Forecasting the natural gas consumption for the period of 2018 to 2028, the south of KZN province.

Abstract

An upward trajectory in the consumption of natural gas around the globe as a source of energy to the industrial, residential, and commercial sector is evident. Gas consumption increase or reduction is closely correlated to the energy demand. Growth in the manufacturing industry can have a similar effect on energy and gas consumption demand. Gas industry in South Africa is relatively new and growing. The current gas consumption growth has led to the formation of the gas market structure which, comprises of exploration, transmission, distribution, and reticulation. The continued gas consumption growth requires additional investment since a rise in the capacity of the gas infrastructure would be expected. End-user gas consumption influences the increase in gas consumption which yields the imbalances in the forecast of natural gas and if actual gas consumption is lower than 80% of the initial forecasted gas, contractual penalties apply. The study is motivated by the value of the penalty that comes with the 80% take or pay contract condition. Spring Lights Gas and other traders are contractually bound by the same contract condition and its critical for the company to submit correct gas quantity to the gas producers to avoid penalties. The study focuses on the South African gas industry, particularly the southern region of KwaZulu Natal. The gas consumption time series is inherently non-stationary due to the different production plans and unplanned equipment repair which can cause a swing in the gas consumption. The study aims to forecast future gas consumption for Spring Lights Gas with minimum error. The actual monthly consumption time series data of the KZN southern region for the period of June 2005 to July 2018 was collected from Spring Lights Gas and used to model and estimate the univariate Autoregressive Integrated Moving Average (ARIMA) method which was compared to the bivariate Vector Error Correction Model (VECM) method. The performance of both models was determined by forecasting the in-sample data from 2016 to 2018. The univariate ARIMA method was chosen because it generated low Root Mean Square Error (RMSE), Mean Absolute Error (MAE) and Mean Absolute Percentage Error (MAPE) compared to the bivariate VECM method. Consequently, the ARIMA model was used to forecast the future gas consumption for Spring Lights Gas and produced RMSE of seven percent compared to the official future gas plans of the company. The results further revealed that gas consumption in the south region is projected to grow by one percent for the next decade and it was recommended that the company revise down its future gas plans.