Uncertainty and risk analyses are important tools for building designs and performance assessment of renewable energy systems. This task requires to account for the variability of the weather data. In this work, we develop a methodology to characterize and simulate stochastic weather data. The stochastic features of each weather input, such as auto-correlation and hourly cumulative distribution functions, are extracted from the dataset at hand. Then, the procedure of Iman and Conover is used to generate stochastic weather inputs. The approach is applied to a sequence of 1 month extracted from the typical meteorological year of the city of Lyon, France. The simulated stochastic weather data are employed to perform the uncertainty and sensitivity analysis of a real passive, low-energy house. The results show that the uncertainty on the predicted energy needs is roughly 20% and is essential due to the stochastic variability of the out- door air temperature.