Abstract:
Automatic detection of the event of switching-on an
electric load, estimation of that switch-on instant and subsequent
automatic identification of that electric load using current
transient signal around the switch-on instant are studied in this
paper. The time variation of the current harmonics is used to
first detect the event of switching-on an electric load and then
spectral features extracted from the current harmonic signal
around the estimated switching-on instant is used for automatic
identification of that load. In particular, feature based on the
second derivative of the magnitude of second harmonic of the
current signal is proposed to be used for automatic detection of
the event of switch-on. This feature provided an error rate of
9.4% of event detection with 0.27 seconds of average error in
estimating the switch-on instant. Finally, the overall system that
combines this detection of the event of switch-on and the
automatic identification of that load gave an accuracy of 87.5%
in a cross fold validation experiment tested with four loads.
