This paper presents an algorithm for the optimaloperable dispatch of distributed battery banks in systems with high integration of variable renewable energies. As a test case, the application of the algorithm is presented in a possible expansion of the Colombian system 2024-2039, subdivided into five regions each with its DBESS. The novelty lies in how to integrate a set of technologies such as stochastic dynamic programming, with reinforcement learning and optimization of the marginal benefit of an agent to obtain an operable operation policy close to the achievable optimum.