A two-stage methodology based on ensemble adaptive neuro-fuzzy inference system to predict carbon dioxide emissions

Abstract

Renewable energy should play a crucial role in increasing energy supplies and achieving the potential target of reducing 50% of CO2 emissions by 2050. The main objective of this study is to propose a neuro-fuzzy modelling entitled ensemble-Adaptive Neuro-Fuzzy Inference System (ANFIS) learning to predict and analyse the interrelationship between renewable energy consumption, economic growth, and CO2 emissions of G8+5 countries. This will help the governments and industry sectors to formulate energy policies and develop energy resources sustainably. The prediction method was constructed by extracting the fuzzy rules from the real-world dataset of World Development Indicators (WDI) and generalising the relationships of the inputs and output parameters for accurate prediction of CO2 emissions. The performance of the proposed method was evaluated, and the results show its efficiency in the prediction of CO2 emissions by incorporating the import indicators, including renewable energy consumption and economic growth. The U test of Sasabuchi–Lind–Mehlum (SLM) was conducted to identify the interrelationship results obtained from the ensemble ANFIS learning and the Environmental Kuznets Curve (EKC) hypothesis. The results of SLM test found an inverse U-shape condition among all countries except Brazil. The prediction of CO2 emissions trends using the soft computing approach (ensemble ANFIS) indicated that the consumption of renewable energy reduces CO2 emissions. The proposed soft computing method was found efficient in predicting CO2 emissions. It was in line with the foreseen targets of increasing the renewable energy generation and achieving the nationally determined contributions (NDCs) objectives.