Welcome to Tesla Motors Club
Discuss Tesla's Model S, Model 3, Model X, Model Y, Cybertruck, Roadster and More.
Register

Gateway 2 and Smart Meter accuracy

This site may earn commission on affiliate links.
I first posted this question on the UK Forum as it concerns a problem with a new UK Smart Meter, which may be over-reading. However, I guess some of you guys in the Sates will have a lot more information about how the Gateway 2 monitors power flow and the relative accuracies of the Gateway and Smart Meter data.

I have had a Powerwall 2 installed since July. I have an existing Sunpower/Solar Edge solar setup, an Tesla Model S and, until a few days ago a mechanical drum type electricity meter. As well as the Tesla Powerwall app, I am logging various parameters via the PVOutput site.
RF5PW 3.924kW | Live Output
Don't ask me why there are gaps in the data - it seems to happen sometimes!
The resolution of the Tesla Powerwall app is 0.1kW. The resolution (not accuracy!) of the PVOutput data is 1 watt (0.001kW). PVOutput is reading the Gateway data directly, but we don't know how accurate the Gateway monitors power flow.

I have dual tariff electricity the Powerwall 2 is set to Advanced Cost Saving mode, so it only charges from solar during the peak period, but can share from the grid or solar during the night period. The Model S has been connected directly to the Gateway input side so that it never charges from the Powerwall, which would deplete it very quickly!

Even without any solar input (it's rather cloudy here in the little ol' UK right now) the single Powerwall 2 has sufficient capacity to satisfy the whole normal house demand with some spare capacity during the peak rate period, provided it is near to fully charged to start with. Since installation in July, on most days the Powerwall has supplied all the peak demand. The old electricity meter, which had a resolution of 0.01kW, hardly ever moved. So measured consumption at Peak rate was maybe 0.1kW maximum. Both this electricity meter and the Powerwall/PVOutput data agreed with each other. Most times there was no peak demand from the grid, unless the Powerwall ran out of charge before the end of the Peak period. I am aware that the Gateway periodically draws a small current from , and also puts back a small current to the grid. This is presumably for calibration purposes and only amounts to bries 0.1kW flow or less. It doesn't measurably affect the overall power demand from the grid.

The new Smart Meter has both a digital display of kWh consumed to the nearest kW, and a Chameleon remote display showing all sorts of useless info like the cost etc, but also the current power consumption with a resolution of 0.001kW, in other words it will display a change of 1 watt. The PVOutput 10 minute live data has been displaying to 0.01KW, but I have just realised that this is configurable and have changed it to display to 3 decimal places- 1 watt, like the Chameleon display.

As soon as the new Smart Meter was installed I noticed that the Chameleon was showing a fairly constant power consumption of about 0.1kW, this while the Powerwall was supposed to be supplying all the house damand. This has varied up to 30% from day to day but shows no sign of reducing, and the meter is clocking up 1-2kW peak demand every day - that amounts to more than £150 per year - Not what I want!

I have tried turning everything off, switching off the consumer unit main switches, the solar and the Model S charger, so that the house demand was ABSOLUTELY ZERO. This had little effect on the 0.1kW demand shown on the Chameleon display. I then turned off the Powerwall and the Chameleon display dropped to ZERO. Then I turned on the Powerwall again and the Chameleon showed 8-11W demand, possible I suppose, but certainly not as high as the 0.1kW normally shown, which reoccurred once I switched on the consumer units.

I have made an initial call to my energy provider to flag up my concern. The lady didn't seem to be too clued up on Smart Meters, let alone home batteries! Anyway we agreed to monitor it for a couple of weeks so that they could get more data.

My questions are:

1. How does the Gateway actually monitor power flow and to what accuracy? Presumably it is a direct measurement transducer, rather than the rather inaccurate remote clamp on devices you can buy for $20.

2. What accuracy, at very low power flow, should we expect from the Gateway monitoring and a Smart Meter?

The Gateway Spec states an AC meter revenue accuracy of +/-0.02%. Quite what that represents I have no idea!

Having been involved in calibration of LVDT's etc I am well aware that measurement of very small power flows, displacements etc tend to be rather less accurate than the full scale accuracy of any monitoring device.

Of course, the old meter may have been significantly inaccurate at low power levels, however, because it agreed with the Gateway data I tend to believe that the new meter is over-reading unless I get accurately measured data verifiable to National Standards.
 
The Gateways in the US use current transformers (CTs) to measure the grid and solar flows. The PW has its own current measurement, but I'm not sure how it is done. The resolution is 1W, but the accuracy is not that good. I've seen a disconnected solar circuit report 5W being produced, with the circuit breaker open! CTs have lower accuracy at lower currents, stray electrical fields interfere and provide false readings. But I've not see accuracy errors of 0.1kW, that is a lot. It's usually under 10W.

My smart meters show 1W resolution also, and track the PW grid numbers quite closely. During the peak with the Powerwall providing energy, there is some small back and forth to the grid as loads shift and the inverter catches up, but not a large accumulation one way or the other. In my six hour peak, it's usually less than 0.1kWh total. Is that what you're seeing?
 
I can see in my intraday charts and totals in the Tesla app that the Powerwall is not offsetting consumption to zero and it is not charging from 100% of the solar generation. If I had to guess, I would say that because the system is not allowed to export battery energy to the grid and not allowed to charge from the grid in the USA, the system is erring on the conservative side and allowing a little solar to go to the household load while the PWs charge and allowing a small grid consumption while the PWs are discharging. I have not looked at my SmartMeter directly while the consumption is supposed to be zeroed out to see what the utility meter says.
 
If you look at my PVOutput data (extended data = right-most blob under the date selector + live data) you'll see that there's always a bit of energy flowing either from or to the grid. I presume this shows that, when online, the PW is unable to completely accurately match its output to the demand. The smart meter only reports the import which, in my experience (since March 2018) typically comes to 1kWh per day when there is no off-peak charging of the battery. This amount includes both the discrepancy in the battery output and occasions, usually of short duration such as boiling the kettle, when domestic demand exceeds the output of battery + solar generation. You should be able to press a button on the meter itself and it will scroll through more data including the amount exported.

PS: I like that particular PVOutput display as it clearly shows the off-peak charging and the battery State of Charge on the same screen.
 
Thanks guys, that's all useful input. @power.saver, yes, I'm seeing about 0.1kW load all the time on the smart meter. Chameleon remote display. That translates to 1-2kWh per day peak rate (17 hours) and agrees with the number of kWh accumulated on the smart meter itself. The kWh meter only displays to 1kWh resolution. If, as you suggest, there was a 10W error I would not be unhappy, but 100W is significant. Interesting that you are not allowed to charge from the grid in the USA. It's permitted in the UK and, with our rather limited solar generation it's the only way, charging from off-peak, that a Powerwall can be marginally viable economically.

I have downloaded today's PVOutput readings and processed the peak rate grid flow, logged every 10 minutes by PVOutput (I assume that's an average of much more frequent readings), if I sum the readings, which are sometimes positive and sometimes negative, I actually get 13 watts exported to the grid! Certainly not 1-2kW imported! Given that you are getting accuracy errors of less than 10W there's something wrong somewhere.

@JohnRatsey, I've followed your posts for some time. You are nearest to my situation geographically and and vour input has been most useful when I was analysing the likely performance prior to ordering. I agree that there always seems to be a bit of flow to/from the grid, as indicated in the para above. It seems to approximately balance out. It's interesting that you are recording a peak import in the order of 1kWh per day. That is not dissimilar to my 1.5-2kWh per day. I don't really see that boiling a kettle should result in a significant import form the grid. Even without any solar input, the 5kW max output capacity of the Powerwall should be well able to cope, together with other c=base home load. My home load never exceeds 5kWh except at night with 2 x 3kW immersion heaters, and I now sequence them so they don't com eon togherht, for other reasons. Certainly if you switch on a kettle there is a momentary import from the grid but I cannot see it adding up to 1 kWh per day and averaging the PVOutput readings (if they are anything like accurate) confirms this. I'll see if I can find any more data on the Smart meter itself, the A and B buttons are not particularly user friendly.

I copied the PVOutput display from a nearby friend whose PVOutput handle is LC, he's got 4 PW2's running a heat pump among other things, plus about 16kW of solar panels, so his input and output are a lot higher than ours. My solar is the standard 4kW array.

This problem is only apparent because those of us with home batteries should not be drawing ANY power from the grid when the battery is supplying the house load. Without a home battery it would not be at all obvious, just a slight increase in home load which most people don't monitor anyway and would not notice it. Eventually one would record an overall annual increase in metered power consumption and wonder why one was using 10% more power. There are documented cases of Smart Meters grossly over-reading so these are not an isolated cases.

My concern is that this problem has only surfaced with the installation of a new Smart Meter. Before that, everything agreed within reasonable error margins. If we really are consuming 1-2kWh per day as a result of running a home battery, fair enough., and an error of £10-20 would be quite acceptable. But my measured cost increase is significant - about 10% of the total normal household electricity bill. If the equipment is over-recording our consumption by 1-2kWh per day the electricity companies are effectively stealing money from us! That's illegal! I want to see real data verifiable to national standards. I have no idea if the Gateway is NAMS calibrated, but the Smart Meter certainly should be and there should be calibration certificates, traceable to the NPL at Kew in the case of the UK, to prove it. I bet getting that out of the electricity companies would be like drawing hens teeth! In any case, test equipment is generally supposed to be recalibrated at yearly intervals, which smart meters obviously are not. I think I'l,l ask to see the original calibration certificate for my Smart Meter and see what reaction I get!
 
I have downloaded today's PVOutput readings and processed the peak rate grid flow, logged every 10 minutes by PVOutput (I assume that's an average of much more frequent readings), if I sum the readings, which are sometimes positive and sometimes negative, I actually get 13 watts exported to the grid! Certainly not 1-2kW imported! Given that you are getting accuracy errors of less than 10W there's something wrong somewhere.

If you're using the default PVOutput Powerwall importer, one thing to note is that PVOutput is integrating the power numbers from the Powerwall, so you can get sampling errors. I believe the sampling is set to a one minute interval.

One thing you can do is to note down the accumulated energy numbers reported directly from the Powerwall in https://xxxx/api/meters/aggregates and cross-reference those against what the smart meter is reporting. That would at least eliminate the sampling error even if the CT measurements are not completely accurate.
 
@cwied mentions the local network address from which you can interrogate your gateway direct. Does anyone know why two of us with the backup gateway in the UK are getting a reported voltage of 110-130 V rather than the 230V we saw on the original Gateway (and is the non Us voltage standard)? Is there a setting which should be amended at gateway setup?
 
Hi,
I'm almost identical to Mike Miller (PW2 + GW2 since July, solar is only dumb 1kWp though, Secure SMETS1 smart meter originally from OVO but now with Octopus Go) so thought I'd just add some relevant points.
I remember that the accuracy of the GW2 was a supposedly big improvement.
My smart meter display flitsbetween +30W and - 30W usually during the day.
My weekly meter readings show I import about 1.2kWh and export 0.7kWh during peak (no solar export) which I guess is the PW2 using the grid to balance loads and then sending this back to net zero. These numbers include brief peaks over 5kWh from house loads, as we have a 4kW oven.
So to sum up, I think there is something amiss with your PW2 as when disconnected you say the smart meter reads zero?

Mark
 
@cwied, thanks for the heads upon the meters/aggregates output, I'll check that.

@MarkJ, the numbers you are getting for Peak meter readings sound like about what I would have expected. The Powerwall 2 and Gateway are in a nice comfortable built in unheated but insulated garage which rarely drops below 10 deg C. As you say, it does suggest that my meter is not telling the truth, Either that or there is something amiss with the Gateway.

The zero reading I obtained from the smart meter was with all house circuits isolated, solar shut down and the switch on the side of the Powerwall 2 in the OFF position. It's slightly odd that with everything but the Powerwall off and isolated there was same 0.1kW demand indicated on the smart meter Chameleon display, but after switching it off and then on again the display only showed 8-11W, fluctuating every fresh seconds as normal. Why the large drop? If it was possible to reliably and accurately test the real power flow in the cable from the meter to the gateway it would solve the issue. Even a test setup accurate to 20-30W at very low power flow would show up where the error is occurring.

Does anyone know if there are contractors in the Gateway that physically isolate the grid when the Powerwall is supplying the house, or is it all done by some electronic balancing system? I did hear clunks from the Gateway when I was going through the demand shutdown test.

I've already discussed the voltage discrepancy shown on PVOutput with simony and we are both puzzled by this.
 
Last edited:
@JohnRatsey, I've followed your posts for some time. You are nearest to my situation geographically and and vour input has been most useful when I was analysing the likely performance prior to ordering. I agree that there always seems to be a bit of flow to/from the grid, as indicated in the para above. It seems to approximately balance out. It's interesting that you are recording a peak import in the order of 1kWh per day. That is not dissimilar to my 1.5-2kWh per day. I don't really see that boiling a kettle should result in a significant import form the grid. Even without any solar input, the 5kW max output capacity of the Powerwall should be well able to cope, together with other c=base home load. My home load never exceeds 5kWh except at night with 2 x 3kW immersion heaters, and I now sequence them so they don't com eon togherht, for other reasons. Certainly if you switch on a kettle there is a momentary import from the grid but I cannot see it adding up to 1 kWh per day and averaging the PVOutput readings (if they are anything like accurate) confirms this. I'll see if I can find any more data on the Smart meter itself, the A and B buttons are not particularly user friendly.
I've downloaded the half-hourly Smart meter for September from Octopus using this facility and produced this graph:
Electricity from Grid September 2019.jpg
It shows that 0.6kWh/day is the typical consumption but the average is boosted by off-peak charging (12th Sep) and the electric shower (3rd, 15th and 18th).

I would note that my battery is currently restricted to a 16A output (~3.68kW) as permission hasn't been obtained from the district network operator to generate more than that (it's a bit daft that the restriction applies to batteries which shouldn't be exporting to the mains but that's the requirement). With electric cooking, washing machine, dishwasher and kettle as well as the electric shower it doesn't take much (eg cooker + kettle) to exceed the battery's output although the actual kWh used by the kettle isn't much. I suspect that about 0.5kWh/day represents a day when there isn't any intentional power from the mains.

My battery is in the garage which never gets very hot (it's on the north side of the house) and probably drops below 10C in cold weather (which reminds me to put a thermometer in there).
 
@power.saver - thanks. That confirms what I suspected. Turning the powerful off causes the contractors to open, with an audible clunk. In normal operation its all done electronically so the Powerwall isn't mechanically isolated from the grid when it' supplying power. I guess that makes sense as there is a bit of to-ing and fro-ing of power to from the grid for calibration purposes, but they should even out s I understand it and as the PVOutput data indicated.

John, thanks, it looks like your 0.6kWh per day is largely a result of occasional higher than 16A demand. I don't have that problem since mine has been approved for the full 5kW output. My solar was already occasionally supplying well over 3kW to the grid on the best days. In any case your 0.6kWh is much lower than my recorded 1.5-2.0kWh per day. It rather confirms that something is wrong with my meter. Is your smart meter a SMETS 1 or 2 meter? I would have thought that there shouldn't be much difference in accuracy between the two though as they are both recent state of the art.
 
John, thanks, it looks like your 0.6kWh per day is largely a result of occasional higher than 16A demand. I don't have that problem since mine has been approved for the full 5kW output. My solar was already occasionally supplying well over 3kW to the grid on the best days. In any case your 0.6kWh is much lower than my recorded 1.5-2.0kWh per day. It rather confirms that something is wrong with my meter. Is your smart meter a SMETS 1 or 2 meter? I would have thought that there shouldn't be much difference in accuracy between the two though as they are both recent state of the art.
I interpret the numbers differently. About 0.5kWh/day may be due to the small imbalances between Powerwall output (+ solar when available) and the household demand. Higher daily totals include the demand spikes. Visually, the mismatch inflows and outflows are roughly balanced, the meter only reports the inflows. Outflows are part of the export total (along with solar generation) but are priced on a much lower tariff. However, with the imbalance being typically being 20W, that amount x 12hrs is only 0.24kWh so accuracy of measurement might also be a factor.

My meter is SMETS1 maybe 4 years old. The remote display has a low quality greyscale display. And the garage temperature is about 7C this morning after a chilly, but not freezing, night.
 
My v1 Gateway has a similar ~20W average offset. I see it in both channels, so it manifests in slightly different ways. While the instantaneous grid draw reported by the GW API is constantly bouncing between slightly positive and slightly negative either when the house is running off of the PW2 overnight (no spikes that get anywhere near 5kW in my house) or all remaining solar is going to recharge the PW2, the average works out to ~20W drawn from the grid (and the cumulative energy input/output readings from the GW also reflect this), so I tend to rack up 0.1-0.2kWh/day even though I should be running 100% off solar+PW2. My PG&E SmartMeter data (via a Rainforest Eagle) matches what the GW reports, ~20W continuous instead of 0W. Like someone else mentioned, it seems most likely that this is to ensure any error is in the direction of not exporting power from the PW to the grid (while there are clearly moments where a bit 'leaks out', it's made up for by a bit of extra grid consumption).

I see the same offset in solar production, which I find a bit more strange. When the solar is in night mode, my GW reports a continuous ~20W solar production. So before the sun even comes up in the morning the app might already be reporting 0.1kWh of solar production. But if I throw the solar AC breakers so that the wires the CTs monitor are de-energized, the GW reading goes to 0, and when the inverters start up in the morning or are about to go to night mode, the reading does get all the way to 0 before it returns to the ~20W value. As a result, my script that mines the data collected each day for PVOutput ignores any readings between 10-30W before the first reading outside this range (which has always been <10W), and similarly ignores any readings after the last one outside the 10-30W range each day, to remove this excess (false) production, which generally results in a total production ~0.2kWh less per-day than the Tesla app reports.

So given that my Neurio (has anyone ever established what is actually inside the GW2, i.e. opened it up?) can read 0 when the wires have no voltage, it seems like perhaps it's some 'noise' picked-up by the AC signal even when there is no net current flow (of course the inverters should represent a slight draw when in night mode, since there's still some circuitry in them powered by the AC side). I tried several times to move wires around to ensure there was nothing else near my CTs or their wires running back to the GW that might induce noise, but nothing seemed to have any impact on these ~20W offsets.
 
  • Like
Reactions: JohnRatsey
There are some pertinent comments here about firmware 1.41.2 reducing the mains energy consumption. That's not sometime I'll be able to verify until the winter half of the year is over and there's no need for off-peak charging of the battery.
 
Simon has pointed out to me that because my Tesla charger is wired on the Non-backup side of the Gateway, it could be drawing power independent of the gateway and house on the backup side. I'll check that out when I get back home in a few days. It still wouldn't explain why the old meter agreed with the Powerwall and PVOutput data whereas the new Smart Meter doesn't.
 
After doing a lot of testing, turning off the house connections etc, I have established that there does not appear to be anything wrong with the Smart Meter. It's the Powerwall inverter and conditioning gear that is drawing a higher than normal power from the grid while supplying the house. I have spoken to Test Energy, who confirm that the specification figure is a maximum of 500Wh per day ( = about 20W average instantaneous power) if not charging from the grid. I'm drawing 3 times that, 1.5kWh. I have been asked to record my meter readings every day for 7 days and report them to confirm that the Powerwall is out pf spec. Hopefully they will then do something about it.

It's very difficult to measure Total power, via AC current due to the problem of phase difference and power factor, You can put digital multimeter probes across the main Gateway fuse and pull the fuse to measure AC current, and hence calculate the power in watts, but this only gives apparent power, which will be much larger than real power because the Powerwall almost certainly uses a switched mode power supply. I estimated power factor of 25% but I cannot measure that so I gave up in the end.
 
Last edited: