This is completely normal for an AC coupled grid tied system. Loads turn on / off instantly but inverters (both by design and by Grid code standards) have a set maximum ramp rate, e.g. if ramp rate is 200w per second, and you turn on a 2kw load, it will take 10 seconds to hit inverter outputing 2kw. And then the same in reverse when turning off load. Multiply that my dozens of devices constantly turning on/off, increase / decreasing load and what you get is a constant ac couple tug of war. Your solar analystics and grid meter data is also only showing you these fluctuations rounded to 10/30m intervals, but its actually calculated internally to 1 second intervals. I have much more accurate data from some of my ac coupled systems which show this and might make a good video if you are interested
This. The control loop in the inverter to alter output current in an effort to zero the grid feed isn't perfect. They're often in the 1 second update range, and quite damped to avoid oscillating. Switching on a load can take several seconds before the inverter cancels its mains draw. You'll also export a bit after the load switches off. Hot plates, ovens, water pumps, refrigerators and other short cycle devices exacerbate the issue.
But wouldn't it be 200w per second in both directions? It would still average to 0w draw on average. To see an average import with 200w per second ramp up the ramp down would have to be faster, say 500w per second?
@BenBrand there is no averaging for metering / billing, you are charged/ recorded for every second, which is then summed (not averaged) into the 30m blocks of data reported. So in a theoretical 30min block if you export 1kw for 15mins and import 1kw for the next 15mins, the average is 0, but you are charged / credited for both 1kw import and 1kw export.
@@CptEcho yes, but it doesn't look like that's what's happening? It's rarely zero and never negative. I can see how if it ever does go negative it would just stay at zero since he's not allowed to export. But if it truly was even in both directions then it would be zero about half the time and positive about half the time?
@BenBrand that's not how this meter records data, it has a seperate export register and an import register, in the video he is only showing the import register which is why it pegs to 0 when exporting, if the data was combined into a single graph you would see half hour intervals of import, half hour intervals of export and most importantly half hour intervals with both import & export.
Try plotting the solar production along with your consumption. You'll find your consumption goes to zero when you are exporting significantly, say >1kW. As for the other spikes, the Deye inverter doesn't respond very quickly to a step change in demand. You can test this by putting the kettle on. You'll consume from the grid while the Deye increases it's output to match the load and then you'll export while it decreases it's output. Don't have a good explanation for why the rest of the input is so high though.
You only get zero in the middle of the day when you likely export to the grid. 30 minute intervals is just way to coarse to see anything. It is most likely the inverter that can not respond fast enough to changes in load power so it briefly imports energy from the grid. It could be just a second or even less but enough for those spikes to be recorded by the smart meter. Thus what seems like fairly random data. You can see that at night imported power is fairly constant and much more variable during the day. So is the inverter that can not guess when a load will switch ON and it needs some time to react and correct the amount of imported power.
The algorithm to select which power to "exported" from a battery to an AC system is always behind the actual usage. Take for example when a fridge turns on. it will introduce a peak then most likely it will overcompensate by providing little bit to much power etc.
Three further reasons for low level usage: 1. The BMS can request charging to balance its cells or to protect the battery from becoming over discharged (on my battery at 10% further use is prevented and at 6% SoC the BMS requests a slow charge to bring it back to 10%). 2.There is inverter lag (the time difference between a load starting / stopping and inverter changing its output power to match). 3. CT / Meter out of calibration. Inverter lag at least for me was improved by a firmware update. There is usually CT compensation settings which can be adjusted. In your case the CT might be reading too low so the inverter not fully catching loads.
With my set up, because I pull data from several systems, they don’t all respond at the same time, in fact my battery data comes from the cloud rather than direct from the inverter itself (yeah I know!) so it can produce some rather unusual wtf readings sometimes.
It can be only synchronising-issues between your home-system and the grid. The electronics have to synchronise quite a lot depending on load but also on the grid. Especially with switching power-supplies in use, it will constantly need to synchronise and therefore put some small amount of load on the grid unless you can mechanically decouple it and later couple it back on. One of my inverter-battery-systems is using this and with a delay of 20 minutes unless significally overloaded. It just switches itself from the grid (relay-based) and checks again after at least 20 minutes. That now makes it really 0W overall. With this off, it was creating small loads all the time enough for the smart-meter.
You have to read the fine print that comes with the smart meter. It comes preinstalled with a electron gremlin. It was trained at the factory to only eat electrons from the grid. Solar electrons give it gas and no one likes a gassy gremlin. What you have measured here is the electron gremlin input.
While I agree with others that 30m is too coarse for a detailed hunt for the cause of the drain, I disagree that it's the ramp rate. That just makes no sense if you actually do the Math. Let's say there were appliances that frequently pull 2 KW for a short time, incurring a ramp rate as was suggested by @CtpEcho of 200 W/s (which seems rather low to me, but I live in a different country with different electrical codes). Since a fully linear ramp is equivalent to half the "ramp delta" for the same duration, that would be 1 KW for 10s, which is 0.0027 KWh. How many of those "ramps" do you expected to have over a day to get to almost 2 KWh total? About 740, so roughly 1 per 2 minutes (ish). Yea right. That would be literally insane appliance behaviour. I mean I don't know your electric loads, but we have everything electric, so no gas or other fossils used or even installed at all, and no AC. Even a dishwasher, washing machine, dryer or whatever will never pulse to such an insane degree. We also have smart plug monitoring for most larger loads (except the cooktop and electric water heater), and can just see that on a per second basis. For value comparisons, we also got a solar installation (only ~7.5 KWp, but smaller battery of only 7.8 KWh or so), and on days where we are as self sufficient as possible we're on 0.1 to 0.2 KWh grid usage. 1.8 is WAY too high for your setup IMHO. I would say throw some smart plugs into the monitoring situation (they are cheap enough with monitoring these days), and see how your actual loads behave. It might be that the reaction delay to any change (before the inverter reacts at all) is misconfigured? Or the ramp time is off by like an order of magnitude for some reason maybe? I would go over all the settings, but without knowing what your load profile looks like that'll probably be hard to identify...
The quite regular spikes are also pretty interesting. Are you doing something energy intensive that consistently? Or maybe it's something on a schedule.
If optimisation towards zero net import is a hugely important goal, would it be worthwhile to prepare a power duration curve & payback calculation on different peak power output battery options?
The "usage" graph is probably aggregating instantaneous samples of supply or demand separately, while the Excel data is all samples aggregated into 30 minute time buckets. What is actually being billed to you?
That graph at 1:34 looks far too consistent unless clouds haven't been invented for the Aussies. For us in Ohio most days are a white haze now, can't see the Sun or Moon most of the time for the past 15 years, now back to normal the past 2 weeks. I won't say why but 757 size planes come in from the NE, leave something behind, and go back to their start point. Closest airport 130 miles that direction.
I'm assuming these are not three-phase values and you're looking at the single phase? Might be time to get some more metering equipment. I tested many many meters working on the network and almost every single one was very accurate.
As neither the smart meter or the inverter/battery system measures power directly the voltage & current variations between them will add additional noise to the data, beyond that of the pure MID accuracy rating, as you pretty much covered already. Arguably the additional need to measure supply voltage and (simplifying here) raise it for export, balance it for 'zero' and lower it for import adds additional noise as the inverter/battery system cannot presume the supply voltage and has to sample it regularly. I presume that the inverter will have a set limit so that it does not chase the supply voltage spikes to above the regulated supply voltage, dropping the output and resampling until within limits. As an aside, here in the UK my supplied voltage is always very near the upper legal limits as it is advantageous to the suppliers. [UK = 230v -6% / +10% (or 216.2 to 253.0v). The lowest I have ever seen is 239.57v and is usually in the 247 to 250v range. Last month's highest supply voltage was a sparky 254.53v. The Dilithium Crystals can take it, apparently. Phase management is another active system at play as the battery/inverter system will need to tune itself to the (relatively minor) phase fluctuations on the supply side. As with voltage and current measurement the inverter/battery system will respond somewhat in lag as it has to sample the incoming before tuning the export to the grid. This 'input sample & export' cycle will also show as a tiny import reading may also add additional noise that trends towards an import measurement. Of course, everything needs to work as intended too... my SolarEdge system has basic mathematical errors in the displayed calculations.
![Kodak Black - Catch Fire [Official Music Video]](http://i.ytimg.com/vi/A4ch-olIuZo/mqdefault.jpg)
This is completely normal for an AC coupled grid tied system. Loads turn on / off instantly but inverters (both by design and by Grid code standards) have a set maximum ramp rate, e.g. if ramp rate is 200w per second, and you turn on a 2kw load, it will take 10 seconds to hit inverter outputing 2kw. And then the same in reverse when turning off load. Multiply that my dozens of devices constantly turning on/off, increase / decreasing load and what you get is a constant ac couple tug of war. Your solar analystics and grid meter data is also only showing you these fluctuations rounded to 10/30m intervals, but its actually calculated internally to 1 second intervals. I have much more accurate data from some of my ac coupled systems which show this and might make a good video if you are interested
This. The control loop in the inverter to alter output current in an effort to zero the grid feed isn't perfect. They're often in the 1 second update range, and quite damped to avoid oscillating. Switching on a load can take several seconds before the inverter cancels its mains draw. You'll also export a bit after the load switches off. Hot plates, ovens, water pumps, refrigerators and other short cycle devices exacerbate the issue.
But wouldn't it be 200w per second in both directions? It would still average to 0w draw on average. To see an average import with 200w per second ramp up the ramp down would have to be faster, say 500w per second?
@BenBrand there is no averaging for metering / billing, you are charged/ recorded for every second, which is then summed (not averaged) into the 30m blocks of data reported. So in a theoretical 30min block if you export 1kw for 15mins and import 1kw for the next 15mins, the average is 0, but you are charged / credited for both 1kw import and 1kw export.
@@CptEcho yes, but it doesn't look like that's what's happening? It's rarely zero and never negative. I can see how if it ever does go negative it would just stay at zero since he's not allowed to export. But if it truly was even in both directions then it would be zero about half the time and positive about half the time?
@BenBrand that's not how this meter records data, it has a seperate export register and an import register, in the video he is only showing the import register which is why it pegs to 0 when exporting, if the data was combined into a single graph you would see half hour intervals of import, half hour intervals of export and most importantly half hour intervals with both import & export.
Try plotting the solar production along with your consumption.
You'll find your consumption goes to zero when you are exporting significantly, say >1kW.
As for the other spikes, the Deye inverter doesn't respond very quickly to a step change in demand. You can test this by putting the kettle on. You'll consume from the grid while the Deye increases it's output to match the load and then you'll export while it decreases it's output.
Don't have a good explanation for why the rest of the input is so high though.
Thanks for the use of the term " Electric Boogaloo" . Needs to be incorporated more often into modern day of language.
Indeed.
You only get zero in the middle of the day when you likely export to the grid.
30 minute intervals is just way to coarse to see anything. It is most likely the inverter that can not respond fast enough to changes in load power so it briefly imports energy from the grid. It could be just a second or even less but enough for those spikes to be recorded by the smart meter. Thus what seems like fairly random data.
You can see that at night imported power is fairly constant and much more variable during the day. So is the inverter that can not guess when a load will switch ON and it needs some time to react and correct the amount of imported power.
The algorithm to select which power to "exported" from a battery to an AC system is always behind the actual usage. Take for example when a fridge turns on. it will introduce a peak then most likely it will overcompensate by providing little bit to much power etc.
Three further reasons for low level usage:
1. The BMS can request charging to balance its cells or to protect the battery from becoming over discharged (on my battery at 10% further use is prevented and at 6% SoC the BMS requests a slow charge to bring it back to 10%).
2.There is inverter lag (the time difference between a load starting / stopping and inverter changing its output power to match).
3. CT / Meter out of calibration.
Inverter lag at least for me was improved by a firmware update. There is usually CT compensation settings which can be adjusted. In your case the CT might be reading too low so the inverter not fully catching loads.
With my set up, because I pull data from several systems, they don’t all respond at the same time, in fact my battery data comes from the cloud rather than direct from the inverter itself (yeah I know!) so it can produce some rather unusual wtf readings sometimes.
It can be only synchronising-issues between your home-system and the grid. The electronics have to synchronise quite a lot depending on load but also on the grid.
Especially with switching power-supplies in use, it will constantly need to synchronise and therefore put some small amount of load on the grid unless you can mechanically decouple it and later couple it back on.
One of my inverter-battery-systems is using this and with a delay of 20 minutes unless significally overloaded. It just switches itself from the grid (relay-based) and checks again after at least 20 minutes. That now makes it really 0W overall. With this off, it was creating small loads all the time enough for the smart-meter.
You have to read the fine print that comes with the smart meter. It comes preinstalled with a electron gremlin. It was trained at the factory to only eat electrons from the grid. Solar electrons give it gas and no one likes a gassy gremlin. What you have measured here is the electron gremlin input.
🤣
Funny how these inaccuracies always turn out with the electricity provider getting more money. Incredible coincidence.
have you considered lead/lag? when you switch a load on the inverter can take a moment to pick it up and compensate by outputting more.
yeh the AGL meters , I gave up working mine out after a month of not getting any matching data to my own data collection
While I agree with others that 30m is too coarse for a detailed hunt for the cause of the drain, I disagree that it's the ramp rate. That just makes no sense if you actually do the Math. Let's say there were appliances that frequently pull 2 KW for a short time, incurring a ramp rate as was suggested by @CtpEcho of 200 W/s (which seems rather low to me, but I live in a different country with different electrical codes). Since a fully linear ramp is equivalent to half the "ramp delta" for the same duration, that would be 1 KW for 10s, which is 0.0027 KWh. How many of those "ramps" do you expected to have over a day to get to almost 2 KWh total? About 740, so roughly 1 per 2 minutes (ish). Yea right. That would be literally insane appliance behaviour.
I mean I don't know your electric loads, but we have everything electric, so no gas or other fossils used or even installed at all, and no AC. Even a dishwasher, washing machine, dryer or whatever will never pulse to such an insane degree. We also have smart plug monitoring for most larger loads (except the cooktop and electric water heater), and can just see that on a per second basis.
For value comparisons, we also got a solar installation (only ~7.5 KWp, but smaller battery of only 7.8 KWh or so), and on days where we are as self sufficient as possible we're on 0.1 to 0.2 KWh grid usage. 1.8 is WAY too high for your setup IMHO.
I would say throw some smart plugs into the monitoring situation (they are cheap enough with monitoring these days), and see how your actual loads behave.
It might be that the reaction delay to any change (before the inverter reacts at all) is misconfigured? Or the ramp time is off by like an order of magnitude for some reason maybe? I would go over all the settings, but without knowing what your load profile looks like that'll probably be hard to identify...
What is the power factor of your home? Could this be the difference between one system measuring real power versus the other measuring apparent power?
Why get data from the provider instead of straight from the meter's interface that it seems to have? Though could be mistaken.
The quite regular spikes are also pretty interesting. Are you doing something energy intensive that consistently? Or maybe it's something on a schedule.
Inconsistent axis over various graphs trigger the OCD.
Have a look at the scale of the graphs 0.45 highest, 0.06 lowest. but that 12th spike wasn't big in comparison.
Are you able to carefully manage appliance usage and go full off grid for ~24 hours?
No, we don't have an AC coupled battery backup.
If optimisation towards zero net import is a hugely important goal, would it be worthwhile to prepare a power duration curve & payback calculation on different peak power output battery options?
The "usage" graph is probably aggregating instantaneous samples of supply or demand separately, while the Excel data is all samples aggregated into 30 minute time buckets.
What is actually being billed to you?
I think you are chasing mo moos Dave (microns) slang😂
could this have sometning to do with reactive power? !
That graph at 1:34 looks far too consistent unless clouds haven't been invented for the Aussies. For us in Ohio most days are a white haze now, can't see the Sun or Moon most of the time for the past 15 years, now back to normal the past 2 weeks. I won't say why but 757 size planes come in from the NE, leave something behind, and go back to their start point. Closest airport 130 miles that direction.
No free lunch here. Somebody gotta pay the dodgy fee.
Turn off the grid for a day or two and see what the meter says.
Glennie Islands
Wonder how much money they're making on that "little" error across all their customers.
Smart meters are notoriously inaccurate especially with harmonic noise... Maybe an opportunity for you to try out a power quality analyzer / monitor
It's elves.
ADW200-D16
Does he still think they don't steal
I just clicked this because I saw the graph and numbers and tables...
All your graphs have different scales, so can’t compare days directly.
1:20 .. whats with that audio? Is it my speakers?
Yah, I can hear some funny business going on with my mic.
Hot water timer??
He has a dedicated current clamp for the hot water.
I have a heat pum that switches on at 10am for a few hours, 1kW draw.
I'm assuming these are not three-phase values and you're looking at the single phase? Might be time to get some more metering equipment. I tested many many meters working on the network and almost every single one was very accurate.
@@EEVblog2 Wouldnt have that issue with a gas HWS.. ;p
@@TobyRobb Yes, single phase.
As neither the smart meter or the inverter/battery system measures power directly the voltage & current variations between them will add additional noise to the data, beyond that of the pure MID accuracy rating, as you pretty much covered already.
Arguably the additional need to measure supply voltage and (simplifying here) raise it for export, balance it for 'zero' and lower it for import adds additional noise as the inverter/battery system cannot presume the supply voltage and has to sample it regularly. I presume that the inverter will have a set limit so that it does not chase the supply voltage spikes to above the regulated supply voltage, dropping the output and resampling until within limits. As an aside, here in the UK my supplied voltage is always very near the upper legal limits as it is advantageous to the suppliers. [UK = 230v -6% / +10% (or 216.2 to 253.0v). The lowest I have ever seen is 239.57v and is usually in the 247 to 250v range. Last month's highest supply voltage was a sparky 254.53v. The Dilithium Crystals can take it, apparently.
Phase management is another active system at play as the battery/inverter system will need to tune itself to the (relatively minor) phase fluctuations on the supply side. As with voltage and current measurement the inverter/battery system will respond somewhat in lag as it has to sample the incoming before tuning the export to the grid. This 'input sample & export' cycle will also show as a tiny import reading may also add additional noise that trends towards an import measurement.
Of course, everything needs to work as intended too... my SolarEdge system has basic mathematical errors in the displayed calculations.