The use of photovoltaics (PV) in electric power networks has increased because of advantages such as power loss reduction, environmental friendliness, voltage profile improvement, and postponement of system upgrades. However, using PVs of an inappropriate size leads to greater power losses due to variations in PV outputs and demand loads. Therefore, it is necessary to determine the optimal size of PVs. When one considers the various weather conditions and demand load profiles, the optimal PV sizing problem (OPSP) becomes a large-scale optimization that is difficult to solve. In this paper, to resolve this computational issue, we formulate the OPSP as a second-order cone programming model and propose a decomposition using the subgradient method for solving the partial Lagrangian dual problem. We implemented the proposed method on a 47-bus distribution feeder model, and the simulation results proved its viability.