Big Data Analyses Show New Ways for the Prediction of Energy Consumption for Sustainability Assessment Reduction to relevant input data for accurate energy predictions of existing residential buildings

The energy consumption during the use phase of buildings is largely responsible for the overall energy necessary over the lifespan of buildings. However, different boundary conditions like climate and weather as well as different indoor conditions can lead to disparities between the predicted and actual energy consumption. A new approach is the use of Artificial Neural Networks (ANN) in order to increase the accuracy in determining the energy consumption of existing residential buildings. Previous studies have already demonstrated the potential of this method. For this work, measurement data of one- and two-family houses were used which allowed the predicted energy consumption by a newly developed ANN algorithm to be directly compared to actual energy consumption. The combined data set includes approximately 14.000 validated data tuples, leading to 1^st hand data. An additional building component condition factor (BCC-factor) was created from the existing data (2^nd hand data). Even though only a limited amount of data and input parameters could be used for this study, the accuracy of the implemented ANN algorithm produced promising results with a weighted mean absolute percentage error (WMAPE) of 22 % and a mean absolute error (MAE) of the energy demand of 41 kWh/(m^2a).