Abstract: To address the low prediction accuracy of single models caused by the non-stationarity of wind speed series, a hybrid forecasting model integrating Chaotic Particle Swarm Optimization (CPSO), Variational Mode Decomposition (VMD), Temporal Convolutional Network (TCN), and Long Short-Term Memory (LSTM) network is proposed. CPSO is used to adaptively optimize VMD parameters, decomposing the wind speed series into multiple intrinsic mode functions. A TCN-LSTM network is then constructed to extract both local and long-term temporal features. Experiments on the Szeged and Jena datasets demonstrate that the proposed model outperforms comparison models in terms of RMSE, and MAE, and ablation experiments verify the effectiveness of the model.