Electricity Consumption Prediction using Energy Data, Socio-economic and Weather Indicators. A Case Study of Spain

Electricity consumption analysis and forecasting is a vital topic of research in the energy sector due to its important role in the reduction of energy waste and improvement of energy-related environmental issues. Spain is one of the largest energy consumers in Europe which consumption rates at the country level are not profoundly examined. This study presents a multivariate modeling methodology for Spain’s electricity consumption prediction based on 14 socio-economic, 3 energy-related, and 2 meteorological variables using machine learning and statistical analysis tools. Historical data covering the period from January 2008 to December 2019 with monthly periodicity is used. The obtained experimental results show that the electricity consumption forecasting model combining a random forest-based feature selection method, a wavelet transform-based denoising scheme and a statistical forecasting tool called SARIMAX produces lower error rates (RMSE=0.125 and MAPE=2.85%) compared to other existing state-of-the-art techniques used for the same dataset.