How certainty means cash in the wind sector

There is a good reason why many wind farm projects underperform financially. And there is a good way to prevent it from happening. By Jason Deign, European Correspondent

Trust the UK tabloid Daily Mail to give a balanced view of why many wind farms in Britain are failing to meet power output expectations.

“It’s because the developers grossly exaggerate potential,” the paper quotes environmental consultant Michael Jefferson as saying in a March 2010 exposé highlighting how 20 UK wind farms were operating at less than 20% of capacity.

The story accuses developers of systematically building in areas where there is not enough wind resource. At the end of it, though, Nick Medic of RenewableUK, the British wind industry association, makes a valid point: “If it’s not windy, it’s not profitable, so why would you build? No business is going to build something that is not profitable.”

Indeed. This is a major concern for the wind farm industry, because poor wind farm performance is not just a feature of the UK. It is also found in markets from the USA to Brazil. And that means a lot of investors are losing a lot of money.

Why do wind farms underperform so often? Wind is naturally variable, so it is only to be expected that a given proportion of projects will fail to operate at full capacity, either for a period of time or even the entire lifespan of the development. However, if the problem were just down to wind variability then a roughly equal proportion of projects ought to be expected to over-perform, and this does not seem to be the case. The bias

is only in one direction.

Wind farm developers are obviously doing something wrong, then.

Last year a John Hopkins University fluid mechanics and turbulence expert called Charles Meneveau suggested turbine spacing might be the issue. He advised locating turbines 15 rotor diameters apart, compared to around seven presently. Reducing the density of turbines might improve the output of each machine, but will clearly have an impact on the total amount of power you can get from a defined size of wind farm.

And that might not be the only factor leading to underperformance. There is at least one other potential candidate, and it could be a lot easier to deal with, because it is all in the mind of the developer. To illustrate it, Matthew Hendrickson, senior director of energy assessment at 3TIER, a global leader in renewable energy risk analysis, asks you to consider five potential wind plants that you could invest US$200 million in.

If each project has a similar price in terms of capital and operational expenditure, and wind resource is the only apparent difference, then logic dictates you will select the one that has the highest net present value.

But now consider that these projects actually all have the same net capacity factor (NCF), and the differences in apparent output are actually the result of the underlying uncertainty in your wind resource assessments. If you picked the project with the highest apparent output then you have essentially picked the project with the most optimistic wind resource assessment and the one that, therefore, is most likely to underperform.

“Examining the historic bias of the industry, it is possible that the assessment process is more accurate than it receives credit for, only that the industry tends to build projects with the error on the side of underperformance,” Hendrickson says.

One way of dealing with selection bias, or indeed any bias, is to try to eliminate it by counterbalancing your calculations in the opposite direction. And, indeed, wind project investors and developers now traditionally give their figures a ‘haircut’ to counteract uncertainties.

Fitch, for example, states in its Global Infrastructure & Project Finance Rating Criteria for Onshore Wind Farms Debt Instruments that it “typically applies a 5% haircut to the results of the wind energy assessment for most greenfield projects.”

There are a couple of drawbacks to the haircut approach, however. The first is that although the haircut levels can be improved through on-the-ground feedback on performance, this information takes three or four years to trickle through, so it is usually relates to older turbine technology.

And the second is that it does not inspire much confidence in the professionalism of the industry when investors have to routinely lop chunks off the value of business plans in order to feel comfortable about putting money into them.

For this reason, Hendrickson says: “The challenge is on us to not only demonstrate we have a grasp on the sources of uncertainty, but to systematically reduce them through guidance and the sophisticated use of observations and advanced models that begin years before the project is built.” Although it is not the only determinant of project locations, he adds: “In the typical model the wind resource is the most sensitive driver of financial return.”

The extent to which it is important to estimate wind resource accurately is dramatically underscored if you model the profitability of net present value returns based on different degrees of NCF uncertainty. For a typical 100MW project, if there is an NCF uncertainty of 10%, which is not unusual, you have a 5% probability that the project will lose $6.1 million, according to Hendrickson’s calculations.

But because of the influence of selection bias the potential loss could be much greater. “Even a modest consideration of the NCF in selection decision yields bias of 4.8% and a further reduction of the P95 by $1.7 million,” Hendrickson says. At 6% NCF uncertainty, however, the P95 value equates to a gain of $4.8 million relative to expectations. “This is a $10.9 million improvement over the 10% case,” Hendrickson points out. And selection bias is minimised.

In fact, Hendrickson estimates that in a 100MW project each 1% reduction in uncertainty improves the P95 downside case by $2.7 million and improves the P50 selection bias by $1 million.

Gaining such reductions is not difficult. It starts with collecting high quality observations. Additionally, companies like 3TIER can combine those observations with advanced numerical weather prediction models to greatly reduce temporal and spatial uncertainty in wind resource estimation, helping to shave off those valuable percentage points.

So maybe it is time the industry developed a new kind of bias: for more accurate modelling.

There is a good reason why many wind farm projects underperform financially. And there is a good way to prevent it from happening. By Jason Deign, European Correspondent

Trust the UK tabloid Daily Mail to give a balanced view of why many wind farms in Britain are failing to meet power output expectations.

“It’s because the developers grossly exaggerate potential,” the paper quotes environmental consultant Michael Jefferson as saying in a March 2010 exposé highlighting how 20 UK wind farms were operating at less than 20% of capacity.

The story accuses developers of systematically building in areas where there is not enough wind resource. At the end of it, though, Nick Medic of RenewableUK, the British wind industry association, makes a valid point: “If it’s not windy, it’s not profitable, so why would you build? No business is going to build something that is not profitable.”

Indeed. This is a major concern for the wind farm industry, because poor wind farm performance is not just a feature of the UK. It is also found in markets from the USA to Brazil. And that means a lot of investors are losing a lot of money.

Why do wind farms underperform so often? Wind is naturally variable, so it is only to be expected that a given proportion of projects will fail to operate at full capacity, either for a period of time or even the entire lifespan of the development. However, if the problem were just down to wind variability then a roughly equal proportion of projects ought to be expected to over-perform, and this does not seem to be the case. The bias

is only in one direction.

Wind farm developers are obviously doing something wrong, then.

Last year a John Hopkins University fluid mechanics and turbulence expert called Charles Meneveau suggested turbine spacing might be the issue. He advised locating turbines 15 rotor diameters apart, compared to around seven presently. Reducing the density of turbines might improve the output of each machine, but will clearly have an impact on the total amount of power you can get from a defined size of wind farm.

And that might not be the only factor leading to underperformance. There is at least one other potential candidate, and it could be a lot easier to deal with, because it is all in the mind of the developer. To illustrate it, Matthew Hendrickson, senior director of energy assessment at 3TIER, a global leader in renewable energy risk analysis, asks you to consider five potential wind plants that you could invest US$200 million in.

If each project has a similar price in terms of capital and operational expenditure, and wind resource is the only apparent difference, then logic dictates you will select the one that has the highest net present value.

But now consider that these projects actually all have the same net capacity factor (NCF), and the differences in apparent output are actually the result of the underlying uncertainty in your wind resource assessments. If you picked the project with the highest apparent output then you have essentially picked the project with the most optimistic wind resource assessment and the one that, therefore, is most likely to underperform.

“Examining the historic bias of the industry, it is possible that the assessment process is more accurate than it receives credit for, only that the industry tends to build projects with the error on the side of underperformance,” Hendrickson says.

One way of dealing with selection bias, or indeed any bias, is to try to eliminate it by counterbalancing your calculations in the opposite direction. And, indeed, wind project investors and developers now traditionally give their figures a ‘haircut’ to counteract uncertainties.

Fitch, for example, states in its Global Infrastructure & Project Finance Rating Criteria for Onshore Wind Farms Debt Instruments that it “typically applies a 5% haircut to the results of the wind energy assessment for most greenfield projects.”

There are a couple of drawbacks to the haircut approach, however. The first is that although the haircut levels can be improved through on-the-ground feedback on performance, this information takes three or four years to trickle through, so it is usually relates to older turbine technology.

And the second is that it does not inspire much confidence in the professionalism of the industry when investors have to routinely lop chunks off the value of business plans in order to feel comfortable about putting money into them.

For this reason, Hendrickson says: “The challenge is on us to not only demonstrate we have a grasp on the sources of uncertainty, but to systematically reduce them through guidance and the sophisticated use of observations and advanced models that begin years before the project is built.” Although it is not the only determinant of project locations, he adds: “In the typical model the wind resource is the most sensitive driver of financial return.”

The extent to which it is important to estimate wind resource accurately is dramatically underscored if you model the profitability of net present value returns based on different degrees of NCF uncertainty. For a typical 100MW project, if there is an NCF uncertainty of 10%, which is not unusual, you have a 5% probability that the project will lose $6.1 million, according to Hendrickson’s calculations.

But because of the influence of selection bias the potential loss could be much greater. “Even a modest consideration of the NCF in selection decision yields bias of 4.8% and a further reduction of the P95 by $1.7 million,” Hendrickson says. At 6% NCF uncertainty, however, the P95 value equates to a gain of $4.8 million relative to expectations. “This is a $10.9 million improvement over the 10% case,” Hendrickson points out. And selection bias is minimised.

In fact, Hendrickson estimates that in a 100MW project each 1% reduction in uncertainty improves the P95 downside case by $2.7 million and improves the P50 selection bias by $1 million.

Gaining such reductions is not difficult. It starts with collecting high quality observations. Additionally, companies like 3TIER can combine those observations with advanced numerical weather prediction models to greatly reduce temporal and spatial uncertainty in wind resource estimation, helping to shave off those valuable percentage points.

So maybe it is time the industry developed a new kind of bias: for more accurate modelling.

There is a good reason why many wind farm projects underperform financially. And there is a good way to prevent it from happening. By Jason Deign, European Correspondent

Trust the UK tabloid Daily Mail to give a balanced view of why many wind farms in Britain are failing to meet power output expectations.

“It’s because the developers grossly exaggerate potential,” the paper quotes environmental consultant Michael Jefferson as saying in a March 2010 exposé highlighting how 20 UK wind farms were operating at less than 20% of capacity.

The story accuses developers of systematically building in areas where there is not enough wind resource. At the end of it, though, Nick Medic of RenewableUK, the British wind industry association, makes a valid point: “If it’s not windy, it’s not profitable, so why would you build? No business is going to build something that is not profitable.”

Indeed. This is a major concern for the wind farm industry, because poor wind farm performance is not just a feature of the UK. It is also found in markets from the USA to Brazil. And that means a lot of investors are losing a lot of money.

Why do wind farms underperform so often? Wind is naturally variable, so it is only to be expected that a given proportion of projects will fail to operate at full capacity, either for a period of time or even the entire lifespan of the development. However, if the problem were just down to wind variability then a roughly equal proportion of projects ought to be expected to over-perform, and this does not seem to be the case. The bias

is only in one direction.

Wind farm developers are obviously doing something wrong, then.

Last year a John Hopkins University fluid mechanics and turbulence expert called Charles Meneveau suggested turbine spacing might be the issue. He advised locating turbines 15 rotor diameters apart, compared to around seven presently. Reducing the density of turbines might improve the output of each machine, but will clearly have an impact on the total amount of power you can get from a defined size of wind farm.

And that might not be the only factor leading to underperformance. There is at least one other potential candidate, and it could be a lot easier to deal with, because it is all in the mind of the developer. To illustrate it, Matthew Hendrickson, senior director of energy assessment at 3TIER, a global leader in renewable energy risk analysis, asks you to consider five potential wind plants that you could invest US$200 million in.

If each project has a similar price in terms of capital and operational expenditure, and wind resource is the only apparent difference, then logic dictates you will select the one that has the highest net present value.

But now consider that these projects actually all have the same net capacity factor (NCF), and the differences in apparent output are actually the result of the underlying uncertainty in your wind resource assessments. If you picked the project with the highest apparent output then you have essentially picked the project with the most optimistic wind resource assessment and the one that, therefore, is most likely to underperform.

“Examining the historic bias of the industry, it is possible that the assessment process is more accurate than it receives credit for, only that the industry tends to build projects with the error on the side of underperformance,” Hendrickson says.

One way of dealing with selection bias, or indeed any bias, is to try to eliminate it by counterbalancing your calculations in the opposite direction. And, indeed, wind project investors and developers now traditionally give their figures a ‘haircut’ to counteract uncertainties.

Fitch, for example, states in its Global Infrastructure & Project Finance Rating Criteria for Onshore Wind Farms Debt Instruments that it “typically applies a 5% haircut to the results of the wind energy assessment for most greenfield projects.”

There are a couple of drawbacks to the haircut approach, however. The first is that although the haircut levels can be improved through on-the-ground feedback on performance, this information takes three or four years to trickle through, so it is usually relates to older turbine technology.

And the second is that it does not inspire much confidence in the professionalism of the industry when investors have to routinely lop chunks off the value of business plans in order to feel comfortable about putting money into them.

For this reason, Hendrickson says: “The challenge is on us to not only demonstrate we have a grasp on the sources of uncertainty, but to systematically reduce them through guidance and the sophisticated use of observations and advanced models that begin years before the project is built.” Although it is not the only determinant of project locations, he adds: “In the typical model the wind resource is the most sensitive driver of financial return.”

The extent to which it is important to estimate wind resource accurately is dramatically underscored if you model the profitability of net present value returns based on different degrees of NCF uncertainty. For a typical 100MW project, if there is an NCF uncertainty of 10%, which is not unusual, you have a 5% probability that the project will lose $6.1 million, according to Hendrickson’s calculations.

But because of the influence of selection bias the potential loss could be much greater. “Even a modest consideration of the NCF in selection decision yields bias of 4.8% and a further reduction of the P95 by $1.7 million,” Hendrickson says. At 6% NCF uncertainty, however, the P95 value equates to a gain of $4.8 million relative to expectations. “This is a $10.9 million improvement over the 10% case,” Hendrickson points out. And selection bias is minimised.

In fact, Hendrickson estimates that in a 100MW project each 1% reduction in uncertainty improves the P95 downside case by $2.7 million and improves the P50 selection bias by $1 million.

Gaining such reductions is not difficult. It starts with collecting high quality observations. Additionally, companies like 3TIER can combine those observations with advanced numerical weather prediction models to greatly reduce temporal and spatial uncertainty in wind resource estimation, helping to shave off those valuable percentage points.

So maybe it is time the industry developed a new kind of bias: for more accurate modelling.