Abstract
Building energy modeling (BEM) is used to support (nearly) zero-energy building (ZEB) projects, since this kind of software represents the only available option to forecast building energy consumption with high accuracy. BEM may also be used during preliminary analyses or feasibility studies, but simulation results are usually too detailed for this stage of the project. Aside from that, when optimization algorithms are used, the implied high number of energy simulations causes very long calculation times. Therefore, designers could be discouraged from the extensive use of BEM to conduct optimization analyses. Thus, they prefer to study and compare a very limited amount of acknowledged alternative designs. In relation to this problem, the scope of the present study is to obtain an easy-to-use tool to quickly forecast the energy consumption of a building with no direct use of BEM to support fast comparative analyses at the early stages of energy projects. In response, a set of automatic energy assessment tools was developed based on machine learning techniques. The forecasting tools are artificial neural networks (ANNs) that are able to estimate the energy consumption automatically for any building, based on a limited amount of descriptive data of the property. The ANNs are developed for the Po Valley area in Italy as a pilot case study. The ANNs may be very useful to assess the energy demand for even a considerable number of buildings by comparing different design options, and they may help optimization analyses.
Highlights
In order to focus on one single illustrative example, this paper describes the development of one artificial neural networks (ANNs) able to generate an accurate estimate of the building’s yearly heating energy demand based on the variation of a large number of parameters that may be known or estimated during the course of feasibility studies
Due to the significant work required to execute the building energy performance simulations and to create the forecasting deep feedforward artificial neural networks (DFANNs), some appropriate tools and automated procedures were developed by the authors in the Python language as described in the following paragraphs
This paper deals with the generation of a set of artificial neural networks as a means to automatically assess the building energy demand from the very early design stages of a project, even in the case of interactive input/output frameworks, without the need of using any dynamic energy simulation software
Summary
Publisher’s Note: MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations. The design of zero-energy buildings (ZEBs) [1] is aimed at the achievement of the highest energy performance level in buildings, leading to the lowest energy demand as well as the lowest operating costs due to energy consumption. On the other hand, reaching ZEB standards implies extra costs in construction, which may increase by up to 15% [2]. An excellent energy performance level similar to ZEB standards may be reached through cost-effective design choices, finding the optimal balance between energy consumption and local generation. Slightly higher energy consumption may be accepted in order to save the cost of construction due to the building envelope or heating, ventilation and air-conditioning (HVAC) system
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
