Monday  |  Salon 9  |  02:20 PM–02:40 PM

#16969–Uncertainty Quantification of Bending Moments with Characterization of Strain Measurement Error on Offshore Wind Turbines

In recent years, a growing number of governments around the world have demonstrated a strong commitment to harnessing the power of renewable energy sources. As part of this effort, there has been a particular emphasis on developing effective strategies for monitoring offshore wind turbines (OWTs) in both the short and long term. This focus on improving the efficiency and reliability of OWTs is driven by a recognition of their potential to generate clean, sustainable energy, and the need to ensure that they are functioning optimally at all times. Strain information is especially important in detecting damage, fatigue, and faults and ensuring the structure's performance over time. At the same time, however, it is known that strain gauges are not extremely robust sensors that could provide unreliable information over time. Strain gauges produce measurements that carry errors, which are attributed to two main sources: the inherent manufacturing variation and a time-dependent drift after the gauges are installed.

This paper proposes a framework to quantify and propagate the uncertainty of strain gauges in wind turbines. The framework has four steps: (1) measure and evaluate the strain calibration factor over time using LSTM neural network with strain and operational data as input, (2) use the calibration factor to derive the corrected strains and bending moments; (3) quantify the calibration fluctuation by comparison with the manufacturing tolerances from literature and data sheets, and (4) estimate the model error from LSTM architecture using deep ensemble learning. The framework can quantify the uncertainty of strain gauges and propagate it to the moments in fore-aft and side-to-side direction as probability distributions. The framework will be tested with data from a 6MW turbine in Block Island Wind Farm.

Eleonora Tronci   Northeastern University
Anna Haensch   Tufts University
Georgios Georgalis   Tufts University
Babak Moaveni   Tufts University