The reliable assessment of building energy performance requires significant computational times. The chapter handles this issue by proposing an original methodology that employs artificial neural networks (ANNs) to predict the energy behavior of all buildings of an established category. The ANNs are generated in MATLAB by using EnergyPlus simulations for testing and training purposes. The inputs are properly set by means of a thorough preliminary sensitivity analysis. The final aim is a reliable assessment of the global cost for space conditioning as well as of the potential global cost savings produced by energy retrofit measures for each category's building. The benefit is a huge reduction of computational times compared to standard reliable simulation tools. Definitely, this can support the diffusion of rigorous approaches for cost-optimal energy retrofits. Beyond the presentation of the methodology, this is applied to the office building stock of South Italy built in the period 1920-1970. The results show a high ANN reliability compared to EnergyPlus simulations, with regression coefficients (. R) always higher than 0.98.
|Titolo:||Artificial Neural Networks for Predicting the Energy Behavior of a Building Category: A Powerful Tool for Cost-Optimal Analysis|
|Data di pubblicazione:||2017|
|Appare nelle tipologie:||2.1 Contributo in volume (Capitolo o Saggio)|