Journals Proceedings

Abstract

The main objective of this study is to present a more accurate method to estimate the energy performance of buildings. This purpose is meant to evaluate the feasibility and relevance of more complex statistical modeling techniques, such as the artificial neural network. The energy performance of buildings may be estimated by their capacity to ensure a healthy and comfortable environment, with low energy consumption during the whole year. The glazed areas have a decisive role in the building energy performance having in view the complex functions that they play in the system. A parametric study, based on another powerful tool - the design of experiment method, which allows us to emphasize the measure in which the geometric and energetic characteristics of glazed areas influence the energy efficiency, estimated by the yearly energy needs, to ensure a comfortable and healthy environment. An artificial neural network - ANN is a computational model inspired by the biological natural neuron. The complexity of real neurons is highly abstracted by mathematical equation when modeling artificial neuron. This transforms the input data in output data depending on the operator’s ability of choosing and connecting more neurons or more layers for obtaining the expected performance. The neuron’s capacity of learning and adapting to operator demands makes a useful tool in math modeling and optimization of nonlinear processes. ANN presents a high potential of adaption to mathematical modeling of processes or phenomena of the black box type, generally with a pronounced nonlinear character and which are difficult to describe and model with simple mathematical models. ANN has the ability to solve new problems by applying information learned from past experience, as the human brain does.