Battery balancing

[CamEllis]

Hi,
I have an 18kWh battery pack in an electric vehicle using a battery technology that allows us to fast charge in under 10 minutes. The configuration is 15 battery modules consisting of 6 battery units in series of 8 cells in parallel. During the rapid charge we are telling the charger to cut off when the max cell voltage MaxCV reaches 2.85V. 2.85 is the top end of the charge curve and represents a small percentage of the battery capacity due to the steepness of the charge curve at this point.

The problem we are facing is how to balance the cells so that we can get all of the cells to approach 2.85, therefore getting a greater capacity out of the battery. As this vehicle is a prototype, I want to do this simply and cheaply with out making an intelligent computer driven relay system.

A thought that I had was to place FET's over each cell so that as the charge approched the gate voltage the drain circuit of the FET would allow burnoff through a resistor thereby allowing the other cells to approach the 2.85V.

My problem is I know nothing about electronics and I hoped that somebody may have a suggestion on an alternative solution or verify that this one may work... and if so, tell me where I might find an FET to do the job. I have included a jpg of the proposed circuit showing only one of the 6 FET circuits proposed.

thanks in advance,
Cam

[pminmo]

Something doesn't add up. Round/rough numbers...If I understand your battery arrangement, fully charged is 256V? So at 18KWh, your battery bank can deliver 70A for an hour. If the discharge is 60%, then your restoring 10.8KWh worth of power in 10 minutes? Assuming 256V charge voltage, your charge rate would be 253A for 10 minutes if my thinking cap is screwed on right. That's a lot more than a simple FET and resistor can handle. What am I missing?

[NC Cams]

We went down that path YEARS ago when "matching" cells for R/C car racing. You are essentially looking for needles in haystacks when trying to "match" the cells BUT, when/if you do, you'll be AMAZEd at what you can do and power with them.

The problem becomes "which characteristic are you trying to match?":

Charge efficiency?
Charge capacity"
Discharge capacity"
Amp hours?
Cell Impedance?
Charge efficiency?
Discharge efficiency?
Cell voltage curves under discharge loads?
Plus other stuff I don't recall or want to discuss

Just what are you trying to "match"???

Once you thing you have some cells "matched", cycle them under the anticipated load and do it again. The damn things don't repeat - at least the nicads we were working with didn't - not until we "conditioned" them which is/waa a whole 'nother technology realm that opened up.

After spending MONTHS both testing and cycling cells, we eventually figured out how to make them do what we wanted/needed them to do - and it did NOT involve charging and/or discharging them in the "Factory" prescribed methods with "trickle chargers". We also "hurt" some cells trying and simply wore some out.

But when we got "good ones", you could damn near arc weld with them and you coudn't hardly hurt them... We usually sold the lamer ones off to other racers to recoup some of our expeneses. Even those "lame" cells were better than some of the "factory" matched cells we came across until we learned how to match our own

All the battery engineers we talked to (including some at NASA) thought we were crazy but it turned out that the R/C car racers taught the factory guys some interesting things about nicads - not the least of which was "how to rate and match them".

Ah the excesses of a misspent youth finding and making power.....

[CamEllis]

Hi again, thank you for yr replies.
The cells have an operating voltage of between 2.0V and 2.8V
The charge/discharge bus connects the batteries as follows:
1 battery
15 battery modules (with child bms) in series
6 cell units in series
8 cells in parrallel

8 x 11amps = 88amps
6 x 15 x 2.3 (Average Circuit Output Voltage) = 207
= 18216 watts


In the image above, I propose that the FET circuit is connected at each cell unit which will hopefully cut off the charge at 2.8V for all of the cells which are being charged in series. It is not feasible to independantly charge the cell units due to the complexity of the bus at th high amps we deliver. Instead I am hoping to drain the current before the highest unbalanced cell unit reaches 2.85, allowing the cell units to catch up. We have had fairly consistent results of unbalance over the last couple of months which suggest that the problem is a prototype error in balancing rather than a cell capacity fault.

If this circuit makes any sense, my question is: Will a feild effect transistor switch to a drain circuit when a nominal cell unit voltage of 2.8volts is reached... or should I use a variable relay? In either case are these items of the shelf or do I have to specify and order them?

I have included a typical charge graph to show the amps and voltage that we apply... the BMS (parent) monitors the cell unit voltage and instructs the charger to switch off when 2.85V has been reached by any cell unit.



The graph below shows the typical state of imbalance after a rapid charge


Our goal is to balance the cell voltage at high end state of charge.

[NC Cams]

I don't know what type of cell you're working with - we did our work with nicads - but the fundamentals remain the same.

You're essentially trying to match the capacity of each cell so that you get ALL the energy into and out of it evenly and efficiently. To do so, you have to somehow figure out cell capabilities (Ie: chemical potential) and "balance" a bunch of random cells so that they all work together.

Been there, done that.

We peak charged and discharged INDIVIDUALLY hundreds of 1.2v nicads and then INDIVIDUALLY discharged them while continuously recording discharge voltage under a constant discharge current. After doing so HUNDREDS of times, we "learned" (they taught) us what the cells wanted/needed in the way of charge cycles and cycling tendencies. NOt all of what we learned was in the text books.

As far as nicads went at the time, you peak charged them HARD and then literally overchared them to the point 0f venting. Then, under HIGH current cycling (usually 10 to 15xC or better), we discharged them until low voltage cutoff. It took a number of cycles for the discharge curves to stabilze but, once they did, the cells were fairly predictable and powerful - but only for a few cycles. YOu could rematch but there were only so many "golden cycles" from a pack of cells.

The key to getting ALL the power into and out of the cells, as far as our application was concerned, was to dynamicall match what we called "psuedo cell resisitance" - sort of a quick E = IR curve for each and every cell.

Since we knew the current draw and the instantaneious volrage, we sort of knew the instanteous "psueodo cell resistance'. When you're trying to pull 60 or more amps out of 5 or 6 nicads, series cell resistance makes a BIG difference when you're trying to get every mah out of the cells.

You probably don't have the time or budget to do the crazy stuff we did to match nicads - we did it as a hobby and not much ielse at the time. I can say, however, that when you do find and get cells that ar TRULY matched, the amount of power and capacity they posess will amaze you.

BTW, when you finally realize that cell charge efficiency and discharge efficiencies are not equal, you'll realize that you just MAY have to use a non-typical charge termination strategy. From that you'll learn more but you already have/know the basics.

Now all you you have to do is listen to what the cells try to tell you.....

[pminmo]

It looks to me like your concept is ok, diverting charge current based on voltage. But implementation is much more difficult. I assume your using some of the newer battery chemistry's which are CCCV based charge?

Switching 508A is no easy chore, but you don't need to switch the full current, just a percentage, but even diverting 10% of the charge current is 51A. 51A at 2.8V is 142 watts of energy that needs to be dissipated.

Kind of reminds me of a problem I was asked to solve many years ago, testing many circuit breakers in series. When breakers started tripping, the test needed those open breakers removed from the circuit. The final solution was MOSFETs with a sense circuit and an isloated floating DC supply. You face a similar problem, getting a mosfet turned on. The second problem is a load. Using N channel MOSFETS with the currents you are proposing, you would need gate to source voltages of 8 to 12V. You could float dc sources in an arrangement like the attached. You would need several MOSFETs depending on the diverted current amount. I'm sure there are other methods, but bottom line this isn't a simple task.

[NC Cams]

When you decide to get SERIOUS about matching, you'll find that sequentially pouring current thru cells starts to do funny things to the "match" - the "dynamic impedance" of the celss change with charge/discharge and they start to react stranglely as current gets drawn thru either high or low "impedance" cells. As cell impiedance changes so does charge/discharge efficiencies and thus goes your "match".

After doing MUCH study and MANY charges/discharges, we finally setttled on INDIVIDUAL cell peak chargiing, followed by partial overcharge to the point of incipient venting and then cycllng under load. Then and ONLY then did we get MAX capacity ouf of the cells. Our efforts to charge "batches" of cells and then selecively charge/discharge/shunt individual cells proved to be complextity in search of a need.

We finally did them individually. Time consuming, boring and tediously slow individually BUT when the cells were matched, they had glass smooth discharge curves, they ALL held nearly identical voltages and you could litearlly compute not only the MAH but also the WATT/sec's of power available from each cell and subsequently the battery. When you can match both POWER and CAPACITY, you've really done something.

And the performance difference is stunningly visible when/if you do it right.

[sswitaj]

That's a heckuva range for conventional chemical batteries. Are you dealing with something unusually exotic, like supercaps?