Bayesian calibration of building energy models for uncertainty analysis through test cells monitoring

Abstract

Improving the energy efficiency of existing buildings is a priority for meeting energy consumption and CO2 emission targets in buildings. Building simulation tools play a crucial role in evaluating the performance of energy retrofit options. In this paper, a Bayesian calibration approach is applied to reduce the discrepancies between measured and simulated temperature data. Through its application to a test cell case study, the incorporation of sensitivity analysis and Bayesian calibration techniques are proven to improve the level of agreement between on-site measurements and simulated outputs, whilst accounting for both experimental and simulation uncertainties. The accuracy of a building simulation model developed using EnergyPlus was evaluated before and after calibration. Uncalibrated models were within the uncertainty ranges specified by the ASHARE Guidelines, with hourly simulation data over-predicting measurements by 3.2 °C on average. After Bayesian calibration, the average maximum temperature difference was reduced to around 0.68 °C, an improvement of almost 80%.