Location representation of single-position fault for power system transient stability intelligent assessment

Fast and accurate transient stability assessment (TSA) is the basis of intelligent security and risk control of large-scale renewable AC/DC power systems. The TSA model based on machine learning needs to adapt to different generation-load patterns and fault locations. In existing power system transient stability intelligent assessment methods, the generation-load pattern was widely adopted as input features. However, there was still a lack of accurate methods to represent fault locations quantitatively. To accurately and quantitatively represent fault location, this paper proposed a novel concept of the electrical coordinate system (ECS) based on the electrical distance. Firstly, ECS is built based on the electrical distance. Then, ECS is optimized by comparing different combinations of reference nodes. Finally, with fault locations coordinates as input features, a TSA model for fault locations is built based on the improved convolutional neural network. The proposed ECS and intelligent TSA model are verified with the New England 39-bus system.