Abstract: In order to improve the response speed of operation and maintenance personnel and reduce the impact of cloud environment exceptions on cloud applications, this paper studied an anomaly intelligent detection method in cloud environment based on LSTM (Long Short-Term Memory).The paper combined traditional time series analysis algorithm with LSTM neural network to implement on-line prediction of cloud environment monitoring data, and defined the prediction fluctuation range by normal distribution and Bayesian inference to quickly and accurately identify cloud environment anomalies. Tested and verified in the railway cloud data center environment, Compared with other time series prediction methods, this method has the advantages of high discrimination accuracy and strong applicability to various scenarios. It can provide an effective anomaly detection method for the implementation of intelligent operation and maintenance of large-scale railway cloud data center.