Statistical Demand Forecasting for Strategic Oil Reserve Planning by the Nigerian National Petroleum Corporation

Accurate statistical demand forecasting is fundamental to the NNPC's strategic petroleum reserve planning, ensuring supply security and minimising emergency procurement costs in a context of volatile global crude prices and domestic refinery underperformance. This study develops and compares statistical demand forecasting models for petroleum product consumption in Nigeria using monthly national consumption data from NNPC from 2010 to 2022. Exponential smoothing, ARIMA, and SARIMA models were developed for premium motor spirit, automotive gas oil, and household kerosene demand series. Stationarity, seasonality, and structural break testing preceded model estimation. Forecast accuracy was compared by root mean square error, mean absolute error, and mean absolute percentage error on a 12-month out-of-sample validation period. All three demand series showed significant seasonal patterns and non-stationarity requiring differencing. SARIMA(1,1,1)(1,1,0)12 provided the best forecast accuracy for premium motor spirit demand (MAPE = 4.2 percent). Exponential smoothing outperformed ARIMA for kerosene demand forecasting. Structural breaks in 2016 and 2023 coincided with subsidy policy changes and were incorporated as dummy variables. Ensemble averaging of SARIMA and Holt-Winters models improved premium motor spirit forecast MAPE to 3.7 percent. The study recommends NNPC adopt an ensemble forecasting approach with quarterly model re-estimation and uncertainty interval reporting to support robust strategic reserve level decisions. Keywords: demand forecasting, SARIMA, NNPC petroleum, strategic reserve, time series