Análises de sensibilidade e propagação de incertezas para modelagem com prioridades na avaliação do desempenho de eficiência energética de edificações pela INI-C

Abstract

The need for global CO2 emission reduction has contributed to the mandatory building energy performance targets. The performance of new buildings or retrofits is predicted by computer simulation according to detailed regulations and protocols procedures, which are often laborious to provide reliable results. The introduction of statistical resources for determining the results' uncertainty and the popularization of metamodels simulation allowed the identification of the most influential building characteristics on energy performance to prioritize the most critical uncertainties without compromising the accuracy of the results. This thesis proposes an approach with priorities to optimize envelope modeling and control the building energy performance assessment using the INI-C energy classification, whose model characteristics' uncertainty and variations become discriminated based on their impact on the results. The behaviors of different envelope combinations and shapes are assessed for the warm and humid climate of Natal/Brazil, with the predominance of cooling loads. The sensitivity and uncertainty analyses are related to each other to optimize the process and identify the priority variables in the cooling thermal load, using the method based on Sobol's variance, scripts in R language, and the ANN metamodel of INI-C for commercial and office buildings. The propagation of uncertainties analyzes the uncertainty of isolated variables and the cumulative effect of uncertainties, including tests of uncertainty limits comparing the approach with priorities within the limits of RAC INI-C. The results demonstrate that the prioritized approach can simplify energy performance evaluations without compromising accuracy once input sensitivities are determined using a metamodel to discriminate them. The procedures required commonly accessible computational resources, essential to identify the impact of characteristics, which depends on the context of the building, absolute value, variation, and combination, among others.