MAIN PUBLICATION :

• APPENDIX
• ABREVIATIONS
• GLOSSARY
• REFERENCES
• EVENTS
• Downloads

Analytical Methods for the Prediction of the Long-Term Wind Regime at a Site

From the above, it is clear that the key element of the assessment of the energy production of a proposed wind farm site is the prediction of the long-term wind regime at the site.  The outcome of the analyses described in this section is a long-term wind speed distribution, together with the wind rose’  Other meteorological inputs to the energy production analysis are the long-term site air density and site turbulence intensity, a measurement of the ‘gustiness’ of the wind. These, while still important, are of secondary influence to the energy production of the wind farm, and therefore their derivation is not considered in detail here.  It should be noted, however,  that the turbulence intensity is very important in determining the loading on a wind turbine, and hence its life expectancy.

OVERVIEW

When assessing the feasibility of a potential wind farm site or where, for strategic purposes, an indication of the variation of wind speed over an area is required, it is unlikely that any wind data from a relatively tall meteorological mast will be available.  If there is no on-site data available, the 'wind atlas method' (Troen and Petersen, 1989) is commonly used.  This method uses modelling techniques to translate the long-term reference station data to the site.  This method can be quite accurate in many cases, but should not replace on-site measurements for more formal wind farm energy assessment.  It is also possible to make predictions of the wind speed at a site using a numerical wind atlas methodology, based on a data source such as the ‘reanalysis’ numerical weather model data sets.  Again, such analyses are generally used to assess the feasibility of a site or sites for development.  

There are essentially two methods that can be used for the prediction of the long-term wind resource at a site where on-site measurements are available:

• Method 1: Correlate on-site wind data with wind data recorded at a long-term reference station; and
  
• Method 2: Use only on-site wind data. 
  

Unless a long data set is already available from a site, it is desirable to use Method 1 for the prediction of the long-term wind resource at a site.  Typically, a reliable result can be obtained with as little as one year worth of site data.  As illustrated by the example presented for Malin Head Meteorological Station above, if Method 1 cannot be used and Method 2 is used with only one year of data, the uncertainty in the assumption that the year of data recorded is representative of the long term is substantial. Therefore, it is normal practice to find a suitable source of longer-term data in the vicinity of the wind farm site.  This allows a correlation analysis to be undertaken and, if only relatively short data sets are available from the site, it is likely to result in an analysis with significantly less uncertainty than that resulting from use of the site data alone.  However, before a data set from a long-term reference station can be used in an analysis, it is vital that thorough checks on the validity of the data for the analysis are undertaken.  

 

Before discussing the details of this approach, it may be helpful to consider the broader picture.  It would be ideal if every site benefited from a long term data set of, say, 10 years, measured at hub-height.  Now and again this happens, but it is very rare. It is, therefore, necessary either to use limited on-site data or to try and use other data to gain a long-term view.  The correlation approach can be thought of in the following way. Data is gathered on the site using good quality calibrated equipment.  This data provides absolute measurements of the wind speed on the site during the measurement period.  If it can be established that there is a close relationship (a good correlation) between the site data and a reference mast, then it will be possible by using the long-term reference data and the relationship to re-create the wind speeds on the site.  Thus, it is possible to ‘pretend’ that the long-term wind speed records exist on the site.  If a good correlation exists, this is a very powerful technique but, if the correlation is weak, it can be misleading and hence it must be used with caution.

Necessary conditions for an off-site wind data set to be considered as a long-term reference are set out below:

• The reference data set includes data which overlaps with the data recorded on site.
• It can be demonstrated that the data has been recorded using a consistent system over the period of both the concurrent and longer-term data.  This should include consideration, not just of the position and height of the mast and the consistency of equipment used, but also potential changes in the exposure of the mast.  For example, the construction of a new building at an airport or the erection of a wind farm near an existing mast will corrupt the data.  The absolute values recorded at the reference station are not important, but any changes to it, in either process or surrounding environment, will render it useless as a reference site.  This investigation is therefore very important and is usually done by a physical visit to the site, together with an interview with site staff.
• The exposure of the reference station should be good.  It is rare that data recorded by systems in town centres, or where the mean wind speed at the reference station is less than half that of the site, prove to be reliable long-term reference data sets.  
• The data is well correlated with that recorded at the site. 
  

Where there have been changes in the consistency of data sample at a reference long-term data source, or where a reliable correlation cannot be demonstrated, it is important that the use of a prospective source of long-term data is rejected.  If no suitable reference meteorological station can be found, then the long-term wind resource can only be derived from the data recorded at the site itself.  It is likely that longer data sets of two or more years are required to achieve similar certainty levels to those that would have been obtained had a high quality long-term reference data set been available.

Experience of the analysis of wind energy projects across Europe has indicated that the density of public sources of high quality wind data is greater in northern Europe than in southern Europe.  This observation, combined with the generally more complex terrain in much of southern Europe, often leads to analyses in southern Europe being based on only the data recorded at the wind farm site, or other nearby wind farm sites. In contrast, for analyses in northern Europe, correlation of site data to data recorded at national meteorological stations is more common.  Clearly, this observation is a generalisation, however, and there are numerous exceptions.  The establishment of a good set of long-term reference masts, specifically for wind energy use in areas of Europe where wind energy projects are likely to be developed, would be an extremely valuable asset.  An EU-wide network of this sort would be highly beneficial.

CORRELATION METHODOLOGIES

The process of comparing the wind speeds on the site with the wind speeds at the reference station, and using the comparison to estimate the long-term wind speed on the site, is called 'measure correlate predict' (MCP).  This process is also described in some detail in Appendix D, along with a more detailed discussion of the merits of different methodologies.  It is difficult to provide definitive guidance on how poor the quality of a correlation can be before the reference station may no longer be reliably used within an analysis.  However, as a general rule, where the Pearson Coefficient (R²) of an all-directional monthly wind speed correlation is less than 0.8, there is substantial uncertainty in using long-term data from the reference station to infer long-term wind conditions at the wind farm site. 

Once the MCP process has been completed, an estimate exists of the long-term wind speed at the site. This stage – Milestone 1 on Figure I.2.2 – is a very important one, since the position has now been reached at which we have knowledge of the long-term wind speed behaviour at a single point (or points if there are multiple masts) on the site. This estimate will contain both the mean long-term expected value and the uncertainty associated with that value.  So far, however, we know nothing of the distribution of the wind speed across the site and neither have we considered the way in which the wind speed values can be converted into energy.