diy solar

diy solar

Design for "ultimate" "fire-proof" 274kWh+ battery shed

To put a LFP fire out takes an extenguisher that is primarily a slury of salt and a foaming agent. When it dries out from heat or evaporation the leftovers dry out and can be swept or vacuumed up. See the tail end of the up in smoke about the 7 bank system in the guy's garage. It toasted several of his banks but didn't burn the structure.
I think his house ended up being a total loss, he was talking about rebuilding.
 
I think his house ended up being a total loss, he was talking about rebuilding.
Some type of monitor, in addition to the BMS that can shut down a pack, or complete ESS based on anyone of several parameters would seem well worth what it would cost to develop and bring to market.
Any location using a LFP server rack could be a potential customer, not just solar.
 
I just realized I misfired this into the wrong thread -- reposting it here and deleting it in the other thread (about to read all posts since my last post and reply to them) --

I will respond to all posts before bed (currently at work) but I did want to drop this in here first just for some additional details and insight into the meticulousness and slow-paced/long-term-planning, engineering/data focused nature of this project. I have done thorough testing of every single one of my cells. My testing protocol first takes the cell I received and discharges it at a slow rate of 45W constant power measuring to see how much energy is in each cell at delivery making sure they have roughly the same (Not just a simple voltage reading that most people do). Then a 40A charge to full, 60W constant power discharge rate test, 40A charge to full again, 40A constant current discharge rate test, and finally charging back to roughly 40% SoC for storage. Each cell takes two whole days to test on one of the common EBC-A40L units that most people use. I have the screenshots of the curves and a second-by-second data log of every single cell for the entire duration of all of the tests. Here are some screenshots of the spreadsheet I use to keep track of the summary of this data:

inv1.png
inv2.png

Note the buy dates starting in 2021 and me getting burned by trash quality cells from Varicore lol. I have been slowly increasing my buy quantities after verifying that cells are quality or not in preparation for this project over years. "Initial SoC" column is the amount of energy in each cell after having received it. Interestingly, ignoring the 8 crap cells I bought, the 32 cells (FE001-032) were not tested until a year after I received them since I didn't have a battery tester yet. It is impressive that their energy contained was ~310Wh after sitting for 1 whole year never being used after receiving whereas compared to cells ordered a year later (FE033+) and tested immediately after receipt from the same manufacturer measured at ~340Wh. Meaning an entire year idle and only 30Wh (~3.3%) of self discharge on a cell that stores about 900Wh to me speaks to very high quality and very low self-discharge. Super happy to have been able to glimpse such a neat piece of data that I can't think of any other person having tested :)

The idea behind this meticulousness is to develop methods to judge the quality of a cell and also be able to forecast cell failure based on previous data vs new data. I plan on pulling random cells from time to time from this system, throwing them on the battery tester and see how their curves and data compares between each 2-3 year time period of being used. Possibly will be able to develop some software and some modelling around expected performance parameters.

The ability to track individual cells (where, almost all battery packs fail due to individual cells failing) is the primary reason I do not really entertain the idea of server rack batteries. Realistically it would be too much of a hassle to constantly be disassembling those to get at the individual cell level. This project is part research, part experiment, part practical, part desire for self-sufficiency, part money-saving (long term vs just paying for electricity), etc. and ultimately a culmination of my interests and passions.
 
I have wondered about the reliability of the BMS's too.
I have not tried this, but I have been thinking about a double BMS on a DIY pack. Bear with me...while I explain.
Say you build a DIY pack, with a JK BMS, the failure mode you are concerned with is the BMS failing to shut down charging while one cell's voltage is rising over the 3.65v limit, the cell's voltage continues to rise as the BMS failed to shut it down and ultimately the cell vents and a fire hazard is created.

Supposed instead you put TWO BMS's in series on your DIY pack. Both are programmed to shut down charging if a cell reaches over voltage protection, and both have their own cell voltage sensor wires Lets agree NOT both will be capable of cell balancing. If Either BMS shuts down then the pack circuit is cut, and you would need both to fail at the same time to have an over voltage and venting condition. Monitoring for BMS failures, we should expect if one fails closed, we should have time to find this fault before the second BMS fails. "Should".

We do a similar thing with water - double check-valves. Not relying on one, but a pair doing the same job.
There is a thing called common-mode failure. Relying on two BMS'es for this in the exact same way risks having both fail in the same way at the same time. For example, if 20kA is too much current to interrupt for the BMS without it burning up, both may burn up/fail at the same time when both try to disengage. I think two BMS'es would definitely add some redundancy or margin of safety but my preference would be to somehow wire the cell monitoring to tell all the loads and MPPTs, chargers, etc. to all shut off instead. A soft trip. Much easier and safer to tell the loads and chargers to turn off cleanly than just do the equivalent of ripping the circuit apart while its live.

The two tiered system could be say 2.7V-3.5V working range of battery for the normal BMS to shut off, and if a cell reaches <2.5V or >3.6V the whole system could be commanded to stop. Maybe an option, just brainstorming.

One other comment or question and comment. What sort of batteries are you using? LFP, LiFePO4, batteries don't generate their own oxygen at least not under anything short of complete and total meltdown with active fire.

The oxygen can only be generated when the electrolyte at temperatures around 600c. This means a cell would have vent, have the hydrogen vented when the electrolyte breaks down into two components at around 350c. One of these breaks down into more hydrogen and oxygen.

To put a LFP fire out takes an extenguisher that is primarily a slury of salt and a foaming agent. When it dries out from heat or evaporation the leftovers dry out and can be swept or vacuumed up. See the tail end of the up in smoke about the 7 bank system in the guy's garage. It toasted several of his banks but didn't burn the structure.

The major downer is the mix is conductive until dry and will short out any other cells pretty quick.
I am using LiFePO4. I was fairly certain all lithium batteries make their own oxygen and if they really catch fire there's not much stopping them. Can you link the post/thread you're talking about and also provide videos or information about this salt slurry / foaming agent stuff being used to put out a lithium battery fire? That downer is a doozie though lol... yikes.

On the subject of contactors. I am considering implementing one in my system... hmm. Can one of y'all who has used these or are familiar with these link me a recommended or common one that are used for these kinds of battery builds?
 
I am using LiFePO4. I was fairly certain all lithium batteries make their own oxygen and if they really catch fire there's not much stopping them. Can you link the post/thread you're talking about and also provide videos or information about this salt slurry / foaming agent stuff being used to put out a lithium battery fire? That downer is a doozie though lol... yikes.
I can't link to a specific thread in question, but I thought it was a commonly accepted fact lifepo4 does not produce it's own oxygen at "self sustaining" temperatures. So all you need to do is to extinguish the electrolyte. At this point it becomes a normal liquid hydrocarbon fire. There is also some hydrogen to deal with if it gets hot enough.

If you're interested in exact composition of gases, the temperature profile etc. This is a good article: https://www.mdpi.com/1996-1073/16/8/3485
 
There is a thing called common-mode failure. Relying on two BMS'es for this in the exact same way risks having both fail in the same way at the same time. For example, if 20kA is too much current to interrupt for the BMS without it burning up, both may burn up/fail at the same time when both try to disengage. I think two BMS'es would definitely add some redundancy or margin of safety but my preference would be to somehow wire the cell monitoring to tell all the loads and MPPTs, chargers, etc. to all shut off instead. A soft trip. Much easier and safer to tell the loads and chargers to turn off cleanly than just do the equivalent of ripping the circuit apart while its live.

The two tiered system could be say 2.7V-3.5V working range of battery for the normal BMS to shut off, and if a cell reaches <2.5V or >3.6V the whole system could be commanded to stop. Maybe an option, just brainstorming.


I am using LiFePO4. I was fairly certain all lithium batteries make their own oxygen and if they really catch fire there's not much stopping them. Can you link the post/thread you're talking about and also provide videos or information about this salt slurry / foaming agent stuff being used to put out a lithium battery fire? That downer is a doozie though lol... yikes.

On the subject of contactors. I am considering implementing one in my system... hmm. Can one of y'all who has used these or are familiar with these link me a recommended or common one that are used for these kinds of battery builds?
Agreed this is why I run two different manufacturers BMS on one string. Fully redundant protection. One cannot be too safe.
 
Last edited:
There is a thing called common-mode failure. Relying on two BMS'es for this in the exact same way risks having both fail in the same way at the same time. For example, if 20kA is too much current to interrupt for the BMS without it burning up, both may burn up/fail at the same time when both try to disengage. I think two BMS'es would definitely add some redundancy or margin of safety but my preference would be to somehow wire the cell monitoring to tell all the loads and MPPTs, chargers, etc. to all shut off instead. A soft trip. Much easier and safer to tell the loads and chargers to turn off cleanly than just do the equivalent of ripping the circuit apart while its live.

The two tiered system could be say 2.7V-3.5V working range of battery for the normal BMS to shut off, and if a cell reaches <2.5V or >3.6V the whole system could be commanded to stop. Maybe an option, just brainstorming.
Yes, just putting ideas out there, and see where this may lead.

My concern was not about relying on a BMS (or two) for interruption of a 20kA spike, that would be the job of the class T fuses.
My concern was the probability of a BMS failing to shut down charging if one cell passes the high voltage cut off, and continues to rise above 3.65 un-checked. The proposed monitor/second BMS/relay option is only to cut off the charging of one Pack - ie the system total ESS would not have a sudden shut down and in fact would operate just at lower capacity with one pack off-line while others continue on their normal operation.

If the probability of a BMS failing to cut off charging above the set point 3.65v is 1:1000, putting two in series could mean reduction to 1:1,000,000 (one in a thousand in a thousand).

To be most effective, the system should also alert us that a BMS has failed, or the back up BMS has triggered. So attention is brought to the situation and the pack is repaired and returned to service. It seems inevitable that a BMS will fail one day, should the result be letting a cell vent and cause a fire? or should the result be triggering a shut down and an alert.
 
The ability to track individual cells (where, almost all battery packs fail due to individual cells failing) is the primary reason I do not really entertain the idea of server rack batteries. Realistically it would be too much of a hassle to constantly be disassembling those to get at the individual cell level. This project is part research, part experiment, part practical, part desire for self-sufficiency, part money-saving (long term vs just paying for electricity), etc. and ultimately a culmination of my interests and passions.
I know with the EG4 rack batteries you can monitor individual cells using their BMS_Test software:
1716294900666.png
 
Interesting reading - thermal runaway of lfp and gasses released


The conclusions near the bottom would seem to support a detached building 15ft or more from the house. The other takaway is that 50% of the released gasses is hydrogen. There are othere that can break into o2 but at higher temps.
 
Last edited:
Yes, just putting ideas out there, and see where this may lead.

My concern was not about relying on a BMS (or two) for interruption of a 20kA spike, that would be the job of the class T fuses.
My concern was the probability of a BMS failing to shut down charging if one cell passes the high voltage cut off, and continues to rise above 3.65 un-checked. The proposed monitor/second BMS/relay option is only to cut off the charging of one Pack - ie the system total ESS would not have a sudden shut down and in fact would operate just at lower capacity with one pack off-line while others continue on their normal operation.

If the probability of a BMS failing to cut off charging above the set point 3.65v is 1:1000, putting two in series could mean reduction to 1:1,000,000 (one in a thousand in a thousand).

To be most effective, the system should also alert us that a BMS has failed, or the back up BMS has triggered. So attention is brought to the situation and the pack is repaired and returned to service. It seems inevitable that a BMS will fail one day, should the result be letting a cell vent and cause a fire? or should the result be triggering a shut down and an alert.


Who watches the watchers? I like the idea of a central external watchdog. One smart enough to know if it can take a string out of the mix given current loads or if it needs to shutdown all power.


Would the midnite solar breakers that can be tripped remotely be functional after being tripped under load or would they weld shut or ruin contacts like a contactor might?
 
Who watches the watchers? I like the idea of a central external watchdog. One smart enough to know if it can take a string out of the mix given current loads or if it needs to shutdown all power.


Would the midnite solar breakers that can be tripped remotely be functional after being tripped under load or would they weld shut or ruin contacts like a contactor might?
The Midnite breakers are designed for multiple consecutive operations at max load and can control the arc appropriately.
 
Interesting reading - thermal runaway of lfp and gasses released


The conclusions near the bottom would seem to support a detached building 15ft or more from the house. The other takaway is that 50% of the released gasses is hydrogen. There are othere that can break into o2 but at higher temps.
Or, install a temp controlled exhaust fan. Turns on around 150 degrees, or when your smoke / heat detectors go off.
 
  • Like
Reactions: JRH
Or, install a temp controlled exhaust fan. Turns on around 150 degrees, or when your smoke / heat detectors go off.
In that test I believe it produced an explosive atmosphere within 10 seconds. Perhaps constant active ventilation to stop the build up of hydrogen vapors wouldn't be a bad idea?
 
It would be better to prevent TR in the first place would it not? or at the very least try.
I would like to see a system in addition to the BMS, that would monitor & shut down a pack, if any one cell in that pack reaches: voltage over 3.65v, current over 100A, or temp over 30c. ie cut off the causes of TR.

Shouldn't this be as simple as a NO DC rated relay connected to a microprocessor monitoring the three parameters listed above ?
 
Seems easier to just monitor current for individual strings and any string that deviates by X% it is assumed something wrong and is then isolated.

The hardest part would be monitoring that rate of current change, comparing it to others and then being able to flag a string as an issue (vs normal load spike) then isolating it. What do you think that time frame would need to be in order to prevent a TR? 0.1 second? 0.001 second?
 
Seems easier to just monitor current for individual strings and any string that deviates by X% it is assumed something wrong and is then isolated.

The hardest part would be monitoring that rate of current change, comparing it to others and then being able to flag a string as an issue (vs normal load spike) then isolating it. What do you think that time frame would need to be in order to prevent a TR? 0.1 second? 0.001 second?
Correct me if this isn't right, but say the BMS fails to shut down over-voltage - and a single cell's voltge climbs above 3.65, and continues increasing un-checked. Current alone will be identical for all cells in the circuit, and monitoring doesn't catch the problem?

How long we have to catch the problem will depend on charging rate and that cell's internal behaviour. Likely to be long enough to allow a monitoring system to catch the issue and trip a relay I expect.
 
In that test I believe it produced an explosive atmosphere within 10 seconds. Perhaps constant active ventilation to stop the build up of hydrogen vapors wouldn't be a bad idea?
Ok. A Hydrogen Gas Detector to turn on the fan.

Better yet: Put the battery in a sealed case, and replace the air in the case with CO2 or Nitrogen. One way valve to vent excess pressure to the outside.
 
Or, install a temp controlled exhaust fan. Turns on around 150 degrees, or when your smoke / heat detectors go off.

I think you would want to run a small fan all the time to constantly change the air and get rid of hydrogen as it vents. Or tie the fan to the battery shutdown from the bms due to low voltage.

Ok. A Hydrogen Gas Detector to turn on the fan.




Better yet: Put the battery in a sealed case, and replace the air in the case with CO2 or Nitrogen. One way valve to vent excess pressure to the outside.


@Hedges suggested a burst disc to the outside and a nitrogen filled case. If that could also turn on a fan even better.
 
Correct me if this isn't right, but say the BMS fails to shut down over-voltage - and a single cell's voltge climbs above 3.65, and continues increasing un-checked. Current alone will be identical for all cells in the circuit, and monitoring doesn't catch the problem?

How long we have to catch the problem will depend on charging rate and that cell's internal behaviour. Likely to be long enough to allow a monitoring system to catch the issue and trip a relay I expect.
What I was thinking was how and over all system monitor that could watch multiple BMS and talk to each other for a fault.

Having a BMS fail to isolate is always a concern for a single string or 16 strings in parallel, the risk of damage for 16 is just much higher.
 
Back
Top