TY - GEN
T1 - An Assessment of Energy Demand Using Short-Term Load Forecasting
AU - Nyathi, Deans David
AU - Bokoro, Pitshou Ntambu
AU - Farrag, Mohamed Emad
AU - Sheik, Adelaide
N1 - Publisher Copyright:
© 2025 IEEE.
PY - 2025
Y1 - 2025
N2 - Renewable energy is crucial for mitigating the greenhouse effect, yet its integration into power systems presents challenges such as energy insufficiency during demand peaks, improper sizing of solar panels and battery storage, and unplanned downtime that can cause energy and economic losses. Accurate short-term load forecasting (STLF) is essential for improving energy demand assessment and optimizing renewable systems. This study evaluates three STLF methods, namely Auto-Regressive Integrated Moving Average (ARIMA), Support Vector Regression (SVR), and Artificial Neural Networks (ANN), using data from a social enterprise in South Africa. The dataset of 646 hourly observations was divided into training and testing sets, and model performance was assessed with Root Mean Squared Error (RMSE), Mean Absolute Error (MAE), and Mean Absolute Percentage Error (MAPE). The ANN model achieved the lowest forecasting error, making it the most suitable approach. Accurate STLF improves system sizing and operation, reducing inefficiencies and supporting clean energy deployment.
AB - Renewable energy is crucial for mitigating the greenhouse effect, yet its integration into power systems presents challenges such as energy insufficiency during demand peaks, improper sizing of solar panels and battery storage, and unplanned downtime that can cause energy and economic losses. Accurate short-term load forecasting (STLF) is essential for improving energy demand assessment and optimizing renewable systems. This study evaluates three STLF methods, namely Auto-Regressive Integrated Moving Average (ARIMA), Support Vector Regression (SVR), and Artificial Neural Networks (ANN), using data from a social enterprise in South Africa. The dataset of 646 hourly observations was divided into training and testing sets, and model performance was assessed with Root Mean Squared Error (RMSE), Mean Absolute Error (MAE), and Mean Absolute Percentage Error (MAPE). The ANN model achieved the lowest forecasting error, making it the most suitable approach. Accurate STLF improves system sizing and operation, reducing inefficiencies and supporting clean energy deployment.
KW - artificial neural networks
KW - auto-regressive integrated moving average
KW - performance evaluation metrics
KW - short-term load forecasting
KW - support vector regression
UR - https://www.scopus.com/pages/publications/105031907246
U2 - 10.1109/PowerAfrica65840.2025.11289118
DO - 10.1109/PowerAfrica65840.2025.11289118
M3 - Conference contribution
AN - SCOPUS:105031907246
T3 - Proceedings of the 2025 IEEE PES/IAS PowerAfrica Conference: Pioneering Sustainable Energy Solutions for Africa's Future, PAC 2025
BT - Proceedings of the 2025 IEEE PES/IAS PowerAfrica Conference
PB - Institute of Electrical and Electronics Engineers Inc.
T2 - 2025 IEEE PES/IAS PowerAfrica Conference: Pioneering Sustainable Energy Solutions for Africa's Future, PAC 2025
Y2 - 28 September 2025 through 2 October 2025
ER -