Residential Harmonic Injection Models Based on Field Measurements

In recent years, harmonic current levels have increased
in residential distribution networks due to the growing
penetration of nonlinear loads at customer premises. To properly
model the harmonic current injection of residences, probabilistic
models that account for the stochastic behavior of domestic loads
are needed. As a part of the study reported in this paper, a field
measurement campaign was performed in 24 dwellings during a
one-week period with a one-minute resolution. Harmonic current
magnitudes and phase angles of odd harmonic orders up to
the 25th as well as the active and reactive power demanded
were recorded. By applying two different unsupervised learning
techniques (i.e., nonparametric density estimation and Gaussian
mixture models) to these measurements, two probabilistic models
of harmonic injection of individual residential sites have been derived.
Statistical probability distributions have been determined
for the magnitude and phase of each harmonic order segmented
in different power demand intervals. The developed models are
validated on a test feeder by comparing harmonic voltages caused
by the injection of the synthetically generated currents with those
obtained from measurements. Both field measurement results
and the detailed data defining the probabilistic harmonic current
models are made public in an Open Science database repository.