Fossil fuel-related CO2 emissions modelling in Ukraine using Multiple Linear Regression and Artificial Neural Networks

Abstract

Abstract The constant growth of carbon emissions is one of the main causes of global warming, which in turn leads to the adverse environmental effects involving a risk of droughts, wildfires, flooding, glacier melting, etc. Ukraine is not among priority countries for greenhouse gases emitters. However, from both, an economic and environmental points of view, monitoring and on-time analysis of CO2 emissions will help beforehand to determine the main drivers of CO2 emissions and, thus, will serve as a base for government to set a number of programs on reducing of greenhouse gases (GHG) or adapting to it. The aim of this paper is to offer the mathematical model for fossil fuel-related CO2 emissions forecasting using statistical technique of Multiple Linear Regression (MLR) and computing method of Artificial Neural Networks (ANN). Three different models are obtained to predict CO2 emissions from coal, oil, and natural gas consumptions taking into account the main carbon drivers. Based on the accuracy assessment analysis, models derived with ANN reveals in more accurate prediction than those obtained with MLR. Therefore, ANN models can be applied while planning several steps ahead and planning out every conceivable worst-case scenario, protecting against it.