Research on adaptive state prediction method for the metering error of capacitor voltage transformer.

This article focuses on developing an adaptive state prediction method for the metering error of capacitor voltage transformers (CVTs), which are widely used in high-voltage power systems but suffer from complex structure and unstable metering accuracy. The proposed approach uses principal component analysis (PCA) to map three-phase CVT measurement data into a combined residual and score (CRS) statistic, enabling real-time self-evaluation of metering error without relying on a standard transformer. An adaptive batch processing strategy adjusts the prediction time based on system operation changes, and an autoregressive moving average (ARMA) time series model predicts the CRS statistics, achieving prediction errors below 15%. Experimental results under laboratory conditions demonstrate the method's effectiveness in accurately detecting and forecasting CVT metering error states, supporting more timely and practical maintenance decisions in power grid operations.