In order to further improve the performance and generalization ability of the power system transient stability intelligent assessment model

this paper carries out research on transient stability intelligent assessment from three dimensions: data enhancement

machine learning model and learning mechanism. Firstly

in order to reflect the global spatial topological characteristics of the power grid and the correlation between variables

a data processing method based on the mirrored encoding of power flow topology is proposed

in which the electrical variables are mapped into a multidimensional matrix feature map containing the power system topology through the coding form. Secondly

a power system transient stability assessment method based on the Swin Transformer model is proposed

in which the global and local transient stability feature information can be effectively extracted through its hierarchical model structure and the self-attention mechanism under the window to enhance the performance of the intelligent assessment model. Finally

a power system transient stability adaptive assessment method based on the dynamic adversarial adaptation (DAA) is proposed

which utilizes an adversarial approach to dynamically adjust the relative weights of the marginal and conditional distributions of the inter-domain data to enhance the migration performance of the Swin Transformer model. Based on the IEEE 39-node standard system

the proposed method is validated by the arithmetic cases

and the results show that the proposed method can effectively improve the accuracy as well as the generalization ability of the intelligent assessment model of transient stability

which provides a new idea for the on-line stability assessment and analysis of the power system.