This PhD thesis studies anomaly detection in wind turbines and wind farms, addressing the need for reliable monitoring under uncertainty and changing operating conditions. The work begins by benchmarking state-of-the-art model-based detection methods for the electrical subsystem of wind turbines and then focuses on the more complex mechanical subsystem, where nonlinear behaviour and wind variability pose significant challenges. 

A central contribution of the thesis is the systematic treatment of uncertainty, both in residual generation and in the residual evaluation stage, leading to improved robustness and reduced false alarms. In addition to physical faults, false data injection cyber-attacks are examined within the same detection framework. The thesis also explores how data-driven techniques, including methods from topological data analysis, can complement model-based approaches in detecting subtle anomalies such as blade icing. Overall, the research contributes methods and insights toward more robust and scalable anomaly detection frameworks for wind energy systems.

After the defence there will be a small reception at Fredrik Bajers Vej 7C/2 - 209