Finally installing PV on my roof in LA: day 1

[SunEagle]

That is good news Dan. Glad your system has been approved and is now producing for you.

[jetsurgn]

did your city inspector crawl up on your roof?

[DanKegel]

Yup. He and the installer and I had a fun conversation about bifacial panels. The inspector had seen them just once before, on a carport.

[DanKegel]

They say they'll be here to "install" the monitoring on Friday (hopefully, that just means pressing a button or two), so I'll make sure to work from home that day... might need to physically prevent them from installing zigbee and pulling out the hardwired ethernet

OK, here it is, the newbie's money shot: the first full day of output, measured from 7pm to 7pm, was 53 kWh.
As the faceplate DC power is 8.95 kW, that's 53 / 8.95 = 5.9 kWh AC / kW DC, not bad.

pvwatts for zipcode 90018 for today with a 8.95 kW premium panel system, pointed south (half of mine is pointed west, but hey), with 2.5% system losses, was only 40kW, so it's doing well compared to that. [Edit: this is not very meaningful, as pvwatts is based on a typical year, not on the average of a bunch of years, so it may have been thinking yesterday was cloudy in that typical year.]

High peak (1-5pm), low peak (8-10am,5-8pm), base (8pm-10am), and total net energy in kWh for the day were
-23 -18 10 -30
so my house used 23 kWh net (I think I ran the laundry and dishwasher, and probably left the old computer on).

Now I just have to switch a few more things in my life from fossil fuel to electricity.
(Water heater seems like the cheapest way to get started on that.... and a heat pump so our bedrooms and home office aren't so nasty hot in summer.
Have to wait for the to dust settle a bit from the reroofing first, though.)

[sensij]

Dan, can you ping the inverter? If so, you don't really need them to set up monitoring, you can do it yourself (or I can help) and you can keep full control.

For goodness sakes, please remove the line where you compared your *daily* output to PVWatts. That is not how PVWatts works... you've been around here long enough to know that!

[DanKegel]

Dan, can you ping the inverter? If so, you don't really need them to set up monitoring, you can do it yourself (or I can help) and you can keep full control.
For goodness sakes, please remove the line where you compared your *daily* output to PVWatts. That is not how PVWatts works... you've been around here long enough to know that!

Yup, I can ping it. (If you can walk me through setting it up, I'd be grateful.)

Yeah, pvwatts is an underestimate, it factors in bad weather, right?
I added up the hourly output it predicted for today.
Since today was good weather, you'd expect pvwatts to be lower than observed.
I don't think it's nutso to use it as a datapoint. Given that I know its limitations, I figure it's fair to show pvwatts' prediction. Do you think it'll confuse people reading the post later?

[sensij]

If you look at the PVWatts hourly output, you might be able to find a clear day plus or minus a couple days from today. Just sum up the daily totals and the clear days should jump out at you. That would be a more useful datapoint to consider than a datapoint that had an arbitrary* amount of weather added in. (*not really arbitrary, read more about how TMY files are created for more info).

Also, email sent.

[DanKegel]

Live & learn! I didn't realize pvwatts was based on a typical year rather than an average year, thought it'd be smoothed.

https://en.wikipedia.org/wiki/Typical_meteorological_year

Thanks!

[SunEagle]

Unfortunately PVwatts while helpful should not be relied on to provide actual daily or even monthly production data. It is more of a guide line based on previous years of weather data.

Too bad it can't predict the future weather.

[J.P.M.]

Live & learn! I didn't realize pvwatts was based on a typical year rather than an average year, thought it'd be smoothed.

https://en.wikipedia.org/wiki/Typical_meteorological_year

Thanks!

Read the help/info screens w/PVWatts, and keep in mind that it is a model for long term average estimated output and not much else. Then, read through the TMY manual for guidance. Then, if still interested, check out something called "Solaranywhere", which is an alternate insolation source to the TMY data.

I believe mine was the first reference on this forum to using PVWatts hourly output and then looking for a clear day at/around any actual clear day(s) to get a SWAG of system performance and/or a loose handle on system losses. Kind of a bastard method of turning the crank backwards, but perhaps useful for getting a feel for system performance or a better/more realistic est. of the sys. loss parameter to use w/PVWatts for a specific system.

[DanKegel]

Finally got access to monitoring. Still don't have the API key, so I can't yet export to pvoutput.org.
I see they didn't set the timezone on the inverter, so it says funky things like I have some output after midnight
I asked them nicely to set that.... or I suppose I could open the case and set it myself, but then they could point fingers at me, so I probably won't.

It's nice to see the per-panel output, even if it's not supposed to be terribly accurate.
It says south-facing panels have generated about 20% more energy than the west-facing ones, no big surprise.
The LG310s seem to be a bit higher output than the GxB300's, no big surprise.
Peak power's been about 7.4kW most days (out of 8.9kW, not bad for half the array pointing west & half pointing south, I suppose).

Oddly, four of the south-facing panels are on the same string as the west-facing panels. I guess that doesn't matter given the optimizers.

I haven't gone up and covered backsides or frontsides of the bifacial panels yet, maybe this weekend I'll try a bit of that.

[J.P.M.]

I haven't gone up and covered backsides or frontsides of the bifacial panels yet, maybe this weekend I'll try a bit of that.

A respectful suggestion: Use an IR thermometer to account for temp. diff. covered vs. uncovered conditions. attaching surface temp. devices. When/if you do that keep in mind that covering a panel either front or back will change its temperature vs. an uncovered panel condition. That temp. diff. will change its efficiencies, both the instantaneous and time integrated values compared to uncovered adjacent panels.

I'd SWAG that covering the front side would kill most of the output but lower the panel temp. Overall, things would get hard to predict partly due to unknown low level irradiance efficiency characteristics of the panel.

On the other hand, covering the backside might actually increase the panel temp. and thus lower the overall efficiencies because covering the back side will probably might inhibit convective heat loss and increase panel temp. more than the relatively small value of the lost irradiance to the back face would have increased the panel temp.

Given the likely small values of the temperature differences involved, and the imprecision and inaccuracies of temp. measurement techniques probably available, and without a pyranometer and some consistent way to convert GHI to POA irradiance, it'll be interesting for me to read what you find.

Every 1 deg. C. error in temp. est. or measurement results in something like ~ 0.4% to 0.5% error in eff. est. or ~ .01% error in raw panel eff. from, say, very roughly 0.18 to 0.179 or 0.181 VERY approx. That's getting down into signal and instrument mud territory.

[DanKegel]

Yep, I'm worried about temperature effects.

And I'm worried about solaredge monitoring accuracy.

I might start by just waiting a week or two, then extract data for each panel for each day from the logs,
write a filter to skip days without full sun,
grab ambient temperature data from the nearest weather station,
and see if I can get a repeatable and stable measure of corrected peak and total power in full sun.
If I get ambitious, I might add temperature sensors glued to the backsides of the panels.

I expect doing this right is harder than it sounds.

[sensij]

The voltage at which the panel operates will give you some insight into the temperature. Even if it isn't accurate on an absolute scale, comparing a single panel's data to itself should be reasonably trustworthy. If the voltage moves much when either side of the panel is covered, that would mostly indicate a temperature effect... there is a small shift from the new mppt point, but temperature should dominate.

[DanKegel]

Thanks, I'll look for that effect.