基于在线压缩重构的非侵入式电力负荷监测

Non-intrusive load monitoring via online compression and reconstruction

对家庭场景下电力负荷的精确监测常依赖复杂庞大的算法模型,难以在边缘设备中部署。同时,海量电力数据给电力网络通信带来了巨大的挑战。本文针对以上问题,提出了一种分布式非侵入式电力负荷监测方法。通过基于注意力机制的长短时记忆网络(LSTM)负荷监测算法计算并识别负荷设备运行状态,借助云边协同技术将负荷监测任务分布式部署于云端以及边缘端中,解决边缘算力不足的问题。针对云边通信带来的高网络带宽需求,通过基于K奇异值分解(K-SVD)双稀疏在线字典学习的压缩感知方法对负荷信号进行压缩和重构,有效缓解通信资源紧张。对比不同负荷场景下监测算法的表现,结果表明:本文负荷监测算法针对不同的负荷类型均可以保持95%以上的准确率。设计实验验证了本文压缩感知方法对电力负荷信号压缩的有效性,确定负荷数据无失真压缩感知最大压缩比。

Accurate monitoring of power loads in home scenarios often relies on complex algorithmic models that are difficult to deploy in edge devices.At the same time,massive power data poses a huge challenge to communication of grid.In response to the above issues,this paper proposes a distributed nonintrusive load monitoring method.This method calculates and identifies the operating state of the load by the LSTM load monitoring algorithm based on the attention mechanism,and distributes the load monitoring task in the cloud and the edge with the help of cloud-edge collaboration technology to solve the problem of insufficient edge computing power.Aiming at the high network bandwidth requirements brought about by cloud-side communication,the compressed sensing method based on K-SVD double sparse online dictionary is used to compress and reconstruct the load signal,which effectively alleviates the shortage of communication resources.Comparing the performance of the monitoring algorithm under different load scenarios,the results show that the load monitoring algorithm in this paper can maintain an accuracy rate of more than 95%.Experiments are conducted to verify the effectiveness of the compressed sensing method on the compression of load signals,and determine the maximum compression ratio of load data without distortion.