reV Losses Module

[ppinchuk]

Added a new reV module for calculating losses. This module facilitates the addition of losses to generation outputs that are more sophisticated than simple haircut losses. Two new types of losses are added in particular.

Power Curve Losses (Wind only)

Users can now specify a target loss percentage and a type of power curve transformation. reV will then attempt to apply the transformation in order to reach the target loss percentage given the annual wind distribution at each site. This results in different transformations for each site.

~Currently, a horizontal power curve translation is the only kind of transformation that is implemented.~ JK, I added a total of three :). I think the latter two should make Owen much happier. New transformations can easily be implemented by sub classing PowerCurveTransfromation and implementing an apply method that calculates the transformation using a single tunable variable, and a bounds property that sets the limits on that transformation variable (and also register the new class to the TRASNFORMATIONS dictionary).

Stochastically Scheduled Outages (Wind and Solar)

Users can now specify information about a set of outages, which reV then attempts to schedule stochastically. Details for each outage include

• count : An integer value representing the total number of times the outage should be scheduled.
  • duration : An integer value representing the total number of consecutive hours that the outage should take.
  • percentage_of_farm_down : An integer or float value representing the total percentage of the farm that will be take offline for the duration of the outage.
  • allowed_months : A list of month names corresponding to the allowed months for the outage. Month names can be unformatted and can be specified using 3-letter month abbreviations.
  • allow_outage_overlap : (optional) A bool flag indicating whether or not this outage is allowed to overlap with other outages, including itself.

Notes

• I did try to apply the stochastically scheduled outages using a recursive function, but the recursion was always tail-recursion, which is not optimized in Python. Therefore, I went with a simple loop model instead.
• The horizontal power curve translation is extremely basic (and not necessarily realistic), but it is a minimal viable product for adding losses while still allowing generation to reach max rated. The code is left open for extension to more sophisticated power curve transformations. (EDIT: This point is somewhat moot and outdated, but technically correct).
• I had to change the AbstractSamPv.gen_profile property to pull the "gen" profile instead of "ac". This change was necessary to overcome a bug in SAM/PySAM < 3.0.0 and can (probably) be reverted when reV compatibility with PySAM 3.0.0+ is integrated.
• I attempted to be thorough in the documentation, but it is always difficult to predict confusing topics/methods/inputs when you are close to the code itself. Feedback is welcome.

[ppinchuk]

Feel free to wait until after our discussion on Tuesday to review.

[codecov-commenter]

Codecov Report

Merging #362 (034c734) into main (a947b6b) will increase coverage by 1.19%. The diff coverage is 98.16%.

@@            Coverage Diff             @@
##             main     #362      +/-   ##
==========================================
+ Coverage   79.14%   80.34%   +1.19%     
==========================================
  Files         135      142       +7     
  Lines       18211    19409    +1198     
==========================================
+ Hits        14414    15595    +1181     
- Misses       3797     3814      +17     

Flag	Coverage Δ
unittests	80.34% <98.16%> (+1.19%)	:arrow_up:

Flags with carried forward coverage won't be shown. Click here to find out more.

Impacted Files	Coverage Δ
tests/test_losses_utils.py	94.91% <94.91%> (ø)
reV/losses/utils.py	96.29% <96.29%> (ø)
tests/test_losses_scheduled.py	96.75% <96.75%> (ø)
tests/test_losses_power_curve.py	98.34% <98.34%> (ø)
reV/SAM/generation.py	93.59% <100.00%> (+0.08%)	:arrow_up:
reV/losses/__init__.py	100.00% <100.00%> (ø)
reV/losses/power_curve.py	100.00% <100.00%> (ø)
reV/losses/scheduled.py	100.00% <100.00%> (ø)
reV/utilities/exceptions.py	100.00% <100.00%> (ø)
reV/config/project_points.py	85.41% <0.00%> (+0.04%)	:arrow_up:
... and 6 more

---

Continue to review full report at Codecov.

Legend - Click here to learn more Δ = absolute <relative> (impact), ø = not affected, ? = missing data Powered by Codecov. Last update a947b6b...034c734. Read the comment docs.

[ppinchuk]

Going to add a few more things while testing this out with Owen so no rush really

[grantbuster]

It looks really good based on my verification of the outputs. My only real concern on the technology side is that it respects the cutout speeds even when shifted.

[ppinchuk]

Yea, so I perform that check in multiple places https://github.com/NREL/reV/blob/3313b708e4b00ebf6ec76367b0b086a9b131b501/reV/losses/power_curve.py#L152

and also: https://github.com/NREL/reV/blob/3313b708e4b00ebf6ec76367b0b086a9b131b501/reV/losses/power_curve.py#L613 which has the following code:

new_gen = self.power_curve(
    self.power_curve.wind_speed - transformation_var
)
mask = (
    self.power_curve.wind_speed >= self.power_curve.cutoff_wind_speed
)
new_gen[mask] = 0

That all being said, I am adding some more commits below to verify and ensure this behavior moving forward