Variable Support Segment-Based Short-Term Wind Speed Forecasting

Accurate short-term wind speed forecasting plays an important role in the development of energy. However, inertia airflow means that has properties time variance and inertia, which pose a challenge task forecasting. We employ variable support segment method to describe these two properties. then propose segment-based model improve accuracy. The core idea is adaptively determine future by self-attention mechanism. Historical series are first decomposed into several components variational mode decomposition (VMD). Then, values each component forecast using modified Transformer model. Finally, summed obtain values. Wind data collected from farm were employed validate performance proposed mean absolute error spring, summer, autumn, winter 0.25, 0.33, 0.31, 0.29, respectively. Experimental results show achieves significant accuracy good performance.