Princeton University released a report late last year which found that global wind speeds are increasing, which could therefore increase the output of wind farms. This naturally places increased importance on turbine uptime and predictive maintenance, since increased wind speeds can only be capitalised on if wind farms are operating well in the first place. This in turn means taking advantage of the latest technological advances in Operations & Maintenance so that wind turbines, no matter how old they are, can be optimised for maximum efficiency.
With this in mind, REM talked to Bruce Hall, CEO of ONYX InSight, and Evgenia Golysheva, Head of Engineering at ONYX InSight, about ongoin meteorological and technological issues and trends in wind power O&M.
Bruce Hall: The wind industry is relatively young and, as with any industry, it goes through phases. Phase 1 is about getting the technology right. Phase 2 is about making that technology ever more efficient. In terms of ONYX InSight, we were in the wind industry at the very beginning, helping the manufacturers to make bigger wind turbines. As the wind industry expanded, there were lots of technological issues to be addressed, because people were moving to 1 MW, then 2 MW, and so it goes on – much bigger turbines. There were lots of technical issues in those earlier days. That’s where ONYX InSight’s business developed. That was primarily solving a multitude of problems that were in there at the design stage, that needed to be ironed out.
A lot of the technical discussion, a lot of the technical improvements were being made to the design and the operation of the wind turbines. You can kind of see that in terms of the number of wind turbines that were developed, the size of the wind turbines that have been developed over the past 20 years, starting off at sub-megawatt class to the latest – the 12 MW HaliadeG, which is scheduled for Dogger Bank – as an example, up and running in Holland right now as a prototype. So the first phase of the wind industry has been the technological challenge and solving that. Increasingly though, our clients have become more mature and are moving to the efficiency dimension, as in, “Hey, I can buy a turbine off the shelf now and I trust it that it’s going to work as the design features have been ironed out” or I’ve got ways of mitigating any other problems. So increasingly we do a lot of work with the clients where we’re looking more broadly at making their sites, their operations, more efficient. That could be focused on maintenance, or it could be looking at the structure of the wind farms, wake effects etc etc. So we can stay strategic or we can be quite tactical.
Evgenia: I think one point to make, that Bruce said, is that in the past wind farm owners tend to react to problems with individual turbines. Now the technology is more mature, with lots of advanced data management available, the task is to start operating wind farms as an energy producing plant, where they don’t deal with individual wind turbine issues but optimise the power production from the whole plant. It is especially crucial for offshore, where access is limited by wind speed as well as the wave height, especially between October and February, the weather data, 30 percent of the overall length of time, which means that for several months, you can’t access the turbines. That makes it very important to install predictive analytics and utilise all of the data and arrange all your operations in such a way that you don’t have any unexpected surprises during this time period. You’re not faced with fairly costly downtime if one of the turbines closes down because of a failure of a gearbox or blade. This becomes even more important as the turbines become larger. The cost of lost production due to downtime from any single turbine becomes greater. In addition, there are fewer turbines at new wind farms that utilise larger turbines, so each turbine will contribute to greater proportion of total output. That again is where predictive analytics can help a lot.
I think what we’re trying to do, I think we’re trying to make owners/operators think in terms of additional value from their assets. In the past, many owners have chosen full service agreement with OEM to maintain their assets. This option typically reduces the risk from technology, but it comes at a price premium. As owners become more confident with turbine technology, digitalisation of their operations enables them to select alternative, often more cost effective service options. This is especially true for utilities and other industry players who have committed to wind industry long-term and started investing in internal technology knowledge. The benefits of digitalisation cover the whole range of wind farm operations from production forecasting and streamlining unscheduled maintenance to scheduled maintenance optimisation.
Bruce: So the issue of maturing if you go into more established, if you go, say, into oil and gas or more traditional power stations, it’s interesting to consider a wind farm as a power station now, to borrow Evgenia’s words, that’s new terminology. A lot of these maintenance practices are more established, in more conventional power plants, let’s call them that, are more mature. You’ve got the integration of different datasets etc whereas they are considered separately, there’s the balance between how do I do my maintenance in the early years or give it to the OEM, which is safe and secure, but ultimately more expensive, or do I bring it in house. Do I do it myself, take a lot of risk, but also take a lot of the upside in terms of savings. Increasingly, operators are choosing to do that as the market is chasing a lower levelised cost of electricity (LCOE). At either end of the scale, those new turbines that are being commissioned now, the debate is about chasing a lower cost. Is it using the OEM, is it doing it themselves, but either way, doing it better than we’ve done before. Integrating a data set, is it application of new techniques like machine learning or AI, and at the very other end of the spectrum, which is some work that Evgenia’s team are doing, there’s an offering also for those people who have got that 20-year old farm. It is smaller turbines, what we are offering to that group of customers, not just us by the way, there are other manufacturers, but we tend to think we’ve got something a little bit different to share with them, if you can squeeze another 2,3,4, years out of, with planning permission of course, out of those wind turbines, then that changes the payback significantly. Depending where our customers are on that spectrum in the wind industry in terms of whether it’s new or whether it’s a 20-year old farm, doesn’t matter, there’s something that can be done.
Evgenia: Another point is that with reduced sensor costs and advanced analytics wind farm owners are faced with large amount of various data sources. They are also bombarded by multitude of analytics companies pitching “big data” and “AI”. Many of our customers find it overwhelming. Our recommendation is to define an efficient operations and maintenance process first, then use the data to support decision making during at each stage of such process. This will allow to unify decision making across the whole fleet which increasingly becoming more global for many of our customers. Data-enabled O&M process will also reduce the pressure on experienced engineering resources which are becoming more scares with rapid industry growth. So it’s not about buying a platform or jumping on the bandwagon of big data, it’s about using engineering knowledge and then making the data work for you.
Bruce: A common theme when we’re at conference with our customers is, they talk about data overload so some of the benefits will come from using existing data, better, not using new data but using existing data better. The data can fit in silos and the managers can sometimes struggle to integrate the datasets to get a better view on how to improve their operations. That’s down to maturity. You can go to other industries, they’ve had more time to seek out those inefficiencies, so also to seek out the efficiencies. By its very nature, it’s a very new industry. This is a message of optimism, not of pessimism. There is a lot to go out. A case study, one of ours that is a favourite at the moment, vibration and oil. So, right now, all our operators are doing all our analytics and they’re doing vibration. They’re complementary, but largely, the datasets are kept separate. This is an analytic technique, separate, highly related, but they’re done separately. Like with any other industry, the benefits will come when you can start integrating those datasets, so bear with me as I carry on with this example. The oil in a wind turbine, or in any machine, is like the blood. By doing the analytics on the oil, you can tell what’s going on. It gives you an early indication. You can pick up wear in the metal, you can pick up traces in the oil, you can pick up contamination, and, and, and. Changes in the oil, like changes in human blood, has an effect upon the machine health, which is also picked up by the vibration. When you start integrating those two datasets, you get a much better picture, so one plus one equals two and a bit at that point. The industry hasn’t been brilliant at doing that. That’s just one example, there are many others.
Evgenia: It’s exactly the same with SCADA data. Typically one department will look at SCADA data and then separate department will look at vibration. There is no long-term consistency of trying to analyse the data jointly to determine if the issue could have been detected earlier or more accurate by combining available data sources. The long-term outlook is often lacking.
Another example is End of Warranty (EoW) process. Very few people think ahead and find themselves in a situation where OEM rejecting all of their EoW claims stating “normal wear and tear” and owner has no evidence to counteract this. Owners need to start thinking about post warranty operations earlier and collect supporting data at least 12 months before EoW. Then when it comes to EoW there is enough data to have a productive conversation with OEM and if there are any issues with warranty, all of this is supported by data and can be sorted out quite quickly. So not only data enables successful EoW campaign, it also builds a significant technology knowledge. This knowledge gives the owners flexibility for the post warranty period to select the best O&M contract model. In a very dynamic service market such flexibility is very important to continue to operate assets in cost effectively while managing the risk.
Bruce: There’s a very important side issue. It’s about the importance of data and who owns it. Ownership of the data and/or access to the data has a fundamental bearing on being able to do maintenance properly. That has been, and continues to be, a battleground. That is a key issue. As Evgenia pointed out, it is only by having the data in a format that can be used that allows you to make different choices. If not, you can be shut out. You can have the machine and plenty of storage with it, but what you will find is that you can’t operate it properly, fully, and totally efficiently until you’ve got access to the data.
It’s a common refrain, that typically what we find is that a lot of our customers don’t have access to the data. The data can be restricted by certain OEMs. Essentially, you’ve bought a turbine, but in order to be able to operate it efficiently, you need to have access to the majority of the data, the SCADA data, the vibration data. In some cases, that data is restricted. Even though you may have bought the turbine, may have operated it for four and a half years, you don’t necessarily have access to all of that data. Without that rearward-looking view, you can’t optimise maintenance. Data is like good wine – it becomes more valuable the older it gets. The business model behind that is quite simple, because as OEMs want to move towards making money on the after market, as in you build a turbine, you sell it once, but it lasts 20-25 years, you’re interested in the service, in the warranty that comes after it. That’s where the battleground is right now. There, data and data access is being used as a control tool.
What effect does all of this have on reducing greenhouse gas emissions, reducing costs and what kind of financial savings can be made?
Evgenia: The exact savings that can be made from data and data usage depends on where you are starting from. With some sites, it could be significant. On average 16 percent of operations and maintenance costs savings could be easily met by using predictive technology. It’s not just about saving costs, it is also about maximising production per unit of cost. Saving costs is important, but ultimately it is about achieving maximum production. In terms of production, we can increase the availability by, on average, 1.3 percent.
Bruce: A little bit of further detail behind that, behind Evgenia’s point, is that the work we do with analytics doesn’t necessarily stop a failure happening, sometimes we can see it if a fault develops early, but what we’re doing is helping our customers, the operators, be in control of their maintenance schedule and their machines. A lot of the work we’re finishing now, the operators now, as we move into the windy season, those wind turbines need to be up and ready for full production all the way through to February-March. Evgenia’s been doing a lot of work to set that up, and that is finishing now. The idea is that if you’ve got a faulty turbine and you’re the operator, you need to know in June whether that is going to fail in November-December and whether you need to fix it now, or whether you’ve got enough life left in that machine such that you can take it through to June 2020. The costs of unscheduled maintenance are significant, such as bringing a crane in etc etc. For example, with my hypothetical turbine, if it fails now it may be that the turbine cannot be fixed until March-April.
That essentially takes it out of operation for a few months and presumably there are costs associated with that....
Evgenia: We’ve done the estimates, so for a 3.6 MW turbine, which is not the largest, if it is down for three months, we’re losing, on average £340,000 in revenue.
Bruce: The other side of the argument is if you can do a full health assessment, that is what is done by the team here, you can look at your fleet of 100 turbines and you can isolate those where you need to do maintenance this year to get you through the windy season and/or you can make some judgements on the middle ground of turbines where you say “look, based on the predictive analytics that we’ve got, the data we’ve got, and the history we’ve got, the historical view, we know those turbines, the remaining 20-30 turbines that are probably in this amber warning area, they can get through the winter period. You can do their maintenance and you can batch it altogether in the calm days of May, June and July, when you’ve got the crane availability, when you’ve got the crews, you can tender that out now ready for June of next year. It is not necessarily that we would prevent the failure, because machines have a finite life, you can try and extend it but you do so within the control parameters of the customer.
Evgenia: Some of gearbox failures can be repaired uptower without deployment of large capital equipment. Now if these defects are not detected in time, they can develop and they can cause secondary damage in the gearbox, so the whole gearbox will need to be replaced. Now, to replace the gearbox, owner needs to call a jackup and the cost of this vessel is £80,000 per day. The owner has to charter from somewhere else which might take three days just to get the jackup vessel in place, so even before the work has started, significant cost is uncured. Now, if owner is doing it every time something goes wrong, the wind farm could go out of business. The whole remit of what we do is to give advanced warning to our customers. Some of these failures can be detected 12 to 18 months in advance. We’re giving them a long-term reliability outlook. We’re also giving them a prognosis on the remaining useful life of their equipment. With this information one maintenance campaign can be scheduled in June and all impending failures can be rectified in one vessel mobilisation resulting in huge savings in operations and maintenance cost.
So are we talking mostly about onshore wind or offshore wind or both?
Evgenia: With preventative maintenance, the same is true for onshore or offshore, the cost is significantly larger with offshore, because the access is limited by wind speed and wave heights. Onshore access to remote sites in Scotland or wales can be very difficult in winter making predictive analysis essential for efficient wind farm operation.
Bruce: We have customers who basically lock everything down, onshore I am talking about, in October and say come back in March.
Evgenia: Offshore is more attractive for optimising the processes because the turbines are much larger so the cost of lost opportunity during downtime is significant, also and the cost of jackup vessels is high.. But the principle of analytics enabling efficiency is exactly the same onshore and offshore.
Bruce: When you get to onshore I think that there’s an additional argument because the offshore fleets, generally speaking, are bigger and newer. Onshore, a lot of the debate is about, same issue, efficiency, but also life extension. That 20-year farm operating for another 2-4 years.
I am thinking that is even more important in places in the UK because of the current government’s de facto ban on new wind farms, therefore you have to do as much as you can to make the existing wind farms much more efficient?
Evgenia: Exactly. And also because much more information is available from older wind farms, customers can get much more confidence behind their economic case.
Bruce: So we’re doing a lot of work at that end of the spectrum with customers that have older, or even smaller, can they run for a little bit longer. Assuming they’ve got the planning permission. The repowering is completely different. It’s a change again. That takes us to your question about what this means for greenhouse gas emissions. It all adds back to what is the Levelised Cost of Energy (LCOE). While there are inefficiencies in the system, maintenance inefficiencies through the system, so, put funding to one side, all the issues around planning and so forth to one side, but with the assets that are in existence, the push is to drive down the LCOE to levels that can compete with gas and coal. We know the figures, we know the comparisons, because if the costs remain high, the only winner is likely to be fossil fuels. Wind is showing that it can compete.
Evgenia: The other point to make here is that the biggest reduction in LCOE was so far achieved by increasing turbine size to around 12 MW. If you make larger turbines, you essentially make less of them. So then it makes less sense for manufacturers to expand their plants which will limit further size increase. This means that further efficiencies will have to lie in using data and analytics to unlock assets’ value. At the moment it almost feels like installing larger turbines is easier than driving Operations and Maintenance process efficiencies. A similar reduction in LCOE can be achieved through existing technology by making it smarter, but it takes more effort and discipline, it takes a bit longer. The potential of unlocking the extra value from assets by analytics is similar to that gained by purchasing larger turbines.
Bruce: There’s an awful lot to go at. We think it’s very interesting. We can drive down costs, and by driving down the cost of maintenance, it makes renewable electricity cheaper, which means lower carbon emissions.
ONYX InSight are a leading provider of predictive maintenance in wind energy. ONYX are pioneers in technical innovations in predictive maintenance and were the first to introduce MEMS sensors to the market for this purpose. The company supports wind farm operators by collecting, monitoring and analysing data in real time, thereby enabling fault prediction in order to make tailored maintenance recommendations. The company also inspects assets at all life-cycle stages to inform improvements to asset management, quality assurance and problem resolution.
For additional information: