Load Curve Classification Method Based on Convolutional Neural Network

Load classification plays a pivotal role in analyzing user electricity consumption behavior characteristics, enabling more refined load modeling. This paper proposes a method for load curve classification by converting load information into two-dimensional (2D) data and utilizing a 2D convolutional neural network (CNN) for feature extraction from load curves, achieving promising classification results. Initially, the CNN is supervisedly trained using existing training samples, followed by fixing the parameters of the convolutional layers. Subsequently, the samples to be classified are input into the trained CNN, and the output of the last fully connected layer is extracted and clustered to generate pseudo-labels. These pseudo-labels, paired with the samples to be classified, constitute a set of training samples. Finally, the newly generated training samples are combined with the original training samples to form an updated training dataset for alternative training of the CNN. Case studies demonstrate the effectiveness of the proposed load classification method.