Statistical Modelling of Electricity Demand Forecasting for the Abuja Electricity Distribution Company

This study develops statistical models for short-term and medium-term electricity demand forecasting for the Abuja Electricity Distribution Company, which supplies power to the Federal Capital Territory and parts of Niger and Nasarawa states in North Central Nigeria. Accurate demand forecasting is essential for efficient generation scheduling, transmission planning, and distribution network management, yet the Abuja Electricity Distribution Company currently lacks a systematic quantitative forecasting framework validated against historical demand data. Hourly metered electricity consumption data at feeder and substation levels are obtained from the distribution company covering the period January 2021 to December 2023, supplemented by meteorological data from the Nigerian Meteorological Agency. Multiple regression models with calendar variables, temperature variables, and lagged demand terms are estimated for short-term daily and weekly forecasting, while ARIMA and exponential smoothing state space models are developed for medium-term monthly forecasting horizons. Model performance is evaluated using mean absolute percentage error and root mean squared error on a held-out validation dataset covering the final six months of the observation window. Results indicate that the regression model incorporating temperature, day type, and previous week's corresponding demand achieves a mean absolute percentage error of 4.2 percent for day-ahead hourly forecasting, outperforming the ARIMA benchmark. Feeder-level forecasting models show significantly higher error rates in areas with high proportions of informal settlements. Keywords: electricity demand forecasting, ARIMA, regression models, distribution company, Abuja