Wind Power Ultra-Short-Term Power Prediction Based on VMD and Weighted Combination Model of BiLSTM and RF

To tackle the predicament of ultra-short-term wind power prognostication and assess the ramifications of turbine modifications on the projected power output, this study proposes a wind power prediction method based on VMD, BiLSTM, and RF weighted combination. Key wind-related parameters, such as changes in blade pitch angle, high-speed shaft braking pressure, and power, are monitored at 30-second intervals and utilized as input features. Initially, the VMD technique is applied to decompose the wind power data, and the optimal number of decompositions is determined using the Pearson correlation coefficient method. Subsequently, the BiLSTM and RF models are employed separately to predict each decomposed component. To enhance prediction accuracy, the inverse of errors is introduced as a weighting factor for the predictions generated by the two models. The predictions for each component are then fused to obtain the final prediction. Finally, the proposed model is validated through simulation experiments to assess its effectiveness.