Electricity price forecasting across Norway's five bidding zones in the post-crisis era

2026-04-29 · Source: Machine Learning · Field: Energy & Utilities — Energy Markets & Policy, Artificial Intelligence & Machine Learning, Data Science & Analytics · Depth: Advanced, quick

Summary

A comprehensive evaluation of electricity price forecasting models was conducted across all five Norwegian Nord Pool bidding zones, addressing the altered price formation post-2021-2022 energy crisis and increased integration with Continental Europe. Researchers developed a multimodal hourly dataset from 2019 to 2025 and tested eight forecasting model families, including LightGBM, ARX, and deep learning architectures. Using strictly causal test sets, robust rolling-origin backtesting, leave-one-group-out feature ablation, and conditional regime analysis, the study found that LightGBM consistently delivered the best performance across all zones, with Mean Absolute Error (MAE) between 1.64 and 5.74 EUR/MWh. The ridge ARX model proved competitive in northern zones. While lagged prices and calendar variables often sufficed for high accuracy, external features like reservoir levels and gas prices were critical for understanding forecast errors during stressed market conditions.

Key takeaway

For research scientists developing electricity price forecasting models in hydropower-dominated markets, you should prioritize LightGBM due to its superior performance across diverse zones. However, ensure your models incorporate external features like reservoir levels and gas prices, as these are vital for interpreting forecast errors and maintaining accuracy during volatile market regimes, even if simpler features provide baseline accuracy.

Key insights

LightGBM excels in Norwegian electricity price forecasting, but external features are crucial for understanding errors in stressed markets.

Principles

• Hydropower dominance impacts price formation.
• Lagged prices are strong predictors.
• External features stratify forecast errors.

Method

The study used a multimodal hourly dataset, evaluated eight model families, and applied rolling-origin backtesting, feature ablation, and conditional regime analysis for robust electricity price forecasting.

In practice

• Prioritize LightGBM for Norwegian zones.
• Include reservoir levels for error analysis.
• Monitor gas prices for market stress.

Topics

• Electricity Price Forecasting
• Norwegian Bidding Zones
• LightGBM
• ARX Model
• Feature Ablation

Best for: Research Scientist, Data Scientist, Machine Learning Engineer, AI Scientist

Related on AIssential

• Electricity price forecasting across Norway's five bidding zones in the post-crisis era — LightGBM excels in Norwegian electricity price forecasting, but external factors are vital for accuracy under market stress.
• Investigating Calibration Challenges in Probabilistic Electricity Price Forecasting — Current electricity price forecasting prioritizes sharpness over calibration, yielding unreliable uncertainty estimates.
• Assessing the Performance-Efficiency Trade-off of Foundation Models in Probabilistic Electricity Price Forecasting — TSFMs generally outperform task-specific models in PEPF, but well-configured conventional models remain highly competitive.
• PeakFocus: Bridging Peak Localization and Intensity Regression via a Unified Multi-Scale Framework for Electricity Load Forecasting — PeakFocus unifies peak timing localization and intensity regression to overcome smoothing and misalignment in electricity load forecasting.
• Product Forecasting through Time Series Analysis (Modelling) — Effective product sales forecasting requires a hierarchical, multi-model approach combining ML, deep learning, and statistical methods.
• Beyond Polarity: Multi-Dimensional LLM Sentiment Signals for WTI Crude Oil Futures Return Prediction — Multi-dimensional LLM sentiment, beyond polarity, improves WTI crude oil futures return prediction.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Machine Learning.