Session: 02-08-02 Reliability of Renewable Energy Devices 2

Submission Number: 176822

Reliability-Based Lifetime Estimation of Offshore Wind Turbines Considering Structural Health Monitoring Data and Operational Variability 

As offshore wind turbines (OWT) are increasingly expected to operate beyond their original design life, a key challenge for operators is to ensure reliable and cost-effective lifetime extension under evolving operational and environmental conditions. Reliability-based fatigue assessment provides a probabilistic framework to account for uncertainties in environmental conditions, load simulation models, and stress analysis [1]. In practice, prior assumptions on these uncertainties have been typically defined during the design phase, often conservatively, due to limited availability of statistical data. With the growing availability of structural health monitoring (SHM) data, it has become possible to refine these prior assumptions and update the long-term load spectrum using measurement-based evidence [2]. Moreover, while traditional design load cases provide a structured representation of environmental and operational variability, they generally overlook potential operator-side decisions, e.g., derating - intentional reduction of power output to balance grid requirements, that can significantly influence structural loading and fatigue progression.

Addressing such scenarios, this study presents a probabilistic framework for updating the long-term load distribution based on monitoring data and assessing the effect of operator-side decisions on the structural reliability and remaining lifetime. The proposed approach is implemented on a real-world case study and monitoring data collected from an operating wind turbine. In particular, the scale parameter of the long-term Weibull load spectrum is modelled as a time-variant random variable in the prior reliability analysis [3]. Monte Carlo simulations (MCS) are performed to propagate uncertainties in the load spectrum and SN-curves through a Miner’s rule–based fatigue damage model. Probability distributions of fatigue damage and time-dependent failure probabilities are computed, forming the baseline for remaining useful lifetime (RUL) estimation of the OWT and reliability updating.

The uncertainty of the long-term load spectrum is updated using several years of site-specific load measurements obtained from the operating turbine. In this work, we rely on dynamic Bayesian networks (DBNs) to integrate load monitoring data and refine the probabilistic model of the load environment. Transition models within the DBN are derived from the MCS-based fatigue damage results, enabling the joint representation of the interacting uncertainties. The measurement uncertainty of SHM is also considered in the DBN observation models. The Bayesian updating procedure refines the scale parameter of the Weibull load distribution, not only reducing epistemic uncertainties but also reflecting the actual site conditions and operational conditions. Through this formulation, the posterior distributions of the Weibull parameter and associated fatigue damage distribution are updated, yielding an evolving estimate of reliability and RUL.

In addition, the collected data also indicate that during the monitoring period,[WW1]  the operator occasionally applied derating, which had not been considered in the design phase. Distinct load characteristics associated with normal and derated operation have been identified from the collected data, demonstrating measurable differences in stress range distributions between control modes. To assess the influence of such operator-side decisions on reliability and lifetime, hypothetical operational variabilities were generated by changing the proportion of derating (e.g., 10%, 20%, etc.). For each scenario, the expected long-term load distribution is reconstructed based on the probabilities of normal and derated conditions, as inferred from the measurement data. The reconstructed load distributions are then propagated through a probabilistic fatigue damage model to evaluate the corresponding end-of-life failure probability and remaining useful lifetime. The case study highlights that operational decisions can substantially influence the remaining lifetime, underscoring the importance of accounting for such decision effects in reliability assessments and lifetime management of offshore wind turbines.

References

[1] Velarde, J., Kramhøft, C., Sørensen, J. D., & Zorzi, G. (2020). Fatigue reliability of large monopiles for offshore wind turbines. International journal of fatigue, 134, 105487.

[2] Hlaing, N., Morato, P. G., & Rigo, P. (2022). Probabilistic virtual load monitoring of offshore wind substructures: A supervised learning approach. In ISOPE International Ocean and Polar Engineering Conference (pp. ISOPE-I). ISOPE.

[3] Hlaing, N. (2024). Data-Driven Virtual Monitoring and Life-cycle Management of Offshore Wind Support Structures. PhD Thesis. Universite de Liege (Belgium).

Presenting Author: Nandar Hlaing Vrije Universiteit Brussel

Presenting Author Biography: Nandar Hlaing is a postdoctoral researcher at Vrije Universiteit Brussel. Her research focuses on probabilistic deep learning and decision-making under uncertainty, with major applications in offshore wind turbine support structures. Her work combines Bayesian inference, uncertainty quantification, and machine learning to support structural health monitoring and data-driven asset management in offshore energy applications.

Authors:

Nandar Hlaing Vrije Universiteit Brussel
Ahmed Mujtaba Vrije Universiteit Brussel
Wout Weijtjens Vrije Universiteit Brussel
Christof Devriendt Vrije Universiteit Brussel