| 32 | 0 | 47 |
| 下载次数 | 被引频次 | 阅读次数 |
针对黄河中游小浪底水库至花园口水文站河段洪水过程的强非线性与突发性特征，以及传统优化调度模型计算复杂、时效性不足的问题，本文提出一种融合物理约束与深度学习的防洪调度模型（CNNBi LSTM-SA）。模型采用一维卷积神经网络（1D-CNN）提取水文序列的局部相关特征，利用双向长短期记忆网络（Bi LSTM）对全局时间依赖关系进行建模，并引入自注意力机制（SA）动态刻画洪水滞后特征。为增强模型的物理一致性，将水量平衡原理构建为物理约束项并嵌入损失函数中。实例应用结果表明，该模型在测试集上的模拟精度优于基准模型：小浪底水库出库流量与花园口水文站流量的纳什效率系数（NSE）分别达到0.9171和0.9691，均方根误差（RMSE）、平均绝对误差（MAE）及平均绝对百分比误差（MAPE）均明显降低，证明了混合网络架构在处理复杂调度逻辑方面的优越性。本研究利用Bi LSTM的双向信息传递机制与自注意力机制的动态权重建模，更精准地刻画了洪水波在长距离河道演进中的非线性、长时程延迟及滞后特征，可为流域防洪调度模拟提供兼顾计算效率与物理一致性的建模方法。
Abstract：In response to the highly nonlinear and sudden characteristics of flood processes of the section from the Xiaolangdi Reservoir to the Huayuankou Hydrological Stationin the middle reaches of the Yellow River, as well as the computational complexity and lack of timeliness associated with traditional optimization scheduling models, this paper proposed a flood control scheduling model(CNN-BiLSTM-SA) that integrated physical constraints with deep learning. The model employed a one-dimensional convolutional neural network(1D-CNN) to extract local correlation features from hydrological time series, utilized a bidirectional long shortterm memory(BiLSTM) network to model global temporal dependencies, and incorporated a self-attention mechanism(SA) to dynamically capture flood lag characteristics. To enhance the model's physical consistency, the principle of water balance was formulated as a physical constraint and incorporated into the loss function. Application results indicate that the model's simulation accuracy on the test set outperforms the baseline model: the Nash-Sutcliffe Efficiency(NSE) for outflow from Xiaolangdi Reservoir and flow at Huayuankou Station reaches 0.9171 and 0.9691, respectively, while the root mean square error(RMSE), mean absolute error(MAE), and mean absolute percentage error(MAPE) are all significantly reduced. It demonstrates the superiority of the hybrid network architecture in handling complex scheduling logic. This study leverages the bidirectional information propagation mechanism of BiLSTM and the dynamic weighting modeling of the self-attention mechanism to more accurately characterize the nonlinearity, long-term delay, and hysteresis of flood wave propagation on long-distance river channels. The proposed approach provides a modeling method for flood control scheduling simulation in basins that balances computational efficiency with physical consistency.
[1]水利部.李国英在第19届世界水资源大会第二场高级别全体会议部长级对话环节上的讲话[EB/OL].(2026-01-10)[2026-03-18].http://www.mwr.gov.cn/xw/slyw/202601/t20260110_2098447.html.
[2]DI BALD ASS ARRE G,SIVAPALAN M,RUSCA M,et al.Sociohydrology:Scientific Challenges inAddressing the Sustainable Development Goals[J].Water Resources Research,2019,55(8):6327-6355.
[3]杨子桐,方国华,叶健,等.统筹防洪、供水、生态的流域水资源系统优化调度研究[J].水资源保护,2025,41(4):71-78+124.
[4]XU W,MENG F,GUO W,et al.Deep reinforcement learning for optimal hydropower reservoir operation[J].Journal of Water Resources Planning and Management,2021,147(8):04021045.
[5]李丹,李志威,曾慧,等.基于深度学习的流溪河生态流量优化调度研究[J].武汉大学学报(工学版),2026,59(3):489-498.
[6]LAI V,HUANG Y F,KOO C H,et al.A review of reservoir operation optimisations:from traditional models to metaheuristic algorithms[J].Archives of Computational Methods in Engineering,2022,29(5):3435-3457.
[7]PHANKAMOLSIL Y,RITTIMA A,SAWANGPHOL W,et al.Deep reinforcement learning for multiple reservoir operation planning in the Chao Phraya River Basin[J].Modeling Earth Systems and Environment,2025,11(2):102.
[8]刘海力,卢勇,王渤权,等.基于物理机制引导的水库群出库流量预测在乌江流域的应用[J].水力发电,2025,51(7):95-102.
[9]高熠飞,陈宏新,陈超.融合机器学习与概念性调度模块的水库群径流预测[J/OL].人民长江,1-15(2026-03-03)[2026-04-15].https://link.cnki.net/urlid/42.1202.TV.20260302.1802.013.
[10]JIANGTAO L,YUCHEN B,KATHRYN L,et al.Probing the limit of hydrologic predictability with the Transformer network[J].Journal of Hydrology,2024,637:131389.
[11]徐刚,舒远丽,任玉峰,等.基于深度学习的三峡水库实时防洪调度模型[J].水力发电学报,2022,41(3):60-69.
[12]LIANG B,GIULIANI M,ZHANG L,et al.Fitted Q-iteration for optimal water reservoir network operation under varying hydro-climatic conditions[C]//EGU General Assembly Conference Abstracts.2020:17461.
[13]李文武,周佳妮,裴本林,等.梯级水库深度强化学习长期随机优化调度研究[J].水力发电学报,2023,42(11):21-32.
[14]孙榕,徐刚,李航宇,等.基于生成式人工智能的水库调度应用研究[J/OL].中国农村水利水电,1-20(2025-11-19)[2026-04-15].https://link.cnki.net/urlid/42.1419.TV.20251119.1533.024.
[15]KRATZERT F,KLOTZ D,BRENNER C,et al.Rainfall-runoff modelling using long short-term memory(LSTM)networks[J].Hydrology and Earth System Sciences,2018,22(11):6005-6022.
[16]朱思鹏.基于可解释机器学习的梯级水库短期精细化调度研究[D].武汉:华中科技大学,2024.
[17]SHEN C.A transdisciplinary review of deep learning research and its relevance for water resources scientists[J].Water Resources Research,2018,54(11):8558-8593.
[18]张建云,谢康,刘艳丽,等.融合物理机制的机器学习水文模型研究进展[J].人民长江,2025,56(10):37-46.
[19]HAO H,ZHAOLI W,YAOXING L,et al.Improving the explainability of CNN-LSTM-based flood prediction with integrating SHAP technique[J].Ecological Informatics,2024(84):102904.
[20]RUAN J L,CUI Y,SONG Y C,et al.A novel RF-CEEMD-LSTM model for predicting water pollution[J].Scientific Reports,2023,13(1):20901.
[21]徐冬梅,王逸阳,王文川.基于贝叶斯优化算法的长短期记忆神经网络模型年径流预测[J].水电能源科学,2022,40(12):42-46.
[22]叶然·道肯.基于BiLSTM深度学习模型的径流预测方法[J].水利科技与经济,2025,31(3):93-99.
[23]VASWANI A,SHAZEER N,PARMAR N,et al.Attention is all you need[J].Advances in Neural Information Processing Systems,2017(30):5998-6008.
[24]丁诚,王兆才,丁伟杰,等.基于可解释多源数据特征融合的深度学习集合径流预测[J].水科学进展,2025,36(4):581-595.
[25]张玮,郑雅莲,刘志武,等.物理机制引导的水库调度深度学习模型研究[J].水力发电学报,2023,42(3):13-25.
[26]KANG L,WU X,ZHOU L,et al.A novel deep learning coupled model for extracting flood control scheduling rules for reservoir groups[J].Journal of Hydrology,2026:135083.
[27]WEN Y F,LI P F,LI M Z,et al.Changes in River Cross-section Morphology and Response to Streamflow and Sediment Processes in Middle Reaches of Yellow River,China[J].Chinese Geographical Science,2025,35(1):161-174.
[28]陈翠霞,刘俊秀,付健.基于黄河中游水库群调控的下游河道输沙能力分析[J].人民黄河,2023,45(6):37-42.
基本信息:
中图分类号:TP18;TV87
引用信息:
[1]赵钊,程寅益,杨小东.融合物理约束与深度学习的防洪调度模型研究及其应用[J].中国水利,2026,No.1034(08):30-37.
2026-05-07
2026-05-07
2026-05-07