Idiot's guide to a self build battery

[Joeboy]

A marked improvement, SOC in pylontechs hit 98% at 02.30hrs and has fallen back to 92% by 08.00hrs with the goodwe software showing a discharge to home of very little, 200W down to 32W. LF280's are reading 3.33V or 80% SOC.

Looks like a decent day too. Might see an increase if SWMBO can control the washing machine frenzy.

[Lincs Robert]

That’s very impressive.

Would it be possible to have just a single Pylon working with the self build battery? I’m assuming so. My logic being that I could keep my single, new, 2000C and sell the existing 4 x 2000s to finance the new battery and end up with a much bigger overall capacity.

I’ll have to have a detailed trawl through the 30+ pages of post to extract further detail on how it all fits together. I’m intrigued as to how BMS functionality works on such an arrangement. I’m assuming that the Pylon looks after itself with the BMS connection to the charger/inverter, and the new stack sits in parallel across the Pylon and looks after itself with the cell balancing module? Hence the “issue” with SoC reporting. But how meaningful is the SoC reported by the Pylon to the inverter? Am trying to get my head round that.

Good job, well done

[Stinsy]

That’s very impressive.

Would it be possible to have just a single Pylon working with the self build battery? I’m assuming so. My logic being that I could keep my single, new, 2000C and sell the existing 4 x 2000s to finance the new battery and end up with a much bigger overall capacity.

The problem is that the Pylontech batteries tell the inverter how fast to charge/discharge. A single US2000 would limit the charge/discharge rate to 25A (1200W) because it wouldn’t know about the additional homebrew battery.

[Lincs Robert]

That’s very impressive.

Would it be possible to have just a single Pylon working with the self build battery? I’m assuming so. My logic being that I could keep my single, new, 2000C and sell the existing 4 x 2000s to finance the new battery and end up with a much bigger overall capacity.

The problem is that the Pylontech batteries tell the inverter how fast to charge/discharge. A single US2000 would limit the charge/discharge rate to 25A (1200W) because it wouldn’t know about the additional homebrew battery.

Gotcha, so I’d really need 3 Pylons to get the 3600W charge/discharge rate. That would free up 2 for sale - which would still realise pretty much most of the funds needed methinks …..

[Caesium]

But how meaningful is the SoC reported by the Pylon to the inverter? Am trying to get my head round that.

It is actually still useful, because the net result of adding more capacity in parallel just means the Pylontechs have less work to do, ie discharge slower, so the SOC drops more slowly than it did before the extra capacity was added.

It's a reasonable assumption to assume that the SOC of the Pylontech and the SOC of the EVE pack is fairly close most of the time, because they'll self-balance if the voltages drift too much (although very slowly, probably).

[Lincs Robert]

But how meaningful is the SoC reported by the Pylon to the inverter? Am trying to get my head round that.

It is actually still useful, because the net result of adding more capacity in parallel just means the Pylontechs have less work to do, ie discharge slower, so the SOC drops more slowly than it did before the extra capacity was added.

It's a reasonable assumption to assume that the SOC of the Pylontech and the SOC of the EVE pack is fairly close most of the time, because they'll self-balance if the voltages drift too much (although very slowly, probably).

Thanks - that makes complete sense & the conclusion that I’d pretty much come to whilst pondering after having asked.

[Caesium]

Gotcha, so I’d really need 3 Pylons to get the 3600W charge/discharge rate. That would free up 2 for sale - which would still realise pretty much most of the funds needed methinks …..

Ideally yes (this is my current situation and it works just fine) however there are ways around it.

You can either tell the inverter the batteres are lead-acid, which makes it rather more dumb and just charges up to a voltage you set; you do lose accurate SOC and some protections like cell under/over voltage shutdown, and temperature related shutdown etc, so this can be a bit risky if you don't have other protections in place.

An alternative is to pretend that your new DIY stack is a Pylontech battery. This can be done with off-the-shelf BMSes and a sort of "gateway" to convert their communications to Pylontech. For example, https://github.com/Uksa007/esphome-jk-bms-can will take RS485 from a JK BMS (£100 on the usual outlets) and spit out Pylontech protocol for your inverter, by slotting an ESP32 in the middle basically.

I'm currently doing the former (lead-acid) which works well enough for my purposes at the moment. I might explore the Pylontech-pretense route in the future as I do have JK BMSes to play with too.

[Tinbum]

It's pretty easy to intercept the CAN messages from any BMS and change them to what ever you'd like.

[Fintray]

That's a tidy installation Joe, be interesting to see how the new batteries perform alongside the Pylontechs.

[Oldgreybeard]

FWIW, I don't think that the SoC reported by the Pylontech BMS is particularly accurate. I know for sure that it's way out for the last ten percent or so from about 89% to 100%. Having spotted some odd behaviour, where the SoC would "stick" at around 89%, then suddenly jump up to maybe 96% to 98%, I monitored what was going on a couple of times. Both times the SoC was under-reading when it was reporting 89%, and even though three or four percent of charge went in the reported SoC didn't change at all.

If I had to guess, then I'd say that the Pylontech reported SoC is probably no better than about 5% accurate. That fits with my limited experience of trying to calculate SoC for home made packs, the closest I got, using a bit of kit that measured true current in and true current out, together with battery voltage, was about 5%. No matter what I tried I couldn't get a more accurate result.

Being within 5% is good enough in practice, anyway, I think.