Titel: Evaluation of the Software program Windfarm and Comparisons with Measured Data from Alsvik, FFA TN 2000-30
Författare: Erik Djerf & Henrik Mattson
Utgivare: Flygtekniska Försöksanstalten
Årtal: 2000
Ämnesord: Teknik
Sökord/keywords: Alsvik, Gotland, wake, aerodynamics, wind farm, vindforsk.
Rapport (Pdf)

Sammanfattning/Abstract: The use of wind power is steadily increasing around the world as a viable, environmentally sound complement and alternative to fossil fuels. For a number of reasons when building wind power plants, the turbines are often placed in groups, socalled wind farms. This method of placement results in the fact that turbines will sometimes be immersed in the disturbed flow, the wake, of other turbines. Today, there exist a number of software programs that have been designed for the purpose of creating, analysing and optimising wind farm layouts. Effective and reliable predictions of wind farm performance is of importance when planning and establishing wind farms, as this reduces the economic risks associated with them. FFA acquired the wind farm analysis program WindFarm for trial purposes. The first part of this report deals with the evaluation and testing of WindFarm. Among the questions asked were; How sensitive is WindFarm to variations in the input parameters? How well does it predict the energy output of a wind farm? The idea was to get a clearer understanding of the program functions, calculation methods, efficiency and the calculation accuracy.

Several test cases were examined to gain a better knowledge of the operation of the
program. Firstly, the program was tested to try and establish how deviations in input
parameters affect the calculated results. WindFarm was also compared to FFA’s own
wind farm analysis program, FFA-MILLY. This was done by remaking a previous
investigation performed with FFA-MILLY on a planned wind farm at Södra Kalmarsund. This site was run in WindFarm with exactly the same data as used with FFA-MILLY. Lastly WindFarm was compared to experimental measurements done at Vattenfalls experimental wind farm at Alsvik, Gotland. This was done to try and establish WindFarms ability to model and predict wind farm efficiency and energy yield.

WindFarm is a relatively new and untried software product, which naturally contains
bugs and faults. In addition, there are a number of parameters, calculation options
and wake modelling options that must be set and defined by the user, thus offering
ample opportunity for mistakes and errors. Further it is concluded that there are some
relatively large differences between the available modelling methods. However, determining which model is to be preferred cannot be done on the basis of investigations made in this report. To make an accurate and fair evaluation of wind farm modelling methods, accurate data from larger wind farms than Alsvik would be needed. However the results of this report should give a hint to which parameters dominate the calculation results and may help WindFarm users judge the validity of calculations and possibly how to improve them.

The second part of the report examines measured data from Vattenfalls experimental
wind farm in Alsvik on the island of Gotland in the Baltic Sea. The focus of this part
of the report was to try to determine the accuracy of these measurements. The data
was treated with a focus on trying to eliminate, or compensate for, as many consistent errors as possible and to make estimates of the remaining uncertainties. This was done so that the models used in WindFarm could be compared to the measured data as accurately as possible.

The main sources of error that were examined were, yawed flow and uncertainties in
the wind velocity- and wind direction measurements.

When the turbine is not aligned perfectly perpendicular to the wind it gives rise to so called yawed flow. This misalignment of the turbine clearly affects the power output, but by removing all data emanating from the most misaligned turbines it was possible to limit the deviations in the output power data. When studying the effects of yawed flow an interesting discovery was made; turbine misalignment need not necessarily lead to a lower energy production as expected. On the contrary, a couple of degrees of positive yaw angle might even increase turbine efficiancy.

When examining the wind direction measurements, the wind vane seemed to be
offset 10-15 degrees from the expected value. This was dealt with using a correction
equation originally created by Jan-Åke Dahlberg. Attempts were made to improve
upon his original correction equation but the results were not good enough to justify
discarding Dahlberg’s original equation.

The wind velocity measured by the anemometers, mounted on booms stretching out
from the measurement masts, is somewhat disturbed by the presence of the mast and
boom itself. This gave rise to uncertainties when measuring the wind speed. Instead
of using the anemometers, the possibility of using the turbines themselves to measure
the wind velocity was examined. This turned out to be a better way of measuring the
wind velocity since the turbine is not influenced by the measurement mast. The error
due to the mast effects is thus eliminated.

The statistical uncertainties were assumed to be independent of the errors from the
yawed flow and were then combined and error bars were added to the data.

The data were thereafter compared to calculations made by FFA-MILLY and WindFarm. The conclusions from the comparisons were that the models, in general, seem to underestimate the wake losses, and thereby over predict the total energy yield of the wind farm. This is not a very surprising result when considering the short distances involved in Alsvik. The models used start their calculations at a distance of about 4-5 diameters downstream of the turbine and this is the shortest distance involved in Alsvik. When the distance to the disturbing turbine increases, the accuracy of the models improves. It would therefore be interesting to see the effect of using the models on a larger wind farm.