There are various ways of wiring up a multiplus to the grid for a PV + ESS system and various ways of monitoring exported power. I have read some conflicting info and most of it is on the old forum so a new thread on this seems useful.
The main requirement is accurate power control so that a) no grid importing occurs unless it is needed and b) there is no export over the export limit. I read that the energy meters (e.g. ET112) respond more slowly than a current transformer (because of the reporting period and serial buses, as opposed to instant analogue response for CT clamp/internal CT). Is this correct? Is there any significant difference in performance?
Is there any difference in terms of the GX interface? If you add an energy meter at the grid exit point do you now get two data values: one from the meter (import/export) and one inside the machine ‘power through the MP II’? As opposed to adding an external CT which replaces the internal CT? Presumably the internal CT measures what goes in/out of the ‘AC IN’ line - right?
There are also other considerations, e.g: I don’t need an energy meter (or external clamp) at all if the multiplus is connected directly to the grid and all power goes through it. That creates lower power limits on the outputs than direct grid connection, so that is one tradeoff.
To clarify the situation this is a grid-connected single-phase ESS system with multiplus II 48/5000, 2.9kWp of AC-coupled PV (derated to 2.65kWp) and 4.2kWp of DC-coupled PV (derated to 3.9kWp).
or just using AC-in ‘single-leg’, using an energy meter or CT clamp, and without any islanding/battery-backup features (no pics for that yet)
Obviously some of these are harder to retrofit than others, but just considering how the import/export monitoring is done are there any differences between the 3 monitoring options (internal CT, external CT, energy meter ET112) which might affect which of these layouts one chooses? There are of course other energy meters and other current transformers. Those might be worth discussing too.
@wookey I dont really know if the delay of the ET112 is appreciable. It uses RS485 ( probably at 9600 baud) but would only be sending a few bytes in a Modbus frame. The delay may be slight but do you have any other choice as I didn’t think you could attach an analogue CT to the CerboGX?
If the entire house is supplied by the multiplus the internal current measuring would age to be the choice but it does limit the house load. Perhaps two
Multis in parralel may be the answer.
It is interesting you have both AC PV and DC solar used. I’m not sure if that gives you some advantage but would be keen to hear about it if it does.
My experience is that the system responds better and calculations are more accurate with energy meters over internal measurement
In my first setup (single phase ess) is used a victron CT on AC PV and an internal on grid (next to dc pv i already had) the CT functioned poorly and i replaced it for a ET112 and an et340 on grid
A year later i expanded to 3 ph and swapped the ET340 for an EM540, and switched from ac pv to all dc, the EM540 is so much faster in response the ess is working much better in my opinion
The topic has already been discussed in great length. For me it came down to reducing the slight ramp up lag with e.g. an ET112 when introducing big loads, which improves noticeably when using a CT. This at the expense of slightly better accuracy of energy meters. There are also meter refresh rates to consider, which you can find in the selection guide on the Victron website. Ease of installation may also may influence your decision as CT clamps are easy to fit, especially if your consumer unit is already fully populated. My rough tests showed a slight grid consumption increase of about 0.4-0.8kWh when using a grid meter, which is as a result of the ramp up lag encountered with mainly high loads. This will vary by household. Finally, there is GX latency, but I don’t see that to be much of a consideration as you don’t have much control over it.
The ET112 uses RTU mode. A modbus RTU frame is 7.5 bytes plus however many the ET112 sends as data. In ASCII mode that’s 9 bits ber byte. In RTU mode it’s 8 bits. Lets say 5 bytes of data (I failed to find the actual numbers, just a list of 13 possible registers. There is a request frame and a response frame each time, so I think that comes to about 12.5 bytes at 8 bits each which is 100 bits, which is about 10ms at 9600 baud, so yes that’s not significant in comparison to the 750ms (0.75s) refresh rate of the meter (another datasheet says 1s, not sure which is right). A 3kW load is 0.83Wh every second, so an average refresh time error of 0.5s is 0.4-odd Wh ‘escaping’ before the new power level can be read.
Turn that load on 100 times a day and that’s up to 40Wh or 0.04kWh of error. Our daily consumption is ~3kWh so that’s a 0.1% error which seems acceptable.
I don’t know what the response latency of VenusOS is, or the size/duration of the VE.direct/CANbus message it sends, but it should be ‘small’ in comparison to 0.75 or 1s so we can probably ignore that bit too.
So it’s not a great instantaneous power meter, but it does seem like it should be ‘good enough’. But equally, there must be modbus meters that read more often than 750ms /1s? The protocol allows for reading every 10ms or so. Certainly 50ms would be fine, and the actual device is sampling at 4kHz (0.02ms). It’s not clear where this 750ms limitation comes from.
When you say a grid consumption increase of 0.4 to 0.8kWh do you mean per day? That would be a lot (13-26% of daily usage in our house), and what I would call ‘very bad’.
When you say it has been extensively discussed can you point to any threads? I could not find any numerical or test data on the performance of these various options when I searched. I did find some interesting discussion on using both an energy meter and a CT clamp: Error 24, External CT is checked in Ve.Config. - VictronEnergy but it wasn’t conclusive.
There are many other modbus meters available:
Rayleigh RI-D35-100-C, Givenergy GEM120, Eastron SDM230D, Sinotimer DDS6619, several no-name Chinese devices. Some protocol interfacing would presumably be needed, but they might have better latency. Sadly none of the datasheets I’ve found actually give that info. I’ve sent email to a few to ask.
My experience with a very slow, but precise meter vs the builtin meter of the MP-II:
In Austria, we have access to the utility meter’s data which get sent every 5 seconds, I inject them via dbus-mqtt-grid.
With this meter, the grid setpoint of the 3-phase ESS is matched precisely, in the sense that with a setpoint of 10W, after an hour the meter will register 0.01kWh exported.
When I had it configured to use the Multi’s internal metering, hourly export was about 30 to 40 Wh.
On the other hand, regulation is much better when using the Multis, with the 5 second grid meter it takes a long time to stabilize when big loads are turned on or off. This looks terrible on the graphs, but in the end, in terms of energy, it doesn’t matter.
Yes. That will however vary depending on various factors e.g. grid setpoint, number of large startup surges. Even set at 0, there will always be slight movement to either side all day long. This is about 5% of my daily consumption and also includes the occasiol call on the grid when loads exceed the 4.4kW of the inverter.
An example on old forum. Even Victron looked at it at some point.
