Uncertainty and modeling energy consumption: Sensitivity analysis for a city-scale domestic energy model

► The overall structure and form of the BEDEM model is presented. ► Local sensitivity analysis is carried out for various dwelling types. ► Input uncertainty effects were linear for a moderate range of input change. ► Input uncertainty effects were superposable for a small range of input change. ► The uncertainty in the stock models predictions can be large. This paper presents the development and evaluation of the Belgrade Domestic Energy Model (BEDEM) for predicting the energy consumption and carbon dioxide (CO2) emissions of the existing housing stock. The distribution of energy use in relation to the end use is estimated as: space heating, 71%; light and appliances, 15%; water heating, 9%; and cooking 5%, while the distribution of CO2 emissions is space heating, 59%; light and appliances, 22%; water heating, 13%; and cooking 6%. Local sensitivity analysis is carried out for dwellings of different type and year built, and the largest normalized sensitivity coefficients were calculated for parameters which almost exclusively influence space heating energy consumption in housing. For all input parameters under investigation, the effects of the input uncertainty were linear for a moderate range of input change (Δx=±10%) and superposable for a small range of input change (Δx=±1%). However, the non-linear and non-additive properties of some input parameters over the wider range hinder the development of a simple but reliable model for estimating energy and CO2 reductions. The findings show that the uncertainty in the stock models predictions can be large and more work is needed in the area of the predictive uncertainty of stock models.