Electric power distribution systems have undergone meaningful changes that motivated the dissemination of specialized analysis tools to improve the studies focused on planning and operation tasks. However, these tools are not effective if the main network elements, such as the loads, are not accurately modeled. Previous works have focused on modeling MV and residential LV load profiles with increased resolutions, but the authors identified a lack of research focused on LV commercial customers. Therefore, this work proposes a probabilistic approach to model load behavior of LV commercial customers with a one-minute resolution, based on data of electrical appliance ownership, probability of use, and energy consumption. The variety of data and the need for providing high-quality input data are among the main challenges faced to apply the proposed methodology. However, once the value of these data is identified, the utilities and regulatory agencies can work to overcome this issue. The created load profiles are compared with real measurements and used in time-series power flows for the evaluation of their applicability.