Short-term Prediction of Distribution Voltage Based on CNN-BiGRU

To further improve the accuracy of voltage prediction for distribution transformers, this paper proposes a combined prediction model based on Convolutional Neural Network (CNN) and Bidirectional Gated Recurrent Unit (BiGRU). The CNN is characterized by local connections and weight sharing, which optimizes the input voltage sequences for the BiGRU model. The BiGRU, with its bidirectional information extraction capability, enhances the predictive power. The proposed method is validated using real measurement data from a distribution transformer in Shanghai in the year 2022. A comparative analysis with GRU and BiGRU models demonstrates the effectiveness of the proposed approach, leading to improved accuracy in short-term voltage prediction for distribution transformers.