Non-invasive load disaggregation method for agricultural power user based on deep learning and improved load behavior correlation graph
Most research works focus on the scenario of urban user,ignoring the relevance between loads of agricul-tural user that leads to a worsen disaggregation for them.This paper proposes a non-invasive load decomposition method for agricultural users based on deep learning and improved load behavior correlation graph.Firstly,one-hot coding was used to construct the characteristic matrix containing discrete and continuous load features.Secondly,the load behavior correlation graph was used to characterize the relationship between loads,and the graph attention mechanism was introduced to optimize the weight of the load correlation.Finally,an agricultural load disaggrega-tion model based on convolutional neural network and long short-term memory is constructed and trained.Simula-tion results show that,compared with the existing methods,the proposed non-invasive load disaggregation method for agricultural users based on deep learning and improved load behavior correlation graph achieves 4.34%and 2.02%load decomposition accuracy respectively,and is more suitable for agricultural electricity scenario.
万方数据
维普
