A probabilistic building characterization method for district energy simulations

To assess the impact of implementing energy efficiency and renewable measures, urban building models are emerging. In these models, due to lack data, natural variability existing stock is often highly underestimated uncertainty on simulated use arises. Therefore, this work proposes a probabilistic characterization method model residential stock. The estimates realistic distributions five input variables: U-values floor, external walls, windows roof as well window-to-wall ratio, based known data (location, geometry construction year). First, quantile regression has been implemented generate uncorrelated Flemish performance certificates database. accuracy marginal good, empirical coverage 50%, 80%, 90% 98% prediction interval deviates 0.6% at most. However, it needed include correlations between variables. Hence, three main methods build multivariate from draw correlated samples extensively compared. Gaussian copula put forward preferred method. Considering mean-maximum discrepancy (MMD), performs eight times better than case (MMD 0.0027 versus 0.0228).