Investigating Lab-scaled Offshore Wind Aerodynamic Testing Failure and Developing Solutions for Early Anomaly Detections

Abstract. Experimental demonstration of offshore wind components and systems plays an increasingly critical role in advancing the industry. As these systems increase in complexity, the likelihood of human error or software malfunction increases, leading to costly equipment damage. This study examines a laboratory incident which occurred during aerodynamic characterization of a 1:50 scale 5 MW reference wind turbine and presents an efficient early anomaly detection method. During the experiment, the model generator disengaged causing a catastrophic rotor overspeed and subsequent blade-tower strike. Applying data-driven approach to predict system's dynamics from measurements, this study investigates the potential to enhance reaction time and prediction quality using single- and multi-principal component models. Early system malfunctions are detected with the single-principal component model showing better performance. Sensitivity analyses show gains in reaction time with increasing sample frequency, lending this work in particular to lab-scale systems that operate at high sample rates to reduce future incidents.