SolarEdge POWER Accuracy? 11,900 AC Watts from 12,000 DC array

[Poway]

During Array Test.

SolarEdge:
AC POWER generation up to 11,900 watts
Panel DC Rating 12,000 watts (40xLG300)
Inverter rating 11,400 Watts AC Nominal (12,000 Max)

I thought this number looked too good and remembered that others had questioned the numbers from SolarEdge. Went back and dug up the following:

James_King and Sensij had been looking into ENERGY generation accuracy for SolarEdge

https://docs.google.com/spreadsheets/d/1GxJF6ut_9q4rls7Ai1IP94orpc2fBVixEQqiWlmC7uM/edit?usp=sharing

Results SolarEdge energy generation up to 5% high

Then I looked at the WattVision data. It too was showing 11,900 watts. And I would tend to believe the WattVision data as it comes directly from my SDGE meter via Rainforest Eagle

I'm fairly convinced that the 11,900 watts is real.
Originally when I read that Sensij and others were saying that SolarEdge was reporting high that this applied to all aspects (Power, Energy...). But after rereading I see that the focus was on energy generation numbers.

Do forum members believe that the POWER (vs. ENERGY) numbers from SolarEdge are inaccurate?

[sensij]

I believe the SolarEdge power data to be much more accurate than the energy data. You can see in this post an example of both compared to revenue grade data.

Yesterday, my array sustained production at over 90% kWp/kW, a new high, with peaks over 93%. Peaks of over 96% kWp/kW were observed in the even cooler weather over a week ago. Cloud effect, cool temps, and sun position create the opportunity for some very high spikes this time of year; several systems in Team San Diego had peaks into the mid 90's. With the optimized orientation and better cooling of your ground mounted array, I would expect that your system will outperform mine (and nearly all others) in most conditions.

[Poway]

I believe the SolarEdge power data to be much more accurate than the energy data. You can see in this post an example of both compared to revenue grade data...

Thanks Again. This forum keeps saving me time!

[makarowski]

FWIW:

my logging on pvoutput shows some real wacky spikes.... guessing due to the sampling frequency (I don't think it was just the clouds going by) ... what ever it is, I think there is definitely some room for improvement, as I can't really take any of the individual numbers ... only averages? maybe some timelag between my meter and solaredge data (should I tweak my offset)? you can see "inverse" spikes in my consumption data...

I have a similar setup as the OP: solaredge, eagle reading my sdg&e meter, uploading to wattvision (may skip this step as I see pvoutput can read directly from eagle?) then to pvoutput...

my system from yesterday ....

http://pvoutput.org/intraday.jsp?id=38005&sid=35075&dt=20150518&gs=3&m=0

Cheers
BMak

[sensij]

The spikes in generation are real. As you suspect, the spikes in consumption are due to mismatches in synchonization between SolarEdge and the Eagle. This is partially due to the fact that SolarEdge uses a forward averaging scheme... data at 11:00 represent the average of 11:00, 11:01, 11:02, 11:03, 11:04. The data from Wattvision (or the Eagle directly, if you go that way) would be a 5 min average from 10:58, 10:59, 11:00, 11:01, and 11:02.

Solving this is hard.... even if you could offset the averages so they are looking at the same time, the power sampling is still not likely to line up well enough to make a smooth chart. A better approach would be to use energy counters such that for each period, Avg Power = (kWh end - kWh beginning) / time. Unfortunately, the Eagle does not report energy on a predictable interval, and SE energy is inaccurate (the subject of the thread). I've been working on how to do this with an external meter, but haven't really got it figured out yet.

The Eagle side of things could be helped by some programming that continuously sums the 7.5 s power readings... this should end up close to the true energy metered. Neither Rainforest Automation nor Wattvision do this now, and PVOutput is unlikely to take that on either. Definitely an opportunity for some 3rd party app development here.

The SE side of the equation is a dead end, as far as I can tell. Without access to data more often than 5 min, and with no explanation of their averaging scheme, it will be extremely difficult to ever generate a spike free gross consumption chart using the inverter data for the generation source.

[J.P.M.]

FWIW:

my logging on pvoutput shows some real wacky spikes.... guessing due to the sampling frequency (I don't think it was just the clouds going by) ... what ever it is, I think there is definitely some room for improvement, as I can't really take any of the individual numbers ... only averages? maybe some timelag between my meter and solaredge data (should I tweak my offset)? you can see "inverse" spikes in my consumption data...

I have a similar setup as the OP: solaredge, eagle reading my sdg&e meter, uploading to wattvision (may skip this step as I see pvoutput can read directly from eagle?) then to pvoutput...

my system from yesterday ....

http://pvoutput.org/intraday.jsp?id=38005&sid=35075&dt=20150518&gs=3&m=0

Cheers
BMak

Yesterday's sun in So. CA was not exactly what you'd call even quasi steady state. Is that what you are referring to ?

[makarowski]

Yesterday's sun in So. CA was not exactly what you'd call even quasi steady state. Is that what you are referring to ?

Its more of the consumption spikes / errors... (agree the generation spikes are realistic, as sensij mentioned above)...
I see a few "0" consumption numbers for my house that has way too many electronics (5 Wi-Fi routers/repeaters/AP's/switches, a few NAS, to say the least) ... I have a pretty high baseline...

Cheers
BMak

[Poway]

FWIW: my logging on pvoutput shows some real wacky spikes.... guessing due to the sampling frequency (I don't think it was just the clouds going by) ... what ever it is, I think there is definitely some room for improvement, as I can't really take any of the individual numbers ...

Sensij has his "Net Delay" set to 15 minutes. I currently have mine set to 0 min (default). And have similar glitches
I'm hoping that It will improve glitches somewhat when I update (may try 10 min since I have my polling interval set to 10 min)
My Fine tuning will have to wait till I get PTO

See attachment from Sensij

PVO Settings2.JPG
"The net delay setting is something of a mystery... 15 min makes sense, but I don't see much difference at 5 min." <== Sensij

What was your "Net Delay" set to??

[J.P.M.]

Its more of the consumption spikes / errors... (agree the generation spikes are realistic, as sensij mentioned above)...
I see a few "0" consumption numbers for my house that has way too many electronics (5 Wi-Fi routers/repeaters/AP's/switches, a few NAS, to say the least) ... I have a pretty high baseline...

Cheers
BMak

Understood. Thank you.

[makarowski]

What was your "Net Delay" set to??

My net delay was 15 min...for fun I changed it back to 0 ..And will see how it looks over the next few days... I will be looking for those false 0 usage data points


Cheers
BMak

[rhedayi]

sunday was funky, at one point my 5.2kw system was reading 5600 !!!

[russ]

sunday was funky, at one point my 5.2kw system was reading 5600 !!!

Some clouds?

[thejq]

sunday was funky, at one point my 5.2kw system was reading 5600 !!!

Haha, you want to see something really funky? A few days ago I saw 5.32KW AC from my 4.8KW system.

Capture3.JPG

I'm sure it's a combination of cloud edge effect + SE SW glitch. It's no way physically it can happen in any sustainable fashion. So I wouldn't pay too much attention to instantaneous measurements, especially in a partially cloudy day.

[russ]

I wouldn't pay too much attention to instantaneous measurements, especially in a partially cloudy day.

You got it - a meaningless number really.

[SunEagle]

Haha, you want to see something really funky? A few days ago I saw 5.32KW AC from my 4.8KW system.

[ATTACH]6748[/ATTACH]

I'm sure it's a combination of cloud edge effect + SE SW glitch. It's no way physically it can happen in any sustainable fashion. So I wouldn't pay too much attention to instantaneous measurements, especially in a partially cloudy day.

Is that 5.32kw a "peak" reading for generation over few seconds or over a longer period of time?

It's like using a meter to measure your kW load where you see a "peak" kW when the AC unit comes on but the actual continuous load is much smaller after the initial start.