A short-term wind power prediction model based on TCN-BiGRU self-attention

Aiming to address the limitations of existing wind power prediction algorithms with insufficient generalization ability and unstable prediction of low-dimensional data, this paper proposes a combined prediction model integrating Temporal Convolutional Network (TCN), Bidirectional Gated Recurrent Unit (BiGRU), and Self Attention mechanism (Self Attention) for short-term forecasting of wind power. Specifically, to address the limitations of BiGRU in capturing longterm dependencies and extracting features, the TCN module is integrated to enhance the model's ability to extract features such as local patterns and long-term dependencies from the data, while also reducing noise interference. In addition, the Self Attention module is further integrated to enhance the model's global feature modelling and dynamic feature weight assignment capabilities. The proposed model is validated by applying the actual wind power dataset, and the experimental outcomes demonstrate its strong predictive performance, achieving an R2 of 0.9595 and an RMSE of 0.0352. In comparison to other combined prediction models, the proposed model achieves an improvement in R2 by 0.337%-0.485%, and a reduction in RMSE by 1.341%-1.909%, which verifies the superiority of the model. The constructed TCN-BiGRU-Self Attention model has excellent short-term prediction effect and strong nonlinear processing ability, which can be extended to other power prediction problems.