Electronic devices have a non-linear characteristic and are sources of harmonic emission. Their massive use pollutes the network and consequently it is needed to measure and characterize those devices. Harmonic current emitted by electronic devices is closely linked to their circuit topology and the distortion of the supply voltage. Different circuit topologies have also different current waveforms. This paper proposes an automatic classification method of steady-state appliances based on current waveform characterization in the higher-
order statistics space. The translation from the time domain to a statistical space enables the automatic identification of individual devices. The algorithm has
been applied to the current waveforms of a large set of household appliances measured under sinusoidal conditions. The classification analysis proves that clusters of circuit topologies can be clearly identified. In addition,
authors show that kurtosis and variance of an individual cycle provide enough information about the distribution of a waveform shape with respect to its average value, while the skewness inform about the half cycle bias. The
method can be a useful tool to identify prevailing circuit topologies in the market. It can also improve automatic load identification, e.g. part of the future intelligent measurement systems such as smart meters.