Abstract: To improve the timeliness of data-based transient voltage stability assessment and to achieve quantitative evaluation, we proposed a transient voltage stability evaluation model based on convolutional block attention module-based convolutional neural network (CBAM-CNN). In the model, the power-flow, fault location, and bus voltage jump information are used as inputs, and hybrid attention mechanism and multi-task learning framework are introduced to output the transient voltage stability level and stability labels of each node in power grid. Studies in the classic IEEE 39-bus system show that the proposed model has sufficient adaptability to changes in power-flow and fault location; compared to several other popular deep-learning models, the recommended model has stronger information representation and generalization capabilities, and is expected to be applied in power system preventive controls.