A short-term wind power prediction based on MCOOT optimized deep learning networks and attention-weighted environmental factors for error correction

Accurate wind power prediction plays an increasingly important role in the safe and economic operation of wind power systems. A two-stage prediction model that combines the modified coot algorithm (MCOOT), deep learning networks, and attention-weighted environmental factors is proposed for short-term wind power prediction. In the first stage, the original sequence is decomposed into subsequences using variational mode decomposition (VMD) with approximate entropy (ApEn) distinguishing high and low-frequency components. Bidirectional gated recurrent units (BiGRU) and Elman neural networks (ENN) are utilized for subsequence prediction. In the second stage, environmental factors are weighted using the attention-weighted method, and the error sequence is decomposed using VMD. BiGRU predicts the error subsequences combined with environmental factors, leading to the second-stage error prediction. Using the MCOOT algorithm, the final prediction result is obtained by weighting and integrating the outputs of both stages. This model effectively captures wind power trends and fluctuations, significantly improving prediction accuracy and effectiveness. Two sets of experiments, conducted with wind power datasets from Xinjiang and Spain, demonstrate that our model outperforms other comparative models in accuracy and stability, enhancing the safety and economic efficiency of wind power systems.