Abstract: Power system transient stability assessment（TSA）mostly uses the data set generated by standard examples at present.However,the number of power components such as buses,generators and lines in actual power grid is huge,which makes it difficult to realize real-time monitoring and online updating of the assessment model. Existing dimension reduction methods often omit important information,resulting in the decline of prediction accuracy. An online evaluation method of power system transient stability driven is proposed by image data,which rearranges the input time series into two-dimensional images,reduces the dimension of original images by using two-dimensional principal component analysis,and establishes a convolutional neural network model to predict the system stability. The verification results in IEEE-39 numerical example show that the evaluation model proposed in this paper based on two-dimensional principal component analysis and convolutional neural network can greatly improve the training efficiency while ensuring the prediction accuracy,and is expected to promote the application of in-depth learning in online evaluation of power system transient stability.