ABSTRACT

Forecasting natural gas demand for Pakistan has become a demanding task due to numerous factors, such as government interventions that decide to shift or halt the supply of natural gas to specific sectors. Secondly, undocumented subsidies in terms of pricing to the domestic, fertilizer and industrial sectors also distorted natural gas consumption patterns. Exploring an appropriate forecasting model for natural gas in Pakistan is challenging. Because forecasting at an aggregated level, such as annually, overlooks the seasonal factors crucial for demand projection, which previous studies have completely ignored. Furthermore, exploring factors driving demand for sectoral natural gas is another complex issue. In this study, Pakistan forecasts natural gas demand on a disaggregated level (monthly) for both the national level and individually for each sector, considering economic features (demand drivers) and excluding them. The study looked at traditional econometrics (SARIMA), machine learning (XGBOOST), and deep learning (LSTM) techniques for making predictions. It found that Extreme Gradient Boost (XGB) from machine learning is the most accurate. Moreover, comparing Extreme Gradient Boost with and without economic features (each sector’s own growth, prices for each sector, urbanization and electricity, etc.) for monthly forecasting analysis shows that without economic features, XGB has a lower forecast error. Finally, this study investigated how to decide whether to forecast natural gas demand at the national level (including all sectors) or each sector separately. Interestingly, the results reveal that forecasting natural gas demand at the national level has higher forecast accuracy because government decisions to allow or halt natural gas supply to different sectors have distorted the data-generating process of the natural gas consumption dataset. So while moving towards last objective ofthis research study, we found that there is enough locally produced natural gas reserves we have till 2027 with the same projected steady demand, and if the government and gas companies decide to meet the peak demand for the domestic sector in the winters (Nov, Dec, and Jan), in this scenario, the government wouldn’t be able to decrease the circular debt that’s currently around Rs 3.022 trillion, sources close to the caretaker Minister for Power and Petroleum told Business Recorder. 19-Jan-2024, and the cost is quiet to build new infrastructure for domestic sector peak demand instead of cutting subsidies for domestic sector. Also, in this study’s last objective, we visualize the gap between demand and supply, which actually shows production is decreasing over the years and demand is steady over the years. Whereas, for a more clear picture, the shortage in winter is actually peak demand in the domestic sector, for which the government needs to revise its policies for effective natural gas demand management. As well as for sustainable and affordable supply for all sectors, instead of importing LNG.