Research on the Method of Online Abnormal Patterns Recognition in the Control Chart of Smart Meter Based on Ml-Svm

In the paper, the author proposes a classification algorithm of control chart abnormal pattern recognition based on migration learning (ML) and support vector machine (SVM) for the problem of slow identification of abnormal patterns in control charts caused by the use of offline data analysis methods due to differences in the distribution of batch data in the manufacture of energy meters. The design idea is to train the network using only small-scale data, and the mathematical abstraction of the model by the discriminant pattern of ISO7870-2:2013 Control Charts-Part 2: Shewhart control charts , and then train the constructed training sample set using SVM classifier, and obtain the strong generalizability SVM model using migration learning after the model effect is verified, and finally by simulating the classification The classification accuracy of the proposed algorithm was determined to be 95.3%, which can achieve fast online detection and has good application prospects