In this paper, we present a methodology to optimally sizing a photovoltaic stand-alone system with battery storage. Its originality resides on taking into account the stochastic nature of the solar radiation in the sizing process using statistical techniques. In fact, the optimal design is based on the inverse conception from only the radiation data and the load profile, the number of photovoltaic panels and the number of batteries were computed.  The radiation data of the chosen location was used to determine its distribution function and then determining the average power generated by the photovoltaic model in every hour of a typical day of every month. Basing on the statistical techniques, the photovoltaic model that most fits the site was chosen among a list of four panels with different characteristics. The loss of power supply probability (LPSP), related to the system reliability was used to evaluate the efficiency of the sizing methodology.