Shunt Equalizer Test.

[waitatian]

Shunt Equalizer Test.
There is a problem with charging series connected batteries. They may start out the same, but they can age differently,and a new replacement battery added to a string is not the same as the rest.
When series connected batteries are charged, the regulator will eventually fix the system voltage, and each battery should have an equal share. If the voltage on some of the batteries goes above their share, the voltage on the rest of them must go down because the total voltage is fixed. Equalizing charges are meant to keep them the same but if the batteries are not equal, no amount of overcharging will make them so.
My house supply is a 48v string of mixed age 6V US2205. When there was plenty of charging power, two of the batteries can be getting overcharged at 7.8v at the same time as two others are being undercharged at 6.8v. To fix this it was either a new set of batteries or some form of active equalizing. I tried a shunt equalzser.
It works by comparing the voltage on two adjacent batteries. A resistor bleeds current off the higher voltage one.
I initially tested it on my 24v shed supply. For the test, I set it up with two US2205 about 1 year old and two about 5 years old. Charging was by four 12V, 40W panels in series through a cheap Chinese MPPT regulator.
  After a month of settling in time, during charge the voltage on one battery was high, correct on another and low on two. So one was overcharged while two were undercharged. The equalizer was connected on a sunny day. One battery was immediately bypassed by about 3A - the maximum the equalizer will do. Over about an hour, as the three remaining batteries got more charge, a second battery started to be bypassed. It took about a week for three batteries to be bypassed, meaning most of the charging current was going into one battery. Over about a month, the bypass current went down as they were more equally charged. Now on float, only about 100mA is being bypassed across 2 batteries, and all the batteries are within 10mV of each other during charge/float.
 The house supply was then set up and a similar equalizing pattern happened, except at the end 6 batteries were being fully bypassed. This means the two un-bypassed batteries were needing a few amps of extra charging current at all times to keep their voltage the same as the rest.
 Load testing the two batteries showed reduced capacity, but they could still provide power. The equalizer allowed them to get charged without ruining the rest of the batteries, and they could stay in service until I could afford to replace them.
The two low batteries remained in service for the summer with no problems, and were replaced with new ones. Due to poor weather, it took about 3 weeks for the new batteries to settle in. Now, only the batteries on each end of the supply have any serious bypassing and a few others are being shunted by a few hundred ma.
 There were no problems with the Chinese MPPT controller or the Tristar PWM regulating the house supply.
Shunt regulators waste power when bypassing so in poor charging weather some of the charging power will be wasted.
This may stop the batteries being fully charged, but they will be equally not fully charged.

[waitatian]

Some pictures.

[SparWeb]

Thank you waitatian for sharing this!
I haven's seen a shunt equalizer (or anything like a BMS) on Fieldlines for a long time.
I think it was another kiwi who posted it, then, too.

I have a similar problem with some of the batteries in my series string - they are always a few % weaker than the rest.  It's not enough to kill the stack, but every year the split between the best and the worst gets wider.

My wind diversion load is going off frequently, so wasting some energy to this form of balancing would not be a concern to me.

Is that a PIC on each board controlling the equalizing current?

[waitatian]

Hi,
It is an op-amp.Doing it with microcontrollers would make it more efficient and probably cheaper, but I needed a quick solution. I could dig up the circuit and post it. It's an op-amp in differential mode driving a darlington transistor/resistor on each battery. The equalizers have been going over a year with no problems.

[SparWeb]

I'd love to see the schematic!

[DamonHD]

Yes, me also.

Rgds

Damon

[waitatian]

It works by amplifying any difference between the battery center connection and 1/2 the volts of the two batteries.
I tried a few different input configurations for the  op-amp. The damped comparator? was the first one that worked. The op-amp was chosen because of its shutdown terminal. The 500 ohm trimpot sets the voltage that it will start equalizing.
The 2 ohm dump resistors add a bit to the price, which a microcontroller system could get rid of. Six 5W resistors were cheaper than 10W or more resistors, and cutting down maximum current would lessen the resistor price.
The BDX34/35 transistors are darlingtons. The other transistors are whatever is laying around, but the two by the 500 ohm pot should be the same.
The led board will show if there is a big fault like a shorted output transistor, but it won't show a bit of leakage.

[SparWeb]

Thank you!

For once I look at a schematic and realize "I have most of that stuff".
Except of course for the TLV op amp, this is done with very pedestrian parts.
The TLV isn't a stocked item at Digikey, but they do have a few left - I'd better order soon!

[waitatian]

I don't like circuits with unobtainable or expensive parts either.
I got most of the parts from Digikey. They didn't show the op-amp as endangered when I got them. It was chosen because it was the cheapest one with a shutdown pin.
If it is discontinued, the output power of the replacement would need checking and the value of the 560 ohm output resistor increased if needed.
If you build one, it is advisable to have covers over the ends of the leads when installing it in case one of them ends up where it shouldn't.

[SparWeb]

My battery stack actually has a strong cell beside each of the bad ones.  Which gets me thinking...

Would it make sense to you if I made only 1 shunt equalizer, and put it on only 2 adjacent batteries?
I see you have put your equalizer across all of the cells in your battery stack, so do you have any warnings about putting an equalizer on just 2?

[waitatian]

It would make the two it is connected to the same but the combination might not match the others. A simple test is to mimic the equalizer by putting a variable resistor across the high battery and keep its voltage equal to the low one while checking the volts on the others. That should show if only one will work.Of course after a while the equalizing current should go down as the low battery gets a full charge.(Doing that was what I based my design on)
Watching my 48V equalizer working, it appears that different rates of charge will make a different order of batteries that are bypassed, but when the regulator fixes the voltage the same ones will eventually be bypassed.

[SparWeb]

The others are rather "average".  There just happens to be a coincidence of a star right beside the slug, in both cases.  All the rest are middle ground, which is very surprising given the age of the entire pack of them.

I could have cell #6 steal a bit from #5 and that would even out all the cells in string 1 to within a few %.
In string 2, having cell #8 steal from cell #7 would also help, even if it might not be enough.

Here's the last reading:

Cell #   Cell V
1   2.287
2   2.283
3   2.279
4   2.278
5   2.285 (above average)
6   2.276 (2nd lowest)
----------------------
7   2.291 (3rd strongest)
8   2.254 (lowest)
9   2.282
10   2.277
11   2.281
12   2.288
----------------------
13   2.285
14   2.278
15   2.275
16   2.273
17   2.292
18   2.300
----------------------
19   2.291
20   2.279
21   2.289
22   2.281
23   2.278
24   2.280
----------------------
Average   2.282

[waitatian]

It would appear no.7 would be limiting the total charge and no.8 is suffering, or, no.8 is pulling more charge, making no.7 get overcharged.
I haven't tested it, but a zener diode and resistor in series across any high cell might work to bleed off excess charge without using any power when the batteries are being used. It should be possible to work out where on the charge curve to select the zener voltage and the resistor value could be periodically changed if the low cell charges up. If that wasn't enough, adding a single zener/resistor from cell 9 to cell 12 would add more charge to no.8

[SparWeb]

Hah.  Well this gets simpler day by day. 
You mean a zener with a cut-off about 2.20 volts?
We are working with currents much less than an Amp, too, aren't we?
I have some 10W/10ohm resistors that might be just right.

Float V is between 27.0 and 27.5 therefore 2.25 to 2.29
Resting V is between 25.0 and 25.8 therefore 2.08 to 2.15
The right zener voltage seems to be 2.20 V.

The whole bank can be kept at float with <2Amps.  With 2 strings then each string needs about 1 Amp to maintain float.  Obviously bulk charge is higher but variable, up to the capability of my combined sources, which is > 20A.

If I bypass one cell with a resistor, I'd start at about 25% current and see what happens first.  For a cell at 2.25V to bypass 250mA then:
2.25V / 250 mA = 9.0 Ohm

Which means I can use one of my 10-ohm resistors in series for my starting point with a zener to shut off the bypass overnight.


Another point of interest:  only as I was writing yesterday, did I start to wonder if there's some inter-play between cells 7 and 8.  You suggest a possibility, too.  Perhaps there is some detriment to cell 8 having it in series with a strong cell like #7.  The cells are mounted in a rigid steel casing, and have swelled a bit, so I can't pull them out any more.

[waitatian]

You might need more bypass power initially to bring up 8 while stopping 7 overcharging.
The most bypassing current will be at full charge.
If this is say 2.55v, minus the zener volts of 2.2 = 0.35V across the resistor.
W=V*A, so a 1W zener at 2.2v will pass 1/2.2 = 450mA
R=V/I so to drop .35V at .45A = about .78 ohms
Power = I*I*R so it would need .45*.45*.78 = .16W resistor.
At float, the voltage across the resistor would be about .07V.
I=V/R so there would be  0.07/.78 about 90mA bypassed at float.
These calculations imply a sharp zener cutoff voltage.I don't know what zeners are available but they can be paralleled for more wattage and the .6 volt drop of a power diode in series can adjust the voltage.
The conditions at different places in the string seem to vary. It probable has to do with inductance or capacitance effects. I used to rotate my batteries before the equalizer was installed. If the intercell connectors are removable is it possible to rewire 7 or 8 to a different position?

[SparWeb]

If the intercell connectors are removable is it possible to rewire 7 or 8 to a different position?

Photo from a previous installation, but the batteries are stacked in the same order.



There's a thought... I crimp two jumpers, remove the bars from 5-6 and 7-8 and use the jumpers to connect from 5 to 7 and from 6 to 8.

[waitatian]

Or balance leads from 6-7 and 5-8?
It would be interesting to know the results if you try a zener/resistor shunt.

[SparWeb]

Just for a first try, I did a temporary setup today.  I clamped a single resistor across cell #7 with the tap on the resistor set to about 1 ohm.  A fairly constant 2A was bypassed as a result.  The resistor was on for about 6 hours.  I didn't include the zener because I planned to remove everything.

Very windy all day, and still breezy even now.
Turbine was producing about 10A with peaks 20A when I started.  When finished wind was giving me 5A with peaks 10A.

When I started, string voltage was about 27.2V, and the diversion load was activating fairly frequently when the peaks would push it up to 27.6V.  Inside its insulated case the batteries were warmed up to over 20C (despite it being below zero C outside) so I could categorically say the stack is "full".

The cells 7 through 12 started with readings about 2.28, 2.25, 2.27, 2.27, 2.27, 2.26.  This varies all over the place as the wind fluctuates.
With a load clipped directly across the terminals of cell 7, its voltage dropped immediately to 2.26 and when I came back later it was down a bit more.  Equal to cell 8 in fact.  I removed the resistor and it immediately began to climb back up.  A few minutes later its usual situation was restored.

To have diverted 2A for 6 hours is only 12 amp-hours.  Taken from a cell that has 400 amp-hour nameplate rating this is just a drop in the bucket I realize.  Also at the same time there was at least 3A still passing through cell 7, so in its full state of charge it was still just producing heat.  To have any long-term effect this will have to be installed for a significant period of time.

It's been an opportunity to witness the effect and measure it.  With a zener I could have left it on overnight. 
I'm really tempted to try switching the cells with the jumpers, though.  I've got them ready to try.

[SparWeb]

Of course I took a picture.  Soon after clipping the resistor to cell #7.

[waitatian]

It looks like 8 has got a bit more charge. If it does pick up condition the bypass current needed to match the two should go down.