Electrical Load Forecasting Based on Random Forest, XGBoost, and Linear Regression Algorithms

Power infrastructure management requires a consistent power supply. One approach of doing this is predicting the power usage. This requires a variety of elements to be considered such as the environment and the spatial and temporal aspects. These tend to demonstrate a considerable fluctuation in the...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:International Conference on Information Technology Research (Online) S. 25 - 31
Hauptverfasser: Abumohsen, Mobarak, Owda, Amani Yousef, Owda, Majdi
Format: Tagungsbericht
Sprache:Englisch
Veröffentlicht: IEEE 09.08.2023
Schlagworte:
Companies
electric power
electrical demand
Linear regression
Load forecasting
machine learning
Machine learning algorithms
Mean square error methods
Power supplies
Radio frequency
ISSN:2831-3399
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:Power infrastructure management requires a consistent power supply. One approach of doing this is predicting the power usage. This requires a variety of elements to be considered such as the environment and the spatial and temporal aspects. These tend to demonstrate a considerable fluctuation in the electrical load pattern depending on the temporal and environmental variables. The primary goal of this research is to develop forecasting models that properly anticipate in predicting the electrical load based on a real and unique dataset of the district energy company (Tubas District Electricity Company - in Palestine). Three machine learning models were used to forecast the electrical loads; namely: (1) Random Forest (RF), (2) XGBoost, and (3) Linear Regression (LR). The models were evaluated, and the RF model was found to achieve the best performance in terms of accuracy. The RF model obtained an R-squared of 87.749%, a Mean Absolute Error (MAE) of 0.03904, and a Mean Square Error (MSE) of 0.00270.
ISSN:2831-3399
DOI:10.1109/ICIT58056.2023.10225968