| 引用本文: | 张晓斌,李抗彬,郝改瑞,等.基于BP神经网络的新安江模型初始土壤蓄水量计算研究[J].灌溉排水学报,0,():-. |
| ZHANG Xiaobin,LI Kangbin,HAO Gairui,et al.基于BP神经网络的新安江模型初始土壤蓄水量计算研究[J].灌溉排水学报,0,():-. |
|
| 摘要: |
| 【目的】为克服传统经验折减系数法在计算新安江模型初始土壤蓄水量方面的缺点，并提高在湿润半湿润地区新安江模型的应用效果。【方法】文章中结合流域初始土壤蓄水量的影响因素和神经网络模型特点，提出构建基于BP神经网络的新安江模型初始土壤蓄水量计算方法。【结果】在三种输入因子组合方式下，当BP神经网络隐含层节点大于11时，模拟训练期模型应用效果达到项目精度评价指标的甲等水平，预测检验期的9个样本，均有6个以上样本检验合格；与采用传统经验折减系数法计算新安江模型初始土壤蓄水量相比，采用BP神经网络模型应用效果明显占优。【结论】①在湿润半湿润地区采用BP神经网络模型计算新安江模型初始土壤蓄水量具有可行性和适用性；②当BP神经网络隐含层节点选择比较合理时，基于BP神经网络的新安江模型初始土壤蓄水量计算方法应用效果要优于传统的经验折减系数法。③该方法在应用过程中可克服采用经验折减系数法计算土壤初始蓄水量需要选择流域一场暴雨或久旱未雨才能开始计算和计算过程不能中断的缺点，为新安江模型在湿润半湿润地区应用时初始土壤含水量计算提供了新的参考方法。 |
| 关键词: 初始土壤蓄水量；BP神经网络；新安江模型；径流模拟 |
| DOI： |
| 分类号:P338.9 |
| 基金项目:国家自然科学基金（51479215）；山西省水利厅科技项目（TZ2019026）；运城学院博士科研项目（YQ-2020003） |
|
| Study on Calculation Method of Initial Soil Water Storage of Xin'anjiang Model Based on BP Neural Network |
|
ZHANG Xiaobin1, LI Kangbin2, HAO Gairui3, ZHANG Xiaopeng1
|
|
1.Yuncheng University;2.Xi’an Land Water and Electricity Measurement and Control CO.LTD;3.Xi’an University of Technology
|
| Abstract: |
| 【Objective】In order to overcome the shortcomings of traditional empirical reduction coefficient method in calculating the initial soil water storage of Xin'anjiang model, and to improve the application effect of Xin'anjiang Model in humid and semi humid areas. 【Methods】 Based on the influencing factors of the initial soil water storage and the characteristics of the neural network model, the calculation method of initial soil water storage of Xin'anjiang model based on BP neural network was proposed in the paper. 【Results】Under the combination of three input factors, when the hidden layer node of BP neural network is greater than 11, the application effect of the model in the simulation training period reaches the first-class level of the project accuracy evaluation index, and 9 samples in the prediction test period, all of which are more than 6 samples qualified; compared with the traditional empirical reduction coefficient method for calculating the initial soil water storage of Xin'anjiang model, the application effect of BP neural network model is obviously superior. 【Conclusion】① it is feasible and applicable to use BP neural network model to calculate the initial soil water storage of Xin'anjiang Model in humid and semi humid areas; ② when the selection of hidden layer nodes of BP neural network is reasonable, the application effect of the calculation method of initial soil water storage of Xin'anjiang model based on BP neural network is better than the traditional empirical reduction coefficient method. ③ In the process of application, this method can overcome the shortcomings of using the empirical reduction coefficient method to calculate the initial soil water storage, which needs to select a rainstorm or a long drought without rain in the basin to start the calculation and the calculation process can’t be interrupted. It provides a new reference method for the initial soil water storage of Xin'anjiang model when it is applied in the humid and semi humid areas. |
| Key words: Initial soil water storage; BP neural network; Xin 'anjiang model; Runoff simulation |
