A Novel Two-Stage Forecasting Framework with Difference-Based Learning for Power Load Forecasting

Accurate short-term load forecasting is critical for the safe and efficient operation of modern power systems, especially in the context of increasing demand fluctuations and the growing penetration of renewable energy sources. However, two key challenges hinder the performance of existing forecasting models: the non-stationarity of load sequences and the limited ability to capture complex temporal dynamics. To address these issues, TSD-Net (Two-Stage Decomposition Network) is proposed as a novel forecasting framework, which explicitly targets these problems through staged prediction, progressive refinement, and structured output fusion. The two-stage architecture first captures coarse-grained temporal trends to establish a stable forecasting baseline, followed by targeted adjustments for short-term fluctuations, thereby mitigating the effects of non-stationarity. To enhance the model’s capacity to learn fine-grained load features, TSD-Net incorporates a series of refinement blocks that iteratively extract previously under-learned patterns. A structured fusion module is further employed to integrate outputs from all stages, improving overall predictive accuracy. Extensive experiments on real-world datasets from Shanghai, Australia, and Tetouan demonstrate that TSD-Net consistently delivers superior forecasting performance, offering a robust and generalizable solution for dynamic load prediction in complex power system environments.