In a microgrid containing wind power generation units, accurate wind power forecasting can enhance the stability of the microgrid system. Considering the numerous factors that affect wind power, forecasting wind power presents certain challenges. In order to address these challenges, a novel hybrid short term wind power forecasting model called SNAO-BITCN-BIGRU-A is proposed in this study, which integrates a multi-strategy improved Swooping Eagle Optimization Algorithm (SNAO), a Bidirectional Time Convolutional Network (BITCN), and a Bidirectional Gated Recurrent Unit network with Attention mechanism (BIGRU-A), aiming to improve the accuracy of wind power forecasting. In the proposed model, BITCN is used to extract features, BIGRU captures bidirectional dependencies, and the attention mechanism emphasizes important information, while SNAO is employed for parameter optimization of the TCN-BIGRU-A. Actual wind power data from a location in Inner Mongolia, China, was used for comparative experiments to verify its forecasting performance. The experimental results show that compared to the BITCN-BIGRU-A forecasting model, the proposed SNAO-BITCN-BIGRU-A short-term wind forecasting model reduces the error metrics MAE, MSE, and RMSE by 44.89%, 66.15%, and 41.86% respectively, and increases R2 by 3.48%. This verifies that the SNAO algorithm can optimize the parameters of the TCN-BIGRU-A forecasting model, and the proposed model can improve the forecasting precision of short-term wind power to a certain extent.