Voltage sag is one of the most common power quality problems in distribution network. Accurately identifying the causes of voltage sag is of great significance to formulate comprehensive prevention and control scheme of voltage sag and division of power grid user responsibility. A method for identifying the cause of voltage sag based on bidirectional long-term and short-term memory（BiLSTM）neural network is proposed. Firstly

the time-domain characteristics of voltage sag and S-transform energy entropy are extracted

and the comprehensive characteristic index of voltage sag cause identification is constructed.Secondly

a BiLSTM network suitable for voltage sag classification is established to identify the causes of voltage sag.Finally

the evaluation index of multi classification problem is set

and the effectiveness of the proposed identification method is verified by simulation. Simulation results show that the proposed method can identify five voltage sag disturbance sources with high accuracy.