摘要
径流集合预报在水资源决策管理中具有重要作用。现有预报系统的性能受限于输入数据、模型参数和结构的不确定性,且预报精度与集合规模密切相关。为了解决此问题,通过耦合信号分解技术与人工智能预报模型,采用CEEMDAN分解技术处理非平稳径流序列,结合人工神经网络和支持向量机进行子序列预报,通过多模型、多参数预报结果随机组合扩展预报样本集形成径流集合预报结果,最终提出一种基于信号分解的多模型多参数集合径流预报方法。雅砻江锦屏一级水库入库径流验证表明:该方法径流预报结果的Nash可提升至0.84,集合预报的覆盖率达到55%,同时将集合规模扩展至1042量级,显著优于传统多模型方法。该方法可精准提取径流序列特征周期项与趋势项,有效提高了预报精度并降低预报的不确定性,可适用于非线性非平稳水文序列预报,可为水库调度决策提供支撑。
Ensemble runoff forecasting plays a crucial role in water resources decision-making and management.The performance of existing forecasting systems is constrained by uncertainties in input data,model parameters,and structures,and the forecasting accuracy is closely related to the ensemble size.To address this issue,this study proposes a signal decomposition-based multi-model and multi-parameter ensemble runoff forecasting method by coupling signal decomposition techniques with artificial intelligence forecasting models.The CEEMDAN decomposition technique is used to process non-stationary runoff sequences,and artificial neural networks(ANN)and support vector machines(SVM)are combined for subsequence forecasting.The forecasting sample set is expanded through the random combination of multi-model and multi-parameter forecasting results to form ensemble runoff forecasting results.Validation using the inflow runoff of the Jinping-I Reservoir on the Yalong River shows that the Nash coefficient of the runoff forecasting results by this method can be increased to 0.84,the coverage rate of the ensemble forecasting reaches 55%,and the ensemble size is expanded to the order of 1042,significantly outperforming traditional multi-model methods.This method can accurately extract the characteristic periodic and trend terms of runoff sequences,effectively improve forecasting accuracy,and reduce forecasting uncertainties.It is suitable for forecasting non-linear and non-stationary hydrological sequences and can provide support for reservoir operation decisions.
作者
李海辰
高洁
刘孟孟
赵增海
益波
朱方亮
郭鹏
张东
张娉
王旭
LI Haichen;GAO Jie;LIU Mengmeng;ZHAO Zenghai;YI Bo;ZHU Fangliang;GUO Peng;ZHANG Dong;ZHANG Ping;WANG Xu(Key Laboratory of Water Safety Assurance in Beijing-Tianjin-Hebei Region,Ministry of Water Resources,Beijing 100038,China;China Institute of Water Resources and Hydropower Research,Beijing 100038,China;HydroChina Planning and Design(Hydroelectric and Hydraulic Planning and Design Institute),Beijing 100011,China;Institute of Marine Energy and Intelligent Construction,Tianjin University of Technology,Tianjin 300384,China;Northwest Engineering Corporation Limited,PowerChina,Xi'an 710065,China)
出处
《西北水电》
2025年第6期4-14,共11页
Northwest Hydropower
基金
国家自然科学基金项目(U2243232)
国家自然科学基金青年科学基金项目(52309031)
中国电力建设股份有限公司核心攻关项目(DJ-HXGG-2024-02)。
