Multiple regression models to predict the annual energy consumption in the Spanish banking sector

This paper presents a regression analysis of energy consumption in the banking sector. In our case study, the target area is the Spanish banking sector, for which we divide the available data into a prediction and a validation subset. Power models were developed using test data from 55 banks. From the analysis, three models were obtained; where the first proposed model can be used to predict the energy consumption of the whole banking sector, while the rest of the models estimate the energy consumption for branches with low winter climate severity (Model 2) and high winter climate severity (Model 3). Models 2 and 3 differ from the first model in that they need independent variables measured in situ. As a result, the uncertainty of the response variable in the function of the independent variables is reduced by 56.8% for the first model and by 65.2% and 68.5% for the second and third proposed models, respectively. The validation of the first model, which is the model with the lowest determination coefficient, shows that this model is appropriate for predicting the energy consumption of bank branches with good energy consumption performance and detecting inefficiencies in bank branches with poor energy consumption performance.

► The paper presents an analysis of the energy consumption in the banking sector.
► Three models for the prediction of the annual energy consumption are presented.
► The first model presented allows for the detection inefficient bank branches.
► The rest of models can achieve a better adjustment.