A Voltage Prediction Method of Distribution Network Based on MMD-MBPResNet-SSA Combined Model

Voltage fluctuations and limit violations in distribution networks may lead to operational disruptions in critical equipment. Therefore, timely and accurate prediction of voltage deviations is essential for improving voltage quality at the end of the user. This paper proposes a hybrid voltage prediction approach for distribution networks based on multi-mode decomposition (MMD), a multi-branch parallel residual network (MBPResNet), and the sparrow search algorithm (SSA). Three modal decomposition techniques are applied to decompose the historical data trajectories of load and voltage, thereby constructing an input matrix that captures the correlations between load sequences and node voltage fluctuations. The prediction model integrates convolutional layers, multi-branch parallel residual blocks, and gated recurrent units (GRUs). Furthermore, the SSA is employed to optimize key model parameters, thereby enhancing computational efficiency. Case studies demonstrate the effectiveness of the proposed method in accurately forecasting node voltages within distribution networks.