摘要
针对电力负荷监测与电能计量数据的多源异构特征,提出基于深度动态概率图模型的数据融合方法。通过构建时空校准机制,消除设备间的采样异步问题。设计贝叶斯神经网络来量化测量不确定性,并引入动态门控单元实现多维度数据的自适应加权。开发PowerFusion边缘云协同系统,集成数据清洗、不确定性建模与融合决策模块。实验结果表明,该系统在复杂工况下显著提升了数据一致性与物理可解释性,为智能电网提供了高精度的量测融合解决方案。
In view of the multi-source heterogeneous characteristics of power load monitoring and electric energy metering data,a data fusion method based on a deep dynamic probabilistic graph model is proposed.By constructing a spatio-temporal calibration mechanism,the problem of asynchronous sampling between devices is eliminated.A Bayesian neural network is designed to quantify measurement uncertainties,and a dynamic gating unit is introduced to achieve adaptive weighting of multidimensional data.The PowerFusion edge-cloud collaborative system is developed,with data cleaning,uncertainty modeling,and fusion decision-making modules integrated into it.Experimental results show that under complex working conditions,the data consistency and physical interpretability of this system are significantly improved,providing a high-precision measurement fusion solution for smart grids.
作者
王佩霞
王震
WANG Peixia;WANG Zhen(Xuejiawan Power Supply Branch,Inner Mongolia Power(Group)Co.,Ltd.,Ordos,Inner Mongolia 010700,China)
出处
《自动化应用》
2025年第18期177-180,共4页
Automation Application
