulate the high and low runoff conditions wit h high accuracy.
