Non-intrusive Load Monitoring Based on Self-attention

Abstract

Non-intrusive load monitoring is an important means to monitor the power consumption of end-user equipment in real time and timely. At present, there are still some problems in non-invasive load identification, such as the analysis of feature quantity is not detailed enough, the subjectivity of feature selection is strong, and the effect of identification algorithm is not good because it does not fully consider the information expression difference of feature quantity. In order to mine local features and the advantage of the latter in capturing the global relationship between sequence elements, a deep self-attention network method based on self-attention calculation is proposed. The research shows that this method has higher accuracy in solving the problem of load decomposition, and the training speed is faster than the traditional deep learning method. It performs well in the power decomposition of continuous variable state equipment and multi state equipment.