水利学报2025，Vol.56Issue(3)：387-397,11.DOI:10.13243/j.cnki.slxb.20240286
基于MDUPLEX-LSTM的非一致条件下的水库出流模拟
Simulation of reservoir outflow under inconsistent condition based on MDUPLEX-LSTM
摘要
Abstract
For the hydrological simulation of river basins with reservoirs,accurately simulating reservoirs'outflow is crucial for the accurate simulation of streamflow.Under the inconsistency among climatic and hydrological condi-tions as well as human activities,it is still a challenge how to reasonably simulate the reservoir discharge process.In this paper,both Longyangxia,a multi-year regulation reservoir,and Liujiaxia,an annual regulation reservoir,in the upper reaches of the Yellow River are taken as research objects.A reservoir outflow simulation method MDU-PLEX-LSTM,which couples deterministic dataset sampling algorithm and machine learning,is proposed and es-tablished.The results show that the MDUPLEX-LSTM can learn the actual reservoir operation patterns under dif-ferent hydrological conditions.MDUPLEX-LSTM achieved NSE values of over 0.84 and 0.73 for the daily outflow simulations of the Longyangxia and Liujiaxia reservoirs,respectively.Additionally,the discrepancy of the simula-tion accuracy among the calibration test and validation datasets is negligible,indicating the robustness of MDU-PLEX-LSTM.关键词
水库出流模拟/机器学习/LSTM/MDUPLEX/水库调度/黄河上游Key words
simulation of reservoir outflow/Machine Learning/LSTM/MDUPLEX/reservoir operation/upper Yellow River分类
建筑与水利引用本文复制引用
余昕鹏,吴书悦,赵建世..基于MDUPLEX-LSTM的非一致条件下的水库出流模拟[J].水利学报,2025,56(3):387-397,11.基金项目
国家自然科学基金项目(黄河水科学研究联合基金)(U2243215) （黄河水科学研究联合基金）
河南省黄河流域生态环境保护与修复重点实验室开放基金项目(LYBEPR202404) （LYBEPR202404）
