Due to the increasing penetration rate of Plug-in Electric Vehicles (PEVs), the infrastructure planning of PEV fast charging stations is under way. Owing to the environmental friendly property of the Renewable Energy Source (RES), Photovoltaic (PV)-integrated fast charging stations are proposed to facilitate the RES utilization and to minimize the station operation cost. However, the intermittent nature of PV adds stochastic property into the station operation process, which increases the possibility of the Quality of Service (QoS) degradation. In this paper, we propose a Markov Chain model to analyze the stochastic process of the PV panels utilization. Then, a PV sizing optimization problem is formulated to determine the optimal number of PV panels in the station by minimizing the station operation cost, while guaranteeing the QoS requirement of the station. Finally, we provide a case study to validate the feasibility of the PV sizing scheme in terms of the station operation cost and QoS.