Cellular access networks need to reduce their dependence on the grid, with the twofold objective to decrease operational cost and guarantee self-sustainability in case of grid unreliability. For doing so, an interesting possibility is to use renewable energy generators. The power supply system considered here consists of small units that power individual Base Stations (BSs) and are composed of solar renewable energy (RE) generators combined with energy storage units to sustain the BS operation during night or low production periods. The simulation and design of the power supply system require the availability of simple, yet accurate, models of the RE production. In this paper, we propose a stochastic model of the daily RE production that is represented by a single random variable, characterized by the mean value and the variance, that multiplies a shape function describing the production in the timeframe of a day. Our results show that the model, despite being simple, is accurate in predicting the system performance. Moreover, our simulations reveal that, while the mean value of production is fundamental to discriminate an insufficient-production from a sufficient-production regime, the variance of the production has an important impact on the system performance. Higher values of the variance imply a reduced BS self-sufficiency; in these cases, larger PV panels are required to guarantee a sufficient RE production regime.