Abstract: The stability assessment of transient voltage is the emphasis in the stability assessment of the power system,and it is also difficult. This paper proposes a method based on deep learning to consider transient voltage stability evaluation considering multiple input feature sets. First,a multi-input fault set including pre-fault,fault occurrence time and fault resection time is established. Second,the convolutional neural network based on deep learning is used to train PMU data offline,so as to achieve fast and accurate evaluation of transient voltage stability. The simulation results show that the proposed evaluation method,compared with the existing neural network and least squares support vector machine method,has improved both accuracy and evaluation speed.