diy solar

diy solar

Backup down under

Today's project was setting up a second instance of Solar Assistant dedicated to monitoring our server rack batteries. We have 1 x Jakiper and 2 x generic 100 Ah 51.2 V server rack units. They all use the same PACE BMS, or at least close enough.

I had a spare Raspberry Pi 4 after I had migrated my Home Assistant onto a mini PC, so thought I'd give it a go.

A little trial and error on using the right battery connections and dip switch settings but have it working now:

Screen Shot 2024-04-12 at 10.51.03 am.png

I already had a Jakiper supplied cable for connecting to the Raspberry Pi USB, so that removed all the hassle of trying to figure that part out.

Here are the two dashboards side by side:

Screen Shot 2024-04-12 at 10.38.39 am.png

Why a second instance?

Well my system battery is a hybrid LiFePO₄ (3 x 100Ah server rack units ~15 kWh) + a bank of sealed lead acid backup batteries (~20 kWh).

I have a Victron Smart Shunt monitoring the whole hybrid battery and that is the source of truth for the whole system.

Unfortunately Solar Assistant does not (at least not yet) enable monitoring of two different types of battery data inputs. So I could have chosen to just monitor the server rack battery, or the Victron shunt, but not both.

This way I can now see both and it will be good to have the server rack battery data constantly monitored and recorded.

If you look at the snapshot above, can see at that moment (it updates every second) there was 2942 W going into the whole battery according to the Victron shunt while the combined battery BMS reports charging at 2895 W. While there may be some accuracy differences between them, that is about expected as the balance would be residual current going into the sealed lead acid.

Also note the SOC differences:
Whole battery showing 84%
LiFePO₄ showing 57%

Since the whole battery SOC is based on 35 kWh, 84% means there is 16% of 35 kWh = 5.6 kWh to fully charged.
LiFePO₄ at 57% means 43% of 15 kWh = 6.45 kWh to fully charged.

Obviously there is a discrepancy there but that is probably down to a few things, once of them being the PACE BMS SOC is probably wrong and needs a recalibration.

I will now have a much better view on the state of charge of the LiFePO₄ during regular daily cycling and see how the energy supply is shared between the LiFePO₄ and SLA at difference states of charge.

Next up is setting up the MQTT link with Home Assistant.
And after a bit of mucking about I managed to get the MQTT connection working with Home Assistant. Mine is not straightforward as I use an MQTT bridge in Home Assistant (which enables multiple MQTT sources).

Anyway, it's working:

Screen Shot 2024-04-12 at 12.44.36 pm.png
Screen Shot 2024-04-12 at 12.45.29 pm.png

So now the fun starts with dashboards, some relevant automations, creation of some custom template sensors (e.g. to track cell voltage ∆) and charts to monitor the total pack vs LiFePO₄ charge and discharge behaviour.

diy solar

diy solar