___

If you are not an electrician, or have some sort of practical experience in that area, I must advise that you employ the services of a professional to do this. What you have to do is "split the DB". Some people install a new separate distribution board and move their backup circuits to this new DB. Others make a physical separation in the existing DB and clearly mark it. Splitting the DB in this manner requires that for each circuit you must identify both the live and the neutral conductor serving that circuit (and hope there are no shared neutrals or other shenanigans left behind by previous cooks in the kitchen). The backup circuits get their own RCD (earth leakage) on the output of the inverter, as well as their own neutral bar. You usually also need to install a changeover switch, so that the backup circuits can be connected to the grid instead of to the inverter. Something like the Hager SFT240.

When connected to the grid, both AC-1 and AC-2 are connected to the grid and are, as you put it, un-inverted. But the Multi will disconnect from the grid if the voltage moves out of the acceptable range, which should help. Personally, I'd put in some SPDs (surge protection devices). Just remember to put them on the grid side BEFORE the RCD (aka earth leakage).

There is always some sort of algorithm in the BMS that it uses to decide when to go to 100%. Usually it is something like 1) all cells above 3.45V and 2) at least one cell just spiked above 3.5V. One reason why seeing per-cell voltages, or at least the lowest and highest voltages, is a very good diagnostic tool.

First, I'd like to encourage you to get the units right, cause it's a little confusing now. The units are kWh (for energy, eg, if you turn on a 2kW kettle and leave it on for an hour, it will consume 2kWh of energy), and kW (a measure of power). It sounds like you want a 5kW inverter, not a 5kWh one Regarding gas: If your goal is to become independent of Eskom, have hot water during load shedding, or you're running a guest house and absolutely must have hot water at all times, then gas is a good idea. If your goal is to save money, then don't do gas. Gas is the most expensive heating option (unless you have a very efficient "geyser" and a natural gas piped to your estate). A heat pump and or solar water heating (depending on how much sun you've got) saves far more money. I agree with the comment about installing an A++ rated fridge. Three pluses is even better, of course, but diminishing returns and all that, each + means 10% less energy, but you get to a point where an additional + doubles the cost... not worth it imho. Bosch A++ unit are common and affordable. As others have said, Victron does not make the only hybrid. But they do make the best one... (yeah, I know... I know... ). Getting back to gas again, seriously. 14.6MJ of energy in a kg of LPG. Convert to kWh, that's around 113kWh in a 9kg bottle. Divide the cost up, around R2/kWh per unit of energy. Compare yo what you pay for electricity, chances are it's very close (Cape Town is R2.56/kWh, so gas is technically a tad cheaper). But a resistive element inside a tank of water is close to 100% efficient, and gas is nowhere near 80%... so it's a simple high school physics question that some teacher really should slip into the matric paper: Gas is not cheaper.

Sounds as if the inverter gets its SOC from the battery. BYD batteries can drift quite a bit, because their current sensors can't sense well below 2A (per module, multiply by number of modules). So if your batteries have been subject to a low charge or discharge currents (in this case I suspect slow charging to be the culprit), the battery SOC might well have been higher than the BMS estimated... until the cell voltage suddenly went up and the BMS realised it had to reset.

Did the wife then also tell you that really, you should maybe have used a professional?

Correct. The LV hub runs the same protocol as the mater battery. Just rememer to set the DIP switches to correctly terminate the CAN-bus. I believe the unit ships unterminated by default, an easy thing to overlook.

What version of Venus are you running? In Veus 2.60, the charge voltage for Pylontech batteries have been increased to 52.4V. Pylontech actually wants us to take it to 52.6V, but that will happen later. In Venus 2.5x, the charge voltage is still 52V. Assuming you have DVCC turned on and all that. The Multi should not be going up that high... Sometimes the spike in voltage is not the cause of the problem, but the result. If the BMS disconnects because it has a problem with a high cell, that often causes a spike in the measured voltage (because now there is no battery to weigh/dampen it down, just the capacitors on the DC bus). So my guess is that you have a high cell somewhere, imbalance issue. And you need the BatteryView software to diagnose that, unfortunately.

It is not only best, it is required! 1. Because if you don't, your earth leakae will trip repeatedly, and 2. Because SANS-10142-1 requires it. There are a few rare cases where you can share the neutral, but none of that applies here

Usually you can get some clue by looking at the output of /data/log/messages, or by just dumping the kernel log ring with the dmesg command.

Not exactly. If you want 0x10 (16) to show up as 1.01 (the closest possible under the Victron scheme), you need to change it into 0x101, or 257. Not that hard luckily, just bitwise AND with 0xF0 to get the MSB, shift left by 8 bits, then add the LSB (which is the original bitwise-AND with 0x0F).

The address is ascii, the idea is to make it flexible so battery makers can use their own representation. Most of them simply use integers. It gets more interesting when you have multiple modules making up a larger battery. Then the high cell might be in module 12, cell 4...

And put the version in /FirmwareVersion. It has to be hex coded though, so you want to send 0x202 (version 2.02), or 514 in decimal. To translate the ascii, take the character value and subtract 0x30 (48, or ascii zero). The difference is the digit number. Then simply shift that number left by 8 bits (512) and bitwise OR on the remainder.

Since you have all the cell voltages, also publish the maximum cell voltage/temperature and the address of the cell to: /System/MaxCellVoltage /System/MaxVoltageCellId /System/MinCellVoltage /System/MinVoltageCellId And put your total installed capacity in /InstalledCapacity.

I've done that, back when I was developing the pulse meter implementation for the Venus-GX (it was for water pumps... but this seemed like a good way to test it): Also, I made this some years ago, or rather, I fixed it up so it works again, an Android app that uses the camera to count pulses. 10649264_10152684751795619_86532852_n.mp4 There is no intelligence in it, it simply uses the average value of the whole photo frame, ie you need to hold the camera close to the LED. I wanted to make it so it can also upload the value to "the cloud" but I kinda lost interest when I found out you can just punch #1# into the conlog to get a power reading.

Yup. And somewhere between 15mA and 20mA you start to get nuisance tripping. Then you have to split the circuits and install two RCDs. There is also leakage on the neutral side, because often there are surge arrestors/filters between earth and neutral too. Since earth and neutral is bonded, there is almost no leakage here... until the bonding relay in the inverter opens (to do the relay test). It is quite a bit harder to find these little leakers... cause opening the breaker does not interrupt the neutral.

Right. So I suppose you were right and I was wrong. Mostly. The leakage is indeed very low. I faked a neutral earth bond, and rapidly connected/disconnected it, and it does increase the leakage, as much as 4-fold, but still only about 40nA... maybe, I doubt the clamp meter is terribly accurate that low down. The clamp meter samples 30 times a second so I figured even if the transient is much higher, the duration is going to be so short I doubt it is worth it to set up the scope. Still, I saw what I saw at the time. Somehow, this unholy gathering of filters and arresters worked together to make my life difficult. I still have the EMI filter from another part of the house. Somewhere. This one actually makes a small spark when you rapidly connect and disconnect the TN bond. Not enough on its own to trip an RCD (already tested), but possibly enough to bother setting up the scope for. This thing:

I moved house in the mean time. And the opener itself also broke and got replaced. I kept the guts of the unit specifically to test it in more depth later. When I find time Edit: for interest sake. It's an ET DC Blue. The entire thing is low voltage (24V battery, motor, etc), with a really decent old-school power supply in the back (toroidal transformer, etc). It needs an earth because it has exposed metal parts. For whatever reason, they decided to also put the little triad of MOVs on a board next to the fuse...

Indeed, which is why a proper current probe is on my wishlist. Makes it so much easier, especially when you apply the Math function, it can show the right unit (VA) instead of V^2. Speaking of scopes doing the right thing. I have an ancient (in the 90s it qualified as high end) Kenwood analog scope. It auto-detects probes, eg if you put on a 10:1 probe, it knows it automatically. Two weeks ago I acquired a brand new Keysight scope. Even though entry level, I was surprised to find that it does not have this feature. You have to manually select the attenuation. But I'm not complaining, it is an impressive scope for the money. It's an EDUX1052 (the edu means this is the castrated one they make for education places).

Since then, my collection of testing kit has expanded somewhat. I actually have an insulation tester now. And I've learned that 99% of the time, an RCD trip is because you have an insulation problem somewhere. It's been a bit of a quest of mine to understand the remaining 1% (Next bit of kit I want... a current probe for the scope. Man these things are expensive. For now I'm using a CT with the normal voltage probe... technically the same thing, if a bit crude).

Indeed, this is what I suspect to be at fault. All I can tell you, is that i did an empirical test: 1. I unplugged the garage door opener, and interrupted the power, forcing the inverter to do a relay test. The RCD did not trip. 2. I plugged in the garage door opener, did the same thing, and the RCD tripped. Obviously it did not do it on its own, but in collaboration with all the other leakers that's already in the house. The gate opener was on the same circuit. I recall the leakage was not insignificant, but I never measured it separately.

*snigger* No such thing in AC, current changes direction 100 times a second. But yeah... I suppose I'm being pedantic

If Someone actually built a transformerless battery inverter, it would mean there is an actual path from the battery to the grid. But as I said earlier, I have not seen such a beast yet. You are indeed right that in most cases, it means the DC PV input is grid connected. I believe the Fronius on my test bench is the same.

The transient when bonding/unbonding TN is much higher. At least, there is a reason I have that one in mind. I had a garage door opener that had three of them in, and it seemed to add just enough to nuisance trip the RCD. So I may not be able to create enough of a standing leak, at a couple of pF or so, but I think it has potential... The garage door opener since broke... and I kept the bits and pieces. Also, I now have a decent storage scope... so maybe I can just do that instead.

I still want to do an experiment with that and perhaps turn it into a video. I want to add 14D431K MOVs until the RCD trips. I also want to test and demonstrate what happens when the TN bond is temporarily disconnected and reconnected, as is done by inverters when they do their relay test.