VMD-IFDA-NAHL模型在某水文站日径流预测中的应用

Application of the VMD-IFDA-NAHL model in daily runoff prediction for the hydrological station

摘要

为提升水资源科学化管理水平，增强洪水灾害预防能力，构建一种基于变分模态分解-改进流向算法-具有增强隐含层的自动神经网络模型（variational mode decomposition-improved flow direction algorithm-network with an augmented hidden layer，VMD-IFDA-NAHL）的日径流预测模型，该模型运用变分模态分解技术对某水文站的日径流数据进行预处理，去除径流数据中存在的噪声并提取日径流数据中潜在的非线性特征；采用改进流向算法对具有增强隐藏层的自动神经网络模型进行优化，提高模型对径流变化的捕捉能力和预测精度。同时，以某水文站为例，开展VMD-IFDA-NAHL预测模型与对比模型日径流预测的研究。结果表明：VMD-IFDA-NAHL模型在径流预测的准确性、稳定性和泛化能力方面均优于对比模型，显示出在径流预测领域的广阔应用潜力。构建的日径流预测模型可以为水资源管理、防洪调控等提供可靠的、科学的数据支撑。

Abstract

Conducting research on high-precision runoff prediction models is crucial in providing a reliable data support for water resource management and flood control. For daily runoff prediction, we designed a variational mode decomposition-improved flow direction algorithm network with an augmented hidden layer (VMD-IFDA-NAHL). This model first employs variational mode decomposition (VMD) to preprocess daily runoff data from the hydrological station, removing noise and extracting latent nonlinear features within the runoff data. Subsequently, an improved flow direction algorithm (IFDA) is applied to optimize the auto-neural network model with an augmented hidden layer, thereby improving the model's ability to capture runoff variations and improve prediction accuracy. Using the hydrological station as a case study, the VMD-IFDA-NAHL model and a reference model are compared in terms of daily runoff prediction. The results show that the VMD-IFDA-NAHL model outperforms the reference model in terms of accuracy, stability, and generalization ability in runoff prediction, indicating that it has a wide range of applications in runoff forecasting. Therefore, the daily runoff prediction model developed can provide reliable and scientific data support for water resource management, flood control regulation, and other related fields.

关键词

日径流预测 / 自动神经网络 / 变分模态分解 / 流向算法 / 改进径流算法

Key words

daily runoff prediction / automatic neural network / variational modal decomposition / flow direction algorithm / improved flow direction algorithm

参考文献

[1] 程家波, 姬厚灵, 陈中举, 等. 基于GWO-PE-VMD-ResNet组合模型的日径流预测[J]. 水电能源科学, 2024(8): 43-47. DOI: 10.20040/j.cnki.1000-7709.2024.20231549.

[2] 史东超. 基于EEMD+RF算法的唐山滦河径流量预测研究[J]. 水利技术监督, 2024(6): 172-175. DOI: 10.3969/j.issn.1008-1305.2024.06.045.

[3] Zhao Qun, Zhu Yuelong, Shu Kai, et al. Joint spatial and temporal modeling for hydrological prediction[J]. IEEE Access, 2020, 8: 78492-78503. DOI: 10.1109/access.2020.2990181.

[4] Jiang Shijie, Zheng Yi, Solomatine D. Improving AI system awareness of geoscience knowledge: Symbiotic integration of physical approaches and deep learning[J]. Geophysical Research Letters, 2020, 47(13): e2020GL088229. DOI: 10.1029/2020GL088229.

[5] Lu Lingjiang, Yan Tao, Chen Yongcan, et al. A lake water level prediction method based on data augmentation and physics-informed neural networks with imbalanced data[J]. Journal of Hydrology, 2026, 665(c): 134660. DOI: 10.1016/j.jhydrol.2025.134660.

[6] Hafezifar E, Shourian M. Groundwater level prediction using deep learning-based recurrent neural network and numerical modeling: A comparative study[J]. Earth Science Informatics, 2025, 18(2): 1-18. DOI: 10.1007/s12145-025-01859-0.

[7] Iqbal J, Su Chunli. Hydrogeochemistry and prediction of arsenic contamination in ground water of Vehari, Pakistan: Comparison of artificial neural network, random forest and logistic regression models[J]. Environmental Geochemistry and Health, 2024, 46(1): 14. DOI: 10.1007/s10653-023-01782-7.

[8] Clemen R T. Combining forecasts: A review and annotated bibliography[J]. International Journal of Forecasting, 1989, 5(4): 559-583. DOI: 10.1016/0169-2070(89)90012-5.

[9] 董欣林, 崔东文. 基于WPT-ITTA-RELM/ELM/LSSVM模型的日径流预测研究[J]. 三峡大学学报(自然科学版), 2024, 46(4): 16-24. DOI: 10.13393/j.cnki.issn.1672-948X.2024.04.003.

[10] Hosseinzadeh Talaee P. Multilayer perceptron with different training algorithms for streamflow forecasting[J]. Neural Computing and Applications, 2014, 24: 695-703. DOI: 10.1007/s00521-012-1287-5.

[11] Liang Xue, Zhu Yuelong. Dynamic hydrological flow prediction with self-iterative spatiotemporal graph neural network: Modeling long- and short-period topological dynamics[J]. Journal of Hydrology, 2010, 663(Part A): 11. DOI: 10.1016/j.jhydrol.2025.134122.

[12] 邹红梅, 朱成涛. 基于LSTM和BP神经网络的水库入库径流中长期预测比较研究[J]. 水文, 2024(4): 6. DOI: 10.19797/j.cnki.1000-0852.20230413.

[13] 胡乐怡, 付晓雷, 蒋晓蕾, 等. 基于LSTM、RF、SVR三种机器学习方法的径流预测研究[J]. 水文, 2024, 44(5): 17-24. DOI: 10.19797/j.cnki.1000-0852.20230347.

[14] Wang Jia, Wang Xu, Lei Xiaohui, et al. Teleconnection analysis of monthly streamflow using ensemble empirical mode decomposition[J]. Journal of Hydrology, 2020, 582: 124411. DOI: 10.1016/j.jhydrol.2019.124411.

[15] 李世林, 王李东, 刘晓阳, 等. 径流式水电站出力预测的学习模型研究[J]. 水利水电技术(中英文), 2025, 56(1): 193-202. DOI: 10.13928/j.cnki.wrahe.2025.01.016.

[16] 孙瑜辉, 王庆杰, 岳春芳. 基于变分模态分解的自适应交叉融合模型及其在月径流预测中的应用[J]. 水电能源科学, 2025, 43(3): 1-6. DOI: 10.20040/j.cnki.1000-7709.2025.20240373.

[17] Kang Yan, Xiang Yue, Zhang Zishang, et al. A novel physical process-ensemble learning model framework with residual error decomposition to upskill daily runoff prediction[J]. Journal of Hydrology, 2025, 651: 132565. DOI: 10.1016/j.jhydrol.2024.132565.

[18] 甘容, 马超鑫, 高勇, 等. 基于二次分解和支持向量机的月径流预测方法[J]. 郑州大学学报(工学版), 2024, 45(6): 32-39. DOI: 10.13705/j.issn.1671-6833.2024.06.003.

[19] Liu Jiadong, Xu Teng, Lu Chunhui. VMDI-LSTM-ED: A novel enhanced decomposition ensemble model incorporating data integration for accurate non-stationary daily streamflow forecasting[J]. Journal of Hydrology, 2025, 653: 132769- 132769. DOI: 10.1016/J.JHYDROL.2025.132769.

[20] Katipoglu M O. Integration of extreme learning machines with CEEMDAN and VMD techniques in the prediction of the multiscalar standardized runoff index and standardized precipitation evapotranspiration index[J]. Natural Hazards, 2024, 120(1): 825-849. DOI: 10.1007/s11069-023-06238-w.

[21] 陈永平, 韩韬, 邱超, 等. 基于调和分析及VMD-LSTM混合模型的甬江河口水位预报方法研究[J]. 河海大学学报(自然科学版), 2025, 53(2): 1-10. DOI: 10.3876/j.issn.1000-1980.2025.02.001.

[22] 梅丹, 张恒. 基于变分模态分解和深度学习算法的污水出水水质预测[J]. 长江科学院学报, 2025, 42(9): 67-74. DOI: 10.11988/ckyyb.20240721.

[23] 李冰艳, 张文煜, 杨成林, 等. 基于VMD和IGWO组合模型的可降水量预测研究[J]. 水文, 2025, 45(1): 68-75. DOI: 10.19797/j.cnki.1000-0852.20240069.

[24] 孙祥瑜, 王超, 杨一旸, 等. CEEMDAN-Pyraformer-LSS模型在永定河径流预测中的应用[J]. 南水北调与水利科技(中英文), 2025, 23(2): 363-374. DOI: 10.13476/j.cnki.nsbdqk.2025.0038.

[25] Berghout T, Benbouzid M, Muyeen S M, et al. Auto-NAHL: A neural network approach for condition-based maintenance of complex industrial systems[J]. IEEE Access, 2021, 9: 152829-152840. DOI: 10.1109/ACCESS.2021.3127084.

[26] Karami H, Anaraki M V, Farzin S, et al. Flow direction algorithm (FDA): A novel optimization approach for solving optimization problems[J]. Computers & Industrial Engineering, 2021, 156: 107224. DOI: 10.1016/j.cie.2021.107224.

[27] 俞阳, 闵雪峰, 王丹, 等. SWAT与GIS耦合的流域多要素水环境管控单元划分[J]. 农业资源与环境学报, 2025, 42(3): 815-824. DOI: 10.13254/j.jare.2024.0101.

[28] 吴小涛, 袁晓辉, 袁艳斌, 等. 基于EEMD-VMD-SSA-KELM模型的汛期日径流预测[J]. 中国农村水利水电, 2023(7): 27-34. DOI: 10.12396/znsd.221872.

[29] Noori N, Kalin L. Coupling SWAT and ANN models for enhanced daily streamflow prediction[J]. Journal of Hydrology, 2016, 533: 141-151. DOI: 10.1016/j.jhydrol.2015.11.050.

[30] Noori N, Kalin L, Isik S. Water quality prediction using SWAT-ANN coupled approach[J]. Journal of Hydrology, 2020, 590: 125220. DOI: 10.1016/j.jhydrol.2020.125220.

基金资助

国家重点研发计划项目(2022YFC3204603)

国家自然科学基金项目(52394234)