This paper investigates the sizing of the Frequency Restoration Reserve (FRR) in a context of increasing penetration of renewable generation. More precisely, we propose (i) a probabilistic method that mimics the current Belgian TSO (Elia) practices and (ii) a Monte-Carlo based procedure that evaluates the corresponding reliability of the system in terms of down/upward reserves activation, wind curtailment and load shedding. Using this method over the IEEE-RTS96 testcase, the impact of wind penetration-low, moderate, high-is studied. In particular, static (annual and seasonal) and dynamic (weekly and hourly) FRR sizing approaches are defined and compared. It turns out that the hourly sizing method is the most robust. It also appears that FRR requirements for upward reserves are almost not impacted by the high wind penetration whereas the downward reserves increase significantly with the wind penetration. Our implementations rely on Julia, Cplex and R and are available in open source.