一种基于EDDL的非侵入式负荷检测模型  

A non-intrusive load detection model based on EDDL

作　　者：李晨 夏立典 章超 叶杨锋 LI Chen;XIA Lidian;ZHANG Chao;YE Yangfeng(Zhejiang Huayun Information Technology Co.,LTD.,Hangzhou Zhejiang 310030,China)

机构地区：[1]浙江华云信息科技有限公司,浙江杭州310030

出　　处：《太赫兹科学与电子信息学报》2023年第11期1381-1386,共6页Journal of Terahertz Science and Electronic Information Technology

摘　　要：针对目前非侵入式负荷检测时存在检测精确度低的问题,提出一种基于事件驱动-深度学习(EDDL)的负荷检测模型。通过零交叉检测电流数据,基于事件驱动机制从大量数据中发现关键事件;将包含关键事件的电流序列转换至图像空间,并代入基于深度学习的负荷检测模型,从而实现端对端的非侵入式负荷检测。实验结果表明,与多分类支持向量机(MSVM)、前馈神经网络(FNN)、卷积神经网络(CNN)和长短时记忆网络(LSTM)模型相比,所提EDDL模型综合性能更优,检测准确率和精确度分别为94.67%和91.76%。仿真结果验证了所提模型可基于事件驱动机制挖掘电流数据,并基于深度学习模型有效提取电流数据特征,从而实现高精确度的非侵入式电力负荷检测。该模型对非侵入式电力负荷检测研究具有一定借鉴作用。A load detection model based on Event Driven and Deep Learning(EDDL)is proposed to address the issue of low detection accuracy in current non-invasive load detection.The current data is detected through zero crossing,and the key events are discovered from a large amount of data based on event driven mechanisms.The end-to-end non-invasive load detection is achieved by converting the current sequence containing key events into image space and incorporating it into a deep learning based load detection model.The experimental results show that compared with the Multi-class Support Vector Machine(MSVM),Feedforward Neural Network(FNN),Convolution Neural Network(CNN),and Long Short Term Memory(LSTM)models,the proposed EDDL model has better overall performance,with detection accuracy and accuracy of 94.67%and 91.76%,respectively.The simulation results verify that the proposed model can mine current data based on event driven mechanisms and effectively extract current data features based on deep learning models,thus achieving high-precision non-invasive power load detection.This model has certain reference value for the research of non-invasive power load detection.

关 键 词：电力系统 非侵入式 负荷检测 深度学习 事件驱动 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]