An open-source framework for developing probabilistic wind forecasting models using deep learning, designed to integrate with wind farm control systems.
WindForecastFramework provides a sophisticated platform for developing, tuning, training, and evaluating state-of-the-art deep learning models for probabilistic wind forecasting. Our framework handles complex temporal dependencies in wind farm operational data and integrates seamlessly with downstream control systems.
Our framework consists of these interconnected repositories:
| Repository | Description |
|---|---|
| wind-forecasting | 🚀 Main Application - Core forecasting framework with data preprocessing, training pipelines, and evaluation tools |
| pytorch-transformer-ts | 🧠 Model Implementation - Implemented time series transformer models for time series forecasting (forked) |
| gluonts | 📊 Base Time Series Library - Customized fork of GluonTS for probabilistic time series modeling |
| wind-hybrid-open-controller | 🎮 Downstream Application - Wind farm control system that consumes forecasts |
- 📈 Probabilistic Forecasting: Generate probabilistic forecasts with uncertainty estimation
- 🔄 Multiple Architectures: Support for various transformer-based models for time-series forecasting in pytorch
- 🖥️ Distributed Training: Scale from local development to HPC environments
- 🧪 Hyperparameter Optimization: Distributed optimization across multiple GPUs
- 🔗 Control Integration: Seamless connection with wind farm control systems
 Aoife Henry |
 Juan Boullosa |
This project is licensed under the MIT License - see the individual repositories for specific details.
- TODO: Add documentation and links to pages/repository docs
Built with ❤️ for renewable energy research
