基于卷积块注意力模型的非侵入式负荷分解算法

Non-intrusive Load Disaggregate Algorithm Based on Convolutional Block Attention Module

非侵入式负荷分解技术可以深度挖掘用户内部用电数据、获取用电信息,具有广阔的应用前景。该文研究了非侵入式负荷监测模式下基于卷积块注意力模型的非侵入负荷辨识算法。该模型首先采用一种序列到点方法,以电源窗口为输入,目标设备中点为输出,然后使用卷积块注意力模型来训练学习目标设备特征,卷积块注意力模块通过引入空间和通道Attention机制可以有效提取有利特征,丢弃无用特征。基于REDD数据集的实验结果表明,该文模型在有效减少训练时间的前提下,显著提升了分解准确率。

The non-intrusive load disaggregation technology can dig deeply into users’ internal electricity consumption data and obtain electricity consumption information, which has a broad application prospects. This paper studies the non-intrusive load identification algorithm based on the convolutional attention block module in the non-intrusive load monitoring mode. The model first adopts a sequence-to-point method. With the power window as the input and the midpoint of the target device as the output, this method uses the convolutional block attention module to train the model to learn the characteristics of the target device. The convolution block attention module can effectively extract the favorable features and discard the useless ones by introducing the spatial and channel Attention mechanism. The experimental results based on the REDD data set show that the model in this paper significantly improves the decomposition accuracy under the premise of effectively less training time.