Evaluating the impact of the shape of school reference buildings on bottom-up energy benchmarking

Geraldi, M.S.; Gnecco, V.M; Neto, A.B.; Martins, B.A.M.; Ghisi, E.; Fossati, M.; Melo, A.P.; Lamberts R.
Journal of Building Engineering

Bottom-up energy benchmarking methods use simulation of reference buildings to reproduce the building stock performance. Regressive methods using those simulated outcomes are employed to obtain typical energy performance values according to certain parameters to make the benchmarking a fair comparison. Those parameters are commonly related to the building construction aspects, operation and climate conditions. However, the building shape might play an important role in its energy performance, especially due to self-shading. This paper aims to evaluate the impact of using different reference buildings’ shapes to develop a bottom-up benchmarking of schools in Brazil. Seven reference buildings were obtained through a building stock analysis and simulated for different scenarios. Regressive benchmarking models were developed for each shape, and the impact of different building shapes was measured through a cross-validation step. Finally, an ANOVA test was applied to determine the variance of using different regressive benchmarking models in different building shapes. Results showed a significant difference in benchmarking buildings using regressions built using different shapes of reference buildings. Average RMSE ranged from 11.79 to 6.79. The best regression model was the one built using all shapes together. Additionally, results included the benchmarking models with adequate R2 (ranging from 0.725 to 0.885), and the sensitivity analysis indicates the city, the air-conditioned area and the shape as the relevant variables for benchmarking. Conclusions pointed out that the shape is an important factor in bottom-up benchmarking models and, by testing the models using actual cases, the benchmarking result might change significantly according to the reference building shape used to build the benchmarking model.