Development of distinct load signatures for higher efficiency of NILM algorithms

A simple and novel methodology for the configuration of robust and distinct load signatures is presented in this paper. A technique based on a spectral analysis of the current waveform of each load is proposed, leading to load signatures consisting of three distinct spectral coefficients. Data required for each load signature come from only one period of the current waveform, under a sampling frequency at the order of kHz. A simple identification algorithm has also been developed in order to estimate the signatures adequacy. Each appliance has been measured and classified as an individual load, using a simple identification procedure. Data obtained from two different measurement sets show that the proposed technique provides unique data formations for each appliance tested by developing an efficient recognition procedure with high levels of identification rate. Due to the simplicity of both the database configuration and the identification procedure, the proposed methodology could be considered appropriate for online monitoring and real time smart grid applications.