Artificial Intelligence Approaches to Estimate the Transport Energy Demand in Turkey

Abstract

In this study eight parameters are selected and their historical data are collected to predict the future of the energy demand of Turkey. The initial eight parameters were the gross domestic product (GDP) of Turkey average annual US crude oil price (COP) inflation for Turkey in percentages (INF) the population of Turkey total vehicle travel in kilometers for Turkey total amount of goods transported on motorways employment for Turkey and trade of Turkey. However after these eight parameters data are analyzed using Pearson and Spearman correlation methods it is found out that five of these parameters are highly correlated. The remaining three parameters are the GDP of Turkey COP and INF for Turkey. Afterward five separate scenarios are developed to forecast the future of the energy demand of Turkey. The first two scenarios involve the third- and fourth-order polynomial fitting the third and fourth scenarios employ static and recurrent neural networks and the fifth scenario utilizes autoregressive models to predict the future energy demand of Turkey. The efficient hybridization of the seagull optimization and very optimistic method of minimization metaheuristic algorithms is carried out to achieve the polynomial fitting of the data. The optimization performance of the hybrid algorithm is assessed by applying the algorithm on benchmark optimization problems and comparing the results with that of some other metaheuristic optimizers. Moreover it is seen that the forecasts of the first scenario agree well with the Ministry of the Energy and Natural Resources estimates.