A combined deep learning application for short term load forecasting

An accurate prediction of buildings' load demand is one of the most important issues in smart grid and smart building applications. In this way, an important contribution is made to improving the reliability of the power system, facilitating the integration of renewable energy sources and makin...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:Alexandria engineering journal Jg. 60; H. 4; S. 3807 - 3818
Hauptverfasser: Ozer, Ilyas, Efe, Serhat Berat, Ozbay, Harun
Format: Journal Article
Sprache:Englisch
Veröffentlicht: Elsevier B.V 01.08.2021
Elsevier
Schlagworte:
Cross-correlation
Electrical power systems
Load forecasting
Transfer learning
Load forecasting
Transfer learning
Cross-correlation
Electrical power systems
ISSN:1110-0168
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:An accurate prediction of buildings' load demand is one of the most important issues in smart grid and smart building applications. In this way, an important contribution is made to improving the reliability of the power system, facilitating the integration of renewable energy sources and making demand response processes more effective. Nowadays, the electricity prediction based on sensor data has become quite common with the increasing popularity of smart meter applications. While such approaches produce very successful results, they often require large amounts of data recorded over long periods of time during the training of machine learning models to make accurate predictions. The fact that smart meter and sensor applications are becoming more widespread in different parts of the world and that the newly constructed buildings and new meters have relatively small historical data is an important constraint for sensor-based approaches. In this article, a cross-correlation based transfer learning approach is proposed on getting data from different parts of the world to obtain more successful predictive results with limited data. Results of application on actual energy system validate the advantages of proposed model.
ISSN:1110-0168
DOI:10.1016/j.aej.2021.02.050