Wind turbine maintenance is big business. The global operations and maintenance (O&M) market is set to be worth $19 million by 2020, from the $3 billion it was worth in 2008, according to recent research.
The importance of this market is further emphasised when you consider that up to 75% of expenditure for large offshore wind projects can be attributed to onsite O&M. As a result, wind-farm operators are under intense pressure to collate and analyse performance data that helps reduce costs and improve return on investment.
Installing monitoring technologies on wind turbines is one way to collect the required data to implement strategies that minimise the risk of wind turbine failure and resultant downtime. Additional value can be unlocked from existing data sources such as condition monitoring systems, Scada and inspection records, which often contain unused data filed in siloed systems.
Addressing O&M costs leads to consideration of the best maintenance strategy. These tend to fall into three categories. At one end is a reactive, or "run-to-failure" strategy, which only takes place in the event of failure. Although it is easy to implement, it is very hard to predict or budget for, and is usually expensive.
At the other extreme is a preventive maintenance or calendar-based programme, where components are routinely replaced before failure to minimise repair costs. Again, this can be counter-productive - replacing parts that could run trouble-free for years to come, while also falling foul of the "Waddington effect".
Between those two extremes lies a predictive or condition-based maintenance strategy that focuses on predicting machinery condition and component failures to ensure maintenance can be performed before failures occur. Specific technologies are required in order for this to be effectively carried out.
Assessing the current condition of a turbine is a little easier; vibration condition monitoring and detailed inspections are widely used and well-proven techniques to assess performance. But evaluating the lifespan of a turbine can be a more problematic process. To enable predictive maintenance, turbine monitoring technology needs to deliver predictions of future component failures with a lead time of at least six months, preferably 12. A number of dissimilar technologies, such as vibration condition monitoring, oil monitoring, remaining useful life models, inspection and maintenance data and measured load data need to be obtained from the turbine.
Typically, these datasets are not analysed using a single predictive model, but it is only by analysing them together that the current and future health of the machinery can be truly understood.
An example of this could be using vibration monitoring to predict main bearing failure. Although this can be very challenging, it can deliver significant O&M costs savings. If an operator is able to plan the maintenance events with sufficient lead time, they are often able to mobilise a crane or large vessel during the low wind season and have all the requisite parts and engineers ready. Additional maintenance can be carried out at the same time.
While the challenges posed by turbine failure are always present, it is important for O&M managers to realise that there are technologies available that can allow them to develop and implement a predictive strategy to assess and prevent failure, and that will reduce the cost of running a wind farm.
Dr John Coultate is head of monitoring and O&M consultancy at Romax Technology