Statistical Modelling of Electricity Consumption Patterns for Load Forecasting at Abuja DisCo

Accurate electricity consumption load forecasting is essential for Abuja Electricity Distribution Company (AEDC) operations, enabling optimal generation dispatch, maintenance scheduling, and infrastructure investment planning in the Federal Capital Territory. This study develops statistical models of electricity consumption patterns and short-term load forecasts using AEDC billing and metering data from 2019 to 2023. Half-hourly load data from five major feeders covering the CBD, Gwarinpa, and Lugbe distribution zones were analysed. Seasonal decomposition, multiple linear regression with temperature and calendar predictors, and SARIMA models were estimated. Neural network forecasting served as a non-linear benchmark. Peak load patterns, seasonal profiles, and weekly periodicity were characterised. Load showed strong daily and weekly seasonality, with evening peaks between 18:00 and 21:00 and Monday load factors 12 percent below Friday peaks. SARIMA(2,1,1)(1,1,1)48 achieved the best short-term forecast accuracy on half-hourly data (MAPE = 3.8 percent). Multiple linear regression with temperature (beta = 0.31), day-of-week dummies, and public holiday indicators explained 81 percent of load variance. The neural network benchmark showed comparable accuracy at MAPE 3.6 percent but required substantially greater computational resources. The study provides AEDC operational planners with validated statistical load forecasting models and recommends deployment of SARIMA-based forecasts on a 24-hour rolling basis with weekly recalibration. Keywords: electricity load forecasting, SARIMA, AEDC, consumption patterns, distribution company