APC UPS and HV LiFePo4 pack

Joined
Dec 5, 2017
Messages
59
Greetings all,
As a long time lurker I decided it's time to start a build thread. The build is still in the early stage of planing and parts collection but it's starting to gain momentum so I need to bounce some ideas off of those with some real world experience rather than just theory and planning.

At the moment we have an off-grid system running on an APC SmartUPS3000VA with 48v PbA pack ~9kWh, 12 x 250w Trina solar panels and a chinesium 40A MPPT charge controller.This was a temporary/testbed system assembled from existing and scrounged parts meant to put theory into practicewhile we assemble the proper system.

Proper System:
We have an APC SmartUPS7500VAwhichuses 32x12V5AH SLA'sfor a nominalpack voltage of 384Vand a charge voltage of 432V.

As the pack is high voltage, even under peak load the current draw should never reach 20A.

The trick is chargingsuch asystem from solar.and finding a BMS formanaging something like 4p120s cell arrangement.

I am looking at building packs from 10AH LiFePo4 cells, something like these,as the nominal and charge voltagenumbers all line up nicely and 4p should give reasonable redundancy if I have a cell failure.

I have seen MPPT solar charger that had a PV input voltage of 300V and a nominal battery voltage of 350V, I'm not sure if it was just a typo but as it was still ~100V under my needs I didn't look into it any further.

In theory, with the 12 solar panels rated at 384Vmp and 8.2Imp,it should be possible with MPPT maintaining at leasta 350v solar outputand a boost converter lifting thesolar output to charge voltage would be less than 25% increase so should be achievable with reasonable efficiency. As charge currents are all under 10A nothing needs to super heavyduty with the exception of warning labels and insulation/isolation.

Question time:
I notice most powerwall builds appear to build cells into parallel 3.2v battery packs then join packs in series to achieve the desired pack voltage eg: 48v = 10p14s.Each cell is connected to the bus via fuse wire incase an individual cell fails,as this will likelynot be detected by the BMS.

I assume this is to keep down the cost andnumber of BMS components required compared to 14s10p where a each cell would be individually managed.

How then do you inspect individual cells for voltage, IR, etc once assembled? Or do you just monitor fuse links and hope for the best?

Are there any readily available BMS system for higher voltage packs, I know they exist in the DIY Electric Vehicle world where 400V or more is not uncommon but these are typically designed around much larger winston/thundersky cells and MUCH larger current draw. Most of the BMS mentioned in the forum seem to peak around 48v

How viable is it to live without BMS, I realise this is a can of worms, but with LiFePo4 systems naturally matching much closer to PbA systems and using new cells which should match in spec's, how likely is it they would wander out of balance? How would you balance them if they did?

Would I be better off with larger 40Ahsingle cells instead of 4x10Ah packs?

Are the any available solar charge controllers that do not require an input voltage near double the nominal pack voltage, like most of the alibaba/ebay ones that appear to be PWM?

Are there any other suggestions/ideas to consider?

Cheers,
Jordan
 
Wow that;s a big UPS. My ABB inverter for my house has a similar issue being 7.5KW as it's DC input starts at 300V.
So I'm thinking of series 100 cells to get the voltage up. 300VDC hurts but not as dangerous as AC but you still need to work carefully and insulate/fuse everywhere.

The BMS checks cells, it can't check any cell that's parallel to another hence why there's a longmon on 1 pack as they are all in parallel.

My skateboard is 10S8P and has a 10 input BMS but it can only check the 8 in parallel as a whole.
You need a BMS for your type of setup because of the constant charge discharge and any variance in capacity of the cells will cause them to charge and discharge at a different rate which you will kill your cells if one is fully discharged or charged and the rest aren't.

the more you parallel the more the BMS is monitoring as a bunch of cells rather than 1 cell 1 bms, etc.

I personally would run 100S#P # being as much parallel as you want. then build more the same size each with their own BMS.

The real problem ist needing BMS that can do 100 cells, maybe custom build one with a multiplexer?
 
Hey Jordan, thanks for the info on your build :)

Most do build in #s#p configuration and it is partly easy of build, and cost of parts to keep down. If cells were arranged in #p#s, each series would need a bms on it to keep track of things. This may be ok on small systems (<1KWh or so). But once you start getting into the large builds, the cost of a bms for each series string is almost, if not, cost prohibitive. Or, at least, currently.

With a system as large as you are planning on building, I would not set it up without a bms. There's way to many things to monitor by hand. You'd drive yourself crazy checking everything.
Now, you could setup a little bit of smart monitoring with your rPi's/arduino's. You could put a shunt between each parallel pack and monitor current going in/out. If a pack isn't delivering as much current as the rest, then there might be something wrong with that pack and at least give you a heads up before it becomes a problem. But that's about the only way to handle parallel packs.

daromer will chime in soon, I'm sure. He's the LiFePo guru and knows a lot of about the more specialty MPPT's and such charge controllers.
 
Welcome! Yeap I will chime in :)

First of all sxpx is more normal as Korishan said. Thats because you only need to monitor 1 string. The other way around you get 1 string per p.

Batrium BMS is the go to for that high voltage and personally i would not go without BMS on such a system. Batrium can easily handle 400V on the battery side. You can hook up as many as 250 cells to same master. http://www.batrium.com

There are a few MPPT chargers out there that handle HVDC but they arent many. It might even be that costly that its easier to just get a 48VDC Inverter instead.
Difference going 48VDC will be that you will need thicker wires and all that but the current per cell will stay the same as long as you use same amount of cells.

Im looking forward to see if you find the charger needed.
 
BaronVonChickenPants said:
How then do you inspect individual cells for voltage, IR, etc once assembled? Or do you just monitor fuse links and hope for the best?

You don't because you have done that prior to assembly or just know by reading the datasheets when the cells are new. And from that point of assembly onwards the cells are connected and age in the same way. The voltage will also always be the same because they are connected in parallel.
 
Thank you for the feedback everyone.

I did intend to place the 3.2v cells in parallel but misinterpreted the ordering of XsYp vs XpYs.

So after some hunting and googling I was having great difficulty finding any sort of boost converter that would accept an input voltage anywhere near the 350 volts expected from the solar string. I was also concerned about the potential downsides of 12 solar panels in series, such as if one panel goes into shade there is basically no output.

I then changed direction and ran down a different rabbit hole, what if each panel was individually stepped up to 450-ish V then connected in parallel, apparently I wasn't the only one to have this thought as I found a very promising looking research paperon exactly this approach, achieved with 98.6% efficiency. The paper also talks about incorporating MPPT into the same circuit so that each panel is individually managed for optimum output, stepped up to the desired charge voltage and connected to a common HV DC bus. Their main issues then are that most traditional PV inverters MPPTdoesnt play nicely with this output, but this not relevant for my use.

If I have read the Batrium specifications correctly, the BMS is also capable of acting as a charge controller, meaning the Batrium controlled battery pack, the APC inverter and the output of the solar panels would all be connected to a common HV DC bus.

Further investigation required but sounds promising so far.

Jordan
 
if each panel was individually stepped up to 450-ish V then connected in parallel, apparently I wasn't the only one to have this thought as I found a very promising looking research paper on exactly this approach, achieved with 98.6% efficiency. The paper also talks about incorporating MPPT into the same circuit

Hrm, this sounds an awful lot like a micro-inverter, but keeping DC as the output instead of sync'd AC wave. So each booster would have a cheap arduino or other mcu as the brains and boost the voltage and mppt for ya. And, you can set up communications so you can see its particular stats and such :) Nice find!
 
And on top of that you need watch out for the specifications of the DC-DC converter, not all of them can be put in parallel.

Also, it isn't such a big issue when panels in series are partly shaded. You can use bypass diodes, the panels might have some built in already, to basically "skip" the shaded cells or modules. You might lose some power, but the output will stay active.
 
I had considered micro inverters to feed the APC UPS or some other charger, but i didn't seem very efficient. Direct DC-DC at high efficiency certainly seems better, just don't lick anything.


I can see how parallel connection to the HV DC bus would need diodes to prevent the output capacitor charging from the bus and also effecting the flyback reading incorrect values from the bus rather than the boost converters own output.

I realise shaded panels isn't a huge deal but when we're struggling to reach chargevoltage already, losing another 50-100v to shade could be an issue.


I have also found a another study with a similar circuit design:http://www.sciencedirect.com/science/article/pii/S2090447914000689

It's some heavy reading with lots of electrical theorybut their test's showed 94% efficiency with 12v panel input and 120v output. I think the circuit in the original article was more elegant and simple but both show promise for further investigation.

Jordan
 
This is similar to how the Solaredge system works, and how their Powerwall v1 DC tie worked. The Solaredge uses optimizer boxes behind each solar panel that when activated by the main inverter will all balance Voltage vs Amps to maintain a roughly 400v DC feed for the main inverter (the panels can be mixed output or shaded without loss). The original Solaredge/Tesla hybrid system was overly complex but it worked and the v1 powerwall could supply or charge directly from the solar feed. Unfortunately Tesla decided to create the v2 powerwall as a direct competitor to Solaredge with an integrated inverter and solar connection so the partnership has been mostly dissolved.

I am in the same boat as you, and working on solutions to keep the excellent performance of my Solaredge grid-tied solar system while also figuring out a way to charge a powerwall efficiently and when their is no grid available(zombie attack). Having 10KW+ of solar be unavailable during an outage is not acceptable.

There are some high voltage inverters and chargers, but as you have seen - not many. The Solarege solution also goes into a safe mode where the optimizers only produce 1v DC if the main inverter is not online. A fantastic solution that means I can work with solar lines in the middle of the say safely, but terrible when you want to just use the power when the inverter and grid are down. So I bought one of their installer tools called a Solaredge "Key" that lets me hook up to the DC line and tell the optimizers to override their safety system and just act like a big string of panels.

Will post if I figure out a good solution but probably will directly charge from AC and Invert from the powerwall for a home sized UPS that I can manually cut the solar over to if we have an extended outage.

David
 
This project has taken a few detours along the way but has basically been abandoned in favour of a 48 V DC LTO system.

While technically possible and superior in regard to current handling, the scarcity andcost offinding high voltage components seems to offset any of the advantages. Not to mention everything becomes a compliance issue once you get into HV territory.


For anyone still interested in HV i have made the following discoveries:
The solaredge and TiGo dc optimizers all seem to operate at relatively low voltages, mostly Voc of the panel they are connected to or of a pair of panels in series, 35-90v DC.

I did find a product that ticked all of the boxes, the eIQ vBoost modules take a single panel and elevate it's voltage to a 410-420 V DC parallel bus.
http://eiqenergy.com/vboost.html
http://eiqenergy.com/assets/eiq-energy-vb300x.pdf

The HV PV buscould be connected to 155 LTO cells in series, and connect to the APC 7500/10000VA UPS's.All of the charge, float, operational voltages checked out as safe on paper and could theoretically achieve an operational system.

I have made enquiries to eIQ and other places promoted as distributors but as yet I have not received a reply. I have also notseen any information publishedsince 2012 so this may be a dead product.

I intended to use the LyteFyba BMS developed by Weber and Coulomb on the AEVA forum for management of the HV pack.

As I never really found a satisfactory way to charge the HV pack any further investigations seemed irrelevant.

Regards,
Jordan
 
Back
Top