EURO 2024 Copenhagen
Abstract Submission

EURO-Online login

2289. Energy Price Modelling: A Comparative Evaluation of Four Generations of Forecasting Methods

Invited abstract in session WD-31: AI for Energy Finance, stream Analytics.

Wednesday, 14:30-16:00
Room: 046 (building: 208)

Authors (first author is the speaker)

1. Stefan Lessmann
School of Business and Economics, Humboldt-University of Berlin
2. Alexandru-Victor Andrei
Bucharest University of Economic Studies
3. Georg Velev
School of Business and Economics, Humboldt-Universität zu Berlin
4. Daniel Traian Pele
Bucharest University of Economic Studies. Institute for Economic Forecasting, Romanian Academy

Abstract

Energy is a critical input in modern economic systems. Much literature has looked into energy forecasting to support critical decisions at various levels and entities; from operational purchasing decisions at individual business organizations to policy-making. In pursuit of maximum accuracy, the field has considered numerous forecasting methods. In this paper, we examine the evolution of methodology across different generations of forecasting methods, starting from classical econometric approaches, over machine learning methods, and earlier sequence learners like LSTMs, to deep transformer networks, which may be regarded as the latest trend in forecasting. We also consider the latest ideas on adopting pretraining and transfer learning, concepts that have revolutionized the analysis of unstructured data and enable modern AI, for time series forecasting. We offer a systematic review of the related literature and categorize different forecasting methodologies. More importantly, using data from EU energy markets, we conduct a large-scale benchmarking experiment, which contrasts the forecasting accuracy of different approaches, focusing especially on alternative propositions for time series transformers.

Keywords

Status: accepted

Back to the list of papers