融合物理机制深度学习的混凝土浇筑仓温度监测数据去噪方法  

A Method for Denoising Temperature Monitoring Data of Concrete Pouring Block Using Deep Learning Combined with Physical Mechanisms

作　　者：李如尧 张毅 何瑞良 张磊[1] LI Ruyao;ZHANG Yi;HE Ruiliang;ZHANG Lei(State Key Laboratory of Basin Water Cycle Simulation and Regulation,China Institute of Water Resources and Hydropower Research,Beijing 100038,China;Batang(Lava)Branch,Huadian Jinsha River Upstream Hydropower Development Co.,Ltd.,Changdu 854085,Xizang,China)

机构地区：[1]中国水利水电科学研究院流域水循环模拟与调控国家重点实验室,北京100038 [2]华电金沙江上游水电开发有限公司巴塘(拉哇)分公司,西藏昌都854085

出　　处：《水力发电》2025年第4期48-53,90,共7页Water Power

基　　金：国家自然科学基金资助项目(51779277);国家重点研发计划项目(2018YFC0406703);流域水循环模拟与调控国家重点实验室资助项目(SKL2020ZY10,SS0112B102016)。

摘　　要：考虑到混凝土浇筑仓温度监测数据背后遵循的物理机制,提出了一种基于物理信息神经网络的混凝土浇筑仓温度监测数据去噪方法,即利用混凝土浇筑仓温度时空变化所遵循的热传导方程,在常规数据驱动的神经网络拟合混凝土浇筑仓温度监测数据去噪的基础上,增加混凝土浇筑仓温度时空变化控制方程的约束来控制神经网络函数拟合的光滑性和合理性,从而实现混凝土浇筑仓温度监测数据的光滑去噪处理。通过工程实例分析对该方法进行验证,结果表明:使用该方法对大坝混凝土浇筑仓温度监测数据进行去噪处理,即保留了数据的局部特征,整体上又比较光滑。该方法能够有效地去除混凝土浇筑仓温度监测数据中的噪声,且具备较好的泛化拟合能力,为混凝土浇筑仓温度监测数据光滑去噪处理提供了一种新思路。Considering the physical mechanism behind the temperature monitoring data of concrete pouring block,a data denoising method based on physical information neural network is proposed,in which,the heat conduction equation followed by the spatio-temporal variation of concrete pouring block temperature is used,and on the basis of the conventional data-driven denoising of concrete pouring block temperature monitoring data by neural network fitting,the constraints of the control equation of the spatio-temporal variation of the concrete pouring block temperature are added to control the smoothness and reasonableness of the neural network function fitting,so as to realize the smooth denoising of the temperature monitoring data of the concrete pouring block.The method is validated through the analysis of engineering examples,and the results show that using this method to denoise the temperature monitoring data of the dam concrete pouring block preserves the local features of the data while overall being relatively smooth.The method can effectively remove the noise in the temperature monitoring data of concrete pouring block,and has a good generalization fitting ability,which provides a new idea for the smooth denoising processing of the temperature monitoring data of concrete pouring block.

关 键 词：混凝土浇筑仓温度监测 监测数据 数据去噪 物理机制 物理信息神经网络 

分 类 号：TV523[水利工程—水利水电工程] TV431