Heterogeneous power assumptions for wake loss module?

[moptis]

Hi!

V3's wake loss algorithm assumes that each turbine at a wind farm is capable of producing the same amount of power as the free stream wind turbines. As you note, that assumption breaks down in heterogenous conditions (e.g. complex terrain) and unrealistic wake loss estimates can result.

Do you envision an upgrade where the user could instead specify relative free stream power production for each turbine by wind sector? This type of info is generally available from pre-construction energy estimates, and could greatly improve the wake loss estimates.

Thanks!

[RHammond2]

Thanks for raising this consideration, @moptis! I think @ejsimley might be better suited to answer this question.

[ejsimley]

Hi @moptis. Thanks for raising this issue. As we discussed briefly in November, we're planning to update the wake loss method to allow corrections for free stream wind variations across the wind plant. The current plan is for users to provide relative wind speeds for each turbine by wind direction sector (possibly with a wind speed dependence). Then we'd use empirical power curves from the SCADA data to estimate the relative free stream power production at each turbine by wind direction sector. This would then be used to normalize the wake losses estimated from SCADA to correct for spatial wind resource variations.

I expect we'll have a pull request to address this by April. Let us know if you have any suggestions for how to apply this correction, and also if there are data formats you think would be convenient for providing the relative free stream wind speeds or powers for each turbine and direction.

[moptis]

@ejsimley Great news! I think the WRG format would be ideal as input. It's a bit clunky, but it does provide wind speed distributions by sector using Weibull parameters. Those could be converted into power pretty easily. You'd probably be fine using a generic power curve since we're only looking at relative differences. Empirical curves from SCADA would be great but would be more computationally intensive and potentially subject to error (i.e., if not cleaned properly).

Keep me posted on the development!