Droop control strategy for microgrid inverters: A deep reinforcement learning enhanced approach

Abstract

To better tap into the potential of distributed renewable energy generation, microgrid system has become an emerging technology. As the bridge of microgrids, the inverters can flexibly convert distributed DC power input into AC power output. It is verified that the traditional droop control strategy for microgrid inverters has inherent defects of uneven reactive power distribution. To this end, this paper proposes a droop control strategy as a multi-objective optimization problem while considering the deviations of bus voltage and reactive power distributions of microgrids. Then, the optimization problem is further formulated as a Markov decision process and solved by a deep reinforcement learning (DRL) algorithm called deep deterministic policy gradient to obtain a dynamic optimal droop coefficient control strategy. Simulation results demonstrated that our DRL-based strategy eliminates the uneven reactive power distribution without voltage drop.