Intelligent Machine-Learning-Driven Load Management for Campus-Scale Energy Optimization and Resilience

The increasing demand for reliable and efficient energy distribution in educational institutions necessitates the adoption of intelligent energy management systems. This research develops a machine-learning-based framework for load management on a university campus using the University of Lagos as a case study. A dataset comprising 3,648 hourly timestamps across 55 occupied structures was analysed: clustering techniques (K-Means, Hierarchical, Gaussian Mixture Models, Spectral, Mini-Batch K-Means and DBSCAN) were applied, with Mini-Batch K-Means achieving the best performance (Silhouette $0.461$, Davies-Bouldin $0.767$, Calinski-Harabasz $42.0$) and segmenting buildings into high-, medium-, and low-demand clusters. Short-term load forecasting (STLF) was then performed with Prophet, SARIMA, ARIMA, LSTM, and GRU on both the whole-campus series and on cluster-specific series; ARIMA produced the lowest point-forecast errors in all evaluations (whole-campus MAPE $7.2\%$ and RMSE $118.7$; cluster-level MAPE 3.8-5.4\%), and consumption-based clustering reduced MAPE by roughly 33-36\% and RMSE by 76-78\% across all algorithms. Overall, the study demonstrates the feasibility of applying machine learning for institutional load management, offering a scalable framework for university campus environments.