U-value data on an urban scale: Outlier detection using comparative thermography to improve data quality

Due to climate change and limited energy resources, it is becoming increasingly important to analyse and optimise the energy performance of existing building stocks. Urban building energy modelling is the most common tool for such tasks, however difficulties arise from incomplete or outdated datasets, which are often updated using subjective assumptions. This study proposes a methodological framework to improve the quality of existing U -value datasets using comparative thermography. As the dataset may contain incomplete or outdated values, it is extended to a comparative U -value assessments based on a thermographic dataset. Outliers are identified at points that show discrepancies between the two datasets. The presented method for a comparative approach provides the user with a strategy to obtain high -quality urban scale U -value datasets with a fraction of the current effort. The comparative approach can reduce the effects of uncertain parameters on thermographic U -value estimation. Within an experimental case study, the concept of outlier detection is applied successfully to improve existing datasets for urban energy modelling. The presented method was able to reduce the workload of detailed U -value investigation to 12% compared to a full archival research. These results represent significant time saving and enabling the user to find inconsistent values in datasets in an objective, unbiased and reasonable way.