Graph Neural Networks in Wind Power Forecasting

2025-06-30

AI Summary

This study investigates the modeling capability of Graph Neural Networks (GNNs) for medium- to short-term wind power forecasting. Addressing the 24–36-hour prediction horizon, we conduct experiments across three real-world wind farms, integrating multi-source Numerical Weather Prediction (NWP) variables and historical power measurements as inputs, and constructing a dynamic spatiotemporal graph structure based on geographical proximity and meteorological correlation. We propose a GNN architecture specifically tailored for wind energy forecasting and systematically evaluate its performance on five years of historical data. Results demonstrate that the proposed GNN achieves accuracy comparable to the best-performing CNN baseline, while offering superior physical interpretability and cross-site generalization capability. To our knowledge, this is the first work to empirically validate GNNs’ effectiveness in capturing complex spatiotemporal meteorological dependencies within operational wind farm settings—establishing a novel paradigm for renewable energy forecasting.

Technology Category

• Machine Learning: Graph-based Machine Learning
• Reasoning under Uncertainty: Graphical Models
• Data Mining & Knowledge Management: Graph Mining, Social Network Analysis & Community

Application Category

• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Web Mining and Content Analysis: Large pretrained models with web data
• Social Networks and Social Media: Influence propagation, information diffusion, and the prediction on networks

Abstract

We study the applicability of GNNs to the problem of wind energy forecasting. We find that certain architectures achieve performance comparable to our best CNN-based benchmark. The study is conducted on three wind power facilities using five years of historical data. Numerical Weather Prediction (NWP) variables were used as predictors, and models were evaluated on a 24 to 36 hour ahead test horizon.

Problem

• Assessing GNNs for wind energy forecasting accuracy
• Comparing GNN performance with CNN benchmarks
• Evaluating models using NWP data across multiple facilities

Innovation

• GNNs applied to wind energy forecasting
• Performance comparable to CNN benchmarks
• Uses NWP variables as predictors