Fault detection and diagnosis, predictive maintenance, wind farms, wind energy
The project aims at increasing productivity and making wind energy generation smarter by optimizing operations of on- and off-shore wind energy generation through monitoring and modelling techniques enabled by Artificial Intelligence (AI) and Digital Twins. To achieve this target the project partners will work on data enrichment towards high- quality annotated ground truth data and the coupling of AI solutions for operational monitoring data of wind farms with a multiscale wind resource model chain for advanced fault detection and diagnosis towards predictive maintenance capabilities. This allows to define smart KPI’s that support wind farm operators during their decision-making processes.
The monitoring and diagnostic needs of wind farms are significantly more complex and demanding every year. As of today, detection and assessment of underperformance in wind turbines are typically executed in a semi- manual top-down approach i.e. fleet, site, turbine. On this basis, monitoring-based fault analysis and diagnosis are time-consuming, expert dependent, and often of insufficient accuracy. As a result, several underperformance issues and failure modes may either remain undetected, get falsely diagnosed or their root-cause stays unidentified.
These limitations are greatly due to little automation, an area where AI could help. Furthermore, AI can assist in deducting smart and understandable insights from the complex wind turbine machine involving many nonlinear mechanical, electrical and thermal interactions. However, the lack of high-quality ground truth data to train and evaluate AI models is problematic. Particularly, this lack of annotated trustworthy data makes that AI solutions today cannot provide precisely quantified and realistic rates on fault diagnosis accuracy and confidence level. We will address this issue relying on the direct access to over 15 GW of operational wind farm monitoring data collected through SynaptiQ, the monitoring and asset management portal from 3E with currently over 10,000 wind and solar farms connected to it. Furthermore, the involvement of a User Committee, composed of selected customers from 3E, will ensure additional insights on technical issues, access to O&M records and to field data, and will further enrich the value of the monitoring data.
The outcomes of the project will provide tools and intelligence for the optimization of on- and off-shore wind farms operations through smart automatic fault detection and advanced diagnosis enabling risk mitigation and empowered decision making. The envisaged solutions include a virtual met mast service, sensor anomaly detection, yaw and blade pitch misalignment analysis, drivetrain fault detection, performance degradation quantification, and remaining lifetime estimation of components. These solutions will be validated in an industrially relevant environment (TRL5) through this project.
May 2021 - April 2024
With the support of