基于卷积神经网络的非侵入式负荷识别  

Non-Intrusive Load Recognition based on Convolutional Neural Networks

作　　者：张碧宁 孙皓月[1] 余志成 ZHANG Bining;SUN Haoyue;YU Zhicheng(Hebei University of Architecture,Zhangjiakou,Hebei 075000)

机构地区：[1]河北建筑工程学院,河北张家口075000

出　　处：《河北建筑工程学院学报》2023年第4期233-239,共7页Journal of Hebei Institute of Architecture and Civil Engineering

基　　金：河北省高校基本科研业务费研究项目(基于机器学习的办公楼用电设备非侵入式监测研究2022CXTD09);河北建筑工程学院研究生创新基金项目(基于机器学习的办公楼用电设备非侵入式监测研究XY202236)。

摘　　要：针对非侵入式负荷识别中识别精度低的问题,提出了一种基于卷积神经网络的非侵入式负载识别方法。通过UK-DALE数据集进行测试和训练,经过数次模拟实验,与传统机器学习对比后,有效地提升负荷识别的正确率,具有良好的抗噪性和鲁棒性,有良好的发展和应用前景。Based on convolutional neural network,a non-intrusive load identification method is proposed,which is mainly aimed at the problem of low recognition accuracy in non-intrusive load identification.The UK-DALE data set is used for testing and training.After several simulation experiments,it is verified that the method can effectively improve the accuracy of load identification,which has good anti-noise and robustness,with good development and application prospects.

关 键 词：非侵入式 负荷识别 卷积神经网络 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]