Forecasting the natural gas demand of Türkiye: A deep learning approach

Abstract

Natural gas is an important input for the production of electricity and other forms of energy. The aim of this study is to investigate the relationship between the natural gas consumption, population and the gross domestic product of Türkiye. For this purpose, deep learning methods are utilized for the modelling of the natural gas consumption for the data covering the period of 1982-2021. First of all, the data required for the modelling are taken from the official sources and then the Granger causality relationship among these variables are studied together with the seasonal and trend decomposition. After assessing the nature of the data that contain high seasonality and nonlinearity, a deep learning network is developed in Python programming language. It is visually demonstrated that the developed deep learning model can be successfully used to describe and forecast the natural gas consumption dependent on the population and the gross domestic product. The performance of the developed deep learning model is also verified using the performance metrics such as the coefficient of determination and the mean absolute percentage error. The developed model is shown to be useful for energy planners and for economists dealing with the energy pricing.