JohanB

You MUST do them in series, if you do in parallel then you will reach your current limit. So the short, long run, more cable combiner boxes argument is mute.

Why do you want to split the string? If you want to expand later, just adding a second string will be easier than adding new panels to 2 strings.
The startup voltage, 150V, is when there is no battery or Escom power input so you will need a minimum of 150V PV for the inverter to start working, in fact mine is rated as 150V as well but in a footnote they mention that without battery or escom then 200V is required from PV. (Goodwe)...Not sure how, why this is so.??????
What I have noticed on my setup is that with 7 Panels, I usually have a lower voltage and higher current than the 8 panels..
But even at 90V, at dusk, 0,1A I'm still getting power produced by the MPPT's on both strings, but I do have batteries and escom..Never tried without batteries.
 
 

I remember being at varsity in the last few months of my studies and just then Microchip released a C compiler/editor for the PIC. That was super exciting! Now I can't remember assembly code much. 

In my minds eye, I see an interrupt handler that services your serial port (going to the BMS), and that simply stuffs the decoded values into memory locations.
The main loop then simply writes those memory locations to the relevant CAN frames, probably with some bit shuffling (ie, a lot of bitwise and-ing to get the half you want and shifting it left or right). You'll probably keep some kind of a wrapping counter variable and write some of the less important frames only when the counter variable reaches multiples of some time value.
Finally, remember to set the mask and filter registers of the can controller (it has two) to zeroes (and something) so it ignores anything incoming from the can bus. I find that if you don't properly handle incoming messages from the CAN-bus, the sketch tends to hang up after some seconds. I have not looked into this in more depth, I assumed it happens because the MCP2515 has only two buffers for received messages and you don't want to overflow those.
I would also play with the watchdog on the chip to make sure it restarts if it hangs up, you know, cause this is a BMS and all 🙂
I also discovered that at 500kbaud, you should try to keep any interrupt handling as short as possible so you don't miss interrupts from the CAN controller. I'm not fully clued up on the ATMEGA328, but if it is anything like most platforms, it masks interrupts while handling one.
Just some ideas off the top of my head.

Hi everyone, I made an easy install script to get an inverter monitor going with Home Assistant. Hope it helps someone.
https://github.com/BionicWeb/AxpertPi

Feature wise the Sunsync has the same options as the victron. If you ever need to expand they can also be connected in parallel. Ease of setup like the Goodwe, 2 strings in series Voc 5000, A=11, 6500 Wp.  Of panels.
Talks natively to pylontechs and can be configured, like the victron for DIY batteries.

Great explanation makes a whole lot more sense now... So the real answer, as with so many things, is it depends. At this stage I think implementing a compatible can-bus interface seems like a better option...
By reading here and doing research elsewhere I understand it as you refer to it as a “true hybrid” but it is so confusing as sellers of the equipment and by extension the manufacturers, refer to it as Hybrids. I found a neat trick is to use the City Of Cape town’s list of approved inverters as they list only true hybrids. and grid-tied.

Well... let me expand on that. If you want to, but I doubt this will be useful to you, you can listen for 0x305. That is sent by a Victron GX control device and is a keepalive. A battery can use that as a signal that the rest of the system is alive, and shut itself down if not. You can also listen for 0x307, which has the ascii charcters VIC in bytes 4-6, that identifies that your battery is in a Victron system. But again, I really doubt either of these features are usable to you. So you don't have to bother with them.
The CAN-receiver chip will however still receive the messages and you have to either read the messages and discard them (the lazy method which is what my sketch above does), or... you must set the MCP2515 masks so that all traffic is ignored.

Victron has an extended protocol that can send information about the minimum and maximum cell voltages and temperatures (but not individual cells... since there could be arbitrarily many cells in a battery depending on the series/parallel make-up). The battery doesn't have to listen to any info on the can-bus, it merely needs to send the frames every couple of seconds. Usually you'll send voltage and current frames relatively often (once a second or once every two seconds), and other frames (charge voltage and current, etc) a bit less often, once every ten seconds usually suffices.
The minimum you must implement is 0x351, 0x355, 0x356 and 0x35A. Adding 0x35E and 0x35F is easy, so you must as well throw those two in as well.
What they mean.... look here from page 10 onwards. Protocol is the same. Get the basics going first before messing with the deeper stuff.