Fraunhofer Institute for Wind Energy Systems (IWES) and AI company Latoda are working together in the "WindKI" research project to develop a diagnostic method that detects performance losses in wind turbines at an early stage. The project, funded by the German Federal Ministry of Research, Technology and Space, aims to deliver an AI-supported system for objective, data-driven performance optimisation.
Wind turbines often generate less electricity than forecast before construction due to factors such as wind conditions, rotor blade settings, or wake effects.
The WindKI project combines measurement data, simulations, and artificial intelligence to build a diagnostic system using heuristic algorithms and machine learning. Initial models are based on high-resolution SCADA data from the 8 MW Adwen AD8 research turbine provided by Fraunhofer IWES. Latoda is developing an analysis system that determines whether a turbine is operating as expected, flags underperformance, and suggests relevant optimisation parameters.
The system not only identifies anomalies but also indicates their likely causes, such as blade angle settings or site-specific conditions. This approach gives operators a tool to address problems more quickly and enables the wind industry to apply AI models to a wide range of time series data.
