Special transformer user load identification method based on deep clustering

Aiming at the limitations of traditional clustering methods in dealing with high-dimensional time series data, this paper proposes a deep clustering method to improve the load identification ability of special transformer users. Firstly, the load data of special transformer users are normalized to improve the convergence speed and performance of the model. Secondly, a deep clustering algorithm combining one-dimensional convolutional autoencoder and fuzzy C-means clustering is used to cluster the user load curves of special transformers and extract the typical load curves of special transformer users. Finally, experiments show that the proposed method can effectively recognize the load of an unbalanced power consumption data set. The accuracy of the special transformer user load identification The accuracy of the user load identification method based on deep clustering is as high as $93.2 \%$, which is suitable for the actual user load identification of special transformers.