I removed the shunts from my systems over 12 months ago and have three different Multiplus units: a 48V, 24V, and 12V, all using the battery monitor for SOC estimation. Each unit is paired with MPPT and internal AC charging from excess grid-tied solar. The Multiplus is extremely accurate at estimating the SOC when you’re loads are AC only, and I run mine down to 0% until the low voltage cutoff activates.
It’s important to stick to one method for estimating SOC whether it’s from the battery monitor, shunt, or BMS and understand its limits.
Additionally, with the MPPT, there’s an option to use MPPT current for better solar performance accuracy, which can help optimize system efficiency.
Both graphs are the same system and the same time period, but the graph readout box is for different times within the graph as I can only display one point per graph. As the legend says it shows system and smartshunt values for SOC and charging current. It’s a screenshot taken from an advanced display I setup on the victron VRM. The colours are as marked in the readout boxes.
Good call on Peukert exponent, whilst I did set this for Li, I have not checked the match with the Quattro. The smartshunt really needs to be more configurable as its really set up fpr LAB. Anyway it was set at 1.05 which I think is close enough not to cause the glitch.
As an engineer and designer things that do not match within either spec or useable range are a FAIL.
Saying just use one system because our design is so poor none of our things gives the same answer (and thus all are wrong) is a total fail. They are different because:
They have been set up incorrectly: solution give clear setup instructions - easy.
There is a flaw in the hardware/software: solution fix software to mitigate problem - quite easy.
When you are on Octopus and setting recharge time and levels its important to know the SOC to within a few percent (-+5 at least) which with safety margins etc will keep me off grid maximally without running batteries low or suffering astronomic peak electricity charges.
Also Victron is expensive and supposed to be Rolls Royce quality, not a clunky rustbucket.
I am not sure exactly how your stuff is connected or which Battery Monitors you are using.
My system is as much off-grid as possible. As a consequence I can run a full week between enforced full absorptions when the SOC cycles between say 50% and 90%. Any significant error on the SOC% measurement can and does push me into actual SOC% falling well below my safety level of 50% to the point where the quattro cuts the system off due low voltage.
Doing this should not be a problem for decent equipment, and basically Victron stuff is pretty good. I think the smartshunt is probably giving me very accurate measurements (perhaps I should set the Peukert to 1.00 as charge is reset on full charge) but the quattro is definitely off, which has taken extra batteries and eventually a smartshunt to see why the system was not behaving as it reported it was doing. There is a provable difference between system and smartshunt, fixable in software IMHO.
Look the Cerbo (I presume) manages to total quattro charge/discharge and solar charge remarkably accurately (system battery current) without the help of the smartshunt. Sadly, it’s not accurate enough and results in an overestimation of the SOC% to an unusable degree (you will remember earlier posts).
I have (as you see) installed a smartsunt and the error (assuming the smartshunt is accurate as I believe it to be) is actually rather small, but is a near-constant offset - not random at all.
However the quattro is measuring/estimating the DC current its doing a very good job indeed EXCEPT for a near constant offset of about 2A (rising to ~3A on higher powers) which is suspiciously close to the self-consumption power of the quattro. Prima facie: that’s what it is.
The SOLUTION IS SIMPLE: Change the “System” algorithm to allow for the Quattro self consumption (I would include a power term too) and correct it.
QED,
No, the Smartshunt measures the consumption of the quattro.
If the Smartshunt SoC is off ( compared to what other information ??) then you must tweak charge efficiency and maybe even peukert, and make sure that you really synchronise with a full charge as often as possible, ofc. with correct settings for charged voltage and tail current.
Dear Ludo,
as before you don’t understand what I am saying, cannot read the graphs, and are unable to offer more than standard replies and do not actually give the figures you would use for the purpose. I presume because, unlike other posters, you don’t actually know what they should be. Continued discussion at that level is pointless.
You can (to some degree) compensate that ‘offset’ that’s annoying you by tweaking Charge Efficiency. There may still be a bit of an error in SOC at the lower end, but it won’t be as much as you think.
I still reckon your plots are mixed up, and you should check that. SOC isn’t included in DVCC, where you can select V, I and T to your preference. SOC is forced based on the status of the measuring device, so your ‘System’ value is most likely coming from your shunt. I guess that’s a deliberate step, so decisions based on SOC are coming from the best source. Protecting the end-user from himself, hey.
The GX doesn’t calculate SOC. Just passes it on.
I use Pylontech batteries, whose SOC can likely be out by as much as 6-7 % at times. But I can tell when that happens and can live with it. I can’t use the Multiplus because of DC loads. I moved my Smartshunt to another system.
When it was used for Pb batteries I had SOC tuned to within 0.2% at sync over a daily cycle and maybe 0.5% over a week. Pretty good.
And when VRM Beta team removed the decimal point I wailed. May not be really needed for observing, but invaluable for tuning.
Hi,
charge efficiency will not work because its essentially a constant offset in the charge current, so it’s time variable, not charge variable. Any errors of a different kind will accumulate, just in a different way.
My graphs are taken from my Victron VRM portal advanced tab and other than the text in yellow are generated by the portal so are unlikely to be wrong.
As far as I can tell, after playing with the available data, the smartshunt (battery monitor 277) in series with the batteries is the reading from the shunt SOC% ONLY. “System” doesn’t seem to change if I remove/disconnect the shunt. I believe its generated by algorithms in the cerbo that estimates (measures) battery current as the difference between the quattro and solar DC currents, can’t be anything else, really.
As you can see from the graph, the “system” battery current (violet) is extraordinarily accurate compared to the BM current (green) except for a near-constant offset. Unfortunately that offset is significant, amounting to around 2.5 kWh/day. Over four days that’s 10kWh, which is hopeless.
Its the “system” algorithm that needs a tweak to be near perfect.
Possibly a fix with software, maybe AI to learn your battery. I can go weeks without a full charge like last month went down to 0%, and a few 5%. The battery monitor is enabled in the ve config when you set the charger settings. I am on grid with ESS. I think most people don’t even test their battery as well to check what is actually there and usable. A bad cell will give you low and high battery voltages that do not reflect 0% and 100% charge.
What are you using for a battery monitor if you have removed all the shunts?
OR are you saying you just have solar/battery/multi at 12 and 24 and 48V gridlinked independently?
I am a little puzzled by your graphs. I can only assume you are not using Victron smart Li, otherwise that’s exactly what I was expecting to achieve, except for smart Li:
Victron demand full charge at least once a week in this application.
Generally seems to be considered bad on batteries to go below 20% SOC let alone to zero.
The smartshunt did do well in a basic accuracy check, so I believe it to be very accurate.
I have had problems where SOC seemed good but switching on a kettle caused a low battery voltage tripout si now I am more careful. I now attribute this to the misreported SOC when I was on system only and had no smartshunt so had a much lower SOC than was reported by the quattro.
12V & 24V in the shed
48V in the house
I use regular LiFePO4 batteries without BMS communication. My 12V system has been cycling between 100% and 0% (14.2V to 12V) for nearly 3 years—about 1000 cycles. I occasionally test one battery and haven’t seen any capacity loss. Since these batteries are rated for 6000 cycles, they’re holding up well.
System Management & Cutoffs
I designed my system so 0% SOC is rarely hit, and if it is, it’s at 7 AM, right before charging begins. Low voltage cutoffs are:
12V system → 11V cutoff
24V system → 22V cutoff
48V system → 44V cutoff
There’s minimal usable capacity below these voltages, and recovery from low voltage takes little charge. I believe low or high voltage is more damaging than SOC itself.
Grid & Internal Balancing
Victron’s sustain voltage option allows a small grid charge to maintain a safe voltage. Since I have three ESS, it’s unlikely they all hit 0%, meaning one can supply the others instead of drawing from the grid.
SOC Accuracy in Victron
Victron offers three different SOC estimation methods, it’s crucial to pick one that works for your setup. If your system isn’t reaching a high charge voltage or discharging low enough, you’re not utilizing your battery’s full capacity. You can only achieve 100% and 0% discharge based on voltage. The system measures the power coming from the battery and the kWh is what matters to me.
Hi,
my apologies for the delay in replying, stuff to do!
Yes, I had enabled battery monitor from the start, that seems to be reported under “system SOC”.
Anyway, I have now resumed using the smartshunt as the battery monitor as this looks to be the more precise of the available options. Unfortunately, whilst this will help with the battery, it has exposed so many other flaws, but another thread for that.
Yes, as others said, and its undoubtedly true although not well expressed in the literature, outside the undercharged and overcharged knees the battery voltage is pretty well random within a band. Luckily, though, leakage current for LiFePO4 batteries is lowt. This is generally quoted as 2-3%/month. That means a 100Ah battery might lose 3Ah/month or 0.1Ah/day or 4mAh/h ie 4mA/100Ah leakage (!!). For a daily or weekly recharge, that’s zero, not even 99%. So measuring SOC% is piss easy, just measure the charging and discharging currents and total. Easy peasy when installing a shunt (just a piece of PCB/copper of known resistance) and stuffing a 20bit (cheap as chips) A-D converted and sample and average periodically (say 1/sec). Cost to end user built in, maybe under $20. Because these are dirt cheap (compared to the item of equipment) it goes in everywhere a DC current is output. Current is conserved so all meters total to zero ± a few tens of mA (a hundred would even do) easy-peasy. I have absolutely no idea why a blue chip company like victron has used grossly inaccurate means to “measure” the most important thing on the system. The bit that allows safe and proper use of the most expensive bit: the batteries. Solar input: no great accuracy needed, Mains in/out no great accuracy needed (except zero setting) DC currents: absolutely needed.
