Wake Steering with Reinforcement Learning Agents in Windparks

What are wake effects?

Wake effects are the turbulent, slower air streams created behind a wind turbine as it extracts energy from the wind. Turbines positioned downstream encounter this disturbed flow, which reduces their power generation and increases mechanical stress. The strength of these effects depends heavily on turbine spacing, wind speed, atmospheric stability, and rotor size. Managing wake interactions through optimized layouts and active wake control is key to improving overall wind farm performance. The key question to answer is how to optimize overall wind park performance while maintaining the longevity and reliability of wind turbines.

KI4Wind - Project Motivation

To achieve climate neutrality in energy supply, it is essential to expand and optimize renewable energy sources - especially the already installed wind turbines contain huge potential to optimize with smarter wake steering algorithms.

Currently, the control systems of modern wind turbines (WTGs) focus solely on the individual turbine, without taking into account how other turbines within the wind farm behave. However, by optimizing their interaction, turbine operation can be improved, mechanical loads reduced, and thus the lifetime extended. This leads to higher overall efficiency and increased net yield of wind farms.

Objectives and approach

The KI4Wind project aims to research the use of machine learning (ML) methods to identify specific operating and load conditions of wind turbines, in order to reduce physical stresses and optimize electrical power output.

To achieve this, sensors from the partner company Fibercheck GmbH will first be integrated into real wind farms to strategically enhance existing measurement datasets. Based on these real measurements and additional synthetic simulation data provided by the research partner Fraunhofer IWES. Turbit Systems GmbH will take the lead in developing and training AI-based control agents for wind turbines. These agents will then be tested and optimized in real-world field trials by the project consortium.

Innovations and perspectives

Compared to existing technologies, the project offers the potential for optimized operation of entire wind farms through Reinforcement Learning Agents (RL-Agents). The models developed in KI4Wind are intended to be integrated either as independent modules into local turbine control systems or applied centrally through a cloud computing approach for operational optimization.

In the medium term, the project will strengthen the competitiveness of German companies in the wind energy sector and thereby make a long-term contribution to achieving climate neutrality.

Project partners

Funded by: