Aiming at the core concept of green development, a short-term load prediction method based on CEEMDAN-Attention-BiGRU is proposed. Firstly, the modal decomposition of the original load sequence is carried out by using Complete EEMD with Adaptive Noise (CEEMDAN), and the modal components of each intrinsic mode function with different frequencies are predicted by Bidirectional Gate Recurrent Unit (BiGRU) based on attention mechanism with different hyperparameters. Finally, the final load prediction results are obtained by combining the prediction values through the full connection layer. The simulation results show that the method proposed in this paper can effectively improve the accuracy of short-term load prediction and has practical application value.