Energy performance assessment of campus buildings using data mining technologies

Buildings consume a large amount of energy during their life cycle. Building performance assessment plays a significant role in improving building energy efficiency and reducing greenhouse gas emissions to form a more sustainable environment and society. This thesis aims at developing data mining-based strategies for performance assessment of campus buildings to better understand their energy usage behaviour, assist their energy management, and improve their overall performance. In this thesis, a data mining-based strategy was first developed to identify typical daily electricity usage profiles (TDEUPs) of individual buildings by using Shared Nearest Neighbours and agglomerative hierarchical clustering (AHC). The developed strategy utilised the advantages of three dissimilarity measures, including Euclidean distance, Pearson distance and Chebyshev distance, to enhance the clustering results. The hourly electricity usage data collected from two university library buildings in Australia were used to evaluate the performance of this strategy. The results showed that the TDEUPs identified via this strategy can reflect the subtle changes in electricity usage behaviours of the assessed buildings in both daily scales and annual scales.