Prediction of energy consumption in residential buildings before and after retrofiting using artificial neural networks

Abstract

This paper presents the development of a new method of energy consumption prediction in residential buildings taking into consideration the great differences between the standard modeling simulations and the real conditions. The novelty of this method is that energy consumption is determined based on real data collected from numerous real cases instead of standard old norms, leading to a more accurate prediction. This method takes into consideration the nonlinearity relations between all the measurable variables and the final energy consumption, without being restricted to standards and norms. To this end, several artificial neural networks were built, trained, and tested, generating computer software that can be used for verifying and proving the accuracy of the new method in predicting energy consumption in retrofitting residential buildings.