Smart Meter Installed

[coppice]

It's cheap electronics engineered to pinch the last penny, it's not like control accuracy of 1%FS or so is that bad - but it still feels wrong and causes real loss of money.

The problem is a large amount domestic consumption is at around something like 1% of full scale. So, if you are only accurate to 1% of full scale you may have a 100% error for much of your consumption. There is not excuse for this in 2024, when an energy metering chip that achieves better 1% accuracy over a 1000:1 range is 20 to 30 cents, and the sensors for it need not cost a lot.

[Siwastaja]

The problem is a large amount domestic consumption is at around something like 1% of full scale. So, if you are only accurate to 1% of full scale you may have a 100% error for much of your consumption. There is not excuse for this in 2024, when an energy metering chip that achieves better 1% accuracy over a 1000:1 range is 20 to 30 cents, and the sensors for it need not cost a lot.

I agree 1%FS is not that great. Then again, 1% accuracy (of value) over 1000:1 range implies 0.001%FS error. I won't buy this claim, especially when you consider real-world AC measurement which requires phase voltage sensing and quite high sampling rates, no phase shift etc. as real world current waveforms have distortion and harmonics.

The reason why inverter manufacturers (and we!) use current transformers is they offer an easy way for retrofit building net metering without having to disconnect and reconnect the main supply phase lines, possibly having to add another box on the wall where to mount a "proper" meter. This, with extra work involved, is easily 200-300EUR worth, so whether a chip costs 20 or 30 cents is totally irrelevant.

We are hoping that utility meters with data output become more commonplace. The transition is currently happening here. The utility meter is both more accurate, and even when it isn't, it's the one used for billing so using the same data for battery control results in zero offset in billing.

[coppice]

I agree 1%FS is not that great. Then again, 1% accuracy (of value) over 1000:1 range implies 0.001%FS error. I won't buy this claim, especially when you consider real-world AC measurement which requires phase voltage sensing and quite high sampling rates, no phase shift etc. as real world current waveforms have distortion and harmonics.

They can't achieve 0.001% accuracy when the load is full scale. Temperature coefficients wouldn't allow that, and calibration would be a nightmare to achieve. They achieve better than 1%, and often easily achieve 0.1%, over 1000:1 range, with a fairly constant error across most of that range. When you get down to really low currents CMRR, and other forms of crosstalk, start to spoil the results. Just look at any app note for an energy metering chip, and the real world results they show. The IEC specs for utility energy meters don't call for maximum accuracy over such a wide range, but its easy to achieve these days.

The reason why inverter manufacturers (and we!) use current transformers is they offer an easy way for retrofit building net metering without having to disconnect and reconnect the main supply phase lines, possibly having to add another box on the wall where to mount a "proper" meter. This, with extra work involved, is easily 200-300EUR worth, so whether a chip costs 20 or 30 cents is totally irrelevant.

If you want a completely non-invasive fit, then accuracy will be limited with a clamp on current transformer. However, if you can disconnect a wire, and slip a CT over it, you can achieve high accuracy over a wide dynamic range. Cheap CTs, driven by the utility meter market, have massively improved in the last 2 decades. One that's DC tolerant up to 10A, with a maximum current of 100A, and maintains it phase and amplitude accuracy down to <100mA is less than a dollar these days.

We are hoping that utility meters with data output become more commonplace. The transition is currently happening here. The utility meter is both more accurate, and even when it isn't, it's the one used for billing so using the same data for battery control results in zero offset in billing.

Utility meters don't really cut it when looking for issues that might be very dynamic in nature, like the impulse response of local generation kit. They are engineered to give longer term information. However, the components they are built from can be used to give the kind of dynamic results you need.

[Electrodynamic]

rteodor

I would think so too. If I watch grid consumption during the night when the mains is more quiet, that value swings from 20W to 80W all the time. During the day when more loads are coming and going from the grid that swing grows to like -30W (export!) to well over 100W.
Maybe Deye is built differently but regulation within 5W margin is not easy.

I agree, I built and tested my own no import/no export programmable hybrid inverter system logging volts, amps, power, energy, frequency and PF. I also tested no export load swapping to power water/space heaters and sand thermal storage.

The problem comes down to resolution and lag time. So by the time my microcomputer measured all the data, made all my calculations and sent the commands to the inverter section the power values had already changed. I found large fast changes in power like AC motor inrush current tend to produce the greatest error rates. The greater the rate of load change or system lag the greater the error rate. I found it very difficult to reduce the lag time and one option is to use a low power always export setting. So we always export a small amount say 50w to compensate for the system lag. If we do import we keep increasing the min export setting until we stop importing if that's important to the user.

I also think the smart meter works against us in a hybrid limited import mode like what EEVblog seems to be doing. From what I gathered the smart meter logs the import and export power separately. So if you do import power you cannot take it back by exporting later, it's a one way street and keeps accumulating. This is also handy for the utility companies to check on what kind of equipment there customer might be operating. If an export reading ever shows up on the smart meter they know about it.

[Siwastaja]

The problem comes down to resolution and lag time. So by the time my microcomputer measured all the data, made all my calculations and sent the commands to the inverter section the power values had already changed. I found large fast changes in power like AC motor inrush current tend to produce the greatest error rates. The greater the rate of load change or system lag the greater the error rate. I found it very difficult to reduce the lag time and one option is to use a low power always export setting. So we always export a small amount say 50w to compensate for the system lag. If we do export we keep increasing the min export setting until we stop exporting if that's important to the user.

I also think the smart meter works against us in a hybrid limited import mode like what EEVblog seems to be doing. From what I gathered the smart meter logs the import and export power separately. So if you do import power you cannot take it back by exporting later, it's a one way street and keeps accumulating. This is also handy for the utility companies to check on what kind of equipment there customer might be operating. If an export reading ever shows up on the smart meter they know about it.

Again, the key question is how import/export is being billed. What I hate in these discussions and product design in general is that companies produce solutions and people discuss them without first defining at all what they should do i.e. requirements. This drives us to solve wrong problems or maybe completely nonexisting problems.

For example if billing is based on net (import + export) over some sensible time period (5 minutes, 15 minutes, hour...) then all those control dynamic issues become trivial, just keep a running count of net and adjust power to zero it out. Even if your impulse response were asymmetrical (e.g. take 1 second to ramp up and 5 seconds to ramp down) even that doesn't matter, then.

But to do something like this we first need to know what are the billing criteria and because they vary by country and time and are sometimes hard to find out, it is hard for product designers to write algorithms for.

[coppice]

Again, the key question is how import/export is being billed. What I hate in these discussions and product design in general is that companies produce solutions and people discuss them without first defining at all what they should do i.e. requirements. This drives us to solve wrong problems or maybe completely nonexisting problems.

For example if billing is based on net (import + export) over some sensible time period (5 minutes, 15 minutes, hour...) then all those control dynamic issues become trivial, just keep a running count of net and adjust power to zero it out. Even if your impulse response were asymmetrical (e.g. take 1 second to ramp up and 5 seconds to ramp down) even that doesn't matter, then.

But to do something like this we first need to know what are the billing criteria and because they vary by country and time and are sometimes hard to find out, it is hard for product designers to write algorithms for.

Most metering measurement units are used in one of 2 ways. One way is the meter takes per second energy results from the measurement unit, and accumulates them according to the direction of the net energy flow over that second. A few meters actually accumulate in 4 quadrant bins, rather than 2 half-plane bins. The other way is they work from the energy pulses. Those 1000/kWh or 1600/kWh type pulses that usually flash an LED these days. These give a somewhat longer average of the energy flow, unless the load is high, but they do properly integrate any moments of import vs export through the pulse period. I think most people work from per second measurements from the front end these days.

[rteodor]

The problem is a large amount domestic consumption is at around something like 1% of full scale. So, if you are only accurate to 1% of full scale you may have a 100% error for much of your consumption. There is not excuse for this in 2024, when an energy metering chip that achieves better 1% accuracy over a 1000:1 range is 20 to 30 cents, and the sensors for it need not cost a lot.

Victron is well known for not cutting corners and still can't - or won't do it more precisely than 50W residual. Accurate measuring does not make it for good grid regulation.

I agree, I built and tested my own no import/no export programmable hybrid inverter system logging volts, amps, power, energy, frequency and PF. I also tested no export load swapping to power water/space heaters and sand thermal storage.

Wow. Did you do it alone for yourself or part of a bigger project ?

I also think the smart meter works against us in a hybrid limited import mode like what EEVblog seems to be doing. From what I gathered the smart meter logs the import and export power separately. So if you do import power you cannot take it back by exporting later, it's a one way street and keeps accumulating. This is also handy for the utility companies to check on what kind of equipment there customer might be operating. If an export reading ever shows up on the smart meter they know about it.

What I find strange in Dave's case is the amount of energy imported when compared with my case. I left the system for a full month with no consumption other than the residual. The meter counted just under 30KWh. That mean my awg. residual is 41W. But Dave is exporting, probably at least 8h/day and still has 1.3KWh/day awg. - if I remember correctly!?. That means the residual power would be double, like 82W. Now I think maybe also the smart meter makes its count a bit "conservative" because I expect that the Deye inverter does not perform that worse as it consumes mostly at night.

Most metering measurement units are used in one of 2 ways. One way is the meter takes per second energy results from the measurement unit, and accumulates them according to the direction of the net energy flow over that second. A few meters actually accumulate in 4 quadrant bins, rather than 2 half-plane bins. The other way is they work from the energy pulses. Those 1000/kWh or 1600/kWh type pulses that usually flash an LED these days. These give a somewhat longer average of the energy flow, unless the load is high, but they do properly integrate any moments of import vs export through the pulse period. I think most people work from per second measurements from the front end these days.

The meter reading guy warned me that with the type of meter I have any export will be billed as consumption. Based on his knowledge of real cases.

[coppice]

The meter reading guy warned me that with the type of meter I have any export will be billed as consumption. Based on his knowledge of real cases.

There are anti-tamper meters which intentionally read energy as consumed whichever way it flows. I think some of the simple energy measurement chips, which simply output a number of pulses per kWh, may not differentiate which way the energy flows. Its a while since I looked at that, and I'm not sure now. Any energy measurement chip sophisticated enough to do more than just generate energy pulses shouldn't be doing that.

[uer166]

It's cheap electronics engineered to pinch the last penny, it's not like control accuracy of 1%FS or so is that bad - but it still feels wrong and causes real loss of money.

The problem is a large amount domestic consumption is at around something like 1% of full scale. So, if you are only accurate to 1% of full scale you may have a 100% error for much of your consumption. There is not excuse for this in 2024, when an energy metering chip that achieves better 1% accuracy over a 1000:1 range is 20 to 30 cents, and the sensors for it need not cost a lot.

As someone who designed smart meters: there's way more to it than this. It's not "$0.2-$0.3" for this kind of dynamic range. There's usually a decently small offset error (certainly less than 1% of FS), and a 1-2-3% error or so that's guaranteed on a much smaller dynamic range. The specific accuracy depends on the load PF, crest factor, etc. There shouldn't be a 80W offset error for a smart meter, but then again we're not talking about a MID certified meter, rather the internals of the Deye inverter that might cause that fixed import.

[Siwastaja]

I find this whole discussion odd, as if you all lived in a chaos. I mean, I can easily explain how this all works in Finland no problem, and have done so before. Finland can't be an exception, don't you have any means to figure out the rules in wherever you live?

I mean, you can't your buy a PV system / battery inverter and connect it to grid without asking anyone and then start wondering how that might affect  your billing or how your energy is measured, can you? At least here you need a permit from the grid company and at that time they check whether the local distribution can handle your export. (This "permit" being a very lightweight process, more like installer just notifying about it). And quite obviously during that permit or notification process they also would make sure that the energy meter at your premise is recent enough so that it can measure import and export and and report that data automatically to the grid operator. In Finland all meters were upgraded to support this probably more than 10 years ago but some early adopters of PV might have required a meter swap, basically putting them earlier in the line of meter changes.

And now we are in the process of the second set of meters that support separate import/export metering. Simply because meters have some technical lifetime. Maybe they are getting out of calibration guarantee timeframe, or maybe this is needed to transition into 15-minute net periods, maybe the older meters from 10-15 years ago only support hourly logging, I don't know exactly. And the new meters have the so-called H1/P1 port, basically a one-way open-collector UART giving ASCII data every 10 seconds, for data integrations into control systems. But no inverter I'm aware of supports it, our control box is one of the few that does. But at least having it available is solving other half of the chicken-egg problem.

So can you do these battery installs without permit and if you can then I think the best would still be to call the grid operator and ask how their billing rules work, is there netting or not and if yes, over what time period?

And if someone still has such ancient meter that it cannot read export at all (or wrongly reads it as import), then it would be obvious you need a meter swap when getting a PV or hybrid inverter or even just a battery inverter because of the dynamics mentioned causing short-time export.

[uer166]

The meter reading guy warned me that with the type of meter I have any export will be billed as consumption. Based on his knowledge of real cases.

This is surprisingly reasonable. IIRC PG&E will actually fine you for it some much larger $$$ if you export without an agreement.

[rteodor]

I find this whole discussion odd, as if you all lived in a chaos. I mean, I can easily explain how this all works in Finland no problem, and have done so before. Finland can't be an exception, don't you have any means to figure out the rules in wherever you live?

You clearly have no idea what 50 years of forced communism did to people's minds.

[tszaboo]

It's cheap electronics engineered to pinch the last penny, it's not like control accuracy of 1%FS or so is that bad - but it still feels wrong and causes real loss of money.

The problem is a large amount domestic consumption is at around something like 1% of full scale. So, if you are only accurate to 1% of full scale you may have a 100% error for much of your consumption. There is not excuse for this in 2024, when an energy metering chip that achieves better 1% accuracy over a 1000:1 range is 20 to 30 cents, and the sensors for it need not cost a lot.

As someone who designed smart meters: there's way more to it than this. It's not "$0.2-$0.3" for this kind of dynamic range. There's usually a decently small offset error (certainly less than 1% of FS), and a 1-2-3% error or so that's guaranteed on a much smaller dynamic range. The specific accuracy depends on the load PF, crest factor, etc. There shouldn't be a 80W offset error for a smart meter, but then again we're not talking about a MID certified meter, rather the internals of the Deye inverter that might cause that fixed import.

1% total error for the Deye inverter and the smart meter.
Smart meters exist with 2%, 1%, 0.5% and 0.2% specifications.
What they install here is domestically Class A, which is 2% accurate. Which is 2% of range. You don't like it, too bad, the DSO doesn't care, nobody understands what it means to care. If you talk about it, you are met with empty stares from people because they have no idea what statistics or "error in measurement" mean.  it could be easily 2KW/day, meaning 160 EUR a year for the entire lifetime of the meter.
When I had my hot water meter replaced, my consumption magically went down 30% in a year. Which is hundreds of EUR in  a year. Also don't look up gas meter accuracy.
Don't complain, be a good sheep.

[coppice]

It's cheap electronics engineered to pinch the last penny, it's not like control accuracy of 1%FS or so is that bad - but it still feels wrong and causes real loss of money.

The problem is a large amount domestic consumption is at around something like 1% of full scale. So, if you are only accurate to 1% of full scale you may have a 100% error for much of your consumption. There is not excuse for this in 2024, when an energy metering chip that achieves better 1% accuracy over a 1000:1 range is 20 to 30 cents, and the sensors for it need not cost a lot.

As someone who designed smart meters: there's way more to it than this. It's not "$0.2-$0.3" for this kind of dynamic range. There's usually a decently small offset error (certainly less than 1% of FS), and a 1-2-3% error or so that's guaranteed on a much smaller dynamic range. The specific accuracy depends on the load PF, crest factor, etc. There shouldn't be a 80W offset error for a smart meter, but then again we're not talking about a MID certified meter, rather the internals of the Deye inverter that might cause that fixed import.

As someone who designed smart meters, its 20 to 30 cents for a chip that does basic functions, and about 50 if you want active, reactive, voltage, current, etc. measurements with 1000:1 range at 0.5% accuracy or better. Who charges more? The parts you need in addition to the measurement chip are mostly cheap. If you want a shunt sensor with a great temperature coefficient that another 20 cents, but for what we are talking about you can use something cheap if you don't need maximum accuracy. If you need isolation that can add a bit, but in the types of products we are looking at, much of the circuitry is connected to the live area, and metrology can be dropped in there.

Outside the US the relevant specification is IEC62053. This defines the accuracy requirements for energy meters to meet what is termed class 0.5, class 1 and class 2. These basically mean 0.5%, 1% or 2% accuracy over the bulk of their operating current range, and a fairly wide power factor range. The spec relates everything to a basis current, which is only a fraction of full scale, but full scale is not defined. Most utilities use 6 times basis as full scale, although some use 10, 12, or even 20. Lets take 6. So, a typical domestic 230V 60A meter has a basis current of 10A. Its a weird way to express an operating range, but that's how it is. Let's take a class 1 (1% accurate) meter. The spec says the meter must be 1% accurate from full scale down to 10% of the basis current, at unity power factor, and down to 20% of the basis current for power factors between 0.5L and 0.8C. It must be 1.5% accurate between 5% of the basis current at unity power factor, and 10% of the basis current for power factors between 0.5L and 0.8C. Below this they don't specify, but no utility will accept a meter with more than a few percent error at its cutoff point, called the anti-creep threshold or starting current. This is 0.4% of the basis current. Below this threshold the meter must not accumulate energy. This is to prevent noise causing the meter to creep on no load, and create bogus consumption for unoccupied buildings. If you do, you have a customer service nightmare.

So, the typical domestic 230V 60A class 1 meter will read anything above 40mA with reasonable acuray, although this is usually implemented as a power measurement, so it ends up dependent on the power factor. Using power helps with noise tolerance. 40mA at unity PF is 9.2W. The meter must read with decent accuracy from 9.2W up to 13.8kW (60A at unity PF), and meet the accuracy template in the IEC spec from 0.5A upwards. In reality, with current metrology solutions, any well designed meter should be 1% accurate from starting current to full scale over a wide power factor range, and a reasonable temperature range. This is now cheap to do, including the related calibration of the sensors. If non-metering products are not achieving results like this, they really need to get their act together.

[coppice]

What they install here is domestically Class A, which is 2% accurate. Which is 2% of range.

A utility meter's energy measurements are not specified as a percentage of range, but as a percentage of reading, as I described in my previous message. If the meter has supplementary functions, like current and voltage measurements, those are typically specified as a percentage of full scale, but the energy measurement accuracy is not.

[Andy Chee]

In Finland all meters were upgraded to support this probably more than 10 years ago

In Australia, the push back against meter upgrades was quite significant.  Of course there were the usual conspiracy theories surrounding radio waves and brain cancer, etc. 

But the biggest reason for the reluctance, is that the cost of the meter upgrade was to be paid by the consumer, not the power company.

https://theconversation.com/smart-meters-dumb-policy-the-victorian-experience-47685

https://stopsmartmeters.com.au/

Even to this day, there are many areas of Australia still not converted to smart meters.

That being said, I do believe that anyone installing solar panels with export capability, are generally not permitted to retain ancient dumb meters.

[uer166]

It's cheap electronics engineered to pinch the last penny, it's not like control accuracy of 1%FS or so is that bad - but it still feels wrong and causes real loss of money.

The problem is a large amount domestic consumption is at around something like 1% of full scale. So, if you are only accurate to 1% of full scale you may have a 100% error for much of your consumption. There is not excuse for this in 2024, when an energy metering chip that achieves better 1% accuracy over a 1000:1 range is 20 to 30 cents, and the sensors for it need not cost a lot.

As someone who designed smart meters: there's way more to it than this. It's not "$0.2-$0.3" for this kind of dynamic range. There's usually a decently small offset error (certainly less than 1% of FS), and a 1-2-3% error or so that's guaranteed on a much smaller dynamic range. The specific accuracy depends on the load PF, crest factor, etc. There shouldn't be a 80W offset error for a smart meter, but then again we're not talking about a MID certified meter, rather the internals of the Deye inverter that might cause that fixed import.

As someone who designed smart meters, its 20 to 30 cents for a chip that does basic functions, and about 50 if you want active, reactive, voltage, current, etc. measurements with 1000:1 range at 0.5% accuracy or better. Who charges more? The parts you need in addition to the measurement chip are mostly cheap. If you want a shunt sensor with a great temperature coefficient that another 20 cents, but for what we are talking about you can use something cheap if you don't need maximum accuracy. If you need isolation that can add a bit, but in the types of products we are looking at, much of the circuitry is connected to the live area, and metrology can be dropped in there.

Outside the US the relevant specification is IEC62053. This defines the accuracy requirements for energy meters to meet what is termed class 0.5, class 1 and class 2. These basically mean 0.5%, 1% or 2% accuracy over the bulk of their operating current range, and a fairly wide power factor range. The spec relates everything to a basis current, which is only a fraction of full scale, but full scale is not defined. Most utilities use 6 times basis as full scale, although some use 10, 12, or even 20. Lets take 6. So, a typical domestic 230V 60A meter has a basis current of 10A. Its a weird way to express an operating range, but that's how it is. Let's take a class 1 (1% accurate) meter. The spec says the meter must be 1% accurate from full scale down to 10% of the basis current, at unity power factor, and down to 20% of the basis current for power factors between 0.5L and 0.8C. It must be 1.5% accurate between 5% of the basis current at unity power factor, and 10% of the basis current for power factors between 0.5L and 0.8C. Below this they don't specify, but no utility will accept a meter with more than a few percent error at its cutoff point, called the anti-creep threshold or starting current. This is 0.4% of the basis current. Below this threshold the meter must not accumulate energy. This is to prevent noise causing the meter to creep on no load, and create bogus consumption for unoccupied buildings. If you do, you have a customer service nightmare.

So, the typical domestic 230V 60A class 1 meter will read anything above 40mA with reasonable acuray, although this is usually implemented as a power measurement, so it ends up dependent on the power factor. Using power helps with noise tolerance. 40mA at unity PF is 9.2W. The meter must read with decent accuracy from 9.2W up to 13.8kW (60A at unity PF), and meet the accuracy template in the IEC spec from 0.5A upwards. In reality, with current metrology solutions, any well designed meter should be 1% accurate from starting current to full scale over a wide power factor range, and a reasonable temperature range. This is now cheap to do, including the related calibration of the sensors. If non-metering products are not achieving results like this, they really need to get their act together.

I agree with most of this analysis other than cost. DC-tolerant CTs or shunts that can handle 5kA surge are not that cheap. And of course a "1%" class A B meter can have error much larger than that for bad PF loads. I've designed for EN50470 (MID metering) specifically which seems relayed to your standard.

[coppice]

I agree with most of this analysis other than cost. DC-tolerant CTs or shunts that can handle 5kA surge are not that cheap. And of course a "1%" class A B meter can have error much larger than that for bad PF loads. I've designed for EN50470 (MID metering) specifically which seems relayed to your standard.

Maybe you use the wrong suppliers. People like VAC charge quite a lot in modest quantities. I don't know what they are like for production volumes. However, you can buy CTs from Chinese suppliers than beat VAC on every performance parameter for a very reasonable price. 20 years ago that was not the case, but from over 10 years ago it has been.

When the power factor gets below about 0.2 even expensive power analysers start struggling to maintain high accuracy. You really need to match the phases of your voltage and current sensors to a very high accuracy for that kind of power factor.

[uer166]

You really need to match the phases of your voltage and current sensors to a very high accuracy for that kind of power factor.

We ended up doing a per-phase calibration of the inter-sensor delays. It was a nightmare but overall avoided needing expensive phase-matched input anti-aliasing filters!

[coppice]

You really need to match the phases of your voltage and current sensors to a very high accuracy for that kind of power factor.

We ended up doing a per-phase calibration of the inter-sensor delays. It was a nightmare but overall avoided needing expensive phase-matched input anti-aliasing filters!

Quite a lot of meter designs screw up those filters, by not keeping them impedance balanced. They work great at room temperature, but the phase shift goes way off as the temperature changes, because most capacitors are not that thermally stable. Get them balanced and any old capacitors will do.

[tszaboo]

What they install here is domestically Class A, which is 2% accurate. Which is 2% of range.

A utility meter's energy measurements are not specified as a percentage of range, but as a percentage of reading, as I described in my previous message. If the meter has supplementary functions, like current and voltage measurements, those are typically specified as a percentage of full scale, but the energy measurement accuracy is not.

Well thanks for correcting me, I learned something.
But you agree that with low currents, cancelled out with ie a home battery the error will be higher than 1 or 2%.

[Njk]

Literally:

[EEVblog]

I've changed the Grid Trickle Feed setting from 5W to 50W, I'll give it a few days and see what happens.

[Siwastaja]

This is now cheap to do, including the related calibration of the sensors. If non-metering products are not achieving results like this, they really need to get their act together.

Again, as a meter designer don't be fooled by the BOM cost of the meter. It's not like an inverter manufacturer could just add those components on the inverter's main PCB. Due to physical requirements the grid point net meter has to be separate. Even if inverter manufacturer started to manufacture their own meters (some do, e.g. Kostal offers both inverters and meters) it is unlikely they could significantly cut prices compared to existing MID compliant meter manufacturers.

The cost of the chips or current sensors in the meter is irrelevant, customer pays for the market price of the meter which is much more than BOM, plus extra install work, which is always at customer's site. These inverters already offer two options for the grid point net metering: using a MID certified (with any accuracy you want) meter using RS485 wiring, or clamp-on CTs. The former option is approx. 150-300EUR more expensive (install cost included), not some cents.

Probably the sweet spot in cost would be to algorithmically improve low-load offset errors, at least if the inverter manufacturer thinks they will be selling in large numbers to compensate for the NRE cost.

Thanks for explaining the accuracy details, the information is not too easy to find without diving deep into the actual standards. % of value indications are most misused metrics, they always require extra attributes like some minimum threshold, otherwise they always imply infinite accuracy and as such are impossible.

One thing to remember is that offset issue is not only at low loads. Say we have two meters which are provably +/-1% accurate in value, so MID compliant (was it class B for 1%, I don't remember). Now let's say one meter is used for billing, other for inverter control, and your consumption is 2000W for 12 hours a day (whatever clothes washing, air conditioner use, cooking). One meter shows 1980W, another 2020W. 12 * 40Wh = 0.5kWh of unwanted import or export is thus billed even if offset error at low loads was completely absent and even when metering is utility-level accurate and compliant to billing requirements. And this 0.5kWh would be enough for some people to complain. So for true no-import-when-production-exceed-consumption, data integration to the meter which does the billing is a must. Even if you designed a perfectly accurate to 0.0001% meter, it won't help because the existing billing meter is not that accurate.

[IanJ]

Just a note……When I decided to build my own centralized home energy monitoring thing I passed on CT’s (current clamps) and went for a proper meter with RS485/modbus output.
CT’s are sooooo easy to fit but meh!

https://owen-brothers.com/singlephase/direct-connected-mid-approved/ob418-multi-function-100a.html

Ian