 This paper proposes a new approach for accurate wind power forecasting. It first analyzes the correlation between different features and selects those with the most significant impact on the forecasts. It then divides the data into groups according to wind speed and wind direction and uses a bi-LSTM neural network to predict the power of each group. Finally, it uses a light GBM algorithm to correct any errors in the predictions. This method has been tested on real-world data and shown to be effective at improving the accuracy of wind power forecasting. This article was authored by Zifu Liu, Sinili, and Haiyan Zhao.