Code and data for Decision-level conformal profit calibration for battery arbitrage in day-ahead electricity markets: a public seven-market benchmark

This WinRAR archive contains the code and result files supporting the manuscript “Decision-level conformal profit calibration for battery arbitrage in day-ahead electricity markets: a public seven-market benchmark,” submitted to the International Journal of Electrical Power & Energy Systems. The study develops and evaluates ProfitGuardStressCP, a decision-level conformal framework that calibrates one-sided scheduled-profit optimism for battery arbitrage in day-ahead electricity markets.

The archive includes the reproducibility materials used to generate the manuscript’s forecasting, operational, conformal-calibration, robustness, and figure results. The workflow combines a public seven-market day-ahead electricity price benchmark covering Germany and six Italian price zones, modern point forecasting models, a validation-weighted CatBoost–NBEATSx ensemble, and a realistic 24-hour battery arbitrage linear programming model with state-of-charge dynamics, charge/discharge efficiency, power and energy limits, terminal state-of-charge recovery, and degradation cost.

The uploaded materials support the reported empirical findings: the CatBoost–NBEATSx ensemble achieves the lowest average forecasting MAE in the compact seven-market benchmark, while ProfitGuardStressCP retains 97.8% of the point-schedule profit, reduces loss days from 73 to 29 across 1,824 market-days, and achieves empirical 90% lower-bound coverage of 0.906. These results are also summarized in the article highlights and manuscript abstract.

The archive is intended to facilitate transparent reproduction of the article’s main tables, figures, and sensitivity analyses. It contains scripts and outputs for the forecasting benchmark, battery scheduling evaluation, conformal lower-bound construction, operational comparisons against hour-wise conformal and abstention baselines, robustness checks, and figure generation. The external market dataset follows the public benchmark structure described in the manuscript and should be obtained from the corresponding public benchmark repository cited in the article.

Suggested use: researchers may use these files to reproduce the reported results, inspect the implementation of the decision-level conformal calibration procedure, compare operational conformal strategies for battery arbitrage, or extend the framework to other electricity markets, storage models, or multi-market trading settings.