Transformer-KAN: A Novel Deep Learning Model for Improving Wind Power Forecasting Accuracy by Integrating Multi-source Data

This study proposes Transformer- Kan,a novel deep learning model for short-term wind power forecasting that synergistically integrates the Transformer architecture with the Kolmogorov-Arnold Network (KAN). The model addresses the limitations of traditional forecasting methods in capturing complex temporal patterns and incorporating diverse data sources. Transformer-KAN employs a Transformer encoder to process time series features and a KAN encoder to extract and fuse external knowledge, with a feature fusion module integrating their outputs. Evaluated on high-resolution operational data from a wind farm in Xinjiang, China, Transformer-KAN demonstrates superior performance over benchmark models. Experiment results underscore the efficacy of our integrated approach in enhancing wind power forecasting accuracy, contributing to the advancement of renewable energy systems and smart grid technologies.