NCN 2025/57/B/HS4/02413

Temporal Hierarchies for electricity pRIce ForecasTing (THRIFT)

Grant no.: NCN 2025/57/B/HS4/02413
Funding agencies: National Science Centre (NCN), Poland
Funding scheme: OPUS
Funding period: 20.01.2026-19.01.2030 (48 months)
Budget: 999 180 PLN
Title in Polish: Hierarchie temporalne na potrzeby prognozowania cen energii elektrycznej (THRIFT)

Research team:

Principal Investigator (Kierownik):

Rafał Weron (Wrocław, POL 🇵🇱)

Partner Investigator (Wiodący partner zagraniczny):

Nikolaos (Nikos) Kourentzes (Skövde, SWE 🇸🇪)

Senior Investigators (Główni wykonawcy):

Investigators (Wykonawcy):

Collaborators (Współpracownicy):

Kaja Bilińska (Wrocław, POL 🇵🇱)
Arkadiusz Lipiecki (Wrocław, POL 🇵🇱)
Katarzyna Maciejowska (PWr, Wrocław, POL 🇵🇱)

– Ph.D. / M.Sc. / B.Sc. student

Aims and scope:

Decision making is at the core of any organization, and highly dynamic operational decisions require accurate short-term forecasts. This is particularly relevant in extremely volatile electricity markets, where the price can increase tenfold in a matter of hours or even minutes. However, decisions may require tailored forecasts that focus on different levels of detail and granularity. For instance, models for quarter-hourly or hourly products can use different information sets than those for average daily prices. As a consequence, these forecasts may not align and can lead to suboptimal decisions.

To cope with this, the forecasts from each temporal and/or spatial level of the hierarchy should be reconciled to be coherent. The last decade has seen an unprecedented growth of interest in forecast reconciliation and the introduction of temporal hierarchies. The latter can be constructed for any time series by means of nonoverlapping aggregation to yield temporally reconciled, accurate and robust forecasts. Interestingly, temporal hierarchies have not yet been applied to day-ahead electricity price forecasting (EPF). However, our preliminary results show that even using a simple temporal hierarchy can reduce errors by 2-3% and increase trading profits by up to 12%!

Tasks:

To fill this gap, the THRIFT project aims to develop and thoroughly assess models that use Temporal Hierarchies for electricity pRIce ForecasTing. It offers an integrated approach consisting of four interrelated tasks performed in parallel:

T1: Developing a framework for temporal hierarchies and reconciliation techniques tailored to day-ahead power
markets;
T2: Using machine learning methods for benchmarking and temporal reconciliation;
T3: Constructing probabilistic hierarchical forecasting methods to support electricity trading;
T4: Developing spatio-temporal hierarchical forecasting models for interconnected power markets.

Publications:

Peer-reviewed articles in JCR-listed journals

2026 (0+)

Peer-reviewed articles in non JCR-listed journals

Edited volumes / special issues

Book chapters

Conference papers

Forthcoming publications, submitted papers and work in progress

A. Lipiecki, K. Bilińska, N. Kourentzes, R. Weron (2025) Stealing accuracy: Predicting day-ahead electricity prices with Temporal Hierarchy Forecasting (THieF), International Journal of Forecasting, submitted. Working paper version available from arXiv: https://arxiv.org/abs/2508.11372

Developed software components

Project team meetings:

Kick-off meeting …