Multivariate Methods for Evaluating Building Energy Efficiency

For an owner or operator of a building, benchmarking can be a useful guide for finding out how energy efficient the building is and identifying what to improve. For successful evaluation of the building energy efficiency, the categorization as well as the parameter identification has decisive importance. That selection can be based on mathematical modeling such as linear regression accompanied with more or less user expert knowledge. The selection, however, is not a simple task since analyses based on statistical data are sensitive to correlations between different measured parameters. For improving that analysis multivariate methods such as Principal Component Analysis (PCA) can be a valuable support. We demonstrate here how PCA can be a useful tool for investigating aggregated statistical datasets. The investigation illustrates how a set of building performance parameters exhibits different relations depending on how the categorization is made, which is relevant to consider when benchmarking. The study is based on a national Swedish database of aggregated energy use and building performance statistics.