Abstract: In order to quickly and accurately evaluate the transient voltage stability after the fault of the receiving-end power system and locate the voltage instability nodes/regions, a transient voltage stability evaluation method of the receiving-end power system based on the graph convolution network (GCN) and the bidirectional long/short-term memory network (BiLSTM) is proposed. Firstly, based on the time series response characteristics and spatial distribution law of the transient voltage, the topological connection relationships of the power systems and the electrical measurement data of each of the nodes are considered, and the input characteristic matrix representing the operating state of power system is constructed to effectively adhere to the temporal and spatial evolution law of transient voltage. Then, a deep neural network combining the GCN and the BiLSTM is developed to extract the spatio-temporal features of transient voltage with the greatest correlation. Then the mapping relationship between the spatio-temporal features and the transient voltage stability states is established to realize the precise positioning of the transient voltage instability nodes/regions. Finally, the proposed method is analyzed and verified by the modified IEEE-39 node test system and an actual power grid system. The results validate the accuracy and effectiveness of the proposed transient voltage stability assessment method.