search

pvlib / pvlib-python

A set of documented functions for simulating the performance of photovoltaic energy systems.
https://pvlib-python.readthedocs.io
BSD 3-Clause "New" or "Revised" License
1.16k stars 985 forks source link

Minor GTI-DIRINT bugs #1902

Open adriesse opened 10 months ago

adriesse commented 10 months ago

Describe the bug gti_dirint() calculates a few things internally that are either not quite right or inconvenient.

  1. airmas is pressure-adjusted for use in DIRINT, but Perez transposition actually needs relative airmass.
  2. The airmass model is hard-coded as "kasten1966" for compatibility with DISC and DIRINT, whereas other calculations tend to use "'kastenyoung1989" as a default and the Perez model coefficients were developed using the latter.
  3. dni_extra is calculated using the hard-coded constant 1370 for compatibility with DISC.

These internal calculation make it harder to do a precise round-trip comparison between forward and reverse transposition and/or restrict the conditions under which a precise comparison is possible.

Work-around

  1. Use pressure=101325.
  2. Pre-calculate airmass and dni_extra for use in the forward transposition with the internal parameters of gti_dirint.

Versions:

prady0t commented 10 months ago

Hey can I try to solve this?

kandersolar commented 10 months ago

Are those errors in our implementation (i.e. we deviate from the reference), or conceptual inconsistencies in the Marion reference itself? If the latter, maybe we should enable these improvements via optional parameters while leaving the current behavior the default.

@prady0t thanks for offering to help! I think we still need to define what should be done here before anyone can start doing it :)

adriesse commented 10 months ago

I'll leave that open for discussion. We tried hard to replicate Marion's published results but couldn't.