Forecasting of Turkey's Total Electricity Consumption in Sectoral Bases Using Machine Learning Algorithms

Thumbnail Image
Journal Title
Journal ISSN
Volume Title
İstanbul Kültür Üniversitesi
Research Projects
Organizational Units
Journal Issue
Turkey is considered to have one of the fastest-growing economies in the world and electrical energy is a milestone in the economic growth of each country. Therefore, it is important to have an idea about the upcoming electricity consumption. Various methods were used in previous studies for forecasting electricity consumption. This study forecasts the sectoral and total electricity consumption in Turkey until the year 2050. This study utilizes two distinct Time series forecasting methods, namely Multilayer perceptron (MLP) and sequential minimal optimization (SMO) as a model to generate the forecasting formulas. The sectoral and total electricity consumption for Turkey from the year 1970 to 2020 was obtained from the Turkish Statistical Institute and fed to the models to forecast the upcoming years. The two models were evaluated and compared using determination coefficient R2 and mean absolute percentage error (MAPE). It is found that SMOReg was superior in forecasting Turkey's total electricity consumption where the R2, MAPE, and root mean square error (RMSE) for the SMOReg model were 91.42%, 2.98%, and 8558.85 GWh respectively, where the forecasted total electricity consumption in sectoral bases reaches 292,142 GWh in 2021 and 777,854 GWh in 2050. SMOReg also performed better in forecasting the industrial and household sectors. Whereas MLP performed better in forecasting the commercial, governmental, illumination, and other sectors.
▪ Yüksek lisans tezi.
Forecasting electricity consumption, Machine learning, Artificial neural network, WEKA, Multilayer Perceptron, SMO Regression