A Generic Energy Disaggregation Approach: What and When Electrical Appliances are Used

With the development of smart meters and data collection techniques, we now can get energy consumption information in real time. Research shows that providing appliance-level energy information can better promote energy conservation than providing total energy information. Previous work on detecting appliance-level energy information either relies on a range of smart meters to detect each appliance's energy information, or specific knowledge about appliances including types, labeling or models that are hard to obtain for domestic household. In order to alleviate this problem, we propose a generic disaggregation approach to calculate the number of appliances, their power and usage information. This approach is mainly based on three common features that most appliances have, including the feature that the ascending or descending edges' power values of each appliance follow the Gaussian distribution. Particularly, we use a Gaussian distribution to model each appliance, and use the Expectation -- Maximization clustering algorithm to identify how many appliances are used, what their generic models are and when they are used. We verified the accuracy of this approach with a real energy data set and find that this approach can successfully disaggregate the total energy data.