Failure Rates in Offshore Wind Turbines: Reducing the Uncertainties and lowering the cost of energy

Fraser Ewing 2 ,1, Benson Waldron1
1DNV GL, London, UK, 2University of Edinburgh, Edinburgh, UK



Operation & Maintenance (O&M) costs constitute a significant part of the total lifecycle cost of an offshore wind farm (OWF), many studies conclude this is around 14-30%. Offshore Wind turbine (OWT) failure rates have a direct and significant effect on these costs; accurately quantifying and assessing the failure rates of OWT's is therefore of great importance to target the biggest reductions in O&M costs. This paper focuses on addressing the uncertainty around failure rate predictions for OWT's. OWT reliability data is scarce given its commercially sensitive nature, so often outdated or surrogate sources are used and then corrected using handbook style ‘fudge' factors. This uncertainty has ramifications for O&M planning and strategy. Also, failure rates for components are typically given as point estimates derived from classical statistical approaches which provide no validated correction for accuracy of the underlying data. These two factors result in a large level of uncertainty associated with the failure characteristics of OWT's.

This paper seeks to address this issue by applying a Bayesian statistical approach that provides a framework for assessing OWT reliability alongside quantified empirical data.



The Bayesian updating method deployed in this work involves the construction of a prior probability distribution that expresses knowledge about the parameters of interest. Then a likelihood probability function is developed that incorporates component specific data from state of the art onshore wind failure databases (Reliawind). One of the merits of the Bayesian approach is that data can be used to ‘update' the models as it becomes available; hence different correction methods are applied to the likelihood function to correct for the offshore environment and thus their effects on the resulting posterior distribution are assessed. This approach enables comparisons to be made between correction methods whilst also inherently providing a quantification of uncertainty in failure rate estimates for key OWT components.


By analysing several critical OWT components (pitch system, frequency converter and generator) through a Bayesian framework, this work provides empirical evidence of failure rates and associated uncertainties. The results show that adopting a Bayesian approach for the reliability assessment of OWT's can help to reduce the uncertainty around component and turbine failure rates by comparing this against a classical approach.



This novel data driven approach for developing failure rates for OWT components could help to provide a step change in the accuracy of reliability predictions used in O&M strategy and planning. Commonly held perceptions of the reliability of OWT components may well be challenged using the Bayesian methods and state of the art data shown in this work. The implications for developers, investors and insurers will be a greater understanding of and confidence in failure rates for OWT's.


 The methods and results presented here will be essential for many different offshore wind stakeholders. Quantified empirical reliability data alongside methods to reduce uncertainty in failure rate estimates will be invaluable for project developers, investors and insurers in the industry.