Photovoltaic (PV) cell powered base stations (BSS) have been widely considered for reducing the cellular network's environmental footprint in the future. An inherent challenge is to match the energy generation profile with the energy consumption profile. In this paper, we develop a Markov chain based algorithm to determine the optimal orientation angle of the PV cell for matching the two profiles, given the energy generation profile of the geo-location, the load profile of the BS and the battery capacity. We investigate the effects of different battery capacities on the optimized PV cell orientation angle for BSS located in a business district in London in summer. Our results show that the optimal orientation angle for a small battery capacity (10000 Joules at a small BS) is in the range from 35° to 60°, whereas a wider range from-50° to 60° could be chosen for a large battery capacity (50000 Joules at a small BS). This reveals that PV cell orientation angle optimization is more important for PV cell powered BSS with small battery capacities than for large battery capacities.