A scenario-based methodology is presented to investigate the probabilistic power flow of distribution systems with stochastic distributed PV outputs and Electric Vehicle (EV) charging loads. First, Monte Carlo simulation (MCS) technology is used to simulate daily EV charging load curves in an EV charging station considering random traffic behaviors and charging intentions of EV drivers. Secondly, EV charging load curves and historical PV output curves are respectively clustered by the K-means clustering method and subsequently the probabilistic scenarios of the EV charging load profile and the PV output profile are given. Then power flow calculation probabilistic scenarios are built based on the probabilistic scenarios of the EV charging load profile and the PV output profile. The forward-backward sweep method is adopted to investigate the power flow in each calculation scenario. The probabilistic power flow results can be obtained by gathering power flow results in each power flow calculation scenario according to scenario probabilities. Simulation results based on IEEE 33-bus distribution systems justify the proposed model and methodology. This work is supported by Science and Technology Project of State Grid Corporation of China (No. SGJSSZ00KJJS 1903295).