More recently, however, owners, operators and developers have been looking to the cost of running the projects, and are recognising the value in preventing failure instead of running turbines until they break.
Two components of the cost of operation and maintenance (O&M) of wind turbines are vitally important and need to be minimised: those for scheduled maintenance and for unscheduled maintenance. If component failures lead to unscheduled stoppages, then the additional cost of loss of electricity sales is introduced. That is why considerable efforts are being made to control and forecast such failures.
There are three ways of expressing O&M costs. The simplest way is to assume that the total annual charges represent a percentage of the installed cost, often quoted between 3% and 5%. More detailed assessments demand scrutiny of the components that make up the total charge, which can be expressed as a cost per yearly power output, or per hourly electricity generated, which is used here.
The inexactitude of the science of calculating O&M costs begins with the variety of ingredients that make up the whole in addition to scheduled and unscheduled maintenance. The diagram below shows data taken from an analysis four years ago from sister publication Windstats, showing that the costs of O&M ranged from €15-26/MWh and, while this is an average, there are significant variations both above and below the estimates quoted. The latest data from the International Energy Agency, reporting from 12 different countries, gives a similar range of €7-26/MWh.
The data indicates that operational costs fall with an increase of turbine size, and a report from The Institut fur Solare Energieversorgungstechnik (ISET) suggests that machines in the 800-100kW range have about 15% lower operational costs than those of machines in the 420-490kW range. Lower values can also be expected from large wind farms, simply because overheads can be spread over more machines.
Wind plant operators have little control over most of the elements of O&M, but they can influence both scheduled and unscheduled maintenance costs. There are sophisticated theories as to how best to minimise these but, in essence, a balance needs to be struck between the superficial attractions of carrying out very little maintenance – with low initial costs but potentially high risks of expensive failures – and carrying out too much maintenance, which would be costly and the incremental benefits may be marginal.
An essential tool when planning maintenance strategies is information about the probabilities of component failures. For many years a valuable source of such information was the German Wind Energy Measurement Programme, funded by the German government and run by the ISET. This tracked the performance of around 1,500 wind turbines in Germany for ten years, from 1997 to 2006. It accumulated 15,400 turbine years of operation and created a detailed picture of failure probabilities.
Equally important is information on the turbine downtime, or outage, associated with the failures of particular components, as this is a crucial bearing on the lost revenue. Although the failure characteristics depend on the precise design of machines, and on the length of time they have been in service, the programme produced a valuable database (see chart below).
The results shows that electrical equipment is the most common cause of stoppages, with approximately 5.5 incidents every ten machine-years. These problems are resolved fairly quickly, however, and the turbines are back in action after around 1.5 days.
Gearboxes, with some well-publicised failures, only account for about 1.5 incidents every ten machine — years, according to the data. But when a gearbox fails, the outage time is much longer, at over six days.
Combining the number of failures with the number of days the turbine is out of action provides an estimate of the average loss of productivity over ten years. Failures in gearboxes and electrical systems both account for just under one day a year loss of operation, and the least troublesome component is the hydraulics, taking out less than half a day a year.
A smaller but more recent database from the agricultural commission of Schleswig Holstein, Germany, covers 5,800 turbine years and shows similar failure rates for the various components, but significantly longer outage periods per failure. In the case of gearboxes the average outage period is 14 days, and this no doubt reflects the move towards larger wind turbines with resulting handling difficulties for the major components.
The consequences and costs of dealing with component failures, particularly in offshore wind farms, are critical, as it takes time to arrange for repair vessels to visit the site and deal with the faults. Failure of a critical component, such as a gearbox, can cause damage to other components and so it is important to obtain advance warning of possible problems. That is why there is a developing interest in condition monitoring systems (CMS), with a number of turbine manufacturers and developers offering measuring equipment that can signal possible problems.
The principles of condition monitoring are not new, but it is the interpretation and analysis of the measured data, with increasingly sophisticated computational analysis, that is now coming to the fore. The example, above, shows vibration levels recorded by CMS on a location within the machine, such as the output shaft from the gearbox. Vibration is often a good indicator of the health of the machinery and, with the normal level known, in this case 4 (arbitrary) units, CMS can monitor it daily, or every minute or hour, depending on the likelihood of rapid changes in the health of the component.
If the vibration level gradually drifts upward towards the critical level of 8 units, as above, an alarm is sent to the operator. If that level is well below the danger level, then occasional excursions – as seen at days 88 and 90 – may be acceptable. However, sustained measurements, as seen from day 92 onwards, would not be acceptable and would trigger an alarm.
CMS and the interpretation of results vary in sophistication, but it is a solution that is likely to become increasingly comprehensive given the increase in size of wind turbines. A recent report for the EU-funded UpWind project, which looks at the design of 8-10MW turbines, suggests that CMS should include measurements of: strain, torque, bending and shear; the physical movements of the rotor shaft; electrical quantities that might change if there were electrical faults; and oil quality.
CMS is also able to detect potential problems through causes other than component failure, such as rotor imbalance due to icing and electrical faults on the network to which the wind turbines are connected.
In common with many items of mechanical equipment, wind turbine faults tend to be at their highest immediately after commissioning. The failure rate declines during the middle years but is expected to increase again, particularly towards the end of its useful life, as shown in the diagram, below left. This is a conventional view, but there is little data to verify this and many early Danish machines are still operating satisfactorily after 20 years, as discussed recently in Windstats, so the precept is not universal.
With data from CMS available, scheduled maintenance intervals can be adjusted in order to strike an optimum balance between the cost of maintenance and the cost of unscheduled fault repairs. A considerable amount of research is in progress on fault analysis, condition monitoring techniques and optimised maintenance procedures that should enable O&M costs of the future to be held at modest levels and thus contribute towards the cost-effectiveness of wind turbine technology.
Most authorities expect operating costs, both onshore and offshore, to continue falling as the industry acquires more experience. The Danish Energy Agency, for example, expects onshore costs to fall by about 22% by the decade starting in 2020. Offshore costs are expected to fall more rapidly, and the agency anticipates a drop by as much as 40% over the same period.