Rainfall Runoff Analysis Using Artificial Neural Network
Rainfall Runoff Analysis Using Artificial Neural Network
复制标题
使用人工神经网络进行降雨径流分析
DOI：
10.17485/ijst/2015/v8i14/54370
复制
发表时间：
2015
期刊：
影响因子：
--
通讯作者：
Himanshu Panjiar
中科院分区：
文献类型：
--
作者：
Ankit Chakravarti;N. Joshi;Himanshu Panjiar
背景/目的：本研究的主要目的是利用降雨模拟器进行降雨径流数据生成的实验室实验。为了验证这一观测数据，利用人工神经网络技术建立了径流观测数据估计模型。方法：利用降雨模拟器进行了12项室内实验，生成了不同坡度和降雨强度下的流域径流水文图。为了验证径流水文观测数据的有效性，采用人工神经网络进行了模拟。利用收集到的1076个数据点建立人工神经网络模型，计算径流流量。为了开发人工神经网络模型，可用数据被分离为70%用于训练，15%用于测试，15%用于验证。结果：人工神经网络模型的预测结果与观测值有较好的拟合效果，可用于水资源规划和管理等。对于模型性能的测试，采用Nash-Sutcliffe效率标准，NSE大于95%。结论：人工神经网络（Artificial Neural Network， ANN）能较好地预测径流观测资料。研究发现，人工神经网络不仅在复杂过程的精确建模方面很有前途，而且在从学习关系中提供洞察力方面也很有前途，这将有助于建模者理解正在研究的过程以及评估模型。
Background/Objective: The main objective of the present study is to conduct laboratory experiment for the generation of rainfall runoff data using rainfall simulator. For the validation this observed data, a model is establish for estimating observed runoff data using Artificial Neural Network (ANN) technique. Methods: A total 12 laboratory experiments were conducted using rainfall simulator to generate runoff hydrograph using various slope and rainfall intensity over the catchment. For the validation of observed runoff hydrograph data were simulate using ANN. The ANN model was developed using collected 1076 data point to compute runoff discharge. For developing ANN model, the available data were separated as 70% for training, 15% for testing and 15% for validation. Results: The predicted results using ANN model performed better estimation with observed values which is useful for water resources planning and management etc. For the testing of model performance Nash-Sutcliffe efficiency criteria were used which gives NSE greater than 95%. Conclusion: The comparison of observed and predicted runoff hydrograph reveals that the Artificial Neural Network (ANN) predicts the runoff data reasonably well in observed hydrograph. It is found that ANNs are promising tools not only in accurate modeling of complex processes but also in providing insight from the learned relationship, which would assist the modeler in understanding of the process under investigation as well as in evaluation of the model.