Abstract Electric vehicles (EVs) will help to decarbonize energy systems. However, their connection to on-board level 2 chargers (7.2 kW) at household facilities brings challenges to Distribution Network Operators (DNOs) as they can affect the power quality of low voltage (LV) networks. In order to truly assess these effects, the electrical behavior of the on-board charger in terms of its non-linear content, power demand, and charge rate must be understood first. Nonetheless, most modeling methodologies with this aim result in circuital approaches, and thus, in heavy computational burdens, or assume simplified representations that do not correspond to the reality of the charge. To overcome this, we present a new methodology to model the power quality characteristics of EVs based on measured data from the harmonic spectra of the charger. The model provides a precise and efficient electrical characterization, where probabilistic models of the harmonic spectra are used to compute the power demand during every stage of the charge. Due to its probabilistic nature, these harmonic spectra are represented using Gaussian Mixture Models. We validate the model contrasting simulated data versus real measured one. Then, we illustrate a case study of the model in a LV network power quality assessment with different EV penetration levels, considering time-series harmonic power flows with 10-min resolution under a Monte Carlo approach. Obtained results revealed an increase in the network chargeability and voltage unbalance, along with an increased content of the third harmonic, which appears to be the most intense.