Principal components null space analysis based non-intrusive load monitoring

Abstract

Non-intrusive load monitoring (NILM) is a commonly known problem where power consumption profile of electrical loads including home appliances connected to the power grid needs to be generated without adding any extra hardware to the appliances. Various approaches have been proposed to indirectly detect and identify electrical home appliences by utilizing the information from their temporal behavior (current and voltage transients), energy and power consumption levels, shape of current, voltage and power consumption, noise level characteristics and electro-magnetic interference (EMI) signatures. In this paper, a principal components null space analysis (PCNSA) based classification method is proposed to identify electrical loads from their time-frequency analysis data or EMI signatures. EMI signatures of a number of home appliances have been measured and collected to determine the performance of the method with actual appliance data. It is shown that the proposed method offers promising results, therefore, further research is motivated.