Smart Meter Fault Diagnosis based on Directional Gradient KNN

This paper presents a novel approach to enhance the effectiveness and precision of smart meter fault diagnosis by utilizing the K-nearest neighbor (KNN) algorithm. Initially, the standard fault monitoring data of smart meters, comprising forward active energy data and reverse active energy data, are collected. Then, the diagnostic data is segregated based on the proposed directional function. The gradient of the data change is computed and the sensitive features of the data are extracted. Subsequently, the sensitive feature parameters are employed to construct the forward K-nearest neighbor algorithm model and the reverse K-nearest neighbor algorithm model of the data, respectively, for the purpose of fault diagnosis of smart meters. By devising a directional function and extracting gradient features, a novel fault diagnosis model is established, which enhances the diagnostic efficiency and accuracy of the model.