基于最优动态下垂系数的分布式光伏快速有功功率调节

To realize the active response for rapid active power regulation of massive distributed photovoltaics (DPVs), a bi-level architecture integrating clustering and optimization is proposed for the coordinated design of dynamic droop coefficients of DPVs in this paper. At the lower level, the adjustable capacity and response time are selected as clustering features according to the rapid active power regulation requirements, and the U-k-means algorithm is applied to implement the clustering of DPVs. At the upper level, a frequency deviation-optimal droop coefficient model incorporating the regulation performance and network losses is developed in a perspective that considers power flow impacts. The droop coefficient regulation strategy under various frequency fluctuations is pre-calculated by a graph attention network (GAT). Simulations are performed on a modified IEEE 33-bus test system, and the results verify the outperformance of the proposed GAT model over the existing neural network models, as well as the effectiveness and superiority of the proposed optimal droop coefficient regulation strategy.