Image-Based Electricity Load Clustering with Deep Embedding Clustering

This paper proposes a novel image-based electricity load clustering method. The numerous load data are converted into annual load images instead of conventional mean load curve, which can reflect the daily and seasonal load fluctuations. The convolutional auto-encoder (CAE) is utilized to extract the features of the load images as the input of the clustering model. The deep embedding clustering (DEC) algorithm is employed to cluster the load images. Since the load images can capture more load features than the traditional daily load curves, the image-based clustering method can achieve more accurate result than the conventional ones. The experiments based on the public load data validate the super-performance of the proposed image-based load clustering approach.