基于深层级联残差图卷积的暂态稳定评估模型及其实际电网应用

Application of Transient Stability Assessment Model Based on Deep Cascading Residual Graph Convolution Network in Real World Power Grids

目前,数据驱动的暂态功角稳定评估模型的研究和测试主要在小规模算例系统进行,在实际电网的应用检验不足,其内在原因是图深度学习中的局部信息提取特点与功角稳定全局性的矛盾尚未解决。为此,设计一种深层级联残差图卷积模型,利用含残差连接的深层级联结构,实现模型层数堆叠性能的有效提升。利用MinMaxPooling模块,使模型参数与系统规模解耦。该模型结构设计与节点数量无关,可以解决数据驱动模型应用于大规模实际电网的问题。在某5419个节点的实际区域电网进行测试,结果验证了所提模型的有效性和实用性。

The studies and tests of the data-driven transient power angle stability assessment models are mainly carried out in small-scale example systems,with insufficient application tests in real power grids.The internal reason is that the contradiction between the local information extraction characteristics and the global power angle stability in graph deep learning has not been solved.In this paper,a deep cascade residual map convolution model is designed to achieve an effective improvement in model layer stacking performance through a deep cascade structure containing residual connectivity.It also proposes the MinMaxPooling module to decouple the model parameters from the system size.The model structure design independent of the number of nodes can solve the problem of data-driven models applied to large-scale real power grids.The validity and performance of the proposed model are tested on a real regional power grid with 5419 nodes.