On deep leaning techniques: Empowering energy sustainability: long-short-term memory recurrent neural network for accurate power forecasting
Type
Supervisor
Subject
Date
2026-01-17
Collections
Research Projects
Organizational Units
Journal Issue
Abstract
The economic growth of nations is closely tied to their electricity infrastructure and availability, as electricity plays a central role in modern life. With increasing global demand for electricity, coupled with fluctuating prices and inadequate generation capacity, accurate load forecasting becomes vital. Estimating future energy demand is crucial for effective planning by electricity providers and promoting energy conservation. However, load forecasting has remained a persistent challenge in the power industry since its inception. In line with the United Nations' Sustainable Development Goals, this study proposes an approach to achieve energy sustainability by predicting future load demands. To test the proposed system, we utilized the outsourced individual household electric power consumption dataset from the University of California-Irvine repository. The Long Short-term Memory-Recurrent Neural Network (LSTM-RNN) algorithm was employed to estimate the overall power consumption of the entire house across different time intervals, including 15 minutes, daily, weekly, and monthly. The system's performance was evaluated using metrics such as Mean Absolute Error (MAE), Mean Squared Error (MSE), Root Mean Square Error (RMSE), and the R-squared score. The accuracy of the model was determined using the Mean Absolute Percentage Error (MAPE). Among the various scenarios for forecasting, the monthly model of the third scenario demonstrated the highest performance. In this scenario the model was trained using data from the previous (n-1) months and tested on the latest month. The obtained scores for this model were 0.034 (MAE), 0.001 (MSE), 0.034 (RMSE), and an accuracy of 97.16%. As a result the constructed model effectively achieved its objectives by accurately predicting the active power consumption of the household.
Department
Publisher
Sponsor
Effat University
Copyright
CC0 1.0 Universal
Book title
Mathematical Modeling for Big Data Analytics