Machine Learning-Based Forecasting of Energy Consumption in a Nigerian Tertiary Institution

O. A. Adebimpe, Department of Industrial and Production Engineering, Faculty of Technology, University of Ibadan, NigeriaFollow
S. S. Showunmi, Department of Industrial and Production Engineering, Faculty of Technology, University of Ibadan, NigeriaFollow
F. T. Adedeji, African Institute for Science Policy and Innovation, Obafemi Awolowo University, Ile-Ife, NigeriaFollow

Keywords

Energy consumption, Forecasting, Machine learning, Nigerian tertiary institution, Prediction

Document Type

Article

Abstract

An accurate electricity consumption forecast is key to effective energy planning and management in Nigerian tertiary institutions. While applications of machine learning (ML) in forecasting have increased, their deployment in predicting energy consumption in Nigerian tertiary institutions has been limited. Also, the significance of weather-related variables, academic and non-academic activities, and the staff and students' population in predicting energy consumption at a tertiary institution has been underexplored. This study addresses these by developing and evaluating six forecasting models, including Seasonal Autoregressive Integrated Moving Average with Exogenous Variables (SARIMAX), Support Vector Regression (SVR), Artificial Neural Network (ANN), Extreme Gradient Boosting (XGBoost), a hybrid (ANN–XGBoost), and TabNet. The models were trained on a dataset comprising temperature, precipitation, relative humidity, academic and non-academic activity days, staff and students' population and daily energy consumption variables from 2017 to 2023. The results showed that TabNet outperformed SARIMAX, SVR, ANN, XGBoost and ANN-XGBoost with a Mean Absolute Error (MAE) of 35,052 kWh, Root Mean Square Error (RMSE) of 42,728 kWh, Mean Absolute Percentage Error (MAPE) of 9.23%, and coefficient of determination (R²) of 0.981. The analysis of feature importance using SHapley Additive exPlanations (SHAP) revealed population variables as the most important feature in predicting the university's energy consumption, particularly the effective population that actively uses electricity. This study has revealed the effectiveness of ML models in predicting energy consumption in a Nigerian tertiary institution and the significant input variables.

Recommended Citation

Adebimpe, O. A.; Showunmi, S. S.; and Adedeji, F. T. (2026) "Machine Learning-Based Forecasting of Energy Consumption in a Nigerian Tertiary Institution," Engineering and Technology Journal: Vol. 44: Iss. 8, Article 4.
DOI: https://doi.org/10.30684/2412-0758.1580

DOI

10.30684/2412-0758.1580

First Page

Last Page

Download

COinS