Using pvwatts losses with ModelChain

[casaout]

Hi

I am trying to use the pvwatts loss model with the ModelChain, similar to Jawairia123 a few years ago. In the meantime, there was a successful PR by @cwhanse and @wholmgren, to make this possible.

Unfortunately, I am unsuccessful at correctly applying the helper-function at this point and haven't been able to find a working example online.

Question 1: Input parameter units? Per documentation, the input and output unit is in%. However, the default parameters range from 0-3 (which would be 0-300%). How can I make sense of this?
Concretely, if a pvsystem is e.g. 8years old, relative to what do I consider it to receive an input %?
Also, to understand the other parameters, what would shading==3 or availability==3 (all the defaults) mean?

Question 2: How do I add the losses_model to the ModelChain? What I am trying today:

system_losses = pvwatts_losses(age=8, snow=0)  # todo: % or number of years? (defaults seem absolute, not %)
print(f"LOSS: {1 - (system_losses / 100)}")
mc = ModelChain(system, location, aoi_model="physical", losses_model=system_losses) 
mc.run_model(times=mc_weather.index, weather=mc_weather)

However, the output is: that the first argument must be callable. An alternative would be what Jawairia123 suggested, but my assumption was that the PR mentioned above made it possible in the meantime to use the pvwatts loss model with ModelChain.

Any pointers are appreciated. thank you!

[kandersolar]

Hi @casaout, hopefully this clears things up a bit :)

Question 1: Input parameter units? Per documentation, the input and output unit is in %. However, the default parameters range from 0-3 (which would be 0-300%). How can I make sense of this?

The inputs are in %, meaning a default value of 3 is interpreted as 3%=0.03. If the inputs were supposed to be unitless then the default value would be shading=0.03.

---

Also, to understand the other parameters, what would shading==3 or availability==3 (all the defaults) mean?

Each of these are percentage losses that get combined into a single overall loss fraction. So for example shading=3 means that 3% of system production is lost to shading, or the system produces only 97% of what it would have if shading=0. And the other losses stack in a similar fashion. Note that PVWatts losses are applied to the overall system output, not at the point in the system that each loss occurs in real life -- so for example, a soiling loss is not "masked" by inverter clipping in the PVWatts loss model.

---

system_losses = pvwatts_losses(age=8, snow=0) # todo: % or number of years? (defaults seem absolute, not %)

The age parameter is not a number of years, but simply another percentage loss like all the other parameters. So for example, if you expect the system's performance to degrade at 0.5%/yr, and the system is 8 years old, then you might specify age=4 to indicate a 4% loss due to system age. Anything more sophisticated than that (e.g. an age loss that changes with time) is outside the scope of a simple model like PVWatts.

---

print(f"LOSS: {1 - (system_losses / 100)}")

Note that pvwatts_losses returns a loss (not the remaining capacity), so e.g. with the default inputs, the return value of ~14 represents a 14% loss, or an 86% remaining capacity. So this print line of code is not correct, as it converts the loss to a remaining capacity but then calls it a loss.

---

mc = ModelChain(system, location, aoi_model="physical", losses_model=system_losses

Check out the ModelChain docs to see what is expected for the losses_model parameter:

losses_model (str or function, default 'no_loss') – Valid strings are ‘pvwatts’, ‘no_loss’. The ModelChain instance will be passed as the first argument to a user-defined function.

So the losses_model value is supposed to be a string (or a function if you want to customize it, but nevermind that) like how you have aoi_model="physical". To specify the loss percentages you would use the losses_parameters input to PVSystem rather than calling the pvwatts_losses function yourself. In fact this is how you specify parameters for any model: set them in the appropriate PVSystem dict, then let ModelChain call the functions for you.

Here is a short example for using the losses_parameters input to PVSystem -- change the value for age, or specify any other pvwatts loss in that dict, and the ac output of the ModelChain will change accordingly.

from pvlib.modelchain import ModelChain
from pvlib.pvsystem import PVSystem
from pvlib.location import Location
import pandas as pd

system = PVSystem(losses_parameters={'age': 4},
                  module_parameters={'pdc0': 1, 'gamma_pdc': -0.004},
                  inverter_parameters={'pdc0': 1},
                  temperature_model_parameters={'a': -3.56, 'b': -0.075, 'deltaT': 3})
location = Location(latitude=32.2, longitude=-110.9)
weather = pd.DataFrame([[1050, 1000, 100, 30, 5]],
                       columns=['ghi', 'dni', 'dhi', 'temp_air', 'wind_speed'],
                       index=[pd.Timestamp('20170401 1200', tz='US/Arizona')])
mc = ModelChain(system, location, losses_model='pvwatts',
                aoi_model='no_loss', spectral_model='no_loss')
mc.run_model(weather)
print(mc.results.ac)

[casaout]

Thank you @kanderso-nrel. This definitely clarified things.
(after being so used of using decimals I've literally not considered that % is really just the percentage 😒😁)

Next stop, looking up on the manufacturer website / factsheet how much degradation I can expect per year 😊.