基于深度学习网络的非侵入式负荷分解方法
Non-Intrusive Load Decomposition Method Based on Deep Learning Network
侯林超 1朱武 1汤德清1
作者信息
- 1. 上海电力大学电子与信息工程学院,上海 200092
- 折叠
摘要
负荷特征的选取能够在很大程度上描述用电设备的电气特征,选择合适的负荷特征是影响负荷分解优劣的关键条件之一,提出基于深度学习网络的非侵入式负荷分解方法.选取有功功率、无功功率和V-I轨迹图作为负荷特征项并获取相关特征,构建GA-BP神经网络和卷积神经网络并加以训练.提取非侵入式负荷的高级特征,并将特征提取结果融合为复合特征.将融合后的复合特征输入至经过训练的GA-BP 神经网络用于负荷分解,实现非侵入式负荷分解.实验结果表明,所提方法的绝对误差均值、均方根误差较小,召回率较大,通过特征处理有效提升了负荷分解效果.
Abstract
The selection of load characteristics can describe the electrical characteristics to a great extent.In this paper,a non-intrusive load decomposition method based on deep learning networks was proposed.Firstly,the active power,reactive power and V-I trajectory were selected as load characteristic items,and then relevant characteristics were obtained.Secondly,a GA-BP neural network and a convolution neural network were constructed and trained.Thirdly,high-level features of non-intrusive load were extracted,and then the feature extraction results were fused into a composite feature,and then the feature was input into the trained GA-BP neural network for load decomposi-tion.Finally,the non-intrusive load decomposition was achieved.Experimental results show that the proposed method has small absolute error mean and root mean square error,as well as high recall rate.And the load decomposition effect is effectively improved by processing features.
关键词
非侵入式负荷分解/轨迹图/神经网络/遗传算法/卷积神经网络Key words
Non-intrusive load decomposition/Trajectory graph/Neural network/Genetic algorithm/Convolu-tional neural network引用本文复制引用
出版年
2024
NETL
NSTL
万方数据