Modelling Electricity Prices in 15-Minute Intervals: A Panel Data Approach for a New Market Regime in the Electricity Pricing
Authors
Yammin, Elissa Kamil ; Pedersen, Andreas Laurids
Term
4. term
Education
Publication year
2026
Submitted on
2026-05-26
Pages
96
Abstract
This thesis examines whether panel data models can be used to model and forecast electricity prices under the new 15-minute settlement regime in the Danish DK1 market. Panel data combine many observations across time and across multiple series or groups. The project focuses on dynamic panel models, where prices may depend on their own past values, and on factor-augmented models that include hidden, common market factors. These approaches aim to capture persistence in prices, regular daily and weekly patterns (seasonality), and dependence across series. The theoretical part introduces fixed effects and random effects as ways to control for systematic differences across units, as well as dynamic panel and factor-augmented approaches. Particular attention is given to modelling temporal persistence and latent, shared market structures during the transition from hourly to quarter-hourly settlement intervals. In the application, the methods are applied to both hourly and quarter-hourly data. The results show that quarter-hourly specifications improve forecasting performance and capture more detailed market dynamics than hourly models. Factor-augmented specifications account for latent dependence structures, but their forecasts become smoother and less responsive to sudden market shocks and extreme price spikes.
Dette speciale undersøger, om paneldatamodeller kan bruges til at modellere og forudsige elpriser efter indførelsen af 15-minutters afregning i det danske DK1-marked. Paneldata kombinerer mange observationer over tid og på tværs af flere serier eller grupper. Projektet ser særskilt på dynamiske panelmodeller, hvor priser kan afhænge af egne fortidige værdier, og på faktorudvidede modeller, der indarbejder skjulte, fælles markedsfaktorer. Målet er at beskrive vedvarende prisbevægelser (persistens), faste mønstre over døgnet og ugen (sæson), og afhængighed mellem serier på tværs (cross-sectional dependence). Den teoretiske del introducerer faste effekter og tilfældige effekter som måder at kontrollere for systematiske forskelle mellem enheder, samt dynamiske panelmodeller og faktorudvidede tilgange. Der lægges særlig vægt på at modellere tidsmæssig persistens og latente, fælles markedsstrukturer i overgangen fra time- til kvartersafregning. I den empiriske del anvendes metoderne på både time- og kvartersdata. Resultaterne viser, at specifikationer baseret på kvartersdata forbedrer prognoserne og fanger mere detaljerede markedsdynamikker end timemodeller. Samtidig håndterer de faktorudvidede modeller de latente afhængigheder, men deres prognoser bliver glattere og reagerer mindre på pludselige markedschok og ekstreme pristoppe.
[This apstract has been rewritten with the help of AI based on the project's original abstract]