Multi-dimensional time sequence anomaly detection method and system based on cloud edge collaboration

The invention relates to the field of security collaborative protection of an intelligent network connection industrial control system, in particular to a multi-dimensional time sequence anomaly detection method and system based on cloud edge collaboration. The system comprises an industrial control data acquisition protection layer for acquiring data of a sensor and an actuator; model deployment is carried out, abnormal detection of ICS real-time data is carried out, and a result and a strategy are issued to an edge calculation layer of the industrial control system; training and updating the multi-dimensional time sequence anomaly detection model, and issuing the model to a cloud computing layer at an edge end; a spectrum residual error data processing method for removing time sequence signal regularity; extracting s-net, u-net and t-net networks of state features, control features and time features; the p-net and d-net networks are used for predicting and decoding the t moment; the Mahalanobis distance cal