Onshore wind turbine technology has matured to the point when assets are now expected to produce availabilities greater than 97%. The wind speed a wind turbine operates in has an impact on its reliability. Hitherto this relationship has not been defined, quantified or used to undertake analysis to assess how wind turbine performance would be affected by conditions at a prospective site. Wind turbine reliability data comes from two modern onshore wind farms, located in Scotland, using multi-megawatt wind turbines. This information is used alongside data from meteorological masts, located on each site, to determine the mean wind speed on the day of each recorded failure that resulted in corrective maintenance. A methodology is proposed in this thesis to define the relationship between wind turbine component failure rates and wind speed using Bayes Theorem. With these relationships known and wind speed dependent failure rates calculated, component reliability is modelled using discrete Markov Chains and Monte Carlo Simulation. The model is used to extrapolate the failure rate and wind speed relationships found within the onshore dataset to a proposed onshore and offshore site. From the generated data, wind turbine annual component failure rates are calculated for each site and analysis is performed to determine how component failure rates are likely to change throughout a year due to seasonal wind speeds at each site. The calculated seasonal failure rates allow wind turbine performance to be analysed more closely than if using traditional annual failure rates. A spares optimisation model is finally proposed using the wind speed dependent failure rate model. The output of this thesis is of particular relevance to operators of offshore wind farms.
|Date of Award||29 Apr 2015|
- University Of Strathclyde
|Sponsors||EPSRC (Engineering and Physical Sciences Research Council)|
|Supervisor||David McMillan (Supervisor) & Graham Ault (Supervisor)|