Victron ESS - Commissioning, Testing and Snags

[NickNou]

 

Hi Gents,

Been tinkering on-off for a long time, but I finally managed to get a Victron ESS system up and running on my domestic system over the weekend. In principal, my ESS setup consists of the following:

Victron Quattro, 5000VA

Fronius 3.5 Primo grid tie inverter (on inverter output)

48V-825AH Battery bank

BMV 700

Beaglebone Black (Venus GX) managing the whole lot

 

I used to run the setup with a SMA SB4000 in a hub 2 configuration. Performance with this was slightly lacklustre, as the Multi/Quattro spends most of its time not connected to the grid to perform the frequency shift on the SMA. I hence decided to embark on the ESS route....

 

Commissioning:

 

Setup of the ESS is done as per the Victron manual - Easy to follow if you have all the bits. The system is quite intuitive to setup. 

 

Started the ESS system up too late in the afternoon to see if Venus GX could actively modulate the power output of the Fronius. After topping off the batts, the system settled into a absorption charge, and stayed there for the rest for the afternoon. Primary PV power modulation is done by Modbus TCP.  If this fails, I'm assuming the Victron should disconnect from mains, and make use of frequency shift control. 

 

Switching of Large loads and pre-paid meter trips:

Don't need to remind anybody here about the rigors of reverse power on prepaid meters. This is one shortfall I'll still need to tune out. 

 

Tested the ESS system with a 2kW kettle on the AC out. This is were ESS really shines (for a brief moment). The system allows the grid to supply the brunt of the initial full load. Gradually, the ESS system works down the amount of power being drawn from the grid and supplements this from either battery and/or PV power, up to a set maximum. Works very well - until the load is switched off. At this point the ESS configuration can't react/modulate fast enough, which sends a reverse power spike into the mains connection. Although very short, this is enough to trip the prepaid meter. Perhaps somebody has had similar experiences? I recall Plonkster posted about this (Hub-4) a while back.

 

My prepaid meter is a old Ecolec 570. I'm assuming I might need to have my meter swapped out and replaced. Is there a prepaid meter available that has a reverse power policy that is slightly better suited to this type of issue (i.e reverse power is determined over a time spans of 10-15 seconds)? Some discussion here - http://powerforum.co.za/topic/816-inverter-tripping-prepaid-meter/

 

Even a prepaid-meter that has Significant Reverse Power metering, and decrements credits for the minor reverse power events that occur would be OK. Landis GEM perhaps? Would appreciate some suggestions. 

 

I could set the grid set-point higher to allow more modulating margin for switching of big loads, but I don't want to be pulling unnecessary power from the grid if I don't need to. Even with this set at 50W, I'm getting meter trips when switching off large loads. I read somewhere that the rate of change of the ESS configuration is set at 400W/s. Perhaps this should be quicker?

 

Mains re-connection:

Each time the Quattro re-synchs to mains and connects, the Fronius disconnects and runs through its grid tests. I think this has to do with grounding of the neutral line of the inverter installation, were it momentarily hovers un-grounded during mains re-connection for a very short time. The Quattro does have an internal grounding relay that connect the outgoing neutral to earth during inverter operation and opens when the unit reconnects to mains. A hard wired N-PE link on the Inverter AC bus might help here, but this might* not necessarily be compliant.

 

Anti-Islanding protection:

Only doing this capital-expenditure once the system actually works.

 

Also, Big thanks to Plonkster who provided me with a development version if Venus GX (2.07) a short while ago with his serial starter patch for the mk3-usb. I've since updated to a development version of 2.08. Really worth while.

 

Nick,

[___]

5 minutes ago, NickNou said:

until the load is switched off. At this point the ESS configuration can't react/modulate fast enough, which sends a reverse power spike into the mains connection. Although very short, this is enough to trip the prepaid meter.

I have a Conlog BEC23 meter. According to its manual it will trip if you backfeed more than 40 watts over a 15 second window. With my small inverter (1600VA) the worst backfeed I've managed was around a kilowatt for a few seconds, and it pulls back fast enough from that to avoid tripping. In addition, when it pulls back it tends to overcompensate and draw power from the grid for a second or so. This means that the average stays nice and low. I've heard that some meters are a bit less forgiving that mine.

11 minutes ago, NickNou said:

Each time the Quattro re-synchs to mains and connects, the Fronius disconnects and runs through its grid tests.

I think it trips on ROCOF, Rate of Change of Frequency. When the multi resyncs with the grid, it does so by changing the output frequency to allow it to catch up with the grid, so that the waveforms are in sync once it switches. It is also possible that the output voltage jumps by a small amount, as the inverter is usually configured for 230VAC while the grid might be higher (mine is at 235V to 242V most of the time). Both these will be correctly detected as an islanding event by the Fronius. In other words, this is "normal" :-)

It could also be the earth-neutral bond. The multi removes the bond a second or so before it transfers back to the grid, leaving earth floating for a really short period. Far as I remember this is not on the list of things a device must do to conform to the various codes, so it's probably not this. Besides, not all earthing systems bond earth and neutral, only the TN ones.

The only solutions I can think of is 1) deliberately set a low inverter limit, or 2) there are devices available from AC/DC, I think it's made be Rhomberg, that can energise a relay when it detects reverse flow.

If you put the reverse-flow detection in a place where the inverter doesn't see it, but the prepaid meter does, you can us it to turn on a couple of bright lights (or something) for a second or so when there is backfeed, thereby absorbing it before it before the meter sees it. This is an idea only, I don't know if it will work and how well.

Or... replace the meter...

17 minutes ago, NickNou said:

Anti-Islanding protection:

The Multi already has this, and as I've argued before, NRS097-2-1 requires an active detection component, while the Ziehl device is passive only. In other words: The magic is in the Multi in any case. The reason for the extra hardware is simply that our regulations require two switches in series, one of which must be mechanical, and therefore you need to wire up only half of the Ziehl to control one contactor (you'll see in the wiring diagram that the UFR-1001e has two connections for redundancy and controlling two contactors, but you need only wire one of them).

[NickNou]

19 hours ago, plonkster said:

If you put the reverse-flow detection in a place where the inverter doesn't see it, but the prepaid meter does, you can us it to turn on a couple of bright lights (or something) for a second or so when there is backfeed, thereby absorbing it before it before the meter sees it. This is an idea only, I don't know if it will work and how well.

ACDC have the SP510 available, Rhomberg. I don't think this unit would be quick enough to respond to a sudden reverse power feed. You can only set the response delay between 1 and 10 seconds. Measurement is also not that accurate. On a 5A CT you can measure down to 100mA. Say you have a 60/5A CT on the line, you getting a sensitivity of 600mA. Rough maths puts that at around 130W. Not really great. I'm sure a bit of googling will reveal purpose made monitors for this sort of application.

19 hours ago, plonkster said:

1) deliberately set a low inverter limit,

Already running the 5kVA Quattro with a maximum inverter limit of 3000W.  I think a big factor here is that the PV is being fed directly into the AC -Output by the Fronius. Even if the Victron can react fast enough (i.e in the case where there are charge controllers feeding into the DC bus) and avoid trips, it still needs to absorb the incoming PV from the Fronius, or modulate power over modbus. All this takes (too much) time!

19 hours ago, plonkster said:

Or... replace the meter...

Will most likely be be simplest option here. But I need to get feedback on which meters are the most forgiving (Conlog, itron, Landis...etc) in absorbing the occasional reverse power spike. The 15 second window on the conlog BEC23 is workable, as the control loop to regulate the output of the Victron does over and under shoot, as you stated above. 

[NickNou]

Have some more time on my hands to test the ESS system again. 

PV modulation:

Modbus throttling of the Fronius works just fine. With no or little AC load, the Fronius is throttled to very low outputs quite effectively. This works quite well even when disconnected to mains. As a redundancy, I checked how well the frequency shift would work in the event of modbus failure. Victron disconnects from grid connection, and immediately starts with frequency shift modulation, pushing frequency up to around 52.7Hz. Fronius scales back PV output accordingly. No problems here.

Reverse power spikes:

Still have a problem with this. A typical example of a trip I had this morning:

• AC load: 3000W
• Battery discharge - approx 800W into system
• PV input +- 2150W
• Grid power - +-150W

AC loads drops to 1200W due to large load switching off - Mains drops out on the Pre-paid. This is the major issue I have on the ESS setup.  Still looking at the various options I have available.

Off-grid Fronius trips:

With the grid disconnected, the Fronius disconnects from the Victrons output every 3-5minutes, or each time that a Large-ish load connects or disconnects from the AC-output. All trips are of type 453, and are due to short term voltage variations on the AC line. Will most likely need to change some of the grid tie settings of the Fronius to prevent this.This is not really an issue when the system is running in parallel with the grid, as the mains connection keeps the system voltage stable.

 

 

[___]

Regarding settings for the Fronius, I suggest leaving a comment at the end of this page. One of the guys usually responds quickly.

Fwiw, the prepaid meter issue also happens with other inverters, though it has to be said that the Multi really is slow to adjust (though no slower than the Infini-with-modbus-meter setup, apparently). I do not know an off-the-shelf solution, though I suspect it should not be too hard to detect reverse flow and light up a load to absorb the backpush. If you had both a current transformer and a normal transformer, you could monitor the phase angle and easily recognise power going the wrong way (at least, this amateur thinks so). I think the trouble with other equipment is they react slowly because of communications- and measurement latency.

[___]

@NickNou, I also noticed just now that there's been some improvements made to the hub4control software. The limiter reaction time has been improved recently, though it seems this was mostly when the load picks up rather than when it goes away. I'm also not sure if this is included in 2.08 (which is a bugfix release for a sign-problem with the ET340 modbus meters) or if you might have to wait for the next release (which is also imminent), just a note that you might want to try the latest firmware if you haven't already.

Something else that might be useful, is either documentation on the prepaid meter itself, or some empirical tests (if possible). With the Conlog meters, the manual indicates that the limit is 40 watts over 15 seconds. This information is useful because it can be scaled to 600W for a second, 300W for 2 seconds, 150W for 4 seconds, etc. A 2000W accidental pushback can be easily compensated for by overcompensating a draw of 2000W towards the other side (assuming enough load is available), as long as the window within which the measurement is done is large enough to allow for that, ie, one might be able to fix this by purposefully overcompensating when backing-off as long as you know what the size of the window is within which you have to work.

A local installer told me that the Solar Edge inverters (with the Watt-Node meter) also works with the Conlog meters, it reacts fast enough to avoid tripping the meter.

[ghatikar]

@NickNou  You can consider reverse power relay ( normally used to protect generator ) such as this - https://mall.industry.siemens.com/mall/en/WW/Catalog/Product/5TT3425

or this - https://www.ebay.com/p/ABB-437W4790-Type-32r-Circuit-Shield-Reverse-Power-Relay-Range-13-150w-24vdc/1293283905

You will have to do the market search typically local generator suppliers will stock these. Instead of tripping the circuit you can use the relay to lsupply  up a large load such as dumping resistors or 500-watt bulb  

 

 

[NickNou]

20 hours ago, plonkster said:

might want to try the latest firmware if you haven't already

I'm currently running with 2.08-13. I did see yesterday that the 2.08 is out as a official release. Changelog does indicate that rev 13 includes the change for the 200W fronius zero feed bug. I'll flash with the release version of V2.08 this weekend. I'm not running with a grid meter, and currently make use if the monitoring inside the Quattro itself.

20 hours ago, plonkster said:

Something else that might be useful, is either documentation on the prepaid meter itself, or some empirical tests (if possible)

The Ecolec 570 was discontinued ages ago. There is little/no documentation available, and all the info I've been able to attain hasn't been of much use.  I do know that the Ecolec is fitted with an internal RCD monitoring circuit, where you can configure the earth leakage trip. I thought this might have been the issue, but I saw tonight that the Earth leakage protection on my specific meter is disabled (You can select from OFF,  or in the range of 10-300mA

I'm going to endevour and have the ECOLEC swapped with a BEC23. Might take forever with my local utility. I've heard from other installers that the BEC23's meters are not that forgiving, so I dont know if I should be considering other options. Nevertheless, you can do alot with a reverse feed trip based on 40W over 15 second window.

The Ecolec trips almost instantly (less than a second after I remove the load) and its been a problem here before.

21 hours ago, plonkster said:

A 2000W accidental pushback can be easily compensated for by overcompensating a draw of 2000W towards the other side

This is true. The ESS controller does overshoot, but settles very quickly. Even though I have meter issues, I must say that ESS does have a very well tuned control loop.

[NickNou]

1 hour ago, ghatikar said:

@NickNou  You can consider reverse power relay ( normally used to protect generator )

One could consider this, but most of the local relays that I've had a look at have time delays starting at 0.5 seconds and up. With my current prepaid meter, the trip occurs almost immediately. It would be great if you could dump the spike into a set of immersion heaters or similar. 

[Louisvdw]

Are there any settings that changed in the ESS since these posts that improves the prepaid meter trip?

I have the Conlog BEC23 meter and in normal instances the ESS can recover just fine so that it does not trip. But about once a day it does trip and it does seem that this happens when the oven is in use and stops.

I have moved my Grid Setpoint up to 300W, but that did not help. Anything else I can set (would like to not go the reverse power relay route if I can avoid it)

[___]

3 hours ago, Louisvdw said:

Are there any settings that changed in the ESS since these posts that improves the prepaid meter trip?

A lot more research has gone into it. Nothing has changed though. Because this needs the entire chain to be improved.

European requirements are that you must pull back 90% of the way within two seconds, and then you can gradually pull back on the remaining 10%. At the moment, the ESS loop takes about 7 seconds to get 90% of the way there and over 15 seconds to get all the way there.

Now the bulk of the slowness is in the control loop on the GX device. By initial testing, it should be fairly easy to get it down to about 5 seconds or so.

The Multi itself takes a good 1.3 seconds to respond after the control loop gives the instruction to pull back on the power. In addition, the fastest meter (EM24) has a refresh rate of 600mS, and the worst one (ET340) has a refresh rate of 2 seconds (I am not making this up, it is THAT slow). The ET112 that most people use in single phase installations have a refresh rate of 750ms.

The refresh rate is how often you get a new reading from the meter, best case scenario.

If you don't use an energy meter, then it gets a little worse. We get a reading from the Multi once every 3 seconds or so. Not going into the detail, cause there are a lot, but there are reasons for this.

So all of this is just to explain that the best that can be hoped for now, without redesigning large parts of it (which will happen eventually) is between 2 and 3 seconds. On  BEC44 (300 joules tripping limit) that's 600W at best...

In other words, I'm telling you not to get your hopes up too highly. There is very likely no rabbit that can be pulled from a hat soon.

But... you can still use the energy diverter option that I've mentioned a few times...

[Louisvdw]

Thanks for the update @plonkster

I have played a bit more now during today's bake-off (skuimpies & brood). It does seem to be a bit better if I put the Grid Setpoint to 600W. This gives it a much larger buffer in which to reduce before the meter trips. Skuimpies (not sure what that is in English??) are baked at only 100Deg C, so the oven element switch on/off much more that say for instance with break which is baked at a higher temp. so the element does not switch off that often.

It will be a pain to switch the Grid Setpoint each time there is a bake-off. Perhaps I should get a hack going on the Pi to do that automatically for me 

Oh, I have the ET112 meter, so that will be 750ms refresh.

[Jaaks]

Skuimpies = Meringue 

I have Multi 2 and am very fortunate to get no trips. 

Only issues iare AC reconnect after switching off the inverter. Incorrect SOC even though I have a BMV (it slowly drops over a few days till I resync) and once in a while overvoltage alarm on my home built lithium battery pack (only for a second or so so no worries there) 

[Louisvdw]

1 hour ago, Jaaks said:

Skuimpies = Meringue

Thanks Jaaks  

There is a way to fix the SOC value if it concerns you, but it is a manual way. You can use a terminal client to log into the VenusOS device using SSH and then run dbus-spy. Go to the Multi and search for the SOC and edit it to the correct value. I had a similar issue (but don't have a BMV).

[___]

33 minutes ago, Louisvdw said:

There is a way to fix the SOC value if it concerns you, but it is a manual way. You can use a terminal client to log into the VenusOS device using SSH and then run dbus-spy. Go to the Multi and search for the SOC and edit it to the correct value. I had a similar issue (but don't have a BMV).

You can also reset the BMV. If you know the vreg, you can write the state of charge to it, eg:

dbus -y com.victronenergy.battery.ttyUSB0 /Devices/0/VregLink SetValue %0xfff '%[0x88, 0x13]'

The above will set it to 50%.

What is going on there? Well, the % prefix tells the dbus command to interpret what follows as a python literal, so that makes it properly interpret the hex values and constructs a list as the second argument. 0xfff is the vreg for the SOC on battery monitors. And the value itself is a 16-bit value (therefore two bytes) with the LSB first. So the above is 0x13 (19) multiplied by 256 + 0x88 (136) = 5000, which is 50% with two decimals.

Why would you use it? Well at least one customer had an issue where on his boat he disconnects everything before transport, and he'd like to put it back to the old values when he reboots next time