Comparison of Long-term Energy Demand Forecasting in Developing and Developed Countries Using Machine Learning-based Algorithms

As the world’s population grows, many nations grapple with the task of supplying sufficient energy. One effective way to manage and plan for this demand is through energy demand forecasting. In this research, we adopt a method employing machine learning algorithms to predict energy demand in various countries, both developed and developing, up to the year 2050. For our long-term forecast covering 2020 to 2050, two types of historical data are utilized: (i) energy consumption (EEC), and (ii) socio-economic indicators, such as Gross domestic product (GDP), energy import and export, and population. An Artificial Neural Network (ANN) based machine learning is employed utilizing 30 years’ worth of socio-economic data (1991-2020). In this manner, the utilization of the ANN facilitates the prediction of long-term energy consumption (EEC) for the planning period spanning from 2020 to 2050. Furthermore, we present an optimization model designed to enhance the precision of our predictions. The outcomes from machine learning algorithms serve as input for our comprehensive model, implemented through ANN across various sections. Ultimately, our forecasts indicate a projected increase in electric energy consumption by $130.1 \%$ for Iran, 37.4% for Portugal, and 58.6% for the United States in 2050 compared to 2020. Significant distinctions exist between developing and developed economies with regard to the anticipated trends in Energy Efficiency Compliance (EEC).