Transient Stability Analysis of AC-DC Hybrid Power Grid under Topology Changes Based on Deep Learning

Abstract

Methods based on physical models are difficult to adapt to the current complex power grids, while methods based on traditional deep learning models have insufficient generalization ability to topologically changing scenarios. The development of graph deep learning provides a new idea for transient stability analysis and control under topology changes. Based on the graph convolution aggregation (GraphSAGE) network, this paper proposes a transient stability assessment method for AC-DC hybrid power grids. According to the principle of the graph neural network, the input features are selected and the graph data processing method is designed, and multiple evaluation indicators are established. Based on GraphSAGE network, a model that can effectively learn the topology information of the power system is constructed. Simultaneous evaluation of power angle stability and voltage stability by the model. Example analysis shows that the proposed method has better performance in the face of running scene datasets with frequent topology changes, and has a stronger generalization ability to new unlearned topologies.