Streamlined computations for applying Marion diffuse-IAM corrections?

[Mark Campanelli]

Hi,

  It is unclear to me if there is a streamlined computational "wrapper" in pvlib that allows one to compute all the IAM-corrected POA components and their total (inc. without computing some things twice using different existing wrapper functions).

  Below is the lower-level way that I do this now (excerpted from https://github.com/markcampanelli/pvpmc2024/blob/main/tutorial/2_dc_performance_prediction.ipynb). Note that this approach may be specific to model choices in perez.

-Mark

# Compute angle of incidence of sun on POA.

aoi_deg = pvlib.irradiance.aoi(

array["surface_tilt_deg"],

array["surface_azimuth_deg"],

weather["apparent_zenith"],

weather["azimuth"],

)

# Compute direct component on POA.

poa_components = pandas.DataFrame(

data={"direct": numpy.maximum(weather["dni"] * numpy.cos(numpy.radians(aoi_deg)), 0)},

index=weather.index

)

# Compute sky-diffuse components on POA using Perez allsitescomposite1990.

poa_components_perez = pvlib.irradiance.perez(

array["surface_tilt_deg"],

array["surface_azimuth_deg"],

weather["dhi"],

weather["dni"],

pvlib.irradiance.get_extra_radiation(weather.index),

weather["apparent_zenith"],

weather["azimuth"],

airmass=pvlib.atmosphere.get_relative_airmass(weather["apparent_zenith"]),

model="allsitescomposite1990", # Cf. albuquerque1988

return_components=True,

)

poa_components["circumsolar"] = poa_components_perez["circumsolar"]

poa_components["isotropic"] = poa_components_perez["isotropic"]

poa_components["horizon"] = poa_components_perez["horizon"]

# Compute ground-diffuse component on POA.

poa_components["ground_diffuse"] = pvlib.irradiance.get_ground_diffuse(

array["surface_tilt_deg"],

weather["ghi"],

albedo=array["albedo"],

)

# Compute total POA irradiance.

poa_components["total"] = numpy.sum(poa_components.to_numpy(), axis=1)

# Compute IAMs for diffuse components.

# pvlib.iam.marion_diffuse cannot make use of our custom fit IAM function,

# so instead use lower level pvlib.iam.marion_integrate for each component.

diffuse_iam = {}

for region in ["sky", "horizon", "ground"]:

diffuse_iam[region] = pvlib.iam.marion_integrate(

lambda angle_deg: iam(angle_deg=angle_deg), array["surface_tilt_deg"], region

)

# Compute IAM-corrected effective POA irradiance.

poa_components["total_eff"] = iam(angle_deg=aoi_deg) * (

poa_components["direct"] + poa_components["circumsolar"]) + \

diffuse_iam["sky"] * poa_components["isotropic"] + \

diffuse_iam["horizon"] * poa_components["horizon"] + \

diffuse_iam["ground"] * poa_components["ground_diffuse"]

[Mark Campanelli]

To simplify and standardize one small part of this workflow, I opened an issue and PR to allow one to pass a custom IAM function (including the existing pvlib.iam.interp) to the high-level pvlib.iam.marion_diffuse function. The PR would also add pvlib.iam.pchip, which generates piecewise-cubic Hermite interpolating polynomial from (AOI, IAM) data. I have found PCHIP to have particularly nice properties for IEC 61853-2 IAM measurement data. Seethe new tests for example of that.

- https://github.com/pvlib/pvlib-python/issues/2049
- https://github.com/pvlib/pvlib-python/pull/2050

I tried to follow the coding style I see already in pvlib, but I have some outstanding questions about some items such as the proper level of input validation and the need to support pandas.Series.

-Mark