Institutional Repository of Vilnius University
Nordpool elektros kainų prognozavimas naudojant dirbtinio intelekto algoritmus /
Abstract
dc:descriptionThis bachelor's thesis dealt with the subject of Nordpool's day-ahead electricity market Elspot forecasting using artificial intelligence algorithms. An analysis of the tests described in the scientific literature was carried out and a neural network of the long-short-term memory model was selected for practical research. A practical study was carried out: initially, the selected model was trained using deep learning methods to predict prices only using past prices as input. One-hour ahead forecasts were accurate and efficient, but when trying to forecast longer times ahead, this type of model showed poor performance. The study was extended with a multi-input variable model which, using deep learning methods, was trained based on past electricity generation, electricity consumption and electricity transmission capacity from other countries to forecast prices one day ahead and achieved 24.59 root mean square error accuracy over a three-month forecast period. The influence of the period of unstable geopolitical situation on the accuracy of the forecasts was observed: in 2020, the forecasting was more accurate and the model could be used for a longer period of time, but from 2022, due to the rapidly changing electricity prices caused by the unstable geopolitical situation, the author recommends that in order to achieve accurate forecasts, the long-short term memory neural network model should be frequently re-trained.
Degree
thesis:*- Grantor dc:publisher
- Institutional Repository of Vilnius University
- Year dc:date
- 2023
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Sinicinas, Aidas,
Rights
dc:rights- Statement dc:rights
-
- info:eu-repo/semantics/openAccess
- Language dc:language
- lit
Identifiers
dc:identifier.*- Repository record dc:identifier
- https://repository.vu.lt/VU:ELABAETD192847293&prefLang=en_US