A Two-Layer Stochastic Model Predictive Control Scheme for Microgrids

A two-layer control scheme based on model predictive control (MPC) operating at two different timescales is proposed for the energy management of a grid-connected microgrid (MG), including a battery, a microturbine, a photovoltaic (PV) system, a partially non predictable load, and the input from the electrical network. The high-level optimizer runs at a slow timescale, relies on a simplified model of the system, and is in charge of computing the nominal operating conditions for each MG component over a long time horizon, typically one day, with sampling period of 15 min, so as to optimize an economic performance index on the basis of available predictions for the PV generation and load request. The low-level controller runs at higher frequency, typically 1 min, relies on a stochastic MPC (sMPC) algorithm, and adjusts the MG operation to minimize the difference, over each interval of 15 min, between the planned energy exchange and the real one, so avoiding penalties. The sMPC method is implemented according to a shrinking horizon strategy and ensures probabilistic constraints satisfaction. Detailed models and simulations of the overall control system are presented