Achieving a more accurate estimate of a wind resource
Many, but not all, wind farm projects do not deliver the energy production (and thus revenue) that was calculated pre-construction. Several explanations have been provided including inadequate O&M, turbines not having the expected availability and also the methodology used to calculate the long-term wind resource. This article focuses on improvements available to calculate the resource, and its potential energy production, which then feed into the financial analysis of a proposed wind farm project.
By Paul van Lieshout, Wind Power Group Manager, Sinclair Knight Merz, UK .
In 2005 SKM (Sinclair Knight Merz) started to implement improved correlation methodology to establish a relationship between met station data and wind data measured on site (either onshore or offshore), in order to establish a long-term resource dataset upon which the AEP of a proposed wind farm can be calculated with greater accuracy. The paper provides a summary of SKM’s work.
In order to calculate the future AEP of a potential wind power project, a three-step methodology is normally used to calculate the long-term wind speed of a proposed wind farm site.
The AEP of the proposed wind power station can then be determined using this long-term dataset, in combination with wind flow tools (like WAsP or CFD programs), wind turbine power curves, thrust curves, electrical layouts and so on.
But the first task is to establish the long-term wind resource at the proposed site, a calculation often based on the MCP methodology.
The three steps of the methodology are briefly described below, then we focus on the second step, namely the correlation process:
Approximately one year of wind resource data, simultaneously measured at the proposed wind farm site and at a suitable weather station is correlated to establish a relationship between the two datasets. A long-term site record can then be calculated based on this mathematical relationship using the long-term wind resource record of the weather station as input data. The resultant long-term wind speed dataset for the proposed wind farm site forms the basis of the AEP calculations.
There are several mathematical curves that can be used to describe the relationship between the weather station data and the site data:
The objective of a correlation analysis is to identify a mathematical function that describes the relationship between the wind characteristics at the meteorological site and the proposed wind farm site. It is noted that this is different from ‘finding a curve that fits the measured data best’. The curve that describes the relationship should be able to forecast site wind speed values from weather station data that has not been included in the establishment of the correlation curve – it should not just provide the highest ‘goodness of fit’ relationship.
Analytical Analysis – Describing the Wind
The IEC 61400 series of standards describe the wind resource at sites with Weibull distribution functions; these functions can be used to model a relationship between site and weather station. These Weibull functions are widely accepted in the wind industry, with numerous papers that depict the distribution of wind with Weibull distribution curves, confirming their ability to describe a wind resource and establish a correlation between measured weather station data and site data.
Correlation Using Weibull Curves
The SKM correlation methodology is based on IEC 61400 (see equation 1 which describes the Weibull probability distribution function (pdf)):
Equation 1
The Weibull cumulative distribution function or cdf is derived from the above equation. The wind resource at both the met station and the site can be described in terms of two cumulative distribution functions (see equation 2):
Equation 2
The correlation analysis establishes a one-to-one relationship between weather station data and site data.
Verifying the Results
Rigorous analysis using prediction and observation comparison techniques of the different correlation methodologies verifies the Weibull correlation methodology.
The original measured data file containing synchronised weather/met station and site data was randomly split forming two separate datasets. Set 1 was used to establish all three correlation curves used to predict three future or synthesised site datasets (one for each correlation methodology). Set 2, like Set 1, contains synchronised site and weather station data (our ‘observed future dataset’).
The weather station data from Set 2 was used together with the three different correlation functions (established based on data Set 1) to calculate/predict three different synthesised site wind speed datasets. These sets were then compared with the actual site data from Set 2 – comparing predicted with observed values.
The comparison of the synthesised/predicted dataset and the actual/observed dataset provides a measure of how well the correlation functions describe the measured data, but more important is the comparison of the energy production based on the predicted wind data with the energy production based on the observed wind data.
This process is randomly repeated 40 times during verification of the three different methodologies. Forty sets, each of three predicted synthesised wind speed datasets were constructed and compared with 40 actual (observational) datasets. These wind speed datasets (observed and predicted) were then translated to AEP values using a wind turbine generator power curve.
The AEP of the actual (observed) dataset was then normalised to one (shown as the straight vertical black line in Figure 1).
The mean AEP value and associated standard deviation based on the three correlation curves are shown as Gaussian distribution curves (RED based on the Weibull correlation curve, GREEN based on the straight line forced through zero correlation and BLUE based on the straight line correlation). These curves show that the AEP based on the Weibull distribution curve is closest to the actual AEP value, thus validating this methodology. The higher accuracy is based on first principles for all potential wind farm sites.
Conclusion
In assignments to calculate AEP values for developers, investors and financial institutions for both onshore and offshore projects, SKM has found that the above Weibull correlation methodology yields a higher accuracy and associated lower uncertainty whenever the site and particularly the weather station data was of high enough quality. This has been valid for onshore projects with different topographical and environmental characteristics as well as offshore projects.
Further Reading
IEC 61400-1, 2005–2008, Wind turbines – Part 1: Design requirements, International Electrotechnical Commission (IEC).
Biography of the Author
Mr van Lieshout has been involved in wind power projects for the past 28 years. During this period he has been involved in the research and development of wind turbines, the implementation of wind power projects around the world and the due diligence of proposed and existing wind power schemes. He manages SKM’s wind consultancy, project implementation and wind O&M teams – and because he is involved in both ‘due diligence’ of existing wind power projects and SKM’s management team, he brings a unique ‘full circle’ understanding of the technologies involved in wind power projects, enabling him to provide in-depth knowledge to development teams for both onshore and offshore projects.{/access}
Many, but not all, wind farm projects do not deliver the energy production (and thus revenue) that was calculated pre-construction. Several explanations have been provided including inadequate O&M, turbines not having the expected availability and also the methodology used to calculate the long-term wind resource. This article focuses on improvements available to calculate the resource, and its potential energy production, which then feed into the financial analysis of a proposed wind farm project.
By Paul van Lieshout, Wind Power Group Manager, Sinclair Knight Merz, UK .
{access view=!registered}Only logged in users can view the full text of the article.{/access}{access view=registered}There are no international guidelines for calculating the Annual Energy Production (AEP) of proposed wind farm projects. AEP calculations are normally based on the Measure, Correlate and Predict (MCP) methodology. The correlation has historically meant assuming a straight-line relationship between site data and long-term met station data. Such correlation methodology (including the matrix methodology) is based on simple mathematical statistics; it does not include nor describe physical properties of wind and this affects the accuracy of AEP results.
This article offers an alternative and transparent methodology using the full physical properties of the wind resource as described in the IEC 61400 series of international standards. The paper addresses the improvements in the central estimate (P50) as well as the uncertainties associated with this value (i.e. P90 and other values).In 2005 SKM (Sinclair Knight Merz) started to implement improved correlation methodology to establish a relationship between met station data and wind data measured on site (either onshore or offshore), in order to establish a long-term resource dataset upon which the AEP of a proposed wind farm can be calculated with greater accuracy. The paper provides a summary of SKM’s work.
In order to calculate the future AEP of a potential wind power project, a three-step methodology is normally used to calculate the long-term wind speed of a proposed wind farm site.
The AEP of the proposed wind power station can then be determined using this long-term dataset, in combination with wind flow tools (like WAsP or CFD programs), wind turbine power curves, thrust curves, electrical layouts and so on.
But the first task is to establish the long-term wind resource at the proposed site, a calculation often based on the MCP methodology.
The three steps of the methodology are briefly described below, then we focus on the second step, namely the correlation process:
- Measure wind speed at the site and weather stations simultaneously (over a period of approximately 12 months).
- Correlate the datasets to find a relationship between the two short-term datasets (site and weather station).
- Predict long-term site wind resource with the long-term dataset from the weather station together with the established relationship between the weather station and site data to calculate the long-term site wind resource.
Approximately one year of wind resource data, simultaneously measured at the proposed wind farm site and at a suitable weather station is correlated to establish a relationship between the two datasets. A long-term site record can then be calculated based on this mathematical relationship using the long-term wind resource record of the weather station as input data. The resultant long-term wind speed dataset for the proposed wind farm site forms the basis of the AEP calculations.
There are several mathematical curves that can be used to describe the relationship between the weather station data and the site data:
- A numerical analysis often using straight lines to depict the relationship.
- An analytical analysis using a mathematical function that describes the physical properties of the wind – this is the SKM methodology that connects the correlation function to IEC 61400.
The objective of a correlation analysis is to identify a mathematical function that describes the relationship between the wind characteristics at the meteorological site and the proposed wind farm site. It is noted that this is different from ‘finding a curve that fits the measured data best’. The curve that describes the relationship should be able to forecast site wind speed values from weather station data that has not been included in the establishment of the correlation curve – it should not just provide the highest ‘goodness of fit’ relationship.
Analytical Analysis – Describing the Wind
The IEC 61400 series of standards describe the wind resource at sites with Weibull distribution functions; these functions can be used to model a relationship between site and weather station. These Weibull functions are widely accepted in the wind industry, with numerous papers that depict the distribution of wind with Weibull distribution curves, confirming their ability to describe a wind resource and establish a correlation between measured weather station data and site data.
Correlation Using Weibull Curves
The SKM correlation methodology is based on IEC 61400 (see equation 1 which describes the Weibull probability distribution function (pdf)):
Equation 1
The Weibull cumulative distribution function or cdf is derived from the above equation. The wind resource at both the met station and the site can be described in terms of two cumulative distribution functions (see equation 2):
Equation 2
The correlation analysis establishes a one-to-one relationship between weather station data and site data.
Verifying the Results
Rigorous analysis using prediction and observation comparison techniques of the different correlation methodologies verifies the Weibull correlation methodology.
The original measured data file containing synchronised weather/met station and site data was randomly split forming two separate datasets. Set 1 was used to establish all three correlation curves used to predict three future or synthesised site datasets (one for each correlation methodology). Set 2, like Set 1, contains synchronised site and weather station data (our ‘observed future dataset’).
The weather station data from Set 2 was used together with the three different correlation functions (established based on data Set 1) to calculate/predict three different synthesised site wind speed datasets. These sets were then compared with the actual site data from Set 2 – comparing predicted with observed values.
The comparison of the synthesised/predicted dataset and the actual/observed dataset provides a measure of how well the correlation functions describe the measured data, but more important is the comparison of the energy production based on the predicted wind data with the energy production based on the observed wind data.
This process is randomly repeated 40 times during verification of the three different methodologies. Forty sets, each of three predicted synthesised wind speed datasets were constructed and compared with 40 actual (observational) datasets. These wind speed datasets (observed and predicted) were then translated to AEP values using a wind turbine generator power curve.
The AEP of the actual (observed) dataset was then normalised to one (shown as the straight vertical black line in Figure 1).
The mean AEP value and associated standard deviation based on the three correlation curves are shown as Gaussian distribution curves (RED based on the Weibull correlation curve, GREEN based on the straight line forced through zero correlation and BLUE based on the straight line correlation). These curves show that the AEP based on the Weibull distribution curve is closest to the actual AEP value, thus validating this methodology. The higher accuracy is based on first principles for all potential wind farm sites.
Conclusion
In assignments to calculate AEP values for developers, investors and financial institutions for both onshore and offshore projects, SKM has found that the above Weibull correlation methodology yields a higher accuracy and associated lower uncertainty whenever the site and particularly the weather station data was of high enough quality. This has been valid for onshore projects with different topographical and environmental characteristics as well as offshore projects.
Further Reading
IEC 61400-1, 2005–2008, Wind turbines – Part 1: Design requirements, International Electrotechnical Commission (IEC).
Biography of the Author
Mr van Lieshout has been involved in wind power projects for the past 28 years. During this period he has been involved in the research and development of wind turbines, the implementation of wind power projects around the world and the due diligence of proposed and existing wind power schemes. He manages SKM’s wind consultancy, project implementation and wind O&M teams – and because he is involved in both ‘due diligence’ of existing wind power projects and SKM’s management team, he brings a unique ‘full circle’ understanding of the technologies involved in wind power projects, enabling him to provide in-depth knowledge to development teams for both onshore and offshore projects.{/access}
