This paper proposes a combined model for ultra-short-term offshore wind power prediction based on principal component analysis (PCA), sparrow search algorithm (SSA), variational modal decomposition (VMD), and bidirectional long- and short-term memory neural network (BiLSTM). The model aims to improve the accuracy of wind power forecasting by reducing data dimensionality, optimizing signal decomposition, and enhancing the prediction model through intelligent parameter optimization. First, PCA is used to reduce the dimensionality of the data by extracting the most significant features. Then, VMD, optimized by SSA, is applied to decompose the wind power time series into multiple frequency components, effectively removing noise. The hyperparameters of the BiLSTM model are further optimized using the SSA algorithm to enhance prediction accuracy. The final prediction results are obtained through simulation experiments, demonstrating that the proposed model significantly improves prediction accuracy compared to existing methods. The model is evaluated using metrics such as RMSE, MAE, MAPE, and R², showing high accuracy and stability. The results indicate that the PCA-SSA-VMD-BiLSTM model outperforms other models in wind power prediction, particularly in handling the high variability and uncertainty of offshore wind power. The study highlights the effectiveness of combining dimensionality reduction, intelligent optimization, and advanced neural networks for accurate and reliable wind power forecasting.This paper proposes a combined model for ultra-short-term offshore wind power prediction based on principal component analysis (PCA), sparrow search algorithm (SSA), variational modal decomposition (VMD), and bidirectional long- and short-term memory neural network (BiLSTM). The model aims to improve the accuracy of wind power forecasting by reducing data dimensionality, optimizing signal decomposition, and enhancing the prediction model through intelligent parameter optimization. First, PCA is used to reduce the dimensionality of the data by extracting the most significant features. Then, VMD, optimized by SSA, is applied to decompose the wind power time series into multiple frequency components, effectively removing noise. The hyperparameters of the BiLSTM model are further optimized using the SSA algorithm to enhance prediction accuracy. The final prediction results are obtained through simulation experiments, demonstrating that the proposed model significantly improves prediction accuracy compared to existing methods. The model is evaluated using metrics such as RMSE, MAE, MAPE, and R², showing high accuracy and stability. The results indicate that the PCA-SSA-VMD-BiLSTM model outperforms other models in wind power prediction, particularly in handling the high variability and uncertainty of offshore wind power. The study highlights the effectiveness of combining dimensionality reduction, intelligent optimization, and advanced neural networks for accurate and reliable wind power forecasting.