| 引用本文: | 邓超,孙培源,尹鑫,邹佳成,王卫光.基于概念性水文模型与长短时记忆模型耦合的汉江上游流域径流模拟.湖泊科学,2025,37(1):279-292. DOI:10.18307/2025.0143 |
| Deng Chao,Sun Peiyuan,Yin Xin,Zou Jiacheng,Wang Weiguang.Runoff simulation in the upper Han River Basin using physics-informed machine learningmodel. J. Lake Sci.2025,37(1):279-292. DOI:10.18307/2025.0143 |
|
| 摘要: |
| 为了探究概念性水文模型（GR4J）与长短时记忆模型（LSTM）耦合过程中物理模型参数反馈机制以及考虑土壤含水量作为中间变量对物理引导机器学习（PIML）模型径流模拟的影响，本研究构建了PIML模型并设置考虑参数反馈、考虑中间变量和两者同时考虑的3种方案（依次简称为H1、H2、H3），以安康站为控制站的汉江上游流域开展实例研究。结果表明：（1）3种PIML模型径流模拟效果均优于LSTM模型，其中验证期纳什系数（NSE）平均提升10.6%，而PIML-H1与PIML-H3模型径流模拟效果优于GR4J模型，验证期NSE平均提升4.2%，其中PIML-H3模型在3种PIML模型中表现最佳，表明基于LSTM和GR4J模型构建的PIML模型对径流模拟效果有所改善，且同时考虑中间变量和参数反馈因素时对径流模拟效果改善最佳；（2）3种PIML模型对低水流量的模拟效果均优于GR4J和LSTM模型，且PIML-H3模型效果最佳，而对于高水流量，3种PIML模型均表现不佳，说明PIML模型往往在模拟低流量事件中更占优势；（3）3种PIML模型的径流模拟效果均表现出较强的季节性变化，PIML-H2和PIML-H3模型相较于PIML-H1模型受到的季节性变化影响更为明显，说明PIML模型模拟径流结果的季节性变化受到中间变量的影响。研究可为汉江上游流域径流模拟预报提供技术支撑。 |
| 关键词: 物理引导机器学习 径流模拟 中间变量 GR4J LSTM 汉江 |
| DOI：10.18307/2025.0143 |
| 分类号: |
| 基金项目:国家重点研发计划项目（2022YFC3202802）；国家自然科学基金项目（51979071）联合资助 |
|
| Runoff simulation in the upper Han River Basin using physics-informed machine learningmodel |
|
Deng Chao1，2，Sun Peiyuan1，2，Yin Xin3，Zou Jiacheng4，Wang Weiguang1，2
|
|
1.College of Hydrology and Water Resources, Hohai University, Nanjing 210098 , P.R.China ；2.The National Key Laboratory of Water Disaster Prevention, Hohai University, Nanjing 210098 , P.R.China ；3.The National Key Laboratory of Water Disaster Prevention, Nanjing Hydraulic Research Institute, Nanjing 210029 , P.R.China ；4.Hydrology and Water Resources Monitoring Center of Lower Ganjiang River, Yichun 336000 , P.R.China
|
| Abstract: |
| This study investigated the impact of coupling the conceptual hydrological model (GR4J) with the long short-term memory model (LSTM) in a physics-informed machine learning (PIML) framework for runoff simulation. Three scenarios (H1, H2 and H3) were designed to examine the effects of the physical model parameter feedback mechanism, the consideration of soil moisture as an intermediate variable, and the former both on the PIML models, respectively. The case study was conducted in the upper Han River Basin, with the Ankang hydrological station as the control station. The main findings were as follows: (1) Compared with the LSTM model, all three PIML models had improved performance on runoff simulation, with a 10.6% increase in average Nash-Sutcliffe efficiency (NSE) during the validation period. Additionally, both the PIML-H1 and PIML-H3 models exhibited better performance than the GR4J model, with a 4.2% increase in average NSE during the validation period. Notably, the PIML-H3 model outperformed other PIML models, indicating that coupling GR4J and LSTM models simultaneously considering intermediate variables and parameter feedback yielded the most significant improvement in the model performance of runoff simulation. (2) For low flows, all three PIML models outperformed the GR4J and LSTM models, and the PIML-H3 model achieved the best performance. For high flows, the performance of all three PIML models was not high, implying that PIML models were suitable in simulating low flows events. (3) The runoff simulations from the three PIML models exhibited significantly seasonal variations during both the training and validation periods. The seasonal variations in the PIML-H2 and PIML-H3 models were more pronounced compared to that in the PIML-H1 model, indicating that the seasonal variations in simulated runoff results of the PIML model were influenced by intermediate variables. This study contributed to a better understanding of the performance differences among various PIML model schemes in runoff simulation, providing technical support for runoff simulation and forecasting in the study area. |
| Key words: Physics-informed machine learning runoff simulation intermediate variable GR4J LSTM Han River |
