New method for non-intrusive data extraction and classification of residential appliances

This paper is focused on the non-intrusive load monitoring(NILM) system, which allows the identification of residential appliances in a non-intrusive way. The previous techniques are not performing perfectly in all aspects of economy, feasibility and accuracy. A new method, via the feature of working style and power characteristic of residential appliances, using human reaction time as the time scale unit (sampling frequency), to complete the classification and identification in a financially viable and easily applicable solution. The attribution of this paper is simplify the identification, which can be integrated into the meters now, and this simplification will improve the application of NILM system in demand side management(DSM). Feature selection, mathematic model and algorithm details are described. Analysis of typical experiments is shown in detail. Classification objects including 12 important power consumption types of the most common residential appliances.