DALNet: A Denoising Diffusion Probabilistic Model for High-Fidelity Day-Ahead Load Forecasting

TL;DR

DALNet introduces a conditional denoising diffusion probabilistic model to probabilistic day-ahead load forecasting by generating the entire load curve rather than predicting point values, addressing error accumulation and forecasting lag. It couples a bespoke Denoising Network with Temporal Multi-scale Attention (TMSAB) to capture intra-day temporal and positional dependencies, and validates generated curves with KDE and KL divergence to ensure distributional fidelity. Empirical results on the GEFcom2014 dataset show DALNet achieving superior sharpness and reliability metrics compared to MLP, LSTM, Transformer, and BNN baselines, with strong performance particularly at common probabilistic coverage levels. The work highlights the practicality of curve-generation diffusion models for PDALF and suggests broader applicability to predictive tasks in power systems.

Abstract

Accurate probabilistic load forecasting is crucial for maintaining the safety and stability of power systems. However, the mainstream approach, multi-step prediction, is hindered by cumulative errors and forecasting lags, which limits its effectiveness in probabilistic day-ahead load forecasting (PDALF). To overcome these challenges, we introduce DALNet, a novel denoising diffusion model designed to generate load curves rather than relying on direct prediction. By shifting the focus to curve generation, DALNet captures the complex distribution of actual load time-series data under specific conditions with greater fidelity. To further enhance DALNet, we propose the temporal multi-scale attention block (TMSAB), a mechanism designed to integrate both positional and temporal information for improved forecasting precision. Furthermore, we utilize kernel density estimation (KDE) to reconstruct the distribution of generated load curves and employ Kullback-Leibler (KL) divergence to compare them with the actual data distribution. Experimental results demonstrate that DALNet excels in load forecasting accuracy and offers a novel perspective for other predictive tasks within power systems.

Figures (8)

• Figure 1: Overview of the conditional diffusion probabilistic models.
• Figure 2: Structure of the proposed denoising network DALNet.
• Figure 3: Illustration of three attention maps. (a): Global attention map; (b): Window attention map with window size $w=2$; (c) Dilated window attention map with dilation $d=1$.
• Figure 4: Visualization of the TMSAB with the sequence length $l$.
• Figure 5: Probabilistic forecasting results of the DALNet for the three loads under three different PINCs.