基于图神经网络的变压器短路电流计算方法OA北大核心CSTPCD

With the increasing complexity of power systems in recent years,transformer operation safety has become a key issue that affects the stable operation of the power system.Currently,calculation methods of short-circuit current through transformers are mostly based on the power grid topology and transformer equivalent impedance.These methods have low flexibility and real-time performance and do not consider the actual operation mode of the system,making it difficult to meet the requirements of real-time operation of power systems.In this paper,a transformer short-circuit current calculation method considering flow conditions is proposed based on a graph convolutional neural network.By introducing the features of transformer bus and regional topology,a transformer short-circuit current calculation model is trained.This method introduces attention mechanisms to make the model more sensitive to dynamic power flow conditions under different operating conditions.Verified by an actual power grid example in a region,the calculated short-circuit current using this method has a small error compared to the reference value,and the distribution of calculation errors is concentrated,which can basically meet the requirements of short-circuit current calculation in practical operation.