Big discrepancy between Powerwall data with SCE usage data

[Redhill_qik]

Thanks for the thoughtful responses. The assumptions underlying the "5/60" method and my method are different, but both are built on assumptions. The "5/60" method assumes that the power is constant within each 5 minute interval. My method assumes that the power varies within the range defined by each group of 5 minute intervals within each 15 minute cluster. I am skeptical about any method that relies on a single randomly selected day.

You appear to be assuming that the SCE and Tesla data is an instant number at the timestamp and it isn't. The reported number is based on second level or sub-second measurements that are added up over the time period (5 mins or 15 mins) and then reported as an aggregate. It might have been 10 kW for a minute and 1 kW for 4 mins and it would report either 2.8 kWh (Tesla) or 0.23 kWH (SCE). For the SCE data you just add the numbers and for the Tesla data you have to factor in the time to kWh.

Oh and BTW, I asked SCE to come out and check the meter. I didn't figure that there was a problem with the meter, but it's a part of their standard deflection procedure. No surprise... the meter is fine. The problem... and yes, I still contend that there is one... is downstream of the meter.

The SCE meter is a revenue grade meter that has gone through extensive testing and is regulated and only has to measure the flow a single point. The Tesla data is based on retail level current transformer and the results are often wrong due to incorrect installation as there are multiple points involved. Yes, there can and have been issues with bad utility smart meters, but that is rarity.

I looked at the data for 10-29, which is really hard to decipher in the tables, and SCE is reporting imports from 7:15am to 2:15pm and again from 5:00pm to 10:30pm. All of these numbers are pretty consistent at 1.98kWh, so maybe an EV charging? And then SCE is reporting exports from 2:00pm to 5:00pm. This seems consistent with the solar charging the Powerwall until it is full and then excess goes to the grid and then higher use in the evening.

If you want to prove that something is wrong you need to focus on the meter instant readouts. Is that matching with the Tesla app? If not then why isn't it matching.

 

[Antonc]

You appear to be assuming that the SCE and Tesla data is an instant number at the timestamp and it isn't. The reported number is based on second level or sub-second measurements that are added up over the time period (5 mins or 15 mins) and then reported as an aggregate. It might have been 10 kW for a minute and 1 kW for 4 mins and it would report either 2.8 kWh (Tesla) or 0.23 kWH (SCE). For the SCE data you just add the numbers and for the Tesla data you have to factor in the time to kWh.

The SCE meter is a revenue grade meter that has gone through extensive testing and is regulated and only has to measure the flow a single point. The Tesla data is based on retail level current transformer and the results are often wrong due to incorrect installation as there are multiple points involved. Yes, there can and have been issues with bad utility smart meters, but that is rarity.

I looked at the data for 10-29, which is really hard to decipher in the tables, and SCE is reporting imports from 7:15am to 2:15pm and again from 5:00pm to 10:30pm. All of these numbers are pretty consistent at 1.98kWh, so maybe an EV charging? And then SCE is reporting exports from 2:00pm to 5:00pm. This seems consistent with the solar charging the Powerwall until it is full and then excess goes to the grid and then higher use in the evening.

If you want to prove that something is wrong you need to focus on the meter instant readouts. Is that matching with the Tesla app? If not then why isn't it matching.

You are correct that I was assuming that the Tesla data was an instant sample at the timestamp. I didn't make that assumption up out of thin air, so I'll have to refresh my memory about where I found that. There are some curious points that I don't seem to be communicating effectively because they keep getting ignored, the biggest one being: If my method is wrong, then why does it match SCE data (and your method) almost perfectly on most days?
Also, our cars are programmed to charge during off-peak hours, so no there's no an unaccounted for load.
I am also puzzled why the SCE data is viewed as unassailable gospel when it shows things like zero solar production when the batteries are full, the sun is shining and consumption is less than solar input? I've still seen no substantive evidence that there is a significant error in my methods. (I did accidentally double-count the overall totals for the 3 days where I compared our two methods, but that doesn't significantly change the overall picture.)

 

[Redhill_qik]

If my method is wrong, then why does it match SCE data (and your method) almost perfectly on most days?

Likely because there is a different load in your house after the meter that is not being measured correctly by the Tesla app. Tesla does not actually monitor the grid flow and it is creating that value out of the sources (solar and Powerwall discharging) and sinks (Powerwall charging, gateway loads, nurio CTs) that it does measure.

Also, our cars are programmed to charge during off-peak hours, so no there's no an unaccounted for load.

You say that, but are you 100% sure. If you are 100% sure, what makes you so sure? Do you have a line drawing that documents all of the connectivity and nodes for your system that you can post? You should have received this during the approval stage or it should be in your Tesla account documents.

I am also puzzled why the SCE data is viewed as unassailable gospel when it shows things like zero solar production when the batteries are full, the sun is shining and consumption is less than solar input? I've still seen no substantive evidence that there is a significant error in my methods. (I did accidentally double-count the overall totals for the 3 days where I compared our two methods, but that doesn't significantly change the overall picture.)

When I was looking at the data you posted it seemed like the system was running in self-powered mode as I was seeing the Powerwalls discharging in the morning and then recharging during the day which can account for why the solar power isn't being exported. The SCE meter has a very very simple job, measure the voltage, current and power continuously across a single point using a mass (millions of units) produced and regulated meter. Yes, a unit can fail and produce bad data, but that is far less common than the Tesla installer screwing up the orientation of one or more current sensors (CTs) or the configuration in the system. You want to trust in the Tesla data and I think your are putting your trust in the wrong place.

My own system had major sensor issues at install for circuits that were non-backed up (not in the gateway). If my A/C turned or I had Christmas lights up the the Tesla app data was wrong and I could tell because I was looking at the meter and the app at the same time with different loads. The A/C part was fixed, but it still doesn't register my Christmas lights, I think that I know why but I also don't want to deal with Tesla again on this issues. Should have had them fix it when they swapped my SolarEdge inverter.

 

[yblaser]

Looking at the data I would have to agree with @Redhill_qik that it seems to be an unmonitored load that is causing the discrepancy. I plotted the data below. Blue line is the "home" power draw as reported by Tesla. The orange line is the discrepancy between the SCE data and Tesla's grid data. For the vast majority of the time the SCE and Tesla agrees. It's just these bursts of 7.5-10 kW that seem to occur mostly during waking hours but not at the same exact time each day that the Tesla data isn't capturing. It seems like you do have something charging at 10 am each day that is being captured by the Tesla data. How many EVs do you have and do you have more than one EVSE?

There is an oddity in the data on 11/6 where for a 30 minute interval the SCE data shows that you are exporting ~6 kW while the Tesla data says you are exporting zero (it's cutoff on the graph below as I limited the graph to -2 kW).