At a glance
- It is not possible to detect the causes of the underperformance of wind turbines reliably with the methods currently available. However, an accurate evaluation of the power curve is a cornerstone for ensuring reliable energy production.
- In the WindKI project, the project partners are developing AI methods to determine the effective wind speed and thus to verify the electrical output in operation depending on the wind conditions at the time. The aim is an AI-supported diagnostics system for performance optimization.
- In addition to the actual and synthetic data from its Adwen AD8 research wind turbine, Fraunhofer IWES is also contributing its experience in model validation methods among other things.
The challenge
Wind turbines do not always generate as much electricity as predicted prior to their installation. There can be a number of reasons for this: typical causes can be found in aerodynamics, including for example erosion of the blade surface and errors in the blade pitch or operation in the wake of other wind turbines. It is not possible for operators to determine the actual reasons with the methods currently available. Although wind measurement data are recorded during operation, the resolution, quantity, and quality are not sufficient for a targeted, data-driven root cause analysis.
The individual power curves, which are provided by the respective manufacturer and generated for each type of turbine prior to its installation and show the relationship between wind speed and power output, are only of limited help and cannot be independently verified using the methods currently available. Without a detailed evaluation of the power curve, it is not possible to ensure the performance of the turbine or to detect and thus rectify any underperformance.
The solution
This is where the WindKI project comes into play: The project partners are developing AI methods to determine the effective wind speed and thus to verify the electrical output of wind turbines depending on the wind speed. In addition, the scientists are investigating data-driven methods to identify possible failure modes as the cause of underperformance. The complex interactions between loads, wind speed, rotor speed, generator torque, and power play a key role here. The aim is an AI-supported diagnostics system for performance optimization.
The method will be developed using a wealth of actual measurements and synthetic data from the Adwen AD8 research wind turbine provided by IWES and the alpha ventus offshore research wind farm. In addition to these data, Fraunhofer IWES is contributing its experience in model validation methods. In the WindKI project, Fraunhofer IWES will therefore be responsible, among other things, for reviewing and processing the data, running simulations, and validating the results.
The added value
An accurate evaluation of the power curve is a cornerstone for ensuring the energy production and utilization of the technology employed. This makes it possible to utilize the limited areas available on land and at sea for the production of electricity from wind energy optimally. Eliminating underperformance is also economically interesting for operators. Last but not least, the development of an AI-supported algorithm for detecting and identifying underperformance will enable industry to offer more efficient and competitive products.