Forecasting Turkey's Energy Demand Using Artificial Neural Networks: Three Scenario Applications

摘要

Energy has become increasingly crucial for countries as we have experienced high economic growth, increases in population together with rapid urbanization in the globalized world. Turkey's energy demand has grown rapidly and is expected to continue growing. In this context many studies have been carried out to forecast energy demand in Turkey. The energy demand forecasts are officially prepared by the Turkish Ministry of Energy and Natural Resources (MENR). However, MENR forecasts are significantly higher when compared with realized demand and the results of other academic studies. In this study, Turkey's energy demand is forecasted by using artificial neural network technique, a type of artificial intelligence application. For this purpose, three different scenarios are developed. These are: 'static scenarios', where economic growth is assumed to be stable, 'sustainability scenarios', where energy intensities are assumed to be decreasing and finally 'periodic-change scenarios', where the economic growth is assumed to change during five different time periods by 2030. Moreover, both static and sustainability scenarios are further investigated under high, medium and slow economic growth assumptions. Periodic-change scenarios also consist of two sub-scenarios, where energy intensities are assumed to decrease and stay the same. All scenarios are applied to the total energy demand of Turkey. The results of the energy demand estimations found by our models are compared with the official estimations of the MENR. It is concluded that the MENR estimations are significantly higher than what we have found with our models.