Enhancement of wind speed forecasting using optimized decomposition technique, entropy-based reconstruction, and evolutionary PatchTST

Accurate wind speed forecasting is critical for ensuring the reliable operation of wind energy systems. This study proposes a hybrid model integrating Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (CEEMDAN), Sample Entropy (SE), Crested Porcupine Optimizer (CPO), a GPT-enhanced database tuning system (GPTuner), and a Time Series Patch Transformer (PatchTST) to address this challenge. Firstly, the CPO algorithm is used to optimize the Number of Standard Deviations (Nstd) and the Number of Realizations (NR) in the CEEMDAN, to obtain the decomposed subsequences. Secondly, the Sample Entropy (SE) method is used to reconstruct the data to reduce computational complexity. Finally, a PatchTST model integrated with GPTuner is used to predict the aggregated components after decomposition, and the sum of the predicted values of each component is taken to obtain the final wind speed forecast result. In contrast to 11 alternative models, the RMSE of the model presented in this study has seen a reduction exceeding 15%, indicating a significant enhancement in predictive accuracy attributed to the integration of CPO-CEEMDAN. The parameter optimization of PatchTST by GPTuner is feasible, and it can enhance the predictive performance of PatchTST.