Multi-parameter joint evaluation method of converter transformer change state based on TET-former-EWMA

The ultra-high voltage (UHV) converter transformer, which integrates transformers, rectifiers, and other essential components, is the core equipment in the power grid. Its operational status is directly linked to the reliability and safety of power transmission. To fully utilize the multi-source condition monitoring data of the converter transformer and achieve an accurate assessment of its status, a statistical analysis was conducted on the monitoring parameters. The results indicate a certain degree of correlation between these parameters. To address this, this paper proposes a UHV converter transformer multi-parameter joint evaluation method (TET-former-EWMA). This method leverages a Time-embedding-Transformer model with a variant transformer structure to explore the interdependencies among condition monitoring parameters and construct a multi-parameter sequence data prediction model. A deviation evaluation index is established by comparing Transformer-based predictions with actual values. Additionally, the exponentially weighted moving average (EWMA) method is employed to monitor the fluctuation trends of prediction errors, thereby enabling real-time operational status monitoring of the converter transformer. The results demonstrate that the multi-state parameter Transformer prediction model effectively visualizes multivariate monitoring data for the converter transformer. By building on statistical analyses of online monitoring parameters—confirming their inherent correlations—this approach seeks to eliminate the limitations of subjective thresholds, which often hinder the ability to uncover hidden relationships within the data.