Cross-Market Meta-Learning for Electricity Price Forecasting

Delaram Pirhayatifard; Terrence Alsup

cs.LGOct 12, 2026

Cross-Market Meta-Learning for Electricity Price Forecasting

Authors: Delaram Pirhayatifard, Terrence Alsup

Organizations: Rice University, Houston, TX · Morgan Stanley, New York, NY

Abstract

Day-ahead (DA) electricity price forecasting is usually posed per market, with a separate model fitted to each grid's own history. This caps accuracy: statistical and tree-based baselines saturate, while higher-capacity sequence models need more training data than a single market's few years of hourly prices can supply. Such methods also treat each node as an independent series, discarding the spatial dependencies induced by network topology and congestion. Because price formation across independent system operators (ISOs) is driven by shared physics and largely common market design, we instead treat cross-market history as a pretraining corpus. We present MetaGrid, a topology-aware transformer, meta-trained across markets so that a single zone-count-independent parameter set adapts to an unseen grid. On a leak-free benchmark of five U.S. ISOs, where every model receives the same prices, DA load forecasts and calendar features, MetaGrid attains the lowest average MAE, improving on the best baseline by 3.1% and on the same architecture trained from scratch by 7.7%.

Figures & tables

		MAE ($/MWh)						AQL
Category	Model	PJM	NYISO	CAISO	MISO	ERCOT	Avg	($/MWh)
Naive	Persistence (24h) †	$11.61$	$11.07$	$6.65$	$8.11$	$12.90$	$10.068$	–
Naive	Persistence (48h) †	$16.15$	$16.41$	$9.19$	$10.78$	$15.66$	$13.639$	–
Linear	LEAR †	$9.24$	$9.09$	$6.34$	$7.64$	$11.07$	$8.675$	$3.544$
Linear	Quantile regression	$11.00_{\pm 0.01}$	$9.94_{\pm 0.01}$	$6.75_{\pm 0.02}$	$7.69_{\pm 0.04}$	$13.79_{\pm 0.02}$	$9.833_{\pm 0.010}$	$4.009_{\pm 0.002}$
GBM	XGBoost	$10.09_{\pm 0.03}$	$9.96_{\pm 0.06}$	$6.27_{\pm 0.01}$	$7.90_{\pm 0.05}$	$11.19_{\pm 0.06}$	$9.080_{\pm 0.018}$	$3.757_{\pm 0.007}$

Table 1: Per-ISO forecasting performance on the 2025 test set (5 seeds).

\dagger

denotes deterministic models; bold indicates the best result.

Appendix figures & tables4 assets

Supplementary material from the paper’s appendix.

Appendix

ISO	Zones	Date range	Freq.	Samples
CAISO	4	2021-05-01 – 2025-12-31	1 h	163,776
ERCOT	8	2021-01-01 – 2025-12-31	1 h	350,592
MISO	8	2023-01-01 – 2025-12-31	1 h	210,432
NYISO	15	2021-01-01 – 2025-12-31	1 h	657,360
PJM	21	2021-01-01 – 2025-12-31	1 h	920,304

Appendix

Table 2: Dataset specs. Samples are zone-hours of day-ahead LMP.

Model	PJM	NYISO	CAISO	MISO	ERCOT	Avg
MetaGrid	$9.04$	$8.78$	$5.64$	$6.43$	$10.27$	$8.03\pm 0.028$
without graph-aware module	$9.43$	$9.51$	$5.84$	$6.87$	$10.43$	$8.42\pm 0.007$

Appendix

Table 3: MetaGrid with and without the graph-aware cross-zone module. Test MAE in

/MWh; the average column is the mean

\pm$ standard deviation over 5 seeds.

	init $\alpha=\beta=0.036$			init $\alpha=\beta=1.000$
Market	$\alpha$	$\beta$	$g$	$\alpha$	$\beta$	$g$
PJM	$0.612_{\pm 0.021}$	$0.388_{\pm 0.017}$	$0.734_{\pm 0.028}$	$0.634_{\pm 0.029}$	$0.401_{\pm 0.022}$	$0.719_{\pm 0.033}$
NYISO	$0.714_{\pm 0.026}$	$0.352_{\pm 0.019}$	$0.681_{\pm 0.031}$	$0.698_{\pm 0.031}$	$0.366_{\pm 0.024}$	$0.695_{\pm 0.036}$
CAISO	$0.283_{\pm 0.018}$	$0.507_{\pm 0.023}$	$0.402_{\pm 0.024}$	$0.301_{\pm 0.025}$	$0.492_{\pm 0.027}$	$0.415_{\pm 0.029}$
MISO	$0.164_{\pm 0.014}$	$0.541_{\pm 0.021}$	$0.286_{\pm 0.019}$	$0.187_{\pm 0.020}$	$0.533_{\pm 0.026}$	$0.298_{\pm 0.025}$
ERCOT	$0.441_{\pm 0.023}$	$0.463_{\pm 0.020}$	$0.588_{\pm 0.026}$	$0.428_{\pm 0.028}$	$0.470_{\pm 0.025}$	$0.601_{\pm 0.031}$

Appendix

Table 4: Learned attention-prior weights

\alpha

(adjacency) and

\beta

(context correlation) and decoder gate

g

per held-out ISO, under two initializations of

\alpha

and

\beta

Initialization	PJM	NYISO	CAISO	MISO	ERCOT	Avg
MetaGrid (FOMAML)	9.04 $\pm$ .009	8.78 $\pm$ .06	5.64 $\pm$ .06	6.43 $\pm$ .005	10.27 $\pm$ .004	8.03 $\pm$ .028
pooled pretraining	9.11 $\pm$ .08	8.89 $\pm$ .02	5.95 $\pm$ .02	6.97 $\pm$ .04	10.67 $\pm$ .13	8.32 $\pm$ .02
no pretraining	9.11 $\pm$ .07	9.18 $\pm$ .12	5.71 $\pm$ .03	8.37 $\pm$ .07	11.12 $\pm$ .26	8.70 $\pm$ .06

Appendix

Table 5: Contribution of cross-market pretraining and of the meta-objective. All three arms share the architecture, the source markets and the evaluation-time adaptation protocol; they differ only in how the initialization is obtained (Test MAE in $/MWh).

Electricity Price Forecasting

Electricity price forecasting predicts wholesale and retail prices to support trading, grid operations, and energy planning.

See all 20 papers

Market-Information-Aware Gated-LoRA of Foundation Models for Transferable Day-Ahead Electricity Price Forecasting
Aug 11, 2026Hang Fan, Wei Wei, Shengwei MeiElectricity Price Forecasting Time Series Forecasting
A Benchmark for Electrical Load Forecasting Across Grid Levels: Time-Series Transformers Outperform Established Methods
Jul 17, 2026Matthias Hertel, Sebastian Pütz, Jonathan Kolar +3Time Series Forecasting Forecasting Benchmarks
Karlsruhe Institute of Technology, Germany · Helmholtz AI, Germany
Deep Learning for Cross-Border Electricity Price Forecasting: A Comparative Study
Aug 17, 2026Hadeer Elashhab, Sai Srijan Papineni, Marvin Dorn +2Electricity Price Forecasting Time Series Forecasting
Institute for Automation and Applied Informatics, Karlsruhe Institute of Technology, Karlsruhe, Germany · Department of Computer Science, Karlsruhe Institute of Technology, Karlsruhe, Germany

Aug 11, 2026J/K move · Enter open · S save

Market-Information-Aware Gated-LoRA of Foundation Models for Transferable Day-Ahead Electricity Price Forecasting

Hang Fan, Wei Wei, Shengwei Mei

Electricity Price Forecasting Time Series Forecasting Low-Rank Adaptation Transfer Learning Time-Series Foundation Models

Open paper AskPDF arXiv

Electricity price forecasting is crucial for market participants but remains difficult because prices are volatile, market-specific, and closely tied to anticipated system conditions. Existing supervised methods depend largely on market-specific historical data, limiting their use in newly established or data-scarce markets. This paper proposes a market-information-aware adaptation framework that transfers the Chronos-2 time-series foundation model to day-ahead electricity price forecasting. It first constructs a multi-source market information (MSMI) interface aligning 7-day price context with pre-clearing supply--demand, reserve, maintenance, generator-capacity, and intertie variables, and then trains a source-domain gated low-rank adapter (LoRA), updating about

1\%

of model parameters without target-market labels. The gate scales the frozen source adapter according to reserve-tightness and operating-state signals. A leave-one-market-out protocol is adopted for evaluating cross-market transferability. Experiments on four Chinese provincial day-ahead spot markets show that the proposed framework reduces the average MAE/RMSE by

6.24\%/7.99\%

relative to market-information-aware zero-shot Chronos-2 and by

3.05\%/3.52\%

relative to vanilla Source-LoRA. Experiments show that the gain is not reproduced by a learned global scalar or by random gate initialization, while the additional improvement over Source-LoRA is limited. These results suggest that market-structured inputs and state-dependent gated LoRA can provide a practical transfer path for data-scarce electricity markets.

Abstract

Electricity Price Forecasting

Explore similar work

Market-Information-Aware Gated-LoRA of Foundation Models for Transferable Day-Ahead Electricity Price Forecasting

A Benchmark for Electrical Load Forecasting Across Grid Levels: Time-Series Transformers Outperform Established Methods

Deep Learning for Cross-Border Electricity Price Forecasting: A Comparative Study