Energy Consumption Profiling Of Appliances Inside Smart Buildings Based On k-means Clustering

Smart Building Energy Management Systems schedule appliances when it is more convenient, e.g. during off-peak times or when power is produced by Renewable Energy Sources (RES). In order to do so, it is necessary for the system to know the power consumption profiles over time of the appliances present in the building. Nevertheless, obtaining this information with sufficient detail is not straightforward, since appliance consumption varies significantly over time as compared to the average consumption. Furthermore, considering appliances with different working cycles, i.e., different power consumption profiles, such as washing machines and dishwashers, the system does not know which cycle will be selected in advance. This paper proposes an appliance power profiling system that analyses the power consumption data collected by smart meters, identifies which features are most relevant for the specific appliance and, using the k-means algorithm, extracts the set of power consumption profiles that are associated with each appliance. Preliminary simulation results show that each profile can be approximated with a single reference consumption cycle representative for the entire cluster with errors that are always lower than 10%, considering both total energy consumption and power values time interval by time interval.