Error prediction for capacitor voltage transformer based on graph convolution and temporal convolution

Accurately predicting the error of capacitor voltage transformer (CVT) is crucial for the stable operation of power system. In order to accurately forecast the error of CVT during long-term operation, this paper proposes a neural network model combining graph convolution network (GCN) and temporal convolution network (TCN) called GC-TCN to predict the measurement error of CVT. The model uses graph convolution to capture the correlation between voltage signal and environmental parameters, and uses temporal convolution to analyze the time series characteristics of voltage signal. Through the complementarity of the two convolution structures, it can process information in spatial and temporal dimensions at the same time, which improves the accuracy of error prediction. Experimental results show that the error prediction model combining graph convolution and temporal convolution has a determination coefficient R2 of ${9 4. 3 6 \%}$, which is 7.95% higher than that of the single TCN model. It has high accuracy when applied to CVT error prediction tasks.