DATA-DRIVEN ELECTRICITY LOAD FORECASTING FOR NIGERIAN UNIVERSITIES: A CASE STUDY OF EDO STATE UNIVERSITY IYAMHO

Reliable electricity demand forecasting is increasingly important for Nigerian universities because campus expansion, digitalisation, laboratory activity, student population growth, and variable electricity supply place pressure on limited energy resources. This study develops and evaluates a data-driven forecasting framework for monthly electricity consumption at Edo State University Iyamho (EDSU), Nigeria. The study uses historical cumulative meter readings obtained from the university's Electrical Section of the Estate and works unit and converted them into monthly consumption observations. The readings cover a period from December 2020 through 2025 as usable sample. The final reconstructed consumption contains 55 monthly consumption intervals, while 48 usable monthly records were retained for modelling. Three forecasting approaches—Seasonal Autoregressive Integrated Moving Average (SARIMA),Prophet, and a hybrid SARIMA–Prophet forecast—were compared using Root Mean Square Error (RMSE) and Mean Absolute Error (MAE). The errors were 55,473 kWh and 39,302 kWh for SARIMA, 39,550 kWh and 27,797 kWh for Prophet, and 42,244 kWh and 29,566 kWh for the hybrid model, respectively. On the numerical metrics, Prophet had the lowest RMSE and MAE, reducing RMSE by 28.7% and MAE by 29.3% relative to SARIMA. The hybrid model improved substantially on SARIMA but did not outperform Prophet. The study contributes a locally grounded forecasting baseline for Nigerian university energy management and identifies the data infrastructure required for more rigorous future forecasting, including automated interval metering, academic-calendar variables, occupancy, weather, tariff, generator operation, and renewable-energy generation. The findings support the use of forecasting as a decision-support component rather than a stand-alone guarantee of supply reliability.