Revving up energy autonomy: A forecast-driven framework for reducing reverse power flow in microgrids

Abstract

Reverse power flow, defined as the continuous flow of electricity in a direction opposite to the normal direction of the power flow in a grid, typically occurs in microgrids when the energy generated by the distributed electric power plants exceeds the local load demand. This phenomenon imposes several risks related to inefficient operation or damage of equipment, grid instability, and energy losses. In order to reduce reverse power flow in microgrids and support energy autonomy, we introduce a forecast-driven framework. The framework builds upon deep learning models that forecast the electricity produced (photovoltaic systems) and consumed by the microgrid and an optimization algorithm that schedules its shiftable loads (electric vehicles) based on said forecasts. We conduct an ablation study to evaluate the effect that optimized scheduling and energy storage has on the autonomy of the microgrid, also investigating the impact of different capacities of batteries and sizes of electric vehicle fleets. Our results suggest that forecast-driven load shifting can significantly reduce reverse power flow, especially for relatively larger amounts of shiftable loads. Moreover, we find that electricity storage can complement load shifting, further improving its beneficial effect. Nevertheless, these improvements are subject to forecast accuracy and storage abilities.