Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Deep Learning for Electricity Price Forecasting: A Review of Day-Ahead, Intraday, and Balancing Electricity Markets

Runyao Yu, Derek W. Bunn, Julia Lin, Jochen Stiasny, Fabian Leimgruber, Tara Esterl, Yuchen Tao, Lianlian Qi, Yujie Chen, Wentao Wang, Jochen L. Cremer

TL;DR

This paper addresses the challenge of forecasting electricity prices across the full market stack by introducing a unified backbone-head-loss taxonomy to compare deep learning approaches. It analyzes trends across day-ahead, intraday, and balancing markets, showing a shift toward probabilistic, multi-country, and foundation-style models in day-ahead, increasing focus on microstructure and trajectory forecasting in intraday, and mechanism-aware, market-specific designs in balancing. The contributions include a structured taxonomy for model evaluation, cross-market trend synthesis, and identification of gaps such as limited intraday and balancing market coverage and the need for market-informed modeling. The findings underscore the practical impact of design choices on forecasting uncertainty, cross-border coordination, and decision-making in power systems.

Abstract

Electricity price forecasting (EPF) plays a critical role in power system operation and market decision making. While existing review studies have provided valuable insights into forecasting horizons, market mechanisms, and evaluation practices, the rapid adoption of deep learning has introduced increasingly diverse model architectures, output structures, and training objectives that remain insufficiently analyzed in depth. This paper presents a structured review of deep learning methods for EPF in day-ahead, intraday, and balancing markets. Specifically, We introduce a unified taxonomy that decomposes deep learning models into backbone, head, and loss components, providing a consistent evaluation perspective across studies. Using this framework, we analyze recent trends in deep learning components across markets. Our study highlights the shift toward probabilistic, microstructure-centric, and market-aware designs. We further identify key gaps in the literature, including limited attention to intraday and balancing markets and the need for market-specific modeling strategies, thereby helping to consolidate and advance existing review studies.

Deep Learning for Electricity Price Forecasting: A Review of Day-Ahead, Intraday, and Balancing Electricity Markets

TL;DR

This paper addresses the challenge of forecasting electricity prices across the full market stack by introducing a unified backbone-head-loss taxonomy to compare deep learning approaches. It analyzes trends across day-ahead, intraday, and balancing markets, showing a shift toward probabilistic, multi-country, and foundation-style models in day-ahead, increasing focus on microstructure and trajectory forecasting in intraday, and mechanism-aware, market-specific designs in balancing. The contributions include a structured taxonomy for model evaluation, cross-market trend synthesis, and identification of gaps such as limited intraday and balancing market coverage and the need for market-informed modeling. The findings underscore the practical impact of design choices on forecasting uncertainty, cross-border coordination, and decision-making in power systems.

Abstract

Electricity price forecasting (EPF) plays a critical role in power system operation and market decision making. While existing review studies have provided valuable insights into forecasting horizons, market mechanisms, and evaluation practices, the rapid adoption of deep learning has introduced increasingly diverse model architectures, output structures, and training objectives that remain insufficiently analyzed in depth. This paper presents a structured review of deep learning methods for EPF in day-ahead, intraday, and balancing markets. Specifically, We introduce a unified taxonomy that decomposes deep learning models into backbone, head, and loss components, providing a consistent evaluation perspective across studies. Using this framework, we analyze recent trends in deep learning components across markets. Our study highlights the shift toward probabilistic, microstructure-centric, and market-aware designs. We further identify key gaps in the literature, including limited attention to intraday and balancing markets and the need for market-specific modeling strategies, thereby helping to consolidate and advance existing review studies.
Paper Structure (19 sections, 4 equations, 2 figures, 2 tables)

This paper contains 19 sections, 4 equations, 2 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Deep Learning Components
  3. Backbone
  4. Head
  5. Loss Function
  6. Trends in Electricity Price Forecasting
  7. Day-Ahead Market
  8. Evolution of model families
  9. Preference for multi-timestep heads
  10. Shift from pointwise to probabilistic setting
  11. Intraday Market
  12. From macro-level features to microstructure
  13. From single-index prediction to trajectory forecasting
  14. Limited accessibility and synthetic data
  15. Balancing Market
  16. ...and 4 more sections

Figures (2)

  • Figure 1: Visualization of day-ahead, intraday, and imbalance prices from 1-15 October 2025 in Austria. The imbalance price exhibits the highest volatility, the day-ahead price is the smoothest, and the intraday price lies in between.
  • Figure 2: Backbone comparison of deep learning models.