Comparison of the Energy Performance of Existing Buildings by Means of Dynamic Simulations and Artificial Neural Networks

The energy efficiency of buildings can be evaluated by using the methodology provided by European regulations; however, this method required a lot of information which is generally not available for the existing buildings. In this paper an alternative method for the energy efficiency investigation of buildings is proposed and tested by means of Artificial Neural Networks (ANNs); an existing building built in 1990 and located in Perugia was chosen as case study and it was investigated by adopting both the mentioned approaches. An experimental campaign was also carried out in order to implement and validate the 3D model developed in TRNSYS. Results showed that the indoor air temperature trend simulated with ANN is closer to the measured data than the one simulated with TRNSYS, with lower mean error and MSE values. The energy consumption simulated with ANN is slightly higher than the one returned by using TRNSYS code of about 20 kWh/m2year (difference lower than 7%). In agreement with the results, the proposed method can be considered as an alternative tool that can be used for the thermal-energy investigation of existing buildings, with important money and time saving.