基于图填补神经网络的配电网稀疏量测数据推演方法OA

Incomplete deployment of measurement devices and data transmission losses can result in sparse mea-surements in distribution networks.To address this issue,this paper proposes an inference method for sparse mea-surements based on a graph imputation neural network(GINN).The proposed method aims to improve the accuracy and reduce the sparsity of existing measurements.First,a GINN-based measurement feature encoder module is de-signed to extract power flow features such as power and voltage from nodal measurements.A transformer network is employed to model cross-feature correlations among different power flow features.Second,a GINN-based graph en-coder module explicitly encodes topological connectivity between distribution network nodes.By incorporating a graph convolutional network(GCN),this module enables the propagation and updating of nodal power flow fea-tures.Subsequently,by leveraging two modules to capture the correlations between different power flow features of node measurements and the topological correlations across nodes,the missing data is inferred and completed using the sparse measurements.Finally,simulation tests are conducted on IEEE 14-,30-,57-,and 118-bus systems to validate the effectiveness of the proposed method.