Back to the programme printer.gif Print



Tuesday, 11 March 2014
16:30 - 18:00 How does the wind blow behind wind turbines and in wind farms?
Science & Research  


Room: Llevant
Session description

Wind turbines operate under highly fluctuating wind conditions. It is thus important to achieve a profound understanding of the characteristic features of the micro scale meteorological conditions. Current research activities focus not only on the inflow conditions and their impact on wind turbines, but also on the wake structures and the wind conditions within a wind farm.

Learning objectives

  • Get a better understanding micro scale wind conditions
  • Learn about new advanced measuring techniques
  • See the possibilities of numerical methods to simulate complex wind conditions
  • Learn about the impact of wind on turbine components
Lead Session Chair:
Joachim Peinke, Uni Oldenburg, Germany

Co-chair(s):
Jakob Mann, DTU Wind Energy
Graeme Wilson University of Strathclyde, United Kingdom
Co-authors:
Graeme Wilson (1) F P David McMillan (1)
(1) University of Strathclyde, Glasgow, United Kingdom

Printer friendly version: printer.gif Print

Presenter's biography

Biographies are supplied directly by presenters at EWEA 2014 and are published here unedited

Graeme Wilson is a PhD student at the Centre for Doctoral Training for Wind Energy Systems at the University of Strathclyde, Glasgow. He obtained his MEng in Civil Engineering from the University of Strathclyde in 2010. His research interests include wind turbine reliability, operation and maintenance and statistical modelling.

Abstract

Quantifying the Impact of Wind Speed on Wind Turbine Component Failure Rates

Introduction

Previous research has shown there is a clear relationship between wind speed and wind turbine reliability (Hahn, 1997; P J Tavner et al., 2013; Wilkinson et al., 2012). However, there have been no efforts made to construct a detailed model of this relationship at component level.
This paper will model the relationship between wind speed and wind turbine component failure rate and investigate the impact of wind speed on reliability. Assessments will be made on the impact of the wind speed offshore on wind turbine reliability and the potential of using this model to plan maintenance schedules and spares holdings will be discussed.


Approach

Often the reliability of engineering systems is described as being discrete in that it can exist in one state until a transition occurs and the system changes to another state. This characteristic means that many systems can be modelled as a Markov process. Markov Chains have been used often to model components and systems, with failure rates denoting the probability of transition between states (Besnard & Bertling, 2010; McMillan & Ault, 2007).

Data used in this analysis consists of 380.9 wind turbine years of reliability data from a dataset which features one model of multi-megawatt turbine all of a similar age. This approach improves on previous research which used a different methodology and smaller databases featuring wind turbines with lower generating capacity and focused exclusively on direct drive topologies (Hahn, 1997; P J Tavner et al., 2013). Wind speed data comes from Weather Station 1 (WS1), which is located nearby to the wind farm.
The methodology used in this paper is described in figure 1. Firstly failures are categorized according the failed component; this data then has a non-parametric function fitted to it which shows the probability of a wind speed occurring given a component failure. Using this historical data from the reliability database and wind speed data from an onsite meteorological mast, the probability of a failure occurring given a wind speed is then calculated.
A Markov Chain Monte Carlo model has been developed using these weather dependent failure rates and determines the impact of wind speeds on the availability of the wind turbine and its components throughout its lifetime. Seasonal component failure trends and predicted site availabilities are calculated using historical wind speed data as a model input.

Figure 2 summarises the simulation method where t represents time in days since the start of the simulation, a denotes the availability of the wind turbine, which is either available (a = 1) or unavailable (a = 0). The downtime and failure rate are, respectively, d and λ. Both d and λ are functions of wind speed, w which is a function of time, t.

Main body of abstract

Simulations produced results which show the relationships between the failure rates of various wind turbine components and average daily wind speed, maximum daily wind speed and daily wind speed variance. This abstract will focus on how the failure rates of the drive train, control system, yaw system and hydraulics are affected by average daily wind speed.

Figure 3 shows the impact of average daily wind speed on controller failure rates. The average failure rate of the controller is 0.38 failures per wind turbine year, however as the wind speed increases to rated speed (12m/s), the failure rate rises to 0.59 failures per wind turbine year and continues to increase until 22m/s.
At high wind speeds the failure rate rises considerably, however as shown in figure 3, the average daily wind speed on the wind farm site rarely surpasses 15 m/s. So despite the failure rate being extremely high past 15m/s, the risk of a controller failure taking place at this wind speed is low.
The model has the ability to filter the reliability data according to the severity of the failure which occurred, this allows for a more detailed investigation of the results. The severity is measured in the number of hours downtime caused by the failure: in this analysis a major failure is considered to cause a downtime of greater than 24 hours (Faulstich, Hahn, & Tavner, 2011).

Figure 4 shows the failure rate profiles for the drive train and the daily average wind speed. The 0 Hour filter consists of drive train failures which caused a downtime of greater than 0 hours. Interestingly the failure rate shows a sharp increase from daily average wind speeds of 3 m/s to the rated speed 12 m/s. From this point the failure rate declines from 0.23 to 0.15 failures per wind turbine year. However when failures which caused downtimes of less than 24 hours are removed the wind speed dependant failure rate shows an upward trajectory from 3 m/s to 20 m/s. This indicates that as the wind speed increases, the severity of any drive train failure also increases, causing longer downtime and reduced availability.

Using wind speed data gathered from a second site in the central belt of Scotland as an input, the model can evaluate the components most at risk of failure throughout the year. The wind resource at Weather Station 2 (WS2) is stronger than the wind resource used in calibrating the model from WS1, as shown in figure 5 and figure 6. The impact of this stronger wind speed is an increase in unavailability due to corrective maintenance from 0.58% to 0.69%. Wind speed on both sites changes seasonally, as illustrated in figure 6, therefore the failure rates of the components change change seasonally as well.

As presented in figure 7, there is a strong seasonal trend in the failure rate of all four components. All fail more often in the winter months compared to the summer which is expected given the lower wind speeds in the summer months shown in figure 6. There are two consequences to the higher wind speeds in the winter months. Firstly there are noticeably more failures, but secondly the high wind speed means there are fewer opportunities to maintain the wind turbine as there is restricted access due to safety at high wind speeds (McMillan & Ault, 2008). Therefore with the foresight of knowing how many more components are likely to fail in the winter months, this model makes it possible to plan preventative maintenance in the summer months, focus on the components likely to cause the most downtime in the winter and keep adequate replacements on the spares list.

Estimates can be made of the failure rate of the whole wind turbine system by summing the wind speed dependant failure rates for every component over the wind speed range. For a typical onshore wind farm with a mean hub height wind speed of 7m/s the model calculates a failure rate of 0.84 failures per wind turbine year as shown in figure 8. For a North Sea wind farm with an average wind speed of 11.5m/s the failure rate increases by approximately 50% to 1.24. This is a conservative estimate though as it assumes that the failure rate curve is linear, which is not the case for wind speeds in excess of 12m/s.



Conclusion

This research has produced evidence that there is a relationship between average daily wind speed and wind turbine failures. A procedure has been demonstrated which uses this relationship to calculate component failure rates based on the wind speed experienced on site. It was found that as the average daily wind speed increases the probability of a failure occurring in the drive train, control system, yaw system and hydraulics also increases.
These wind speed dependant failure rates have then been used in a Markov Chain Monte Carlo simulation which calculates the seasonal effects of the wind speed on component reliability. All four components evaluated displayed seasonal trends.
This approach at modelling wind turbine component failure rates differs to those used by previous authors and presents new applications and a different perspective on wind turbine reliability. This research could be used potentially by operators and manufacturers for better informed preventative maintenance and to help stock spares holdings. It could also be used to predict the operation and maintenance cost of a site based on the wind resource.
Offshore reliability estimates have been made, however the dataset does not contain any offshore wind turbines which are larger and likely to be less reliable than those used in the dataset.
Future work will aim to:
• Make more reliable analysis using a larger and richer dataset containing wind turbines of various ages.
• Extrapolate wind speed dependant failure rates beyond the limits of the calibration data.
• Calculate the economic benefit of using this model to decide maintenance strategy both onshore and offshore by building a maintenance model.
• Improve reliability of model for estimating offshore failure rates and account for vessel logistics and accessibility.
This research has improved on previous research in this field by modelling how the reliability of a wind turbine, at component level, changes according to daily average wind speed. Additional research not presented in this abstract includes assessments of the impact of changeable wind speed conditions and maximum daily wind speeds on component reliability.



Learning objectives
• Component failure rates increase as wind speed increases, in the case of the drive train, control system, hydraulics and yaw system.
• There is the potential that preventative maintenance can be planned with wind speed forecasts in mind.
• The effects of offshore wind speed on component reliability can be estimated using this model.



References
Besnard, F., & Bertling, L. (2010). An Approach for Condition-Based Maintenance Optimization Applied to Wind Turbine Blades. IEEE Transactions on Sustainable Energy, 1(2), 77–83.
Faulstich, S., Hahn, B., & Tavner, P. J. (2011). Wind turbine downtime and its importance for offshore deployment. Wind Energy, 14(3), 327–337. doi:10.1002/we
Hahn, B. (1997). “ Zeitlicher Zusammenhang von Schadenshäufigkeit und Windgeschwindigkeit .” In FGW-Workshop Einflub der Witterung auf Windenergieanlagen,. Leipzig.
McMillan, D., & Ault, G. W. (2007). Quantification of Condition Monitoring Benefit for offshore wind turbines. Wind Engineering, 31(4).
McMillan, D., & Ault, G. W. (2008). Condition monitoring benefit for onshore wind turbines : sensitivity to operational parameters. IET Renewable Power Generation, 2(1), 60 – 72. doi:10.1049/iet-rpg
Slimacek, V., & Lindqvist, B. H. (2013). Process with covariates and unobserved heterogeneity. In ESREL 2013 (pp. 1229–1233). Amsterdam.
Tavner, P J, Greenwood, D. M., Whittle, M. W. G., Gindele, R., Faulstich, S., & Hahn, B. (2013). Study of weather and location effects on wind turbine, (May 2012), 175–187. doi:10.1002/we
Tavner, P. J., Xiang, J., & Spinato, F. (2007). Reliability analysis for wind turbines. Wind Energy, 10(1), 1–18. doi:10.1002/we.204
Wilkinson, M., Hassan, G. L. G., Vincent, S., Lane, S., Bs, B., Delft, T. Van, & Harman, K. (2012). The Effect of Environmental Parameters on Wind Turbine Reliability. In EWEA 2012.