Wind Power Prediction Based on Data and Model Joint Drive

Wind power is an important part of new energy, but its development is subject to meteorological factors, which has stronger uncertainty and intermittency. To enhance the efficiency of wind energy utilization, this study introduces a wind power forecasting approach that integrates dynamic time warping (DTW), convolutional neural networks (CNN), and bidirectional long short-term memory networks (BiLSTM). By combining both data-driven and model-driven strategies, the proposed method addresses the challenges of low accuracy and high uncertainty commonly encountered in wind power prediction. From the data-driven aspect, this method calculates the similarity of two sequences through DTW curved alignment to construct similar wind condition daily data for model training, which effectively improves the quality of data-driven and avoids the early leakage of power information. The CNN-BiLSTM combination model is adopted, This model integrates CNN’s capability in local feature extraction with BiLSTM’s long-term dependency modeling through bidirectional information flow. Multi-dimensional comparative results validate the performance advantage of the proposed approach.