The paper presents a method for the co-optimization of energy storage systems allocation and line reinforcement in active distribution networks. The objective is to guarantee the capability of an active distribution network to follow a dispatch plan by appropriately coping with the high uncertainties of loads and stochastic renewable generation while ensuring the secure operation of the grid and minimizing the power grid losses. The proposed formulation relies on a modified formulation of the so-called Augmented Relaxed Optimal Power Flow (AR-OPF) method, which considers the exact (i.e., non-approximated) AC-OPF in a convexified way (the AR-OPF is proven to provide the global optimal and exact solution of the OPF problem for radial power grids). To tackle the complexity and computational burden of the proposed planning problem, the Benders decomposition algorithm is used and, in order to enhance the convergence speed of the numerical solution of the proposed problem, the Benders decomposition has been suitably modified to determine the energy storage systems site and size sequentially. To assess the performance of the proposed method, simulations are conducted on a real Swiss distribution network composed of 55 nodes and hosting a large amount of stochastic photovoltaic generation. The sensitivity analysis with respect to the photovoltaic capacity is also carried out to assess the effectiveness of the proposed method.