Hands-on tutorial about data science in wind resource assessment. You can explore this tutorial interactively via binder at github.com/flrs/predicting_the_wind
The tutorial is about the fictitious scenario of building a wind farm on the hills around the AI incubator The Sandbox San Diego. Using Python code, it explains how the wind can be measured and how the measurement can be used together with climate models and ground station data to generate a long-term estimate of the wind. Subsequently, the tutorial explores how to use that estimate to predict wind turbine power output. Finally, the tutorial puts the output of the fictitious wind farm into the broader context of the California power grid.
From a data science perspective, the tutorial touches on data exploration, modeling and validation, and clarifies where domain knowledge of wind and fluid dynamics can help improve the wind estimate.
The tutorial uses the following tool stack:
- Python for performing calculations
- Jupyter Notebook as an IDE
- RISE Jupyter extension for presenting the notebook
- hide_code extension for hiding some (long-ish) code to build maps
- Folium to display maps
- SciPy's orthogonal distance regression
- scikit-learn's RandomForestRegressor
- brightwind to support wind resource assessment tasks