Power equipment system anomaly detection method based on spatial-temporal feature segmentation and reconstruction

The invention discloses a power equipment system anomaly detection method based on spatial-temporal feature segmentation and reconstruction, and the method comprises the steps: dividing multi-source time sequence data generated by a plurality of sensors for monitoring the state of a system according to a specific length, and then forming a multi-source data set; pre-processing the multi-source data sequence and then dividing the multi-source data sequence into a training set and a test set; obtaining a priori graph structure by utilizing priori knowledge of a mechanical equipment structure; establishing an adaptive graph structure by using a graph structure parameterization learning mode, constructing a spatio-temporal feature extraction network based on graph convolution, and extracting spatio-temporal features of multi-source data in combination with the two graph structures; constructing a segmentation and reconstruction anomaly detection network, and optimizing the constructed model by using the training