Abstract:

Energy consumption presents great interest in modern society due to the impact it has both on the climate change mitigation and on the smooth functioning of the economy. Building energy monitoring and planning has been particularly studied in order to find solutions to reduce energy loss and greenhouse gas emissions. Additionally, the European Commission has set ambitious goals for building energy efficiency and the use of renewable sources as part of reaching climate neutrality in 2050. In the context of the recently announced action plan REPowerEU, the stakes of having powerful predictive tools to support the use of renewable energy have risen once again. This research compares the performance by calculating the values for R-Squared, Mean Absolute Error and training time for the best models of two artificial neural networks architectures, FFNN and LSTM.