Probabilistic electricity price forecasting through conformalized deep ensembles

Probabilistic electricity price forecasting (PEPF) is subject of an increasing interest, following the demand for proper prediction uncertainty quantification, to support the operation in complex power markets with increasing share of renewable generation. Distributional neural networks ensembles (DE) have been recently shown to outperform state of the art PEPF benchmarks. Still, they require reliability improvements, as fail to pass the coverage tests at various steps on the prediction horizon.In this work, we tackle this issue by extending the DE framework with the introduction of a Conformal Prediction based technique.Experiments have been conducted on multiple market regions, achieving day-ahead probabilistic forecasts with better hourly coverage.