Running summarizing tools on a new article

Abstract: Accurate load forecasting is essential for the reliable and economic operation of modern power systems. However, the growing integration of renewable energy sources and electric vehicles has introduced substantial uncertainty into load dynamics, making traditional deterministic forecasting methods increasingly insufficient. To address this issue, this paper proposes a probabilistic load forecasting framework based on a fractal-guided Bayesian neural network (BNN). First, the internal temporal components of the load series, including level, growth, and seasonal information, are extracted through dedicated attention-based modules. Meanwhile, multifractal detrended fluctuation analysis is employed to derive compact statistical descriptors that quantify finite-scale heterogeneity, persistence, and scaling behavior over the observed range. These temporal and fractal features are then fused through a cross-attention mechanism and fed into a BNN for probabilistic load forecasting. In addition, SHAP-based analysis is introduced to improve model interpretability and quantify the contributions of different feature components. Experimental results on multiple datasets demonstrate that the proposed method improves CRPS by approximately 15% to 50% and reduces PINAW by up to 40% compared with strong baseline models.