Abstract
Due to building thermal inertia and delayed user behavioral responses, power load often lags behind meteorological changes, particularly drops in temperature and humidity. Existing models tend to overlook this dynamic lag relationship. To address these challenges, this study proposes a power load forecasting model that integrates three key mechanisms: a dynamic forward lag mechanism, a turning point attention mechanism, and a temporal and channel hybrid mechanism. These components collectively enable the forecasting model to adaptively align meteorological lags, focus on abrupt load transitions, and capture both global temporal dependencies and local feature interactions, thereby enhancing its ability to model complex load behaviors and improve prediction accuracy. Seven experiments were conducted using load data from Singapore and Calgary, Canada, to validate the effectiveness of the proposed model. Numerical results show that the proposed method achieves higher forecasting accuracy compared to existing load forecasting approaches.
Funding Information
  • National Natural Science Foundation of China (52377076)

This publication has 34 references indexed in Scilit: