High Efficiency BMS Project

[Glenn0010]

Hi all,

I want to design a more efficient battery management system. Rather then dumping the energy of a higher voltage cell in a resistor, I want the energy of the high voltage cell to be used to charge the low voltage cell.

When balancing would be required, with my current concept, the cells would be removed from the series configuration and be and appropriate cells would be put in parallel to  balance. Now I am in the concept stage and have come up with something that might work but will have a lot of issues as seen below.



1 - Intrinsic body diode for charging and discharging would mean that I cannot truly separate the cells and make them in parallel as see by the 'boxed switches'
2 - High number of switches
3 - High loses due to multiple switches in series.

Now if we were to ignore losses and just focus on something that could work how could we change/optimize the design. I have only started brainstorming and welcome your ideas as this would be a pretty cool project.

[ajb]

The body diode problem can be solved by a pair of back-to-back MOSFETs, that's what BMS systems already use.  Of course that means two transistors for every switch.  I'm not sure that the loss due to switch resistance would be that bad, Dewalt actually does something similar with their "flexvolt" batteries, that allow a 20V 3P5S pack to be reconfigured as a 60V 15S pack.  I guess that the losses in the switches must be less than the efficiency/performance gained by running the connected device at higher voltage, although in their case they don't need switches on every single cell. 

The current flow between higher- and lower-state-of-charge batteries won't be controlled in your scheme, except by the resistance of the switches and internal resistance of the batteries.  That means higher I²R losses in the balance portion of the cycle, and potentially harder on the batteries unless you stop charging to balance quite frequently.

I think the ideal solution is to keep all of the cells in series, and connect essentially a two-quadrant buck converter in between each pair of cells.  By controlling the overall pack charge current and the current into or out of each of the inter-cell nodes you can essentially control the charging current on each cell individually.  That means you can continuously balance while charging, and you don't have to add any switches that need to handle the full pack current.  You could even balance while discharging if you wanted to.  The overall efficiency will of course depend on the efficiency of the converters involved.  But it's a lot of complexity to add for dubious benefit.  Maybe it would be worthwhile for a high-capacity pack that needs to be recharged and balanced as fast as possible.

[station240]

The usual way of doing this is to have isolated DC-DC converters, which means lots of little transformers.
Your idea of using lots of mosfets, is bound to be less complex, and probably more compact.

Bear in mind you cannot charge/discharge and balance the cells at the same time. So may need two sets of batteries.

[NiHaoMike]

The cells are matched well enough that the balancer rarely comes into play.

[Glenn0010]

I'd still need loads of little isolated DC/DC converters to provide the gate voltage for the MOSFETS since most of the MOSFETS would be  'high' side mode in this configuration and that would be pretty expensive.

[Glenn0010]

I'm  going to have to put in some research into two quadrant dc dc converters and see how I can use them as you suggest.

To be honest with my solution, gate drive is going to be and issue since Isolation is required for almost every swithc

[jbb]

I think you really don’t want to put any extra devices in series with the batteries; they would need to handle the full current and would add resistance to the pack.

Linear Tech do a chip for ‘active’ balancing . http://www.analog.com/en/products/ltc3300-1.html It uses bidirectional synchronous flyback converters to shuffle energy around.

[Glenn0010]

Oh wow I did not realize this existed. This is amazing I will probably set my design round this part number. It's good as I will get to know the new MCU that I am using before I move on to a VSD which is more complex

[capt bullshot]

For a project that is still in progress, I've built this active balancer for a Li-Ion battery block with 4 series and many parallel cells.

The balancer is a synchronous rectified DC/DC using a coupled inductor with 4 equal sections. From the 4 series cells, the mid tap goes to the center of the inductors, the "inner" taps and the "outer" (full pack voltage) connections go to an H-bridge each. The H-bridges are driven by a 50% duty cycle signal.

See the rough sketch to (hopefully) get an idea. I haven't drawn a proper schematic yet, but the balancer works.

The red blocks are some pulse transformers I had at hand, used to drive the gates of the MOSFETs.

[Glenn0010]

Hi guys,

I found this amazing source:

https://pdfs.semanticscholar.org/39a8/bcda90c9b0c5ea2495a3fbf679a4488d2cf5.pdf



It is the concept I am going to proceed with.

However I cannot get it to work properly on LT spice for some reason





The voltage on the secondary source is remaining at 0. I am definitely missing something. Can anybody help?

[Siwastaja]

This is an absolute classic!

Don't do the same mistake everyone (me included) did.

Of course you want to get rid of energy loss, everyone does. When you see a resistor, you want to replace it with an active circuit "without loss".

But before committing to designing a system to "reduce the loss", which will be a lot more complex and use more expensive, and heavier parts, do yourself a favor and try to analyze, or at least estimate the expected savings.

Don't optimize meaninglessly small costs. Optimize the biggest part you can.

Spoiler: Redistributive balancing in li-ion cells, always and almost unexceptionally, results is increased total system cost in all measurable aspects (energy, total lifetime CO2, cost, environmental...), by a wide margin.

So it basically never makes any sense - nor isn't even close to making any sense.

This is why in practical systems, such redistributive balancing is almost never used, and it remains in two fields:
1) academic research,
2) beginner projects

I did it with synchronously rectified bidirectional flybacks. It was a fairly nice design, able to transfer charge from any cell to any other, through an isolated 24V "balancing bus", or from any single cell to full pack, or from the full pack to any single cell or group of cells, through extra converter at the end of chain. Sad it was completely meaningless - even when designed to minimize cost (with over x10 reduction compared to a competitive distributive balancer). I dropped the project midway. The BOM cost for the dissipative version ended up around $0.50 per cell - the redistributive would have been around $5 per cell - the competitive over-engineered reference point was around $50 per cell.

A typical li-ion pack consumes balancing energy in order of about a few percents max of the total capacity during the lifetime.

It's not unusual to have no balancing at all. Including some balancing capability, while making sure it doesn't accidentally imbalance the cells, offers you some reassurance that you won't be losing a few percent of capacity after several years of use due to slight differences in self discharge (typically around 1-3%/year) or coulombic efficiency (typically around 99.98%).

For this level of imbalance, resistive is just fine. The amount of energy is so small everything else dominates: from the energy viewpoint, if this is a vehicle, moving around extra components, even just some hundred grams, has an energy cost as well!

Failed cells are an exception to this rule, but failed cells would require very powerful converters to supply required power to bypass significant charge all the time, making it even more expensive, and they would most likely continue failing beyond usable very soon anyway.

Strongly mismatched (by capacity) cells - for example, let's say one is 100Ah and another is 110Ah -  is another possible exception, but utilizing all the capacity from all the cells would again require impractically powerful converters, since they would need to supply the difference over the single short cycle. In this example, 5A balancing current would be required for two-hour discharge. Such 5A converter for each cell would cost so much that you could double the capacity by just buying more cells. In any case, you just shouldn't use strongly mismatched cells, so the point is moot.

You'll find a lot of sources and topologies you can read about, and I remember that designing this was very fun and educative and gave me a lot of experience in designing switching converters and custom transformers. I hope I won't ruin your fun.

[Glenn0010]

I definitely agree. Closely matched cells would not really need that much balancing. However I want to proceed just the same for the following reasons:

1 - As you said gain experience with switch mode power supply design
2 - Gain experience with a new Uc I will be using before moving on to the drive.

I am not very experienced with LTspice so if you have some input on that I would highly appreciate it.

Thanks

[capt bullshot]

I found this amazing source:
https://pdfs.semanticscholar.org/39a8/bcda90c9b0c5ea2495a3fbf679a4488d2cf5.pdf

This is an inded an interesting reading. As far as I remember, the flyback charge redistribution method is also described somewhere in Linear Technologies app notes by Jim Williams.

However I cannot get it to work properly on LT spice for some reason
The voltage on the secondary source is remaining at 0. I am definitely missing something. Can anybody help?

This is due to the catch diodes that you've placed across the transformers windings. These short out the flyback operation. You'll have to remove them and fight the inductive kickback caused by stray inductance with some kind of snubber.
And there lies the problem of the flyback based approach, the stray inductance will cause losses in the snubbers or lead to more complex circuits using active snubbers.
Also note that this flyback charge redistribution will require a rather complex control circuitry.

How many cells will your battery pack have?

[Glenn0010]

Thanks for the reply!

As stated previously this is just so I can gain some experience with isolated SMPS. Hence battery pack size is not really an issue. However I was going for a 4S or a 6S to start with.

I will also take a look at the app note and remove the catch diodes as you suggested.

Thanks

[capt bullshot]

This is the app note I remebered:
http://www.analog.com/media/en/technical-documentation/application-notes/an112f.pdf
Alas, it's not about balancing the cells but measuring the cell voltages.

[Siwastaja]

I definitely agree. Closely matched cells would not really need that much balancing.

Matched on which parameter?

This is an important distinction when talking about matching.

To extract full capacity out of cells poorly matched by capacity, you need a very specific form of redistributive balancing with:
1) Very high balancing power
2) Very high efficiency

This means shuffling charge with rate high enough so that the cell balance will be dynamically changed during one single cycle.

This is typically not much discussed (because it would be prohibitively expensive); discussion on balancing, including most discussion on redistributive balancing as well, is the "normal" sort of balancing, where cell capacity matching is irrelevant: the balancing power is used to only keep the balance at a fixed point (typically so that every cell is at 100% SoC at the same time). In this case, balancing current only compensates for the internal loss of charge in cells.

In these cases, cell matching by capacity is totally irrelevant from the balancing viewpoint.

Balancing is there to help with mismatches on:
1) self-discharge rate
2) coulombic efficiency
only.

Now, with self-discharge typically near zero, and coulombic efficiency near 100%, the absolute differences thereof will be very limited as well. If you have one cell self-discharging at 1%/year and another at 2%/year, even if they are relatively very different (i.e., non-matched on that parameter), the absolute value is still negligible.

I understand there are sources of low-quality (often not low-cost, though, even though people think so) cells where there are significant differences between capacity and/or internal resistance. There parameters, however, are totally irrelevant for balancing.

Sadly, when any cell shows high enough self-discharge to warrant "efficient" redistribution balancing devices over simple resistive solutions (such as exceeding 20-30%/year), it's highly likely that this particular cell is about to totally fail soon.

Simple dissipative solution will be able to prolong the life of some corner case packs with some cells starting to get "slightly bad", before they are over the edge of quick decay.

[digsys]

.... A typical li-ion pack consumes balancing energy in order of about a few percents max of the total capacity during the lifetime ..... 

Having designed / worked with racing packs for solar cars for years, this is pretty much what we all find, except for 2 differences.
Because we push our packs way past their limits, we do go out of balance more frequently. As Mr S says though, in normal use, they should rarely drift over a year.
If they do, you have done something WRONG. The other difference is - we NEVER ever put ANY type of cell balancing in the pack !! It's a recipe for danger.
I wire a "utility" socket, which connects all the cells (~10A rated) to a secure serviceable location in the pack. IF the BMS then detects an issue, just connect up
the active circuits (same as Mr S describes). Plus you end up with a very handy "maintenance / test" port.
I've posted a few relevant research documents on my ftp site for reference - www.digsys.com.au/ftp/ActiveCellBalancing_06422665.pdf  then -
BMSCellEqualizer_06422663.pdf , MultiPhaseEqualizing_06165846.pdf , ActiveCellBMS_05289837.pdf , TequniquesCellBalancing_05289663.pdf
Heaps of ideas and explanations

[Siwastaja]

Thanks digsys for sharing your experience.

Pushing packs "over limits" may mean copper dissolution in overdischarge, or lithium plating in overcharging (including charging too quickly, often even within datasheet values(!), or charging at low temperatures). Both operations mess up with coulombic efficiency; they "move electrons" (charge) out of the cell / in to the cell, without performing the desired (reversible storing or releasing) chemical reation with said charge. Now, imagine charging multiple cells in series, and same current going through all of them, if one cell decides to not store the charge, but to do something else (plating metallic lithium, for example) with it, another cell storing the same charge properly gets to higher SoC, thus imbalancing the pack. The cell that didn't store the charge properly gets damaged by it, decreasing its capacity.

Thus, imbalance in li-ion cells is almost always at least partially related to the cells permanently aging and losing capacity at the same time.

Copper dissolution due to severe overdischarge can permanently increase self-discharge.

Yes, I agree it's sometimes a good idea not to include permanently-connected balancing at all, even though it's supposed to be 100% normal everyday thing, supported by a large IC industry, for over two decades now.

I have just seen too many problems as well.

If you can, add a cell-by-cell connector, and monitor the balance manually with a multimeter. Once a year is definitely OK for non-abused packs. You can even rebalance manually. Once per product lifetime is just fine (for example, after 3-4 years). Of course, this is often only acceptable for DIY projects or products aimed for tinkerers / people who know what they are doing.

In many cases, connecting a separate balancer every now and then, is not an option (users expect maintenance-free, completely integrated products). Then, the options are to either not provide balancing at all (and take the risk that in certain packs, the product lifetime could be slightly shorter, for example 5 years instead of 7, due to accumulated imbalance leading to excess capacity drop hitting your End-Of-Life limit (e.g., 70% capacity) earlier); or to provide integrated, always connected balancing circuitry, which has a risk that it doesn't work as intended, possibly causing way more dead units in far shorter time if everything doesn't go well. I see this as a choice between "small risk, small consequence" and "small risk, medium consequence", and hence often choose not to implement balancing at all.

[digsys]

... Yes, I agree it's sometimes a good idea not to include permanently-connected balancing at all, even though it's supposed to be 100% normal everyday thing, supported by a large IC industry, for over two decades now ....
... If you can, add a cell-by-cell connector, and monitor the balance manually with a multimeter ...
... In many cases, connecting a separate balancer every now and then, is not an option ... 

I've seen the after effects of even "simple" top balance circuits going wrong .. we lost a $1-2 million race car many years ago, others have as well.
Also, if you transport the pack by air or even ship and the coroner determines that it was a poor design decision ... It won't be ME answering :-)
The on-board BMS keeps accurate track of the entire battery pack and history - I run 3 load lines at 3 power levels. It will then "let you know" to go a and "add"
a balance at your next service !! I am trying to standardize a multi pin plug / socket so all they simply do, is plug-in the multi-balancer and the BMS instructs
which cells need attention. I'm working on a simpler version currently.
 ALL vehicles should have a service interval, and a quick plug in-out is very efficient.