How does WindSim treat climatology data
Uploading data
When uploading climatologies (meteorological masts, lidars or otherwise) they can either be stored as timeseries or as a frequency table.
Here follows a simplified journey through the differences.
Frequency
In the frequency approach, which was historically used due to computational limitations, the measured conditions are stored in bins. In each of these bins a count of how many times this windspeed bin was measured is recorded.
| Windspeed | Frequency |
| 0-1 | 3 |
| 1-2 | 50 |
| 3-4 | 57 |
| ... | ... |
This is then used to compute for example the Annualized Energy Production (AEP), this lead to inaccuracies due to everything in a bin is grouped together.
To calculate the AEP at a turbine position WindSim then takes the relative speedup at the position compared to the climatology. Then for each wind direction it multiplies the relative frequency and the windspeed in the bin and this is then paired with the power table to produce AEP.
Time Series
In the time series approach each measured entry is represented by a timestamped entry. This leads to a more correct computation, with less loss of precision.
To compute the AEP for each timestep and climatology the following steps are done- each entry is interpolated to wind sectors
- it is multiplied with the time between timesteps
- the speedup factors are applied
- paired with the power table to show energy
Differences
Due to how a measurement of 4.9 m/s at 10 degrees are treated in the two different approaches we expect differences.
In the frequency approach it is stored as 1 entry in the 4-5 bin at 30 degrees and information is lost.
Conclusion
In general we always recommend to use time series if possible.
In some cases there are differences in how WindSim computes certain values and there can be seen a discrepancy in Windspeed between different parts. This is due to not all parts having the timeseries approach fully completed, and they are using the frequency tables.
Some of the formats we are exporting is based on frequency tables so there may be significant differences in values.