Towards real-time energy generation scheduling in microgrids with performance guarantee

Microgrids are local electricity systems that integrate demand and distributed energy resources. They represent a basic building block for future electricity systems that are more cost-effective than conventional grids. Real-time energy generation scheduling is an important issue in microgrid operations. However, common approaches in conventional grids suffer from various limitations and do not apply to microgrids: (1) stochastic optimization approaches rely on stochastic modeling, which is difficult to obtain in microgrids with significant wind penetration and varying demand responses participation; (2) robust optimization approaches provide no cost-efficiency performance guarantee. In this paper, we present a real-time scheduling solution without stochastic modeling, considering a flexible time-window of prediction and providing performance guarantee with minimal risk. Our solution is an generalization of our previous work in [13] from homogeneous local generation to heterogeneous settings. We evaluate the performance of our proposed solution using the operating trace of a pilot microgrid on the Bornholm Island, Denmark. We show that our solution achieves a performance close to perfect dispatch, and is robust to prediction errors.