Longevity experience

Could you define "geometric computation" as you apply that term to solar energy applications? I'm ignorant of that term with respect to solar energy or PV system monitoring.

Have you done a PVWatts run on your system ?

If not, check it out. It's a PV model from NREL that's free, user friendly, quite helpful and as I and others have found to be reasonably to surprisingly accurate with respect to stochastic handling of residential array output.
heck the help screens for a toe dip into probability density functions for array output.

JPM – My "geometric" calculations were just that, no probabilities of cloud coverage etc.
I considered only the path (angle) of the sun relative to normal at the panels, calculating the resultant power reduction.
Numerically integrated. Repeated for various tilts to choose my optimum.
Looking into my archive, I think the primary tool I used was what you see in this Excel spreadsheet –

Note – the sheet was created under MacOS 10.6.8, and uses macros. My memory on its use is dim. Pieces may have came from some NASA and NOAA sources.

Also I guess I could have included the actual kWh per year numbers –
Geom 9902
2008 7453
2009 7398
2010 7201
2011 7288
2012 7419
2013 7233
2014 7423
2015 7299
2016 7189
2017 6870
2018 7072
2019 6595
2020 6822
2021 7072
2022 7053

And I did a run of PVWatts for fun, here is the result –
PVWatts Calculator.pdf
That 6637 is a bit pessimistic compared with actual performance!

I also did a PVWatts run using your location, tilt and azimuth and got 7007 kWh/yr.
The mean for your data is 7159 kWh/yr.
That seems reasonable to me.

Thank you for the explanation. I know that type of data as solar geometry and is the type of data I've programed in FORTRAN back in the day, later in BASIC and the LOTUS and finally in Excel using open literature algorithms and formulas.
If you're modeling irradiance for PV modeling NREL has "TMY" file (Typical Meteorological Year) data. The irradiance data is mostly modeled for all the TMY locations and originally based on data from 26 U.S. sites. The other climatological data in the TMY is updated once in a while for the several hundred locations in the TMY data base. It's been substantially revised over the years That's what the PVWatts and other models commonly use for irradiance modeling.
Info on the TMY database, its history an updates is available per www.nsrdb.nrel.gov/data-sets/tmy.

Thanks!
My 6637 is lower than your 7159 because I threw in a couple percent of loss for presumed degradation ...

My programming goes back 60+ years to start with Fortran on an IBM 1620 desk (literally) computer that probably had far less power than our phones today.

Here is a version of my first graph, with an overlay of the 6637 PVWatts curve in weird blue-green.
PVWatts overlay.jpg
And a year-by-year history (with linear trend line). We did some serious panel washing a couple of years ago. Very hot temps in 2019 & 2020.
Trend.jpg

Ahhhh !!!, The joys of the 1620, machine language and punch cards.

Use a 10% system loss parameter and don't try to second guess the model. I adjusted that system loss parameter for my system and site so that the PVWatts annual output matched the annual SAM output for my system including shading and other stuff SAM can handle. That adjusted (PVWatts) system loss parameter (with 3.5 % in it for a shading loss) is 12.8 %.
The actual system running 365 day output matched that output within ~ +/- about 5% over the 10+ years of operation from 10/13/2013 to the present.
Separately, from array insitu measurements since then I've estimated the annual panel output degradation at somewhere between 1.0 and 1.3 W/panel-yr. (~ 0.3 to 0.4%/yr. based on STC panel rating) at what seems to be a pretty steady rate.