Effect of the meta parameters on a model predictive algorithm for renewable micro-grid optimal control

Abstract

Renewable energy micro-grids are often coupled with energy storage systems to balance the production variations. A common way to control such micro-grids is the Model Predictive Control (MPC) strategy that optimizes the grid behavior at a given time-step for a rolling horizon. This computation is refreshed periodically so that the algorithm can work with up-to-date measurements. This paper focuses on how the time-step, the horizon and the refresh period affect the optimal solution. The MPC algorithm is made up of a Linear Programming optimizer and an Auto Regressive Moving Average forecast. It is applied to a micro-grid with a PV array and a Battery storage that has to supply a residential complex and is able to trade energy with the main grid. Several configurations (equipment size, load profile, cost profile) are studied to ensure the results’ robustness. The MPC can reduce the energy exchange cost and can almost reach the optimal solution in most of the cases. The results show that an optimization with a time-step of 30 min and a 12-h horizon gives good performances. However, the refresh should happen at least 4 h before reaching the optimization horizon.