基于时空特征图卷积网络的配电网实时状态估计方法

Real-time State Estimation Method for Distribution Networks Based on Spatial-temporal Feature Graph Convolution Network

为了在考虑负荷时空特性以及不健全量测的场景下实现配电网的实时、高容错状态估计,本文从时空图的角度提出一种基于时空特征图卷积网络的动态状态估计算法。首先将配电网量测数据组成三维张量;然后利用时空特征图卷积网络分别提取量测数据的空间拓扑、时间序列和节点属性上的特征信息,通过特征融合得到实时状态估计结果;最后根据状态估计结果生成虚拟量测以消除不良数据的影响。实验结果表明,本状态估计方法具有较高的鲁棒性,且准确性和计算速度均优于传统方法。研究结果可为基于机器学习的电力系统状态估计方法提供参考。

We propose a real-time state estimation method for distribution networks based on spatial-temporal feature graph convolution network(STFGCN),to address the issues of load spatial-temporal characteristics and defective measurement.Firstly,a three-dimension measurement tensor was constructed by the measurement data of distribution network.Then,the STFGCN was adopted to extract the features of spatial topology,time series and node attributes from measurement tensor,and real time estimation result was obtained through features fusion.Finally,a virtual measurement was generated to offset the interference of bad data.Numerical results clearly validate the efficiency and robustness of the proposed state estimation method compared with the traditional algorithms,which provides a valuable reference to the estimation algorithms based on machine learning.