Abstract: A prediction model of pipeline corrosion rate was established through training the monitoring data (90%) from corrosion probe using recurrent neural network (RNN), and the validity of the model was verified by the remaining 10% monitoring data. The results showed that the model based on RNN could predict the corrosion rates of pipelines accurately. The mean square error between the predicted values and the monitoring data was 0.008%. The method can provide early warning information for pipeline corrosion monitoring.