Victron Quattro Grid back-feed.

[Harryc]

Hi

My Victron Qautro feeds back to the grid. This causes my prepaid meter to trip all the time

I have disabled feed-back in the ESS setup and have set the grid limit to 150 watt which defeats the whole purpose of having a system as it still consumes Eskom power.

Even at a 150 watt it still feeds back every now and then.

Is there a way to overcome this?

Thanks

[Jaco De Jongh]

28 minutes ago, Harryc said:

Is there a way to overcome this

2 ways only. Run in HUB1 or replace the prepaid. Last option will be the best.

 

[___]

49 minutes ago, Jaco de Jongh said:

replace the prepaid

These prepaid meters are becoming a real problem. The trouble is that even though the system is somewhat slow to back off (7 - 15 seconds), even the strictest speed requirements (parts of Europe wants 90% reduction in <2 seconds) will still trip these meters. The meter manufacturers really need to be a bit more lenient. I understand that you want to sell anti-tamper features to a municipality, but surely it doesn't have to be so darn hairline sensitive.

Take for example the Landis GYR meters with the SRE detection feature (significant reverse energy). It can still be programmed to go into tamper mode, but it only triggers if you are feeding hundreds of watts in for many minutes...

[phil.g00]

I think this MAY work. Morningstar makes a PWM charge controller, well actually two a 60A and a 45A version. ( Not the Morningstar MPPT versions !)

This unit can also be used in load diversion mode. What is that you ask?

It is generally used for wind generation, where the wind turbine is connected directly to the batteries. Generally the battery voltage clamps the wind turbine voltage, but in high winds the turbine voltage wins out. That is when load diversion steps in, and it a PWM fashion at a certain threshold the excess voltage is bled off into a rated dummy load.

According to morningstar they do not have a rated time for how responsive this feature is other than to say it is pretty much instantaneous.

Outback MPPT's have a similar feature, but their PWM output must control an SSR onto a dump load.

You are feding back into the grid because of a short term voltage rise, this MAY be able to prevent that.

I do not have this issue and subsequently will probably never prove/disprove this theory.

But, for someone that it's becoming a real PITA, it maybe worth experimenting.

[Harryc]

Hi guys, I see there is a VICTRON ENERGY METER ET112 SINGLE PHASE available, do you know if this will help?

[Jaco De Jongh]

5 minutes ago, Harryc said:

Hi guys, I see there is a VICTRON ENERGY METER ET112 SINGLE PHASE available, do you know if this will help?

Regretfully no. No meter will stop the system from dumping into the grid when a load switches off. Most electronics is to slow in making adjustments. I had one of these meters. We have build noumorous devices to try and act fast enough to counter this, but nobody thus far could build something that could react faster than the prepaid. I had to go to the munisipality to get the prepaid replaced. 

[___]

34 minutes ago, phil.g00 said:

You are feding back into the grid because of a short term voltage rise, this MAY be able to prevent that.

Regretfully this is not what is happening. Simply, the inverter feeds an amount of energy into the grid to cancel out any of your local loads. Then one of those loads switches off, so now the inverter is feeding more energy into the grid than the loads are using. The remainder goes into the grid. It now takes a certain amount of time before the inverter reduces power. This is for two reasons, the one is that closed control loops (which is a deep technical discussion, I mean deeper than I'm already going here, that I want to avoid now) needs to be dampened otherwise they may become unstable. The second is that the grid measurements are only taken about once a second (and the best case is 600ms... which is not much better).

The amount of watts multiplied by the number of seconds that this is going on gives you a number of joules (a joule is literally a watt-second). If that number gets too high, the prepaid meter trips. For Most Conlog meters this number is 600 joules (BEC23) or 300 Joules (BEC44). As you can see... WAY to darn sensitive.

To expand a bit more, for the benefit of the OP (which is short for original poster... ), AC is a bit of a difficult beat to measure. It is not as easy as DC where you just measure which way the current is going, and then you know which way the power is (also) going. With AC the current literally changes direction 100 times a second, and so does the voltage. Now as long as the voltage and the current goes the same way on average, then power is going in the "forward" direction, towards the consumer, but if the current goes in the opposite direction (out of phase, in other words) to the voltage, then power is going towards the grid.

Now we get to the cost of the meter part. Cheaper meters (eg the chip in the Sonoff POW) gives you a value averaged over several seconds. Really good meters (Smappee Powerbox) gives you a value every 100ms. The former device costs R350, the latter costs R4500. You can get even more accurate... but then the price goes to 45k and more... and you still have to account for the communication delays, the adjustment delays, etc.

And I haven't even gone into the matter of non-linear loads which seriously complicates power measurements...

So long story short, to avoid the issue these meters create would require very expensive metering devices and fast control loops. And it's frankly just not worth it.

[___]

Here is a video explaining how to measure power with an oscilloscope, which gives you an idea of what it takes. It gets very technical. But from 7 minutes on, it shows how to use the math function in many modern DSOs to multiply the two measurements. Around 08:30 you can see the red result of the multiplication, and you will note that on every cycle a little bit goes in the reerse direction (where the red line goes under the zero line, reactive power). But you can visually see that the power is on average going in a forward direction.

I've actually repeated this on a DSO, with AC power, and it works perfectly. On my scope I can even ask the "math" module to work out the average of that waveform.

Now every power meter, including prepaid meters, does this sort of math and then averages it over a certain window of some milliseconds or seconds.

[phil.g00]

@plonkster, I am aware of the of circumstances under which this occurs, but it stills boils down to a sudden change in potential difference ( voltage) between the grid and the load side.

By definition power cannot flow without a potential difference across the connecting impedance, with AC this can be a phase difference or a magnitude difference or both.

Power flow is not a cause it is a response. The connecting impedance is a constant, the current flow has changed, therefore the voltage must have changed - Ohm's law.

It happens too fast that the feedback loop cant keep up, I know. 

Which I why I proposed what I think is the fastest clipping mechanism.,( and does change the impedance).

(If I recall this voltage spike is also in Victron documentation and is the basis of the 1:1 rule).

Edited August 11, 2020 by phil.g00

[___]

1 minute ago, phil.g00 said:

@plonkster, I am aware of the of circumstances under which this occurs, but it stills boils down to a sudden change in potential difference ( voltage) between the grid and the load side.

Oh I know, that's more for the benefits of others who want to dive that deep You very probably know more about this skubala than I do...

1 minute ago, phil.g00 said:

Which I why I proposed what I think is the fastest clipping mechanism.

I agree, but it needs to be clipped on the AC side. an energy diverter, which has been mentioned a few times, at least by myself. Like this one.

Trouble is, this level of DIY is beyond the reach of most people that have this problem.

With such an energy diverter, it essentially does a crude form of the above measurement, taking a number of samples of both current and voltage, averaging, and switching some kind of switch (probably a TRIAC) with it. Because it does not have to be very accurate, nor does it have to do digital communication or calculate actual energy, even a small MCU is fast enough to do the job. The only trouble, as I said, is that DIY needs skill, and whatever is commercially available is 1) not in this country, and 2) not cheap.

[phil.g00]

21 minutes ago, plonkster said:

I agree, but it needs to be clipped on the AC side.

To this I say "maybe, maybe not".

Although again the Victron documentation (ala the 1:1 rule) the batteries also experience a voltage spike and I would expect it.

I believe, (though like I say, I am not prepared to prove it), that clipping that DC voltage rise instantaneously will have the same affect as doing it on the AC side as we are dealing with a bi-directional device.

I agree that messing about on the AC side a fools errand, but the DC side that's a battery voltage usually somewhere between stable limits and of course no phase differences to contend with. So its far easier to deal with.

I did caveat my proposal, but it MAY work, and it is an off the shelf solution, that is designed to dump DC energy very quickly.

Essentially, what it does is instantly present a DC load absorbing that DC voltage spike, giving that feedback loop time to respond.

 

 

Edited August 11, 2020 by phil.g00

[___]

12 minutes ago, phil.g00 said:

Although again the Victron documentation (ala the 1:1 rule) the batteries also experience a voltage spike and I would expect it.

Aaah I get it. You're talking about AC-tied PV. I'm talking about DC-tied... a bit more common in these parts

I know that in an off-grid system with AC-tied PV the transformer coupling results in a DC voltage increase, this is in fact how the PV-inverter assistant regulates the frequency. It changes the AC frequency relative to the battery voltage.

A reduction in load does not necessarily cause a rise in the DC voltage when the PV is DC-tied and the inverter is feeding a constant amount of power into the grid based on a control loop that updates it roughly every 3 seconds, and hence clipping it on the DC side will not work.

Messing around on the AC side is your only option, sadly. Commercial energy diverters do exist. In the UK, people use them to turn on the hot water cylinder once the PV starts pushing into the grid.

Edited August 11, 2020 by plonkster

[Jaco De Jongh]

11 minutes ago, phil.g00 said:

Essentially, what it does is instantly present a DC load absorbing that DC voltage spike, giving that feedback loop time to respond.

Just a question, taking the reaction time of an normal MPPT into consideration, do you think this would be faster than the detection time of the prepaid. They are really fast in detecting reverse current, believe me, I had one in my house and had it replaced. Luckily for free. 

[___]

2 minutes ago, Jaco de Jongh said:

really fast

I work from the premise that if 300 joules trips the meter, then it means 300W for 1 seconds, 600W for 0.5 seconds, 1200W for 0.25 seconds.

An AC cycle is 0.02 seconds, and I doubt even these pesky prepaid meters measure power to the nearest 20mS. They will average it over a number of cycles, possibly at least 5 or so. That gives you 100ms to work with.

So if you can build something that can average power into 100ms buckets and turn on a TRIAC for a few cycles (as many as is required to get the average down to below 300 joules), you're going to be fine.

You also don't have to hold it at zero. You can compensate for 1000W going out for 0.5 seconds by dumping 2000W into a kettle for 0.25 seconds. The prepaid meter will probably still see zero if you make your time intervals short enough.

[phil.g00]

5 minutes ago, plonkster said:

Aaah I get it. You're talking about AC-tied PV. I'm talking about DC-tied... a bit more common in these parts

No, I was drawing the analogy of losing the grid ( a step change in AC feed-in conditions) to a sudden change in load ( another step-change in AC feed-in conditions) and the inertia inherent in the feedback loop. 

9 minutes ago, plonkster said:

A reduction in load does not necessarily cause a rise in the DC voltage when the PV is DC-tied and the inverter is feeding a constant amount of power into the grid based on a control loop that updates it roughly every 3 seconds, and hence clipping it on the DC side will not work.

Maybe you are right about this, but I don't think its cut and dried,  increasing voltage on one side of a transformer is mirrored on the other-side.  Really need a scope trace to say for sure.

[Jaco De Jongh]

4 minutes ago, plonkster said:

I work from the premise that if 300 joules trips the meter, then it means 300W for 1 seconds, 600W for 0.5 seconds, 1200W for 0.25 seconds.

An AC cycle is 0.02 seconds, and I doubt even these pesky prepaid meters measure power to the nearest 20mS. They will average it over a number of cycles, possibly at least 5 or so. That gives you 100ms to work with.

So if you can build something that can average power into 100ms buckets and turn on a TRIAC for a few cycles (as many as is required to get the average down to below 300 joules), you're going to be fine.

You also don't have to hold it at zero. You can compensate for 1000W going out for 0.5 seconds by dumping 2000W into a kettle for 0.25 seconds. The prepaid meter will probably still see zero if you make your time intervals short enough.

This is all great in theory and we mentioned this a few times in the past, but to build something like this proves to be another story all together. The cost involved as well makes me believe that asking the municipality to change the meter might remain the best option overall, especially if it can be done at no cost like with mine. 

[phil.g00]

16 minutes ago, Jaco de Jongh said:

Just a question, taking the reaction time of an normal MPPT into consideration, do you think this would be faster than the detection time of the prepaid. They are really fast in detecting reverse current, believe me, I had one in my house and had it replaced. Luckily for free. 

I'd have to read the documentation, but I think the sampling rate is in the kHz, least-ways that's where I'd expect it to be for PWM.

Edited August 11, 2020 by phil.g00

[___]

4 minutes ago, phil.g00 said:

Maybe you are right about this, but I don't think its cut and dried,  increasing voltage on one side of a transformer is mirrored on the other-side.  Really need a scope trace to say for sure.

I agree about the AC/DC coupling, but perhaps to help you get an idea of what we're dealing with: Imagine what happens at night. There is no PV power coming in, it's just an inverter running fro a battery feeding power into the grid, with a closed control loop that runs from a grid meter, attempting to get the meter to zero. The control loop is slow...

 

[phil.g00]

2 minutes ago, plonkster said:

agree about the AC/DC coupling, but perhaps to help you get an idea of what we're dealing with: Imagine what happens at night. There is no PV power coming in, it's just an inverter running fro a battery feeding power into the grid, with a closed control loop that runs from a grid meter, attempting to get the meter to zero. The control loop is slow...

 

You can overcome attempting to get the meter to zero by setting a marginal in feed value.

The control loop is fine for an accumulation of small and gradual changes, it's that sudden step change that throws it a wobble, when that margin is not sufficient.

And obviously its not prudent to set that margin sufficiently to cater for step changes or we'd be constantly drawing power from the grid unnecessarily.

Yes,  I appreciate the dilemma.

 

[___]

7 minutes ago, phil.g00 said:

You can overcome attempting to get the meter to zero by setting a marginal in feed value.

Indeed. And that is also done to improve signal/noise ratio. At zero it becomes horrible. Conceptually it is easier to think in terms of "zero", but of course the default is to import 50W.

8 minutes ago, phil.g00 said:

The control loop is fine for an accumulation of small and gradual changes, it's that sudden step change that throws it a wobble, when that margin is not sufficient.

Spot on. Which is why it helps to increase the setpoint. It gives you a margin to work with. For example, if you set it to 150W, and its been running at roughly 150W for the last 7-8 seconds (the BEC23 has a 15-second window), that means you have 7*150 ~= 1000 joules in "credit"

I was in a meeting the other day with people who wanted a more accurate meter. The short answer at the end was: If it was cheap and easy, you'd be able to buy one from the shelf.

[Louisvdw]

5 hours ago, Harryc said:

Hi guys, I see there is a VICTRON ENERGY METER ET112 SINGLE PHASE available, do you know if this will help?

I do have the ET112 and that does not help. 
What I have found is that this is normally a problem when a large load switch on/off like my stove. 

My setup is just to assist with power, so if I use more power with the stove and use the solar/battery in another place I still will the same.
So I have set my grid setpoint to 150W and when there is baking happening I set it temporary to 600W. This seems to fix the issue for my meter (BEC23)
It's very manual process at the moment and while I am working from home that can work. I'm still looking at a way to automate that for when we return to the office.

[daniemare]

Didn’t read through all the technical replies. But I spoke to the technical people at Conlog (the pre-paid meter supplier for S’Bosch). Whilst the tripping of the meter is due to the various anti tampering measures in the meter, this specific issue is easily disabled by a meter specific code generated by Conlog. 

They need a written request and authorisation from the Municipality. Which is now treating this as a practical double check to check if you got your solar installation approved. No approval no code. 

Funny thing is that the reputable solar installers did not know the cause of this tripping, or the solution when I enquired into this phenomenon as part of my ongoing system research. To such an extent that my neighbour’s system is set to 500W (as opposed to 0) export limit. His power bill is R200 less than my no solar system R1,200 per month bill, after a R210k 14kWh battery investment. 

[Harryc]

Thanks everyone.

It seems like the only way to sort it out is to replace the pre-paid meter.

Thanks

Harry

[___]

4 hours ago, daniemare said:

They need a written request and authorisation from the Municipality. Which is now treating this as a practical double check to check if you got your solar installation approved. No approval no code. 

You're lucky with your municipality. Some areas in the country... the eyes turn a little glassy... and then that's the end of it.

[daniemare]

31 minutes ago, plonkster said:

You're lucky with your municipality. Some areas in the country... the eyes turn a little glassy... and then that's the end of it.

They also didn’t know. But I got the Conlog guy and the Municipality guy to talk to one another. A light went on for the electricity department (because he personally was scratching his head because of 3 tripping case complaints he was dealing with) as he got his answer for the tripping and a way to check to see who applied or not