Analysis of Electricity Consumption in Poland Using Prediction Models and Neural Networks

Abstract

The challenges of the modern world require transformations in the energy market towards the possible reduction of consumption and greater use of renewable sources. The conducted research of consumers of this market confirms that the behaviour in the field of increased use of renewable energy is burdened with cognitive errors and motivational factors, which makes it difficult to conduct quantitative research. Electricity demand forecasting can be modelled using selected quantitative methods. In this way, not so much the behaviour, but the result of the consumer’s behaviour is predicted. The research presented in the article has been divided into two parts. The aim of the first one is to study the prospects of a greater share of renewable sources in obtaining energy in Poland, based on the attitudes and opinions of consumers on the retail energy market, legal regulations and the energy balance. The aim of the second part is to build forecasts of daily, weekly, monthly and quarterly electricity consumption in Poland, including the prediction of the RES share, using selected machine and deep learning methods. The analyses used the time series of daily electricity consumption in Poland from 2015–2021; the ENTSO-E data was obtained from the cire.pl website. Depending on the adopted forecast horizon, the forecasting method with the lowest MAPE error was exponential smoothing, SARIMA and NNETAR. An evolution of energy consumers’ attitudes towards pro-ecological and pro-social sensitivity and understanding of the importance of RES for the economy was also observed.

Keywords

• energy consumer research
• energy consumption forecasting
• RES
• machine and deep learning