
1.贵州电网电力调度控制中心,贵州 贵阳 550000
2.中国南方电力调度控制中心,广东 广州 510700
Received:04 July 2026,
Revised:2026-09-11,
Accepted:15 September 2026,
Online First:23 September 2026,
移动端阅览
王融融,苏华英,王凌梓等.耦合气象信息和EEMD-BNN的多核并行中期径流预测方法[J].人民珠江,DOI:10.3969/j.issn.1001-9235.XXXX.XX.001.
Wang Rongrong,Su Huaying,Wang Linzi,et al.Medium-term Runoff Forecasting Method Coupling Meteorological Information and EEMD-BNN with Multicore Parallel Computing[J].PEARL RIVER,
王融融,苏华英,王凌梓等.耦合气象信息和EEMD-BNN的多核并行中期径流预测方法[J].人民珠江,DOI:10.3969/j.issn.1001-9235.XXXX.XX.001. DOI:
Wang Rongrong,Su Huaying,Wang Linzi,et al.Medium-term Runoff Forecasting Method Coupling Meteorological Information and EEMD-BNN with Multicore Parallel Computing[J].PEARL RIVER, DOI:10.3969/j.issn.1001-9235...001.
中期径流预测是水库运行管理与调度决策的重要依据。针对天然径流过程复杂、非平稳,且现有方法在气象信息利用、预测精度和计算效率方面不足的问题,提出一种耦合气象信息、集合经验模态分解-贝叶斯神经网络(EEMD-BNN)与多核并行计算的中期径流预测方法。首先,根据不同变量的水文作用和时间响应特征,对气象信息进行空间加权和分类筛选,结合滞后关系、相关性、显著性及共线性分析确定模型输入。其次,利用EEMD分解径流和气象序列,经频率特征检验后组合频率相近的同阶分量并构建BNN子模型,以减弱序列非平稳性和多频信息干扰。最后,将参数组合和分量模型划分为独立任务,采用多核并行计算完成参数寻优与模型训练。以南方某流域大型水库未来7 d入库径流预测为例,结果表明,所提模型的5项指标均优于传统BNN及仅分解径流的消融模型,气象因子分解使RMSE降低3.5%。模型训练与参数寻优时间由单进程的11 010s缩短至32个进程下的523 s,加速比为21,兼顾了预测精度与计算效率。
Medium-term runoff forecasting is essential for reservoir operation and scheduling
providing important support for water allocation
hydropower generation planning
flood control
and the coordinated operation of reservoir systems. However
natural runoff is affected by complex interactions among antecedent hydrological conditions
precipitation
atmospheric moisture
cloud processes
and other meteorological factors
resulting in pronounced nonlinearity
non-stationarity
and uncertainty. Existing forecasting methods often face three major limitations: insufficient utilization of multi-source meteorological information
interference among signals with different temporal frequencies
and high computational cost during repeated model calibration and parameter optimization. To address these problems
this study proposes a medium-term runoff forecasting method integrating meteorological information
ensemble empirical mode decomposition-Bayesian neural networks (EEMD-BNN)
and multicore parallel computing. First
a meteorological information processing framework is developed by considering both the spatial distribution and temporal response characteristics of meteorological variables. Meteorological data from multiple grid points are spatially weighted
while lag analysis
correlation analysis
significance tests
and multicollinearity analysis are combined to identify effective predictors
thereby improving the representation of hydrological and meteorological interactions while reducing redundant information. Second
EEMD is introduced to jointly process runoff and meteorological series rather than decomposing runoff alone. The decomposed components are analyzed according to their frequency characteristics
and components of the same order with similar temporal scales are jointly fed into the corresponding BNN submodels. This component-wise modeling strategy separates mixed-frequency information into relatively stable subseries
reduces non-stationarity and random interference
and strengthens the ability of the forecasting model to capture relationships between runoff and meteorological signals at different temporal scales. The component forecasts are subsequently reconstructed to obtain the complete multi-step runoff process. Finally
considering the substantial computational burden associated with multiple parameter combinations and component models
independent computational tasks are executed in parallel on multiple processor cores for parameter optimization and model training
improving computational efficiency without changing the forecasting structure. The proposed method is evaluated for seven-day-ahead inflow forecasting at Hongjiadu Reservoir in the Wujiang River basin of southern China. Results show that it consistently outperforms conventional BNN and the ablation model in which only runoff is decomposed. On the test set
the proposed model achieves a correlation coefficient of 0.92 and a Nash-Sutcliffe efficiency coefficient of 0.90. Compared with the ablation model
decomposing the meteorological factors further reduces RMSE by 3.5%. In addition
training and parameter optimization time decreases from 11 010 s with one process to 523 s with 32 processes
corresponding to an approximately 21-fold speedup. The proposed method therefore improves the utilization of meteorological information
runoff forecasting accuracy
and computational efficiency
providing an effective approach for medium-term reservoir inflow forecasting and operational scheduling.
冯仲恺 , 付新月 , 张景帅 , 等 . 水电日前发电能力智能预测方法 [J]. 人民长江 , 2025 , 56 ( 10 ): 214 - 224 . doi: 10.16232/j.cnki.1001-4179.2025.10.028 http://dx.doi.org/10.16232/j.cnki.1001-4179.2025.10.028
Feng Zhongkai , Fu Xinyue , Zhang Jingshuai , et al . Research on intelligent day-ahead hydropower generation ability forecasting method [J]. Yangtze River , 2025 , 56 ( 10 ): 214 - 224 . (in Chinese)
任康 , 周子航 , 李昕媛 , 等 . 西江流域水库群调度对非一致性入库径流的敏感性响应特征 [J]. 水资源保护 , 2025 , 41 ( 6 ): 139 - 148 .
Ren Kang , Zhou Zihang , Li Xinyuan , et al . Sensitivity response characteristics of reservoir group operation in the Xijiang River Basin to non-stationary inflow runoff [J]. Water Resources Protection , 2025 , 41 ( 6 ): 139 - 148 . (in Chinese)
He Shaokun , Chen Kebing , Liu Zhangjun , et al . Exploring the impacts of climate change and human activities on future runoff variations at the seasonal scale [J]. Journal of Hydrology , 2023 , 619 : 129382 .
Guo Shaolei , Wen Yihao , Zhang Xianqi , et al . Runoff prediction of lower Yellow River based on CEEMDAN–LSSVM–GM(1, 1) model [J]. Scientific Reports , 2023 , 13 : 1511 .
Ji Hong kang , Mirzaei M , Lai S H , et al . The robustness of conceptual rainfall-runoff modelling under climate variability–A review [J]. Journal of Hydrology , 2023 , 621 : 129666 .
朱祖龙 , 陈华勇 , 袁仔洋 , 等 . 基于机器学习的堰塞坝溃决峰值流量预测模型 [J]. 人民长江 , 2025 , 56 ( 4 ): 56 - 64 .
Zhu Zulong , Chen Huayong , Yuan Ziyang , et al . Prediction models for peak discharge caused by barrier dam failures based on machine learning [J]. Yangtze River , 2025 , 56 ( 4 ): 56 - 64 . (in Chinese)
廖胜利 , 王在能 , 程春田 , 等 . 耦合气象信息和可解释深度学习方法的日径流预报 [J]. 水利学报 , 2025 , 56 ( 7 ): 831 - 843 .
Liao Shengli , Wang Zaineng , Cheng Chuntian , et al . Coupling meteorological information and interpretable deep learning methods for daily runoff forecasting [J]. Journal of Hydraulic Engineering , 2025 , 56 ( 7 ): 831 - 843 . (in Chinese)
Liao Shengli , Liu Zhanwei , Liu Benxi , et al . Multistep-ahead daily inflow forecasting using the ERA-Interim reanalysis data set based on gradient-boosting regression trees [J]. Hydrology and Earth System Sciences , 2020 , 24 ( 5 ): 2343 - 2363 .
麦紫君 , 曾小凡 , 周建中 , 等 . 基于偏互信息法遴选因子的长江中长期径流预报 [J]. 人民长江 , 2018 , 49 ( 3 ): 52 - 56 .
Mai Zijun , Zeng Xiaofan , Zhou Jianzhong , et al . Medium and long-term runoff forecasting based on input variables selected by partial mutual information [J]. Yangtze River , 2018 , 49 ( 3 ): 52 - 56 . (in Chinese)
邓智予 , 崔东文 . 基于二次分解技术与十种“鸟”群算法优化的OSELM月径流预测 [J]. 人民珠江 , 2025 , 46 ( 11 ): 44 - 54 .
Deng Zhiyu , Cui Dongwen . Improved monthly runoff prediction of OSELM based on secondary decomposition technique and optimization of ten “bird” swarm algorithms [J]. Pearl River , 2025 , 46 ( 11 ): 44 - 54 . (in Chinese)
蔡亮 , 包艳飞 , 崔东文 . EWT多重分解与若干新型元启发式算法优化的多层感知器月径流预测 [J]. 人民珠江 , 2025 , 46 ( 9 ): 72 - 83 .
Cai Liang , Bao Yanfei , Cui Dongwen . Multi-layer perceptron monthly runoff prediction optimized by EWT multiple decomposition and several new metaheuristic algorithms [J]. Pearl River , 2025 , 46 ( 9 ): 72 - 83 . (in Chinese)
Cui Zhongjie , Qing Xiaoxia , Chai Hongxiang , et al . Real-time rainfall-runoff prediction using light gradient boosting machine coupled with singular spectrum analysis [J]. Journal of Hydrology , 2021 , 603 : 127124 .
张闻胜 , 王银堂 . 时间序列分解模型在水文要素中长期预报中的应用 [J]. 水文 , 2001 , 21 ( 1 ): 21 - 24 .
Zhang Wensheng , Wang Yintang . Application of the time series decomposable model in medium and long-term hydrologic forecasting [J]. Journal of China Hydrology , 2001 , 21 ( 1 ): 21 - 24 . (in Chinese)
Wang Wenchuan , Chau K W , Qiu Lin , et al . Improving forecasting accuracy of medium and long-term runoff using artificial neural network based on EEMD decomposition [J]. Environmental Research , 2015 , 139 : 46 - 54 .
Zuo Ganggang , Luo Jungang , Wang Ni , et al . Decomposition ensemble model based on variational mode decomposition and long short-term memory for streamflow forecasting [J]. Journal of Hydrology , 2020 , 585 : 124776 .
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 , 42 ( 8 ): 43 - 47 .
Cheng Jiabo , Ji Houling , Chen Zhongju , et al . Daily runoff prediction based on the GWO-PE-VMD-ResNet combined model [J]. Water Resources and Power , 2024 , 42 ( 8 ): 43 - 47 . (in Chinese)
Siqueira V A , Fan F M , de Paiva R C D , et al . Potential skill of continental-scale, medium-range ensemble streamflow forecasts for flood prediction in South America [J]. Journal of Hydrology , 2020 , 590 : 125430 .
杨强 . 耦合CFS集合预报和深度学习的中长期径流预测方法研究 [D]. 大连 : 大连理工大学 , 2019 .
Yang Qiang . Research on medium and long term runoff forecasting method coupled with climate forecast system and deep learning [D]. Dalian : Dalian University of Technology , 2019 . (in Chinese)
蔺星娜 , 牛健植 , 贾京伟 , 等 . 张家口清水河上游流域植被指数时空变化特征 [J]. 中国水土保持科学 , 2018 , 16 ( 1 ): 123 - 130 .
Lin Xingna , Niu Jianzhi , Jia Jingwei , et al . Spatial-temporal variation of vegetation in Qingshui River Basin, Zhangjiakou [J]. Science of Soil and Water Conservation , 2018 , 16 ( 1 ): 123 - 130 . (in Chinese)
Esha R I , Imteaz M A . Assessing the predictability of MLR models for long-term streamflow using lagged climate indices as predictors: a case study of NSW (Australia) [J]. Hydrology Research , 2019 , 50 ( 1 ): 262 - 281 .
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 .
Ding Cheng , Wang Zhaocai , Ding Weijie , et al . Deep learning ensemble streamflow prediction based on explainable multi-source data feature fusion [J]. Advances in Water Science , 2025 , 36 ( 4 ): 581 - 595 . (in Chinese)
陈小泽 , 王忠静 , 刘丹 , 等 . 径流序列两段动态分解-预测-重构中长期预报模型 [J]. 水利学报 , 2025 , 56 ( 7 ): 933 - 944 .
Chen Xiaoze , Wang Zhongjing , Liu Dan , et al . A two-stage dynamic decomposition-prediction-reconstruction model for medium-long term runoff forecasting [J]. Journal of Hydraulic Engineering , 2025 , 56 ( 7 ): 933 - 944 . (in Chinese)
Schober P , Boer C , Schwarte L A . Correlation coefficients: appropriate use and interpretation [J]. Anesthesia and Analgesia , 2018 , 126 ( 5 ): 1763 - 1768 .
Ahmad K , Iqbal M , Tariq M A U R , et al . Exploring alternate coupling inputs of a data-driven model for optimum daily streamflow prediction in calibrated SWAT-BiLSTM rainfall-runoff modeling [J]. Frontiers in Water , 2025 , 7 : 1558218 .
郭婷婷 , 崔东文 . 基于STL-WPT-RFO-HLSTSVR模型的月径流时间序列预测 [J]. 人民珠江 , 2026 , 47 ( 2 ): 56 - 67 .
Guo Tingting , Cui Dongwen . Monthly runoff time series prediction based on STL-WPT-RFO-HLSTSVR model [J]. Pearl River , 2026 , 47 ( 2 ): 56 - 67 . (in Chinese)
秦泽宁 , 黎曙 , 周祖昊 , 等 . 分布式水文模型区域分解并行计算方法及其应用 [J]. 水电能源科学 , 2020 , 38 ( 10 ): 1 - 4,12 .
Qin Zening , Li Shu , Zhou Zuhao , et al . Domain decomposition parallel computing method of distributed hydrological model and its application [J]. Water Resources and Power , 2020 , 38 ( 10 ): 1 - 4,12 . (in Chinese)
Niu Wenjing , Feng Zhongkai , Feng Baofei , et al . Parallel computing and swarm intelligence based artificial intelligence model for multi-step-ahead hydrological time series prediction [J]. Sustainable Cities and Society , 2021 , 66 : 102686 .
0
Views
0
下载量
0
CSCD
Publicity Resources
Related Articles
Related Author
Related Institution
京公网安备11010802024621