AI-POWERED ELECTRICITY LOAD PREDICTION SYSTEM FOR OLD BOYS HOSTEL, IGBINEDION UNIVERSITY, OKADA

Abstract This research presents the development and implementation of an electricity load forecasting system for the Old Boys Hostel at Igbinedion University, Okada, using advanced machine learning techniques. The study addresses the growing challenges of energy management in Nigerian institutional settings, where fluctuating electricity demand, rising energy costs, and limited infrastructure necessitate more efficient resource planning. Data was collected on key influencing factors, including hourly appliance usage, occupancy levels, environmental conditions (temperature and humidity), and historical consumption records. Using the XGBoost regression model, the system was trained to predict hourly electricity load, achieving a Mean Absolute Error (MAE) of 0.424 kWh, a Root Mean Square Error (RMSE) of 0.588 kWh, and an R² score of 0.908, indicating strong predictive accuracy. Visual analyses, including line plots, heatmaps, and feature importance rankings, revealed that peak demand typically occurs between evening hours and is significantly influenced by temperature and block type. The forecasting system offers practical applications for hostel management, enabling better scheduling of generators, reduction of fuel costs, and improved energy budgeting. Furthermore, the research highlights the feasibility of applying machine learning-based solutions in data-constrained environments, providing a foundation for future integration with real-time monitoring systems and expansion across other campus facilities. By combining rigorous data analysis with actionable recommendations, this study contributes to the advancement of energy efficiency strategies within Nigerian universities.