Short-Term Electric Load Forecasting Based on CNN-GRU-Attention

Electric load forecasting is vulnerable to various complex factors, such as extreme weather caused by climate change impacting electricity demand, industrial electricity structure adjustments driven by economic development, and diverse electricity consumption patterns due to regional differences. These factors make electric loads unstable and nonlinear, significantly increasing the difficulty of improving prediction accuracy. To address this issue, this paper proposes a CNN-GRU-Attention forecasting model. The model uses CNN to mine spatial features of electric load data, GRU to capture long-term dependencies in time series, and an Attention mechanism to accurately allocate feature weights. Analysis of actual electricity load data from a specific region shows that the CNN-GRU-Attention model achieves an RMSE of 25.803, a goodness-of-fit (R<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup>) of 0.979, and an MAE of 19.402. These metrics demonstrate significantly improved prediction accuracy compared to the GRU-Attention model (RMSE = 26.533, R2 = 0.977, MAE = 19.194). This study lays a foundation for the model’s application in complex electricity consumption scenarios and marine electric load forecasting.