Calibration of building model based on indoor temperature for overheating assessment using genetic algorithm: Methodology, evaluation criteria, and case study

With the increased severity, intensity, and frequency of heatwaves due to climate change, it has become imperative to study the overheating risks in existing buildings. To do so, a building simulation model needs to be calibrated based on measured indoor temperatures under the current weather conditions. This paper presents a robust automated methodology that can calibrate a building simulation model based on the indoor hourly temperature in multiple rooms simultaneously with high accuracy. This methodology includes a variance-based sensitivity analysis to determine building parameters that significantly influence indoor air temperatures, the Multi-Objective Genetic Algorithm to calibrate different rooms simultaneously based on the significant parameters identified by the sensitivity analysis, and new evaluation criteria to achieve a high-accuracy calibrated model. Maximum Absolute Difference (MAD), a new metric, that calculates the maximum absolute difference between simulated and measured hourly indoor temperatures, Root Mean Square Error (RMSE), Normalized Mean Bias Error (NMBE) were used as the evaluation criteria. Another new metric is introduced, 1 degrees C Percentage Error criterion that calculates the percentage of the number of hours with an error over 1 degrees C during the calibration period, to select the best solutions from the Pareto Front solutions. 0.5 degrees C Percentage Error criterion is also used for the level of accuracy the model can achieve. It was found that the calibrated model achieved these metrics with RMSE of 0.3 degrees C, and MAD of 0.8 degrees C, and 85% of data points with an error less than 0.5 degrees C for a school building case.

Automated summary

This paper presents a robust automated methodology for calibrating a building simulation model based on indoor temperature data. The methodology utilizes sensitivity analysis, multi-objective genetic algorithm, and new evaluation criteria to achieve high accuracy calibration. The results show successful calibration for a school building case using the proposed methodology.