What caused my electric car fire?

My thoughts on the Tesla battery fire

It’s been exactly two years since our electric Volvo Amazon wagon caught fire during our trip in Sweden. A question that came to my mind very often, and still does is : Why? Like me, many people want to know ‘What was the cause of the fire, do you know?’. I did a lot of research and reading about electric car fires. My answer was that I had some theories but probably would never know for sure. Even though I still want to do some experiments to find more answers, time has come to share my ideas on what might have caused the fire.

29-07-2023: New insights, thanks to input of people reading my story. Check chapter 15. Addendum.

  1. Recap of the context
  2. Why do lithium-ion batteries fail?
  3. Battery management system faults?
  4. Particularities worth mentioning
  5. Understanding battery chemistry
  6. Hypothesis 1 : External short circuit
  7. Hypothesis 2 : Internal cell short circuit
  8. Did my capacity fade or did I not use it?
  9. Pack temperature aka Rapidgating
  10. Internal resistance perspective
  11. Wrapping it all up….
  12. Conclusion
  13. Epilogue : Lessons learned and prevention
  14. Acknowledgments
  15. Addendum : key insights based on feedback

1. Recap of the context

Most cases of EV’s catching fire are known to occur during charging. This was not the case here. The car had been charging the day before on June 29th 2021. I monitored the charging process on the charge controller and saw the end current was about 0,7A and the pack reached 388V. This matched the 4,05V per cell average that I had defined as the target voltage and thus 100% state of charge. For these cells 4,05V is relatively low but I wanted to be on the safe side to be sure not to be overcharging.

This is the last photo I took on June 29th around 23:45, the evening before the car burned down.

Parked not charging

I used eight Tesla Model S battery modules to convert the car. Three were in the front en five were installed in the back.

On June 30th around 7:55 someone came to our tent and yelled ‘Your car on fire’.

For sure the fire started within the front battery box. When I arrived at the car there was white smoke coming from the car and I heard popping sounds from venting or exploding cells.

Electric car on fire

The answer on why my converted car caught fire is not an easy one to answer. Let’s first look into failure modes of electric vehicle batteries. In other words, why or how does an electric car fire start?

2. Why do lithium-ion batteries fail?

In the 2019 report Fire Safety of Lithium-Ion Batteries in Road Vehicles by Roeland Bisschop et. al. from Research Institutes of Sweden (RISE) a lot of information is summarized and explained regarding lithium battery fires. They describe six battery failure causes and I did an assessment of how likely this cause is in my situation:

 CauseCircumstancesMy situationScore
1Internal cell short circuitMay occur very suddenly and without previous warning.Applicable++
2Mechanical deformation and impactSevere deformation may be a result of certain crash or ground impact conditions.Not likely, car was parked for more than a day.
3ChargeWhen overcharging dendrites may grow on the anode and penetrate the separator causing an internal short circuit.
At the cathode overcharging may result in thermal decomposition and thus generation of heat.
Not likely, car was not plugged in and the highest cell was at 4,05V
4DischargeOverdischarge abuse occurs when discharging battery cells below their minimum voltage.Not applicable, cells were way above the minimum voltage.
5External short circuitMay occur in case the battery is exposed to for example severe mechanical deformation, immersion in water, corrosion, etcetera.Could be applicable+
6Exposure to high temperaturesWhen exposed to high temperatures (> 190 °C) internal degradation mechanisms and exothermic reactions may lead to problems.Not applicable, coolant and ambient temperature was around 25 degrees.

For my assessment I defined overcharge as more than 4,2V and overdischarge as less than 2,5V as per the datasheet of a Panasonic NCR18650B which is similar to a Tesla cell. However, it can be the case that due to degradation cells were already under stress within this bandwidth. I’ll discuss degratation further in paragraph 7, Hypothesis 2 : Internal cell short circuit.

So that leaves us two remaining possible root causes for my battery pack fire:

  • Internal cell short circuit
  • External short circuit

Have there been any ‘early warning’ signs? I had a decent battery management system (BMS) from Lithium Balance. Could that not have mitigated the fire? Or are there any indications something was wrong or broke down and started to go wrong? Any insights from the BMS log?

3. Battery management system faults?

In addition it could of course be that the BMS was measuring things wrongly and the data I was seeing was partially incorrect. Or perhaps a loose cell tap?

No CAN-BUS log

While the guys from the fire department managed to pull the USB memory stick from the display (the Ecumaster ADU) I used for logging, luck was not on my side.

The day we left to Sweden the Vehicle Control Unit (GEVCU) broke down. New Electric supplied a replacement and that one had other issues. The new VCU did not send an RPM signal. At that time I concluded we could do without.

No CAN   BUS log due to GEVCU fault

Only on June 30th it turned out that this signal was the trigger for the ADU to start logging. So unfortunately I did not have any data from June 6th onwards.

Lithium Balance battery management system log

The good news is that the BMS master was in the back and undamaged.

Rear end all good and BMS master OK

So after opening the rear battery box I hooked up the BMS on the bench and looked at the error log.

The Lithium Balance sBMS does not have a realtime clock so the events are not timestamped. But on June 23rd (so one week before the fire) I downloaded a log from the BMS to my computer so I could compare those two and find out what the recent errors were. I only found two types or errors:

  • Overcurrent IN, e.g. Current: 3,5 A, Threshold: 1 A
  • Contactor error, Main –

These are not significant. The Main – error was a low 12v battery so the Main – contactor did not want to close. The Overcurrent IN was due to the fact that the charge controller (Thunderstruck EVCC) could not use and interpret max. allowed charge current parameter from the BMS. As a result the BMS would open the contactors when the charge current driven by the EVCC would be too high.

But no findings like “Leak detected” or “Cell unmanaged” or similar. I decided to double check that and disconnect some cell taps on the bench.

For testing cell tap1 disconnected
For testing cell tap1 disconnected visible in errors

Which immediately was visible in the Lithium Balance dashboard since it then reported a ‘Cell over voltage’ and ‘Cell end of life voltage’ error.

And based on this and my contactors off settings, the contactors would have been opened of not closed at all if there would have been a loose cell tap connection.

Contactors off settings

Final check I did is if the cell voltages I measured with a multimeter matched with those reported in the BMS software and with those send via CAN-BUS and show on the display.

This was indeed the case.

Voltages on display correct

So that still leaves us with two remaining root causes for my battery fire, an internal cell short circuit or an external short circuit. Before diving in to those I want to bring three things to the table that are ‘unusual’ in my case.

4. Particularities worth mentioning

There are three topics that worth mentioning.

1. Evans waterless coolant

Regular coolant is conductive. On average the conductivity is about 3 to 4 mS/cm. Therefore I did quite some research on alternatives. If there would be a leak in the system I did not want to introduce an external short circuit. I found Evans who claim:

Evans Waterless Coolants contain no oxygen and are virtually non-conductive effectively preventing corrosion.

On their page ‘no electrolysis’ this is quantified as an conductivity of 0,7 mS/cm.

Furthermore they mentioned that electric truck builder Emoss uses it as well so I decided to give it a go.

Evans waterless coolant

But what I still want to test: if there has been a leak and the waterless coolant gets in contact or has been in contact with a Tesla battery cell, what happens?

2. Modified Tesla modules

As mentioned, I only used eight Tesla model S battery modules. That was the maximum I could fit in terms of volume and weight. However to be able to run at 96s (355V nominal) they were modified from 6s74p to 12s37p by the seller.

Modified Tesla modules

Additional cell taps were installed, cell tap wires were covered with an extra sleeve and glued in place with special non-conductive glue. Furthermore the cell taps were fused. But one questions that kept coming to my mind is: Can the modification have been the cause of the fire?

He used a CNC’d path and a router with two vacuum cleaners to prevent particles falling into the module. After a careful inspection the sealed the cut paths with special sealant.In theory there could have been debris or perhaps a chip that remained on the busbar and could have dropped later. Knowing how thorough the performed the modification this is not likely, but you never know.

3. Sudden increase in cell voltage delta

Third thing worth mentioning is that the evening before the fire the delta between the highest and lowest cell was slightly bigger than it used to be.

Normally it was 0,02V but at that point it was 0,06V.

Back then I was not alarmed and did not log into the BMS to check further unfortunately. With today’s knowledge that probably is THE indicator something was going wrong already.

Cell voltage delta increase before the fire

But what does that exactly mean if ‘something is going wrong’? And if something starts to go wrong, will it always end in a thermal runaway? To answer those questions we need to explore the (implications of) the battery chemistry.

5. Understanding battery chemistry

There are three common battery chemistries used in electric vehicles nowadays. The majority of the manufacturers use NMC. Or in full Lithium Nickel Manganese Cobalt Oxide (LiNiMnCoO2). Tesla uses NCA (in full Lithium Nickel Cobalt Aluminum Oxide (LiNiCoAlO2)). The third one is LFP Lithium Iron Phosphate(LiFePO4). They all have their benefits and disadvantages.

In their 2010 report “Batteries for Electric Cars. Challenges, Opportunities and the Outlook to 2020” the Boston Consulting Group (BCG) visualized tradeoffs between various battery chemistries:


There is no chemistry that has the maximum score on all parameters so it’s always a compromise. The NCA chemistry Tesla used in my modules facilitates high power output and high power density but has a lower score on safety. At the other hand of the spectrum LFP is intrinsically more safe but has a lower energy unit per mass. But how does ‘battery safety’ look like? How can we determine or quantify that?

As concluded in the Vehicle Battery Safety Roadmap Guidance report by Daniel H. Doughty for the National Renewable Energy Laboratory (NREL), thermal stability is perhaps the most important of several parameters that determine safety of lithium-ion cells, modules, and battery packs.

Thermal stability

In the article Electrical safety of commercial Li-ion cells based on NMC and NCA technology compared to LFP technology the ‘thermal stability’ is quantified as self-heating rate dependent on the cell temperature measured at the cell’s casing.

The maximum temperature rates of LFP are much lower than the ones of the NMC and NCA cells. NCA and NMC cells show temperature rates of more than 400 °C/min. This means, that the investigated LFP cells show a significantly higher thermal stability and that thermal stability of the NMC cell is higher than of the NCA cell.

Thermal stability of lithium   ion batteries

This means that once a NCA thermal runaway starts it propagates very very fast.

With the above in mind, let’s look at the two remaining possible root causes for my battery pack fire.

6. Hypothesis 1 : External short circuit

So let’s call that hypothesis one, that the fire was caused / initiated by an external short circuit.

An external short circuit can happen at pack, battery box, module and cell level. At higher level however, fuses are in place to blow before the short turns into a fire. When off/idle, the front and rear battery boxes were disconnected and due to the design and use of contactors, no two live cables were coming out each of the boxes. This means the highest voltage short would have been at the front battery box level. In my view it is not likely a module battery cable would cause a gradual short between a module + and -. Maybe via the battery box enclosure but that would have resulted in isolation monitoring faults (which did not occur). Plus the cables all had their original isolation, heat-shrink with glue and an extra sleeve. A cell tap short between two wires from a battery module to the BMS also is not likely since that would have tripped a cell tap fuse and resulted in an unmanaged cell. Zooming in further it could be a (new) cell tap wire on the module itself. But I expect then the 0,34mm2 wire would act as a fuse itself and melt/evaporate. That leaves an external short circuit at 37p cell group or cell level.

Still I cannot get my head around completely at how this would actually look like. At the positive side of the 18650 the plus and minus are close together since the outside of the cell is the minus. Perhaps a coolant leak where coolant got between the plastic top and bottom cover of a battery module and shorting cell groups or cells?

Then the Evans should be too conductive after all. Or perhaps coolant got into the cell causing protective barriers within the cell to dissolve and result in a short?

18650 Cell

How conductive is (Evans) coolant and what is the impact?

As mentioned above, Evans coolant has a conductivity of 0,7 mS/cm and regular coolant of 3 to 4 mS/cm. Conductivity is the inverse of resistivity (ρ). We can use Pouillet’s law to calculate the resistance (R) in Ohm from the resistivity and the surface (m) and length (m).

\[ R = \rho \frac{l}{A}\]

If we assume there is a drop of coolant of 1 x 1 x 2 mm on top of a battery cell shorting + and – then we can calculate the resistances. Using Ohms law I = V / R we can then calculate the current at 4,05V. Building on the conductivity of Evans that would be 0,14 mA and regular coolant would result in a current of 0,6 mA. Not something to directly worry about in terms of heat development.

Other type of external short could be that due to the modification of the module the isolation barriers changed. Or perhaps cooling tubes now acted as a conductor between two cell groups.

Unfortunately inside the battery box all I found was a big black mess and empty battery cell shelves. So no pointers or clues regarding external shorts or other interesting insights.

Inside the battery box with burned Tesla modules

So let’s move to the other possibility, the internal cell short.

7. Hypothesis 2 : Internal cell short circuit

The internal short circuit is the most tricky one. There is not much you can do to prevent it and when it happens the damage is often severe. An internal cell short circuit may occur very suddenly and without previous warning. The cell discharges its energy through the short circuit which results in heat. Once this gets beyond the tipping point, we’ve seen that temperature rates of NCA can be extreme and thus resulting in a full thermal runaway quickly.

So one of the questions is, what can be de cause of an internal short?

The RISE report mentions this can be a result of manufacturing defects, physical damage due to dendrite growth or mechanical deformation. In my case manufacturing defects or mechanical deformation do not seem likely cause the cells had been used in a Tesla and in my car and were not crashed shortly before the fire started. So let’s have a further look at dendrites. But first let’s look at battery aging and degradation in general.

Battery aging and capacity fade

A lithium ion cell consist a.o. of a anode, cathode an electrolyte which all are carefully designed and manufactured. But how do lithium batteries degrade?

In the Journal of Power Sources I found the article ‘Modeling of lithium plating induced aging of lithium-ion batteries: Transition from linear to nonlinear aging‘ in which the authors conclude:

Dominated by the solid electrolyte interphase (SEI) growth, the capacity decay of lithium-ion batteries exhibits a square root dependency on time if in storage. If under load, capacity decay is found to be linear with the charge throughput at the beginning, indicating SEI-dominated aging. After prolonged cycling, however, the cell capacity dropped abruptly and cell impedance increased sharply, indicating that some other mechanism took over the aging process.

Also the article A closer look at how batteries fail by Paul Voelker mentions “Extended storage of the lithium-ion batteries is another condition that results in an incremental increase in solid electrolyte interphase (SEI) film thickness and capacity fade.”

SEI growth, say what?

So you need a certain amount of ‘natural’ SEI growth, but not too much. But what is it and where does it ‘grow’? In another article in the Journal of Power Sources called ‘Mechanistic understanding of Li dendrites growth by in- situ/operando imaging techniques’ I found some great info on what this means.

The solid electrolyte interphase (SEI) spontaneously formed at the interface of the lithium (Li) metal and electrolyte is not mechanically stable enough to accommodate the high volume changes of the lithium anode during repeated lithium deposition/dissolution leading to the localized formation of lithium dendrites and excess consumption of fresh lithium and electrolyte. In addition, during prolonged cycling, the lithium dendrites with narrow roots can easily detach from the electrically conducting deposition substrate and form electrochemically inactive (dead) lithium. Under extreme circumstances, lithium dendrites can make internal short circuit between the electrodes and cause catastrophic phenomena and explosion of the cell.

SEI growth and Li dendrites
Schematic of the main challenges associated with Li metal anode: (a) short circuit induced by Li dendrites, (b) formation of inactive (dead) Li, and (c) development of thick and mechanically unstable SEI.

8. Did my capacity fade or did I not use it?

The consumption in the last leg was 156 Wh/km. Our overall trip to and in Sweden was 1678 km and we used 308 kWh so that is 183 Wh/km. That is very much in line with what I expected at those average speeds. Longest leg we did was 114 km using 18 kWh so that was only 158 Wh/km. In terms om voltages and state of charge that was from average 4,0V per cell (384V) and 95% SOC to 3,6V per cell (346V at pack level) and 46% SOC. At that point we would see undervoltage warnings which were triggered if one cell(group)was below 3,4V. So while the consumption was spot on with my estimate of 180 Wh/km the usable capacity was not. Eight 5,3 kWh Tesla modules should give us a usable capacity of around 40 kWh and thus a range of 200+ km. However, I charged only up to 4,05V per cell. So the question is, is the capacity not used or not usable?

Usable and used capacity of a Tesla Model S battery module

Below you see a Tesla Model S battery module discharge graph made by Jack Rickard of EVTV.

Tesla Model S battery module discharge curve
Source: EVTV Motor Verks Tesla Model S Battery Module Charge and Discharge Curves on Youtube, June 5th 2018.

It’s an interesting graph. As you can seen the graph goes down quite steep on the right hand side. There only is 8 Ah of energy between 3,3V down to 3,0V per cell. So from 3,3V per cell onwards the voltage will drop quickly and there is not much use going that low.

When starting at 4,05V per cell there would be about 20 Ah unused capacity. So per module that is 0,6 kWh (in total 5,3 kWh). The above graph already anticipates a usable capacity of 4,88 kWh instead of the full 5,3 kWh for which these modules are known.

So in my case when going from 4,05 to 3,3V per cell there should be a usable capacity of 4,2 kWh per module so in total 33,7 kWh of usable energy.

This was also what the BMS indicated when fully charged.

Based on the actual average consumption my EV Peripherals controller calculated a range based on that. Taking some margin in consideration.

A usable remaining energy of 34 kWh would give the 153 km of range when using 222 Wh/km. But we used much less, on average 180 Wh/km so it should be able to do more.

Usable capacity 34 kWh

However, back in The Netherlands we already found out the hard way that that certainly was not the case. On a test drive on May 30th we experienced this a couple of times. Good news is that back then the logging still worked so I could analyze that data and can share some graphs.

Already at 39% SOC and 15 kWh remaining energy we had low cell voltage warnings so at least one cell group was dipping below 3,4V.

Undervoltage warnings at 36 percent SOC

On our way back we expected we’d make it across the Dijk Lelystad – Enkhuizen but we didn’t, again due to undervoltage warnings. So due to voltage sag under load, it was not possible to get the last 10 or 12 kWh out.

After being dropped off at a fast charger we could charge and resume our trip. But again, much less usable range due to undervoltage warnings. As you can see in the image below on the left, that already started around 3,6V per cell. Under load there was a big difference in the amount of voltage sag between cells. The delta was up to 270 mV in this example.

At IDLE the balance was very good, but under load it wasn’t. As a result the usable capacity was not 33,7 kWh but rather around 20 kWh. So it seems some cells had a hard time keeping their voltage up under load. Unfortunately I did not log which string reported the lowest voltage.

At that point I did not realize, know and understand what I know now. Perhaps the fact that I had a much lower range should have triggered some alarm bells. Final indicators I’d like to discuss are temperature and after that internal resistance.

9. Pack temperature aka Rapidgating

When the tow truck dropped us off at a fast charger we did two charging sessions of approximately 30 minutes in total at 80A and thus a C-rate of 0,67. From the log I have the following data.

ParameterSession startSession endDelta
Battery coolant temperature22 °C30 °C8 °C
Exterior temperature21 °C21 °C0 °C
Highest temperature in the pack31 °C39 °C8 °C

We charged 14 kWh and since the battery pack was getting quite warm we decided to drive a bit again first. With an exterior temperature of now around 19 °C that brought the battery coolant temperature back to 23 °C again. The highest temperature of the pack remained high at 34 °C. We had to charge once (again with 80A, 0,67C) more to get home, now with the following data.

ParameterSession startSession endDelta
Battery coolant temperature23 °C28 °C5 °C
Exterior temperature19 °C20 °C1 °C
Highest temperature in the pack34 °C36 °C4 °C

We only charged 6 kWh as the BMS started to taper the current and we just got sufficient to get home.

So what can we conclude from this temperature data? Temperature increases because of heat generation because of internal resistance. So let’s have a look at that.

10. Internal resistance perspective

Unfortunately I do not have any internal resistance data, but what is interesting to study is the temperature increase given the amount of energy that went into the battery.

While building the car and designing the cooling and heating system I did some calculations on the energy needed to increase the pack temperature by one degree and during bench tests these turned out to be fairly accurate.

Electric Volvo Amazon cooling system

For now I ignore the temperature increase of the coolant and the energy dissipated by the radiator and just use the 8 °C increase in the session of 30 minutes and 4 °C increase in the session of 15 minutes.

According to my calculations the average internal resistance of the cells must have been around 45 mOhm. Knowing some heat has been dissipated, in real this is more. So how much more?

Some calculations I did earlier on Newtons Law of Cooling and the temperatures shown above the amount of heat dissipated via the radiator is around 1% of the amount of heat needed for the elevated temperatures I saw. So I think we can assume the order of magnitude of the 45 mOhm average internal resistance is about OK.

Though I originally anticipated it would be around 34 mOhm, that 45 mOhm does not sound alarming. In the post ‘Model X 100D Battery Internal DC Resistance‘ on teslamotorsclub.com someone reports 37 mOhm at cell level.

So that raises the question, is the internal resistance all good then?

Ohms law to the rescue

Let’s look a bit closer at the undervoltage warnings. We can use Ohms law to understand what is happening under load. Ohm’s Law states that the voltage drop across a resistor (in this case, the internal resistance) is equal to the current flowing through it multiplied by its resistance. Mathematically, it can be expressed as:

\[V_{drop} = I_{load} \times R_{internal}\]

So let’s compare the internal resistance of the weakest string and the healthiest string.

 Cell voltage minCell voltage max
No load voltage3709 mV3728 mV
Load voltage3225 mV3498 mV
V_drop484 mV230 mV
Load current (pack)290 A290 A
Load current (per cell)7,8 A7,8 A
Internal resistance62 mOhm29 mOhm

The above table shows there is a significant difference between internal resistances within the pack.

And since the overall link is as strong as the weakest link this means that these undervoltage warnings may be triggered because of one or some of the 96 strings of 37 cells in parallel each underperformed. It might even be that within those 37 cells that are in parallel there are truly weak ones and better ones. I cannot see from the logs which CMU or CMU’s report the cell(group) with the lowest voltage. I also do not have log data of how all 96 string voltages behaved under load so I’m not able to tell how widespread the problem is.

Voltage sag and sag delta under load

An interesting question still is, why are there such internal resistance differences at all? One would expect aging and degradation due to storage would be equally distributed throughout the pack. Maybe some coolant was spilled onto the module during pack teardown? Or a short circuit occurred during the modification?

11. Wrapping it all up….

There are still uncertainties and many of them will always remain. One thing that it still on my to do list is to find out how a Tesla cell (or perhaps a generic 18650 li-ion cell) will respond to the Evans waterless coolant.

By building on all information I gathered and shared in this blogpost I tend to conclude an external short circuit has not been the root cause of my electric car fire and thus most likely it has been one weak cell that went thermal.

A bit more on why I tend to draw that conclusion.

Emptying the burned front battery box

12. Conclusion

Due to all safety measures that were in place such as disconnects when off, pressure tested cooling system, non conductive coolant, fuses, isolation monitoring and a decent BMS in which no relevant errors were present I don’t think an external short circuit has happened.

That leaves an internal cell short circuit as the most likely root cause of my electric car fire. Of course that is the ‘easy answer’, but here is why I believe this is plausible.

  • Usable capacity was much lower than anticipated due to undervoltage warnings so there was degradation
  • Cells have been in storage for a long time before I started to use them again and from literature it is known cells degrade faster when stored
  • Internal resistance of at least one cell (group) was twice as high compared to the healthiest cell (group)
  • The higher internal resistance of that cell(group) makes it even worse as it causes extra local heat degrading that group even faster / further
  • Higher internal resistance can be an indicator for solid electrolyte interphase (SEI) growth and dendrites formation
  • Dendrites can cause internal cell short circuits
  • An internal cell short circuit may occur very suddenly and without previous warning

If there is anything you want to comment, add or see things differently, please let me know and contact me or post a comment below.

13. Epilogue : Lessons learned and prevention

Final question perhaps is: with the knowledge and information I have today would it have been possible to prevent the fire? Did I miss things? And also interesting to share what things I’m glad with I did.


Adding all up it seems heat and higher state of charge (can) lead to more stress within a cell. So if there are internal damages, deterioration or the mentioned dendrites it is more likely that is can result in problems at higher state of charge and/or temperature. In my case 100% SOC was still only at 4,05V but even this could have been stressful given the condition of the weakest cell(s).

Pre-commissioning test

Having said that, perhaps doing a pre-commissioning test would have been a good idea. This could have been:

  • Bring the module to a safe area
  • Charge to 4,15 or even 4,17V
  • Monitor the module over time
  • Be prepared to take action if a cell heats up.
  • If nothing happens in the first couple of hours, leave it in the safe area for some more time
  • Introduce some heat into the module for example by discharging it with a load
  • Monitor over time
  • If all good, discharge to around 60%
  • Repeat for all modules

If the “stress at higher SOC” theory is correct then it is likely any internal defects might be disclosed already. Though this is all about increasing the chance to discover issues at an early stage. But is it also possible to identify (potential) problems during use?

Blind spots during usage?

There are two events that I noticed but did not trigger alarm bells to do further research and take action:

  1. Much lower usable capacity as an indicator there were bad cells in the pack
  2. Sudden increase in cell delta the evening before the fire

But it remains difficult to say if this could have prevented the fire. Perhaps if I did find which cell(group) was bad and I’d replaced it? This is under the assumption that indeed the bad group was in the front pack and triggered the fire.

And perhaps if I had switched on the car again and let the cooling pump run the evening before the fire the heat of the internal short that perhaps was developing could have been dissipated preventing the thermal runaway (though not likely given the nature of NCA).

Success factors

From the beginning I knew Tesla battery modules are relatively dangerous and I always treated them with respect and caution. Since I wanted to have as much time as possible in case a battery fire would occur following a car crash I wanted to build heavy duty boxes.

Both the front and the rear battery box were made out of 2 mm stainless steel. While it is more difficult to machine and weld compared to aluminum or steel, in terms of melting temperature it is superior.

I added plastic vent valves taken from a Tesla model S battery pack as wel as a breather valve. You can see the vent valves on the picture on the right, all the way towards the front of the box on the right and left hand side.

That’s where the flames and heat could get out. Perhaps pointed towards the ground would have been even better.

But the fact that the box was sturdy, fully sealed and could vent at the front is probably the reason there now is a car to rebuild.

Heavy duty battry boxes from stainless steel
Front battery box scrapped

The front battery box contained the fire very well. All three Tesla modules burned up until there was nothing left to burn inside. Kudos to the firemen who cooled the outside of the box, otherwise the impact probably would have been bigger. While being prepared that the car burned down to the ground when it started I’m happy it did not. Even the front tires still held (and hold) air.

For my rebuild I will not be using Tesla battery modules for sure. Nowadays there are plenty of other options that give me a high voltage setup of 96s at a very reasonable volume, weight, energy density, quality and availability. Biggest challenge is to find time and resources. The car has been dry-ice cleaned and is waiting inside again.

Donate and support my rebuild

It turned out the Whydonate platform I use does not (temporarily did not?) facilitate (international) Paypal payments.
So if you cannot donate using the form in the sidebar and want to support me, please use my personal Paypal.me page or my Paypal account lars@rengersen.nl.

14. Acknowledgments

There are many people I talked and exchanged with over the last two years who have contributed to the insights I have today that allowed me putting together this story. If your name is not on the below list and we did speak, sorry for that and thanks!

But I do want to explicitly name and thank Joost, Steven, Martijn, Perttu, Roeland, Ruben, Mischa, Minos, Daniel, Jurry and authors of the blog posts, reports and papers I used. Thanks to my family, Saskia, Sven and Jiri for giving me the support and time to do this. And thanks to everybody who supported me back in 2021 when it had just happend!

If there is any news regarding the test I still want to do, I’ll update this blogpost. But from now on, I’ll focus on the rebuild, stay tuned!

15. Addendum : key insights based on feedback

First of all, many thanks for all the responses and feedback on this post. There is one response that was particularly valuable. I was able to link an event that I had not linked before.

Someone shared his case where he accidentally shorted two LiFePO4 cells when assembling a pack. He was able to remove the short circuit before things got out of hand. He mentioned that later he did see higher internal resistances and more heat development for those cells and these stayed the ‘weaker cells’. 

It is too long ago since the modification on my Tesla modules was done and there is nothing left to study and research but I tend to believe a similar situation perhaps has happened when the cell tap studs were added. A wrong connection was made and the stud welder was damaged. This can have been a similar short circuit that caused internal damage to cells. It’s a bit far fetched since I wasn’t there and don’t have enough details to get beyond educated guessing. But it can be an explanation for the differences in internal resistance in my case as well.

Another story that reached me was also interesting. One EIG cell in a custom pack was discharged to 2V (so below the minimum of 2,5V). It was recovered by slowly charging that cell. All seemed to work out fine. However two years later that cell caught fire. In that case luckily in an isolated and contained way so the car could even still drive.

The two above stories indicate that events that seem to be innocent or just went well can have an impact at longer term.

Main takeaway for me is that it is likely to believe that the cause of the fire in my electric car was less ‘random’ that I assumed so far.

Leave a Comment


OldVolvo is a classic Volvo hobby blog by Lars Rengersen.

Also electric?

Do you also want to convert your car (Volvo or another brand) to full electric? As a spin-off of my own EV conversion project I started EVcreate for 'enabling EV conversions'

More info www.evcreate.com