Price co-movement and price information diverge across Nordic electricity zones after the energy crisis

My Thi Diem Phan, Tuyen Trung Truong, Hoai Phuong Ha, Dat Thanh NguyenarXiv 2026 · arXiv

TL;DR

The 2021-2022 energy crisis permanently split Nordic electricity zones into two groups: those that move in lockstep with continental Europe and those that remain isolated. This study reveals a surprising divergence: while price movements have split, the information needed to forecast those prices has not—a zone's coupling to the continent is a poor predictor of its forecasting needs, which are instead driven by the local generation mix (hydro vs. wind/solar).

What Problem Does This Paper Address?

Electricity markets are undergoing two simultaneous shifts: increased physical interconnection (market coupling) and a rising share of variable renewable energy (VRE) like wind and solar. Traditionally, researchers used price convergence as a proxy for market integration. However, the 2021-2022 energy crisis created a "regime break" where northern Nordic zones became insulated from continental price shocks due to grid bottlenecks, while southern zones remained tightly coupled.

The authors ask: Does this divergence in price movement change the information carried by prices? Specifically, if a zone is "coupled" to Germany, does it suddenly need German wind forecasts or neighboring price data to achieve accurate forecasts? Understanding this helps System Operators and traders decide which forecasts are worth producing and publishing before the day-ahead auction closes.

Figure 1: Bidding zones and study design. (a) The zones as a network. Nodes are placed by hand at approximate zone locations and coloured by market type from the 2024 coupling measure. Grey nodes are price-only zones, and Great Britain (open) is not in the data. All links are drawn with the same width. The four named interconnectors from NO2 are NordLink to DE-LU and North Sea Link to Great Britain (1,400 MW each) [4], NorNed to NL (700 MW) [26] and Skagerrak to DK1 (about 1,600 MW) [27]. NordLink is drawn curved only for legibility. The other lines mark neighbouring areas with a direct link [28]. Links to Finland, the Baltic states and Poland are not shown. Map lines delineate study areas and do not necessarily depict accepted national boundaries. Base map from Natural Earth, 1:50m countries (public domain). (b) Study design, with the section of each experiment.

Methods

The researchers employed a rigorous out-of-sample forecasting framework across 10 bidding zones.

  • Co-movement Analysis: They measured the correlation of intraday price profiles between Nordic zones and the Germany-Luxembourg (DE-LU) reference zone from 2018 to 2025.
  • Forecasting Model: They utilized a per-hour LightGBM model. To handle the extreme price volatility of the crisis years, they implemented a "level-and-scale" target normalization. Instead of predicting absolute prices, the model predicts the deviation from the previous day’s mean, scaled by the previous day’s standard deviation. This allows the model to learn from "expensive" crisis data and apply those patterns to "cheaper" periods.
  • Information Value: Following the predictive logic of Granger causality, they measured the "value" of an input (like VRE forecasts or neighboring prices) by observing the increase in Mean Absolute Error (MAE) when that input was removed.
  • Double Machine Learning (DML): This was used to diagnose which market conditions (e.g., reservoir levels, temperature) were associated with higher forecast errors.

Datasets

The study uses a comprehensive dataset spanning 2018–2025, with 2024 used for model screening and 2025 for the final rolling test.

  • Price Data: Day-ahead prices for 13 zones (Nordics + DE-LU + NL).
  • TSO Forecasts: Load, wind, and solar forecasts from ENTSO-E.
  • Physical Data: Weekly hydro reservoir filling (Norway/Sweden) and hourly weather reanalysis (Open-Meteo).
  • Commodities: TTF gas, EUA carbon, coal, and oil futures.

Key Insights and Ideas

  • The Persistence of the Split: The "Nordic Split" wasn't a temporary crisis artifact. As shown in Figure 2, the divergence in co-movement that began in 2020/2021 has persisted through 2025, creating two distinct market types: "Coupled" (NO1, NO2, SE3, SE4) and "Isolated" (NO3, NO4, SE1, SE2).
  • Coupling $\neq$ Information Needs: Counter-intuitively, southern Norwegian zones (NO1, NO2) move almost exactly like Germany but gain very little from VRE forecasts. Conversely, northern Swedish zones (SE1, SE2) are isolated from Germany but gain significantly from VRE forecasts.
  • Generation Mix is King: The value of wind and solar forecasts is dictated by the local generation system, not the interconnectors. In hydro-dominated Norway, reservoirs act as a buffer, absorbing local VRE fluctuations before they hit the price. In Sweden, which has more wind and nuclear, the price is far more sensitive to weather-driven supply.
  • Neighboring Prices are Redundant: Adding the previous day's prices from neighboring zones improved accuracy by less than 2.2% across the board. The zone's own price history already captures the necessary continental signals.
Correlation with DE-LU of day-ahead prices

Key Results

  • VRE Forecast Impact: Removing wind and solar forecasts increased the relative MAE (rMAE) by 31% in Germany, 8–16% in Sweden, but only 0–6% in Norway.
  • Forecasting Accuracy: The LightGBM model with normalization reduced the MAE of a naive D-1 forecast by 24–37% in Nordic zones and 47% in DE-LU.
  • Isolated Zone Paradox: In isolated zones, prices are actually easier to forecast on high-wind days. The authors suggest that when local wind supply exceeds export limits, prices fall "low and flat," making them highly predictable.
Loss of accuracy without VRE forecasts

Limitations and Future Work

  • Point Forecasts vs. Tails: The study focused on Mean Absolute Error. Continental shocks might still be highly relevant for predicting "extreme" price events or probabilistic tails, which this study did not explore.
  • Publication Timing: The authors assumed TSO forecasts are available at 11:00 D-1. In reality, VRE forecasts are often published later (18:00), meaning their "value" in this study represents a potential gain if publication rules were changed.
  • Causality: The relationships found are associations. Future work using event studies (e.g., the exact moment a new cable like NordLink went live) could better isolate causal drivers of the split.

Summary updated

Abstract

After the 2021-2022 energy crisis, day-ahead prices in Nordic bidding zones no longer move together. We ask whether this divergence also changes the information carried by prices. The answer matters for integrating variable renewable energy (VRE), because it can help identify where wind and solar forecasts are worth producing, publishing, and using. We measured the co-movement of each Norwegian and Swedish zone with Germany-Luxembourg (DE-LU) in 2018-2025. We then used a single model, selected on 2024 data, to forecast 2025 prices at 11:00 on the previous day and tested which inputs improved accuracy in each zone. The zones moved alike until 2019, but co-movement diverged in Sweden in 2020 and in Norway in 2021, and the split has persisted. In 2024, four southern zones co-moved strongly with DE-LU (intraday correlations of 0.54-0.71), while four northern zones co-moved weakly (0.25-0.30). Yet price information did not follow this split. Previous-day neighbouring prices reduced mean absolute error by at most 2.2% in any zone, equally in coupled and isolated zones. Removing the transmission system operators' wind and solar forecasts increased mean absolute error by at most 6% in Norway, 8-16% in Sweden, and 31% in DE-LU, and the ordering held in 2024. Within Norway, the value of these forecasts showed no detectable dependence on reservoir filling. The divergence in co-movement was therefore not matched by a divergence in price information, which differed mainly between national generation systems. Thus, a zone's coupling to the continent does not indicate the value of VRE forecasts there.
BibTeX
@misc{phan2026price,
  author = {My Thi Diem Phan and Tuyen Trung Truong and Hoai Phuong Ha and Dat Thanh Nguyen},
  title = {{Price co-movement and price information diverge across Nordic electricity zones after the energy crisis}},
  year = {2026},
  eprint = {2611.16657},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2611.16657},
}
Figure