ArticlePDF Available

Safety of Grid Scale Lithium-ion Battery Energy Storage Systems

Safety of Grid Scale Lithium-ion Battery Energy Storage Systems
EurIng Dr Edmund Fordham MA PhD CPhys CEng FInstP
Fellow of the Institute of Physics
Dr Wade Allison MA DPhil
Professor of Physics, Fellow of Keble College, Oxford University
Professor Sir David Melville CBE FInstP
Professor of Physics, former Vice-Chancellor, University of Kent
Sources of wind and solar electrical power need large energy storage, most often provided by
Lithium-Ion batteries of unprecedented capacity.
Incidents of serious fire and explosion suggest that the danger of these to the public, and
emergency services, should be properly examined.
5 June 2021
– 2 – June 5, 2021
Executive Summary
1. Li-ion batteries are dominant in large, grid-scale, Battery Energy Storage Systems (BESS) of
several MWh and upwards in capacity. Several proposals for large-scale solar photovoltaic (PV)
“energy farms” are current, incorporating very large capacity BESS. These “mega-scale” BESS
have capacities many times the Hornsdale Power Reserve in S. Australia (193 MWh), which was
the largest BESS in the world at its installation in 2017.
2. Despite storing electrochemical energy of many hundreds of tons of TNT equivalent, and several
times the energy released in the August 2020 Beirut explosion, these BESS are regarded as
“articles” by the Health and Safety Executive (HSE), in defiance of the Control of Major Accident
Hazards Regulations (COMAH) 2015, intended to safeguard public health, property and the
environment. The HSE currently makes no representations on BESS to Planning Examinations.
3. Li-ion batteries can fail by “thermal runaway” where overheating in a single faulty cell can
propagate to neighbours with energy releases popularly known as “battery fires”. These are not
strictly “fires” at all, requiring no oxygen to propagate. They are uncontrollable except by
extravagant water cooling. They evolve toxic gases such as Hydrogen Fluoride (HF) and highly
inflammable gases including Hydrogen (H2), Methane (CH4), Ethylene (C2H4) and Carbon
Monoxide (CO). These in turn may cause further explosions or fires upon ignition. The chemical
energy then released can be up to 20 times the stored electrochemical energy. Acute Toxic
gases and Inflammable Gases are “dangerous substances” controlled by COMAH 2015.
Quantities present “if control of the process is lost” determine the applicability of COMAH.
4. We believe that the approach of the HSE is scientifically mistaken and legally incorrect.
5. “Battery fires” in grid scale BESS have occurred in South Korea, Belgium (2017), Arizona (2019)
and in urban Liverpool (Sept 2020). The reports into the Arizona explosion [8, 9] are revelatory,
and essential reading for accident planning. A report into the Liverpool “fire” though promised
for New Year 2021, has not yet been released by Merseyside Fire and Rescue Service or the
operator Ørsted; it is vital for public safety that it be published very soon.
6. No existing engineering standards address thermal runaway adequately, or require measures
(such as those already used in EV batteries) to pre-empt propagation of runaway events.
7. Lacking oversight by the HSE, the entire responsibility for major accident planning currently lies
with local Fire and Rescue Services. Current plans may be inadequate in respect of water
supplies, or for protection of the local public against toxic plumes.
8. The scale of Li-ion BESS energy storage envisioned at “mega scale” energy farms is
unprecedented and requires urgent review. The explosion potential and the lack of engineering
standards to prevent thermal runaway may put control of “battery fires” beyond the
knowledge, experience and capabilities of local Fire and Rescue Services. BESS present special
hazards to fire-fighters; four sustained life-limiting injuries in the Arizona incident.
9. We identify the well-established hazards of large-scale Li-ion BESS and review authoritative
accounts and analyses of BESS incidents. An internet video [10] is essential initial instruction.
10. We review engineering standards relating to Li-ion BESS and concur with other authorities that
these are inadequate to prevent the known hazard of “thermal runaway”. We conclude that
large-scale BESS should be COMAH establishments and regulated appropriately. We respectfully
request evidence from the HSE that “mega-scale” BESS are not within the scope of COMAH.
11. We seek the considered response of relevant Government Departments as well as senior fire
safety professionals to these concerns.
– 3 – June 5, 2021
Executive Summary p 2
1. Introduction p 4
2. Leading concerns p 10
3. Thermal runaway (Battery “fires”) p 11
4. Toxic and flammable gas emissions p 14
5. Total energy release potential p 15
6. Applicability of the COMAH (Control of Major Accident Hazard) Regulations 2015 p 17
7. Engineering standards for BESS p 18
8. Fire Safety Planning for BESS “fires” p 19
References p 22
Appendix 1: Battery capacity calculations for grid-scale BESS at “Sunnica” p 24
Appendix 2: Applicability of the COMAH Regulations to large-scale BESS p 26
Appendix 3: Shortcomings of existing engineering standards for large-scale BESS p 29
Appendix 4: Fire Safety Planning in the Councils’ Response p 30
– 4 – June 5, 2021
1. Introduction
Lithium-ion (Li-ion) batteries are currently the battery of choice in the ‘electrification’ of our
transport, energy storage, mobile telephones, mobility scooters etc. Working as designed, their
operation is uneventful, but there are growing concerns about the use of Lithium-ion batteries in
large scale applications, especially as Battery Energy Storage Systems (BESS) linked to renewable
energy projects and grid energy storage. These concerns arise from the simple consideration that
large quantities of energy are being stored, which if released uncontrollably in fault situations could
cause major damage to health, life, property and the environment.
Table 1. Comparison of some recent “battery fires” since 2014.
Note: this is not a comprehensive list of all Li-ion BESS battery fires.
“Battery fire” cause
Time to
bring under
for cooling
Texas, April
Driverless vehicle crash
4 hours
30,000 (US)
Tesla Model S
South Korea
21 fires
Not known to Korean
Ministry of Trade Industry
and Energy
Not known
522 out of 1490
ESS facilities in
Korea suspended
(Korea Times 2
May 2019)
Belgium. 2017
1 MWh
Not known.
Not known
Occurred during
commissioning of
system by ENGIE
Arizona, USA.
2 MWh
Thermal runaway in a
single rack out of 27 that
were in the cabin hence
74 kWh electrochemical
energy released less
than the Tesla Model S
2 hours from
first report to
Explosion as H2
and CO mixed
with air and
ignited. Critically
injured 4 fire-
fighters. Extensive
forensic report.
Carnegie Rd,
Liverpool, UK,
Not known
11 hours
Full report [1]
delayed 4 months;
still unpublished.
Even battery electric vehicle (BEV) batteries store energy sufficient for “fires” that have taken
hours to control. A Tesla Model S crashed In Texas on the weekend of 17-18 April 2021 igniting a
BEV battery fire that took 4 hours to control with water quantities variously reported [2] as 23,000
(US) gallons or 30,000 gallons (87 -115 m3). Yet the energy storage capacity in even the latest Tesla
Model S vehicles is only 100 kWh. This is 1/20 the size of the BESS in Arizona [3] which failed in
2019, and 1/200 the size of the BESS in Liverpool [4] which caught fire [5] in September 2020, and
1/7000 the capacity of the Cleve Hill Solar Farm and Battery Store [6] approved in May 2020.
The past decade has seen a number of serious incidents in grid-scale BESS, which are
summarised in Table 1. Despite these incidents, and our growing understanding of these, these
large scale Li-ion BESS are not currently regarded by HSE as regulated under the COMAH
– 5 – June 5, 2021
Regulations 2015. The legal basis for this attitude is unclear simple calculations summarised in this
paper argue that they should be and the issue may yet be challenged in judicial review.
The reason the COMAH regulations should apply is the scale of evolution of toxic or
inflammable gases that will arise in BESS “fires”. In the Drogenbos incident (2017, Table 1), the
inhabitants of Drogenbos and surrounding towns were asked to keep all windows and doors shut;
50 emergency calls were made from people with irritation of the throat and airways
. A chemical
cloud which “initially had been enormous”, was charted by helicopter. The Belgian Fire Services
could not control what was described as “the chemical reaction” and filled the cabin with water.
Fears of an explosion with 20 metre flames kept people confined for an hour. Although the initial
visible flames were controlled quickly, cooling continued over the next 36 hours.
Figure 1: Remains of
the Tesla Model S
crash and fire, 17 Apr
2021, after 4 hours
and 30,000 gallons.
Battery capacity 100
Figure 2: Remains of a
Korean BESS destroyed
by a “battery fire”. An
energy storage system
was destroyed at the
Asia Cement plant in
Jecheon, North
Chungcheong Province,
on Dec. 17. Courtesy of
North Chungcheong
Province Fire Service
Headquarters (Korea
Times 2 May 2019)
Tom Vierendeels (2017) “Explosiegevaar by brand in Drogenbos geweken : 50-tal oproepen van mensen die zich
onwel voelen door rook.” Het Laatste Nieuws, 11 November 2017
– 6 – June 5, 2021
Figure 3: “Battery Fireat Drogenbos, Belgium 11 Nov 2017. Taken at the start of the incident and
15 minutes later (eye-witness footage). 1 MWh facility; fire occurred during commissioning.
Figure 4: The 2 MWh McMicken (Arizona) BESS after the explosion on 19 April 2019
– 7 – June 5, 2021
Figure 5: The 20
Carnegie Rd,
Courtesy Ørsted.
Figure 6: The fire at Carnegie Road, 15 Sep
2020. Liverpool Echo report, which took 11
hours to control.
– 8 – June 5, 2021
The incidents recorded in Table 1 are all in relatively small BESS or a single BEV. Yet “mega-
scale” BESS are now planned on a very large scale in many current proposals in the UK, listed in
Table 2 and illustrated in the subsequent Figures.
And no engineering standards are currently applied to pre-empt future accidents in grid-scale
BESS, the most critical of which would be design features aimed at preventing the phenomenon of
“thermal runaway”, the process whereby failure in single cell causes over-heating and then
propagates to neighbouring cells so long as a temperature (which can be as low as 150 °C) is
BEV batteries do now include thermal barriers or liquid cooling channels between all cells to
safeguard against this phenomenon, but no such engineering standards exist for grid-scale BESS. A
large BESS can pass all existing engineering design and fire safety test codes and still fail in thermal
runaway by now a well-known failure mode. This must be urgently addressed.
The consequences of major BESS accidents could be significant and emergency services need
adequate plans in place to handle any such incident.
Table 2. “Mega” scale solar plant and/or Li-ion BESS in Australia and the UK*
Solar PV
Scheme Size
S. Australia
Not directly
Single site
193 MWh
Cleve Hill
Solar +
350 MW; land
coverage 890
Single site
700 MWh
Solar +
500 MW; land
approx. 2792
31.5 ha of land
over 3
compounds [7]
of 5.2, 10.7 and
15.6 ha
3000 MWh
Solar +
500 MW; land
approx. 1400
Stated as 3.7
acres: number
of sites TBD
150 MWh
* Li-ion technology has been assumed in all these proposals as Li-ion battery electrochemistry is
dominant in grid-scale BESS applications (deployment at this scale is unlikely to involve
technologies with lesser experience). Estimated values for Battery Capacity for the Sunnica are
calculated based on the McMicken facility in Arizona (Appendix 1) and the Cleve Hill DCO. For the
Longfield site it is estimated from Energy Institute guidance on energy density [25] at about 100
MWh ha-1. The exact specification for the battery units has not been disclosed by the developers at
this present time.
– 9 – June 5, 2021
Figure 7: The Hornsdale Power
Reserve (South Australia) in the
process of expansion from 100
MW/129 MWh to 150 MW/193.5
MWh, as of November 2017.
Figure 8: a “typical” BESS
compound (abstracted from
Sunnica PEIR, Ch 3)
Figure 9: Artists impression
of Tesla 250 MWh
Sunnica may have 3 ´this
capacity in just one of its
three BESS compounds.
10 June 5, 2021
2. Leading Concerns
The main concerns regarding large scale Li-ion BESS are:
1) The potential for failure in a single cell (out of many thousands) to propagate to neighbouring
cells by the process known as “thermal runaway”. Believed to be initiated by lithium metal
dendrites growing internally to the cell, a cell may simply discharge internally releasing its
stored energy as heat. Even sound Li-ion cells will spontaneously discharge internally if heated
to temperatures which can be as low as 150 °C, releasing their stored electrical energy, thus
overheating neighbouring cells and so on. Temperatures sufficient to melt aluminium (660 °C) at
least have been inferred from analyses of such thermal runaway accidents. Eye-witness reports
consistently speak of repeated “re-ignition” which is inevitable, even in the complete absence of
oxygen, so long as the temperature anywhere exceeds the thermal runaway initiation threshold.
2) The emission of highly toxic gasesprincipally Hydrogen Fluoride for prolonged periods, in
the event of thermal runaway or other battery fires. At a minimum, respirators and complete
skin protection would be required by any fire-fighters. Measures to protect the public from toxic
plumes would also be necessary.
3) The emission of large quantities of highly inflammable gases such as Hydrogen, Methane,
Ethylene and Carbon Monoxide even if a fire suppression system is deployed. These gases will
be evolved from a thermal runaway accident regardless of such measures, with explosion
potential as soon as they are mixed with air and in contact with hot surfaces. Such an explosion
was the cause of the “deflagration event” at McMicken, Arizona in 2019 in a 2 MWh BESS,
which critically injured four fire-fighters and was triggered simply by opening the cabin door.
4) The absence of any adequate engineering and regulatory standards to prevent or mitigate the
consequences of “thermal runaway” accidents in Li-ion BESS.
5) The potential for thermal runaway in one cabin propagating to a neighbouring cabin. In Arizona
[3] there were reports of “fires with 10-15 feet flame lengths that grew into 50 - 75 feet flame
lengths appearing to be fed by flammable liquids coming from the cabinets”.
6) The significant volumes of water required to thoroughly cool the system in the event of a “fire”,
and how this water will be contained and disposed of (since this will be contaminated with
highly corrosive hydrofluoric acid and, therefore, must not be allowed to drain into the
surrounding environment).
Such incidents are routinely and repeatedly described in the Press as “battery fires” though they
are not “fires” at all in the usual sense of the word; oxygen is completely uninvolved. They
represent an electrochemical discharge between chemical components that are self-reactive. They
do not require air or oxygen at all to proceed.
Hence the traditional “fire triangle” of “Heat, Oxygen, Fuel” simply does not apply, and
conventional fire-fighting strategies are likely to fail (Figure 10, over).
Thermal runaway events are uncontrollable except by cooling all parts of the structure affected
even the deepest internal parts below 150 °C. This basically requires water, in large volumes.
11 June 5, 2021
Figure 10: The traditional “fire triangle” does not apply to “thermal runaway”.
3. Thermal Runaway (Battery “fires”)
Li-ion batteries are sensitive to mechanical damage and electrical surges, both in over-charging
and discharging. Most of this can however be safeguarded by an appropriate Battery Management
System (BMS) and mechanical damage (unless deliberate and malicious) should not be a hazard.
Internal cell failures can arise from manufacturing defects or natural changes in electrodes over
time; these must be regarded as unavoidable in principle. Subsequent escalation into major
incidents can propagate from such apparently trivial initiation.
In July 2020 a thorough failure analysis by Dr Davion Hill of DNV GL [8] was prepared for the
Arizona Public Service (APS), following the April 2019 thermal runaway and explosion incident in
the 2 MWh Li-ion BESS facility at McMicken, Arizona. This report is revelatory and more detailed
than any other failure analysis known to us. It is essential reading for any professional involved in
fire safety planning for major BESS. (Figures 11 to 13).
Figure 11: Cells
stack into
Modules into
Racks; Racks into
Strings; Strings
into Systems.
Propagation of
single Cell
failure through
cascade to
entire Rack.
12 June 5, 2021
A report by Underwriters Laboratories (UL) on the same incident [9] is less technical on the
physics and engineering of the underlying causes and failure modes, but more comprehensive in
terms of practical situations and consequences found, and suffered, by the “first-responders”. Two
fire-fighters suffered life-limiting brain injuries, one suffered spinal damage and fourth facial
lacerations. This report is similarly essential reading for any fire and emergency response planning.
Figure 13: Destruction of Rack at McMicken.
Detail: molten aluminium pools (exceeded 660 °C)
Forensic analysis [8] of the 2019 Arizona “fire”
identified a failure mode different from mechanical
abuse or electrical mis-management. The initiating
failure was localised to a single cell at a known
position in the rack. Although the cell itself was of
course destroyed during the incident, the failure
mode is believed to have been lithium metal deposition and abnormal growth of lithium metal
dendrites. These phenomena were also found in randomly selected undamaged cells from the
same BESS and also from a different BESS of the same manufacture elsewhere. These phenomena
must be regarded as common, and inherent to the cells themselves.
The lithium metal deposits will react with air moisture, causing overheating and smoke. Battery
swelling, electrolyte degradation, and internal short circuits are all possible modes of failure with
internal discharge and generation of locally intense heat.
Because of the known thermal breakdown of even non-faulty cells, above a threshold
temperature (which can be as low as 150 °C), the loss of even a single individual cell can rapidly
cascade to surrounding cells, resulting in a larger scale fire.” This is “thermal runaway” in which
failures propagate from cell to cell within “modules” and from module to module within a “rack”.
This is what happened at McMicken [8], with temperatures sufficient to melt Aluminium (660
°C) being reached. Such “fires” can be extremely dangerous to fire fighters and other first
responders because, in addition to the immediate fire and explosion risks, they would have to deal
with toxic gases (principally hydrogen fluoride HF, also hydrogen cyanide HCN and other fluorine
compounds such as phosphoryl fluoride POF3) and exposure to other hazardous materials.
Rack to rack propagation fortunately did not happen at McMicken, though an explosion did [8].
A local conventional fire involving the plastics materials or gases evolved from them could have
13 June 5, 2021
initiated rack-to-rack propagation; the only essential factor would have been sufficient heat to
trigger thermal breakdown in just one cell in a neighbouring rack. Li-ion cells have been observed to
eject molten metal during thermal runaway, another possible mode of propagation over distance.
Propagation through a subsequent rack would then occur by exactly the same thermal runaway
mechanisms, and potentially beyond between neighbouring cabins in large-scale BESS.
Thermal runaway is illustrated in dramatic fashion with tiny commercial Li-ion cells in a useful
internet video [10] (Figure 14). The commercial cells involved in this demonstration have tiny
capacities: a mere 2.6 Ah or about 10 Wh for typical terminal voltages.
A Tesla Model S would have the capacity of about 10,000 such cells.
A 20 MWh BESS has the capacity of about 2 million such cells.
In the video, the cell is deliberately over-heated on a small electric stove. The fully charged cell
goes into thermal breakdown, eventually rupturing the can. The cell flies off as a rocket and
seconds later is discharged but red hot and will burn anything combustible. Although not
illustrated, it is evidently hot enough to produce the same thermal breakdown in an adjacent cell
within a battery.
This illustrates the damage done to a non-faulty cell, simply by overheating externally.
Figure 14: (a) A
charged 2.6 Ah cell
being deliberately
overheated. (b) at
the point of rupture
(c) the cell takes off
as a rocket (d)
seconds later the
discharge is
complete, and the
cell is red hot.
14 June 5, 2021
4. Toxic and flammable gas emissions
During a Li-ion “battery fire,” multiple toxic gases including Hydrogen Fluoride (HF) [11],
Hydrogen Cyanide (HCN) [13] and Phosphoryl Fluoride (POF3) [11] may be evolved. The most
important is Hydrogen Fluoride (HF), which may be evolved in quantities [11] up to 200 mg per Wh
of energy storage capacity.
HF is toxic in ppm quantities and forms a notoriously corrosive acid (Hydrofluoric Acid) in
contact with water. It is toxic or lethal by inhalation, ingestion and by skin contact. The ERPG-2
concentration (1 hour exposure causing irreversible health effects) given by Public Health England is
just 20 ppm; the workplace STEL (15 minute Short-Term Exposure Limit) is just 3 ppm [12]. Major
emissions of HF would form highly toxic plumes that could easily threaten nearby population
centres, workplaces and schools.
Appendix 3 contains calculations of projected toxic gas quantities for 3 grid-scale battery stores
that have been approved or are pending review by the Planning Inspectorate (Table 2).
The calculated capacities at the “mega-scale” sites listed in Table 2 are tens, or even hundreds,
of times larger than the facilities in Table 1, which experienced significant fires or explosions.
In addition to evolution of toxic gases, even in an inert atmosphere (without Oxygen), multiple
flammable gases (such as Hydrogen H2, Carbon Monoxide CO, Methane CH4, and Ethylene C2H4)
would be evolved during thermal runaway. These are “typical of plastics fires” [8] and have been
measured in sealed vessel tests [13]. As noted by Hill/DNV [8] and others [13], the proportions of
H2, CO, CH4 and C2H4 do not in fact vary greatly between different cell technologies, simply because
the chemical nature of the envelope polymers, separators, electrolyte solvents and electrolytes
themselves do not differ greatly. The variations between Li-ion technologies are in the electrode
systems, which are typically not polymeric.
Such inflammables can clearly create (ordinary, air-fuel) fires or explosions once mixed with
air/oxygen. It is important to note that the Heats of Combustion of the inflammables may be up to
15 20 ´ the rated electrical energy storage capacity of the BESS. This has been demonstrated by
the same tests which determined the quantities of HF evolved [11]. These were fire tests, not
sealed vessel tests [13]. The stored electrical energy is therefore by no means a conservative
estimate of the total energy release which could be released in a major (air-fuel) fire in a BESS,
irrespective of whether the initiating cause was a conventional fire or Li-ion cell thermal runaway.
Appendix 2 estimates the inflammables potentially evolved from the BESS given in Table 2.
15 June 5, 2021
5. Total Energy Release Potential
Any large energy storage system has the risk that energy released in malfunction will be
uncontrollable in ways that will do major damage. BESS can release electrochemical energy in the
form of thermal runaway or “battery fires”. In addition they can release chemical energy in the
form of explosions or conventional fires of inflammable gases, or of polymer components. Many
thermal runaway “fires” have now happened, as has explosion of evolved inflammable gases.
An important indicator of the foreseeable scale of a “worst credible hazard” is provided by the
total stored energy in the system. For BESS, this comprises two components:
(i) The stored electrical energy which might be released in the event of thermal runaway incidents,
a self-reactive electrochemical energy release not requiring oxygen at all, and
(ii) Stored chemical (fuel) energy which might be released in complete combustion of the
inflammable gases which might be released by (i).
Electrochemical energy release is uncontrollable once started, by any measure except cooling
of all cells and cell parts below about 150°C. Water is the only fire-fighting substance with the
necessary heat capacity. Concurrent conventional fire would first heat cells above the thermal
runaway temperature, causing more thermal runaway. Chemical energy release from inflammable
gases is also uncontrollable once those gases are mixed with air and ignited: explosions result.
What might be the scale of such energy releases? The Sunnica proposal is estimated to have a
stored energy between 1.5 3.0 GWh in total, spread across 3 separate sites called Sunnica East A,
Sunnica East B and Sunnica West A (see calculations in Appendix 1). It is between 2 4 times the
capacity projected for Cleve Hill (700 MWh). It is 8 15 times the capacity (193 MWh) of the
“Hornsdale Power Reserve” in Australia, at installation (2017) the world’s largest.
Compared to other energy storage technologies, the Dinorwig Pumped Storage Scheme in
Snowdonia stores about 9 GWh [14]; the Sunnica BESS corresponds to 17 33 % of Dinorwig.
Compared to major explosions, the energy released in the Beirut warehouse explosion of
August 2020 has been estimated [15] by Sheffield University at about 0.5 kilotons of TNT (best
estimate) with a credible upper limit of 1.12 kilotons. A totally independent estimate [16] (based on
seismic propagation instead of eye-witness footage) gives the same range, without specifying a
“best” estimate. The popular measure of major explosions in “kilotons of TNT” has an agreed
of 1.162 GWh of released energy; in this paper we shall take “one Beirut” to be an
explosive energy of 0.5 kilotons of TNT or about 580 MWh of released energy.
The projected BESS storage at Sunnica corresponds to 1.4 2.7 kilotons of TNT in total, across
all three sites. In the “low” case, this would be0.92 Beiruts” at the Sunnica West A site alone, or
“2.7 Beiruts” over the whole scheme. In the “high” case “2.7 Beiruts” could be stored in the Sunnica
East B site alone. Note that these are stored electrochemical energy only; the potential for
conventional fire or explosion of evolved inflammables could be up to 20
larger [11]. See Table 3,
Appendix 1.
This is plainly a quantity of stored energy which, if released uncontrollably, could do major
damage. Explosions and fires at individual BESS are matters of record. They can propagate from
failure in a single cell out of many thousands. Cell-to-cell and module-to-module propagation
occurred at McMicken. Rack-to-rack propagation was avoided, but could readily occur if continuous
See e.g. Wikipedia.
16 June 5, 2021
fires start. Cabin-to-cabin propagation of a major BESS “battery fire” would be the critical link that
would escalate major but manageable fires into catastrophes.
Yet this propagation route remains unanalysed. Significantly, Commissioner Sandra D Kennedy of
the Arizona State Commission [3] reviewed reports on the 2019 McMicken battery fire and also a
2012 battery fire at the APS Eldon substation facility in Flagstaff, AZ. She quotes the Flagstaff fire
department report on the latter incident as referencing :
Fires with 10-15 feet flame lengths that grew into 50 - 75 feet flame lengths appearing to be
fed by flammable liquids coming from the cabinets”.
Finally, in the context of BESS, “Stranded Energy” will remain a hazard at any affected BESS
cabins even assuming an initial incident is controlled. The accident investigation at McMicken
required nearly 3 months, simply to discharge “stranded energy” safely [8].
‘”Mega-scaleLi-ion BESS should, in all prudence, require the highest level of regulation. The
COMAH regulations are designed for this, including establishments where dangerous substances
may be generated “if control of the process is lost” [17] in a thermal runaway accident.
17 June 5, 2021
6. Applicability of the COMAH (Control of Major Accident Hazard) Regulations 2015
The governing criteria for application of the COMAH Regulations [17] are:
1. The presence of hazardous materials, or their generation, “if control of the process is lost.”
2. The quantity of such hazardous materials present or that could be potentially generated.
There is no doubt that hazardous substances such Hydrogen Fluoride (an Acute Toxic controlled
by COMAH) would be generated in a BESS accident (i.e., in “battery fires”). Similarly highly
Inflammable Gases (also controlled by COMAH) would be evolved even if the atmosphere remained
oxygen-free. Depending on the size of the “establishment” these could be produced in sufficient
quantities to be in the scope of COMAH. In Appendix 2 we estimate quantities guided by the
literature, where fire tests have directly measured evolution of the hazardous gases.
For small capacity BESS installations, under 25 MWh capacity, the quantities (“inventory”) of the
evolved hazardous substances might be outside COMAH. This paper however addresses the recent
trend towards “mega-scale” Li-ion BESS (Table 2) with very large quantities of stored energy, where
the inventory should be large enough to bring the installation within scope.
Broadly speaking, the threshold for applicability of COMAH will be dependent on the precise
BESS technology chosen, but likely to be for BESS in the region of 20 50 MWh. See Appendix 2.
A letter to the HSE regarding applicability of COMAH to large-scale BESS (dated 25 Nov 2020
[18]) received no reply until follow-up letters were sent addressed personally to the Chief Executive
on 7 February 2021, with the intervention of Mrs Lucy Frazer MP. The reply from the Chief
Executive [19] dated 22 February 2021 stated that “Li-ion batteries are considered articles and are
not in scope of COMAH”.
We believe the current attitude of the HSE that even large-scale Li-ion BESS are “articles” best
regulated by operators – is not consistent with the law.
Unless tested in the Courts however, this throws the entire responsibility for ensuring the safety
of major BESS battery fires” onto the Fire and Rescue Services. Currently the HSE makes no
representation to the Planning Inspectorate in respect of BESS hazards.
18 June 5, 2021
7. Engineering standards for BESS
As with any hazard, the basic principles of Prevention and Mitigation must be applied to minimise
the risk to life, property and the environment. A major contribution of the Hill/DNV report [8] is a
review of current engineering and fire protection standards. This did not concern planning, siting
and electrical standards, but simply addresses the question: which standards, if any, offer
Prevention or Mitigation of the phenomenon of thermal runaway? The answer appears to be none.
“Thermal runaway” is an electrochemical reaction, well-known in Li-ion BESS, that is largely
uncontrollable once started. Since failures in single cells (among many thousands) can be sufficient
to initiate thermal runaway, the only known Prevention measure is that adopted by the BEV
industry, viz. thermal isolation of neighbouring cells, so that if failure occurs in any one cell,
insulation or water cooling prevents easy thermal spread to neighbouring cells. Various design
strategies have been adopted in BEV Li-ion batteries, usually involving some form of thermal
However these are not widely used in grid-scale Li-ion BESS. Current practice is the assembly of
stacks of cells, typically “pouch” cells which are externally flat polymer bags, that are stacked side
by side in low profile modules with no thermal isolation. This is not the construction adopted in
current generation BEV batteries; BEV practice (with thermal isolation) extended to grid-scale BESS
would obviously increase costs and complexity considerably.
The engineering standards reviewed by Hill/DNV [8] included NFPA 855, UL 1973 and UL
9540/9540A. UL 9540A is a US standard that is widely used in grid-scale BESS engineering, is
routinely recommended by insurance and risk consultants [20] and was appealed to by the
developer of the Cleve Hill solar farm (Table 2). The problem is that UL9540A is fundamentally a
test procedure. It mandates no design features. It requires absolutely nothing that would prevent
thermal runaway in any BESS design. This means that an operator can say truthfully that a given
BESS is fully compliantwith UL9540A, yet this would provide no assurances at all regarding
thermal runaway prevention. It is therefore wholly insufficient as a safeguard to either the
operator, the public, or to emergency services.
NFPA 855 [21], uniquely, requires evaluation of thermal runaway in a single module, array or
unit and recognises the need for thermal runaway protection. However, it assigns that role, with
complete futility, to the Battery Management System (BMS). Thermal runaway is an
electrochemical reaction which once started cannot be stopped electrically. It is uncontrollable by
electronics or switchgear. A BMS can locate faults, report and trigger alarms, but it cannot stop
thermal runaway.
The Hill/DNV report [8] highlights the many shortcomings of existing standards, see Appendix 4.
The basic issue is simple:
(1) Thermal Runaway has very few means of Mitigation once started.
(2) It is therefore essential to address Prevention as a priority.
(3) No current engineering or industry standards require the Prevention of thermal runaway
events by thermal isolation barriers.
Nothing in existing standards prevents runaway incidents happening again, requiring for initiation
only single-cell failures from known common defects in cell manufacture.
19 June 5, 2021
8. Fire Safety Planning for BESS “fires”
Taking the recent Sunnica BESS proposal as an example, a joint statutory consultation response
has been submitted by the four Local Authorities concerned. The Local Authorities in this case are
Cambridgeshire and Suffolk County Councils, and West Suffolk and East Cambridgeshire District
Councils. This joint consultation response [22] included a section on Battery Safety (pp 74-75) and
states as follows:
Suffolk Fire and Rescue Service (SFRS) will work and engage with the developer as this project develops to
ensure it complies with the statutory responsibilities that we enforce.
Sunnica should produce a risk reduction strategy as the responsible person for the scheme as stated in
the Regulatory Reform (Fire Safety) Order 2005. It is expected that safety measures and risk mitigation is
developed in collaboration with services across both counties.
The response also later states: As with all new and emerging practices within UK industry, the SFRS
would like to work with the developers to better understand any risks that may be posed and
develop strategies and procedures to mitigate these risks.
It is clear that local Fire and Rescue Services have been given the lead responsibility for
independent emergency planning, in concert with the developers. Because of the attitude of the
HSE refusing to exercise regulatory control over BESS safety, local Fire and Rescue Services become
the sole independent public body able to influence BESS safety issues at the planning stage.
Many detailed recommendations have been made by the Local Authorities in the case of
Sunnica. It is unclear how much opportunity or input Suffolk FRS has had in these. However the
recommendations offered betray some serious misunderstandings and a complete lack of
awareness of the lessons and recommendations made in publicly available documents such as the
Hill/DNV report [8] into the McMicken explosion.
These are taken point by point in Appendix 4 but some general points are made here.
1. Thermal runaway cannot be controlled like a regular (air-fuel) fire. The only way to mitigate re-
ignition” (a regular report of eye-witnesses) is by thorough cooling. Water is the only fire-fighting
material with the necessary thermal capacity. Sprinkler systems, though with good records in
conventional building fires, are likely to be completely inadequate. The purpose of the water is
absorbing a colossal release of energy. The Hill/DNV report [8] called for so-called “dry pipe”
systems allowing first responders to connect very large water sources to the interior without having
to access the interior.
It is critical to appreciate that all parts of the battery system must be cooled down. Playing
water on a battery “fire” may cool the surface, but so long as Li-ion cells deep inside the battery
remain above about 150°C, ”re-ignition” events will continue. It is not sufficient to estimate water
requirements on the basis of calculations assuming water reaches everywhere, uniformly.
For example, in the recent Tesla car fire [2] the BEV battery kept re-igniting, took 4 hours to
bring under control and used 30,000 (US) gallons of water (115 m3). This was for a 100 kWh BEV
battery, designed with inter-cell thermal isolation barriers.
In the case of Sunnica, the Local Authorities have suggested that water supplies of 1900 litres
per minute for 2 hours (228 m3) will be needed [22]. But this is grossly inadequate. Using the above
Tesla BEV fire experience, this amount of water would suffice for just two Tesla Model S car fires.
Scaling this up to even the smallest 2 MWh BESS (such as that in McMicken [8]), which contains
20 June 5, 2021
stored energy equivalent to twenty Tesla Model S cars, it is clear to see that a much greater
amount of water would be needed.
The actual amount of water required will depend on the energy storage capacity per cabin
which, in the case of Sunnica, is still unstated. Some simple estimates are, however, made below.
The requirements suggested to date by the Local Authorities for the Sunnica installation are
completely inadequate and, if not addressed, would leave Suffolk FRS without the means to
control a major BESS “fire”.
Taking a storage capacity of 10 MWh in just one of the Sunnica cabins (see Appendix 1), a
complete thermal runaway accident in such a BESS would release that stored electrochemical
energy, plus an indeterminate quantity of heat from combustion of hydrocarbon polymer materials
or inflammable gases evolved from them. Such Total Heat Release may be up to twenty times the
amount of the stored electrochemical energy in the BESS [11].
The thermal capacity of water is 4.2 kJ kg-1 K-1 or in kWh terms, about 1.17 kWh m-3 K-1. If
heated from 25 °C to boiling point about 87.8 kWh m-3 of thermal energy is required.
Hence the water volume required to absorb 10 MWh of released energy without boiling is about
114 m3 or 30,000 US gallons, the same amount as required in practice to control a fire in a single
Tesla Model S car with a mere 100 kWh battery, 100 times smaller than a 10 MWh BESS.
The quantity suggested by the Local Authorities’ joint response is 228 m3 (1900 L min-1 for 2
hours), twice the above estimate, which would naively be sufficient for a 20 MWh BESS fire.
However, from the experience of recent BEV fires, it could be insufficient by a factor of 100.
No such calculations were presented in the Examination of the 700MWh Cleve Hill BESS [6].
2. “Clean agent” fire suppression systems are a common fire suppression system in BESS, but are
totally ineffective to stop “thermal runaway” accidents. The McMicken explosion was an object
lesson in this: the installed “clean agent” system operated correctly, as designed, on detection of a
hot fault in the cabin [8]. There was no malfunction in the fire suppression system. But it was
completely useless because the problem was not a conventional fuel-air fire, it was a thermal
runaway event. Only water will serve in thermal runaway.
Indeed in the McMicken explosion the “Novec 1230” clean agent arguably contributed to the
explosion by creating a stratified atmosphere with an air/Novec 1230 mixture at the bottom and
inflammable gases accumulating at the cabin top.
The most probable cause of the explosion was mixing caused by the opening of the door by first
responders. The explosive mixture contacted hot surfaces and ignited [8].
3. A further recommendation of the Hill/DNV report [8] into the McMicken explosion is for a
means of controlled venting of inflammable gases before first responders attempt access. In the
Local Authority response to the Sunnica consultation, ventilation is listed as a BESS requirement
[22] but the reason given, bizarrely, is “to control the temperature” at which ventilation or air-
conditioning (also listed) would be totally ineffective, lacking any significant thermal capacity.
The critical reason for controlled ventilation is the removal of inflammable gases before an
explosive mixture forms. Deflagration panels (to decrease the pressure of explosions that do occur)
are also recommended.
21 June 5, 2021
It should be noted that although controlled venting provisions would mitigate the consequence
of inflammable gas evolution, they would also require simultaneous venting of Hydrogen Fluroide
that would be evolved concomitantly.
Toxic gas hazard would continue to present a risk to the community and the environment for
the duration of the incident. Fire-water will be contaminated with, inter alia, highly corrosive
Hydrofluoric Acid. Contamination of water supplies and waterways must be prevented.
It is strongly recommended that Fire Services study the Hill/DNV report [8], and the related
Underwriters Labs report [9], act upon their recommendations, and make realistic, physics-based,
calculations of the water quantities required to be available at every single BESS cabin. There
could be as many as 150 BESS cabins at the Sunnica East B site alone see Appendix 1; each of
these would need a sufficient water supply.
22 June 5, 2021
[1] Major Emergency Report MER 49652 (Liverpool City Council, Environmental Health Dept) Report from
Merseyside Fire and Rescue Services and operator Ørsted into the battery fire at Carnegie Rd, Liverpool,
14-15 September 2020.
Promised January 2021 but still not released as of May 2021.
MER 49652 is a Liverpool City Council file code.
[2] Washington Post 19 April 2021
[3] Letter 2 August, 2019 from Commissioner Sandra Kennedy, of the Arizona Public Service Commission.
Docket E-01345A-19-0076, State of Arizona Public Service Commission.
[4] Energy Storage News.
[5] Liverpool echo, 15 September 2020.
[6] Cleve Hill Development Consent Order.
[7] Sunnica Preliminary Environmental Information Report, Ch 3: Scheme Description
[8] D. Hill (2020). McMicken BESS event: Technical Analysis and Recommendations, Arizona Public Service.
Technical support for APS related to McMicken thermal runaway and explosion.”
Report by DNV-GL to Arizona Public Service, 18 July 2020. Document 10209302-HOU-R-01
[9] McKinnon, M B, DeCrane S, Kerber S (2020).
“Four fire-fighters injured in Lithium-ion Battery Energy Storage System explosion Arizona”.
Underwriters Laboratories report July 28, 2020.
UL Firefighter Safety Research Institute, Columbia, MD 20145
[10] Li-ion batteries video.
[11] Larsson F, Andersson P, Blomqvist P, Mellander BE (2017).
Toxic fluoride gas emissions from lithium-ion battery fires.
Scientific Reports 7, 10018 doi: 10.1038/s41598-017-09784-z
[12] Public Health England (2017).
Hydrogen Fluoride and Hydrofluoric acid Incident Management, PHE gateway number 2014790
[13] Golubkov A W, Fuchs D, Wagner J et al. (2014).
Thermal runaway experiments on consumer Li-ion batteries with metal-oxide and olivin-type cathodes.
RSC Advances 4, 3633-3642 doi: 10.1039/c3ra45748f
[14] D J C MacKay (2009) “Sustainable Energy Without the Hot Air” UIT Cambridge Ltd p329
23 June 5, 2021
[15] Rigby S E, Lodge T J, Alotaibi S et al.
Preliminary yield estimation of the 2020 Beirut explosion using video footage from social media.
Shock Waves. doi:10.1007/s00193-020-00970-z
[16] Pilger, C, Hupe P, Gaebler P, Kalia A, Schneider F, Steinberg A , Henriette S & Ceranna L (2020). Yield
estimation of the 2020 Beirut explosion using open access waveform and remote sensing data.
Bundesanstalt für Geowissenschaften und Rohstoffe - submitted
[17] Understanding COMAH a Guide for new entrants
Appendix 1 Figure 1 flowchart.
[18] Letter from Dr E J Fordham to HSE, 25 November 2020.
Addressed impersonally but sent Recorded Delivery and receipted.
[19] Letter from Ms Sarah Albon, Chief Executive of HSE, to Dr E J Fordham, 22 February 2021.
[20] Allianz Risk Consulting (2019) Battery Energy Storage Systems (BSS) using Li-ion batteries,
Technical Note Vol 26.
[21] National Fire Protection Association (2020)
Standard for the Installation of Stationary Energy Storage Systems Standard 855, Table 9.2
[22] Response of affected Local Authorities
(Cambridgeshire and Suffolk County Councils, West Suffolk and East Cambridgeshire District Councils) to
Sunnica Consultation.
[23] Note (2/11/20) from Councillor Andrew Douch, Freckenham Parish Council, meeting 30 October 2020
[24] Power Engineering, 4/18/2017, “What you need to know about energy storage”.
[25] Energy Institute (2019) Battery Storage Guidance Note 1: Battery Storage Planning, Sec 4.2 page 16
[26] A guide to the COMAH regulations (2015):
[27] A guide to the COMAH regulations (2015):
Schedule 1,Part 1, Col 2.
24 June 5, 2021
Appendix 1: Battery Capacity Calculations for the Grid-scale BESS proposed at the “Sunnica” site.
The Sunnica scheme will be taken as an example of a mega-scalesolar plant with BESS. If
approved, it would cover approximately 2800 acres and will include BESS on 3 separate sites.
The proposed BESS capacity in the Sunnica scheme has not been specified. Estimates of storage
capacity can be made on the basis of the land areas allocated to the BESS compounds, assuming full
use (per meeting with Parish Councillors, 30 October 2020 [23]). Li-ion battery technology has also
been assumed because it is the most widely used in the industry today. Li-ion batteries have a high
energy density, and the costs of these have fallen significantly over the past few years [24] .
Land areas and cabin size are quoted in the Sunnica Scheme Description as:
Sunnica East A: 5.23 ha
Sunnica East B: 15.6 ha
Sunnica West A: 10.65 ha
Total: 31.48 hectares.
One storage cabin size is 15 m length ´ 5 m width ´ 6 m height. This height is double that of a
so-called “hi-cube” shipping container and has a larger footprint (75 m2 vs 30 m2 for a standard 40-
foot shipping container).
Storage capacity can be estimated based on other BESS and storage cabin volumes:
Single cabin energy storage capacity:
The McMicken, Arizona, Li-ion BESS was a single cabin, footprint of 60 m2 andshipping container
height. The Sunnica BESS cabins are 75 m2, with ‘double shipping container height (6 m).
Energy storage at McMicken was 2 MWh.
Scaling by footprint and height yields a single cabin energy storage capacity estimate of 5 MWh
for each of the “Sunnica” BESS cabins.
The Arizona cabin had empty space for expansion racks, so a larger single cabin energy storage
capacity, up to say 10 MWh, is entirely conceivable.
Density of BESS cabins on allocated land:
This is unstated by Sunnica. We assume that 7.5% of the allocated land area will be occupied by the
BESS cabins themselves (this allows for safety separations, fire access routes, Battery Management
Systems (BMS) and other electrical plant, bunding for firewater in the event of incidents). This
implies a total of 315 BESS cabins allocated over the three sites.
Total scheme storage capacity:
5 MWh (single cabin capacity) x 315 cabins yields a total energy storage capacity of 1575 MWh (or
1.574 GWh), distributed over 3 separate battery compounds of unequal size (31.48 ha total). If the
single cabin capacity were 10 MWh, the total doubles to 3150 MWh.
A storage capacity between 1500 3000 MWh is therefore credible for the Sunnica proposal,
depending on single cabin storage and the density of cabins on the land.
The area density of storage at this cabin density would be 50 MWh ha-1 for a single-cabin
storage of 5 MWh. This figure of 50 MWh ha-1 is independent of the total area allocated; it depends
only on the assumed fraction (7.5%) occupied.
For comparison, the corresponding density at Cleve Hill [3] is a very similar 69.2 MWh ha-1.
25 June 5, 2021
The Energy Institute [25] gives 100 MWh ha-1 as ‘typical’ for Li-ion BESS planning. This density
would be reached in our assumptions if the single cabin capacity were 10 MWh. The latter figure is
entirely conceivable because the “base estimate” derives from an incompletely populated cabin. It
is also readily achievable if the spacing of cabins is closer than implied by the assumption of 7.5%
land occupancy.
The base caseestimate of 315 cabins and 1574 MWh is an overestimate only if the project
does not fully occupy the allocated land (i.e. BESS cabin density is less than the 7.5% assumed), but
this would be contrary to advice from the developer in meetings with local Councillors.
It is also an overestimate if the single cabin storage capacity is less than 5 MWh. This is unlikely
because it is estimated from a BESS cabin still incompletely populated.
These estimates are summarised in the following Table.
Table 3. Estimates of electrical stored energy under various assumptions at Sunnica.
Note: “1 kiloton TNT” is equivalent to 1.163 GWh. “One Beirut” is equivalent to 580 MWh.
No. of
at area
of 7.5%
Energy storage capacity
( Single cabin )
75 m2
5 MWh
10 MWh
Per cabin
(per cabin land)
1000 m2
Sunnica East A
5. 23 ha
260 MWh
520 MWh
Per compound
estimates of stored
Sunnica East B
15.6 ha
780 MWh
1560 MWh
Sunnica West A
10.7 ha
535 MWh
1070 MWh
Whole Scheme
31.5 ha
1575 MWh
1.575 GWh
1.36 kilotons
3150 MWh
3.150 MWh
2.71 kilotons
energy only.
Does not include
chemical energy
from inflammables.
26 June 5, 2021
Appendix 2: Applicability of the COMAH Regulations to large-scale BESS
The COMAH regulations (2015): COMAH regulates establishments with quantities of dangerous
substances (categorised as toxic, flammable or environmentally damaging) that are present above
defined thresholds. The substances do not need to be present in normal operation. If dangerous
substances could be generated “if control of the process is lost”, the likely quantity generated
thereby must be considered. If the mass of dangerous substances that could be generated in loss
of control exceeds the COMAH thresholds, the Regulations apply.
There are two “tiers” to COMAH, the “upper tier” imposing more stringent controls. Thresholds
of hazardous substances are listed with thresholds for both tiers.
The regulations specify aggregation rules when more than one substance in a hazard category
(e.g. flammables) may be present; even if all such substance are below the COMAH thresholds,
others in the same hazard category must be quantified and the proportions of the threshold
aggregated. If the total exceeds one, the establishment is subject to COMAH. It is also clear that the
inventories of all “installations” including pipework must be considered as a whole.
Extracts from COMAH Regulations [26] 2(1) (definitions):
“establishment” means the whole location under the control of an operator where a dangerous substance
is present in one or more installations, including common or related infrastructures or activities, in a quantity
equal to or in excess of the quantity listed in the entry for that substance in column 2 of Part 1 or in column 2
of Part 2 of Schedule 1, where applicable using the rule laid down in note 4 in Part 3 of that Schedule;
“presence of a dangerous substance” means the actual or anticipated presence of a dangerous substance
in an establishment, or of a dangerous substance which it is reasonable to foresee may be generated during
loss of control of the processes, including storage activities, in any installation within the establishment, in a
quantity equal to or in excess of the qualifying quantity listed in the entry for that substance in column 2 of
Part 1 or in column 2 of Part 2 of Schedule 1, and “where a dangerous substance is present” is to be
construed accordingly;
Application to grid-scale BESS: The Regulations refer to a dangerous substance which it is
reasonable to foresee may be generated during loss of control of the processes”. Both Flammable
Gases (P2) and Acute Toxics (H1 and H2) are certainly reasonable to foresee” in thermal runaway
incidents which are now well-documented. The evolution of regulated, named and categorised
hazardous substances from Li-ion battery cells in thermal runaway is also well-documented. A
“worst credible accident” would have to consider that the entire inventory of Li-ion cells would be
destroyed in a single BESS cabin at least. Cabin-to-cabin propagation should also be considered.
The Regulations apply to the entire “establishment”, controlled by a single operator. Whilst the
individual BESS compounds at Sunnica might be regarded as separate establishments, it is less
reasonable that individual BESS cabins should be regarded as separate “establishments”. They are
separate “installations” but “establishment” means the entire area under control of an “operator”.
Only if the most stringent safeguards were in place to ensure that the disastrous consequences
of cabin-to-cabin propagation of “battery firescould not conceivably occur, could it be argued that
dangerous substances, exceeding the COMAH thresholds in quantity, were not “reasonable to
foresee [being] generated during loss of control of the process”.
We believe the COMAH regulations apply to BESS and that the approach of HSE is wrong in law.
Dangerous substances “reasonable to foresee … generated during loss of control of the
processes”: The literature and known experience of BESS accidents is clear that dangerous
27 June 5, 2021
substances in the hazard categories H1 and H2 (Acute Toxic) and P2 (Flammable Gases) are
foreseeable in the event of thermal runaway accidents. One of the Flammable Gases is Hydrogen,
which is a “Named Dangerous Substance” in Part 2 of Schedule 1 of the COMAH Regulations 2015.
Lower thresholds are specified for Hydrogen than for other P2 Inflammable Gases.
It remains therefore to consider the quantities of dangerous substances which could be
generated if “control of the process is lost” in a thermal runaway incident. Published literature
sources quantify evolution of flammable gases from tests of various Li-ion cells in sealed vessels.
Open “fire tests” quantify the evolution of toxic gases particularly Hydrogen Fluoride. Many other
test results exist in the records of specialist test laboratories, but here we rely upon two primary
published sources.
Golubkov et al. (2014) [13] report quantities of evolved inflammables from Li-ion cells of three
different electrode chemistries in thermal runaway situations. The proportion of Hydrogen (H2),
Methane (CH4), Ethylene (C2H4) and Carbon Monoxide (CO) does not in fact vary greatly between
different types of Li-ion cell, reflecting an underlying inventory of hydro-carbon material (plastics,
electrolyte solvents etc) that remain similar in all Li-ion technologies. This is consistent with DNV/GL
test data cited in the Hill/DNV report [8]. The quantitative estimates here are taken from results
derived from cells with Nickel-Manganese-Cobalt (NMC) electrodes, as used in the McMicken BESS.
It was not possible in the apparatus of Golubkov et al. to determine the concentrations of HF
Larsson et al. [11] report evolved quantities of Hydrogen Fluoride (HF) from Li-ion cells in open
“fire tests”, and also the Total Heat Released (THR) from combustion of the inflammables. Again
these vary between cell technologies and “form factors”, especially whether the cells have an outer
metal cannister or are in the “pouch” format. Quantities between 20 200 mg / Wh are reported.
The worst case figure is used in the following estimates; the lowest evolution reported for “pouch”
cells was 43 mg/Wh.
Both sources report evolved gas quantities on a per Wh basis. We scale these to a Li-ion BESS
cell size on the basis of stored energy since this will be roughly proportional to the electrolyte
solvents and other polymer materials in the cell. Scaling on a per mass basis would be preferable,
but this would require further information on the exact composition of the cells in the literature
tests, and indeed those for the BESS in question. During the McMicken investigation, the cell
manufacturers refused to release such data.
H1 and H2 Acute Toxics. The applicability of COMAH is easiest to determine in respect of Hydrogen
Fluoride (HF). This has a dual hazard classification [12] as H1 Acute Toxic (skin exposure) and H2
Acute Toxic (inhalation) and both exposure routes would apply to the general public nearby. The
lower tier COMAH threshold for H1 Acute Toxics is 5 tonnes [27]; using the upper estimate of 200
mg/Wh from Larsson, the BESS capacity at which a BESS enters the scope of COMAH (lower tier) is
25 MWh.
This is far below the projected storage capacities given in Table 3 (Appendix 1). With high
storage capacity cabins (of e.g. 12.5 MWh), it would require propagation of a fire from just one
cabin to a second, to generate HF above the COMAH threshold. It is not necessary to foresee a
major conflagration involving multiple cabin-to-cabin propagation to bring the establishment within
scope of COMAH; just two cabins would suffice. If 25 MWh were stored in a single large cabin, the
question of cabin-to-cabin propagation is irrelevant.
28 June 5, 2021
The upper tier for “H1 Acute Toxics” is entered at four times higher capacity (100 MWh), which
is well below the estimated capacity of Cleve Hill, and is also below each of the three Sunnica BESS
compounds individually.
Even on the lowest evolution figure of 43 mg/Wh reported by Larsson et al. for “pouch” cells,
the lower tier of COMAH is entered at a storage capacity of 120 MWh, again well within the “low
case” capacity of each of the Sunnica BESS compounds, and Cleve Hill.
There is little doubt that either the lower or upper tier of COMAH is applicable to Cleve Hill and
all three of the Sunnica BESS compounds, on the basis of “H1 Acute Toxics” (HF, skin route) alone.
Carbon Monoxide (CO) is categorised as an H2 Acute Toxic as well as a P2 Inflammable Gas, and
will also be evolved, but in application of the aggregation rule its presence does not materially alter
these conclusions. It is sufficient to consider HF alone.
P2 Inflammable Gases. Assessing applicability of COMAH on the basis of inflammable gases is more
complicated because of the evolution of Hydrogen (H2), Methane (CH4), Ethylene (C2H4) and Carbon
Monoxide (CO) in significant quantities, and because Hydrogen is a “named dangerous substance”
for which different COMAH thresholds apply. These must be taken into account when applying the
Aggregation Rule. Although proportions are generally similar, quantities do depend on the different
electrode chemistries in the different Li-ion cell types.
Taking the largest evolutions reported by Golubkov et al. [13] for the LCO/NMC electrode type
tested by them these are equivalent to 335 mg/Wh of P2 inflammables. For the NMC cells tested
(the McMicken cells were NMC) the evolution was 214 mg/Wh. Taking the higher figure and
applying the aggregation rule, grid-scale BESS enter the lower tier of COMAH at about 30 MWh
capacity. Taking the lower figure, they enter the lower tier at 45 MWh capacity.
Hence there is little doubt that grid-scale BESS are lower tier COMAH establishments on the
basis of “P2 Inflammable Gases” at storage capacities between 30 45 MWh.
Because of the variability between cell types, and the difficulty of scaling laboratory tests to
actual BESS cells without detailed composition data, there is room for adjustment. However the
calculated estimates of the thresholds for applicability of COMAH are so far below the projected
capacities that it is inconceivable that the Cleve Hill and Sunnica BESS compounds would not be
COMAH establishments, in lower tier at the very least, and probably the upper tier also.
Conclusion: Grid-scale Li-ion BESS should be considered COMAH establishments in the lower tier on
the basis of “H1 Acute Toxics” (HF) alone, at energy storage capacities in the region of 25 MWh.
Upper tier would apply at about 100 MWh. They should be lower-tier COMAH establishments on
the basis of “P2 inflammable gases” alone, at storage capacities between 30 45 MWh. Again
larger establishments could become upper tier COMAH. Laboratory closed vessel and fire tests on
actual Li-ion BESS cells proposed to be deployed would be required to refine these estimates
It is difficult to see how these conclusions could be avoided if tested in litigation.
29 June 5, 2021
Appendix 3: Shortcomings of Existing Engineering Standards for Li-ion BESS
The July 2020 report for the Arizona Public Service by Dr D Hill [8] provides a comprehensive
discussion of existing engineering standards relating to BESS, and how they are inadequate to
address the known hazards of “thermal runaway” incidents in Li-ion BESS. This was the failure
mode leading to the explosion at McMicken, Arizona.
Unfortunately, when the UK’s first “mega-scalesolar plant and battery storage site was granted
approval in May 2020, this paper had not been published. The Cleve Hill solar developers cited one
standard, UL 9540A [3]. This is also cited by some insurance and risk consultants [20].
It is important to be clear that nothing in UL 9540A addresses thermal runaway, and as a test
method standard, it can provide no “safety certification” for Li-ion BESS.
Specific criticisms made in the Hill/DNV report include the following:
1. UL 1973 allows for the complete destruction of a BESS and the creation of an explosive
atmosphere so long as no explosion or external flame is observed. An installation can do all
these things but still “pass” UL 1973. At McMicken one rack was completely destroyed and an
explosive atmosphere created but no flame or explosion occurred until first-responders opened
the cabin door.
2. UL 9540A is merely a test method for generating data. It does not define any “pass/fail” criteria
for interpreting results. Specifically, it does not address cell-to-cell cascading in thermal
runaway, nor the evolution of a potentially explosive atmosphere. It does not even prescribe
that the cell-to-cell cascading rate be measured.
It allows that thermal runaway may proceed to an entire rack (as at McMicken) and offers
testing of fire suppression systems (which operated correctly at McMicken but cannot prevent
thermal runaway, and did not prevent an explosion).
Presentation of data generated under UL 9540A to an “AHJ” (Authority Having Jurisdiction) does
not translate to a succinct understanding of potential risks.
3. NFPA 855 [21] does require evaluation of thermal runaway in a single module, array or unit and
does acknowledge the need for thermal runaway protection. However, it assigns that role to
the Battery Management System (BMS). Yet thermal runaway is an electrochemical reaction
that once started cannot be stopped electrically. It is uncontrollable by electronics or
switchgear, only by water cooling.
The evolution of engineering and safety standards has not yet incorporated the lessons of
experience arising from the McMicken explosion [8] or explosion incidents in the UK like the
Liverpool explosion and fire of 15 September 2020 [1]. Compliance with existing standards does
not prevent such incidents happening again.
Articles in the industry press
do now recognise and discuss the problem of thermal runaway
but make proposals such as: If off-gases can be detected and batteries shut down before thermal
runaway can begin, it is possible that fire danger can be averted”.
Such statements betray a dangerous misunderstanding. Batteries cannot be “shut down”,
except by complete discharge, which cannot be done quickly. Taking cells “out of circuit” is useless;
thermal breakdown and runaway will still occur.
30 June 5, 2021
Appendix 4 – Fire Safety Planning requirements in the Local Authorities’ Joint Response to the
Sunnica statutory consultation
This Appendix deals point by point with the BESS requirements in the Local Authority response (text
in blue) pp 74 75.
Sunnica should produce a risk reduction strategy as the responsible person for the scheme as stated in the
Regulatory Reform (Fire Safety) Order 2005. It is expected that safety measures and risk mitigation is
developed in collaboration with services across both counties.
The Local Authorities require that the Fire Services work with Sunnica to prepare fire safety and risk
mitigation measures. The Cambridgeshire and Suffolk Fire Services are therefore the only public
bodies with independent oversight of BESS safety.
The use of batteries (including lithium-ion) as Energy Storage Systems (ESS) is a relatively new practice in the
global renewable energy sector. As with all new and emerging practices within UK industry, the SFRS would
like to work with the developers to better understand any risks that may be posed and develop strategies
and procedures to mitigate these risks.
This paper is provided as input to this process, which appears to be insufficiently understood.
The promoter must ensure the risk of fire is minimised by:
Procuring components and using construction techniques which comply with all relevant legislation.
This overlooks the points made in this paper that (i) existing legislation is being ignored by the
statutory regulatory body, the HSE (ii) no adequate engineering standards exist to exercise
Prevention measures over what is by now a very well-known hazard, viz. thermal runaway. Public
Health and Safety cannot be assured whilst either of these situations continues.
Developing an emergency response plan with both counties fire services to minimise the impact of
an incident during construction, operation and decommissioning of the facility.
Ensuring the BESS is located away from residential areas. Prevailing wind directions should be
factored into the location of the BESS to minimise the impact of a fire involving lithium-ion batteries
due to the toxic fumes produced.
This is impossible to satisfy. All the BESS compounds in the Sunnica proposal are sufficiently close to
residential areas to present a major danger of toxic fumes in the event of an accident. Plume
dispersal modelling should be performed to ensure that concentrations of HF cannot exceed
dangerous thresholds in the event of the worst credible accident in a BESS compound.
The emergency response plan should include details of the hazards associated with lithium-ion
batteries, isolation of electrical sources to enable firefighting activities, measures to extinguish or
cool batteries involved in fire, management of toxic or flammable gases, minimise the environmental
impact of an incident, containment of fire water run-off, handling and responsibility for disposal of
damaged batteries, establishment of regular onsite training exercises.
This requirement is very broad but insufficiently detailed. Means of cooling would require water
volumes many times in excess of those requested. Management of inflammable gases is best
addressed by venting, but that exacerbates the hazard of toxic gas plumes. Large water volumes
may lead to unrealistic or impossible requirements for the containment, and subsequent disposal,
of the contaminated water resulting from the fire-fighting activity. Other sections of this paper
address these points.
The emergency response plan should be maintained and regularly reviewed by Sunnica and any
material changes notified to SFRS and CFRS.
31 June 5, 2021
Environmental impact should include the prevention of ground contamination, water course
pollution, and the release of toxic gases.
Preventing the release of toxic gases is all but impossible. A thermal runaway event WILL release
toxic gases. If inflammables are vented to avoid /mitigate explosion risk, toxic gases WILL be
vented. Ground contamination and water course pollution is almost certain to occur if sufficient
water to control a major thermal runaway event is deployed. It will pose a significant challenge to
contain, and safely dispose of, such large volumes of contaminated fire water.
The BESS facilities should be designed to provide:
Automatic fire detection and suppression systems. Various types of suppression systems are
available, but the Service’s preferred system would be a water drenching system as fires involving
Lithium-ion batteries have the potential for thermal runaway.
This is a correct precaution, but no specification is made of likely water volume requirements, nor
for a “dry pipe” system allowing water to be deployed without cabin entry. We provide some water
estimates elsewhere in this paper.
Other systems, such as inert gas, would be less effective in preventing reignition.
This is also a correct insight. The so-called “clean-agent” fire suppression system at McMicken was
triggered correctly, but was useless to control thermal runaway. Moreover the stratified
atmosphere created allowed the build-up of inflammables to a dangerous level, before the
explosion occurred.
Redundancy in the design to provide multiple layers of protection.
Design measures to contain and restrict the spread of fire through the use of fire-resistant materials,
and adequate separation between elements of the BESS.
This comment only vaguely considers the true essentials. The “elements of the BESS” could be:
cells, modules, racks, strings, and the entire system. As discussed in the Hill/DNV report what is
required is for the industry as a whole to accept that thermal runaway in an unacceptable hazard,
and demand engineering standards that Prevent thermal runaway by design, or if it occurs, Prevent
its cascade or escalation to larger system elements. This requires
a. Thermal barriers (i.e. Low thermal conductivity barriers, not merely refractory barriers,
ideally with water cooling, between all cells, so that propagation from cell to cell cannot
occur. This is precisely the requirement the industry has so far NOT made in the
development of its engineering standards.
b. Separation of modules by similar barriers to Prevent module-to-module cascade.
c. Separation of Racks to prevent rack-to-rack cascade, even with ejection of molten metals.
d. Spacing of BESS cabins such that even with “75 foot flame lengths” cabin to cabin escalation
is impossible. This is probably the most critical of all, since cabin-to-cabin escalation could
turn a major fire incident into an unprecedented catastrophe, on the scale of the Beirut
explosion or a small nuclear weapon.
Provide adequate thermal barriers between switch gear and batteries,
Install adequate ventilation or an air conditioning system to control the temperature. Ventilation is
important since batteries will continue to generate flammable gas as long as they are hot. Also,
carbon monoxide will be generated until the batteries are completely cooled through to their core.
This comment is very strange. There is no possibility whatsoever that air conditioning could be
adequate “to control the temperature”. The importance of ventilation is however recognised, as is
32 June 5, 2021
the generation of carbon monoxide (toxic as well as inflammable). However the generation of
Hydrogen Fluoride will also continue until the batteries are “completely cooled” and HF (H1 Acute
Toxic by skin exposure) is much more toxic than CO (H2 Acute Toxic).
Install a very early warning fire detection system, such as aspirating smoke detection.
The “very early warning” fire detection system required should be thermocouples to report
continuously on the local temperature at every cell in the entire system. A single cell overheating
can escalate via thermal runaway. By the time smoke is generated, this will be a “very late, rather
thanvery earlydetection system. Just as thermal runaway events do not necessarily generate
flame, neither do they necessarily generate smoke, until nearby combustibles are ignited.
Install carbon monoxide (CO) detection within the BESS containers.
This is a good straightforward measure, but detectors for other gases expected (HF, H2, CH4) could
equally well serve and multiple gas detection would provides additional security.
Install sprinkler protection within BESS containers. The sprinkler system should be designed to
adequately contain and extinguish a fire.
The excellent record of sprinkler systems in ordinary building fires shows they would help contain
fire in regular combustible parts of the structure. However as discussed earlier in this paper, a
mere sprinkler system would be useless to contain thermal runaway. Much larger water quantities
would be needed.
Ensure that sufficient water is available for manual firefighting. An external fire hydrant should be
located in close proximity of the BESS containers. The water supply should be able to provide a
minimum of 1,900 l/min for at least 2 hours. Further hydrants should be strategically located across
the development. These should be tested and regularly serviced by the operator.
As discussed elsewhere, we believe these water requirements to be under-specified by a factor of
100, based on real experience with BEV fires. “Strategic location” is inadequate. Every single BESS
cabin (potentially up to 150 of these at Sunnica East B alone) should have such a hydrant.
We remark elsewhere on the recommendation made by Hill/DNV for a “dry pipe” system to deploy
water drenching inside via external connections, without cabin entry being needed.
A safe access route for fire appliances to manoeuvre within the site (including turning circles). An
alternative access point and approach route should be provided and maintained to enable
appliances to approach from an up wind direction. Please note that SFRS requires a minimum
carrying capacity for hardstanding for pumping/high reach appliances of 15/26 tonnes, not 12.5
tonnes as detailed in the Building Regulations 2000 Approved Document B, 2006 Edition, due to the
specification of our appliances.
The requirement for safe access routes and space for appliances to manoeuvre could usefully be
expanded into requirements for safe spacing of BESS cabins and thermal or flame barriers between
cabins, to Prevent the “disaster scenario” of cabin-to-cabin propagation.
Final Comment: (over)
33 June 5, 2021
Final Comment:
The fundamental failure mode of Li-ion batteries presenting major hazard is thermal runaway.
This paper is far from the first to identify the risk which is now well-known.
However the BESS industry as a whole has still not agreed or implemented adequate engineering
standards to address basic Prevention measures to pre-empt thermal runaway accidents.
Until it does, Mitigation of major accidents by the Fire Services will remain the sole recourse for
public protection and safety.
... What is more important to be concerned about than those challenges is the cost of production of batteries. However, there is a chance in the future that production costs would be lower as the method of battery production is improved [12][13][14][15][16][17][18][19][20][21][22][23]. Technology Description Basic Process is a theory which describes the thermodynamic cycle for the steam turbine which is known as the Rankine cycle [24][25][26]. ...
Full-text available
An ever-increasing demand for electrical power and soaring levels of energy consumption around the world have led to an energy crisis. Thus, this paper aims to review the conventional technologies against those of newer developments in electrical power generation such as using nitrogen generators. The nitrogen generator method is most appealing as it is a seemingly free energy already existing in nature. A nitrogen generator with a 5000 (Nm3/h) capacity has the potential to be used to analyze gas composition and the results are compared with the gas composition of a conventional steam turbine, which is used to pressurize 6000 (kWh) injection steam turbines. The magnetic bearing must be installed in both systems to modify all centrifuged systems which reduces all energy consumption in all systems by more than 50%. Artificial intelligence is used with the machine to analyze and control nitrogen gas flow to provide a more precise evaluation resulting in a more efficient technology. It should further be noted that the nitrogen turbine is superior to the steam turbine because it does not require the burning of fossil fuel to generate power. Hence, it is crucial to modify conventional technologies to improve energy sustainability and begin the long task of tackling environmental issues.
Full-text available
We report on a multi-technique analysis using publicly available data for investigating the huge, accidental explosion that struck the city of Beirut, Lebanon, on August 4, 2020. Its devastating shock wave led to thousands of injured with more than two hundred fatalities and caused immense damage to buildings and infrastructure. Our combined analysis of seismological, hydroacoustic, infrasonic and radar remote sensing data allows us to characterize the source as well as to estimate the explosive yield. The latter ranges between 0.8 and 1.1 kt TNT (kilotons of trinitrotoluene) equivalent and is plausible given the reported 2.75 kt of ammonium nitrate as explosive source. As there are strict limitations for an on-site analysis of this catastrophic explosion, our presented approach based on data from open accessible global station networks and satellite missions is of high scientific and social relevance that furthermore is transferable to other explosions.
Full-text available
Rapid, accurate assessment of the yield of a large scale urban explosion will assist in implementing emergency response plans, facilitate better estimates of areas at risk of high damage and casualties, and will provide policy makers and the public with more accurate information about the event. On 4th August 2020 an explosion occurred in the Port of Beirut, Lebanon. Shortly afterwards, a number of videos were posted to social media showing the moment of detonation and propagation of the resulting blast wave. In this article, we present a method to rapidly calculate explosive yield based on analysis of 16 videos with a clear line-of-sight to the explosion. The time of arrival of the blast is estimated at 38 distinct positions, and the results are correlated with well-known empirical laws in order to estimate explosive yield. The best estimate and reasonable upper limit of the 2020 Beirut explosion determined from this method are 0.50 kt TNT and 1.12 kt TNT, respectively.
Full-text available
Lithium-ion battery fires generate intense heat and considerable amounts of gas and smoke. Although the emission of toxic gases can be a larger threat than the heat, the knowledge of such emissions is limited. This paper presents quantitative measurements of heat release and fluoride gas emissions during battery fires for seven different types of commercial lithium-ion batteries. The results have been validated using two independent measurement techniques and show that large amounts of hydrogen fluoride (HF) may be generated, ranging between 20 and 200 mg/Wh of nominal battery energy capacity. In addition, 15–22 mg/Wh of another potentially toxic gas, phosphoryl fluoride (POF3), was measured in some of the fire tests. Gas emissions when using water mist as extinguishing agent were also investigated. Fluoride gas emission can pose a serious toxic threat and the results are crucial findings for risk assessment and management, especially for large Li-ion battery packs.
Full-text available
Li-ion batteries play an ever-increasing role in our daily life. Therefore, it is important to understand the potential risks involved with these devices. In this work we demonstrate the thermal runaway characteristics of three types of commercially available Li-ion batteries with the format 18650. The Li-ion batteries were deliberately driven into thermal runaway by overheating under controlled conditions. Cell temperatures up to 850 °C and a gas release of up to 0.27 mol were measured. The main gas components were quantified with gas-chromatography. The safety of Li-ion batteries is determined by their composition, size, energy content, design and quality. This work investigated the influence of different cathode-material chemistry on the safety of commercial graphite-based 18650 cells. The active cathode materials of the three tested cell types were (a) LiFePO4, (b) Li(Ni0.45Mn0.45Co0.10)O2 and (c) a blend of LiCoO2 and Li(Ni0.50Mn0.25Co0.25)O2.
This is the electronic version of the book which is also available in hardback and paperback.
McMicken BESS event: Technical Analysis and Recommendations, Arizona Public Service. Technical support for APS related to McMicken thermal runaway and explosion
  • D Hill
D. Hill (2020). McMicken BESS event: Technical Analysis and Recommendations, Arizona Public Service. Technical support for APS related to McMicken thermal runaway and explosion." Report by DNV-GL to Arizona Public Service, 18 July 2020. Document 10209302-HOU-R-01
Four fire-fighters injured in Lithium-ion Battery Energy Storage System explosion -Arizona
  • M B Mckinnon
  • S Decrane
  • S Kerber
McKinnon, M B, DeCrane S, Kerber S (2020). "Four fire-fighters injured in Lithium-ion Battery Energy Storage System explosion -Arizona". Underwriters Laboratories report July 28, 2020. UL Firefighter Safety Research Institute, Columbia, MD 20145
Hydrogen Fluoride and Hydrofluoric acid -Incident Management, PHE gateway number
  • Public Health England
Public Health England (2017). Hydrogen Fluoride and Hydrofluoric acid -Incident Management, PHE gateway number 2014790
Addressed impersonally but sent Recorded Delivery and receipted
Letter from Dr E J Fordham to HSE, 25 November 2020. Addressed impersonally but sent Recorded Delivery and receipted.