A Monthly Urban Electricity Consumption Prediction Algorithm Based on Self-Attention

Urban electricity consumption prediction is an important basis for power grid enterprises to formulate power dispatch plans, plan future distribution network development, and promote power reform. However, the temporal characteristics of urban electricity consumption are highly nonlinear and seasonal. Previous urban electricity consumption prediction models do not fully analyze and utilize features. Deep Learning is a type of algorithm that can adaptively and self-organize learn non-linear features of data, and can effectively predict and analyze urban electricity consumption. This paper uses the Self-Attention model to analyze and extract the characteristics of urban electricity consumption data, which is a type of deep learning algorithm. By using the XGBoost algorithm, the prediction results of Self-Attention are integrated with other models to achieve better predictive analysis of urban electricity consumption. Afterwards, the electricity consumption model proposed in this paper is validated using urban electricity consumption data, demonstrating good predictive accuracy of the model.