Electronic Voltage Transformer Metering Error Prediction Model Based on VMD and Correction Model

The deployment of Electronic Voltage Transformers (EVTs) within ultra-high voltage smart grids exceeding 110kV is progressively expanding, rendering their operational stability paramount for precisely acquiring voltage signals across power grids. This paper proposes a dual-stage metering error prediction model for EVTs, which integrates Variational Mode Decomposition (VMD) with a Convolutional Neural Network-Bidirectional Gated Recurrent Unit-Attention Mechanism (CNN-BiGRU-AM). Initially, an optimization algorithm is applied to determine the optimal decomposition parameters in the variational mode. Non-stationary and stochastic raw ratio error sequences are then decomposed into different modal components through VMD. Subsequently, each component is modeled with a predictive framework incorporating BiGRU and an Attention Mechanism to obtain error deviation data based on the first stage's predicted ratio error results. Next, highly correlated features with EVT metering errors are identified using a random forest algorithm to construct a feature set. These features, combined with the error deviation obtained from the first stage, are used as joint inputs to establish an error correction process (second stage), resulting in the final prediction value. Empirical analysis indicates that the proposed method improves prediction accuracy and stability.