EG4 batteries drained to zero overnight

[jjbillings]

Hey all! I have an off grid system in my shop with 2xEG4 6000XP inverters and 6x48v 100ah eg4 ll-s batteries. I don't have solar connected yet, so I charge with a generator and chargeverter periodically (~once per week). Because the inverters draw a fair bit at idle, I turn the EPS Output switch OFF when I leave the shop for the night, which has served to limit the overnight discharge. Usually 1% overnight. Last night, I did the same thing I always do, turned the EPS off and left. SOC of the bank was ~33%. Came back this afternoon and all the batteries were at 0%! . I double checked the data in the EG4 monitor webapp and it corroborates my own memory that the SOC was ~33% when I turned off the output. It also shows that the battery voltage was down around 47.5v at that point. So, I have two questions:
1) Did the batteries go from 33% to 0% so fast because the tail on the voltage curve drops off in the lower SOC percentiles? Or some other reason? Even with the inverters EPS off, they still draw a little bit (0.5-1amp).
2) I have the EOD cutoff on the inverters set to 15%. I _thought_ that this meant the inverters would shut off and stop drawing power at that point. However, given that the bank discharged to 0%, that doesn't seem to be the case. Am I misunderstanding that cutoff? Or is it a situation where EPS would need to be ON for the EOD cutoff to work?

 

[aussiesam]

six batteries all drained from 33 to 0?
First off are you using voltage to determine SOC? or are the batteries communicating with the inverter closed loop?
you're going to have to show the data, because that's not making any sense.
Share screenshots of battery voltage chart and current state data

 

[42OhmsPA]

SOC was ~33% when I turned off the output. It also shows that the battery voltage was down around 47.5v at that point.

SOC is way off. They are darn near empty at that voltage.
I'd charge them up to 56V to reset SOC to 100%.
If you aren't charging them to 100% frequently SOC will go way out of whack. You'll also have issues with SOC with low loads, like the inverter idling, due to the BMS doing a shitty job at tracking low current.

 

[jjbillings]

The batteries are connected to the inverter communication port. Im using the state of charge numbers on the batteries/inverter, which have always matched. Im attaching a couple of images with the data from yesterday, which ends when I shut off the EPS output OFF (and thus, the wifi).
@42OhmsPA I'll give that a shot. I charged them up to 52.9V (BMS reads 43%) before I had to leave for the day. I'll charge them up to 56v tomorrow. I appreciate your insight on this

 

[42OhmsPA]

Notice how flat the voltage curve is.

52.9V could be a wide spread for SoC.

 

[42OhmsPA]

You could add a shunt to the system for accurate SOC tracking.

 

[jjbillings]

Noted about the flat voltage curve.
I'll look into adding a shunt to the system. I'd very much like to have an accurate picture of the SOC, and it's a bummer that it's not working out of the box with this equipment. If a shunt will get me that, it seems worth some additional cost.
It sounds like my immediate next steps are to charge back up to 56v to reset the SOC and switch the inverter cutoff from SOC to voltage params so it doesn't drain the batteries to zero again.

 

[aussiesam]

47.3V is not 33%, that's your issue right there. For storage you should leave them above 50%, while the exact voltage is hard to determine with lifepo4, 50% should be above 52V.



This chart is not 100% accurate but it gives you a bit of an indication of the the actual charge especially at the upper and lower ends

 

[42OhmsPA]

Almost every BMS is inaccurate.
That's why I've been happily running open loop for years and trusting my shunt.

 

[Skypower]

Even the best shunt will drift SOC with numerous short cycles that don’t reach full charge reset. I’d suggest you reach 56.2 bulk voltage more often. The battery also needs adequate time above balance voltage to keep the cells in line. If the balance gets out of hand you will lose capacity due to cell(s) triggering high and/or low voltage protections sooner than normal. You can’t allow a battery with low balance current (150ma) to have long periods of short cycles, not reaching full charge PLUS absorption/balance time without a potential delta problem and the consequential steps needed to correct it.

 

[rogerheflin]

make sure you have the latest firmware. And be aware that most BMSes count charge and you typically charge up at typically much higher currents than you discharge, and at lower currents (discharge) the currents tend to be under counted.

So charge up at say 30A rate and add 50% to the battery and discharge(at say 5A) and remove 50% (but the current counter only sees 45%) so it becomes 5% too high. Repeat a few times and 33% SOC is actually 0%.

A few people pushed JK BMS to fix their issues on their BMS, and now they don't report 100% until the voltage spikes correctly to be real 100%, before that they reported 100% when charge counting said the battery SHOULD be full (and the inverter stopped charging) and slipped a few % each cycle. I believe most of the other BMS'es still have this defect (report 100% when it SHOULD be full without verifying it is full via voltage).

On the bms I have changing internal parameters would reset SOC to 70% (no matter the voltage) so when I saw SOC slipping (before the fw upgrade) I would let the battery hit near 100%, and then change something to reset SOC to 70% and let it charge another 20-30% to actually be 100%. It was a major pain.

 

[houseofancients]

Almost every BMS is inaccurate.
That's why I've been happily running open loop for years and trusting my shunt.

True unless the BMS has an actual shunt build in..
Mine does

 

[rogerheflin]

True unless the BMS has an actual shunt build in..
Mine does

Just about all of the BMSes have some sort of actual shunts built it. The issue is with the full range calibration and number of bits of accuracy of the built-in shunt. If you only use 10bits and full range is +350A to -350A then the lowest current you can measure is .7A, use 12 bits and you have .17A accuracy (assuming everything about the shunt is perfectly calibrated--probably a bad assumption).

I wonder if one built a 2-range shunt (ie 2 measurement ranges built in, one -350A to 350A, and one say -20A to 20A, or similar) and you always used the -20 to 20 one when it was in that range to improve low current accuracy.

 

[Skypower]

What I’m seeing is an issue with directional calorie counting in bms’s. Their petit sampling shunt isn’t very good at counting calories when the current is different. Most of use don’t drain the battery at the same rate as we charge it and that’s where I believe the problem lies. You can try lying to them, for instance, setting a different value from what a real shunt is actually indicating. This can help to a certain extent until the season changes so I split the difference. One tweak on one JK doesn’t necessarily work on the other, manufacturing and component variances throws a wrench into that equation. My JK’s aren’t terrible but after a few weeks of not hitting full charge they’ll easily be 10% off from each other with the lowest being closest to reality.
An old episode of Off Grid Garage had Andy working on his three battery shelf and he was surprised to see that two of the JK’s had a correct SOC and the Victron shunt was off.
I have a collection of different brand shunts and Hall effect battery monitors. The closest meter I have now is a Junctek and I didn’t have the adjust this particular one at all. Aside from the glitches which were corrected by the alternative power diagram, it’s spot on even after a week and a half without full charge. Would I recommend this to others? No, I think I got lucky with this one. It’s at your own risk with these things, do you feel lucky?
Victron is probably the gold standard for shunts and for connectivity.

 

[42OhmsPA]

I'm using the Juntek 600A (WonVon clone). It's been very accurate and mostly reliable, I had to reset it twice in a few years. I have a cheap tablet upstairs dedicated to the bluetooth display.