Electricity production based forecasting of greenhouse gas emissions in Turkey with deep learning, support vector machine and artificial neural network algorithms

Bakay M. S., Ağbulut Ü.

Journal of Cleaner Production, cilt.285, 2021 (SCI-Expanded) identifier

  • Yayın Türü: Makale / Tam Makale
  • Cilt numarası: 285
  • Basım Tarihi: 2021
  • Doi Numarası: 10.1016/j.jclepro.2020.125324
  • Dergi Adı: Journal of Cleaner Production
  • Derginin Tarandığı İndeksler: Science Citation Index Expanded (SCI-EXPANDED), Scopus, Academic Search Premier, PASCAL, Aerospace Database, Business Source Elite, Business Source Premier, CAB Abstracts, Communication Abstracts, INSPEC, Metadex, Pollution Abstracts, Public Affairs Index, Veterinary Science Database, Civil Engineering Abstracts
  • Anahtar Kelimeler: CO2 emission, Electricity production, GHG, Greenhouse gases, Machine learning algorithm
  • Yıldız Teknik Üniversitesi Adresli: Hayır

Özet

Today, the world's primary energy demand has been met by the burning of fossil-based fuels at a rate of 85%. This dominant use of fossil-based fuels has led to an accelerating increase in the release of greenhouse gases (GHG) all across the world. The largest share in total GHG emissions belongs to the electricity and heat production sector with a rate of 25%. With this viewpoint, this paper is aiming to forecast the GHG emissions (CO2, CH4, N2O, F-gases, and total GHG) using deep learning (DL), support vector machine (SVM), and artificial neural network (ANN) algorithms from the electricity production sector in Turkey. The dataset is supplied from the Turkish Statistical Institute and covers the years 1990–2018. In the study, the last four years (2015–2018) is being forecasted. To evaluate the performance success of the algorithms, five metrics (RMSE, MBE, rRMSE, R2, and MAPE) are discussed in detail. In the results, this research is reporting that all algorithms used in the study are giving separately satisfying results for the forecasting of GHG emissions in Turkey. Based on the forecasting results, it is seen that the highest R2 value for the emissions varies from 0.861 to 0.998 and all results are categorized as “excellent” in terms of rRMSE (all rRMSE values < 10%). Besides, MBE changes between −2.427 and 2.235, and all MAPE values are smaller than 1.2%. Total GHG emission is forecasted in DL algorithm with very satisfied R2, RMSE, MBE, rRMSE, and MAPE of 0.998, 2.046, 0.419, 0.406%, and 0.021%, respectively. On the other hand, CO2 accounted for 69.05% of total GHG emissions of Turkey in 1990 but rising by 80.48% in the year 2018. In comparison with those of 1990, electricity production and total GHG emissions of Turkey in 2018 increased by 429.7% and 137.4%, respectively. Total GHG emission corresponding to electricity production is calculated to be 0.3813 Mt-total GHG/MWh in 1990 and 0.1709 Mt-total GHG/MWh in 2018. In conclusion, GHG emissions have recently increased at a high rate, but it is noticed that this increase is considerably higher as compared to the increase in energy production for Turkey.