A new design of wind power prediction method based on multi-interaction optimization informer model.

Area of Science:

• Renewable Energy Systems
• Artificial Intelligence in Energy
• Machine Learning for Power Grids

Background:

• Accurate wind power prediction is crucial for grid stability.
• Traditional neural networks and the Informer model face limitations in complex conditions due to feature coupling and data health issues.
• Existing methods struggle with prediction accuracy and computational efficiency.

Purpose of the Study:

• To propose a novel prediction framework, MFIO-Informer, for enhanced wind power forecasting.
• To improve prediction accuracy and computational efficiency by optimizing multi-source feature interactions and data health perception.
• To address the limitations of the Informer model in complex operational environments.

Main Methods:

• Feature screening using Lasso and Pearson correlation to identify key multi-source features.
• A fully-connected neural network (FNN) to extract the Dynamic Synergistic Coefficient (DSC) reflecting equipment performance.
• A data health assessment using historical power data and DSC to generate an optimization matrix for the Informer model.
• Implementation of the MFIO-Informer framework integrating feature optimization and data health perception.

Main Results:

• The MFIO-Informer model demonstrated superior performance on two public wind power datasets.
• Achieved approximately 20% higher prediction accuracy compared to the traditional Informer model.
• Realized a 54.85% faster prediction speed, indicating significant computational efficiency gains.

Conclusions:

• The proposed MFIO-Informer framework effectively addresses the limitations of existing wind power prediction models.
• Integrating physical feature collaborative analysis and data health status perception significantly enhances prediction accuracy and speed.
• MFIO-Informer offers a promising solution for stable and efficient wind power grid integration.