A new hybrid gravitational search–teaching–learning-based optimization method for energy demand estimation of Turkey

Abstract

In this study, energy demand estimation (EDE) was implemented by a proposed hybrid gravitational search–teaching–learning-based optimization method with developed linear, quadratic and exponential models. Five indicators: population, gross domestic product as the socio-economic indicators and installed power, gross electric generation and net electric consumption as the electrical indicators, were used in analyses between 1980 and 2014. First, the developed models were trained by the data between 1980 and 2010, and then, accuracy of the models was tested by the data between 2011 and 2014. It is found that the obtained results with the proposed method are coherent with the training data with correlation coefficients in three models as 0.9959, 0.9964 and 0.9971, respectively. Root mean square error values were computed 1.8338, 1.7193 and 1.5497, respectively, and mean absolute percentage errors were obtained as 2.1141, 2.0026 and 1.6792%, respectively, in the three models. These values calculated by the proposed method are better than the results of standard gravitational search algorithm and teaching–learning-based optimization methods and also classical regression analysis. Low, expected and high scenarios were proposed in terms of various changing rates between 0.5 and 1.5% difference in socio-economic and electrical indicators. Those scenarios were used in the EDE study of Turkey between 2015 and 2030 for a comparison with other related studies in the literature. By the proposed method, the strategy in energy importation can be regulated and thus more realistic energy policies can be made.