Online Metrological Assessment of Voltage Transformers via Multidimensional Feature Matrix

To address the intrinsic and temporal correlations in voltage transformer outputs, this study proposes a metrological assessment framework utilizing multidimensional feature matrices. By exploiting group-wise correlation features among transformers, multidimensional feature matrices is constructed to characterize metrological deviations. The T-values derived from the matrix are standardized via a proportional scaling method and subsequently fed into an LSTM network, enabling precise anomaly detection. Empirical validation demonstrates an accuracy rate as high as 94.83%.