Powerwall 2 and Smart Meter accuracy

I just posted this in the Questions section in reply to a topic about PVOutput accuracy, but i think it’s worth raising here too. I’ve also been discussing this problem on the TMC Tesla Energy forum. My Powerwall 2 was installed in July and my old mechanical meter generally recorded Zero peak power demand per day, except, rarely, when the battery ran out of charge. The Powerwall usually supplies all the Peak demand. I have dual tariff, Peak and night rates.

I had a new Smart meter installed on 30th October, with a remote display showing current power consumption to 10W resolution, and immediately noticed that there was a fairly constant demand of 60-100W. That doesn’t sound very much, but it’s 1.5kWh per day, over 500kWh Peak tariff per year = >;£150 - 20% of my electricity bill!

We have done fairly extensive testing (by turning everything off in sequence) and it appears that the Smart Meter is recording accurately and that the Powerwall itself is responsible for the demand. I am in discussion with Tesla Energy at present. They told me that the Spec was for a maximum daily demand of 500Wh so mine is way out of spec. I’ve just sent the requested 7 days of meter readings to them and am waiting with baited breath for the response!

One thing occurs to me - PVOutput grid flow shows continuous very small demands and exports, usually <10W. I understand that this is the Gateway calibrating things. Tesla also told me that the inverter and internal systems sometimes draws power from the grid. Why it can’t use it’s own power I’m not sure since it will operate off grid with no problem. However, the Smart Meter only records Import. It can theoretically record Export as well but that hasn’t been activated on my meter. In any case, it would be impossible to separate out the Solar export, which we get paid for, (assumed 50%) and Powerwall checking export, which we wouldn’t get paid for. So the meter may be recording the import side of the Gateway balancing algorithm stuff and not the return export.

Maybe the old meter was not sufficiently sensitive to record very low power demand although it should surely have recorded a couple of light bulb’s worth of power?

Has anyone else noticed excessive Powerwall grid demand while the battery should be supplying everything?

I have two PW2s originally on the non-backup UK gateway and now on the newer Gateway 2. I have always seen small back-and-forth flow with the grid, of the order of +/-10W. Looking at yesterday’s usage:

and after the cheap rate, the hourly consumption is recorded as of the order of 3Wh:

You can query the Tesla G/W directly for real time flow and side-step the pvoutput data for your own sanity :slight_smile: See GitHub - vloschiavo/powerwall2: Tesla Powerwall 2 - Local Gateway API documentation

Good luck

Thanks Peter, that’s very useful. Yesterday I did manage to download raw data using the api/meters/aggregtes call and was trying to decipher the results. Your Github link explains it quite well. My Site ‘instant power’ values are similar to yours, usually less that 10W and sometimes negative. That agrees with the PVOutput Grid Flow values I’m getting. It should do because it’s using the same data.

But it doesn’t explain why my Smart Meter is recording nearly 1.5kWh peak demand per day. The aggregates data from the Powerwall, averaged over a day, is close to zero, the positives and negatives, cancelling out. If I sum all the positives and all the negatives separately and convert them to cumulative Wh, it comes to about 100-200Wh positive and 50-100Wh negative over a complete 17 hour peak period. That doesn’t correlate with the Smart Meter’s 1.5kWh! Something is wrong somewhere. Even with the Smart Meter summing the demand only, not the export, either the Smart Meter is very inaccurate at low power flow rates or the Powerwall is drawing a demand that is not showing up in the api/meters/aggregates reports.

Since the values reported by the aggregates output are labelled “instant”, I assume they are not averaged values over the Status Interval. This is set to 5 minutes in my case. Even so, for that amount of data it should be fairly accurate.

Your smart meter may be wrong. It’s not impossible.

Can you live without power in the house for an hour? Maybe if you go out over the weekend for a shopping trip? Turn off all the supply on your consumer unit and see what the smart meter records over at least two 30 minute periods (to allow for timing etc.) …

You didn’t say if your gateway is the original non-backup or the newer backup one? If the latter, then you can also test that :slight_smile:

Also, on the non-backup gateway, have you checked and double checked your consumer unit to make sure nothing is wired on the wrong “side” of the PW2 grid CT (which is integrated in the backup one) ? An alarm system or other system that may have been added outside the consumer unit? Garden/outdoor or fridge circuit perhaps?

The amount you are seeing feels like a “fridge” to me and they are sometimes wired in before any RCD or other safety breakers to ensure 24 hour running in case of tripped circuits.

The api data is instant power, as you say, but there are registers that record the total since installation I think.

Hi Pete, thanks for your suggestions.

I thought the Smart Meter was in error at first, particularly since the old meter and the PVOutput grid flow figures agreed with each other. But, while I now still don’t discount an error in the Smart Meter, my testing suggests to me that it is reading fairly correctly. If it isn’t I’m hoping that Tesla will flag it up in their investigations.

It’s a new backup Gateway 2. During my testing I turned off everything, mainly by switching off my two consumer units and including the car charger on the non-backup side, and then the powerwall itself and pulling the main Gateway 100amp fuse, at which point the meter demand reading dropped to zero. Replacing the fuse induced an 8-11W demand, presumably the gateway electronics themselves. Then switching the Powerwall back on initially pulled about 300W, which slowly declined to the 60-100W constant figure. I guess this is the inverter control circuit drawing power. So I’m satisfied that it’s not recording power from any other source and the there is nothing else wired in on the non-backup side. The meter tails go straight to the gateway main fuse. The non-backup car charger is wired in to a CB inside the Gateway. There is also a CB for the Solar inverter, but that’s on the backup side and is for something other than the solar output, probably the electronic control circuit and data link. I did isolate this as well to make sure it wasn’t the culprit.

One thing I’m planning to try is to change over to Self Powered mode on the app during the day and see if that stops the drain, but I’ll what to see what Tesla say before doing that. It takes between half an hour and 2 hours for changes to the backup reserve to take effect so there’s probably a similar delay when changing to self powered mode. However, it might be a lot longer. When it was installed I was told it would take 48 hours for the advanced modes to appear and that is what happened. In self powered mode the Powerwall should be disconnected from the grid and power it’s inverter purely from the battery. Any thoughts on that? It’s not something I want to do regularly to save some demand. If it works like I think it should I don’t see why Tesla can’t program it to work more or less the same way.

I did test the current draw across the main fuse with a multimeter, before I remembered the thing about phase angle, apparent power and load factor with AC current measurement. The Powerwall probably uses a switched mode power supply. I got about 1.3amps which, if the 60-100W figure from my remote display is to be believed, would suggest a power factor of about 0.25. There clearly is some continuous power demand going on but to check the real value independently would require some rather expensive test equipment. We tried the same experiment on my friend’s 4 Powerwall setup, which has just been upgraded to a Backup Gateway 2. (during the day with the Powerwalls supplying all the house demand).The current measured for his 4 Powerwalls was about 4 times what I measured at my own main fuse, which seems to indicate that my reading is not unique. However, his old visible disc meter wasn’t rotating at all, probably not sensitive enough and may oscillate back and forth slightly if the demand is balanced by the export, but my smart meter won’t record export.

There do seem to be large energy_imported numbers in the raw api/meters/aggregates downloads that suggest totals since installation. I’ll try toting them up and comparing them with the meter readings since install. On that topic, does anyone know why the instant_average_voltages reported for the Site, Load and Solar sections (and displayed on PVOutput) are about 143 volts while the battery section shows 247v on mine?

I just checked the API data this morning against meter readings since Powerwall installed. The site:energy_imported figure does agree within 2kWh with the sum of peak and night rate meter readings over 4 months. The meter readings at initial install may have been taken a few hours after commissioning, but that’s quite encouraging.

Something sounds odd. Mine are similarly configured and show no real draw. I used to use self powered in the summer but as my Gateway 2 was upgraded only recently it’s been in Cost Saving mode pretty much all the time.

Let’s see what Tesla Tech have to say!

I tried changing to self powered mode for 10 hours yesterday and, rather surprisingly, it made no difference at all to the phantom grid demand while the Powerwall was supplying the house. I would have thought it should operate completely independently. There difinitely isn’t anything else causing the demand.During testing a while ago now I turned absolutely everything off. and the 60-100W demand was still there. Turning the Powerwall side switch off resulted in a drop to about 8W. The meter dropped to Zero when I pulled the 100amp fuse in the powerwall. as ir should do because there was nothing at all connected to the meter. Fuse back in - 8-11W (Gateway electronics drain?) Switching the Powerwall switch back on - 157W-152W-eventually dropping to 100W. It must be the Powerwall inverter and systems taking the power. That’s what the Tesla helpline said but it should be much less than that, max 500Wh per day. Maybe it always takes something from the grid if connected, regardless of the mode of operation. Maybe it’s set like that to keep maintaining the powerwall systems if the battery is empty - a bit like the 12 volt essential services battery in the Model S.

Keep hassling Tesla, something sounds off. Mine do not do this.

Thanks for the prod, @pgalbavy, I emailed a week a go and they asked if there was anything else drawing power during the day, answer -NO, my Model S on the non-backup side only charges at night rate, I\ve checked. Must give them another rattle.

Here’s something I’ve realised, and just posted on the TMC Tesla Energy forum.

One thing I have established from downloading the api raw output directly from the gateway is that when the Powerwall is supplying all the house power, the ‘phantom’ drain from the grid comprises a very small total power (instant_power), usually 5-10W, but very large reactive and apparent power, typically 100-200VA. That means a very low power factor, I think. Reactive power is wasted power which mainly goes up in heat and does no useful work, so it means that whatever in the Powerwall is drawing that power is very inefficient - not good environmentally. It also means that, since my new Smart Meter appears to be metering reactive power, which the old mechanical meters can’t do, I’m paying for this wasted power, but not being credited for anything it puts back as export, which is not metered. Maybe an electrical expert among you can confirm or refute my assessment. here’s part of a typical api download:


My nearby friend with 4 Powerwall 2’s just sent me his api output and it shows the same pattern - 4 times as large.

Why it can’t draw the power from the battery instead of the grid is a mystery to me. I can force it to do that by pulling the main 100amp CB, which disconnects it from the grid and puts it into backup mode.

And here is the reply. It rather confirms my suspicions about reactive power metering - don’t get a smart meter installed!

"Thank you for your reply.

From our investigation we have noticed that most of the times you only import energy from the grid overnight and since the solar is really poor the battery usually offsets your house loads during the peak hours.

I suspect that this discrepancy is due to the fact that your meter reads also the ‘reactive power’ while the App does not. The app shows only the ‘real power’.

To make this difference clear here it is a good and easy example:

The App only shows the Real power while I suspect your meter shows also the reactive power.

You can also check online the specific differences between the two types of “power”.

By the way, the system does not seems to record any import from the grid except the tolerated 0.4/0.5kW that basically is the amount of energy that the inside inverter uses to keep the modules DC to AC and AC to DC ready to be used.

I also suggest you to contact your energy supplier to check the meter and the installer of the battery to maybe do some checks on the CT clamp positioned on the site side.

I will anyway escalate the case to our Tech. Support to let the double check because I do not see this data discrepancy but maybe they can help me."

@pgalbavy, who is your energy supplier and what make of smart meter do you have? I am EDF with a Landis & Gyr smart meter. Your typical hourly consumption of 3Wh is what I would have expected, or hoped for anyway. Is it possible that your meter is reading only Total power while mine is factoring in Reactive power, which the Powerwall AC-DC converters seem to be rather hungry for? Very difficult to know without inside knowledge I guess. Perhaps I could ask my supplier.

Here’s a screenshot of my EDF power graph. During peak period it’s recording a consistent 0.08kWh. That is hugely higher than your 0.003kWh. At present I’m trying to engage in an intelligent discussion with EDF about whether my Landis & Gyr Smart Meter is recording reactive power. Your 0.003kWh looks consistent with the typical “site” average instant_power api figures produced by my Powerwall. Mine look like some sort of integration of real and reactive power. I haven’t yet got onto someone at EDF who knows what reactive power is…

Reactive power is the power that “passes through” reactive loads such as motors, transformers and PV inverters. This power is not consumed but is utilized to produce magnetic fields in the coils of these devices then is returned to the grid. There are losses which are due to the resistance of the coils they pass through and is dissipated a heat. Generally only heavy industrial users are charged for this type of power or are required to have “power factor” correcting devices to keep the voltage and current in phase making the power factor as close to 1 as possible by adding a capacitive load to the circuit. In the USA some power companies are beginning to charge for this type of power because they have to make up the losses they cause. If you maintain a power factor of .96 or above you may not be charged. Most homeowners don’t have enough reactive loading to produce power factors below .96.

I’ve now spoken to two departments at EDF, the Smart Meter department, who put me onto the Green Team (Solar). Unfortunately, none of them are electrical engineers and they don’t even know what reactive power is. They weren’t being unhelpful, they just don’t know - slightly worrying since EDF run the UK’s nuclear power stations! As far as they are concerned the meters just record energy consumed which they invoice. I think they are just administrators and accountants who don’t speak to engineers and couldn’t put me through to one. I could try putting in a complaint about my smart meter and see if it stirs up the beehive!

@iwsmiser, what is your opinion about the power factor of the powerwall inverters? From my api downloads, for instance, a typical reading is:


I calculate the power factor as 17/158 = 0.108.

That’s very low. Is my calc correct? If so, would that be typical for an AC-DC converter? PS - I have just read that SMPS’s typically exhibit a power factor of about 0.6.

What Tesla appears to be saying is that this grid demand is to keep the inverter ready for use if there is a power outage, or maybe with the DC-AC converter, ready to start charging when there is excess solar. I presume it might be keeping capacitors charged up. I’m not an expert on switched mode power supplies, if that’s what’s in there. That would mean a very low power factor by default?

What are the units given for reactive power? Since you didn’t give the units I assumed they were kVAR. If they are Volt-Amperes rather than kiloVolt-Amperes I wouldn’t be concerned at all. That amount of power is very small.
I monitor my power with CT’s on the mains and on the PV output. Here is a clip showing the instantaneous values:

Note that the units are Watt-hours and Volt-Amperes-Reactive. Also the reactive power at the inverter is -185 VAR. Check the units on your PW output.

1.5 kWh per day seems excessive but the 500 Wh per day number seems reasonable as that is about 20 watts continuous draw.

AIUI the gateway needs to draw a small amount of power to keep itself and the PowerWall AC inverter synced with the grids 50 hz frequency and monitor for power outages etc etc. you would only want the gateway drawing power from the battery or solar system during a grid outage.

Good point, @iwsmiser, I’m assuming that in the data line “instant_power”:11,“instant_reactive_power”:638,
instant_power is Watts and instant_preactive_power is in VAr, but I don’t have a reference for that.

Looking at other people’s api downloads the data I’m getting seems quite consistent, but the smart meters of some people, like @pgalbavy, seem to only record the instant_power value, about 3W average. That’s about what my old mechanical meter was recording. Since my new Smart Meter is recording a constant 80W average I am guessing that itis recording a different way and adding in some component of reactive power. I live in hopes of being able to talk to a real electrical engineer at EDF who knows what reactive power is. If I draw a blank I may try contacting the meter manufacturer, Landis & Gyr. But I suspect that Smart Meters can be programmed to record different parameters.

As @djos1475 says, 100Wh per day or 20W continuous would be reasonable and in spec. There seems to be something fishy about how different energy suppliers are recording and charging for domestic energy in the UK.

1 Like

Mike, here is a link to how EDF charges for reactive power.