EnergyBench: a live, gate-aligned benchmark of day-ahead forecasts for European power markets

Evgenii Genov; Ander Martinez Alonso; Christoph Bergmeir

cs.LGNov 5, 2026

EnergyBench: a live, gate-aligned benchmark of day-ahead forecasts for European power markets

Authors: Evgenii Genov, Ander Martinez Alonso, Christoph Bergmeir

Organizations: EVERGi Research Group, Flanders Alliance for Climate Technology (FACT), Vrije Universiteit Brussel, Pleinlaan 2, 1050 Brussels, Belgium · DTU Wind and Energy Systems, Technical University of Denmark · Department of Computer Science and Artificial Intelligence, University of Granada, Granada, Spain · Department of Data Science & Artificial Intelligence, Monash University, 25 Exhibition Walk, Clayton, VIC 3168, Australia

Abstract

Energy forecasting models are usually compared on backtests and fixed public datasets. Backtests can use data that did not yet exist when the forecast would have been made, and a fixed test set, reused across studies and possibly contained in the pretraining data of foundation models, invites overfitting to the benchmark. EnergyBench evaluates every entry live instead. Each day at 11:30 CET/CEST, thirty minutes before the day-ahead (SDAC) gate closes, every model forecasts the next day's price, wind and solar generation and load on the 15-minute grid for six bidding zones (BE, NL, FR, DE-LU, DK1, DK2). Unlike existing live benchmarks, EnergyBench runs the models itself, on one information set that it freezes at the cut-off and publishes, including three numerical weather prediction models, so that foundation models can be compared with and without the same weather covariates under the actual market deadline. Errors are pooled per series against a seasonal-naive baseline, and models are ranked by a Bradley-Terry rating on the Elo scale with a day-bootstrap interval. The field spans baselines, linear and gradient-boosted models on the frozen weather, eighteen time-series foundation models from eleven families, ensembles, and the grid operators' own forecasts. Over the first thirty-two live days, the covariates reduced the error of each of the ten foundation models run with and without them by 37-51% on matched races (64-73% on wind, 48-61% on solar). Used zero-shot, these models match or outperform a LightGBM model trained on the same covariates. Model size, from 3M to 331M parameters, has a much smaller effect, and median ensembles of the covariate models rank highest. The grid operators' pre-gate forecasts remain ahead on wind and solar, where the best single foundation models have 16-22% more error; on load the gap is 4-6%, within the intervals.

Figures & tables

	TS-Arena [ 17 ]	Energy-Arena [ 18 ]	EnergyBench (this paper)
Task timing	from each series’ last published value; no market gate	submission before the 12:00 day-ahead gate, whole next day	11:30 cut-off, locked before the 12:00 gate, whole next day
Zones and targets	14 challenges (SMARD, Gridstatus, Fingrid)	wind, solar, load: AT, DE-LU; price: 15 zones	price, wind, solar, load in each of BE, NL, FR, DE-LU, DK1, DK2
Who runs models	participants (containers)	participants (API submissions)	the benchmark, from submitted code
Information set	series history only	anything before the gate; self-declared, unverified	frozen and published: history, three NWP models, load forecast, fuel
Covariates for foundation models	none	if a participant adds them	eleven foundation models also with the same covariates
Foundation models	the focus of the benchmark	none named; depends on submissions	eleven families, and ensembles

Table 1: Live energy forecasting benchmarks compared, as described in their papers and on their websites (October 2026).

Panel of Figure 1 — Figure 1: One benchmark day (not to scale). Everything a model sees is fixed at the 11:30 cut-off; the forecast must be locked before the 12:00 gate. The TSOs’ final day-ahead wind and solar forecasts appear only after the gate and enter as references.

	Price	Wind	Solar	Load
ENTSO-E document	A44	A75 (B18 + B19)	A75 (B16)	A65
BE	ENTSO-E	Elia (quarter-hourly)	Elia (quarter-hourly)	ENTSO-E
NL	ENTSO-E	NED (national)	NED (national)	NED (gross)
FR	ENTSO-E	ENTSO-E	ENTSO-E	ENTSO-E
DE-LU	ENTSO-E	ENTSO-E	ENTSO-E	ENTSO-E
DK1, DK2	ENTSO-E	Energinet	ENTSO-E	ENTSO-E

Table 2: Primary source of the actuals per zone and target. Gaps are filled from the TSO’s own data (Elia, RTE éCO2mix), then from Energy-Charts; the Dutch NED series have no fallback.

Panel of Figure 2 — Figure 2: NWP runs admitted at the cut-off, in UTC. Each bar runs from a run’s initialisation (dot) to the time it is admitted, initialisation plus the lag (3.5 h for ICON-EU, 7 h for GFS and IFS); blue: the run the covariates use. The cut-off stays at 11:30 local time, so it moves by an hour in UTC with the clock change, but the same runs are admitted in both regimes: ICON-EU 06 UTC (exactly at the summer cut-off) and GFS and IFS 00 UTC.

Zone	Onshore wind	Offshore wind	Solar	Load	Registers
BE	8	2	8	6	OpenStreetMap turbines scaled to Elia’s capacity; RIVM offshore turbines; Fluvius and ORES municipal PV scaled to Elia; Statbel population
NL	8	4	8	6	RIVM turbines; CBS municipal PV and population
FR	8	4	8	6	ODRÉ national register of installations (commune level); INSEE population
DE-LU	8	4	8	6	Marktstammdatenregister; GeoNames population
DK1	8	4	8	6	Energinet capacity per municipality; Danish Energy Agency offshore farms; GeoNames population
DK2	8	3	8	6	as DK1

Table 3: Weather sampling points per zone and bucket (capacity-weighted

k

-means centroids; load weighted by population) and the public registers they are built from.

Figure 3: The 153 weather sampling points: capacity-weighted

k

-means centroids per zone and bucket, from public registers (load: population). Marker size is the point’s share of its zone and bucket.

Entry	Params	Covariates	Developer, set-up
Baselines
naive_weekly			same quarter-hour one week earlier; the reference of all relative errors
naive_daily			same quarter-hour one day earlier
climatology			same local time of day within $\pm 14$ days, 2023 to August 2026
lin_lags , lgbm_lags			ridge and LightGBM on the series’ own lags; two years, retrained monthly
Models trained on the covariates a

Table 4: The field, 31 October 2026 (43 active entries). Parameter counts are those of the checkpoints the benchmark loads. Covariates: native , the model’s own known-future covariate input; ridge , an in-context ridge regression on the covariates whose residual the model forecasts; full , the whole covariate table.

Panel of Figure 4 — Figure 4: The live record, issue days 2 October–2 November 2026. Top: when each series is scored from, with its number of scored issue days (Danish wind lost three; Danish load is forecast but left out, Section 4.1 ). Bottom: when each entry first forecast, with the issue days it ran; Toto left after 27 October. TimeH2O, with six days, is not yet ranked.

	Model	Elo	95% interval	Races	Rel. MAE	Rel. RMSE	Price	Wind	Solar	Load
					all series		Rel. MAE per target
1	ens_hybrid	1259	[1222, 1303]	306	0.240	0.276	0.195	0.130	0.314	0.546
2	ens_fm_cov	1223	[1189, 1264]	306	0.247	0.285	0.197	0.136	0.324	0.569
3	t0_gamma_cov	1207	[1171, 1247]	390	0.275	0.311	0.224	0.162	0.344	0.586
4	timestral_cov	1193	[1165, 1225]	390	0.273	0.306	0.250	0.151	0.321	0.597
5	timesfm3_cov	1192	[1167, 1219]	525	0.282	0.307	0.229	0.160	0.336	0.686

Table 5: Live standings, issue days 2 October–2 November 2026, all six zones (22 series). Elo with its day-bootstrap 95% interval; Rel. MAE and Rel. RMSE pooled per series against naive_weekly , geometric mean over the series a model forecasts (“all”) or over the zones of one target. Ranked entries; covariate models in bold.

Panel of Figure 5 — Figure 5: Live Elo ratings with day-bootstrap 95% intervals, entries of Table 5 . Blue: models using the frozen covariates; grey: history only. The pre-gate TSO forecast is not rated over all series (Section 4 ). The dashed line is the field average.

Figure 6: Pooled Rel. MAE per series for the references, five covariate models, two history-only foundation models and naive_daily (live, 2 October–2 November 2026; colour on a log scale, 1 is the seasonal-naive baseline). Empty cells: no forecast, or (for tso_final in DK1 and DK2) a forecast of a different scope.

				Rel. MAE		Change with covariates
Model	Covariates	Days	Races	without	with	all	price	wind	solar	load
Chronos-2	native	31	525	0.569	0.295	$-48\%$	$-34\%$	$-67\%$	$-57\%$	$-4\%$
TimesFM 3.0	native	31	525	0.578	0.282	$-51\%$	$-33\%$	$-73\%$	$-61\%$	$-1\%$
t0-alpha	native	29	511	0.596	0.334	$-44\%$	$-15\%$	$-69\%$	$-54\%$	$-2\%$
t0-beta	native	29	511	0.588	0.320	$-46\%$	$-33\%$	$-64\%$	$-55\%$	$-3\%$
t0-gamma	native	18	390	0.527	0.275	$-48\%$	$-31\%$	$-69\%$	$-53\%$	$-13\%$

Table 6: The same foundation model without and with the frozen covariates, on the races both forecast (live, all zones, issue days 2 October–2 November 2026). “Native”: the model takes known-future covariates itself; “ridge”: an in-context ridge regression on the covariates, the model forecasting its residual (Tabby forecasts load without covariates). Rel. MAE pooled per series, geometric mean over series; a change is the geometric mean of the per-series ratios minus one.

Panel of Figure 7 — Figure 7: Pooled Rel. MAE per zone and target group (live, 2 October–2 November 2026, log scale; the dashed line is the seasonal-naive baseline). Every point of a zone and panel is computed on the same races: those that all models shown there forecast, the TSO references included where they exist ( tso_cutoff : BE and NL wind and solar, and load; tso_final : wind and solar outside Denmark, Section 4.6 ). Danish load is left out (Section 4 ).

Panel of Figure 8 — Figure 8: Two delivery days: actual output, Chronos-2 without and with covariates (shaded: the covariate version’s 10–90% band), and a TSO forecast (Elia’s pre-gate forecast for BE; RTE’s final forecast, published after the gate, for FR).

Figure 9: Pooled Rel. MAE against parameter count for fifteen foundation models, without covariates (open) and, where a variant exists, with the frozen covariates (filled, joined to the same model without), on the issue days 16 October–2 November 2026 (all zones, 390 races per model). Parameter counts are those of the checkpoints the benchmark loads; TiRex-2 uses about 38M of its 95M in univariate mode.

Panel of Figure 10 — Figure 10: Foundation models with covariates against the grid operators’ pre-gate forecast, by model release date, with day-bootstrap 95% intervals; the line is the best model released so far. Open markers: covariates through a ridge regression. Geometric mean of the per-series error ratios on the same 141 races (issue days 16 October–2 November 2026, eight series); every model ran on the same days, so the figure compares model generations, not progress over time.

Model	All	95% interval	Wind, solar	Load
Timestral	$+11\%$	[ $0$ , $+23$ ]	$+16\%$	$+6\%$
t0-gamma	$+13\%$	[ $+1$ , $+26$ ]	$+22\%$	$+4\%$
Chronos-2	$+20\%$	[ $+6$ , $+35$ ]	$+32\%$	$+8\%$
TimesFM 3.0	$+21\%$	[ $+9$ , $+33$ ]	$+21\%$	$+21\%$
t0-beta	$+24\%$	[ $+9$ , $+41$ ]	$+34\%$	$+15\%$
Tabby	$+29\%$	[ $+14$ , $+46$ ]	$+48\%$	$+13\%$

Table 7: Foundation models with covariates and lgbm_prod against the pre-gate TSO forecast tso_cutoff , on the same 141 races (issue days 16 October–2 November 2026; BE and NL wind and solar, load in BE, NL, FR and DE-LU): extra pooled error, geometric mean over series, for all eight series with its day-bootstrap 95% interval and per target group. Negative: the entry has the lower error.

Forecasting Benchmarks

Forecasting Benchmarks are datasets and evaluation protocols for comparing predictions of future time-series values, events, and outcomes.

See all 237 papers

Foundation models for electricity price forecasting and battery arbitrage: Can they replace market-specific forecasting models?
Aug 31, 2026Arkadiusz Lipiecki, Rafał WeronTabular Foundation Models Electricity Price Forecasting
Department of Computational Social Science, Wrocław University of Science and Technology, Poland · Department of Operations Research and Business Intelligence, Wrocław University of Science and Technology, Poland · Center for Research in Energy (CoRE), Aarhus University, 8000 Aarhus C, Denmark
FETS Benchmark: Foundation Models Outperform Dataset-specific Machine Learning in Energy Time Series Forecasting
Apr 24, 2026Marco Obermeier, Marco Pruckner, Florian Haselbeck +1Time Series Forecasting Forecasting Benchmarks
Julius-Maximilians-Universität Würzburg, Modeling and Simulation Lab, Am Hubland, Würzburg, 97074, Germany · Weihenstephan-Triesdorf University of Applied Sciences, Smart Farming, Am Staudengarten 1, Freising, 85354, Germany · Weihenstephan-Triesdorf University of Applied Sciences, Digital Energy Transition, Hans-Carl-von-Carlowitz-Platz 3, Freising, 85354, Germany
Energy-Arena: A Dynamic Benchmark for Operational Energy Forecasting
Apr 27, 2026Max Kleinebrahm, Jonathan Berrisch, Philipp Eiser +11Benchmark Design Time Series Forecasting
Karlsruhe Institute of Technology, Institute for Industrial Production, Karlsruhe, Germany · University of Duisburg-Essen, House of Energy Markets and Finance, Essen, Germany · Karlsruhe Institute of Technology, Institute for Automation and Applied Informatics, Eggenstein-Leopoldshafen, Germany +5

Aug 31, 2026J/K move · Enter open · S save

Foundation models for electricity price forecasting and battery arbitrage: Can they replace market-specific forecasting models?

Arkadiusz Lipiecki, Rafał Weron

Department of Computational Social Science, Wrocław University of Science and Technology, Poland · Department of Operations Research and Business Intelligence, Wrocław University of Science and Technology, Poland · Center for Research in Energy (CoRE), Aarhus University, 8000 Aarhus C, Denmark

Tabular Foundation Models Electricity Price Forecasting Time Series Forecasting Probabilistic Forecasting Time-Series Foundation Models

Open paper AskPDF arXiv

Foundation models promise accurate forecasts with little or no task-specific training, but whether they can replace models designed specifically for electricity price forecasting remains unclear. We compare nine variants from five foundation model families, evaluated in zero-shot mode, with two state-of-the-art electricity price forecasting benchmarks in Germany, Poland, and Spain over 2021-2025. Their performance is assessed in terms of point and probabilistic forecasting accuracy, as well as economic value in battery energy storage arbitrage. Only the TabPFN models consistently and significantly outperform the benchmarks across all three markets and all statistical measures. However, this statistical dominance does not translate directly into economic dominance: TabPFN performs best under unlimited bids and riskier quantile-based strategies, whereas the Distributional Deep Neural Network benchmark is more profitable when risk tolerance is lower. Thus, foundation models cannot universally replace market-specific models, and their value depends on both model architecture and the decision problem.

Abstract

Forecasting Benchmarks

Explore similar work

Foundation models for electricity price forecasting and battery arbitrage: Can they replace market-specific forecasting models?

FETS Benchmark: Foundation Models Outperform Dataset-specific Machine Learning in Energy Time Series Forecasting

Energy-Arena: A Dynamic Benchmark for Operational Energy Forecasting