PVoutput mismatch vs SolarEdge monitoring portal

[CharlieEscCA]

Anyone else seeing PVoutput showing lower generation values than SolarEdge portal? I'm seeing PVOutput showing about 2.7% below the SE portal.

[sensij]

SE Portal energy values are inflated. PVoutput data is based on the 5 min power values, which are more accurate. This is a known characteristic of Solaredge, confirmed with a revenue grade meter.

[CharlieEscCA]

Now I get it. When I had read before that the SolarEdge production values were high / inflated, I had assumed that it was the values that the inverter reports [which may also be the case]. And then I assumed that the data being retrieved by PVoutput data pulls / queries would be the same data used by the SE Portal -- thus expecting these values to by in sync.

Based on what you said, it would seem like both SE Portal and PVoutput are retrieving an instantaneous production rate number (xx.xx kWh) and then doing an integration (i.e. area under a production curve) to yield total production or perhaps taking the average between last read and current read times the interval between reads to calculate production for the period and then keeping a running total for the day.

I would have thought that the inverter itself would be keeping its running total (after all it should be able to integrate/sum the instantaneous production rate the best, accurate or not), and SE Portal and PVoutput would query this total (along with any other data it wants), and then report this total value retrieved from the inverter. Perhaps the SE Portal total is indeed this number from the inverter directly -- whereas PVoutput does it's own integration of data points.

In any event, this is all moot when in the end the only meter that counts is the utility's meter

[J.P.M.]

Now I get it. When I had read before that the SolarEdge production values were high / inflated, I had assumed that it was the values that the inverter reports [which may also be the case]. And then I assumed that the data being retrieved by PVoutput data pulls / queries would be the same data used by the SE Portal -- thus expecting these values to by in sync.

Based on what you said, it would seem like both SE Portal and PVoutput are retrieving an instantaneous production rate number (xx.xx kWh) and then doing an integration (i.e. area under a production curve) to yield total production or perhaps taking the average between last read and current read times the interval between reads to calculate production for the period and then keeping a running total for the day.

I would have thought that the inverter itself would be keeping its running total (after all it should be able to integrate/sum the instantaneous production rate the best, accurate or not), and SE Portal and PVoutput would query this total (along with any other data it wants), and then report this total value retrieved from the inverter. Perhaps the SE Portal total is indeed this number from the inverter directly -- whereas PVoutput does it's own integration of data points.

In any event, this is all moot when in the end the only meter that counts is the utility's meter

As you write, the meter that belongs to SDG & E is the one that counts and is the standard against which other devices reporting use and generation are usually measured.

FWIW, it's certainly possible to get a separate revenue grade meter and use that as a standard, but the meter that's used to charge you for power is probably as good in terms of absolute accuracy as any that's commonly available. With 2 revenue grade meters, if they don't have reasonable agreement, you're like the person with two watches who never knows what the actual correct time is.

I considered getting a revenue grade meter and may someday. In the meantime, and FWIW for those who are as eccentric as me about such things, I did an experiment on 2 separate occasions while on 2 extended vacation trips a few years ago to attempt a calibration (comparison) between reports of what the Sunpower equipment I have says what the equipment is producing and what the SDG & E meter thinks the system is producing. I think Ian may have done something similar with his stuff.

Both times, I read the SDG & E meter immediately prior to leaving (car packed and running in the driveway before sunrise). Prior to that, I had opened all the circuit breakers except those running the fridge and solar thermal water heater pump (and the PV array of course), and put kill-a-Watt meters on those two devices. When I returned 2 or 3 weeks later, I read the SDG & E meter immediately upon pulling in but before unlocking the house. A little math and kWh accounting with the kill-a Watt meters, the Sunpower monitor, the inverter output screen and the SDG &E meter left me with the result for both trials that the inverter screen data and the information from Sunpower agree pretty much dead nuts with one another. However, the difference between what the equipment says the array produced vs. what the SDG & E meter says the equipment produced different numbers. The first trial had the equipment reporting 1.0024 of the SDG & E meter, that is ~ 0.24 % high. The second trial had the equipment reporting 1.0078 of the SDG & E meter, that is ~ 0.78 % high. I average the two results and divide all the production and power output numbers from the array equipment by 1.0051, so that all equipment reporting numbers get multiplied by 1/1.0051 = .9949 of equipment reading before I use them.

If/when I get a revenue grade meter, I may find it does not agree with that ~ 0.5 % correction, but that .9949 multiplier gives results that seem to agree with other stuff I measure.

[sensij]

Yeah, what Solaredge is doing here is hard to explain. I've described it this way in the past:

SolarEdge inverters appear to have two internal data sources, one of which is calibrated about 3% higher than the other. The daily energy reported by the inverter uses the higher number, the live power reported uses the lower number. PVOutput's uploader is using the live power, and averages created from it. As a result, on a clear day, the energy generated for the day reported by PVOutput will be about 3% less than the energy reported by the SE portal.

By a couple methods (revenue grade meter, gross consumption calculated from generation and net consumption), it is clear that the power value is the more accurate of the two.

In other words, it isn't really the portal's fault, the inverter provides energy data and power data that don't match. At one point, PVOutput was going to switch to using the energy data only, and I convinced the developer (BB) to keep the power data as default, and let users choose energy if it was important to them to see agreement between the portal and PVOutput.

With respect to what J.P.M. writes, another way to verify accuracy of the revenue grade meter is to look at *calculated* consumption, by adding generation numbers (as reported by the RGM) to the net consumption seen by the poco meter. Generation varies with time, while true consumption can (in some cases) be fairly predictable over several hours as the same appliances cycle on and off. If there is a calibration error on the RGM (relative to the poco meter), it will show up as error in the calculated consumption... 1% error on 6000 W of generation would result in 60 W error on consumption, very noticeable when actual consumption is 200-300 W. The script I've written is very careful to keep the data polling as synchronized as possible, and I see less than 0.5% error in my setup. On very cloudy days, even polling net consumption once every 10 sec isn't always fast enough to keep additional synchronization error out, since conditions can rapidly fluctuate over any 5 min interval.

[GSV3MiaC]

I have a (revenue grade) feed in meter attached to the Solar Edge inverter, and this shows a 3% lower reading than the 'energy production' reported by Solar Edge's API .. I wonder if that 3% is what the inverter self-consumes when it is actually in production mode? The 5 minute data also has a power number, but that appears to be instantaneous power at some moment when the data is sent to the portal, so doesn't align with the Energy values (it correlates to an R^2 of .95 or so, but there are plenty of 5 minute points where recorded energy is 5X higher or lower than the power number would suggest).

I'm using excel/VBA to query the portal, record the date (if you look at it for the inverter, you can get 5 minute data, whereas everything else is 15 minute resolution), and then compare that with models of what I might expect (here in the UK .. 52 degrees N). The array is inset into a (black) slate roof and the cell temperatures regularly hit 50-60c (or slightly more) in full summer sun, which puts a serious dent in performance. It does look nice though. 8>.