Abstract: To enhance the universality of data-driven transient stability assessment models and the completeness of stability assessment information, a hierarchical transient stability assessment model based on representation enhanced graph convolutional networks is proposed. Firstly, based on the adjacency matrix, the graph convolution operation is improved to construct enhanced features, and gate units are used to obtain informative graph feature representations, thereby improving the feature extraction ability of the model. Then, a stability evaluation model structure for system and generator classification is proposed, which enables the model to output multiple stability evaluation results such as system stability, generator rotor angle difference, and stability. Meanwhile, the independent evaluation mode for the generator level is designed to enable the model to perform stability evaluation in scenarios where the power grid unit commitment mode changes, improving the universality of the model. Finally, based on the pre-fault generator rotor angle difference, the numerical range of the rotor angle difference label is compressed, further improving the performance of the model in the rotor angle difference evaluation task. The test results in standard systems and actual large-scale power systems show that the proposed model has high evaluation accuracy and strong engineering application performance.