To exploit local energy resource flexibility and relieve congestion in distribution grids, this paper introduces a two-stage stochastic allocation model for energy storage systems (ESSs) that leverages decomposed distribution locational marginal pricing (DLMP). The planning stage determines the location and size of ESSs, while the operation stage employs a linearized Alternating Current Optimal Power Flow (ACOPF) model, to clear the local power market. A rigorous mathematical framework is provided to derive decomposed DLMP directly from the linearized OPF, enabling computationally efficient and scalable price-clearing processes. The proposed method eliminates the need for extensive price computations after solving the ACOPF, making it applicable to large-scale systems with no approximations in the grid model. The Benders decomposition is applied to solve this investment problem in view of its inherent hierarchical structure. The numerical experiments demonstrate the proposed method's effectiveness, achieving a 16.4% cost reduction compare to businesses-as-usual model over 20-year planning horizon, effectively mitigating congestion costs and reducing price fluctuations by providing lower and more stable DLMP values.