Data-Driven Probabilistic Modeling of EV Harmonic Emissions Based on Field Measurements

The accelerating transition toward sustainable transportation is expected to significantly increase the penetration of electric vehicles (EVs) into distribution networks. EVs, as nonlinear loads with a AC-to-DC convertors, introduce harmonic distortion that may exceed regulatory limits or impact power system components. Accurate prediction of resulting voltage harmonics requires probabilistic models capturing the nonlinearity and inherent variability of EV harmonic emissions. In this work, an extensive field measurement campaign was conducted to record harmonic current phasors for odd harmonic orders up to the 50th, across more than 54 charging cycles from 14 EVs representing nine different models. The measurements show significant differences in harmonic injection profiles across EVs, attributed to variations in design, power electronic, and charging technologies. A Multivariate Gaussian Mixture Model (GMM) is applied to characterize EV harmonic current injections and to generate synthetic samples consistent with observed distributions. The proposed model accurately captures the underlying patterns and correlations between harmonic magnitudes and angles. Model validation has been performed by comparing the probability distributions obtained from measured and synthetically generated harmonic currents, as well as the harmonic voltage levels resulting in a distribution feeder. The field measurement data has been made publicly available to support further research.