Stochastic model predictive control of community energy storage under high renewable penetration

This paper focuses on the robust optimal on-line scheduling of a grid-connected energy community, where users are equipped with non-controllable (NCLs) and controllable loads (CLs) and share renewable energy sources (RESs) and a community energy storage system (CESS). Leveraging on the pricing signals gathered from the power grid and the predicted values for local production and demand, the energy activities inside the community are decided by a community energy manager. Differently from literature contributions commonly focused on deterministic optimal control schemes, to cope with the uncertainty that affects the forecast of the inflexible demand profile and the renewable production curve, we propose a Stochastic Model Predictive Control (MPC) approach aimed at minimizing the community energy costs. The effectiveness of the method is validated through numerical experiments on the marina of Ballen, Samso (Denmark). The comparison with a standard deterministic optimal control approach shows that the proposed stochastic MPC achieves higher performance in terms of minimized energy cost and maximized self-consumption of on-site production.