摘要
以广东省滃江流域为典型研究对象,旨在通过先进的数据驱动方法改进传统的水文模型,从而提高洪水过程模拟准确性。文章提出一种新型LSTM-SWMM混合模型,其综合了SWMM物理模型和LSTM神经网络的优点。结果表明,LSTM-SWMM模型模拟的洪水过程与实际观测流量具有良好一致性,模型的R^(2)值为0.97,MAE和RMSE依次为81.3m^(3)/s、148.6m^(3)/s;而LSTM对照模型的R^(2)值为0.91,MAE和RMSE依次为100.8m^(3)/s、192.4m^(3)/s。LSTM-SWMM网络因其独特的设计,在预测洪水序列过程中上表现出显著优势。
Taking the Hanyu River Basin in Guangdong Province as a typical research object,this paper aims to improve the traditional hydrological model through advanced data-driven method,so as to improve the accuracy of flood process simulation.This paper proposes a new hybrid model of LSTM-SWMM,which combines the advantages of SWMM physical model and LSTM neural network.The results show that the flood process simulated by LSTM-SWMM model is in good agreement with the actual observed flow.The R^(2)value of the model is 0.97,and the MAE and RMSE are 81.3 m^(3)/s and 148.6 m^(3)/s respectively.The R^(2) value of LSTM control model was 0.91,and the MAE and RMSE were 100.8 m^(3)/s and 192.4 m^(3)/s,respectively.Because of its unique design,LSTM-SWMM network shows significant advantages in the process of flood prediction.
作者
刘科佑
Liu Keyou(Hydrological Branch of Shaoguan,Guangdong Provincial,Shaoguan 512000,Guangdong)
出处
《陕西水利》
2024年第6期45-48,共4页
Shaanxi Water Resources
