Coordinated Voltage Regulation of Distribution Network and Multi-Microgrids Via Attention-Based Transfer Reinforcement Learning

The integration of large-scale distributed photovoltaics (PVs) causes voltage violations in distribution networks, adversely affecting power quality. As components of flexible resource optimization, microgrids (MGs) provide critical support for the secure and economical operation of distribution networks. However, challenges such as the difficulty in real-time identification of model parameters and measurement data make traditional model-driven optimization methods increasingly unsuitable. Therefore, this paper proposes a coordinated voltage regulation method for distribution networks with multiple microgrids via attention-based deep reinforcement learning. Moreover, to enhance optimization performance upon the addition of a new microgrid, a transfer reinforcement learning approach is introduced to enhance training efficiency. Finally, the effectiveness and superiority of the proposed method are validated through analysis on the standard IEEE 33-bus system.