Research on Short-Term Wind Power Prediction Based on an ICEEMDAN-FuIILinear-TSMixer-CNN Combined Model

In response to the challenges of high randomness, significant fluctuations, and low prediction accuracy in wind power, we propose an improved complete ensemble empirical mode decomposition with adaptive noise (ICEEMDAN) method to decompose wind power data. The decomposed components are then sorted and regrouped using a multi-scale permutation entropy algorithm. Subsequently, these regrouped components are analyzed and predicted using a combination of fully connected linear (FullLinear), time-series mixer (TSMixer), and convolutional neural network (CNN). The final prediction values are obtained by combining the results of these predictions. Validation using SCADA data from a Turkish wind turbine indicates that this method significantly improves prediction performance compared to single models, with an overall improvement of approximately 40% to 50%. This approach enables accurate short-term wind power prediction, providing a solid foundation for the stable operation and optimal scheduling of the power grid.