United States

United States

A critical gap in the knowledge bank

Google Translate

The wind industry suffers from a persistent gap between predictions of electricity output and actual performance. The average shortfall is 10% in North America. The engineers say they are closing the gap, but investors say not all the way and not soon enough to provide the confidence needed to attract new investment in the current financial climate

One of the most vexing and consistent problems to have plagued the modern wind power industry since its birth is the persistent gap between predictions of electricity output from wind farms and actual production from the turbines.

The problem is particularly acute in North America, where despite the entire fleet's broad distribution of various turbines across different regions and weather patterns, wind farms generate about 10% less on average than pre-construction modelling predictions have said they should. Progress has been made towards closing the gap, say the major wind resource assessment specialists, who are confident they are on the right track.

The wind farm investment sector, however, is not fully convinced. Neither are its members pleased with the shortfall. "It's not that it's dramatically harming the economics, but it's just disappointing," says John Eber at J.P. Morgan Capital Corporation. "And it's disappointing because we've been pushing the industry for a couple years now on these observations and we would hope they could just get better at it." J.P. Morgan Capital Corporation is one of the largest owners of wind plant in the US, having been an equity player in the wind business since 2003. It has ownership stakes amounting to 4200 MW -- about a sixth of US capacity.

No one single smoking gun is to blame for the shortfall, but wind power plant availability -- the percentage of time that wind turbines are available to operate provided the wind is blowing -- is unanimously considered the largest culprit. Unusually poor winds in certain regions compared with the historical norm is the next most cited reason for the shortfall.

In response to the clear discrepancy between predicted and actual production, wind resource assessment companies have been making adjustments and improvements to the modelling formulas they use for predicting wind plant performance in an attempt to understand the gap and close it.

"I think other investors probably have similar observations. There just may be very few that have this size portfolio that we have," says Eber. "We were one of the first institutions that got into making these investments as a financial investor rather than a strategic owner-operator. We've been at it a little longer and a little broader than many of the other institutions, so we have a good eye on this, given the size of our portfolio."

Eber says the uncertainty about what a wind farm is actually going to produce makes it more challenging to bring in new investors -- particularly in the current financial climate -- at a time when the wind industry needs a lot more money to continue growing. J.P. Morgan's wind investment model is to use its long-time experience with wind to bring in other investors and the company has leveraged about one-and-a-half times the sum of its own investment in the wind business from outside sources.

Uneasy investors

"It makes it harder to bring in new investors when you say, well they really don't hit the numbers -- on average they're about ten percent off. It isn't helpful for the industry in the long run in terms of raising the large amounts of capital the industry needs, that we can't be a little better at this," says Eber.

Investors, whether providing equity or debt, could decide to do other things with their money that give more predictable returns. Eric White at New York-based wind forecasting company AWS Truewind says many investors have structured deals that shield them somewhat from fluctuating returns, but that the performance gap brings up issues around the time-value of money. "You stay in until you get the target return, so now instead of getting that money, you get it later than expected and you might have invested in something else in the meantime."

Aside from making investors more wary of investing in wind, White says the performance gap slows the ability of utilities to comply with state law demanding an increasing proportion of renewables power. Green power mandates have been passed in 28 states so far. Depending on the wind power content of a utility's green energy portfolio -- and it is usually a high percentage of wind -- that can mean an unanticipated shortfall in meeting the legal requirements.

"That may mean needing to commission another wind farm to meet the targets, which isn't the end of the world, but it's a ripple down effect if you're not generating what you thought you were," says White. The problem also has a troubling public relations angle, he points out. "There's the whole anti-wind side that says wind plants don't work like advertised. Even if it's marginal and not the end of the world, it still can be picked on."

AWS Truewind and the other major consultancies involved in making wind production estimates for potential investors have been working to improve the accuracy of their projections. Among the market leaders is Garrad Hassan, a British-owned global consultancy. After completing a major global research and validation exercise for its wind production models in mid-2008, Garrad Hassan established there was indeed a gap. The company adjusted some aspects of its approach and its methodology to reflect most recent experience, particularly in North America.

Specifically, Garrad Hassan adjusted its P50 base line representing the primary power prediction figure for the expected mean energy production of a wind plant. The P50, with "P" meaning "probability," represents a 50% probability that the energy production will be higher, or lower, than the prediction. It is the most commonly used estimate across the wind industry of the long-term average power value expected from a given project.

Garrad Hassan's raw global results have shown that production predictions, on average, have been overestimates. At the 2008 European Wind Energy Conference, results were presented from 486 wind plant years throughout Europe where the average annual production has been 94% of pre-construction projections. In North America, a smaller sample reveals that wind farm production on average has been 90% of pre-construction estimates, news reported by Garrad Hassan at the 2008 American Wind Energy Association annual conference.

The long term predictions of probable wind farm production and resulting revenues are used by project developers as the basis for securing equity and debt financing. Equity investors who own the project, often into the long term, are more prepared to take on risk than debt providers for the project loan. While P50 is the primary measure used by equity investors for how a wind plant will perform, debt providers require greater security and commonly use a P90 or even a P95 standard, meaning that there is a 90-95% probability of the prediction being exceeded. Predictions for debt providers have proved to be very accurate, says Garrad Hassan boss Andrew Garrad.

Debt pays for construction and turbine supply loans. "The debt is only interested in making sure it's paid back," says Garrad. "That debt is issued on the P90, which has a high probability of exceedance. Everybody has exceeded that level. There is not one project that has gone into default and all the low ends of the spectrum are way above what they need to satisfy the debt. As far as estimating for debt, we've done an incredible job for something like fifteen or twenty years. We've done a really good job of protecting the lenders."

Eber says a typical loan structure uses debt coverage ratios (expected cash to service debt) of 1.2x, meaning that a 20% margin is required. "So usually a project can operate as low as P95 before there is insufficient cash to cover the scheduled loan payments," says Eber. "The current average ten percent shortfall being observed in the US is equal to about a P80 performance, so debt service would not likely to be impacted until the shortfall grew closer to P95 performance."

The main problem lies with equity investors and the P50 estimates. At seminars in North America, experts from Garrad Hassan and other wind resource assessment firms have been pooling data and research. "It's pretty alarming on the surface so some big soul searching went on," says Clint Johnson of Garrad Hassan, referring to a series of presentations at recent wind resource assessment workshops.

What is to blame?

"I think we've done a lot of work as an industry in the last six to twelve months to understand what's going on. The single biggest factor has to do with wind farm availability," says Johnson, referring to the time for which a wind turbine or a wind farm is not shut down for scheduled or unscheduled maintenance. "In short, availability in North America has not been very good. It's been below expectations at around 93% to 94% versus about 97% in Europe, which is really unfortunate because it means we're leaving a lot of money on the table and it means we're over predicting how much wind farms are producing."

Johnson and others believe that a confluence of factors is leading to lower availability in the United States. This includes the relative rush of building that puts a squeeze on the number of technicians available to keep wind plant running optimally. Turbines typically have lower availability during start-up and, with so many new wind farms coming online in past years in the US, it is bound to skew the availability curve. According to AWS Truewind, poor availability accounts for 3-5% of the total production gap, which it agrees lies in the 6-9% range.

Added to the start-up problems, hundreds of new turbines in America have suffered blade quality failures. The consultancies are not naming names, but there have been much publicised series retrofits of Suzlon and Clipper turbines and less publicised down time for repair work on other makes.

Unusually poor wind seasons play a role too, and this can be worsened in America by the tendency of the US fleet to be clustered in certain geographic areas. Many wind plant were built in West Texas over the past few years and 2005 and 2007 were especially low wind speed years in this region, with 2007 the worst in 15 years. Garrad believes this played a big role in the production gap.

But J.P. Morgan's Eber does not buy the case that poor wind years could have this much of an effect on the P50 because his company's investments are so broadly dispersed between various regions that experience different weather patterns. "Whatever the cause of the fluctuations, if you have forty-three different wind farms spread from Maine to Hawaii, with eleven different sponsors, that you've been investing in for over five years, you would think we would be much closer to the mean, the P50, than where we are," says Eber.

Methodology improved

Other reasons for the production gap are each given 1% of the blame by AWS Truewind (table). These include resource assessment biases, often caused when wind speed measuring is conducted at the windiest corner of a site rather than at a location seeing average winds for the area. Older assessments also failed to properly understand complex terrain modelling issues and wake effects in wind farms from turbines disturbing the wind flow for other machines. Unexpected problems during construction, such as crews encountering stubborn bedrock, can mean the wind farm layout being changed to a less than optimal configuration for the prevailing winds. Lastly, sub-optimal operation, such as an improperly calibrated turbine anemometer or other faults preventing a turbine correctly yawing into the wind, is also a factor. Better understanding of all of these issues is leading to improvements.

When it comes to causes and solutions, Garrad says his firm and the others disagree only on minor details. He breaks the gap down between three factors contributing about one-third each to the 10% gap: the wind, wind plant availability, and methodology of the measurement and prediction work (box).

Most seem to agree the gap being seen today is partly a vestige of less refined approaches to measurement and prediction years ago compared to what is de rigueur today. "A little bit of where we're at is not as bleak as it looks. Most of us have been making changes as we go, trying to improve our methods," says White. "The bulk of the projects we're evaluating...those were designed in 2000 and that's not how we design them today. We've already corrected several percent of the issue at least," says White.

In an attempt to understand the performance gap, AWS Truewind has been conducting a series of studies of past cases that re-evaluate older pre-construction wind power plant data under newer, stricter modelling. The improvements in the newer modelling are clear. "Looking back at that data, I don't get too worked up over how bad I did then, because I wouldn't accept that data today," muses White.

Most of the wind measurement community believes that the gap has largely been solved through the incremental changes made to prediction methodology, but Eber is still not convinced. "They're going in the right direction. But my people cannot find adjustments that appear to be on the scale of the shortfall." Eber says J.P. Morgan has a number of engineers who specialise in banking who have been studying the changes being made to P50 by the independent engineers.

"They are adjusting maybe a third of the shortfall we are observing, so time will tell, but for now, we have not been able to confirm that they really are making the full adjustments," says Eber. "I think part of the problem is they don't really know why the entire shortfall is being observed...and our sense is that until they can identify it, they don't want to start incorporating it into their analysis."

Benjamin Bell, who heads Garrad Hassan in the US, confirms that only the issues and changes that are fully understood and quantified are those being implemented into the modelling protocol. "What we're trying not to do at Garrad Hassan is to give arbitrary haircuts that are not based on science," says Bell. "That's the knee-jerk reaction and the easy thing to do, but actually trying to figure out what's going on and improve the models and get a better understanding of it, that's two different approaches."

Meantime, Eber says his division is looking "a lot harder" at the P50 cases today than in the past and is starting to make adjustments on its own. "I know the engineers haven't been denying what we're observing -- and we appreciate that. We just want to keep the message out there that it seems like they are still not there in terms of making these adjustments because it's in the best interests of the industry that they get this right -- and it will make it a lot easier to keep the money coming in if we can make these P50 cases more of a reality."

Garrad is more optimistic. "I think the whole industry is tightening up. Ironically, part of this credit crunch will be to make sure that everything everybody does is done better. I think it's been very cathartic to address the problem and everyone is reassured to know that corrective action is being taken on three fronts: better data, better calculations and more attention paid to proper operations."

Have you registered with us yet?

Register now to enjoy more articles
and free email bulletins.

Sign up now
Already registered?
Sign in