50 % seminar: Stability-constrained energy management for converter-dominated microgrids

Microgrids, which integrate renewables, energy storage system and loads, are converter-dominated systems which suffer from converter-driven stability issues. Existing energy management systems (EMSs) optimize operating cost under static power and energy constraints, while ignoring converter-driven stability, and thus cannot guarantee stable operation under wide operating conditions. We propose a stability-constrained energy management framework that embeds the microgrid small-signal stability region into the EMS. First, a deep neural network (DNN)-based stability region characterization is developed based on a module-based small-signal model and eigenvalue sweeping. Second, we propose a stability region-embeded safe deep reinforcement learning (DRL) to ensure stability and safety of microgrids. Simulation and experimental results both demonstrate that the proposed EMS framework can achieve economic energy scheduling while maintaining stable system operation.