K-means Traceability Study of Abnormal Electricity Based on Graph Neural Network

With the advent of the big data era, the analysis of data and the identification of abnormal data have become the core issues. In this paper, for the tasks of abnormal power identification and user localization, we propose a K-means clustering method using the graph-attention self-encoder model for dimensionality reduction to complete the identification of abnormal power. Subsequently, based on the probabilistic graph structure and random walk algorithm, we constructed the user power relationship network to realize the localization of abnormal users and the warning of potential abnormal users. We conducted experimental analysis on the electricity consumption data of a total of more than 300,000 users in Shan dong Province. The experimental results show that the clustering results of our model have significant improvement compared with the results of the original dimensionless model, which can accurately identify the abnormal electricity data and can locate the abnormal users more accurately to accomplish the tasks of abnormal electricity identification and user localization.