Precise power calculation when running with WindRose objects

[misi9170]

@pjstanle pointed out that when running FLORIS binned by condition, as we do when running with the wind_data set as a WindRose object, we lose some resolution (in the power output) by evaluating the power curve only at the mean wind speed in the bin (or whatever that produces in the wakes).

We could, instead, evaluate the power curve at the bin edges, as well as any points that the power curve is defined at in between the bin edges, and then combine those powers to get a better representation of the power produced across the entire bin.

A good example of when this might matter is near cut-out. For example, consider a turbine with a cut-out wind speed is 25.0 m/s, with rated power produced at 24.0 and 25.0, and zero power at 25.1 and 26.0. For a bin spanning 24--26 m/s, our current approach will evaluate the power curve at 25.0 and assume that the power contribution for the bin is rated. However, in fact, the bin is half at rated and (almost) half at zero power, and really the power contribution for this bin should be about half of rated power.

Theoretically, the same is true for thrust. However, altering how thrust is computed is more complicated, since it happens within the flow calculation (whereas power calculations only happen after the flows are computed). My feeling is that our current method is fine for thrust, but that this suggestion could be used for power evaluations only when running with a WindRose object (or similar "binned" data object).

One part of this that is not clear to me, though, is what bin edges to assign to waked turbines. Once the flow is evaluated (at the mean wind speed of a given bin), the wind speed at a waked turbine will be some other value. To implement this approach, we would then need to define "edges" on the waked wind speed. An option would be to simply assume that that waked wind speed is the mean over a bin of the same width as the base binned wind speeds, which would allow us to follow the same approach---but that is somewhat heuristic.

Note that the error introduced by our current method reduces with the width of wind speed bins (with the cut-out wind speed example mentioned above, there would be no problem if wind speed bins are of width 0.1 m/s).

[Bartdoekemeijer]

It would be interesting to quantify the impact of this, e.g., a simple convergence study with AEP calculated at different wind speed step sizes would be insightful. Your example at cut-out makes sense, but my expectation would be that the effect on AEP is small because the frequency for that wind speed range is so low. Perhaps this effect is larger around the 'knee' of the power curve, right when region 2 transitions to region 3?