Day ahead forecast of complex seasonal natural gas data to enhance procurement efficiency

Abstract

Natural gas is one of the important commodities used by industry. With the volatility in market prices of this resource, it is essential to accurately forecast the consumption of natural gas at an organisation level. This will not only aid in effective procurement but also help in reducing various penalties associated with it. The forecasting task becomes more complicated with the existence of multiple seasonality in consumption data. A daily natural gas consumption data of a manufacturing plant with multiple and non-integer seasonality is analysed. The selected forecasted model provides recognition and identification of the existence of seasonality. RMSE and MAPE are compared for distinct forecasting horizons to examine the performance of forecasting techniques. TBATS model represents excellent forecasting results for individual prediction horizons with minimum error. The forecasted outcome suggests the daily contracted quantity of natural gas that falls within the limits of procurement without any penalty.