DNV Considerations For ESS Fire Safety 2017

Final Report
Considerations for ESS Fire

Safety
Consolidated Edison
New York, NY
Report No.: OAPUS301WIKO(PP151894), Rev. 3

January 18th, 2017
Consolidated Edison
Considerations for ESS Fire Safety
Project Name: Considerations for ESS Fire Safety DET NORSKE VERITAS (U.S.A.), INC.
(DNV GL)
Customer: Consolidated Edison
Materials & Corrosion Technology Center
Contact Person: Britt Reichborn-Kjennerud O&G Materials Compatibility / Energy
5777 Frantz Road
Date of Issue: January 18th, 2017 Dublin, OH 43017-1886
Project No.: PP151894 United States
Tel: (614) 761-1214
Organization Unit: O&G Corrosion Control/Energy Fax: (614) 761-1633
Report No.: OAPUS301WIKO(PP151894), Rev. 3 www.dnvgl.com
Task and Objective:
Please see Executive Summary.
Prepared by Verified by Approved by
Davion Hill, Ph.D. Nick Warner, M.S. William Kovacs III, P.E.
Energy Storage Leader, Americas Engineer, EAA Laboratories Senior Engineer
☒ Unrestricted Distribution (internal and external) Keywords

☐ Unrestricted Distribution within DNV GL Battery safety, fire testing, FTIR, thermal
☐ Limited Distribution within DNV GL after 3 years runaway, toxic gas, fire extinguishing, ventilation
☐ No Distribution (confidential)
☐ Secret
Reference to part of this report, which may lead to misinterpretation, is not permissible.
Rev. No. Date Reason for Issue: Prepared by: Verified by: Approved by:
0 2016-11-01 First Issue Davion Hill
1 2016-11-14 Second Issue Davion Hill
2 2016-11-29 Third Issue Davion Hill
3 2017-1-18 Final Issue Davion Hill Nick Warner William Kovacs
DNV GL – OAPUS301WIKO(PP151894), Rev. 3 ii

January 18th, 2017
Consolidated Edison
Executive Summary
This report summarizes the main findings and recommendations from extensive fire and
extinguisher testing program that evaluated a broad range of battery chemistries1. The
testing was conducted through much of 2016 on behalf of the New York State Energy
Research & Development Authority (NYSERDA) and Consolidated Edison, as they engaged
the New York City Fire Department (FDNY) and the New York City Department of Buildings
(NY DOB) to address code and training updates required to accommodate deployment of
energy storage in New York City. This executive summary can be read as a standalone
summary of the main project findings and recommendations.
The main conclusion from the program is that installation of battery systems into buildings
introduces risks, though these are manageable within existing building codes and fire
fighting methods when appropriate conditions are met. This statement comes with caveats.
There is a need to clarify a universal finding in this program: in the case of heating by fire
or thermal abuse all batteries tested emitted toxic gases. It should also be noted that the
average emissions rates of equivalent masses of plastics exceed those of batteries. Every
battery tested emitted toxic gases (Table 3 on page 29); however, this can be expected
from most fires.
The toxicity of the battery fires was found to be mitigated with ventilation rates common to
many occupied spaces. While it was found that all batteries tested emitted toxic fumes, the
toxicity is similar to a plastics fire and therefore a precedent exists. The batteries exhibited
complex fire behaviors that led to abundant water use; however, it was found that the
extinguishing requirements for batteries need not be excessive if an intelligent, system-level
approach is taken that includes external fire ratings, permits direct water contact, and
implements internal cascading protections. The general outcome of the work is that fire
safety considerations are applicable to all the batteries tested in this program, even though
vanadium redox and lead acid electrolytes were not observed to be flammable. The data
presented in this report supports these findings.
All energy systems carry with them a risk in their deployment; however, the risks identified
in this study are manageable within the limits of today’s engineering controls for safety
when appropriate conditions are met. The resulting requirements in codes, if implemented,
are within the boundaries of the typical built environment.
The batteries tested in this program are as follows:

1. Li-ion NCM (4 vendors)
2. Li-ion LiFePO4 (2 vendors)
3. Li-ion LTO
4. Lead Acid
5. Vanadium Redox
6. An additional Li-ion chemistry described as BM-LMP
1
Chemistries are listed in the Appendix on page 107
DNV GL – OAPUS301WIKO(PP151894), Rev. 3 iii

January 18th, 2017
Consolidated Edison
In addition, at the request of FDNY the following extinguishing agents were tested:
1. Water
2. Pyrocool
3. F-500
4. FireIce
5. An aerosol agent
Greater detail is found within the report. It is suggested the reader use cross references
provided in the report to see where technical information can be found that supports these
findings. This report extensively uses cross references so that the reader can begin reading
at any point in the document and quickly find relevant supporting information in other
sections of the document, similar to a handbook.
Sections Directly Informing Code Development and Training
1. Locations (see Locations and Ventilation on page 48)

2. Ventilation rate (see Locations and Ventilation on page 48, as well as the Appendix,
page 64)
3. Enclosures, fire rating (see Fire Rating, page 40)
4. Capacity limitation dependent on space (see Room Capacity Limitations on page 56)
5. Clearances (see Clearances page 55)
6. Monitoring, Detection, and Alarms (see page 55)
7. Fire suppression and Water Requirements (see Extinguishing, page 45 as well as the
Appendix, page 68)
8. Emergency Response (see Guidance for First Responders on page 34 and Frequently
Asked Questions, page 8)
DNV GL – OAPUS301WIKO(PP151894), Rev. 3 iv

January 18th, 2017
Consolidated Edison
Table of Contents
1.0 ACKNOWLEDGEMENTS ............................................................................... 1
2.0 ACRONYMS............................................................................................... 2
3.0 HOW TO USE THIS DOCUMENT ................................................................... 3
4.0 TESTING METHODOLOGY ........................................................................... 4

4.1 Cell Testing .............................................................................................. 4
4.2 Module Testing ......................................................................................... 5
5.0 CONSIDERATIONS FOR SYSTEM TYPES AND LOCATIONS ............................... 6

5.1 Large versus Small Systems ....................................................................... 6
5.2 Occupied versus Non-Occupied Spaces ........................................................ 7
5.3 Challenges with Proposed Codes ................................................................. 8
6.0 NEW FINDINGS AND ANSWERS TO FREQUENTLY ASKED QUESTIONS

IN BATTERY SAFETY .................................................................................. 8
7.0 PRESENT DAY INDUSTRY-ACCEPTED SAFETY PRACTICES FOR

ENERGY STORAGE PROJECTS ................................................................... 31
8.0 GUIDANCE FOR FIRST RESPONDERS ......................................................... 34

8.1 Considerations for Permitting and Siting ..................................................... 34
8.2 Considerations for Operations at the Scene ................................................ 35
8.3 Guidance for Isolation and Overhaul .......................................................... 38
9.0 FINDINGS RELATED TO CODES AND TRAINING .......................................... 40

9.1 Fire Rating ............................................................................................. 40
9.2 Extinguishing .......................................................................................... 45
9.2.1 Class D and Deep Seated Fires.................................................................. 45
9.2.2 Cooling and Collateral Damage ................................................................. 45
9.3 Locations and Ventilation ......................................................................... 48
9.3.1 Outdoor Locations ................................................................................... 48
9.3.2 Indoor Locations (Penthouse or Dedicated Room) ........................................ 48
9.3.3 GPM and CFM Requirement....................................................................... 50
9.4 Inspection and Monitoring ........................................................................ 55
9.5 Clearances ............................................................................................. 55
9.6 Room Capacity Limitations ....................................................................... 56
DNV GL – OAPUS301WIKO(PP151894), Rev. 3 v

January 18th, 2017
Consolidated Edison
9.7 Project Development Considerations for Interaction with First

Responders and AHJs .............................................................................. 57
9.8 Considerations for Battery Chemistries that are not Li-ion ............................ 58
10.0 SUMMARY OF MAIN FINDINGS AND RECOMMENDATIONS ............................ 59

10.1 Conclusions ............................................................................................ 61
11.0 REFERENCES .......................................................................................... 62
12.0 APPENDIX 1: REFERENCED DATA .............................................................. 64

12.1 Assumptions for Air Changes per Hour (ACH) Calculations ............................ 64
12.2 Water Mass Requirement Calculation ......................................................... 68
12.2.1 Sizing the Water Requirement to Peak HRR ................................................ 69
12.2.2 Alternative Strategy for Sizing the Water Requirement: Preventing
the Peak HRR Altogether .......................................................................... 69
12.2.3 Additional Consideration: Heat of Vaporization ............................................ 70
12.2.4 Summary of Water Extinguishing Calculations............................................. 73
12.3 Why Bowtie Models? ................................................................................ 74
12.4 Heat Load from Li-ion Battery Failures ....................................................... 76
13.0 APPENDIX 3: TESTING PLAN AND APPROACH ............................................. 76

13.1 Design of Experiments ............................................................................. 77
13.2 Combustion Gas and Particulate Matter Analysis ......................................... 77
13.3 Heat Release Rate ................................................................................... 78
13.4 Procedure............................................................................................... 78
13.4.1 Testing Procedure: Lithium Battery Gas Sampling, No Extinguishing .............. 79
13.4.2 Testing Procedure: Lithium Battery Gas Sampling With Extinguishing ............ 79
13.4.3 Testing Procedure: Flow and Lead Acid Battery Electrolyte (liquid) ................ 80
13.4.4 Testing Procedure: Lead Acid Battery Lead (solid) ....................................... 81
13.5 Large Scale Burns ................................................................................... 84
DNV GL – OAPUS301WIKO(PP151894), Rev. 3 vi

January 18th, 2017
Consolidated Edison
List of Tables
Table 1 Statistics on cell failure rates for the entire test program including all Li-
ion variants. ........................................................................................... 21
Table 2 Average release rate for battery materials over a 30 minute time period. ............ 23
Table 3 Inventory of Toxic and Flammable Hazards found in this Study ........................... 29
Table 4 Non-battery related codes for energy systems in buildings. ................................ 43
Table 5 Battery specific codes for battery systems in buildings, existing and
proposed. ............................................................................................... 44
Table 6 Immediately Dangerous to Life and Health (IDLH) for the emphasized
toxic gases identified in the testing work. ................................................... 49
Table 7 Example calculation to determine the minimum water requirement per

kg of burning cell. ................................................................................... 51
Table 8 Benchmarks for airflow and water flow for typical structures. ............................. 51
Table 9 Values derived from probabilistic analysis of water flow rates (GPM) and
air flow rates (CFM) per system energy (kWh) or mass (kg). ........................ 52
Table 10 Example implications based on extrapolated findings from testing. It

can be seen that the calculated water requirement is within the
bounds of what was described for libraries, pharmacies, warehouses,
and laboratories; similarly the air flow requirements can be at or
below unless the room volume is too small. ................................................ 53
Table 11 Air change rates based on HCl emissions as a function of room size and
quantity of failing cells. ............................................................................ 65
Table 12 Air change rates based on HCN emissions as a function of room size
and quantity of failing cells. ...................................................................... 65
Table 13 Air change rates based on CO emissions as a function of room size and
Table 14 Air change rates based on HF emissions as a function of room size and
Table 15 Conversions from ACH to CFM based on the example room volumes for
energy storage systems. .......................................................................... 66
Table 16 Aggregation of data regarding battery fires, extinguishing, emissions

rates, and extinguishing. Distributions in the column labeled “Dist”
are triangular; the mean is shown. ............................................................ 68
DNV GL – OAPUS301WIKO(PP151894), Rev. 3 vii

January 18th, 2017
Consolidated Edison
Table 17 Summary of methodologies and results of the water requirement

calculation. ............................................................................................. 74
Table 18 Battery burn tests without extinguishing, combustible batteries only ................. 82
Table 19 Battery burn or heat tests with extinguishing .................................................. 83
DNV GL – OAPUS301WIKO(PP151894), Rev. 3 viii

January 18th, 2017
Consolidated Edison
List of Figures
Figure 1 Diagram of the abuse chamber used for fire testing of batteries
in the BESS program. ............................................................................. 4
Figure 2 Configuration of module burn site. ............................................................. 6
Figure 3 A proactive, system level approach to extinguishing need not

prescribe excessive levels of water if the system also
contains a high external fire rating as well as internal
barriers to prevent cascading. ................................................................ 8
Figure 4 Peak ppm per kg (in a 0.44 m3 volume) for all batteries tested
as compared to plastics. ....................................................................... 10
Figure 5 Average emissions per kg per minute of test mass for plastics
vs. batteries. ........................................................................................ 11
Figure 6 Emission of HCl was observed from the vanadium redox flow
electrolyte. ........................................................................................... 12
Figure 7 Heating of Pb acid electrolytes yielded SO2. .............................................. 12
Figure 8 Heating of Pb acid electrolytes also yielded HCl. ....................................... 13
Figure 9 Observation of delayed cascading during extinguishing in a

module without cascading protections. ................................................ 14
Figure 10 Internal temperatures between two sandwiched pouch cells

remained 300OC higher than external faces after aerosol
extinguishing. ...................................................................................... 15
Figure 11 Some battery modules still had residual voltage after fire
testing. ................................................................................................. 16
Figure 12 Progression toward lower water requirements as testing

progressed. .......................................................................................... 17
Figure 13 Residue analysis from a coupon hung in the headspace of the

vanadium redox boiling test. ................................................................ 19
Figure 14 Percent of mass loss as a function of cell form factor. ............................ 20
Figure 15 Relationship of heat release rate (kW) per gram of mass lost. ............... 22
Figure 16 Estimated ventilation rates (air changes per hour) as a function

of room volume and mass of battery undergoing failure for
HCl. ...................................................................................................... 23
DNV GL – OAPUS301WIKO(PP151894), Rev. 3 ix

January 18th, 2017
Consolidated Edison
Figure 17 Estimated air changeover rate (air changes per hour) as a

function of room volume and battery mas undergoing failure
for HCN. ............................................................................................... 23
Figure 18 Estimated air change over rate (air changes per hour) as a
function of room volume and battery mass undergoing failure
for CO. Because the IDLH of CO is much higher, there is little
dependency on battery masses at these scales. ................................... 24
Figure 19 Estimated air change rate (ACH) as a function of room volume

and battery mass during failure for HF. ................................................ 24
Figure 20 Representative emissions histogram from a Li-ion battery. .................... 25
Figure 21 The combined LEL and autoignition temperature of mixed gases

emitted from Li-ion batteries may be as low as 3,000 ppm
and less than 400oC, as flammable gases were ignited and
burned off above this temperature. ...................................................... 26
Figure 22 For Li-ion batteries, the mass loss is directly proportional to the
state of charge prior to failure. ............................................................ 27
Figure 23 It is generally true that LiFePO4, LTO, and BM-LMP batteries

demonstrate lower than average temperatures during failure.
The temperatures indicated for Pb acid and vanadium redox
batteries is the peak heating temperature, as these
electrolytes did not demonstrate flammable or exothermic
properties as they were tested. ............................................................ 28
Figure 24 Smoke plume rising through door gaps and out of top vents in
the burn container. .............................................................................. 31
Figure 25 Project development timeline and the implementation of FMEA

or other safety review for the site. ....................................................... 33
Figure 26 Submerging batteries in water resulted in cooler temperatures,

but slow shorting and persistent CO generation (bubbles). ................. 39
Figure 27 Simplified diagram of fire impinging on the external wall of a

battery energy storage system. ........................................................... 41
Figure 28 Direct fire exposure model to a steel wall with a 1/16”

thickness and 1” air gap between the wall and battery
modules. .............................................................................................. 42
Figure 29 Performance of water compared to other agents as water

additives, top temperature of battery cell. ........................................... 46
Figure 30 Cooling performance of water compared to other extinguisher

types, bottom temperature of cell. ....................................................... 46
DNV GL – OAPUS301WIKO(PP151894), Rev. 3 x

January 18th, 2017
Consolidated Edison
Figure 31 Distribution of gallons per minute for a 1 MWh battery,

calculated from cell testing and extrapolating with the latent
heat value, which demonstrates that the 0.1 GPM/kg
estimation is highly conservative. ........................................................ 54
Figure 32 Example of the air flow requirement for a 1 MWh Li-ion system,
demonstrating that the distribution of values is strongly
skewed leftward. .................................................................................. 54
Figure 33 The FDNY Building Information Card (actual example) contains

emergency contact information for fire safety and building
engineers. ............................................................................................ 58
Figure 34 A reactive cooling approach requires an oversized water flow

requirement; whereas a system-level proactive approach
enables a reduced water requirement. ................................................. 72
Figure 35 Regression coefficients of the ventilation requirement. .......................... 72
Figure 36 Regression coefficients for the water flow rate in GPM/kg

demonstrate that the duration of the event and the water
contact efficiency are the strongest drivers in reducing the
water requirement. .............................................................................. 73
Figure 37 BowTie analysis permits the visualization of threats to a top

event, such as loss of battery control, and ties these threats
to consequences................................................................................... 75
Figure 38 Battery weight and the peak room temperature are positively
correlated............................................................................................. 76
Figure 39 Large abuse test chamber design for battery fire and
extinguishing testing. .......................................................................... 78
DNV GL – OAPUS301WIKO(PP151894), Rev. 3 xi

January 18th, 2017
Consolidated Edison
DNV GL – OAPUS301WIKO(PP151894), Rev. 3 xii

January 18th, 2017
Consolidated Edison
Consolidated Edison and NYSERDA Disclaimer

This report was prepared by DNV GL in the course of performing work contracted for and sponsored by Con Edison
along with the New York State Energy Research and Development Authority (NYSERDA (hereafter the "Sponsors").
The opinions expressed in this report do not necessarily reflect those of the Sponsors or the State of New York, and
reference to any specific product, service, process, or method does not constitute an implied or expressed
recommendation or endorsement of it. Further, the Sponsors and the State of New York make no warranties or
representations, expressed or implied, as to the fitness for particular purpose or merchantability of any product,
apparatus, or service, or the usefulness, completeness, or accuracy of any processes, methods, or other
information contained, described, disclosed, or referred to in this report. The Sponsors, the State of New York, and
the contractor make no representation that the use of any product, apparatus, process, method, or other
information will not infringe privately owned rights and will assume no liability for any loss, injury, or damage
resulting from, or occurring in connection with, the use of information contained, described, disclosed, or referred
to in this report.
1.0 ACKNOWLEDGEMENTS
The battery and extinguishing companies identified below all donated their products for testing in this
project. The Con Edison - NYSERDA team and DNV-GL are extremely grateful for the generosity and
engagement with this project by these companies. We also thank the key stakeholders for their
significant contributions to this effort.
NCM 1: LG Chem
NCM 2HE and NCM 2HP: Samsung SDI
NCM 3: Kokam (donated by Sunverge)
NCM 4: Electrovaya
LFP 1: BYD
LFP 2: XO Genesis
T 1: Toshiba
BM-LMP: C4V
VR 1: UET
PBA 1: EnerSys
Aerosol agent: Fireaway Inc. (product Stat-X)
DNV GL – OAPUS301WIKO(PP151894), Rev. 3 1

January 18th, 2017
Consolidated Edison
2.0 ACRONYMS
ACH – Air Changes per Hour
AHJ – Authority Having Jurisdiction
BESS – Battery Energy Storage Safety
BIC – Building Information Card
BMS – battery management system
BM-LMP – Bio-mineralized Lithium Mix-Metal Phosphate
BSCAT – Barrier-Based Systematic Cause Analysis Technique
BTM – Behind the Meter
CFM – Cubic Feet per Minute
CFR – Code of Federal Regulations
CHP – Combined Heat and Power
CID – Current Interrupt Device
CO – Carbon Monoxide
COF – Certificate of Fitness
C-rate – charge rate
DCE – Duty Cycle Eccentricity
DMC – Dimethyl Carbonate
DOB – New York City Department of Buildings
DOD – depth of discharge
EC – Ethylene Carbonate
EDS – energy dispersive spectroscopy
ERPG – Emergency Response Planning Guidelines
ESS – Energy Storage System
FAQ – Frequently Asked Questions
FID – Flame Ionization Detector
FDNY – New York City Fire Department
FEA – Finite Element Analysis
FMEA – Failure Mode Effects Analysis (sometimes FMECA to include “Criticality”)
FTA – Fault Tree Analysis
FTIR – Fourier Transform Infrared Spectroscopy
GPM – Gallons Per Minute
HAZID – Hazard Identification
HCl – Hydrochloric Acid
HRR – Heat Release Rate
HCN – Hydrogen Cyanide
HF – Hydrofluoric Acid
HVAC – Heating, Ventilation, and Air Conditioning
IDLH – Immediately Dangerous to Life and Health
IE – Independent Engineer(ing)
IEC – International Electrotechnical Commission
IFC – International Fire Code
IPP – Independent Power Producer
LEL – Lower Explosion Limit
LMO – Lithium Manganese Oxide
LTO – Lithium Titanium Oxide
NYSERDA – New York State Energy Research and Devlopment Authority
NAVSEA – Naval Sea Systems Command
NCA – Nickel Cobalt Aluminum
NCM – Nickel Cobalt Manganese

January 18th, 2017
Consolidated Edison
NIOSH – National Institutes for Occupational Safety and Health

NFPA – National Fire Protection Association
NHTSA – National Highway and Traffic Safety Administration
NRTL – Nationally Recognized Test Laborator
PBA – Lead Acid
PC – Polypropylene Carbonate
PID – Photo Ionization Detector
PPE – Personal Protective Equipment
PVC – Polyvinyl Chloride
SCBA – self contained breathing apparatus
SEI – Solid Electrolyte Layer
SOC – State of Charge
SOP – Standard Operating Procedures
SME – Subject Matter Expert
UL – Underwriter’s Laboratories
UN – United Nations
UPS – Uninterruptible Power Supply
VOC – Volatile Organic Compounds
VRLA – Valve Regulated Lead Acid
3.0 HOW TO USE THIS DOCUMENT

This document is designed to inform codes writing procedures and first responder training.
It can be considered a reference and handbook for this purpose. To that end, the document
is structured around key ingredients to codes as determined by a survey of building and fire
codes for energy-related machinery and devices.
Executive Summary: This section can be considered the consolidated list of findings and
recommendations from the NYSERDA/Con Edison Battery Energy Storage System (BESS)
Program.
Frequently Asked Questions (FAQ): This may be considered the main guide of the
document, cross referencing to relevant sections of the report, and also serving as an
introduction to the topic.
Recommendations: This is the main deliverable of the document. Essential data is
provided to support recommendations, detail is left to the appendix. Recommendations and
main findings are within the document text in bold.
Appendix: Supplementary reference data needed to communicate the recommendations,
but as useful reference for detailed background. The Appendix begins on page 64. The
appendix is separated in two parts that represent supporting information: a literature review
on past fire incidents and data, and a confidential appendix which can be omitted for the
public version of the report.
Literature References: Whenever possible, literature references are provided for
independent confirmation of facts, figures, or assertions. Literature references are found in
“References” on page 62.
Cross references: Whenever possible, cited data or key conclusions that are relevant to
other sections of the report are cross referenced by section title and page number.

January 18th, 2017
Consolidated Edison
4.0 TESTING METHODOLOGY

Four different lithium chemistries (LTO, LFP, NCM, BM-LMP), lead acid, and vanadium redox
batteries represented by nine unique battery types from eight different manufacturers were
tested. For the Li-ion batteries, these included prismatic cells as well as pouch cells, but no
cylindrical cells. For the lead acid and vanadium redox batteries, testing was largely focused
on the battery electrolytes. Modules were also provided for large scale burn testing. A more
explicit description of the test plan is included in the Appendix.
4.1 Cell Testing

The cells tested ranged from 1.2 to 200 Ah with an average of 52 Ah, excluding the
electrolytes from vanadium redox and Pb acid cells that were tested separately. All cells
were heated with 4 kW of radiant electric heat in DNV GL’s Large Battery Destructive
Testing Chamber (see Figure 1). All cells were placed inside the chamber and exposed to
heat until they vented. Upon venting, some cells self ignited. For those that did not, hot
point ignitors were placed in the upper half of the chamber and were activated once lower
explosive limit (LEL) reached 50% to prevent an explosion. Many cells vented enough gas to
lead to a flashover in the chamber upon activation of ignitors. In addition to heaters and
ignitors, the chamber also contained ambient and inlet air temperature thermocouples, two
thermocouples on each cell (top and bottom) and eight thermocouples in a cube shape
around the cell to act as a thermopile for Heat Release Rate (HRR) calculations; four were
level with the cell while four more were eight inches above the cell. There was one
additional thermocouple in the center exhaust stack of the abuse chamber. In addition,
swatches of Morning Pride personal protective equipment (PPE) material were placed in the
unit above the cell to assess the effect of the fire and offgas on firefighter PPE. Cells were
tested at 25, 50, 75 and 100% state of charge (SOC).
To filtration, metering and exhaust pump
Destructive Testing
Containment Chamber
(30”x30”x30”)
Flue Gas Direction
Flue Gas Gas Bag Sampling

Duct
Thermocouples
Gas Sampling Ports*

(real time)
Radiant HRR * O2, CO2, CO, H2S, SO2,
Heater and HF, F2, VOC, LEL
Thermopile
Ignitor Surface Deposition
Coupons
Morning Pride
Test Specimen
PPE Swatches
Air Inlets
Figure 1 Diagram of the abuse chamber used for fire testing of batteries in the
BESS program.

January 18th, 2017
Consolidated Edison
Finally, gas sampling was performed by a Gasmet DX4000 Fourier transform infrared
spectroscopy (FTIR) gas analyzer. This analyzer monitored HCl, HF, HCN, CO, CO 2, O2, SO2,
NO, NO2, and a range of hydrocarbons including methane, ethane, ethylene, benzene,
toluene, and others. In line with the FTIR analyzer were MSA Ultima sensors for O 2
(redundant measurement), H2, and F2/Cl2. A final MSA sensor was placed directly off the
chamber for flammability measurements. The sensor was of the catalytic bead type and was
factory calibrated to non-specific gas for total LEL measurement. This was deemed suitable
as a range of flammable gases were expected and calibration to one may show improper
bias. In addition to the gas sensors, gas capture bags were set up off of the exhaust stacks.
Select gas bag samples were taken periodically and were used to verify the FTIR
measurement.
For extinguishing, the abuse chamber was fitted with a 2.5-gallon water can with an
extinguishing trigger. The can was pressurized and engaged by a temperature trigger, with
an in-line electronic solenoid valve for actuation. Once a single temperature exceeded
350°C, the solenoid was opened and the extinguisher released. The can was typically filled
with 1 gallon of liquid and the entirety of the can was emptied. In one test, an 8-second
pulse of water was used and the solenoid reclosed. The extinguisher nozzle was fixed
approximately 10 inches from the battery, to the side and about 3 inches above. The nozzle
was a fogging mist nozzle, and because of proximity, tank pressure was reduced to 75 psi
to allow better saturation. All cells for extinguisher testing were tested at 90% SOC.
For large scale testing, a purpose built propane torch was constructed by Fire Force Inc, a
builder of aircraft fire simulators. This torch was used to apply a direct propane flame to
battery modules which were placed in a walled off shipping container shown in Figure 2 and
Figure 24. The “room” was approximately 10 feet into the trailer, with one end being the
trailer door and having a man door installed into a double sheeted drywall wall on the
interior wall. A series of ventilation ports were cut into the room to allow for ventilation
testing (two high, two low, one roof) and positive and negative ventilation were tested. In
addition, two sprinklers were piped into the room for suppression testing. Most tests were
conducted with doors open; however, two tests were conducted with the container closed to
test ventilation. In addition to the sprinklers, hose suppression was used at times as well to
assess effectiveness.
4.2 Module Testing

DNV GL and Rescue methods constructed a partially enclosed outdoor burn facility for
module testing for all Li-ion battery types where modules were provided. The module sizes
ranged from 7.5 to 55 kWh. Burns were conducted directly with a propane torch. A steel
grate was hung from the ceiling of the burn enclosure at a height of approximately 4 feet.
Below the grate a pan was constructed to catch water runoff from extinguishing. Two
sprinkler heads were installed above the burn location and were fed with a 2.5-inch line
reduced to a ½-inch pipe from a hydrant and pumper truck at the burn site.
Venting ports were constructed above and below the burn platform to control ventilation
and also provide sampling locations. The doors to the burn chamber could be opened or
closed to test the effect on oxygen, toxicity, and heat release of the fire (Figure 2).

January 18th, 2017
Consolidated Edison
Figure 2 Configuration of module burn site.
5.0 CONSIDERATIONS FOR SYSTEM TYPES AND LOCATIONS

As of 2016, energy storage systems to be deployed in the near-term market will have
differentiating characteristics dependent on size and location.
5.1 Large versus Small Systems

The testing results have been translated to scalable metrics for ventilation and fire
suppression such as cubic feet per minute of air flow per kilogram of battery mass
(CFM/kg), and gallons per minute of water flow per kilogram of battery mass (GPM/kg).
The reasons for this are several:

- Large systems and small systems should have an intelligent means of addressing
ventilation and fire suppression with a scalable metric that correlates to size or mass
is preferable to meet this challenge, rather than an arbitrary kW, kg, or kWh number
as what is proposed in some codes as shown in Table 5.
- Energy and power densities for systems are perpetually evolving and improving.
Arbitrarily prescribing a kW, kWh, or kg number to limit system installation threatens
the value proposition of energy storage as energy density increases in the future
- With an energy density metric, it is possible to translate CFM/kg or GPM/kg to
CFM/kWh or GPM/kWh with a single calculation. The same can be done for power
density such as CFM/kW or GPM/kW. Lastly, it is possible to translate these numbers
to CFM/ft3 or GPM/ft3 as is used by the fire service. All of these metrics are scalable
and can be calculated depending on context. Because battery mass and energy

January 18th, 2017
Consolidated Edison
density will continue to evolve, these metrics will capture that evolution as codes
follow the market.
- Many small systems are dependent on the ventilation and fire suppression in the
space, and there should be a means to check if the host-infrastructure is adequate.
- Large systems may have standalone ventilation and fire suppression equipment.
- One of the main stakeholders of this report is the New York Fire Department (FDNY)
and consequently most United States (US) fire departments, and they are familiar
with GPM and CFM units of measure for firefighting and codes.
5.2 Occupied versus Non-Occupied Spaces

As discussed, the proposed codes in many standards organizations shown in Table 5
become increasingly prescriptive as energy equipment becomes installed in occupied
spaces. Non-occupied spaces (such as outdoor energy storage containerized systems, for
example) may have less restrictive codes for ventilation or clearance.
The water flow calculations presented in this document are addressing a key issue in battery
safety. Over-reaction to the threat of thermal runaway has led to recommendations for
“copious amounts of water” [12] for the extinguishing of Li-ion battery systems. Such
recommendations inflate the perceived water requirement. The reasoning for this is logical;
it is better to err on the side of caution and advise first responders to use as much water as
possible to indirectly cool the battery system.
This work has demonstrated that excessive water need not be the design criteria but should
instead be considered part of an intelligent set of safety systems including external fire
ratings, internal cascading protections, and fixed suppression systems to slow the
propagation of heat in a combined manner such as in Figure 3. If a systems approach to
safety is taken, the water requirements may be far less severe. If and when first responders
need to react to a system fire, it may be the case that these systems be overridden or
overcome, and a “copious amounts of water” approach may be desired. Therefore water
requirements for the codes and water requirements for first responders are
separate issues.

January 18th, 2017
Consolidated Edison
Figure 3 A proactive, system level approach to extinguishing need not prescribe

excessive levels of water if the system also contains a high external fire rating as
well as internal barriers to prevent cascading.
5.3 Challenges with Proposed Codes

The findings from this program indicate that scalable metrics are appropriate for sizing
ventilation and water requirements for building sites. A summary of proposed codes is
shown in Table 5. For example, in proposed changes to the International Fire Code IFC 608,
20 kWh is cited as a threshold for battery sizes or 600 kWh in a room. The code also
proposes 3 feet of clearance between battery arrays. Such prescription threatens the value
proposition of energy storage as energy and power density metrics have been increasing
rapidly over the last 5-8 years. Limitations placed on kWh or kW will directly limit the
energy service function of the device and will therefore limit the market. Providing scalable
safety metrics, however, will allow the market to be flexible within safety limits.
6.0 NEW FINDINGS AND ANSWERS TO FREQUENTLY ASKED

QUESTIONS IN BATTERY SAFETY
The findings of this program directly address some common misperceptions in battery
safety. It is therefore helpful to address some of them directly in this section. These
questions are an aggregation of questions posed during the testing program by FDNY,
battery vendors, and other stakeholders. Reading through this section may serve as an
adequate introduction to the topic and will also guide the reader through the report and its
logic.

January 18th, 2017
Consolidated Edison
Question: Are the commonly cited battery fires in the media due to spontaneous
ignition events?
Finding: No. The Literature Review (an addendum to this report) covers several incidents in
detail. In the context of fire risk and firefighting for batteries, it is helpful to summarize the
abuse tests that are performed in United Nations (UN) 38.3, the required testing scope in
order to ship and transport Li-ion batteries. The eight separate tests in UN 38.3 are a
checklist of nearly all physically conceivable abuses that could cause a Li-ion battery to
catch fire. These abuse events are:
1. Low ambient pressure
2. Overheating
3. Vibration
4. Shock
5. External short circuit
6. Impact
7. Overcharge
8. Forced discharge
All of the safety incidents commonly reported in the general media can be traced to one of
these abuse mechanisms. In some cases, contaminants in the battery (as a result of
manufacturing defects) weaken the ability of the battery to withstand instances of these
eight abuse factors. In general it is good practice to avoid any scenario that may introduce
the threat of any action on the above list. Three items in particular (overheating, external
short circuit, and impact) are the abuse mechanisms that have increased probability of
occurring to a battery during and after a fire. The fire is the most obvious heat source, but
subsequent heating may occur internally once batteries reach critical temperatures
(typically > 120oC). Short circuiting may occur by contact with tools or equipment or by
water. Items #7 and #8 are electrical stimuli that are typically monitored and controlled by
active safety barriers in the battery management system (BMS).
Question: How is the battery industry handling safety today?

Finding: For most energy storage projects that are not paid for on the “balance sheet”, the
typical independent engineering (IE) verifications that are required in the wind and solar
industries apply to energy storage projects as well. During the technology review,
performance and safety analyses are performed. This may include a review of accredited
testing, certifications, and other hazard-consequence analyses. DNV GL routinely supports
this with risk analysis to look at the overlap between energy storage system (ESS) safety
functions and the site (see “Why Bowtie Models?” on page 74); particularly for energy
storage projects that are a portfolio of behind-the-meter devices deployed across a
geography in a mix of commercial and industrial applications. In some cases for larger
installations a heat and plume study is performed to determine clearances. Such practices
are common to energy and petrochemical sectors prior to the commissioning of any new
project. IE practices are described in “Present Day Industry-Accepted Safety Practices for
Energy Storage Projects” on page 31.
Question: Are battery fires more toxic than plastics fires?

Finding: In general, no, with conditional exceptions. The average emissions rate 2 of a
battery during a fire condition is lower per kilogram of material than a plastics fire, as
shown in Figure 5. However, the peak emissions rate (during thermal runaway of a Li-ion
2
Emissions concentration in ppm averaged over total minutes of burn time

January 18th, 2017
Consolidated Edison
battery, for example) is higher per kilogram of material than a plastics fire, as shown in
Figure 4. This illustrates that a smoldering Li-ion battery on a per kilogram basis can be
treated with the same precautions as something like a sofa, mattress, or office fire in terms
of toxicity, but during the most intense moments of the fire (during the 2-3 minutes that
cells are igniting exothermically) precautions for toxicity and ventilation should be taken. It
should be noted that if Li-ion battery modules are equipped with cascading protections, the
cell failure rate may be randomized and staggered. The randomized failure rate limits the
toxicity and heat release rate of the fire.
Figure 4 Peak ppm per kg (in a 0.44 m3 volume) for all batteries tested as
compared to plastics.

January 18th, 2017
Consolidated Edison
Figure 5 Average emissions per kg per minute of test mass for plastics vs.
batteries.
Question: Is standard firefighter turnout gear adequate protection from a battery

fire?
Finding: DNV GL and the provider of turnout gear (Honeywell Morning Pride) did not note
any degradation in PPE as a result of exposure to fire test conditions when the gear was fit
on a mannequin and exposed to the fire directly. Therefore first responders equipped with
standard issue turnout gear may have protection against the toxic gas species observed
under these tested conditions. Limited electrical protection was also observed without
modifications to PPE, based on the conditions tested. Rescue Methods used common MSA
Altair four- and five-gas sensors during full scale testing. Rescue Methods worked with
Honeywell to test turnout gear, and one recommendation from Honeywell was that the
general materials integrity of the jacket should withstand most species measured in this
testing, cautioning that sustained exposure to Chlorine can have a degrading impact on
Kevlar. It should be noted that HCl was observed in the battery fire testing and is also a
common byproduct from combustion of most plastics in similar or greater volumes per
kilogram of burning material.

January 18th, 2017
Consolidated Edison
Question: Are any batteries excluded from the ventilation requirement?

Finding: Because the volume of the room plays a key role in dictating the ventilation rate,
batteries in larger rooms will have lower air changes per hour (ACH) requirements and the
size of the room will have a buffering effect on the peak emission rate. The vanadium redox
and lead acid batteries tested both emitted HCl upon heating, starting as low level emission
around 100-150OC (see Figure 6 for vanadium redox and Figure 8 for lead acid). The
findings in this program demonstrate that HCl plays a dominating role in ventilation rates
for battery systems in enclosed spaces, and because it is common for all battery types
tested, ventilation recommendations (in section “Locations and Ventilation” on page 48) are
universal for all battery types. However, it should be noted that in the smallest unit of
failure scenarios, the recommended ventilation rate of 0.25 ACH is well below the
typical rating of 3-4 for most general spaces which means that vanadium redox
and Pb acid batteries, as well as single cell failure modes for Li-ion, are already
within the implied code requirements [27]. Laboratories and server rooms can have
ACH ratings > 10. Therefore the DNV GL recommendation for air change rates > 0.25 ACH
is already exceeded by the building code in most instances.
Figure 6 Emission of HCl was observed from the vanadium redox flow electrolyte.
Figure 7 Heating of Pb acid electrolytes yielded SO2.

January 18th, 2017
Consolidated Edison
Figure 8 Heating of Pb acid electrolytes also yielded HCl.
Question: What kind of testing is required to certify the safety of battery systems?
Finding: The most commonly referenced system level safety testing the US is Underwriters’
Laboratories (UL) 1973. For marine and automotive applications, International
Electrotechnical Commission (IEC) 62619 covers many of the same requirements and has a
more stringent pass/fail criteria to demonstrate limited cascading between cells. The US
market appears to be moving toward UL 9540 which includes aspects of UL 1973 and UL
1642 (for cell safety) in addition to an up-front failure mode effects analysis (FMEA) on the
system. As mentioned, such a risk analysis should also include the site under unique
circumstances. It is also conventional to have a third party inspect the field installation and
provide a sign-off for the local authority having jurisdiction (AHJ). Globally, UN 38.3 is the
most widely recognized safety testing for Li-ion battery cells and is a requirement for
transport. The results of accredited safety testing are an indicator of the strength of the
barriers in a risk model.
Question: Do battery systems have an external display of error or health?

Finding: Yes, in a limited way. The present codes in NYC for uninterruptible power supplies
(UPS) require a system health display panel. A primary concern for first responders is lack
of knowledge about what is happening inside the battery system upon being called to the
scene, which impacts their ability to deem a site under control and then hand off control of
the site to the property owner. Some engagement between the systems integration, project
development, and first responder community is needed to discuss viable solutions for such a
panel, or whether the intent of the panel is met through other means (such as an
emergency hotline or remote data access by phone or other means).
Question: Do battery fires re-ignite?

Finding: The term “re-ignition” is a misnomer due to the factors described in the incident
history of Li-ion battery fires. Upon extinguishing, great care must be taken to assure that
all electrical, thermal, and mechanical abuse factors are neutralized. If any remain, it

January 18th, 2017
Consolidated Edison
poses a hazard for continuing (not reigniting) the fire. Therefore, it is technically inaccurate
to classify this as re-ignition if the primary cause of the hazard is never removed. After a
fire, a battery module or system may contain intact cells that still have DC voltage, meaning
there is a persisting electrical hazard (Figure 11). Water shorting out cells, for example, is a
genuine risk (such as was witnessed in the Chevrolet Volt crash test or the flooded Fisker
cars [15,17 ]). In addition, if the heat deep within the module has not been removed, that
heat poses a continued thermal hazard. DNV GL and Rescue Methods witnessed this effect
during testing as shown in Figure 9, the cause of which was lack of thermal barriers
between cells. DNV GL replicated this effect in more controlled laboratory tests in Figure 10
and observed that temperatures between battery cells can be 300 oC higher than the
exterior during extinguishing unless there is a means to remove internal heat or prevent its
transfer between cells. First responders should be cognizant that all electrical, thermal, and
mechanical hazards have been mitigated before deeming a battery fire fully extinguished.
Figure 9 Observation of delayed cascading during extinguishing in a module

without cascading protections.

January 18th, 2017
Consolidated Edison
Figure 10 Internal temperatures between two sandwiched pouch cells remained

300OC higher than external faces after aerosol extinguishing.
Question: What is the time frame for delayed ignition?

Finding: As mentioned previously, this is entirely dependent on whether the residual
hazard is electrical, thermal, or mechanical. If these measures are successfully taken then
no delayed ignition should occur. In the case of thermal abuses, DNV GL witnessed the
residual heat cause a delayed cascading event within 10 minutes (Figure 9). In the case of
the Chevrolet Volt that shorted across the battery pack terminals after the National Highway
and Traffic Safety Administration (NHTSA) crash test, the shorting event occurred 3 weeks
later and was a separate hazard event.[15] Again, the delay was due to the time it took for
the coolant to leak and eventually short the battery; it is not the battery that caused this
event but the electrical short hazard introduced by the coolant. Therefore if all electrical,
thermal, and mechanical hazards are monitored, controlled, or mitigated, first responders
should be able to assess the risk of delayed cascading during the first encounter and the
minutes or hours after extinguishing. The signature of any abuse due to shorting, crush or
penetration, or residual heating is climbing temperature on the battery, which can be
monitored by the system thermocouples (if they are still intact and the data is provided
remotely) or by handheld thermal sensors or infrared (IR) monitors.
Question: How long does it take for a Li-ion battery to go into thermal runaway if
it is being heated?
Finding: This is entirely dependent on the rate of heat absorption into the cell. DNV GL
observed in this work and other projects that a Li-ion cell can smolder for more than an
hour if the heat transfer rate is slow. By the time temperatures near 120oC (248oF) were
reached, all Li-ion batteries tested (including LiFePO4 and LTO chemistries) offgassed and/or
ruptured. If the threshold near 120oC is never crossed, the battery may smolder and gas
but never ignite unless an external spark ignites the flammable gases emitted from it. It
was common for LiFePO4, LTO, and the BM-LMP cells to offgas without flame, but their
offgas composition contains the same flammable and toxic constituents as batteries with
higher temperature failures.

January 18th, 2017
Consolidated Edison
Question: Are there risks of electric shock?

Finding: During extinguishing, Rescue Methods did not observe transfer of electricity from
the battery system to the first responder through the water stream. Some sparks were
observed to be thrown during the active burning of some modules. Arcing was also
observed when the batteries were disturbed, such as when they shorted to metal tools or
the metal support structure upon which they sat. The turnout gear worn by the firefighters
provided adequate protection such that no evidence of shock was observed in the conditions
of this test program. The shock hazard, as shown in Figure 11, is presented by stranded
energy in the form of DC voltage in the remaining intact cells.
Figure 11 Some battery modules still had residual voltage after fire testing.
Question: Is water a sufficient extinguisher?

Finding: DNV GL’s testing indicates that all extinguishers have benefits and drawbacks,
including water. Every extinguisher that DNV GL tested put out the flame on battery cells,
including the aerosol. During module testing, all extinguishers tested knocked out the flame
but in some cases the flame rekindled once the stream was removed because the battery
was still hot enough to ignite the remaining fuel. The ideal battery fire extinguisher would
be both highly thermally conductive and highly electrically insulating. Water is the former
but not the latter. Deionized water is both until it dissolves contaminants from the fire,
including ash and soot. In DNV GL’s testing, it was found that other extinguisher types could
have equal or poorer heat removal capability to water, but all were electrically conducting
due to their reliance on water as a dispersion medium. (Figure 29) Gases or aerosols—due

January 18th, 2017
Consolidated Edison
to lack of thermal mass, poorer thermal conductivity, and restricted access to the deep
seated heat source—were not observed to cool as quickly as water can. Water has been
historically recommended because of its ability to cool. It was found in this program that
water cools best, with the potential unwanted side effect of shorting other cells.
Question: Do battery fires require “copious amounts of water” to be extinguished?

Finding: If appropriate precautions are not taken to limit propagation between cells in the
module design, then the water requirement could be described as “copious” as NHTSA
coined in 2012. [12] The total content of water is entirely dependent on the water contact
efficiency with the battery cells (see the regression coefficients in Figure 36 and the GPM
example calculation in Figure 31). This language is anecdotal, however, and requires some
quantification. As mentioned previously, lack of barriers between cells results in a deep
seated and inaccessible fire (Figure 10). In practice, this would result in the use of more
water to cool and contain a battery fire. The use of “copious amounts of water” potentially
introduces the unwanted effect of shorting out other cells, thereby perpetuating the fire.
The water amount need not be so excessive if heat can be removed from the between cells,
and cells have limited ability to transfer heat to nearest neighbors. DNV GL found through
testing that this water amount could be increasingly reduced as strategies to direct cooling
were learned (Figure 12).
Figure 12 Progression toward lower water requirements as testing progressed.
Question: What about fire suppressants other than water?

Finding: DNV GL found that all suppressants put out the fire including an aerosol. The most
effective agent for cooling the fire is water. (Figure 29)
Question: Is FM-200 sufficient as an extinguisher?

Finding: FM-200 was not included in the test scope of this program. DNV GL did, however,
obtain permission from an aerosol manufacturer to test their product, which succeeded in
putting out the cell fire. The testing demonstrated that the cooling rate for the aerosol is
less than the liquids (a direct consequence of less thermal mass in a gas versus a liquid,
and a reduced effect from latent heat of vaporization). If gases have less thermal mass to
take heat from the batteries, then it is the assumption that all gas-based agents are likely
to cool less effectively than water. For this reason DNV GL recommends a staged
extinguishing approach as demonstrated in “Extinguishing” on page 45.
Question: Do the other extinguishing agents produce slippery conditions?

Finding: DNV GL and Rescue Methods did not observe slippery conditions with the use of
the other agents during full scale testing.

January 18th, 2017
Consolidated Edison
Question: Is a 2 ½” hose line with 250 GPM sufficient to put out a battery fire?
Finding: This is dependent on the battery size. DNV GL translated the findings to both
GPM/kg and GPM/kWh of battery mass (Table 9). In general, however, if the water can be
targeted at the deep seated, highest temperature areas of the fire, it will be most effective
and the water requirement will be reduced. It is demonstrated as an example in Table 7
that 250 GPM is more than sufficient for typical battery systems on the market, provided
that cascading protections and external fire rating requirements are also met.
Question: How much water is required?

Finding: DNV GL found in Table 17 and Table 9 that a minimum of 0.07-0.1 GPM/kg of
battery mass can accomplish both extinguishing and cooling for a battery fire.
Accommodation for increasing energy density can be accomplished by dividing this number
by the energy density (in Wh/kg) and multiplying by 1000 Wh/kWh.
Question: Is the water extinguisher requirement for batteries significantly larger

than what is already prescribed in the built environment?
Finding: This need not be the case if battery firefighting is considered at a system level.
Residual heat within a battery module was observed in this program, demonstrating that
battery modules equipped with cascading protections will have a reducing effect
on water flow rate requirements because less water will be needed to cool them.
This has direct economic impacts on the overall system installation cost. In Figure 3 it is
demonstrated how the combined effect of external fire rating and internal cascading
protections works to limit the heat transfer rate, thereby reducing the overall water
requirement. Overall, DNV GL found that it is conceivable that water flow requirements
would not exceed what is already seen in the built environment when appropriate room
volumes are considered (compare Table 10 with Table 8).
Question: Will the ventilation rates for battery systems be excessive?

Finding: They need not be excessive if the appropriate room volume is considered.
(Compare Table 10 with Table 8.)
Question: What are the differences in safety considerations for outdoor vs. indoor
systems?
Finding: Please cross reference to “considerations for System Types and Locations”.
Outdoor systems may have standalone safety equipment such as fixed suppression systems
and self-contained heating, ventilation, and air conditioning (HVAC). The risk to the site
should be considered in all cases, which is intended to be addressed in the FMEA required
by UL 9540. A risk analysis should guide stakeholders toward a probable risk consideration
during project commissioning. This probability-driven analysis helps avoid over- or under-
prescribing safety systems. Outdoor systems may have different or lower ventilation
requirements, but their size and proximity to inhabited structures may dictate heat or plume
considerations in the event of fire (see Figure 25 and considerations in” Present Day
Industry-Accepted Safety Practices for Energy Storage Project” on page 31). Indoor systems
may be dependent on the building infrastructure for ventilation and fire suppression. If that
is the case, the risk analysis should identify if these systems are adequately sized, using the
guidance identified in Table 9.

January 18th, 2017
Consolidated Edison
Question: Are residues left behind after a battery burns?

Finding: There is potential for residues. DNV GL found traces of vanadium after boiling the
vanadium redox electrolyte. In the EDS (energy dispersive spectroscopy) scan from the
scanning electron microscope, coincident detection of both V and O could indicate vanadium
oxide dust (Figure 13). The vanadium peak is low; however, there is no other component of
the test that would contribute it other than the vanadium redox electrolyte. Oxygen can also
be sourced from various oxides that form on metals. In addition, some Pb residue was
swabbed from the burn container where Pb acid batteries were tested, but it was in low
amounts and limited to immediate proximity of the burn specimen. Traces of metals were
observed in the interior of the battery abuse chamber after Li-ion testing. In addition, the
pH of runoff water from the module burn tests was measured to be anywhere from pH 6 to
pH 11. However, many of the same contaminants found from plastics fires were common to
those found from battery fires. In any case, the precautions recommended for PPE and self-
contained breathing apparatus (SCBA) during overhaul apply to solids residues and dusts as
well. Bare skin contact with residues should be avoided, as is good practice in the aftermath
of most fires.
Figure 13 Residue analysis from a coupon hung in the headspace of the vanadium
redox boiling test.
Question: Are certain form factors of cells safer than others?

Finding: DNV GL saw that unconstrained pouch cells, if given the opportunity, will inflate
and then burst catastrophically under extreme heating conditions (Figure 14). However,
pouch cells are compressed when engineered into modules, so a free-floating pouch cell is
not a realistic representation of a field system. DNV GL did notice, however, that controlled
venting of cells is necessary to reduce their volatility. The ability to vent and relieve
pressure is critical to whether the cell’s failure is benign or sudden. This illuminates the
fact that trapped gases are the cause of explosive failure. It should be noted that
DNV GL did not directly witness any exploding battery cells during testing. However,
flashovers of the contained gases within the test chamber were a frequent occurrence for all
Li-ion batteries tested.

January 18th, 2017
Consolidated Edison
Figure 14 Percent of mass loss as a function of cell form factor.
Question: Will Li-ion batteries explode?

Finding: In this program DNV GL tested dozens of Li-ion batteries and could not
conclusively say that any of them “exploded.” DNV GL has conducted hundreds of abuse
tests on cells in other programs and has not conclusively observed an event where a battery
exploded or was the source of a rapid energy event. What is a highly repeatable condition,
however, is the degree to which the test chamber fills with flammable gases before those
gases ignite. The flashover event could be very rapid. The explosion hazard is not the
battery itself, but the gases it may generate. Therefore the requirements for stress-relief by
venting of the cells (described above) and the ventilation of the space are emphasized
throughout this report.
Question: How long would it take for flammable gases to explode?

Finding: This is entirely dependent on the emissions rate and the ventilation of the
surrounding area. It is shown in Figure 5 that the emissions rate varies for all batteries but
the diagram indicates the upper and lower boundaries of emissions limits. The emissions
rates corresponded to 0-57% mass loss over a period of 13-83 minutes. CO is the primary
signature of flammable gases. Sensors detecting CO may be cross-sensitive to hydrogen.
Many flammable gas sensors are non-equally cross sensitive across a family of
hydrocarbons and provide a general “LEL” audible warning. The emissions rates observed
from batteries are included in Table 1.

January 18th, 2017
Consolidated Edison
Table 1 Statistics on cell failure rates for the entire test program including all Li-ion
variants.
Single Cell Emissions Statistics

Average Std Dev Min Max
Mass Loss 18% 14% 0% 57%
Duration 41.7 17.1 13.0 83.0
(min)
Average ppm per kg per min in HCl 0.057 0.150 0.000 0.719
1 m^3
HF 0.009 0.010 0.000 0.032
HCN 0.003 0.005 0.000 0.027
CO 0.279 0.440 0.000 2.341
Question: What is the energy of the explosions from battery offgas?

Finding: DNV GL did not observe batteries exploding directly, but did observe the energy of
flammable gas flashovers. The energy of these events is proportional to the concentration of
gases in the enclosed volume. The power of these events (or the heat release rate) is
significantly variable depending on the volumes of gases, the duration of their release, the
resulting mixture, and the rates of their ignition, DNV GL observed considerable scatter in
the HRR (Figure 15). The HRR was observed to be anywhere from 2-8 kW with 100-800 g of
released materials. This brackets the value from 2.5-80 kW/kg. By comparison, burned
specimens of common furniture items have demonstrated a mass weighted HRR of 32-260
kW/kg. [51] It was found during testing that long periods of smoldering for the batteries
resulted in reduction in mass prior to the peak event, which likely produced much of the
scatter observed in the measurements.

January 18th, 2017
Consolidated Edison
Figure 15 Relationship of heat release rate (kW) per gram of mass lost.
Question: Is the ventilation rate governed by the LEL or Immediately Dangerous to

Life and Health (IDLH) limits ?
Finding: IDLH. The concentrations of HCl reach a threatening level much faster than the
concentrations of flammable gases. Therefore by sizing the ventilation requirement to the
IDLH of HCl, the flammability concern is also mitigated. See Figure 16 and related figures
starting on page 23.
Question: What are the ventilation requirements for batteries?

Finding: DNV GL quantified and produced suggested ventilation rates in Table 9. The
suggested ventilation rates range from 0.02-0.03 CFM/kg or 0.2-0.32 CFM/kWh. This
translates to roughly 0.25 ACH in many cell failure scenarios, climbing to as high at 10-14
ACH in the worst case (see Figure 16 on page 23 and related figures). It should be noted
that laboratory spaces, pharmacies, or some manufacturing environments can also have an
ACH of 10 or higher (compare Table 15 and Table 8). Therefore, the ventilation rates in
most buildings will meet or exceed the ventilation required for the battery system in single
cell or low mass failure modes.

January 18th, 2017
Consolidated Edison
Table 2 Average release rate for battery materials over a 30 minute time period.
Materials 30 min Release

Rate (kg/s)
HCl 2.36E-07
HF 1.74E-07
HCN 1.74E-07
CO 2.00E-07
Figure 16 Estimated ventilation rates (air changes per hour) as a function of room
volume and mass of battery undergoing failure for HCl.
Figure 17 Estimated air changeover rate (air changes per hour) as a function of
room volume and battery mas undergoing failure for HCN.

January 18th, 2017
Consolidated Edison
Figure 18 Estimated air change over rate (air changes per hour) as a function of
room volume and battery mass undergoing failure for CO. Because the IDLH of CO
is much higher, there is little dependency on battery masses at these scales.
Figure 19 Estimated air change rate (ACH) as a function of room volume and
battery mass during failure for HF.
Question: Is HF emitted from batteries?

Finding: Yes. HF was observed in all of the Li-chemistries. Vanadium redox also
demonstrated HF emissions in 2 out of 3 tests, even after a complete overhaul of the test
equipment to remove the possibility of contaminants affecting the result (see Figure 20 as
well as Figure 5). However, it is HCl, not HF, that governs the ventilation and toxicity
consideration. It was found that on a per kilogram basis, the average emission rate of HF in
a plastics fire can be higher than the average emission rate of a battery fire (compare
Figure 4 to Figure 5). From this study it was found that the ventilation requirements for

January 18th, 2017
Consolidated Edison
anything less than 15 simultaneously burning battery cells are the same for HCl and HF (see
Figure 16 and related figures starting on page 23).
Figure 20 Representative emissions histogram from a Li-ion battery.
Question: Is the combined LEL of the flammable gases lower than any of the gases
alone?
Finding: Yes. This phenomenon is described by Le Chatelier's Mixing Rule which states that
the combined LEL of a mixture of gases is the sum of the weighted ratios of volume to LEL
for each individual gas species. Because the emissions rates are constantly varying and
therefore never in a prolonged chemical equilibrium such that this simplified textbook
solution may apply, DNV GL was able to observe that ignitions occurred as low as 400oC at
CO concentrations as low as 3,000 ppm. (Figure 21) Frequently observed gases of C2H4, CO,
and CH4, if coexisting in a mixture, have the lowest autoignition temperature of 490oC and
100,000 ppm, respectively (see below):
- C2H4 = 2.7% (27,000), 490oC

- CH4 = 5% (50,000), 537oC
- CO = 10% (100,000), 609oC
As expected the combined LEL is indeed lower than the individual components but as
mentioned above, the ventilation requirements should be set by the IDLH, which
should exceed and override LEL considerations. Therefore LEL is less of an
immediate concern than IDLH.

January 18th, 2017
Consolidated Edison
Figure 21 The combined LEL and autoignition temperature of mixed gases emitted
from Li-ion batteries may be as low as 3,000 ppm and less than 400oC, as
flammable gases were ignited and burned off above this temperature.
Question: What is the explosion risk?

Finding: The battery is not the source of an explosion risk, but the flammable gases
generated from it are. These gases need to be vented to reduce the risk. Because the
ventilation rates are dictated by the lower IDLH thresholds than the LEL thresholds,
ventilation sized to the IDLH should exceed the ventilation requirement for explosion
hazards.
Question: Are Li-ion batteries more volatile with higher states of charge?
Finding: Yes. There is a very direct increasing relationship between mass lost and the SOC
before failure as shown in Figure 22. However, the BMS limits the SOC of the battery
intentionally for both longevity and safety reasons. As shown in the figure, the decline in
mass loss is significant as the SOC of the battery is decreased from 100% to 90% or 80%.
As many battery systems limit the upper electrochemical SOC range to 80-90%, a
significant safety precaution has already been made. It should be noted that the GPM/kg
and CFM/kg metrics found in this program are inherently conservative because they include
the peak emission rates observed at 100 % SOC and they also capture the short lived peak
emission events. In reality, a system fire spends most of its time smoldering, and if the BMS
is properly functioning, no cells should be at 100% SOC.

January 18th, 2017
Consolidated Edison
Figure 22 For Li-ion batteries, the mass loss is directly proportional to the state of
charge prior to failure.
Question: Are some battery chemistries safer than others?

Finding: No battery tested in this program is excluded from toxicity concerns in a fire. In
general, it is good advice to treat a battery like any fuel should be treated, and make note
that risk is context specific and weighted. In Figure 23 it is evident that iron phosphate, BM-
LMP, and titanate batteries have lower heat release rates and less flammability, as does
vanadium redox and lead acid which did not exhibit flammability. However, it was shown in
Figure 5 that all batteries have varying degrees of emissions of HCl, HF, CO, HCN, and
potential SO2 and H2S. Because many of them have plastic casing, the plastic itself is a
toxicity and flammability hazard. Therefore, there is no single battery chemistry in this
testing program that should be excluded from toxicity considerations in an enclosed space
or near a populated building. Furthermore, the source of toxicity may be as much plastic
componentry as it may be attributed to electrolytes. Because the toxicity risk is similar
to plastics, it is DNV GL’s recommendation that toxicity be treated equally across
chemistries. In the case of batteries with non-flammable electrolytes, adequate
precautions should be demonstrated that polymer cases or other flammable materials are
sufficiently protected against external fire in order to warrant any reduction in the water
requirement, if any. It should also be noted that the water requirements for Li-ion batteries
need not be excessive if the fire safety measures are viewed as a system rather than
standalone requirements. Lastly, it should also be noted that the low level ACH
requirements for vanadium redox and Pb acid are well below the typical 2-4 ACH ventilation
requirement in most occupied spaces, so the existing infrastructure may be adequate in
many instances.

January 18th, 2017
Consolidated Edison
Figure 23 It is generally true that LiFePO4, LTO, and BM-LMP batteries demonstrate
lower than average temperatures during failure. The temperatures indicated for
Pb acid and vanadium redox batteries is the peak heating temperature, as these
electrolytes did not demonstrate flammable or exothermic properties as they were
tested.
Question: What is the solubility of liberated gases, and are some of them
consumed by fire?
Finding: The solubility of the gases observed is shown in Table 3. Those consumed by fire
have an indicated flammability limit and autoignition temperature.

January 18th, 2017
Consolidated Edison
Table 3 Inventory of Toxic and Flammable Hazards found in this Study

Concentration (ppm unless NFPA Codes
otherwise noted) (F=flammability,
H=health,
R=reactivity,
S=special)
Chemistry Relevant Detected LEL (Lower IDLH Solubility Auto Ig. F H R S Ref.
Batteries State Explosion (Immediately in Water Temp
Limit) Dangerous to (mg/L) (°C)
Life and
Health)
Methane CH4 Li-ion Gas 50,000 5,000 22.7 537 4 1 0 NJ DOH
Carbon CO All Gas 12,500 1,500 27.6 609 4 2 0 CDC.gov

Monoxide
Benzene All except Gas 12,000 3,000 3 2 0 CDC.gov
PbA
Ethane Vanadium Gas 30,000 4 1 0 CDC.gov
Redox
Ethylene C2H4 Li-ion Gas 27,000 - 2.9 490 4 2 2 Matheson
MSDS
Hydrogen H2S Pb Acid, Li- Gas 40,000 - 4 0 0 CDC.gov
ion
Hydrogen H2S VR, PbA Gas 4,000 300 4,000.0 260 4 4 0 CDC.gov
Sulfide
Hydrogen HF All except Gas - 30 miscible - 0 4 0 CDC.gov
Fluoride PbA
Hydrogen HCl All Gas - 100 720.0 - 0 3 1 CDC.gov
Chloride
SO2 SO2 VR, PbA Gas - 100 94,000.0 - 0 3 0 CDC.gov
Hydrogen HCN All except Gas - 50 miscible - 4 4 2 CDC.gov

Cyanide PbA
Nickel Ni Li-ion Residue / 1 3 0
Powder
Manganese Mn Li-ion Residue / 3 3 3
Powder
Cobalt Co Li-ion Residue / - - Insoluble 0 1 0
Powder
Lithium Li Li-ion Residue / 2 3 2 W
Powder
V2O5 Dust V2O5 VR Residue (V) - 35 mg/m^3 0.8 - 0 3 0 CDC.gov
Pb Vapor, Pb PbA Residue - 700 mg/m^3 10^-5 to - 0 2 0 CDC.gov

salts, dust 4400
Question: Can batteries be “neutralized” by immersing them in water after an

incident?
Finding: Partially. Immersion in water provides adequate cooling to prevent violent thermal
runaway, but it may not neutralize voltage. DNV GL found the following results should be
considered prior to doing so:
- Batteries may have residual voltage on damaged and exposed terminals.
(Figure 11) Handling of the battery may produce a shock hazard.
- Batteries persistently gassed even under water. The primary measured
component of that gas was CO, though the handheld CO sensors are cross
sensitive to H2.

January 18th, 2017
Consolidated Edison
- For most tests the water runoff was slightly acidic measuring pH 6-7. In one
case, however, the water became alkaline climbing to pH 10-11 after a few
hours of submersion. This case was observed for a battery that was highly
consumed in the fire.
- Batteries did not climb in temperature after submersion, indicating that even
if cells short circuited, their temperature was never permitted to climb to
thermal runaway conditions.
- Some battery cells still had voltage on them after 24 hours of submersion.
While some cells may have shorted, not all shorted. The water did not have
any additives such as salt to make it more conductive.
Question: Was hydrogen generated as a result of electrolysis during submersion?

Finding: Possibly, high levels of CO (10-100ppm) were detected on the four and five gas
meters right above the submersion pools. These electrochemical sensors are cross sensitive
to H2. High levels of CO were also detected on the FTIR during and after testing though,
suggesting that CO generation is real and any cross sensitivity from H 2 is low as CO is the
dominant gas. This was further supported by data from cell testing not involving
submersion.
Question: How much hydrogen was emitted?

Finding: During cell testing DNV GL witnessed > 1000 ppm (sensor max value) on a few
occasions. Hydrogen was not observed directly during submersion, though CO was
measured. CO sensors can be cross sensitive to hydrogen. The lower flammability limit for
ethylene and related species is 3.6%, which is lower than hydrogen at 4%. Therefore the
greater flammability risk is presented by ethylene carbonate decomposition due to its
greater volumes, higher emissions rates, and similar volatility. This is supported by gas bag
sampling, which showed far higher levels of hydrocarbon gases than H2, which was still well
below the LEL.
Question: Are the liberated gases lighter or heavier than air?

Finding: The molecular weight of air is generalized at 29 g/mol. By comparison the
molecular weights of the main gases observed from battery fires are shown below (in
g/mol). It can be seen that HCl is heavier than air. Another observation from testing is that
the gases are typically hot, which means they are rising as part of the plume.
- CO: 28.01
- HF: 20.01
- HCl: 36.4
- HCN: 27.02
Question: Should exhaust fans be intrinsically safe or grounded?

Finding: DNV GL used an exhaust fan during module testing (Figure 24). After several
consecutive fire tests the heat and smoke eventually overstressed the fan. However, it was
not observed that the fan ignited the gases. Consideration of intrinsically safe fans may be
necessary in sensitive locations.

January 18th, 2017
Consolidated Edison
Figure 24 Smoke plume rising through door gaps and out of top vents in the burn
container.
Question: Should exhaust fans be variable speed?

Finding: Based on the nonlinear behavior of emitted gases (Figure 16 and related figures) a
variable speed fan should be a consideration. This may be a more cost effective solution
than a highly rated fan running continuously which may exceed the minimum ventilation
requirement of ~0.25 ACH. A variable speed fan can accommodate the low level ventilation
rate for the majority of the time, with the capability to ramp up in the event of failure.
Question: How were gases measured in this testing?

Finding: DNV GL used a Gasmet DX4000 FTIR gas analyzer during cell testing supported by
MSA Ultima sensors for IR transparent gases and LEL. The FTIR was used again for module
testing. In addition, for module testing, Rescue Methods used MSA Altair handheld four- and
five-gas sensors. These tools were used for both LEL and toxicity monitoring. LEL was
measured via a photo ionization detector (PID) (10.6eV bulb) on the handheld sensors.
7.0 PRESENT DAY INDUSTRY-ACCEPTED SAFETY PRACTICES

FOR ENERGY STORAGE PROJECTS
It is important to place this report in the context of what is actually occurring in energy
storage project development today. Presently there are over 400 stationary storage systems
comprising 1,200 MW operating around the world, with 600 MW of electrochemical energy
storage in the United States [5].
The types of battery energy storage systems being deployed are both utility solutions at the
multi-MW scale in consolidated sites, typically with energy storage batteries housed in

January 18th, 2017
Consolidated Edison
shipping-container like systems with integrated BMS, ventilation and cooling, and fixed fire
suppression. Smaller, behind the meter energy storage systems are designed to be
deployed near the customer and controlled as an aggregate fleet. These smaller systems
have a BMS and may have active cooling, but rarely have integrated fire suppression.
Shipping container systems are typically located outdoors and are MW scale, whereas
behind the meter systems are typically sited at a commercial site (or potentially residential)
and may be indoors, and will have ratings in 10’s of kW.
Energy storage can be utility owned or it can be owned by an independent power producer
(IPP). Much of the US energy storage market is presently being driven by IPPs. The IPP may
monetize the energy storage asset through utility contracts or a commercial power purchase
agreement. Some IPPs have the balance sheet to pay for energy storage projects
themselves, but many seek financing. With financing comes insurance to underwrite risk in
both the finance and safety of the project. Because of these additional parties that are
exposed to financial risk, a performance and safety review are a critical piece for
financing an energy storage project, which is performed by an independent
engineer (IE).
Independent Engineering is a field of service where independent third-party engineers

review the technical specifications of energy projects and provide an assessment of financial
or technical project risk. The practice of hiring an IE is common in the wind and solar
industries and is now industry practice for energy storage projects. Many insurers and
lenders require an IE report – and must feel comfortable with the findings of the report – in
order to finance or underwrite an energy storage project. The first step in most IE reports
on storage is a review of the technology which will include performance and safety aspects.
The IE functions are typically performed during or prior to permitting and before project
commissioning, as shown in Figure 25.
In the context of safety the IE is tasked with independently evaluating the adequacy of
safety systems appropriate to the project. With the business case and project site(s)
identified, the AHJ is likely to become aware of the project when the project developer is
seeking permit(s). The AHJ will typically respond with requirements, which may be few or
many, at which point the project developer takes actions to fulfill them in order to secure
the sites as quickly as possible. Project finance may be secured or will be sought in parallel
to this process. Because the project developer is encouraged to obtain the IE report by the
financial stakeholders, it is most cost effective for the project developer to use the IE report
to simultaneously satisfy requirements for the AHJ and the financial parties. The objectivity
of the report should increase comfort in the transaction(s) between parties. Therefore at the
request of the project developer, the IE report is written in the context of generalized
project specifications so that it may enable as many transactions as possible. Therefore it is
never the case that more than one IE report is generated for identical projects because it is
a cost that can be practically avoided. It is also common practice for large aggregated
projects of similar system types to have inspections performed on a subset of sites.

January 18th, 2017
Consolidated Edison
Figure 25 Project development timeline and the implementation of FMEA or other

safety review for the site.
For an outdoor container system, the IE may provide the following services related to
safety:
 Assessment of the adequacy of the safety systems
 Assessment of the safe perimeter around the site
 Emergency response plans
 Review or recommendations of materials to be provided to the local AHJ
for permitting or code review
 Adequacy of firefighting equipment
 Impact of a fire scenario on the site or surrounding area, which may
include a plume study if residential or populated areas are nearby
 Risk model for the site
For an indoor system, the following IE services related to safety may be requested:
 Review of safety testing
 Assessment of the adequacy of safety systems
 Recommendations on the requirements for indoor room locations
 Adequacy of cooling and venting
 Review of fire ratings
 Inspection of installation
 Risk analysis related to the system and its site(s)
In the case of behind the meter systems, this review is usually done at the project
portfolio level unless specific site considerations require local review.

January 18th, 2017
Consolidated Edison
8.0 GUIDANCE FOR FIRST RESPONDERS

Many of the questions in the FAQ were intended for first responders, who wish to know what
should be done upon encountering a fire that involves a battery.
The first and foremost finding from this report is that the equipment available to present
day first responders can be considered adequate for battery fire fighting with
additional considerations.
8.1 Considerations for Permitting and Siting

If a building or site information summary is available, it should state whether there is a
battery on site and its chemistry. The primary concern upon approaching the scene should
be HCl toxicity and rising temperatures, and the potential for the fire to expand if it has not
already.
 FMEA, siting, and standard operating procedure (SOP) development: UL 9540
requires an FMEA for ESS permitting and siting. In addition, DNV GL recommends an
FMEA be performed on any system or project portfolio, particularly for behind-the-
meter applications.
o A risk analysis involves review of all potential failure modes for their likelihood
of failure and the resulting consequence to determine the total risk. As this
process serves as a deep dive into the design and operation of the unit, this
process would provide valuable insight for code officials and first responders
to better understand the risks and potential faults they may be dealing with
during emergency situations.
o Requesting participation in this process would serve as the best opportunity
to become involved in the development process and would allow AHJs and the
fire service to best understand the system in the least intrusive way to the
project developer (since an FMEA may be required regardless of AHJ
participation).
o In addition to FMEA involvement, DNV GL recommends all AHJs and fire
departments perform a walk through for all large ESS in their jurisdictions
and develop SOPs according to their level of comfort with the electrical risks.
Though small home systems may not exceed 48 VDC and be easily
disconnected from the AC source, larger utility scale systems may exceed
1,000 VDC and 10,000 VAC. Again, even prompt disconnect of AC voltage
does not eliminate voltage on the DC side.
o DNV GL recommends all fire departments with large ESSs or ESS portfolios in
their jurisdictions work with project developers or system manufacturers to
provide emergency contacts and readily available subject matter experts
(SME) who can quickly advise fire departments on system status and risks
associated with the current fire environment.
o Finally, DNV GL recommends fire departments and first responders work with
system and project developers to understand the level of risk and their
appropriate response. A single cell failure in a large containerized system
need not require the entire system be destroyed with water. However, a

January 18th, 2017
Consolidated Edison
system with an unknown internal hazard may pose risks to the surrounding
environment or to fire fighters and may be better handled via a defensive
posture than entry and attack.
8.2 Considerations for Operations at the Scene

Upon arriving at a fire scene, the following considerations should be made:
- Has on-site extinguishing already been triggered?
- Is the system gassing?
- Is the temperature of the system rising?
- Are flames visible?
- Is there a site representative or SME available?
Answers to the above questions will indicate whether the system fire has already peaked or
if it is expanding. Support from an SME, an information display panel, or other form of
emergency contact will greatly aid in assessing the risk.
If the system is gassing but onboard suppression (if any) has already triggered, and
temperatures are remaining stable, it is likely that a single cell or module fire has occurred
and been isolated, and may have been managed by the onboard system. Additional
suppression may not be required in this case. Eventually, the system will need to be
ventilated to remove the internal atmosphere, but only if temperatures have remained
stable for approximately 60 minutes.
The list below summarizes key points from this study that are directly relevant to
firefighters and other first responders. This section may stand on its own as an independent
part of this report and may be distributed to fire departments and first responders
nationwide independent of this document. It is not intended to serve as an SOP on its own,
but should inform the response and development of SOPs for situations involving ESS.
There has been much said about ESS fires in the past which has led to several myths about
these fires. DNV GL wishes to dispel the falsehoods while promoting real world, data driven
facts when dealing with these systems. Ultimately, findings suggest that while these
systems are unique in the combination of threats posed, none of the threats on their own
are unfamiliar to firefighters, and they remain manageable so long as certain points are
known and followed.
 Toxicity: In general, battery fires resemble plastic fires in terms of emission of toxic
gases including CO, HCl, HF, HCN, Benzene, and Toluene
o The average toxicity of the fire is equivalent to many plastics on a per mass
basis. Li-ion fires will have short peaks of toxicity as individual cells randomly
fail.
o However, battery fires, even once extinguished, continue to emit CO as long
as the batteries remain hot.
o DNV GL and Rescue Methods (RM) recommend continued monitoring of CO
from ESS fires, especially in enclosed spaces, and the continued use of
personal protective equipment (PPE), including self-contained breathing
apparatus (SCBA), until CO levels are shown to be at normal levels. These
practices may include monitoring for HCl, if applicable or possible.

January 18th, 2017
Consolidated Edison
 Ventilation: Though integrated ventilation will be recommended for indoor systems,

it may not always exist or prove adequate to remove heavy smoke, especially in
cases where the surrounding environment is fully involved or the battery is rapidly
overtaken.
o DNV GL and RM recommend sufficient firefighting ventilation, ideally negative
pressure, to remove fire gases from enclosed areas.
o The batteries themselves emit flammable gas and fully involved or improperly
ventilated systems may pose a lower explosive limit (LEL) or flash hazard.
o DNV GL and RM recommend monitoring of LEL levels in the fire ground and
surrounding environment to determine if intrinsically safe ventilation is
required.
o Partially burned systems may continue to emit flammable gas even after the
fire is extinguished as long as the cells remain hot. Proper cooling of the
system is key to remove prolonged fire risks.
 Temperature: Climbing temperature is an indicator of increasing risk.
o If flames are visible and temperature is rising, the system may have more
than one battery cell or module engulfed.
o If temperatures are rising rapidly (>1 °F per minute) and temperatures on the
battery are approaching anywhere near 100 °C (212 °F), cooling will be
required with water.
o Monitoring with handheld infrared (IR) thermometers, if available, should
provide an assessment of risk.
 Delayed Cascading Ignition: On site responders should assess that all thermal,
electrical, or mechanical stimuli that may act on the system have been mitigated.
o In the short term, when cells appear to “reignite” after seconds or minutes, it
is almost always a result of incomplete removal of heat from the system, or
an electrical short due to liquids or water. Prevention of cascading between
cells may be addressed by proper cascading protections in the system, which
may retard extinguishing and external cooling but also mitigates the free
movement of heat internally in the batteries which can ignite previously
undamaged cells. DNV GL refers to this phenomena as delayed ignition.
o In some cases, the only way to halt this process is to let the system burn
itself out (but this may not be practical) or continue to drown the battery until
this process stops as the battery finally cools. This decision should be made
based on the circumstances of the fire ground.
 Shock Hazards: Cells that have not been burned may remain intact in systems and
modules.
o Shock during water suppression (via conduction into the water spray) was not
observed in this program.
o Beware of arcing if batteries are disturbed. Turnout gear was observed to
provide shock protection under the conditions tested in this program, but do
not touch arcing equipment.
o Stranded energy in partially burned batteries will likely remain an issue in any
system that is extinguished unless it has consumed itself entirely. DNV GL

January 18th, 2017
Consolidated Edison
found that even in systems that appeared thoroughly damaged, live cells and
stranded energy can remain. These cells may pose a shock/arcing risk and
can reignite if physically damaged, reheated, or allowed to short.
 Extinguishing: DNV GL tested several water based extinguishing agents and found
none to be as effective for cooling as water. These included PyroCool, F500, and
FireIce.
o The most challenging aspect of the battery fire is its deep-seated nature.
Access to the heat source is necessary to provide adequate cooling.
o Cooling the battery once flames are knocked down is the most important
aspect of containing battery fires. The tested agents proved slightly less
effective than water at cooling the cells. On a module level, there was no
evidence to suggest these agents perform better than water.
o Because many encapsulating agents, including foam (AFFF) are intended to
blanket the fire, and a battery fire needs to have heat removed as quickly as
possible, DNV GL generally recommends against using foam for ESS fires.
Foam has been tested in other projects and used in real world ESS fires. In
testing in other projects, it failed to perform better than other agents.
o The aerosol may prove effective at knocking down flames from ESS. Gas
based agents may suppress the flammability of contained atmospheres with
high explosive gas content; however, in the case of severe ESS fires where
these agents would be tasked to suppress flammability, cells may be
producing heat above the autoignition temperature of their flammable gases.
This may result in fire if oxygen were reintroduced to the system. DNV GL
recommends gas-based systems be backed up by water-based suppression
when cooling becomes a necessity, in combination with cascading protections
in the modules and systems.
o Though water proved most effective for cooling, water and any water-based
agent introduces shorting risks when applied on a full system. This may
exacerbate the situation in addition to presenting a collateral damage risk.
o Several entities, including DNV GL, have advised that class D fire
extinguishers and agents be investigated for use during the incipient stage of
the fire. Based on the findings from this program, DNV GL views the
deployment of classical class D agents as impractical due to the short lived
peak of a cell fire and its deep seated nature, which prevents direct access.
o RM’s experience during suppression testing suggests forced access to the
interior of battery systems may be difficult or inadvisable for first responders.
In this case, water should be used to provide indirect cooling on the outside
of the system to prevent spreading.
o Water use inside the system, if applicable, should be done with care to avoid
shorting neighboring and surviving cells, i.e., the failing module should be
isolated and targeted. Fully involved systems may be compromised enough to
allow better water penetration. Fully involved systems posing a risk to
surrounding life and property, or neighboring systems, should be suppressed
immediately and heavily to avoid spreading.

January 18th, 2017
Consolidated Edison
o If the fire appears to be stable and not expanding, periodically stop water
flow and monitor temperatures. Note that the temperature may “spring back”
after water extinguishing stops, but it should plateau and stabilize if the fire
fuel has been consumed.
o Observe for water shorting other cells. They may begin to heat, meaning the
deep seated heat remains.
o Repeat extinguishing process as needed, while ventilating the area as much
as possible.
o If the battery system has closed doors, do not open them unless absolutely
necessary or it has been determined that opening the doors will not introduce
new hazards. Forced entry is discouraged unless a prior access plan has been
described.
o Suppression of large, fully involved systems may take more time than fires of
similar size with different fuels. It is recommend fire service personnel
continue to suppress with water for as long as required and then ensure the
system is fully cooled throughout once suppression appears complete.
8.3 Guidance for Isolation and Overhaul

After burning, the removal and isolation of the batteries demonstrated real-world hazards
that may be encountered in the overhaul stages of fighting a battery fire. Residual live DC
voltages in intact battery cells, and damaged but still live bus bars within modules after a
fire represent an electrical shock hazard (see Figure 11). During testing, it was found that
firefighters were not shocked while wearing standard turnout gear when arcs and sparks
resulted from disturbance of the debris. For this reason it is recommended that
whenever possible, first responders need not open or otherwise disperse burned
battery modules and wait for an experienced liaison to arrive on site and take
ownership of the site after extinguishing has been achieved.
As shown in Figure 26, submerging battery modules in water provided adequate cooling to
slow and prevent delayed cascading thermal runaway in the remaining battery cells;
however, the batteries persistently off-gassed even under water. The primary gases
detected in the bubbles generated were CO and possibly hydrogen. The figure demonstrates
the bubbles observed even after submersion for over 30 minutes.
Even after submerging, some batteries generated a severely alkaline solution climbing to pH
10-11. Other solutions gradually became slightly acidic (pH 6). There was not a clear
explanation for the pH behavior of the solutions, other than one of the most severely burned
batteries created the most basic solution. Therefore, if water submersion is used by
first responders for isolating spent modules, preparation to deal with alkaline or
basic water for disposal should be a consideration.
Lastly, it was found that after extinguishing the persistent emission of CO was sometimes in
quantities large enough to trigger threshold alarms on the gear worn by fire fighters. The
persistent emission is perhaps a more insidious risk than the emissions during the fire, as
the apparent climax of the fire has passed, and first responders may be inclined to remove
their masks. After extinguishing, continued ventilation and monitoring of the area
with gas monitors is highly recommended.

January 18th, 2017
Consolidated Edison
Figure 26 Submerging batteries in water resulted in cooler temperatures, but slow

shorting and persistent CO generation (bubbles).
As the climax of the fire has concluded, but continued ventilation and monitoring is
underway, first responders are left with the final challenge of determining when they can
relinquish control of the area. There are several risks that first responders wish to avoid and
they are prioritized by the list demonstrated on page 57. It is highly recommended that
battery systems installed within buildings have an established emergency contact
list and a SME who can arrive on the scene to take over containment, cleanup, and
eventual disposal of damaged battery equipment. This recommendation requires
involvement from the project development and systems integration community.
This is a necessary risk transfer procedure to mitigate the first responder concern that they
are responsible for damaged battery systems for hours or days after they have been
involved in a fire or catastrophic safety event.
The following summarizes recommendations for overhaul procedures:

 Overhaul and Stranded Energy: As mentioned, stranded energy in the surviving
cells remains a risk to first responder during overhaul and post fire operations.
o Live or damaged but surviving cells may contain voltage that will cause arcing
when shorted by debris or metal tools. This arcing may also serve as an
ignition source to localized gases if hot batteries are still venting.
o Firefighters should thoroughly avoid penetrating, cutting, or otherwise
damaging batteries in the ESS, especially during overhaul, as live cells that
are physically damaged or penetrated are subject to rapid venting.
o Firefighters should avoid blindly reaching into cabinets to remove damaged
batteries as DC energy may still remain active even if AC and site power is
cut. It was observed that typical turnout gear provided adequate protection

January 18th, 2017
Consolidated Edison
against shocks in this testing; however, high voltage DC may penetrate PPE in
cases where it is damaged or otherwise compromised, such as a torn glove or
a exposure to sharp metals. These kinds of hazards were not studied in this
testing program.
o DNV GL and RM recommend fire fighters continue to wear PPE and SCBA even
during overhaul as CO levels may remain elevated even after flames are
extinguished as batteries remain hot and continue to offgas. DNV GL
recommends CO levels, especially in enclosed or unventilated spaces be
monitored and SCBA worn until levels are shown to be safe.
o Complete submersion of damaged batteries in water provides cooling for
damaged batteries; however, batteries continued to offgas CO. Because
handheld sensors are cross sensitive to CO, H2 may have also been emitted
while submerged. In addition, this did not always entirely neutralize the
voltage on surviving cells. However, cells seem to remain stable once pulled
from water and dried. Caution should be exercised when removing damaged
batteries from enclosure/containment per the risks discussed above.
Whenever possible, a relevant subject matter expert from the site, project
owner, or manufacturer should provide guidance or control of removal.
9.0 FINDINGS RELATED TO CODES AND TRAINING

The following summarizes key recommendations from the report study. The findings are
sorted in their relevance to sections of precedent codes.
9.1 Fire Rating

DNV GL testing has shown that naked cells3 and modules exposed to direct fire are
susceptible to failure within 10 minutes. However, systems deployed in the field, when
exposed to external flame, are likely to sustain much longer durations because of the
shielding and air gap provided by the enclosure since the cells and modules are not likely to
be installed “naked” in an installed system. Because many code precedents such as
those shown in Table 4 and Table 5 require 1-hour fire ratings, and more
conservative precedents require 2-hour fire ratings dependent on height above the
ground floor, DNV GL recommends a minimum 1-hour fire rating with a 2-hour
rating in areas with critical population density, and that the fire rating be
considered as part of a system level approach to avoid cascading fires. Exceptions
to this general rule may include 1-hour requirement for outdoor locations, similar to
combined heat and power (CHP) and backup generator requirements. The finite element
analysis (FEA) model in Figure 28 demonstrates an Abaqus FEA model of a fire impinging on
a generic battery system.
3
Cells not integrated into modules or systems

January 18th, 2017
Consolidated Edison
The model demonstrates the heating effect on a

battery module after 60 minutes of a 1000oF fire
(811K or 537oC) impinging on a steel wall of 1/16”
thickness, with a 1” air gap between the wall and
the nearest inner battery module. For simplicity, the
battery is assumed to be constructed of entirely
aluminum or polypropylene in order to bracket the
low and high temperature scenarios, because many
battery modules are a composite of these or similar
materials. After 60 minutes of exposure the model
predicts the battery temperature to be 84oC for the
aluminum and 231oC for the polypropylene4.
Because a critical temperature for Li-ion batteries is
Figure 27 Simplified diagram of ~120oC, a conservative 2-hour rating on the system
fire impinging on the external metal enclosure would slow heat absorption for the
wall of a battery energy storage worst case polypropylene estimation.
system.
The boundary conditions are a fixed wall temperature of 811K (537oC or 1000oF). Model
components are a steel wall with temperature-dependent conductivity, an air gap (1 in) with
temperature-dependent conductivity, and a composite battery case made of aluminium with
fixed conductivity and polypropylene with fixed conductivity. The heat transfer modes are
natural convection and conduction.
4
This may seem counterintuitive; the aluminum conducts heat away faster and therefore maintains a lower
temperature than polypropylene.

January 18th, 2017
Consolidated Edison
Figure 28 Direct fire exposure model to a steel wall with a 1/16” thickness and 1”
air gap between the wall and battery modules.

January 18th, 2017
Consolidated Edison
Table 4 Non-battery related codes for energy systems in buildings.

Non-Battery Codes
Code Item CHP Backup Diesel Confined Spaces OSHA Flammable
Generator Liquids 1926.152
Ventilation Rated to maintain gas Effective engineering Should be constructed
concentrations below 25% LEL of controls required to keep vapor at or
the fuel gas, or at an exhaust or rather than below 10% of the LFL.
makeup rate equal to 80 times dependence on Shall have pressure
the maximum leakage rate respirators release capability to
relieve pressure during
a fire.
Fire suppression Fully sprinklered Automatic fire Sprinkler, water spray,
sprinkler system or CO2 or other system
approved by nationally
recognized test
laboratory (NRTL).
Monitoring: Detection, Gas detection and alarm in Monitor and display
alarm, display supervised location that potential
hazardous
atmosphere can be
mitigated by forced
ventilation
Capacity limitation 1 MW in dedicated room, Fuel stored 25 gallons outside
dependent on space 0.5 MW in boiler room limited to 10 storage cabinet, 60 or
gallons 120 gallons in cabinet
depending on
flammability category
Clearance 5 ft from other 3 ft wide aisle access
structures
Thermal runaway NA NA NA NA
protection
Fire rating If indoors or in a dedicated Up to 2 hours Compatible with NFPA
room, 2-hour fire rating on 251-1969, 1-2 hour
external walls rating
Location Outdoor, penthouse, boiler Outdoor, Electrical rated for Class
room, dedicated room penthouse, boiler I, Division I Hazardous
room, dedicated Locations
room
Seismic rating Appropriate for zone Appropriate for Appropriate for zone
zone
Accountable parties
FMEA/HMA
Inspections
Signage Flammable, keep away
from open flames

January 18th, 2017
Consolidated Edison
Table 5 Battery specific codes for battery systems in buildings, existing and
proposed.
Battery Related Codes
Code Item IBC IFC 608 NFPA 1 Chapter 52 FDNY Certificate of Fitness (COF) B-29
(Uninterruptible Power Supply)
Ventilation 307.1.1 Rooms shall None for Li-ion Required for Valve Regulated Lead Acid
have ventilation, (VRLA) only, designed to limit H2
batteries shall have buildup to 1% of the entire room
venting caps volume; Continuous ventilation = 1
CFM per ft^2 of room
Fire suppression 403.3 and 903.2 Proposed Chapter 5 Sprinklers not required but
Not required in of NFPA 13. recommended. Portable Class ABC on
external structures hand.
with fire detection
Monitoring: Detection, 907.2.23 Smoke Hydrogen monitoring or handheld
alarm, display detection system detector for COF holder, system health
status
Capacity limitation > 50 gallons 100 gallons of 50 gallons of electrolyte for Pb acid,
dependent on space electrolyte or 1000 electrolyte VRLA, NiCd, or 1000 lbs. for Li-ion
lbs. Li-ion. Proposed (sprinklered) or
20 kWh limit for 50 gal electrolyte
single units, 600 (unsprinklered) or
kWh limit for total 1000 lb. Li-ion
in a room.
Clearance Proposed 3 ft
between arrays no
larger than 50 kWh.
5 ft from lot lines
for outdoor.
Thermal runaway Required Required for both VRLA and Li-ion
protection
Fire rating Table 509 1 and 2
hour ratings
Location Proposed no more
than 75 ft above or
30 ft below fire
access, exceptions
on non-combustible
rooftops
Seismic rating Appropriate for Required for zone Appropriate for Appropriate for zone
zone zone
Accountable parties Equipment shall be under "general
supervision" of certificate holder, in
case of emergency there shall be a
hazardous materials liaison, contact
info available to fire command center
FMEA/HMA HMA required
Inspections Performed by COF holder. Record
keeping on site.
Signage Warning against electrolyte or voltage.
Battery information on Building
Information Card.

January 18th, 2017
Consolidated Edison
9.2 Extinguishing
DNV GL tested a number of extinguishers during cell and module testing. During testing,
DNV GL found that all extinguishers tested5 could put out the fire if applied immediately
upon detection of a thermal spike (indicating the immediate onset of thermal runaway).
While extinguishing was accomplished with all extinguishers, water demonstrated the best
ability to cool and maintain cool temperatures on the battery.
9.2.1 Class D and Deep Seated Fires

During testing, DNV GL witnessed firsthand how residual heat between batteries can lead to
delayed cascading and prolonged extinguishing for battery modules. This highlights the
importance of cascading protections between cells and inter-cell cooling in battery modules.
Cascading protections can be tested by the UL 1973 internal fire test, the IEC 62619
internal propagation test, SAE J2929 propagation test, or similar standards. DNV GL
recommends more stringent criteria such that a single cell failure cannot propagate to
neighboring cells, with the intent of maintaining heat release rates that can be otherwise
managed by the water extinguisher flow rate and the system external fire rating. This
recommendation illuminates that the extinguishing solution and the module design are
interlinked; a module with adequate cascading protection is more likely to be appropriately
designed with a gas-based suppression system.
Because the consumption of a single cell is rapid, the metal fire fuels (Class D) are rapidly
consumed and the fire evolves to Class A, B, or C quickly. Because of the rapid evolution
of a cell fire, DNV GL does not see an advantage to using a Class D extinguisher on
a single cell or system fire. This has direct implications for first responders who are
accustomed to using water as their primary extinguishing agent. In the event of a single cell
fire, cascading protections should limit propagation to other cells. First responders may still
respond to a call reporting smoke, but in the best case scenario the fire has consumed itself
and burned out. If a fixed suppression agent is installed within an enclosed environment
containing the single failed battery cell, it may suppress flammability in the enclosed space.
The use of water may be unnecessary at this point unless the fire has progressed. A key
issue to be addressed in later sections is how the first responder is able to determine if this
single cell fire has been mitigated or if further action (and water extinguishing) is needed,
and hence some system health information, an emergency response phone line, or some
other means to gain information on system health is a need that requires industry
engagement to overcome. The first responder is not comfortable deeming the site
extinguished and is technically responsible for the scene until this information allows them
to make the decision to leave the control of the scene with a responsible party.
9.2.2 Cooling and Collateral Damage

Cooling is a secondary component of extinguishing that has not been previously discussed in
the literature. In 2011, the NHTSA recommended “copious amounts of water” in an official
release concerning the extinguishment of battery fires in hybrid and electric vehicles. [12]
The intent and purpose of this recommendation was to introduce cooling to the fire.
DNV GL found that water extinguishes, cools, and maintains lower temperatures on a
battery fire than other tested agents. As shown in Figure 29, water consistently maintained
a 50-100oC sustained cooling advantage over equivalent volumes of other water borne
5
For the complete list of extinguishers tested, see the Appendix, page 82.

January 18th, 2017
Consolidated Edison
agents in the seconds and minutes following extinguishing. The behavior demonstrated in
the figure is consistent for all battery types, with the heat decay duration, “reheat” period,
and peak temperatures varying as a function of cell mass.
Figure 29 Performance of water compared to other agents as water additives, top

temperature of battery cell.
Figure 30 Cooling performance of water compared to other extinguisher types,

bottom temperature of cell.

January 18th, 2017
Consolidated Edison
The initial cooling rate is nearly equivalent for all extinguisher types, but the thermal mass
of the battery causes the extinguishing agent to evaporate as temperatures climb back to
250-275oC. Extinguishers were triggered the moment the battery fire climbed above 350 oC.
In each case 1 gallon of water was applied. In all instances the total extinguishing time
spanned less than 60 seconds, or about 1 GPM.
The duration of this “reheating” is approximately 200s for non-water agents, whereas water
is shown to reheat for about 100 seconds. Therefore, DNV GL saw no particular cooling
advantage of water borne agents such as F-500, FireIce, or Pyrocool over water alone.
(Figure 29) Some of these agents are encapsulators, which are designed to blanket a fire
and insulate surrounding areas from heat; in an exothermic battery fire, trapping heat is
undesirable. The figure demonstrates that cooling with water persistently achieves lower
sustained temperatures after extinguishing, with as much as a 50-100oC advantage within
1-2 minutes of extinguishing (See appendix on page 76). This data demonstrates that water
and all water borne agents reduce cell temperatures from > 400 oC to near 50oC within 10-
30 seconds. Water can maintain cell temperatures after extinguishing below 100oC even as
the initial mass of water evaporates.
An additional vendor provided an aerosol agent to be tested. The aerosol was observed to
extinguish the cell fire. The aerosol provides some initial cooling but does not reduce cell
temperature until the exothermic reactions of the battery begin to decay. It was shown that
the cooling ability of the aerosol was significantly less than water.
Because cooling is an inevitable need, a fixed suppression gas agent may reduce
or mitigate flammability in an environment until ventilation and/or cooling
strategies are implemented.
While the use of water demonstrates excellent cooling capability, it also potentially shorts
out undamaged cells or neighboring modules. The use of water is a fully committed
extinguishing tactic that is highly likely to result in a total loss of the asset.
Because it was noted that the aerosol test demonstrated extinguishment of the fire upon
execution, aerosols can potentially serve as an initial attack for the fire followed by water as
a backstop.
Therefore, DNV GL recommends the following:

 Stage 1: If a system can limit cell cascading, a gas based suppression
system may be considered for the first stage of fire fighting to extinguish a
single cell fire and prevent flashover in a contained environment.
 Stage 2: If temperatures continue to rise or if an increasing level of smoke
and gas is detected, forced ventilation and water extinguishing should be
considered to cool the system and prevent further propagation of fire.
Stage 1 provides an opportunity for avoiding collateral damage and total asset loss. Stage 2
provides a backstop for a situation when more than one battery cell is on fire. Both stages
may also include some form of alarm or notification external to the battery system that
notifies first responders of elevated risk.

January 18th, 2017
Consolidated Edison
9.3 Locations and Ventilation

DNV GL quantified that the gases emitted from a battery fire have somewhat differing
toxicity and flammability risks across chemistries. However, mitigation of toxic or flammable
gases is addressed with ventilation in all cases.
9.3.1 Outdoor Locations

Toxicity of the fire should be modeled to account for the impact on neighboring areas. The
fire may be modeled in scenarios of increasing severity, such as a single cell fire of short
duration, a module fire of short and long duration, and a total system fire.
The probability of fire, size of the system, plume contents, proximity of nearby buildings,
wind direction, and duration of the fire will have an impact on the location of fencing and
safety perimeters. It is the discretion of the project owner to consider these hazards.
DNV GL deploys a tool called PHAST for plume models [58] and uses the output to inform
the risk analysis. This model directly impacts a FMEA, Bowtie, HAZID, or other hazard
analysis as required by UL 9540 or standards with the same intent. It is implied by ANSI
and IEEE 1547 updates that UL 9540 will be a requirement for energy storage projects,
which includes FMEA for the system and related ancillary equipment. [25] As shown in
Figure 25, it is common practice for a safety review to occur during permitting and prior to
installation. This review may include the FMEA as required by UL 9540, or it can be part of
an independent engineering review on behalf of the lender, project developer, or insurer.
9.3.2 Indoor Locations (Penthouse or Dedicated Room)

Emissions from batteries are simultaneously flammable and toxic during failure. The
emissions characteristics of a Li-ion battery are shown in Figure 20. In all of the tests
conducted in this program, this behavior was consistent among all Li-ion batteries. The
figure indicates that 40-90% of the time, a single battery cell emissions rate corresponds to
less than 10 ppm in a 0.44 m3 volume. The peak event can exceed 200 ppm in this volume
for a single cell, and it is short lived (2-3 minutes).
Similarly, it was found that vanadium oxide electrolytes emit HCl and HF, with HCl occurring
in greatest quantities (see Figure 6 on page 12). Lead acid battery electrolytes emit SO2 and
HCl when heated (see page 12, Figure 7, and Figure 8). The mass and volume equivalent
concentrations of emissions from all battery types are included in Figure 4 (peaks) and
Figure 5 (average ppm per kg per minute).
A common toxic emission from all battery types was HCl. This is also common with plastics
fires. Because the IDLH rating for HCl is low and the quantity of HCl emission is typically
largest among the four toxic constituents monitored, the ACH rating is therefore governed
by HCl. As shown in Figure 5 all battery types average lower than 2 ppm per kilogram per
minute in the categories of CO, HF, HCN, and HCl emissions.
IDLH and Emergency Response Planning Guidelines (ERPG) values for HCl, HF, HCN, and CO
are shown in Table 6. The term immediately dangerous to life or health (IDLH) is defined by
the US National Institute for Occupational Safety and Health (NIOSH) as exposure to
airborne contaminants that is "likely to cause death or immediate or delayed permanent
adverse health effects or prevent escape from such an environment."
 ERPG-1 is the maximum airborne concentration below which nearly all individuals
could be exposed for up to 1 hour without experiencing more than mild, transient
adverse health effects or without perceiving a clearly defined objectionable odor.

January 18th, 2017
Consolidated Edison
could be exposed for up to 1 hour without experiencing or developing irreversible or
other serious health effects or symptoms which could impair an individual's ability to
take protective action.
could be exposed for up to 1 hour without experiencing or developing life-threatening
health effects.
Table 6 Immediately Dangerous to Life and Health (IDLH) for the emphasized toxic
gases identified in the testing work.
IDLH (ppm) ERPG-1 ERPG-2 ERPG-3

HCl 50 3 20 150
HF 30 2 20 50
HCN 50 n/a 10 25
CO 1200 200 350 500
SO2 100 0.3 3 25
This dynamic and varying emissions rate was time-averaged and then charted as a function
of air change over rate (air changes per hour, or ACH), of the battery mass undergoing
failure, and the room volume. Because this time averaged calculation includes the nonlinear
effect of higher emissions during the peak, this ACH calculation is overly conservative for
40-90% of the duration of the battery failure event. As mentioned previously and as shown
below, HCl (Figure 16) governs the dominating air change over requirement because of the
low IDLH value. The chart in Table 15 on page 66 converts ACH to CFM based on room size
and approximate room footprint. In all cases the ACH rate is calculated to maintain gas
concentrations below IDLH.
An air change rate of 0.25 ACH is sufficient for limited cell failure scenarios to mitigate HCl
in the room sizes considered (see Figure 16). The peak emissions rate for up to 1.5 Li-ion
modules (typical masses assumed) would require up to 11.5 ACH. This is within normal
laboratory building ACH requirements, by comparison (Table 8 on page 51), and ASHRAE
notes that 1 – 4.4 ACH is common in residential and commercial environments. [26] This
clarifies DNV GL’s recommendation that ventilation requirements are within
established limits of the built environment as long as the system demonstrates it
can limit propagation of cell failures with cascading protections
 CO (Figure 18 and Table 13 on page 66) can be mitigated in all scenarios with only
0.25 ACH.
 HF (Figure 19 and Table 14 on page 66) can be mitigated with 0.25 ACH in the most
probable failure scenarios and may require up to 14.5 ACH in the smallest room
considered.
 HCN emissions rates can be mitigated for the most probable failure scenarios (a
single or multiple cells) with only a 0.25 ACH. In the worst case scenario of 1.5
failing modules, the ACH is 7.5.

January 18th, 2017
Consolidated Edison
Note that HCl and HF govern the ventilation requirements, which implies that the
ventilation requirement is determined by toxicity, not flammability. This is because
toxic gas IDLH limits are between 30-50 ppm, while flammability limits for many gases are
in the 1,000-10,000+ ppm range. The assumptions used in this calculation are shown
below. The emissions rates assumed for the ACH calculations are the average of the
emissions measured during cell testing. A 30-minute release rate is conservative, and
accounts for an average of emissions rate that is higher than the low level emissions leading
up to peak failure, and lower than the peak emissions.
9.3.3 GPM and CFM Requirement

It is shown in Figure 15 that the heat release rate has a weak positive correlation to mass
lost because the linear fit has a positive slope but the R2 is low due to scatter in the data. As
discussed on page 4, the scatter is due to the nonlinear behavior of battery fires. As shown
in Figure 20 the battery spends between 40-90% of the time in a smoldering state, meaning
that the exothermic contribution to the fire is low during this period and much of the battery
mass is lost during this time, which means there is less to contribute to the peak HRR
event. It was also shown in Figure 12 that it was possible to reduce the water requirement
as testing progressed on modules and systems. This data was directly measured from the
masses of the cells and modules and the water used.
The theoretical minimum water requirement for the battery mass (not the system mass) is
calculated in Table 7. It should be noted that the water calculation is determined in units of
GPM/kg; dividing this number by the energy density (commonly given in Wh/kg) will
convert the result to GPM/Wh, and multiplying by 1000 Wh/kWh will convert the result to
GPM/kWh. A cross check for these conversions will be needed as energy density of batteries
will inevitably increase over time.
For context and benchmarking, typical ventilation and water sprinkler requirements are
shown in Table 8 on page 51. The range of possible values for the GPM/kg of battery are
shown in Figure 31 on page 54. Table 15 on page 66 shows conversion factors between
ACH, CFM, and CFM/ft2.
The aggregate of such data is shown on page 67, which demonstrates the means to
estimate water flow and ventilation flow requirements based on system size. In some cases
it can be seen that the ventilation rates and GPM requirements are within the norm of
building codes. This is translated in Table 10 on page 53. However, the factors that affect
this most are the mass of batteries, their energy density, and the volume of the room where
they are installed. The air volume in larger rooms will dilute emissions, resulting in lower
requirements for air change.
DNV GL and Rescue Methods found that the water requirement per kg of battery material
decreased as the quantity of modules became larger (Figure 31). It is acknowledged that
initial testing began with an arbitrary water volume at the cell level, and it was found that
this quantity was more than sufficient—and is therefore excessive—for a practical
application. DNV GL recommends that further study be considered to find the
minimum water requirement for extinguishing and measure the physical
parameters impacting water contact efficiency

January 18th, 2017
Consolidated Edison
Because outdoor systems are likely containerized they are also likely to include on-board
gas-based fixed suppression systems. As recommended by DNV GL in the extinguishing
section (see page 45), a gas based suppression system may serve as a first line of
extinguishing. Adequate sizing of nearby fire hydrants should be considered in the context
of the maximum possible heat load during a system fire.
Table 7 Example calculation to determine the minimum water requirement per kg

of burning cell.
Theoretical Minimum Water Requirement to

Cool a Battery
Battery burn time (min) 42.25
water density (kg/gal) 3.7
m battery (kg) 2.87
c water (kJ/kgC) 4.1
c battery (kJ/kgC)6 1.4
ΔT battery (deg C) 525
ΔT water (deg C) 75
Q battery (kJ) 2,107.0
m water (kg) 6.9
vol water (gal) 1.9
GPM 0.044
Theoretical Minimum GPM/kg 0.015
Table 8 Benchmarks for airflow and water flow for typical structures.
Benchmarks CFM/ft2 GPM/ft2 Sources

Libraries 0.12 0.05-0.3 ASHRAE Addendum n to
ANSI/ASHRAE Standard
Warehouses 0.06 0.05-0.3 62-2001 "Ventilarion for
Pharmacy 0.18 0.05-0.3 Acceptable Indoor Air
Quality" and NFPA 13
Laboratories 0.18 0.05-0.3 Area Density Curves
Based on the known test data, DNV GL is able to recommend the following across the
aggregate of battery chemistries. The values in Table 9 are derived from Table 16. These
are converted to example CFM/ft2 and GPM/ft2 values in Table 10 on page 53.
6
Estimated by phenolic, given that the battery is a composite of multiple polymers, liquids, and some metals.

January 18th, 2017
Consolidated Edison
Table 9 Values derived from probabilistic analysis of water flow rates (GPM) and air
flow rates (CFM) per system energy (kWh) or mass (kg).
Scalable Metrics for Systems based on Electrochemical Battery Mass and Energy
Content
25th Percentile Mean 75th Percentile
Water Flow Rate GPM/kg 0.07 0.10 0.20

Water Flow Rate GPM/kWh 0.70 0.99 2.09
Air Flow Rate CFM/kg 0.01 0.02 0.03
Air Flow Rate CFM/kWh 0.11 0.18 0.31
A sensitivity analysis is demonstrated with calculated regression coefficients. The

uncertainty in the calculation is captured by triangular probability distributions created in
Table 16. In regression analysis, the coefficients calculated for each input variable measure
the sensitivity of the output to that particular input distribution. The sensitivity of the
calculation of the ventilation rate is shown in Figure 35.
The energy density, cell mass, and emissions rate from the cell are the greatest influencing
factors in 90% of the calculated outcomes. DNV GL recommends that when calculating
the air flow and water extinguishing rate, one must account for battery energy
density (only the battery cells, not the entire system) as well as the duration of
the event. The sensitivity of the calculation of the water flow rate is shown in Figure 36.
The two main factors influencing the calculation are the range of flow rates found during
testing and the range of possible energy densities of the battery system.
These uncertainties demonstrate the following:
 Energy density and the emissions duration should dictate the ventilation requirement
 Energy density and the duration of the event affect the extinguishing requirement
In Table 10 some example ventilation and water extinguisher ratings are calculated based
on hypothetical systems. The values in Table 10 are calculated from Table 7 and
demonstrate the mean of probability distributions generated from Table 16. The
distributions of the water requirement is skewed to the left, as shown in Figure 31. The
table demonstrates how these findings translate to codes development via examples. The
table demonstrates that the ventilation and water requirements are within the scope of
present requirements for the built environment when the system is placed within adequate
room volumes (compare with Table 8). Considerations of whether the spaces are occupied
or whether they are outdoors apply.
In practice, these results will depend on the actual system weight and room size on a per
project basis. When considering a containerized system, the following additional
considerations may occur:
 A containerized system may not be considered a livable or occupied space and
therefore may have different code considerations.
 The water requirement in Table 10 is equivalent to about 4-5 garden hoses and is
less than half the GPM rating of a typical 2.5” line (250 GPM). If the system is
already equipped with a gas-based fixed suppression system, a parallel water

January 18th, 2017
Consolidated Edison
connection on the exterior would accomplish the goal for first responders to create a
cost effective internal sprinkler system as a backup to the fixed suppression system.
The calculated airflow requirement can be oversized with a variable speed fan that meets
the minimum air change requirement and may peak upon detection of smoke or
particulates.
The leftward skewness of the distributions for both the GPM requirement and the ACH
requirement is demonstrated in Figure 31 and Figure 32.
Table 10 Example implications based on extrapolated findings from testing. It can

be seen that the calculated water requirement is within the bounds of what was
described for libraries, pharmacies, warehouses, and laboratories; similarly the air
flow requirements can be at or below unless the room volume is too small.
Example Code Requirements
System System Estimated Estimated Ventilation Theoretical Median GPM GPM CFM/ft2 Min Median
Size Chemistry Mass (kg) Room Size Requirement Minimum Requirement Requirement GPM/ft2 GPM/ft2
(kWh) (ft2) (CFM) GPM at 0.1
Requirement GPM/kg
20 Li-ion 133.3 100 2.3 2.0 2.2 13.3 0.02 0.02 0.02
100 Li-ion 666.7 146 11.7 9.8 11.2 66.7 0.08 0.07 0.08
100 Pb Acid 3,333.3 200 58.4 UPS 0.29
Requirement
1000 Li-ion 6,666.7 300 116.8 98.1 111.8 666.7 0.39 0.33 0.37
1000 Vanadium 20,000.0 1500 350.4 Standard 0.23
Redox Commercial
Note: Fire flows in excess of 3,000 GPM per buildings are considered impractical for many state fire codes. Consideration of battery
flammability, cascading protections, and building water supply should be considered. For containerized systems, a parallel system may be
fed externally by fire hose.

January 18th, 2017
Consolidated Edison
Figure 31 Distribution of gallons per minute for a 1 MWh battery, calculated from
cell testing and extrapolating with the latent heat value, which demonstrates that
the 0.1 GPM/kg estimation is highly conservative.
Figure 32 Example of the air flow requirement for a 1 MWh Li-ion system,
demonstrating that the distribution of values is strongly skewed leftward.

January 18th, 2017
Consolidated Edison
9.4 Inspection and Monitoring

DNV GL’s testing revealed that, besides lingering offgases such as CO, lingering (unseen)
internal temperatures and residual voltages on unburned cells presented a hazard to first
responders.
Similar to the code for uninterruptible power supplies, which recommends hydrogen
monitoring and a system health status display (see Table 5), DNV GL recommends at a
minimum that an error status panel, emergency response contact, or other form of
error notification be available to first responders, and that the energy storage
supply chain engage with first responders to propose a viable solution. Current
codes for UPSs include a display panel for inspection and error notification purposes.
If a system has been in a fire which has been contained by internal fire suppression, such a
display panel is enough to alert first responders that the system has sustained damage.
They may be able to call in a specialist to handle the hazard and relieve FDNY of their
responsibility for the site.
There are technical parameters that have direct impact on the volatility of the system,
though it is debatable whether they should be the responsibility of the first responder.
Recall that the ultimate objective of the first responder is to protect life, preserve property,
and ultimately secure the scene. The intent of system health notifications or an emergency
response network is to alleviate the concern of the first responder that he/she will somehow
be obligated to own an unknown hazard. The project development community would serve
its own interest to support first responders in creating a means to facilitate a hand-off from
the first responder to a project owner with good certainty that the hazard is under control.
9.5 Clearances
As referenced in Table 4 and Table 5, the majority of codes identify a 3-5ft clearance on
energy devices within enclosed spaces. [4]
In addition, from an economic and technical standpoint, limitation of footprint of energy

storage systems directly undermines one of the key value propositions of energy storage,
i.e., high density stored electrical energy in a small space where it is most needed.
Therefore there is a need to weigh overly prescriptive recommendations against the actual
hazard. Recall that in DNV GL’s model (also supported by testing) a system could withstand
60 minutes of direct fire at 1000oF with only a metal barrier and an air gap (see Figure 28
on page 42).
The current rule structures (NFPA 855, IBC, and IFC updates may allow for local AHJ
exceptions for the spacing and quantity of energy storage systems provided they pass a risk
analysis).
With these considerations, DNV GL recommends that all system installations

undergo a risk analysis, with particular attention paid to:
- Cascading protections between cells and modules
- Clearances to structures above the energy storage systems
- Fire rating of the enclosure
- Most probable expected failure mode

January 18th, 2017
Consolidated Edison
This recommendation is consistent with the IEEE 1547 and ANSI recognition of UL 9540
because of its FMEA process. Because cascading protections have been overlooked in safety
incidents (see Literature Review) it is highly important that this consideration be
emphasized in the up-front risk analysis. Clearances to nearby structures are presently
being recommended on a kWh basis, which may inadvertently limit the effectiveness of
energy storage by artificially increasing its footprint and therefore its effective functional
power and energy density. The risk analysis should provide a foundation for stakeholder
agreement on when the risks are deemed acceptable to exceed these requirements.
Similarly, the fire rating of the enclosure, if exceeding specification, may create
opportunities to reduce spacing or clearances. And the most probable failure mode is the
most important part of the risk analysis; it helps differentiate risks that seem significant but
are actually low probability, versus risks that are probable and measurable, and then design
with cost effectiveness and practicality.
9.6 Room Capacity Limitations

The holding capacity of an enclosed space is dependent on a number of factors:
- As shown Figure 20 the total quantity of emissions from burning batteries is
dependent on the mass available and the nonlinearity of its emissions rate.
- As shown in Figure 20, the battery fire is largely a smoldering event until a 2-3
minute peak.
- It was also mentioned in “Extinguishing” on page 45 that cascading protections
between cells have a direct impact on the propagation of the event to the entire
system.
- It was also found in this work that peak room temperatures in a fire are directly
correlated to the mass of the battery (see Figure 38 on page 76).
Present guidance is suggesting limitations on battery systems as a function of kWh capacity.

It should be noted that energy density (kWh/kg) in battery cells is continually increasing as
new generations are released. Prescribing a code based on mass (kg), would present the
challenge of increasingly higher amounts of energy being deployed under the same mass
constraint. The precedent is a limitation of 1000 lbs. (453 kg) of Li-ion batteries in a space
without suppression, which at today’s typical Li-ion energy density of ~150 Wh/kg,
corresponds to about 67.9 kWh. Proposed IFC language will reduce this to 20 kWh for single
units with a total limit of 600 kWh in an enclosed space.
As previously stated, DNV GL recommends that a risk analysis be performed on

any basis where battery systems larger than 20 kWh and assembled in aggregate
shall be installed in an enclosed space, with the intent of answering these
questions:
 Is the system functionally limited by the code rating?
 Does the system have design features that prevent cascading failure
between cells and modules? (See fire test, UL 1973 test, or IEC 62619 test
data.)
 Is the baseline and peak ventilation capacity adequate for the potential off
gas? (Example, Figure 16)
 Is the sprinkler system adequately designed for the potential heat load and
battery chemistry? (Example, Figure 29)

January 18th, 2017
Consolidated Edison
 Does the protective casing provide adequate insulation and fire blocking?
(Example, Figure 28)
The output of this analysis should determine if the rules are too prescriptive for the case
being considered, or alternatively, if the rules have not adequately captured a safety risk.
9.7 Project Development Considerations for Interaction with First

Responders and AHJs
DNV GL surveyed several handbooks for fire departments in large cities across the country
and found a universal theme in fire fighter training concerning extinguishing. Fire fighters
are trained to achieve the following objectives when arriving at the scene:
 Objective 1: Remove endangered person(s) and treat the injured.
 Objective 2: Stabilize the incident and provide for life safety.
 Objective 3: Provide for the safety, accountability, and welfare of personnel
(this priority is ongoing throughout the incident).
 Objective 4: Protect the environment.
 Objective 5: Property conservation.
Note that Objective 5 is often the primary concern of the property owner. It is on the
priority list of the first responder, but safety of life at the scene takes precedence. The
following recommendations for emergency response specific to batteries refer to these
objectives. These are based on the UPS battery system precedent that already exists in New
York City.
 Battery systems should be described in the Building Information Card (BIC) (see
example, Figure 33). This greatly aids in first responders meeting Objective 2.
 A building should have an assigned liaison who works with FDNY to update
emergency response plans. This liaison may be the same as the certificate of fitness
(COF) holder for the battery system, or may be a different individual. This Liaison
should be listed in the BIC. This aids first responders in meeting Objectives 2
and 3, and also protects the property owner’s interest relating to
Objective 5.
 Battery systems should have a COF similar to what is required for UPS systems.
Again, this aids in Objectives 2, 3 and 5.
 The recommendations for monitoring and system health display are consistent for
codes for uninterruptible power supplies. The method of system health display and
monitoring should be proposed by the system integrator or project owner.

January 18th, 2017
Consolidated Edison
Figure 33 The FDNY Building Information Card (actual example) contains

emergency contact information for fire safety and building engineers.
9.8 Considerations for Battery Chemistries that are not Li-ion

Much of the data in this report pertains to Li-ion because the majority of battery cells tested
are variants of that chemistry. However, the data contained in this report should concisely
demonstrate the following:
- Vanadium redox and Pb acid electrolytes are not flammable.
- Vanadium redox and Pb acid electrolytes do represent a toxicity hazard when heated.
- Polymer cases for any battery are flammable and will contribute to a fire as fuel and
a source of toxic emissions.
While not tested explicitly in this study, it is also worth mentioning that under rare
circumstances lead acid batteries are also capable of so-called thermal runaway, i.e., an
exothermic failure. Because the members of the battery industry have taken great care to
differentiate themselves in the area of safety, with nearly all chemistries that are not Li-ion
using marketing language such as “safe”, “nonflammable”, “thermally stable”,
“environmentally benign” or “incapable of thermal runaway”, there is a need to clarify a
universal finding in this program: in the case of external fire, all batteries emit toxic gases.
It should also be noted that the average emissions rates of equivalent masses of plastics
exceed those of batteries. Every battery tested either emitted a gas or left a residue that
has a varying degree of hazard (Table 3 on page 29); however, this can be expected from
most fires. The general findings of this work conclude that water and ventilation
requirements are within the technical limitations of legacy building codes, i.e., there are
precedents for managing these hazards.
All of the batteries tested carry with them a risk in their deployment; however, all of the
risks identified are manageable within the realm of today’s engineering controls for safety.
In addition, the toxicity and flammability risks identified are not insurmountable or highly
unique when compared to the challenges of burning hydrocarbons or plastics, and the
resulting requirements in codes, if implemented, are within the boundaries of the typical
built environment.
DNV GL’s recommendations are the following:

- If a battery is demonstrated to have a non-flammable electrolyte, there may
be considerations for a reduced water extinguisher requirement, or at a

January 18th, 2017
Consolidated Edison
minimum a water requirement equivalent to that required for the space

without battery systems installed.
- The ventilation requirements should be the same for all battery chemistries
tested in this program because they all have varying degrees of HCl or
similar toxic emission upon heating.
Lastly, the emissions rates of equivalent amounts of plastics during a fire, including
common every day materials that are found in office environments, commercial and
industrial settings, and even residential homes, can exceed the quantity of emissions from a
battery fire and will emit HCl as well.
10.0 SUMMARY OF MAIN FINDINGS AND RECOMMENDATIONS

The below directly summarizes key findings and recommendations from this study. These
are placed in list form up front in the document for access and readability. The reader is
strongly encouraged to use the cross references in the report to learn the reasoning behind
the recommendations, or read the FAQ section for clarifications.
 The toxic emissions from fires in this study are not necessarily excessive in content
or quantity, and can be managed by today’s engineering controls.
 The code requirements for battery systems have the potential to fall within the
boundaries of legacy codes, provided that technical and practical engineering
considerations are made concerning room volume and battery size.
 DNV GL recommends a minimum 1-hour fire enclosure rating with a 2-hour rating in
areas with critical population density.
 For the intent of delaying the escalation of the fire, all systems with individual cells
as part of their assembly should demonstrate that a single cell failure cannot
propagate to neighboring cells in a module design, which may exceed the acceptance
criteria for UL 1973 or the IEC 62619 internal propagation test.
 Because of the rapid evolution of a cell fire, DNV GL does not see an advantage to
using a Class D extinguisher on a single cell fire, given the difficulties of access and
timing. While technically appropriate, the deep seated nature and window for access
present technical challenges; the need for cooling should be prioritized.
 Fixed suppression gas agents may reduce or mitigate flammability in an environment
until ventilation and/or cooling strategies are implemented, though their actual
cooling capability should be scrutinized in comparison to water.
 DNV GL recommends the following for extinguishing:
o Stage 1: If a system can limit cell cascading, a gas based suppression system
may be considered for the first stage of fire fighting to extinguish a single cell
fire and prevent flashover in a contained environment.
o Stage 2: If temperatures continue to rise or if an increasing level of smoke
and gas is detected, water extinguishing accompanied by forced ventilation
should be considered to cool the system and prevent further propagation of
fire.
 For nearly all chemistries, the ventilation of HCl, CO, and HF govern the ACH
requirements, i.e., toxicity considerations dominate the ventilation need rather than
flammability.

January 18th, 2017
Consolidated Edison
 The gases emitted are also found in plastics fires in greater time-averaged
quantities. This should be considered in the context of prescriptive codes because
these hazards are likely to already exist in the built environment.
 DNV GL recommends that the lowest level ventilation rate - if prescribed - be
continuous under normal systems operation. The study concludes this may be as low
as 0.25 ACH, which is lower than what is required for most occupied spaces.
 Regardless of chemistry, DNV GL recommends sizing for ventilation and extinguisher
systems as the following (these may be translated to GPM/ft 2 and CFM/ft2 or ACH
starting on page 50).
 DNV GL recommends that minimizing the water requirement be an area of further
study as it has likely been overstated in these recommendations for
conservativeness.
 DNV GL recommends at a minimum that an error status from an operating energy
storage system be readily apparent to first responders for the following parameters,
and recommends that a dialog be opened with system integrators to
determine the most effective and economic way to address this need:
o Internal atmosphere (normal or gas detected)
o Temperature (above normal or normal)
o Current (normal or threshold exceeded)
o Voltage (normal or threshold exceeded)
 During and after fire extinguishing, it is recommended that if first responders choose
to use water submersion to cool and isolate battery modules, that preparation to
deal with alkaline or acidic water be considered.
 After extinguishing, continued ventilation and monitoring of the area is highly
recommended to protect first responders from continued toxic and flammable gas
emissions. The first responder team can monitor the area with handheld sensors to
determine the appropriate time to stop ventilation.
 It is highly recommended that an emergency contact list and/or subject matter
expert be available for all battery systems installed in buildings in order to introduce
the opportunity for first responders to relinquish control of the scene to the system
developer or a designate after the site has been secured and extinguishing has been
completed. This is likely to require involvement from the project development and
systems integration community.
 If a battery is demonstrated to have a non-flammable electrolyte, there may be
considerations for a reduced water extinguisher requirement, or at a minimum, a
water requirement equivalent to that required for the space without battery systems
installed.
 The ventilation requirements—if prescribed—should be the same for all battery
chemistries tested in this program because they all emit similar HCl levels.
 In order to meet or exceed UL 9540 requirements, DNV GL recommends that a risk
analysis be performed on any basis where a battery system or portfolio of systems
shall be installed in an enclosed space near occupants. The analysis should look at
the general safety picture of the project(s) —in aggregate if possible—with a focus
on these risks:

January 18th, 2017
Consolidated Edison
o Does the system have design features that prevent cascading failure between
cells and modules? (See fire test, UL 1973 test, or IEC 62619 test data.)
o Are ventilation systems at the intended site(s) adequately rated to handle the
most probable failure mode? (Example: Table 7)
o Are sprinkler systems at the intended site(s) adequately designed for the
potential heat load and battery chemistry? (Example: Figure 29 and Table 7)
o Does the protective casing provide adequate insulation and fire blocking?
(Example: Figure 28)
10.1 Conclusions
 Many historic battery incidents are due to external damage factors which have
created confusion and overreaction to the topic of battery safety.
 Existing building codes and engineering controls can be adequate in many cases to
handle battery safety issues.
 The toxic emissions from fires in this study can be managed by today’s engineering
controls and are not anomalous or excessive when compared to a plastics fire.
Plastics fires can generate similar gases in larger quantities over the average
emissions duration on an equivalent mass basis.
 The water requirements from this study can be lessened for building fire extinguisher
systems when combined system-level safety approaches are implemented.
 Legacy codes could provide insightful interim requirements for battery systems used
in energy management, provided that technical and practical engineering
considerations are made.
 Gas-based agents that can reduce flammability in an enclosed environment can put
out single battery fires, but should not be considered an adequate cooling measure.
 Water demonstrated the highest cooling efficacy of all extinguishing agents tested.
The use of water should only be considered if there is an acceptable risk of shorting
additional cells or collateral damage to the remainder of the system.
 Water volumes for cooling can be minimized based on the expected duration of a
failure event. Systems with adequate internal cascading protections will minimize the
water volumes required for extinguishing.
 Staged extinguishing with fixed aerosol or gas suppression agents first, followed by
water in the event of a cooling need, is recommended. It may be possible to use
parallel water inputs on fixed suppression systems for containerized battery systems.
 Forced ventilation is recommended for first responders, even after the fire has been
extinguished.
 The historical legacy of safety concerns has validity, though understanding of the
root causes and failure modes is necessary in order to understand the true threats
and failure modes.
Appropriate mitigation of risk shall include a pre-commissioning design review per accepted
industry practices that are presently being used in California and other states. Overall
DNV GL’s findings are that these hazards are manageable for building code officials and first

January 18th, 2017
Consolidated Edison
responders. No significant technology barrier exists that prevents code officials or first
responders from doing their duty when encountering battery energy storage systems.
11.0 REFERENCES
1. US Department of Labor, “Confined Spaces”. Occupational Safety and Health Administration
(OSHA) https://www.osha.gov/SLTC/confinedspaces/
2. 1 RCNY §50-01 Chapter 50 Distributed Energy Resource Standards.
3. “Home Fires”, Federal Emergency Management Agency (FEMA). http://www.ready.gov/home-
fires. Accessed May 10, 2015.
4. Hopper, Howard. “Energy Storage Systems Fire Safety Considerations”. UL Codes and
Advisory Services.
5. United States Department of Energy , DOE Global Energy Storage Database,
www.energystorageexchange.org. Accessed 11/11/16
6. Milke, Kodur, Marrion. “Overview of Fire Protection in Buildings”. Federal Emergency
Management Agency (FEMA). Appendix section from “World Trade Center Building
Performance Study: Data Collection, Preliminary Observations, and Recommendations”. 2002
7. Tyler, David. “Battery Related Fire Damages Famed Hybrid Tug, Puts it Out of Service.”
Professional Mariner, Nov 8, 2012.
8. Hayman, Susan. “Final Report on Battery Re-Installation; Campbell Foss and Carolyn
Dorothy”. Foss, 2015.14
9. Junod, Thomas. “Gaseous Emissions and Toxic Hazards Associated with Plastics in Fire
Situations – A Literature Review.” NASA Technical Note, NASA TN D-8338. Lewis Research
Center, Cleveland, OH. Oct 1976.
10. Hoff, Steeves. “New Insights into Thermal Runaway of Valve Regulated Lead-Acid Batteries”.
Battcon 2005.
11. Hill, et al. “Sensor Enhanced and Model Validated Life Extension of Li-Ion Batteries for Energy
Storage” APRA-E Advanced Management & Protection of Energy Storage Devices (AMPED)
Annual Program Review. January 8-10, 2014. Denver, CO
12. Tran, Lyda. “Statement of National Highway Traffic Safety Administration On Possible Fires in
Lithium-Ion Vehicles Involved in a Crash.” November 11, 2011.
13. Andersson, et al. “Investigation of Fire Emissions from Li-ion Batteries” Fire Technology, SP
Report 2013: 15.
14. NTSB Case Number: DCA13IA037, Interim Factual Report. Boeing 787-8, JA829J, Japan
Airlines. National Transportation Safety Board, Office of Aviation Safety. March 7, 2013
15. “Chevrolet Volt Battery Incident Overview Report” DOT HS 811 573. US Department of
Transportation, National Highway Traffic Safety Administration (NHTSA). January, 2012.
16. Shepardson, David. “Tesla Recalls 29k Vehicle Chargers That Could Spark Fire” The Detroit
News, January 14, 2014.
17. Chernova, Yulia. “Electric Car Maker Fisker Lost More Than 300 Cars in Hurricane Sandy
Flooding” The Wall Street Journal, Nov 6 2012.
18. Blanco, Sebastian. “GM Battery Lab Explosion Cost could Reach $5M” April 18th, 2012.
Autoblog.
19. Webster, Harry. “Lithium Battery Fire Tests” FAA William J Hughes Technical Center
20. Webster, Harry. “Flammability Assessment of Bulk-Packed, Nonrechargeable Lithium Primary
Batteries in Transport Category Aircraft. June 2004
21. DOT/FAA/AR-04/26
22. Lithium-Ion Batteries Hazard and Use Assessment. Fire Protection Research Foundation
Exponent. July 2011

January 18th, 2017
Consolidated Edison
23. Recommendations on the Transport of Dangerous Goods. Manual of Tests and Criteria, 5 th Ed.
Amendment 1, United Nations, Section 38. 38.3: “Lithium metal and Lithium Ion Batteries.”
24. Navy Sea Systems Command Advanced Change Notice for Lithium Battery Firefighting
Procedures.
25. Florence, Laurie. “Establishing Safety of Energy Storage – An Overview of UL Safety
Standards.” CPUC Energy Storage Workshop, August 19, 2015.
26. Fennel, Haehnel. “Setting Airtightedness Standards” American Society of Heating,
Refrigeration, and Air Conditioning Engineers (ASHRAE), 2005.
27. "ANSI/ASHRAE Standard 62.2-2013: Ventilation and Acceptable Indoor Air Quality in Low-Rise
Residential Buildings". Atlanta, GA: American Society of Heating, Refrigerating and Air-Conditioning
Engineers. 2013
28. Stout, Roger. Beyond the Data Sheet: Demystifying Thermal Runaway. ON Semiconductor,
Nov 1 2007. Power Electronics.
29. Stout, Roger. Power Electronics System Thermal Design: Thermal Runaway. IEEE Power
Energy and Industry Application Courses, Feb 2007.
30. Ultracapcitor Safety Data Sheet, Maxwell Technologies.
31. Allcell Technologies, LLC. 2014 “Prevent Thermal Runaway Propagation”.
32. Winsun Energy Solutions, LiFePO4 specifications
33. “PBES announces world’s first commercialized thermal runaway suppression system”. Plan B
Energy Solutions (PBES).
34. SurePower, ”What is Thermal Runaway?” 2015 http://www.sure-power.com/2013/07/what-is-
thermal-runaway/
35. CD Technologies Technical Bulletin 41-7944: “Thermal Runaway in VRLA Batteries – Its Cause
and Prevention.” 2012
36. Rydh, Car Johan. “Environmental Assessment of Vanadium Redox and Lead Acid Batteries for
Stationary Energy Storage.” Journal of Power Sources 80 (1999) 21-29
37. Redox “Safety Data Sheet for Vanadium Pentoxide” Rev. 2 Jan 15 2014.
38. (cited with permission) DNV GL report, on behalf of ICL Industrial Products. “Safety Review of
Bromine-Based Electrolytes for Energy Storage Applications”, ICL Industrial Products, Report
No 1 Rev 1 Feb 26, 2016
39. “Guidelines and Considerations for Safer Deployment of ZnBr Based Battery Systems”. DNV GL
Feb 2016
40. Roman, Jesse. “The Li-ion Conundrum: Ignition and Reignition Issues with Batteries in Electric
Vehicles are Also Concerns for Energy Storage Systems.” Accessed March 4, 2016
41. DNV GL Guideline: Cell Level, Risk Based Testing and Modeling of Li-Ion Batteries for
Maritime, Energy Storage, and Other Applications Rev. 5.1, JANUARY 2016
42. Neubauer, et al. “The Effect of Variable End of Charge Battery Management on Small-Cell
Batteries.” NASA/TM-2007-215044, AIAA-2007-4789. December 2007
43. Dynamic Stress Test Power Profile. USABC Electric Vehicle Battery Test Manual Rev. 2 1996
44. Pinson, Bazant. “Theory of SEI Formation in Rechargeable Batteries: Capacity Fade,
Accelerated Aging, and Lifetime Prediction.” MIT, Department of Physics. 2012
45. Kuami, et al. “Gas Generation Mechanism due to Electrolyte Decomposition in Commercial Li-
Ion Cell.” Journal of Power Sources 81-82 (1999) 715-719
46. Hill, et all. “Sensor Enhanced and Model Validated Life Extension of Li-Ion Batteries for Energy
Storage”. Advanced Research Projects Agency – Energy (ARPA-e). DNV GL Final Report. July
1, 2015.
47. DNV GL Recommended Practice (GRIDSTOR). DNV GL RP 0043. “Safety, Operation, and
Performance of Grid Connected Energy Storage Systems.”
48. “Qualification of New Technologies” DNV Recommended Practice. DNV RP A203.

January 18th, 2017
Consolidated Edison
49. International Standard IEC 60812, “Analysis Techniques for System Reliability – Procedure for
Failure Mode and Effects Analysis.” 2nd Edition 2006, International Electrotechnical Commission
50. Huggett, Levin. “Toxicity of the Pyrolysis and Combustion Products of Poly(Vinyl Chlorides): A
Literature Assessment. Fire and Materials, Vol 11 131-142 (1987).
51. Ames, et al. Chapter 15: “Upholstered Furniture Predictions by Correlations”. Excerpt accessed
at the National Institute of Standards and Technology (NIST),
http://fire.nist.gov/bfrlpubs/fire92/PDF/f92017.pdf
52. Carbon-Monoxide – Hydrogen Combustion Characteristics in Severe Accident Containment
Conditions. NEA/CSNI/R(2000)10. Nuclear Energy Agency, Committee on the Safety of
Nuclear Installations. March, 2000.
53. Canadian Center for Occupational Health and Safety (CCOHS). Cheminfo, MSDS, Hydrogen
Chloride. Accessed 7-12-16.
http://www.chem.utoronto.ca/~pmeindl/labs/msds%20files/hydrogen%20chloride.htm
54. Dupont Teflon AF. Amorphous Fluoroplastics, Safety in Handling and Use. 2016.
55. Liyu Li, et al. “Vanadium Redox Flow Battery Using Mixed Acid Electrolytes”. US DOE Energy
Storage Systems (ESS) Program Review, Washington DC Nov 2 2010
56. Vanadium Redox Flow Batteries (VRB). Electricity Storage Association (ESA). Accessed 3-9-16.
57. Wang, Wei. Vanadium Redox Flow Batteries. US Department of Energy, Electricity Delivery
and Energy Reliability, Energy Storage Program. Pacific Northwest National Laboratory (PNNL).
58. Witlox, et al. “Validation of PHAST Dispersion Model as Required for USA LNG Siting
Applications.” Chemical Engineering Transactions, Vol 31, 2013. ISBN 978-88-95608-22-8
59. Dupont, Gauthier. “NGK’s NAS Batteries Long Recovery” August 23, 2012.
60. Fayet, et al. “Safety Issues Related to Stationary Electrochemical Energy Storage on Industrial
Sites.” CEt Chemical Engineering Transactions, Vol 48, 2016.
61. NAATBatt International, UET. March, 2015. NAATBatt Annual Meeting, Litchfield Park, AZ.
62. Trickett, David. NREL/TP-460-25553. “Current Status of Health and Safety Issues of
Sodium/Metal Chloride (ZEBRA) Batteries.” 1998
63. Willet, Ken. “Safety Validation in Grid Energy Storage.” Energy Storage Association, May 2014.
Washington DC.
12.0 APPENDIX 1: REFERENCED DATA
12.1 Assumptions for Air Changes per Hour (ACH) Calculations

Statistics for mass loss, duration of failure, and HCl, HCN, HF, and CO emissions are shown
in Table 1. This data is taken directly from all of the cell tests. It can be seen from the data

January 18th, 2017
Consolidated Edison
that the mass loss ranges from 0-57%, the duration of the event lasts from 13-83 minutes,
and the emissions rate (in ppm per kg per min) in the 0.44 m3 chamber ranges from zero to
0.719 for HCl, 0.032 for HF, 0.027 for HCN, and 2.341 for CO. This data demonstrates that
CO is emitted in greatest quantity and HCl is emitted in the second greatest quantity, but
because HCl has a lower IDLH this threshold is met first in most scenarios.
The following tables demonstrate the calculated ACH as a function of burning battery mass
and room size. The tables below are the same data that is visually presented in Figure 16
and related figures. It is clear from the visual representation of the data that these
relationships are nonlinear. The estimations limit the failure to 1.5 modules, with the
presumption that the system should demonstrate adequate separations, cascading
protections, and suppression systems to limit failure to a single cell or at least a single
module. The probability of failure for multiple modules should be very low for systems with
these active and passive barriers to catastrophic failure. Catastrophic failure scenarios can
be examined by risk analysis to determine which barriers are in place to prevent it and the
relative strength of those barriers. The risk analysis places practical boundaries on the
probability of high consequence events, and should either 1) tame the deployment of
extreme safety measures with a low probability of utilization or 2) identify likely failure
scenarios that have been overlooked in the context of the site and system.
Table 11 Air change rates based on HCl emissions as a function of room size and
quantity of failing cells.
80 ft X
20 ft 40 ft 80 ft
container container room
HCl 33.1 67.6 3624
1 cell 1.54 0.25 0.25 0.25
5 cells 7.70 0.25 0.25 0.25
10 cells 15.40 0.25 0.25 0.25
15 cells 23.10 1.25 0.25 0.25
20 cells 30.80 3.25 0.25 0.25
24 cells 36.96 4.75 0.25 0.25
1 module 44 6.5 0.6 0.25
30 cells 46.20 7 1 0.25
35 cells 53.90 8.5 2 0.25
1.5 modules 66 11.5 3.5 0.25
Table 12 Air change rates based on HCN emissions as a function of room size and
80 ft X
20 ft 40 ft 80 ft
HCN 33.1 67.6 3624
1 cell 1.54 0.25 0.25 0.25

January 18th, 2017
Consolidated Edison
5 cells 7.70 0.25 0.25 0.25

10 cells 15.40 0.25 0.25 0.25
15 cells 23.10 0.25 0.25 0.25
20 cells 30.80 1 0.25 0.25
24 cells 36.96 2.25 0.25 0.25
1 module 44 4 0.25 0.25
30 cells 46.20 4 0.25 0.25
35 cells 53.90 5.5 0.25 0.25
1.5 modules 66 7.5 1.5 0.25
Table 13 Air change rates based on CO emissions as a function of room size and
80 ft X
20 ft 40 ft 80 ft
CO 33.1 67.6 3624
1 cell 1.54 0.25 0.25 0.25
5 cells 7.70 0.25 0.25 0.25
10 cells 15.40 0.25 0.25 0.25
15 cells 23.10 0.25 0.25 0.25
20 cells 30.80 0.25 0.25 0.25
24 cells 36.96 0.25 0.25 0.25
1 module 46.20 0.25 0.25 0.25
30 cells 53.90 0.25 0.25 0.25
35 cells 44 0.25 0.25 0.25
1.5 modules 66 0.25 0.25 0.25
Table 14 Air change rates based on HF emissions as a function of room size and
80 ft X
20 ft 40 ft 80 ft
HF 33.1 67.6 3624
1 cell 1.54 0.25 0.25 0.25
5 cells 7.70 0.25 0.25 0.25
10 cells 15.40 0.25 0.25 0.25
15 cells 23.10 2.5 0.25 0.25
20 cells 30.80 5 0.25 0.25
24 cells 36.96 7 1 0.25
1 module 44 9 2 0.25
30 cells 46.20 9.5 2.5 0.25
35 cells 53.90 11.5 4 0.25
1.5 modules 66 14.5 5.5 0.25
Table 15 shows conversion factors from air changes per hour to CFM and CFM/ft 2 for the
modeled energy storage rooms and enclosures.
Table 15 Conversions from ACH to CFM based on the example room volumes for
energy storage systems.

January 18th, 2017
Consolidated Edison
ACH (row) to CFM (column) CFM/ft2 @ ACH

m3 ft3 ft2 0.25 1 5 10 30 0.25 1 5 10 30
Shipping Container, 20 ft 33 1,168 146 5 19 97 195 584 0.03 0.13 0.67 1.33 4.00
Shipping Container, 40 ft 68 2,407 301 195 781 3,906 7,811 23,434 0.65 2.60 12.98 25.96 77.88
Room, 80x80 ft 3,624 128,290 16,036 417,549 1,670,195 8,350,973 16,701,946 50,105,838 26.04 104.15 520.76 1041.52 3124.55
Notes: Occupied laboratories = 4-12 ACH, emergency ventilation ~ 30 ACH.
In Table 16, useful metrics derived directly from testing are provided. As mentioned
previously these values are input into a probabilistic model 7 to generate the sensitivity
analysis demonstrated in Figure 35 and related figures. The min, average, and max values
are used to generate triangular probability distributions. The GPM/kg measurement is a
direct measure of the water used to extinguish fires across the entire spectrum of cell to
module testing. The cell masses, mass loss, emissions range, HRR, and duration are the
ranges of values observed from cell testing. The energy density is calculated directly from
the cells. The estimated peak cell temperature is directly sourced from the cell data. The
fraction of cells simultaneously burning is a factor used to estimate the impact of total
emissions rate and account for the observed fact during module testing that cell failures
were rarely simultaneous and occurred as discrete events. It should be noted in the table
that the water contact efficiency averages 1-2%. This highly conservative number greatly
drives the water requirement estimation. Any method by which a battery manufacturer or
system integrator can demonstrate that the water contact efficiency is higher will reduce the
water requirement overall.
7
Palisade @Risk

January 18th, 2017
Consolidated Edison
Table 16 Aggregation of data regarding battery fires, extinguishing, emissions

rates, and extinguishing. Distributions in the column labeled “Dist” are triangular;
the mean is shown.
Probabilistic Inputs
Parameter min avg max Dist Notes
Cell Mass kg 0.5 1.6 6.5 2.867 From cell test data
Peak Cell Temperature (°C) 350 525 700 525 From cell test data
Duration (min) 2 47 83 43.882 From cell test data
Water Contact Efficiency 0.001 0.01 0.04 0.017 Estimated
Energy Density (Wh/kg) 30 120 150 100 Pb Acid to Li-ion
Probabilistic Outputs
Delta T to Cool Battery to 25 C 325 500 675 500 Calculated from Above
Energy to cool battery (kJ) 227.50 1120.00 6142.50 2006.67 mcdT
Required Water Mass including 0.09 0.44 2.39 0.78 Q battery = Q water, m_water = Q
battery / (energy to heat water to
heat of vaporization (kg) 100 C + dHv)
Required Water Volume (gal) 0.02 0.12 0.65 0.21 divide by 3.7 kg/gal
GPM 0.012 0.003 0.008 0.005 divide gal by duration
GPM/kg 0.024 0.002 0.001 0.002 divide GPM by battery mass
GPM/kg with water contact 23.916 0.158 0.030 0.099 Divide by water contact efficiency
efficiency
Additional Probabilistic Parameters

Testing GPM/kg 0.105 0.881447 1.65789474 0.881 From cell, module, and system test
data
Emissions range kg/min per cell 0.0002 0.0077 0.0152 0.0077 From cell test data
HRR kW/kg (of mass lost) 17 31 45 31 From cell test data
Fraction of cells simultaneously 0.12 0.16 0.2 0.16 From module testing, 1-3 out of 8-
15
burning
12.2 Water Mass Requirement Calculation

There are two ways to calculate the water mass requirement. One method is to calculate
the rate of energy released, which assumes that the extinguishing event is perfectly timed
with the peak energy release of the cell. The other method is to size the water requirement
to the battery mass, acknowledging that the cell failure event is not a single peak event, but
is instead characterized by long periods of smoldering (40-90% of the total event duration)
and a 2-3 minute peak event (accounting for 1-15% of the event duration).
The latter method was observed to be effective during testing as the water use in DNV GL
and Rescue Methods’ testing became progressively smaller (on a GPM/kg basis) as the
timing of the extinguishing event became decoupled with the peak HRR. In other words,

January 18th, 2017
Consolidated Edison
extinguishing the module or large pack was an exercise of removing distributed heat and
preventing perpetuating failure modes.
12.2.1 Sizing the Water Requirement to Peak HRR

Sizing the water requirement to the peak HRR involves the calculation of the amount of
energy required to raise the temperature of 1 kg of battery8 by 500 degrees Celsius. Using
the average HRR in Table 16, 1 kg of battery emits 31 KJ/second. This would be the heat
release rate �̇ and the formula used to calculate the mass flow rate of water is �̇ = �̇�� ∆�.
This results in a flow rate of 1.7 GPM per kg of battery. This is an oversized water
requirement, as shown below.
12.2.2 Alternative Strategy for Sizing the Water Requirement: Preventing

the Peak HRR Altogether
There are minutes of opportunity to simply cool the cell and avoid the peak event
altogether. Recall from Figure 20 that a Li-ion battery smolders for minutes before
eventually failing. Also recall from Figure 28 that the metal enclosure around a battery
system can provide a window as long as 60 minutes to respond to a fire. Thermal runaway
risk builds, but can be arrested by cooling and preventing the battery from reaching
temperatures near 120oC. A strategy in the marine sector is exactly this: cool the battery
and prevent thermal runaway temperatures from ever being reached, resulting in very
benign cell failure even during aggressive overcharge. [33]
Therefore the extinguishing strategy should be arrest the climbing temperatures before they
reach the transition temperature at 120oC. This more practical approach takes into account
that automatic fixed suppression systems typically lack the intelligence to sense and trigger
according to specific gas species or gas emission rates; i.e., they are discharged upon
detection of smoke via a sensor that is generally sensitive to multiple particulate and
hydrocarbon species. As a result, fixed suppression will trigger very early in the cell failure
process. This would be the case for all battery types tested, as smoldering and gaseous
emissions from the plastics used for containment began as early as 60 oC. Just the fumes
from the plastics may be enough to trigger a smoke alarm.
If the module has adequate cascading protections and a 1-hour fire rating, there is
an opportunity to contain the cell failure and avoid the issue of oversizing the
water requirement to the peak and instead size the water requirement to the
battery mass.
As a result the water calculation is simplified by sizing the water flow to the battery mass
rather than the HRR at thermal runaway. This strategy is only valid if the cascading
protections are demonstrated to contain single cell failures and prevent cascading from cell
to cell and module to module, and the fire rating of the system provides adequate time to
address an external fire.
Following this method, the energy to be removed from the system is:
Q = mcΔT
8
Simulated as phenolic due to its specific heat which is near the average of the battery composition by material

January 18th, 2017
Consolidated Edison
And thus the thermal equilibrium requirement is:

Qwater = Qcell
For a 1 kg battery cell with an estimated composite specific heat similar to phenolic
(1.4 kJ/kg°C), and a temperature change of 525oC – 25oC = 500oC, the energy of heat
transferred is 700 kJ. This calculation neglects the additional removal of heat by water from
the heat of vaporization, which is addressed below.
The specific heat of water is 4.1 kJ/kg°C. The objective is to use the minimum amount of
water before water flashes into steam. If we target a volume of room temperature water
necessary to prevent the water from flashing off into steam, we assume ΔT = 70 oC (70+25
= 95oC, or just under the boiling point). This translates to
mwater = 700 kJ / (4.1 kJ/kgoC * 70oC) = 2.43 kg
This states that 2.43 kg of water is required to cool a 1 kg battery from 500 oC to 25oC, and
the water will have risen in temperature to 95oC. This calculation should be very
conservative, as it neglects the vaporization of water into steam and assumes the entire
mass of the battery is contributing to the heat.
The density of water is 3.7 kg/gal, and therefore the theoretical conservative minimum
volume of water required is 0.65 gal. However recall that this reaction occurs over 1-3
minutes during the peak, and up to 40 minutes over a slow duration, and therefore the
gallons per minute required is 0.02-0.6 GPM/kg with the latter being
conservatively sized to still address the peak. The major factors driving the GPM/kg
requirement are the battery mass and the duration of the event.
The water requirements need not be excessive if the battery system employs
simple, industry proven safety measures such as an external fire rating and
cascading protections between cells and modules. Most of the batteries tested had
masses from 0.5-1.5 kg, with one battery being particularly large at 6 kg, which skews the
average to 2.8 kg and therefore makes this calculation more conservative. The values in the
table are probabilistic and the resulting distribution of water flows is shown in Figure 31.
The skewness of the distribution demonstrates that the theoretical minimum water
requirement mean is actually 0.019 GPM/kg, or very near the minimum.
12.2.3 Additional Consideration: Heat of Vaporization

When the heat of vaporization of water is included, the water volume requirement is further
reduced. The latent heat of vaporization is the energy required to accomplish the phase
change from liquid to gas. This property is given in kJ/kg and there is no change of
temperature to make the transition at 100o C at atmospheric pressure. This factor is
important is because the latent heat of vaporization is larger than the energy required to
heat water from 25 to 100o C.
The required energy to heat water from 25o C and then vaporize to steam at 100oC is:
� = ��∆� + �∆��

January 18th, 2017
Consolidated Edison
The specific heat of water C is 4.187 kJ/kgC and the latent heat of vaporization ΔHv is
2257kJ/kg. Using these numbers, the energy required to heat and boil one kilogram of
water from 25o C is:
� = �� ∗ . ∗ − + ∗
�� + , �� = , ��
It can be seen from the calculation that the latent heat of vaporization is 7x greater than
the energy required to heat from 25-100o C. This is important for cooling considerations
because the heat energy of the fire is transferred from the fire to the heating and boiling of
water; water withdraws energy from the fire, reducing its destructive power and energy.
Every kilogram of room temperature water that that is heated and flashed into steam draws
2,571 kJ from the fire.
Energy is most efficiently drawn from the fire when water contact is as complete as
possible. The method of delivery for the water will affect this contact efficiency such as mist,
spray, and jet. Access to the deepest seated batteries will govern the water contact
efficiency as well. When more water is in contact with the hot surfaces of the battery, the
rate of the water-to-steam conversion process increases, which saps energy from the fire
and reduces overall temperature as a result.
Expanding on the prior section, if the following assumptions are reconsidered with the
inclusion of latent heat of vaporization, the calculation follows:
mcΔT + mΔHv = Qcell
Where Qcell = 700 kJ. Therefore for 1 kg of battery cell:
mH2O = Qcell / (cΔT + ΔHv) = 700 kJ / (4.1 kJ/kg oC * 75 oC + 2257 kJ/kg)
= 700 kJ / (307.5 kJ/kg + 2257 kJ/kg) = 0.27 kg
Using the conversion factor 3.7 kg/gal, the resulting water volume is 0.07 gal. Again
assuming 1-3 minutes of battery burn duration, and up to 40 minutes for a slow duration
failure, the water requirement is 0.07 gal over 1-40 minutes or 0.001-0.07 GPM per
kilogram of battery. Note that this requirement is nearly 10x less than the thermal
mass balance calculation in the previous section. The latent heat of vaporization is
therefore a significant contributor to the cooling of the battery fire.

January 18th, 2017
Consolidated Edison
Figure 34 A reactive cooling approach requires an oversized water flow

requirement; whereas a system-level proactive approach enables a reduced water
requirement.
Figure 35 Regression coefficients of the ventilation requirement.

January 18th, 2017
Consolidated Edison
Figure 36 Regression coefficients for the water flow rate in GPM/kg demonstrate
that the duration of the event and the water contact efficiency are the strongest
drivers in reducing the water requirement.
12.2.4 Summary of Water Extinguishing Calculations

It can be seen in Table 17 that sizing the water requirement to the peak HRR leads to a 4-
170x oversizing of the water extinguishing system, when proactive and integrated safety
approaches are more efficient and reduce the water requirement.
The water contact efficiency of the extinguishing method is highly relevant to the overall
cooling effectiveness. The calculations demonstrate physically possible water flow rates,
however the testing is the most telling. As testing progressed, DNV GL was able to reduce
the water requirement from 1.7 GPM/kg at the module level to 0.1 GPM/kg. Conservative
factors accounting for water contact efficiency have resulted in DNV GL’s recommendations
in Table 9.

January 18th, 2017
Consolidated Edison
Table 17 Summary of methodologies and results of the water requirement

calculation.
Method Water Requirement Cross Reference

(GPM/kg)
Module Testing 0.1 – 1.7 Figure 12 on page
17
Calculated by Peak HRR 1.7 Page 69
Calculated Minimum Static 0.015 Page 51, Table 7
Thermal Mass Balance
Calculated Time-Dependent 0.02 - 0.6 Page 69
Thermal Mass Balance
Calculated by Thermal 0.001 – 0.07 Page 70
Mass Balance and the
Latent Heat of Vaporization
of Water
Based on the testing results and the calculations, 0.07 GPM/kg (including latent heat of
vaporization) and 0.1 GPM/kg (observed in testing a multi-module configuration) brackets a
significant range in heating and cell failure rate scenarios. A value of 0.1 GPM/kg appears to
be a highly conservative extinguishing rate as it does not account for the added benefit of
latent heat of vaporization and it provides a substantial compensation for water contact
efficiency.
12.3 Why Bowtie Models?

Cell level safety and system level safety are two different things. Assessing the risk of
external abuse factors can be accomplished with a risk analysis at the site. This technique
permits the visualization of all possible threats that may cause a top event, such as battery
failure, to occur. Putting barriers in place to prevent such events may increase safety of the
system overall. The diagram illustrates a generic battery failure model, illustrating that a
number of threats (left side of the diagram) can be prevented from leading to the top event
– which is loss of battery control – with barriers in place such as active monitoring and
proactive controls.
An example shown is mechanical damage by the red arrows progressing from the left of the
diagram to the right. In this example, there may be monitoring methods in place that did
not react quickly enough to identify and prevent consequences of mechanical damage, and
other barriers (such as physical barriers) may have failed. If these barriers are breached
and the top event occurs, then a possible consequence is thermal runaway. There may also
be reactive controls such as fire alarms, automatic module disconnects, or emergency
cooling systems to draw heat from the battery before the thermal runaway threshold is
reached. Either side of the Bowtie model may be expanded into multiple threat or
consequence layers, depending on the detail of the model.
The Bowtie model is the highest level analysis that can be done and may be performed in
tandem with or in lieu of a failure mode effects and criticality analysis (FMECA). The FMECA
process involves a listing of all possible failure modes and a relative ranking of the
probability of their occurrence. The Bowtie model adds a visual representation of the
incident paths, the consequence of their occurrence, the barriers that are in place to prevent

January 18th, 2017
Consolidated Edison
the occurrence, and the escalation factors that can either defeat barriers or increase the
probability of the event occurring. Escalation factors are typically included on the left hand
side of a Bowtie model and demonstrate how outside factors increase the likelihood of a
barrier failure. Barrier defeating mechanisms can occur on either side of the top event in the
figure, but are more commonly included in the right hand side. The list of possible failure
modes in the FMECA analysis is a rank order list of all possible incident pathways diagramed
in the Bowtie model. Thus the Bowtie model is descriptive and qualitative in nature, while
the FMECA analysis is more quantitative. The Bowtie output can easily be converted to a
FMEA output and vice versa. Together, the Bowtie and the FMECA listing can be used to
address risks and outline recommendations for improvement in safety systems.
Figure 37 BowTie analysis permits the visualization of threats to a top event, such
as loss of battery control, and ties these threats to consequences.

January 18th, 2017
Consolidated Edison
12.4 Heat Load from Li-ion Battery Failures
Figure 38 Battery weight and the peak room temperature are positively correlated.
13.0 APPENDIX 3: TESTING PLAN AND APPROACH

The total project scope for the Consolidated Edison-New York State Energy Research and
Development Authority (NYSERDA) BESS program is shown below. It includes four project
tasks with a final report, which also includes the development of guidelines and training
materials.
Literature Review
A literature review concisely summarizing the findings from previous safety testing
conducted on the specific battery chemistry families tested in this scope of work.
Additionally, a review of sodium sulfur and nickel sodium chloride batteries, not being tested
in this scope of work, was completed.
Chemistries Participating in the Program
1. NCM (4 vendors)
2. LiFePO4 (2 vendors)
3. LTO
4. Lead Acid
5. Vanadium Redox
6. An additional Li-ion chemistry described as BM-LMP

January 18th, 2017
Consolidated Edison
Small Scale Testing Parameters Measured

1. Heat release rate
2. Species and rate of release of gasses liberated during a burn and as a result of
application of suppression agents
3. Species and volume of liquids or solids released during a burn and as a result of
application of suppression agents
4. Perform limited suppression agent testing of a small number of suppressants: Water,
F-500, FireIce®, and PyroCool®. Testing of suppression release rates for water, or if
water is deemed ineffective or unsafe the next best candidate suppressant identified,
will also be performed.
5. Observe for presence of electrical arcing or mini-explosions and post burn re-ignition
Computer Modeling
Computer modeling was used to extrapolate small scale burn test results to larger scale fire
scenarios involving battery racks. A model at the system scale (rack level) was constructed
for each of the chemistries tested. Model predictions were validated through comparison
with burn testing of small units.
Final Report
The final report (this document) includes the following for each family of chemistries:
findings from the literature review, results from the small scale cell level tests, results of the
system size modeling, an assessment of risk at the system scale, effectiveness of
extinguishers and techniques, and any other code relevant findings that emerge. First
responder training materials and guidelines are also a deliverable from this report.
The testing program is designed to address two hazards: 1) toxic or flammable off gases as
well as solids and liquids released during the burn and during fire suppression, and 2) heat
load and release rate. The testing program is designed to determine what toxic and
flammable gases are present as a function of chemistry and when they are released during
the fire. The heat release data provides scalable data as a function of chemistry to
determine passive fire protection requirements (as part of container or room design), as
well as the quantity and duration of release for fire extinguishers.
13.1 Design of Experiments

Extinguisher tests were performed on cells that demonstrated the best burn properties for
testing. All module tests were also subject to extinguishing. Vanadium redox and lead acid
electrolyte tests were performed in an autoclave (without direct fire) to examine the
volatility of the electrolyte in high heat conditions. There were seven donated battery
chemistries to the program as well as two volunteer participants.
13.2 Combustion Gas and Particulate Matter Analysis

Of chief concern to the fire services and first responders are CO, O2, H2S and
LEL/combustible values. There are additional risks of fluoridated compounds (F2 and HF),
SO2, VOCs and H2. DNV GL monitored these during the tests using an FTIR gas analyzer
from Gasmet (Figure 39) as well as gas chromatography bags for post-test analysis.
Additionally, coupon sampling was performed to measure ash, soot and particulate matter
emitted and deposited during the fire, in addition to analysis of the battery debris. These
coupons and debris measurements will inform hazmat risks during overhaul and after fire
ground operations.

January 18th, 2017
Consolidated Edison
13.3 Heat Release Rate

ASTM9 tests were modified and combined to measure the heat release rate of the batteries.
Heat release rates as a function of time and fire stage were calculated using a thermopile
built around the battery as well as thermocouples around the chamber including at inlet and
outlet. DNV GL was able to quantify heat release rate (kJ/s or kW or BTU/min) and fire load
per mass of battery (BTU/lb. or kWh/kg). As standardized sizes and footprints do not yet
exist, these parameters provide better insight into the fire hazard than the typical ASTM
approach per unit area (per ft2 or per m2).
The power and energy of the fire per unit mass of battery provided data to estimate the
required extinguisher flow rates or mass. The heat removal potential of the extinguisher
was estimated by calculation prior to the extinguisher test by matching the battery mass to
the required extinguisher mass (mcT) with an added safety margin.
13.4 Procedure
The setup for all tests is depicted in the figure below. Additionally, all batteries underwent
multiple tests and state of charge (SOC) was varied to account for differences in energy
levels10. Battery voltages were measured during and after each test to determine their
potential for re-ignition, if any.
To filtration, metering and exhaust pump
Destructive Testing
Containment Chamber
(30”x30”x30”)
Flue Gas Direction
Flue Gas Gas Bag Sampling

Duct
Thermocouples
Gas Sampling Ports*

(real time)
Radiant HRR * O2, CO2, CO, H2S, SO2,
Heater and HF, F2, VOC, LEL
Thermopile
Ignitor Surface Deposition
Coupons
Morning Pride
Test Specimen
PPE Swatches
Air Inlets
Figure 39 Large abuse test chamber design for battery fire and extinguishing
testing.
9 ASTM 906: Standard Test Method for Heat and Visible Smoke Release Rates for Materials and Products Using a
Thermopile Method, ASTM 1354: Standard Test Method for Heat and Visible Smoke Release Rates for Materials
and Products Using an Oxygen Consumption Calorimeter, and ASTM E1623: Standard Test Method for
Determination of Fire and Thermal Parameters of Materials, Products, and Systems Using an Intermediate Scale
Calorimeter (ICAL)
10
Current plan is for testing at 50% and 100% SOC

January 18th, 2017
Consolidated Edison
13.4.1 Testing Procedure: Lithium Battery Gas Sampling, No Extinguishing11

Heating was accomplished with a radiative electrical heating element and hot point ignitor to
heat the lithium batteries to the point of sublimation or off gassing and ignite any
flammable gases produced or released. Testing was recorded with regular image and
thermal video.
 Step 1: Put battery in chamber, verify function of all sensors. Begin filming.
 Step 2: Compile gas sensor baselines, capture ambient gas bag for baseline
 Step 3: Initiate radiative heating element. Monitor temperature battery.
 Step 4: Gas bag sample. Monitor temperature and gas sensors. Heat rise may last
10 minutes to one or more hours. Monitor for flammables such as hydrogen and
VOCs.
 Step 5: Record increasing heat with thermocouple measurements. Gas bag sample
as appropriate
 Step 6: Monitor for peaking heat. Gas bag sample as appropriate. Monitor gas
sensor and thermocouples. This may occur in durations < 5 minutes.
 Step 7: Monitor decaying heat. Gas bag sample as appropriate. Monitor and record
gas sensor and thermocouple data.
 Step 8: Monitor decaying heat until temperatures reach safe levels. May take hours
or overnight.
 Step 9: Once battery remains are deemed safe, collect surface swabs, coupons, PPE
swabs, and secure battery in flame resistant enclosure for posttest observation.
 Step 10: Battery remains will be secured and monitored (video and temperature) for
24 hours. If re-ignition does not occur, batteries will be observed and intentionally
re-ignited the following day to observe remaining fire load. 12
13.4.2 Testing Procedure: Lithium Battery Gas Sampling With

Extinguishing13
With basic off gas testing complete, a range of extinguishing agents as well as water were
tested for effectiveness and reaction. Heat release rates were used to estimate required
extinguisher flow rates and volumes. Battery remains were stored in flame resistant
enclosures for 24 hours and monitored with video and thermocouples for re-ignition. Testing
was video recorded with regular image and thermal video.
 Step 1: Put battery in chamber, verify function of all sensors. Begin filming.
 Step 2: Compile gas sensor baselines, capture ambient gas bag for baseline
 Step 3: Initiate radiative heating element. Monitor temperature battery.
 Step 4: Gas bag sample. Monitor temperature and gas sensors. Heat rise may last
10 minutes to one or more hours. Monitor for flammables such as hydrogen and
VOCs.
as appropriate
11
Items in bold are actions to be determined as a function of testing progress – requires attentive monitoring by
technician.
12
No such events were observed.
13
Items in bold are actions to be determined as a function of testing progress – requires attentive monitoring by
technician.

January 18th, 2017
Consolidated Edison
 Step 6: Execute extinguisher based on recommended extinguisher use 14. Monitor and
record temperature and gas sensors. Gas bag sample immediately after.
or overnight.
 Step 9: Once battery remains are deemed safe, collect surface swabs, coupons, PPE
swabs, and secure battery in flame proof enclosure.
 Step 10: Battery remains will be secured and monitored (video and temperature) for
24 hours. If re-ignition does not occur, batteries will be observed and intentionally
re-ignited the following day to observe remaining fire load. Battery will be allowed to
burn out on its own to ensure complete destruction and remove change of re-
ignition.
13.4.3 Testing Procedure: Flow and Lead Acid Battery Electrolyte (liquid)
A sealed autoclave with heater was used to contain the test. A sample of either liquid
(vanadium redox) or acid soaked glass mat (Pb AGM) was placed in a smaller container
within the autoclave. The autoclave was heated and off gases measured.
 Step 1: Put electrolyte (liquid or wet glass mat) in autoclave, verify function of all
sensors. Electrolyte for each test will be taken from batteries charged to different
SOCs to maintain SOC variance in testing.
 Step 2: Compile gas sensor baselines, capture ambient gas bag for baseline before
heating
 Step 3: Initiate radiative heating element. Monitor liquid and ambient temperature.
 Step 4: Collect gas bag sample. Monitor temperature and gas sensors. If
electrolyte is not expected to heat exothermically, monitor that heat rise is
consistent with controller setting. Monitor for flammables such as hydrogen, VOCs,
and sulfuric gases (SO2 and H2S).
as appropriate
 Step 6: Continue heating to predetermined temperature 15. Collect gas bag sample
as appropriate. Monitor gas sensor and thermocouples. This may occur in durations
< 5 minutes.
 Option Step 7A: Attempt spark ignition. If fluid vapor is known to be inert, this step
shall be skipped.
 Option Step 7B: Execute extinguisher. Monitor and record temperature and gas
sensors. Gas bag sample immediately after.
14
Different extinguishers, including automated extinguishers, have different guidelines for use and deployment.
Execution of extinguisher will be based on FDNY recommendations and use cases.
15
Peak temperature for flow batteries may vary. Temperature may be based on common class A/B/C/D fire
temperatures to determine fluid behavior during boiling or combustion. Max testing temperature may specified
by NYSERDA or Con Ed.

January 18th, 2017
Consolidated Edison
or overnight.
13.4.4 Testing Procedure: Lead Acid Battery Lead (solid)16

As multiple risks exist with burning lead and lead oxides, a different approach was taken to
test the lead acid batteries. This test involved burning a small amount of lead in a simple,
class “A” fire17 to determine the amount of lead vaporized and deposited on the surrounding
surfaces. This testing took place in a tightly contained enclosure to minimize lead
contamination. As class “A” fires are not uncommon to the fire service, the focus was to
quantify the risk posed by lead and lead oxides.
 Step 1: Place small, known quantity of lead plate and lead oxide in class A material.
Material will be taken from batteries charged to different SOCs to maintain SOC
variance in testing.
 Step 2: Place coupons and ensure swab areas are clean, ensure container sealed
except for air inlets
 Step 3: Ignite class “A” materials
 Step 4: Collect gas bag sample, monitor temperature. Allow fire to burn out on its
own.
 Step 5: Let container sit, allowing lead vapor to settle
 Step 6: Open container with appropriate PPE, collect sample coupons, all solid waste,
and surface swabs.
 Step 7: Reseal container for disposal or re-use.
16
No HRR will be performed on the lead acid or flow components as the energy storage portions of these
technologies are non flammable, only the balance of system will add to the fire load.
17
Likely PTFE or PET plastic or basic construction materials (wood).

January 18th, 2017
Consolidated Edison
Table 18 Battery burn tests without extinguishing, combustible batteries only18

Stage Gases Analyzed Information Information Information from Heat Testing Standards used
of Fire Gained From Gas from Coupons Release Rate
Analysis and PPE
samples
Before (Background) Background gases, Turnout gear

fire baseline “as is”
measurement. condition. Bare
Coupon and coupons before
turnout gear prior contamination
to damage.
Incipien Sensors: CO, CO2, Toxic or flammable Early stage heat release Modified ASTM 906
t/ O2, H2S, HF, F2, gases during fire rate, potential (thermopile); modified
Ignition SO2, VOCs, H2, LEL incipient stage. combustibility of ASTM 1354 (O2
radiantly heated batteries consumption
calorimetry) ; modified
ASTM 1623 (intermediate
Gas scale calorimetry)
Chromatography
Bags, post test
analysis: VOCs,
fluoride
compounds, CO,
CO2, heavy
metals19
Rising Same Evolution of gases Accelerating heat release Same

heat as fire climaxes rate, O2 consumption,
CO production20,
thermopile temperatures
Heat Same Gas composition Peak heat loads, O2 Same

Climax during fire climax consumption
Decayin Same Gas composition as Heat decay rate Same

g fire fire evolves and
decays
Fully Same Background gases Determination of Same

decaye after fire has potential for re-ignition
d fire decayed completely
Debris (Background) Residues and Residues and Turnout gear after

HAZMAT HAZMAT exposure. Coupons for
conditions. considerations, SEM/EDAX/XRD. Ion
degradation to chromatography may be
PPE performed with swabs
from turnout gear.
18
See below test procedures for flow battery electrolytes
19
If contained within battery, based on MSDS
20
Compliments gas analysis

January 18th, 2017
Consolidated Edison
Table 19 Battery burn or heat21 tests with extinguishing

Stage of Fire Gases Information Gained Coupons and PPE Heat Release Testing
Analyzed From Gas Analysis samples Rate Standards
Used
Before fire (Background) Background gases, Turnout gear “as is”

baseline measurement. condition. Bare
Coupon and turnout gear coupons before
prior to damage. contamination
Incipient Sensors: CO, Toxic or flammable gases Early stage heat Modified ASTM
CO2, O2, H2S, during fire incipient stage. release rate, 906
HF, F2, SO2, potential (thermopile);
VOCs, H2, LEL combustibility of modified ASTM
radiantly heated 1354 (O2
batteries consumption
calorimetry) ;
Gas modified ASTM
Chromatography 1623
Bags, post test (intermediate
analysis: VOCs, scale
fluoride calorimetry)
compounds, CO,
CO2
Rising heat Same Evolution of gases as fire Accelerating Same

climaxes heat release
rate, O2
consumption, CO
production22
Heat Climax Same Gas composition during Peak heat loads, Same
fire climax O2 consumption
Extinguisher Same23 Changes in gas Changes in residues Heat removal Same

Deployment composition as a result as a result of rate24 achieved
of extinguishing extinguishing, with
HAZMAT impact extinguisher
Decaying fire Same Gas composition as fire Heat decay rate, Same
evolves and decays. ability to
Changes in gas sustain cooling
composition as a result with
of extinguishing extinguisher
Fully decayed Same Background gases after Accelerated

fire fire has decayed cool down rate
completely. Changes in with
gas composition as a extinguisher
result of extinguishing.
Debris (Background) Residual fumes. Changes Residues and HAZMAT Turnout gear
in gas composition as a considerations. after exposure.
result of extinguishing. Changes in residues Coupons for
as a result of SEM/EDAX/XRD.
21
Flow battery electrolytes may be heated to achieve the simulation of external heating due to a fire. Some flow
battery electrolytes are not expected to be exothermic.
22
Complements gas analysis
23
To be compared against benchmark “without extinguishing”
24
Evaluation of heat management as a result of extinguishing will inform firefighter extinguisher guidelines

January 18th, 2017
Consolidated Edison
extinguishing, Ion
HAZMAT impact, chromatography
degradation to PPE may be
performed with
swabs from
turnout gear.
Change in
residues as a
result of
extinguishing.
Liquid samples
for IC will
determine if
extinguisher
liquid residues
are toxic.
13.5 Large Scale Burns

Upon completion and evaluation of the small scale burn tests, and following or in parallel to
the modeling of the data from those tests, large scale tests, at the module or pack level or
bigger, were conducted to verify modeling results and determine unforeseen risks posed by
larger systems. This phase of testing was performed in conjunction with Rescue Methods
(RM) and involved the complete ignition of a full system or subsystem of an energy storage
unit comprised of cells of the previously tested chemistries. These tests took place in a
designated burn trailer used for the development of guidelines and training material for first
responders as well as testing the effectiveness of extinguishing agents on a larger scale.
Test units were secured overnight for observation of re-ignition and then intentionally re-
ignited 24 hours later to determine remaining fire load as well as to ensure complete
destruction for safe disposal. Samples of the remaining battery, as well as residual run-off
from the extinguisher and coupon samples from within the burn area were collected after
each test. Thermal and regular video was taken.

January 18th, 2017
ABOUT DNV GL
Driven by our purpose of safeguarding life, property and the environment, DNV GL enables
organizations to advance the safety and sustainability of their business. We provide
classification, technical assurance, software and independent expert advisory services to the
maritime, oil & gas and energy industries. We also provide certification services to customers
across a wide range of industries. Combining leading technical and operational expertise, risk
methodology and in-depth industry knowledge, we empower our customers’ decisions and
actions with trust and confidence. We continuously invest in research and collaborative
innovation to provide customers and society with operational and technological foresight.
Operating in more than 100 countries, our professionals are dedicated to helping customers
make the world safer, smarter and greener.

DNV Considerations For ESS Fire Safety 2017

Uploaded by

Document Informationclick to expand document information

Document Informationclick to expand document information

Copyright:

Available Formats

DNV Considerations For ESS Fire Safety 2017

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

DNV Considerations For ESS Fire Safety 2017

Uploaded by

Copyright:

Available Formats

Final Report

Considerations for ESS Fire

Report No.: OAPUS301WIKO(PP151894), Rev. 3

Task and Objective:

Please see Executive Summary.

Prepared by Verified by Approved by

☒ Unrestricted Distribution (internal and external) Keywords

DNV GL – OAPUS301WIKO(PP151894), Rev. 3 ii

The batteries tested in this program are as follows:

DNV GL – OAPUS301WIKO(PP151894), Rev. 3 iii

Sections Directly Informing Code Development and Training

1. Locations (see Locations and Ventilation on page 48)

DNV GL – OAPUS301WIKO(PP151894), Rev. 3 iv

3.0 HOW TO USE THIS DOCUMENT ................................................................... 3

4.0 TESTING METHODOLOGY ........................................................................... 4

5.0 CONSIDERATIONS FOR SYSTEM TYPES AND LOCATIONS ............................... 6

6.0 NEW FINDINGS AND ANSWERS TO FREQUENTLY ASKED QUESTIONS

7.0 PRESENT DAY INDUSTRY-ACCEPTED SAFETY PRACTICES FOR

8.0 GUIDANCE FOR FIRST RESPONDERS ......................................................... 34

9.0 FINDINGS RELATED TO CODES AND TRAINING .......................................... 40

DNV GL – OAPUS301WIKO(PP151894), Rev. 3 v

9.7 Project Development Considerations for Interaction with First

10.0 SUMMARY OF MAIN FINDINGS AND RECOMMENDATIONS ............................ 59

11.0 REFERENCES .......................................................................................... 62

12.0 APPENDIX 1: REFERENCED DATA .............................................................. 64

13.0 APPENDIX 3: TESTING PLAN AND APPROACH ............................................. 76

DNV GL – OAPUS301WIKO(PP151894), Rev. 3 vi

Table 4 Non-battery related codes for energy systems in buildings. ................................ 43

Table 7 Example calculation to determine the minimum water requirement per

Table 10 Example implications based on extrapolated findings from testing. It

Table 16 Aggregation of data regarding battery fires, extinguishing, emissions

DNV GL – OAPUS301WIKO(PP151894), Rev. 3 vii

Table 17 Summary of methodologies and results of the water requirement

Table 19 Battery burn or heat tests with extinguishing .................................................. 83

DNV GL – OAPUS301WIKO(PP151894), Rev. 3 viii

Figure 2 Configuration of module burn site. ............................................................. 6

Figure 3 A proactive, system level approach to extinguishing need not

Figure 7 Heating of Pb acid electrolytes yielded SO2. .............................................. 12

Figure 8 Heating of Pb acid electrolytes also yielded HCl. ....................................... 13

Figure 9 Observation of delayed cascading during extinguishing in a

Figure 10 Internal temperatures between two sandwiched pouch cells

Figure 12 Progression toward lower water requirements as testing

Figure 13 Residue analysis from a coupon hung in the headspace of the

Figure 14 Percent of mass loss as a function of cell form factor. ............................ 20

Figure 16 Estimated ventilation rates (air changes per hour) as a function

DNV GL – OAPUS301WIKO(PP151894), Rev. 3 ix

Figure 17 Estimated air changeover rate (air changes per hour) as a

Figure 19 Estimated air change rate (ACH) as a function of room volume

Figure 20 Representative emissions histogram from a Li-ion battery. .................... 25

Figure 21 The combined LEL and autoignition temperature of mixed gases

Figure 23 It is generally true that LiFePO4, LTO, and BM-LMP batteries

Figure 25 Project development timeline and the implementation of FMEA

Figure 26 Submerging batteries in water resulted in cooler temperatures,

Figure 27 Simplified diagram of fire impinging on the external wall of a

Figure 28 Direct fire exposure model to a steel wall with a 1/16”

Figure 29 Performance of water compared to other agents as water

Figure 30 Cooling performance of water compared to other extinguisher