基于智能组合模型的大坝变形预测预报研究  

Studies on dam deformation prediction and warning based on an intelligent combination model

作　　者：李双平 刘祖强 张斌 郑俊星 王华为 李永华 苏森南 Li Shuangping;Liu Zuqiang;Zhang Bin;Zheng Junxing;Wang Huawei;Li Yonghua;Su Sennan

机构地区：[1]长江空间信息技术工程有限公司(武汉),武汉430010 [2]湖北省水利信息感知与大数据工程技术研究中心,武汉430010 [3]国家大坝安全工程技术研究中心,武汉430010 [4]华中科技大学,武汉430074

出　　处：《中国水利》2025年第2期65-72,共8页China Water Resources

基　　金：国家重点研发计划(2022YFC3005504)。

摘　　要：针对数字孪生水利工程大坝安全“四预”中预测预报的准确性和可靠性需求,提出智能组合模型系统方法。该方法通过深入分析大坝变形的多重影响因素,结合信号处理技术,智能地分离出主导性的大坝变形趋势分量。随后,采用智能算法精确匹配最优拟合模型,并结合灰色模型、时间序列模型及神经网络等多种建模技术,构建了一个高度集成、具有自适应能力的智能组合模型。通过丹江口大坝变形时间序列的训练和优化,并与传统统计模型预测结果对比验证,实验表明,智能组合模型在预测精度、数据适应性和鲁棒性方面具有显著优势,尤其是在处理非线性关系和长时序依赖性方面表现突出,同时有效提升了外延预测精度和泛化能力。此外,该模型能够提前1个周期(时长为一年)准确预测大坝关键部位的潜在变形趋势,为工程人员提供充足的时间采取预防措施,减少潜在风险。利用智能组合模型开展丹江口大坝的变形预测与预报,不仅提升了监测系统的智能化水平,还为大坝的安全评估、风险预警和科学管理提供了有力技术支撑。To satisfy the requirements of accuracy and reliability of“forecasting,early-warning,rehearsal and emergency planning”for dam safety in digital twin water resources project construction,an intelligent combination model was developed.This method separates the dominant trend component of dam deformation by evaluating multiple influential factors and combining signal processing technology.Intelligent algorithms were then used to accurately match the optimal fitting model.Various modeling technologies including grey model,time series model and neural networks were adopted to build a highly integrated and adaptive intelligent combination model.Through training and optimization of deformation time series of the Danjiangkou Dam,and comparing with the prediction results of traditional statistical models,the experiment indicates that the intelligent combination model has significant advantages in prediction accuracy,data adaptability and robustness,especially in dealing with nonlinear relationships and long-term dependencies.At the same time,it effectively improves the accuracy of extended prediction and generalization ability.In addition,the model can accurately predict the potential deformation trend of key parts of the dam one cycle in advance(with a duration of one year),providing sufficient time for prevention and risk reduction.Application of the model for deformation prediction and forecasting of the Danjiangkou Dam provides a strong support for dam safety assessment,risk warning and scientific management.

关 键 词：智能组合模型 泛化能力 鲁棒性 大坝变形 预测预报 

分 类 号：TV698.2[水利工程—水利水电工程]