My A123 LiFePO4 - tearing them down and rebuilding them in a different configuration

[Roil]

I exposed you guys to my challenge of how to utilise my large stock of A123 LiPo modules back in November 2018
It led to this interesting topic with lot's of good information. "Lithium-ion multiple batteries on parallel - same challenge as Lead Acid or not", the first topic I made on these batteries are here: A123 Fisker Karma Battery module - 7 cells with built in BMS

After surviving the winter in cold Norway using my ageing AGM 2V 1100Ah cells in a 24V configuration (Discussed here: AGM) They still serve me well, with acceptable performance but I want more storage capacity. I have 9 x 300 watt solar panels and daily power production does reach 15 Kwh if I'm able to utilise it.
Norway has become an EV country, I have 2 of them and no fossile fuel burner any longer. It would like to be able to charge my car from a 50Kwh battery that has been charge from the sun during the week when I am not at my cabin.

So much for the introduction - here is the challenge

These A123 cells appear to be very well built, with a fuse on each cell. But they are unfortunate in a 7s3p configuration, and after the discussion back in November I have concluded that I have to reconfigure them if I am going to be able to use them. The modules are built with 20Ah prismatic pouch cells where each 3p is "flipped" and laser welded together. On each welding point 6 pouch cell tab's are welded together. The laser welding to the pouch tab's are welded for life an not possible to undo. The same is true for the A123 cell modules. They are very sturdy built, the prismatic cells are glued together with steel heat conductors, and held together by steel ribbons.

This leaves me with one option. Separating the series connection of the 3p packs and rewiring the hole module in a 7s x 3s = 21s configuration making each A123 module a 3.3V 420Ah (21 x 20Ah). A good module for making a 24V or even a 48V large LiPo battery of 40 cells in a 5p8s design of more than 50Kwh. Each of the 40 cells (the old modules) with a proper BMS solution.

The big question is how to connect these 3p modules wich are already laser welded together in a nice an profession way. Spotwelding, connect them mechanically with nuts and bolt (they would have to be tiny), soldering or by any other means.

Your views gentlemen is highly appreciated

A123 ombygging (1).jpg

A123 Original konstruksjon.jpg


IMG_20190512_124244.jpg

[Mike90250]

Spot Weld with nickel ribbon, is the usual method.

Soldering generally takes too long, and the conducted heat ruins the seals in the cell.

After re-configuring, what will you do for BMS and safety monitoring ?

[Roil]

Spot Weld with nickel ribbon, is the usual method.

This is doable - I am able to connect 10mm of 0.15mm Nickel strips to each 3p of cells. if I got it right this strip can handle close to 7A and my calculated worst case scenario based on my 5000W Victron inverter is 5A per connection

After re-configuring, what will you do for BMS and safety monitoring ?

If this rebuild is a success the next would be to put them in a 24V configuration of 5 strings in parallel og 8 modules (each of 21s prismatic cells) in series. 5p8s giving a total of 40 batteries each in need of a separate BMS control module.


Batteri bank config.jpg

I am currently looking into a solution from Battrium.com or REC both capable of balance and "protect" and communicate over CANbus to my Victron Multiplus Inverter and my Victron MPPT charge controller.

The obvious issue is the capacity of the BMS per cell module. With my cell modules consisting of 21s 20Ah they are massive 420Ah, however this table from Batrium indicate that it is possible .

Screenshot 2019-05-12 at 20.38.13.png

[Roil]

I have done some rethinking of batterybank design Will try to build only one string 1s5p(21p)Batteri bank config 1s5p(x21p).jpg

[Ampster]

I have done some rethinking of batterybank design.....]

That make a lot more sense and you will only need one BMS for the entire pack. Each of the buddy parallel cells will be at the same voltage.

[Mike90250]

I have done some rethinking of batterybank design Will try to build only one string 1s5p(21p)Batteri bank config 1s5p(x21p).jpg

This method (lots of parallel cells) requires that the individual cells, be tested and binned according to capacity. You can imagine what happens with a 18ah cell paralleled with 23ah cells., the BMS won't have a chance to figure out what's happened till the 18ah cell starts burning. At least that's the way I'd configure it.

[Ampster]

Binning would help insure the quality of the pack. It will also inform you where to set the charge and discharge limits. Fortunately each cell has a fuse and they are LFP chemistry so less risk of fire. If a cell goes bad most likely it will short which will blow the fuse before the others dump much current into it. Then you will have a group of parallel buddy cells that have less capacity than other groups. That group of buddy cells will tend to overcharge or over discharge but the BMS alarms should be set to cut off charge or discharge to protect the weaker group of buddy cells.

I have a more volatile pack consisting of Nissan Leaf cells in a 7P 12S configuration and I set my charge and discharge parameters conservatively.

[Roil]

Screenshot 2019-05-18 at 19.11.48.png
I have good documentation of these A123 Cells, and I also plan a rather conservative utilisation.

By keeping the cell voltage between 3.25V - 3.34V I will use 76% of max capacity. Available capacity will be 40.9Kwh

I assume securing "perfect balance" of all cells is the ideal solution, the question is how to go about getting there. One way could be to leave the modules in it's original string configuration of 7s3p. and balance every module like than, then balance each set of 5 modules (intended to be connected in parallel in the final back) by connecting them in serie, an again using the BMS to balance them. As the cells has a rather sharp Knee at 3.34V that's potentially a good point for top balance. I should probably cycle every module at least once to be sure that the cells have the same capasity.

Comments?A123 Fisker Battery Moule (Cell) build up - Spenning vs SOC.pdf

[Ampster]

Screenshot 2019-05-18 at 19.11.48.png
I have good documentation of these A123 Cells, and I also plan a rather conservative utilisation.

By keeping the cell voltage between 3.25V - 3.34V I will use 76% of max capacity. Available capacity will be 40.9Kwh

I assume securing "perfect balance" of all cells is the ideal solution, the question is how to go about getting there. One way could be to leave the modules in it's original string configuration of 7s3p. and balance every module like than, then balance each set of 5 modules (intended to be connected in parallel in the final back) by connecting them in serie, an again using the BMS to balance them. As the cells has a rather sharp Knee at 3.34V that's potentially a good point for top balance. ......

Sounds like a good plan. You also appear to understand the characteristics of those cells and have set your parameters accordingly. That is a good lesson for any one using Lithium chemistry safely. What I like about using lithium cells is the charge efficiency. You don't lose much energy while charging and the cells are well suited for an environment where they might be left unattended for a while.

[Roil]

The big question is how to connect these 3p modules wich are already laser welded together in a nice an profession way. Spotwelding, connect them mechanically with nuts and bolt (they would have to be tiny), soldering or by any other means.

Your views gentlemen is highly appreciated

Any comments of the way to connect these cells?

AC-380V-5000A-8000A-Semi-Automatic-LiFePO4-Li-ion-Battery-Spot-Welding-Machine.jpg

[NewBostonConst]

Without seeing the bottom connections I would guess the closest strip is plus and the farthest strip is neg.

[Roil]

Without seeing the bottom connections I would guess the closest strip is plus and the farthest strip is neg.

This photo was just a way to show welding cells - the 3 photos in the binning of this thread are the ones in need of welding together

[HollySprings]

I am intrigued by the change Roil made from 5p8s to 8s5p. I'd like some more insight on the rationale one way or the other. Here are the tradeoffs I see. In the initial 5p8s (5 parallel sets of 8 cells in series) the implementation has the capability to monitor/charge/track via BMS on all of the cells... with one 8S BMS for each of the 5 parallel modules (5 BMS units total.) The second version is 8s5p (8 series sets of 5 parallel cells) which would use 1 BMS for all 40 cells.
Ampster , regarding your two posts: "That make a lot more sense and you will only need one BMS for the entire pack. Each of the buddy parallel cells will be at the same voltage." and "I have a more volatile pack consisting of Nissan Leaf cells in a 7P 12S configuration and I set my charge and discharge parameters conservatively." Are you running 7 BMSs or is your setup a 12S7P running (presumably) a single 12S BMS?
I am planning on trying a Nissan implementation with an SMA inverter and my gut says to have a BMS 'touch' each module.

[Ampster]

.........
I am planning on trying a Nissan implementation with an SMA inverter and my gut says to have a BMS 'touch' each module.

My BMS touches each module. Since my modules are in parallel the voltage in each set of modules {buddy pairs) is the same because they are connected in parallel by the buss bars.The voltage sensor from the BMS is attached to the buss bars on each group of modules.

I have reconfigured so my pack is now 9P12S. To be clear I am using Nine modules in parallel (buddy cells) and six modules in series. Since each modulle consists ot 2P2S if you go down go the individual cell count my pack would be 18P24S. I am running a single BMS.