Real-reference buildings for urban energy modelling: a multistage validation and diversification approach

Abstract

Urban energy models are required to establish baselines, map buildings' performance, and explore energy and CO2 emissions reduction strategies. The authors propose a new reductive bottom-up model to estimate the final energy demand for educational building stocks. For this purpose, this model relies on data-driven validated real-reference buildings and probability up-scaling. This model has three stages: (1) multivariate clustering techniques identify real-reference buildings from building stocks, (2) calibrated energy simulation estimate demand, uncertainty and allow to test scenarios, and (3) probability up-scaling diversifies building-level results regarding variations in their urban emplacement. Afterwards, the authors applied this model to two educational building stocks: a conditioned stock in Barcelona, Spain and a free-floating stock in Quito, Ecuador. Two real-reference schools represent 62% of the stock in Barcelona, with a final energy demand of 15.96 GWh/year. Likewise, two schools describe the entire educational stock in Quito with a final energy demand of 29.83 GWh/year. Results reflect Barcelona stock is more heterogeneous than Quito despite its lesser population. Specifically, their energy use intensity was 85.92 kWh/m2 with a 4.25% deviation in Barcelona and 40.23 kWh/m2 with a 0.11% deviation in Quito. This model is transferable among stocks because it relies on the characterization of buildings' thermal balance in their as-is state and provides good accuracy with building and urban energy records. This diversification procedure gave back part of the lost variability because of the reductive approach in the real-reference building definition. Also, it produced lower prediction errors for aggregated final energy use.