Gachevski, Nikola and Kostov, Mitko and Atanasovski, Metodija (2025) Time Series Prediction of Electricity Consumption in North Macedonia Based on LSTM and Feature Analysis. 60th International Scientific Conference on Information, Communication and Energy Systems and Technologies, Ohrid, North Macedonia, June 26–28, 2025. ISSN 979-8-3315-2655-9 (IEEE)
![[thumbnail of 0804_ICEST-2025-1974.pdf]](https://eprints.uklo.edu.mk/style/images/fileicons/text.png) |
Text
0804_ICEST-2025-1974.pdf - Published Version Download (860kB) |
Abstract
In this paper, a model for hourly electricity load forecasting for the Republic of North Macedonia has been presented using deep learning. A soft attention-based LSTM network was trained on the period 2016-2020 and later validated through the usage of the consumption for 2021. It uses meteorological as well as temporal inputs and delivered a MAPE value of 4.93%. Feature importance analysis revealed that day type and temperature had the highest impact on prediction. The model is very accurate, with more deviations observed at national holidays and during changes of seasons.
| Item Type: | Article |
|---|---|
| Subjects: | Scientific Fields (Frascati) > Engineering and Technology > Electrical engineering, electronic engineering,information engineering Scientific Fields (Frascati) > Engineering and Technology > Other engineering and technologies |
| Divisions: | Faculty of Technical Sciences |
| Depositing User: | Prof Mitko Kostov |
| Date Deposited: | 10 Aug 2025 21:05 |
| Last Modified: | 10 Aug 2025 21:05 |
| URI: | https://eprints.uklo.edu.mk/id/eprint/11005 |
Actions (login required)
 |
View Item |