基于图注意力网络的配电网分布式光伏实时消纳能力评估方法

Real-time distributed PV consumption capacity assessment method for distribution network based on graph attention network

大规模分布式光伏接入加剧了配电网运行的复杂度与不确定性,实时准确评估配电网分布式光伏消纳能力成为系统安全运行的迫切需求。针对传统评估方法依赖大量仿真计算难以满足实时性的问题,同时应对传统人工智能方法难以适应配电网拓扑频繁变化的情况,基于图注意力网络模型提出一种配电网分布式光伏实时消纳能力评估方法。该方法结合网络拓扑结构,充分挖掘有关光伏消纳的节点特征,提高模型在拓扑结构频繁变化时的适应性。通过节点光伏消纳能力回归模型和节点光伏消纳制约条件分类模型,实时评估配电网各节点光伏消纳能力,并确定各节点光伏消纳的制约条件。仿真算例表明,所提方法在面对已知拓扑时具有较好准确性与实时性,面对未知拓扑时拥有较好的泛化能力。

Large-scale distributed photovoltaic(PV) access aggravates the complexity and uncertainty of the distribution network operation.Real-time and accurate evaluation of the distributed PV consumption capacity of the distribution network has become an urgent need for safe operation of the system.Aiming at the problem that traditional evaluation methods rely on a large number of simulation calculation and cannot meet the realtime requirement.At the same time,it is difficult for the traditional artificial intelligence methods to adapt to the frequent changes of distribution network topology.A real-time PV consumption capacity evaluation method is proposed based on the graph attention network model.Combined with the network topology,this method fully exploits the node features related to the PV consumption,and improves the adaptability of the model when the topology changes frequently.The regression model of node PV consumption capacity and the classification model of node PV consumption constraints are used to assess the realtime PV consumption capacity of each node in the distribution network and determine the constraints of PV consumption at each node.The simulation example shows that the proposed method has good accuracy and real-time performance in the face of known network topology,and has good generalization ability in the face of unknown topology.