Application of Neural Networks for Forecasting Energy Security Indicators

Elena Bykova, Irina Vasiliev, Valeriu Bosneaga, Victor Suslov

Abstract: Methodological approaches to the use of neural network models for constructing short-term forecasts of energy security indicators and energy balance indicators have been developed. Experimental neural forecast models for 2 years ahead (for 24 points) have been constructed based on actual monthly data during 7 years. The best models for the indicator "Gross consumption of natural gas" showed convergence at the level of 7%, and for the indicator "Gross consumption of electricity" - 2.14%.

Keywords: Energy security, indicators, neural network analysis, forecast