N-S trackers with manual sky generation

[Lukas-dvonc]

Hi,

I am using bifacial_radiance and leveraging the gendaylit2manual() function to generate skies based on the Perez All-Weather model. This approach works very well for fixed-tilt systems, since I can fully control the inputs (DNI, DHI, sun elevation, sun azimuth) and loop over timestamps (e.g., from a pandas DataFrame). This allows direct comparison with other simulation tools such as PVsyst on a per-timestep basis.

A simplified version of my workflow for fixed-tilt systems is:

for i, row in df_filt.iterrows():
dni = row['DNI']
dhi = row['DHI']
sunalt = row['ELEV']
sunaz = row['AZIM']
albedo = row['albedo']

mysky = demo.gendaylit2manual(
    dni=dni,
    dhi=dhi,
    sunalt=sunalt,
    sunaz=sunaz,
)

demo.setGround(albedo)
octfile = demo.makeOct(demo.getfilelist())

This workflow gives me full control over the sky definition and works as expected.

---

However, I am struggling to apply the same approach to single-axis (N-S) tracking systems.

From my understanding, the typical bifacial_radiance workflow for tracking relies on built-in functions (e.g., set1axis / gendaylit-based workflows), which internally handle sun position and tracking angles. This seems to conflict with the manual sky generation approach using gendaylit2manual().

My questions are:

1. Is it possible to use gendaylit2manual() together with N-S tracking simulations in bifacial_radiance?
2. If yes, what is the recommended workflow to:
   • define a custom sky per timestamp, and
   • correctly update tracker geometry (tilt/rotation) for each timestamp?
3. If not, is there an alternative way to achieve full control over irradiance inputs (DNI/DHI) while still simulating tracking systems?

My goal is to run timestep-by-timestep simulations with externally provided meteo data and compare results directly with other tools.

Any guidance or example workflows would be greatly appreciated.