Deep Reinforcement Learning for offshore wind farms layout optimization
Description
Layout optimization of wind turbines is a complex problem that requires the analysis of multiple variables such as wind speed, wind direction, and the power curves of the turbines. The optimization of the layout of wind turbines can lead to an increase in energy production and a reduction in the cost of energy production.
The key aspects of this project are:
- The use of Deep Reinforcement Learning to optimize the layout of wind turbines.
- The analysis of the wind speed, wind direction, and power curves of the turbines.
- The optimization of the simulation process to reduce the computational cost.
Value provided by Taniwa
Taniwa developed:
- A system to optimize the process already provided by Pywake.
- A framework to simulate millions of layouts in a reasonable time.
- A system to analyze the results and identify the best layout.
- A Deep Reinforcement Learning model and Curribulum to optimize the layout of wind turbines.
We had to understand pywake, the data, and the problem. Then we had to optimize the simulation process to reduce the computational cost and develop a Deep Reinforcement Learning model to optimize the layout of wind turbines.
Technologies
- Pywake for the simulation of wind turbines.
- Ray for the orchestration of the simulations.
- RayLib for DRL.
- Python and Jupyter Notebooks for data processing.
- React for the interface.