Energy management of a grid-connected photovoltaic plant coupled with a battery energy storage device using a robust approach

This paper addresses the energy management with a robust approach of a grid-connected photovoltaic plant coupled with a battery energy storage device, within the capacity firming specifications of the French Energy Regulatory Commission. A tailored multi-output deep-learning model is used to generate quantile forecasts of PV generation. It leads to an integrated forecast-driven strategy modeled by a min-max-min robust optimization problem with recourse, solved by using a Benders decomposition. The case study is the PV generation monitored on site at the university of Liège (ULiège), Belgium. This approach achieves better performance than its deterministic counterpart, and the subsequent risk-aware robust planning tool allows finding a trade-off between conservative and risk-seeking policies