An Incremental Non-Intrusive Load Disaggregation Model Considering Dynamic Changes of Appliances

In recent years, substantial increase in electricity demand has led to dynamic loads in the grid. The unpredictability of these incremental loads often hampers traditional load disaggregation techniques. Thus, we propose an incremental non-intrusive disaggregation model. It aims to effectively disaggregate additional appliances. We systematically outline the pros and cons of existing NILM techniques, and then use a sequence-to-point model combined with transfer learning to disaggregate the incremental loads. Validation is conducted on the REDD dataset. Results demonstrate effectiveness of the proposed model in disaggregating incremental loads.