Transformer model for behavior-driven seasonal electricity consumption forecasting
Аннотация
Electricity consumption forecasting (ECF) is crucial for effective energy management and requires powerful models that allow accurate prediction of future electricity demand. The latter depends on seasonal variations influenced by factors such as consumer behavior, weather conditions, and extreme temperatures. Accounting for these seasonal fluctuations enables more accurate predictions of consumption patterns, improving the energy management process. In this paper, we propose a Transformer-based forecasting approach to predict electricity consumption using a real-world dataset, leveraging consumers’ seasonal behavior. This method is integrated into a novel Internet of Behavior (IoB) system that harnesses consumer behavior insights to provide personalized feedback, optimize electricity usage, and enable adaptive demand management tailored to individual needs. The performance of the proposed approach is evaluated against five existing baseline models, which achieves the lowest Root Mean Squared Error (RMSE) of 0.01862 GWh and the lowest Mean Absolute Error (MAE) of 0.00746 GWh. By improving forecast accuracy, this study contributes to optimizing energy management, reducing costs, and facilitating decision-making for companies operating in regions with diverse consumption behaviors.