Appliance Recognition Using V-I Trajectories based on Deep Learning

One aim of the non-intrusive load monitoring is disaggregate the total power consumption to the power consumption of a single device by analyzing the change in voltage and current measured in order to realize recognition of appliance loads. The appliance identification is the core of the non-intrusive load monitoring (NILM). In this paper, a methodology for characterizing appliances and identifying appliances in a 2-dimensional V-I trajectory is proposed for actual measured appliances data. And a method is proposed to filter the sampled data using Empirical Mode Decomposition (EMD). A deep learning method is applied to automatically extract features from the built V-I trajectory maps. After experiments, the accuracy of load identification is relatively high.