Reflecting energy use patterns and lifestyles in home using data mining techniques

Most methods to analyse and understand the residential energy use features rely on invasive measurements, such as energy monitoring systems, which eventually affects the reliability of pattern classifications. This paper, thus, adopts a non-invasive method using unsupervised data mining algorithms to analyse hourly energy consumption data in order to learn the occupant's lifestyle and energy consumption behavioral patterns. The study analyses hourly energy use of 298 households in Texas in 2015, using an online open source data set - Pecan Street Dataport. This study scientifically identified household's energy use features and associated behavioural patterns through a multi scale observation of the clusters. As the contribution, this study takes the house age and size into account as these variables may significantly affect building energy use patterns. Second, it takes dissimilarity measures into account by using TSclust R package for clustering time series. And third, introduces a method of multiscale observation of clusters in order to interpret the lifestyle patterns. Finally, the results demonstrated how data mining techniques might be utilized to help investigating energy use data from the behavioural perspective.