Probabilistic prediction of green roof energy performance under parameter uncertainty

Studies on the quantification of energy benefits of a green roof have so far treated its parameter values only deterministically. In reality, however, these values may scatter over different ranges due to the inherent variation of vegetation and soil properties and also because of the unavoidable deviation from designated values during construction and/or actual operation of a green roof. Under such parameter uncertainty, green roof performance can no longer be predicted deterministically but rather probabilistically. The present study attempts to integrate the whole building energy simulation with a parametric uncertainty analysis. An example office building is used to systematically examine how the cooling and heating energy demands can be reduced by a green roof that replaces a conventional roof, when values of the most significant green roof parameters determined by sensitivity analysis are treated as random variables with prescribed probability distributions. An ensemble of green roof configurations is generated using Monte Carlo simulation with a Latin hypercube sampling technique. The coefficient of variation of the calculated energy savings is found almost linearly related to (with a slope of about 0.4) that of green roof parameters. Finally, implications of probabilistic energy analysis for more reliable green roof design are emphasized.