A Hierarchical Classification Method for Type II Appliance Recognition in NILM

Abstract

Complex load patterns of multi-state appliances, also known as type II appliances, make trouble for appliance recognition by non-intrusive load monitoring (NILM). We solve this problem by decreasing the impact of intra-class variety (IACV) and inter-class similarity (IECS) caused by type II appliances. In this paper, an agglomerative hierarchical clustering (AHC) method is used to overcome the IACV while a comprehensive feature set is used to overcome the IECS. Considering the impact of IACV and IECS together, a hierarchical classifier is introduced to improve performance on appliance identification. The experimental results on public NILM datasets validate the effectiveness of the proposed method.