Comparisons of Deep Learning Models to predict Energy Consumption of an Educational Building

According to the latest United Nations Environment Programme report, in 2020, the construction and operation of buildings globally accounted for more than a third (36%) of final energy used and 37% of carbon dioxide emissions. This has given rise to research interest in energy efficiency in buildings in recent years, considering different approaches. In this work, black-box approaches based on bio-inspired techniques to predict the energy consumption of a university building are proposed. Three artificial neural network (ANN) architectures. Feed Forward Neural Network (FFNN) such as multilayer perceptron (MLP) and recurrent neural network (RNN) such as long short-term memory (LSTM), and gated recurrent unit (GRU). Three scenarios with different resolution timesteps and forecast horizons multi-step are tested for each architecture. The results according to the comparison of analyzed metric and computational execution times, the MLP model ranks first. These models can be very useful in predictive control systems considered in buildings to forecast the behavior of buildings' electric load in the short term with very good precision.