Abstract: To enhance the wind speed prediction accuracy in wind farms and address the challenges associated with multi-source data fusion and parameter optimization, a combined prediction model is proposed based on the improved hybrid cuckoo search quantum-behaved particle swarm optimization (ICSQPSO), a convolutional neural network (CNN), a gated recurrent unit (GRU), and an attention mechanism. A multimodal feature fusion framework is constructed, in which the CNN extracts spatial features, the GRU captures temporal dependencies, and the attention mechanism dynamically weights key factors. The ICSQPSO algorithm is designed to integrate quantum search and Levy flight strategies for optimizing model hyperparameters and avoiding local optima. Experiments based on 75 days of wind farm data show that the model achieves a root mean square error (RMSE) of 1.537 m/s and a mean absolute percentage error (MAPE) of 1.6% on the test set. Compared with the sparrow search algorithm (SSA), grey wolf optimizer (GWO), and particle swarm optimization (PSO) models, the proposed method reduces the RMSE by 10.1%, 60.9%, and 62.4% respectively, and achieves a determination coefficient of 0.9947 for the prediction results. The results demonstrate that this model effectively improves wind speed prediction accuracy in complex scenarios through feature fusion and parameter optimization, providing technical support for new energy accommodation.