Model Estimation of ARMA Using Genetic Algorithms: A Case Study of Forecasting Natural Gas Consumption

Energy is accepted as a vital strategic issue all over the world due to the important hesitations/concerns about energy reliability, sustainability and affordability. The future of the any country's economy entirely depends on energy because it is the major input and indispensable resource for all sectors. Particularly, natural gas is a common used energy source for electricity generation, heating and cooking. Natural gas dependency on the foreign countries leads to economic damages for developing countries like Turkey, due to the high import costs. In this respect, precise forecasting of natural gas consumption plays crucial role in energy projections and economic progress. Underestimating natural gas demand leads to unsatisfied demand for both industrial and residential needs. In this study, we propose a forecasting method integrating Genetic algorithms (GA) and Autoregressive Moving Average (ARMA) method to take advantages of the unique strength of ARMA and genetic algorithms model. In order to predict natural gas consumption of Ä°stanbul, which is the most important metropolitan city of Turkey, with a lower percentage error and with a greater sensitivity based on penalty function. According to the experimental results, the developed combined approach is more robust and outperforms classical ARMA models in terms of mean absolute percentage error (MAPE) and cost function values.