Combining TPE-PSO and BiGRU for High-Precision Short-Term Wind Power Forecasting

Given the significant volatility and unpredictability of wind power generation, accurately forecasting its future power output has become a key element in ensuring the stability and reliability of energy systems. This paper proposes a network model based on TPE-PSO hyperparameter optimization, aimed at maximizing the search for optimal parameters globally and locally to achieve the highest precision. The predictive performance of the model was evaluated using two sets of wind power data with a resolution of 1 hour, employing four commonly used evaluation metrics in the field. The results show that all power forecasting results within 24 hours have R2 decision values consistently above 0.973. Under the same environmental conditions, our model significantly outperforms other models in predicting wind power generation, as evidenced by its superior MAE, RMSE, and NMAE. This significant research outcome provides strong support and precise guidance for the safe and stable operation of power grids in practical applications.