By Richard Stubbe, BloombergNEF. This article first appeared on the Bloomberg Terminal.
The math of wind turbines creates a sizable target for the companies that keep the rotors turning.
Operating expenses account for about 60% of the cost of running a wind farm, and unexpected breakdowns account for about 60% of operating expenses, said Bruce Hall, CEO of Onyx InSight, which monitors more than 5,000 turbines around the world. For Onyx and other businesses in predictive maintenance, reducing those breakdowns makes wind power ever cheaper to produce. Predictive maintenance uses data from sensors to detect vibrations and other indications that the turbine may fail, enabling operators to schedule downtime rather than deal with costly surprises.
“The reason it’s so expensive is because it’s unscheduled,” Hall said in a phone interview with BloombergNEF in late August. “You need to get a replacement part urgently, which means you won’t have your regular bargaining power with the manufacturer. You’ll have a loss of production, possibly in a high-wind season.”
The promise of reducing the length and frequency of those breakdowns, as well as perhaps getting extra years of service from older turbines, is a huge opportunity for predictive-maintenance companies including Onyx, based in Nottingham, U.K. The company, an independent provider of software and predictive maintenance, also has offices in the U.S. and Asia.
Hall answered questions from BNEF in a phone interview in late August.
Q: Most wind turbines now are sold with a maintenance contract as part of the purchase. How do you make the sale of your product to wind operators?
A: We’re deeply embedded in and around the wind industry in and around predictive maintenance. We do software and we do engineering, and we’re independent of an original equipment manufacturer. Wind owners come to us when they want a maintenance offering that is independent of an OEM and therefore allows them to operate multibrand fleets.
Q: When in a turbine’s life cycle are you getting into the business with asset owners?
A: We would go in with a wind-farm operator for a new asset if they wanted shadow monitoring. Some owners want an independent set of eyes on the asset. They may go for a full-wrap service agreement with Siemens or Vestas, for example, and a lot of what we do, particularly in offshore, is shadow monitoring. We will also monitor the asset and feed that back to the owner-operator. It’s kind of a competition of who spots the faults the fastest.
But the aftermarket is also very, very big, and it includes turbines that weren’t fitted with condition monitoring systems. We upgrade older turbines that may have an older version of a CMS and didn’t have access to the new technologies.
We focus on the aftermarket because an owner might want to put these systems in. So we do shadow monitoring on new turbines and we go for full monitoring on older turbines. When older turbines come out of warranty, we take the data off and work to set up a maintenance schedule.
Q: How fast is the business changing?
A: Very quickly. The prices have come down. The turbines are getting bigger and more complicated. All of those turbines are coming fitted with CMS systems, and the data feeds off of them are much greater than they were. That is one market that we’re addressing: We work with the OEMs and the asset operators.
There’s also a huge market of smaller turbines that don’t have any CMS at all, and those are operating right here, right now. There’s a huge opportunity to make those older, smaller turbines that at the time could never have afforded a proper analytics system more efficient.
Asset owners are saying they need to make those assets run for longer. They are now starting to apply more sophisticated maintenance practices that weren’t feasible three or four years before.
Q: How is the addition of merchant risk changing the market and your business?
A: The wind industry started with pioneers and the government provided help for the industry, subsidies to encourage people to invest. The primary risk was technological. As the market developed, those subsidies gradually started to disappear and competitors started to show up. Now the governments are fast-tracking subsidies away from the industry.
As we consider wind auctions and the way wind assets will be awarded in the future, you have to go after the top line and the bottom line. On the top line, owners and operators are dealing with that by signing power-purchase agreements. But they also help the bottom line by improving O&M practices. Predictive maintenance can help operators fulfill PPAs by reducing their downtime.
Q: How much can predictive maintenance do to reduce those costs?
A: We can’t necessarily stop a failure. We can stop a failure from becoming a catastrophic failure, or a catastrophic failure from getting even worse. We can look at how the machine is degrading over time and give our customers greater clarity on how their assets are operating such that no unplanned events, unscheduled events, can take place. This lets them schedule and professionalize their maintenance strategies.
Q: How is the maintenance industry developing? What’s the next stage?
A: We see three main areas that will bring about more reductions in operating costs. The first is the abundance of data coming off those turbines, to some extent too much data. There are so many sensors on the newer turbines, and they’re being put on older turbines in retrofits.
Second, you’ve got better analytic systems available, whether that’s machine learning or artificial intelligence. The issue for us is that the data is kept in silos rather than readily mapped on top of each other to find trends or cluster.
And third, we have much faster and more powerful computing. Combining those three features will combine to enable O&M costs to come down even faster.
There are some challenges there. Not all of this data is readily accessible.
Q: What’s the issue with data access?
A: It’s a big frustration in the industry that if you buy a wind turbine and it produces data, sometimes the operator is not allowed access to the data. There’s a debate about who owns the data and who gets access to it.
Q: Who does?
A: Here’s an analogy from the automotive industry: When the new electronic management systems came out 20 years ago, you had to go to the dealership to have your car serviced because they were the only ones who could read what was going on in the engine management system. They owned the data and they owned the data access. The authorities stepped in and changed the format to open-source, so you could take the car to any garage and get it serviced. It brought the cost of motoring down.
That analogy also works for the wind industry. We’ve had plenty of discussions with lawyers about who owns the data. And I don’t think it’s to do with ownership but with access. We believe the owner and operator have the right to use all the data to drive costs down.
The issue is also about mapping the datasets on top of each other. For example, the vibration data is in one silo, and the oil analysis data is in another. The two are related. If we as predictive maintenance are picking up vibration in the machine, it means there’s damage to a bearing or a gear, and there will probably be debris in the oil. The datasets are separate, and we’re not able to integrate them.
We can give our customers much more accurate views of how their assets are operating if we can integrate the data. Combine that with edge computing and cloud computing, and the opportunities are huge.
Q: Is there an instance where your information has changed turbine design?
A: Yes, there have been, not just ours. Bearings will be upgraded, or surface materials or loads. The drivetrain continues to improve, and the designs continue to get more complicated. Right now, we and the OEMs are looking at pitch bearings. A number of operators are starting to see higher failure rates in pitch bearings.
This comes back to merchant risk. If I have a 20-year-old turbine, can I operate it for 22 years or 25? And that fundamentally affects the ROI models or allows operators to rewrite them, just by taking datasets and applying them to other aspects of the turbine.
Those centers can also pick up tower vibration. Then we’re asked whether we can start building models for fatigue loads on the tower that may allow the turbine to operate for 25 years. There’s plenty more to go at.
Q: The wave of replacing the first generation of turbines has begun. What does that mean for you?
A: We can make a sensor package that is much more affordable than three or four years ago, which means it can now be applied to older, smaller turbines.
We have an investor who’s buying up a lot of wind farms with older, smaller turbines. Their strategy is to buy the farms with older assets and run them for longer by upgrading their O&M practices. That would not have been possible four or five years ago. They may then be able to replace them with much more efficient, larger models.