With the continuous increase in wind power penetration, the power fluctuation and uncertainty of power systems have significantly intensified, posing severe challenges to traditional transient stability assessment methods in terms of both accuracy and efficiency. To address this issue, a transient stability assessment model based on an improved one-dimensional convolutional neural network with attention (CNN-Attention, CNNA) is proposed. First, by fully exploiting the feature extraction capability and key temporal sequence focusing ability of the CNNA, residual blocks are introduced to alleviate the gradient vanishing problem, and Wiener filtering is incorporated to suppress noise interference in the samples. Second, a transient stability assessment framework is designed based on a combined criterion of rotor angle stability and voltage stability, and a corresponding evaluation index system is established. Finally, simulation studies are carried out using PSASP software on the IEEE 39-bus and IEEE 118-bus systems with wind turbine generator. The results show that the proposed method maintains high accuracy under different noise levels and wind turbine operating conditions, effectively reduces the risk of “false stable” misjudgment, and exhibits strong capability in identifying unstable states.