Hochschule Offenburg University of Applied Sciences
Thermal runaway of lithium batteries
2006: DELL recalls 4 mio. laptop batteries
(China)
2013: Boeing dreamliner battery
May 2010: Hewlett Packard recalls 54.000 laptop batteries “Since the May 2009 recall, HP has received 38 additional reports of batteries that overheated and ruptured resulting in 11 instances of minor personal injury and 31 instances of minor property damage”
14.12.2011
Thermal runaway of lithium batteries
15.02.2013
Wolfgang G. Bessler, Hochschule Offenburg
Nanako Tanaka, Deutsches Zentrum für Luft- und Raumfahrt, Stuttgart
Michael Danzer, Harry Döring, Zentrum für Sonnenenergie- und Wasserstoffforschung, Ulm
Julian Mehne, Felix Bode, Wolfgang Nowak, Universität Stuttgart
Project funded by VolkswagenStifung, 01.08.2011-31.07.2014
Hochschule Offenburg University of Applied Sciences
Motivation: Battery safety
2013: Boeing dreamliner battery
2006: DELL recalls 4 mio. laptop batteries
Material decomposition Heat
Thermal Runaway
Triggering event
Manufacturing (e.g. particles)
Dendrite formation
Internal short-circuit
Thermal runaway mechanism:
Crash
Over- charge
Over- discharge External
short-circuit
Over- heating
Runaway = Chemistry + Heat transport
[email protected] 15.02.2013
Hochschule Offenburg University of Applied Sciences
The highly energetic active materials are separated by a porous film that is thinner than a human‘s hair.
CT image of a Li-ion battery, DLR Stuttgart
Repeat unit Separator Cylindrical cell
150 µm 15-25 µm 2.6 cm
Arora & Zhang, Chem. Rev. 2004
Safety versus energy
15.02.2013 [email protected]
Hochschule Offenburg University of Applied Sciences
15.02.2013 [email protected]
Li+e–
Microscale
Macroscale
3D thermal model
Electrochemistry & heating model
T TQ
Electro-chemical and micro-structural parameters
Thermal and macro-structural parameters
Trigger and runaway simulation
Monte Carlo
Stochastic para-meter variation
Scale interface
Experiments
Trigger of extreme events
Model validation
Model validation
Goal: Early-alert risk-aware battery management system Stochastic and optimal model-predictive control with constraints
Our approach
Hochschule Offenburg University of Applied Sciences
Li+e–
Microscale
Macroscale
3D thermal model
Electrochemistry & heat model
T TQ
Electro-chemical and micro-structural parameters
Thermal and macro-structural parameters
Trigger and runaway simulation
Monte Carlo
Stochastic para-meter variation
Scale interface
Experiments
Trigger of extreme events
Model validation
Model validation
Microscale heat source model
15.02.2013 [email protected]
Hochschule Offenburg University of Applied Sciences
Microscale model components
• Performance model for standard operating conditions
– Heat sources due to charge, discharge, cycling
Runaway trigger
• High-temperature degradation model
– Additional heat sources due to thermal decomposition reactions
Runaway chemistry
15.02.2013 [email protected]
Hochschule Offenburg University of Applied Sciences
Performance model for standard operation
• Thermodynamics • Half-cell potential
• Kinetics • Butler-Volmer kinetics • Concentration overpotential
• Transport in particles • Mass conservation • Spherical diffusion
• Electrolyte transport • Nernst-Planck equation • Charge neutrality
• Cell voltage
−−−
= actact0)1(expexp ηαηα
RTF
RTFii
concLieqact )()( ηφ∆φ∆η −−= ct
( ) Viiii
iii
i sMy
cDyRT
FzycD
ytc +
∂∂
+
∂∂
=∂∂
∂∂
∂∂ φε
Diffusion Migration Chemistry
( ) 0=∑i
ii zc
=
)(ln 0
tcc
zFRT
concη
E = φcathode – φanode
zFcSTcH
zFGc )(Δ)(ΔΔ)( LiLi
Lieq−
−=−=φ∆
izFM
rDr
rrtLiLi2
2Li 1
−
∂∂
∂∂
=∂∂ ρρ
Diffusion Chemistry
Hochschule Offenburg University of Applied Sciences
Results: Discharge at various C-rates
• Different discharge (C) rates
• Relatively flat discharge curve, voltage variation mainly from C6 electrode
• Decrease of voltage and slight capacity loss upon increasing discharge rate
0.0 0.5 1.0 1.5 2.0 2.52.0
2.2
2.4
2.6
2.8
3.0
3.2
3.4
3.6
Experiment Simulation
Capacity [Ah]
0.1C 1C 2C 4.6C
Cell V
olta
ge [V
]
LiFePO4 cell
15.02.2013
Hochschule Offenburg University of Applied Sciences
Results: Discharge at various temperatures
• 1C discharge at different temperatures
• Loss of performance at decreasing temperature:
– Increasing polarization losses
– Decreasing capacity
• Good agreement between model and experiment
0.0 0.5 1.0 1.5 2.0 2.52.0
2.2
2.4
2.6
2.8
3.0
3.2
3.4
3.6
50°C 30°C 20°C 10°C 0°C -10°C -20°C
Cell V
olta
ge [V
]
Capacity [Ah]
Polarization losses
Capacity losses
LiFePO4 cell
15.02.2013
Hochschule Offenburg University of Applied Sciences
High-temperature degradation model
• There are a large number of potential high-temperature degradation reactions.
• Included so far in our model:
– Solid electrolyte interface (SEI) decomposition (CH2OCO2Li)2 Li2CO3 + C2H4 + CO2 + 0.5 O2
– SEI re-formation 2 C3H4O3 (EC) + 2 e– + 2 Li+ (CH2OCO2Li)2 + C2H4
– Electrolyte evaporation C3H4O3 (liquid) C3H4O3 (gas)
• Parameterization of thermodynamic and kinetic parameters performed
15.02.2013 [email protected]
Hochschule Offenburg University of Applied Sciences
Solid electrolyte interface (SEI)
• What’s SEI (Solid Electrolyte Interface)? − Passivating layer between electrode and
electrolyte − Arises from the reductive decompositions
of a small amount of organic electrolytes − Composes mostly during the first several
cycles of a working cell
• Why SEI decomposition?
− Triggering event for reaction of electrolyte and electrode which possibly leads thermal runaway •http://www.cmt.anl.gov
US department of Energy
Graphite
[email protected] 15.02.2013
Hochschule Offenburg University of Applied Sciences
Results: Heat sources upon cell heating
• Numerical experiment: Constant heat-up of cell, simulation of heat source (differential scanning calorimetry, DSC). Cell is assumed isothermal.
SEI decomposition
SEI re- formation and de- composition
15.02.2013 [email protected]
Hochschule Offenburg University of Applied Sciences
Thermal runaway example simulation
• Runaway induced by thermal SEI decomposition:
(CH2OCO2Li)2 Li2CO3 + C2H4 + CO2 + ½ O2
• Self-accelarating reaction until SEI is completely consumed
[email protected] 15.02.2013
Hochschule Offenburg University of Applied Sciences
Li+e–
Microscale
Macroscale
3D thermal model
Electrochemistry & heat model
T TQ
Electro-chemical and micro-structural parameters
Thermal and macro-structural parameters
Trigger and runaway simulation
Monte Carlo
Stochastic para-meter variation
Scale interface
Experiments
Trigger of extreme events
Model validation
Model validation
Macroscale heat transport model
15.02.2013 [email protected]
Hochschule Offenburg University of Applied Sciences
Modeling approach
• 3D, 2D and 1D model of single cell
• Solution of heat transport equation
• Boundary conditions: Convection and radiation
• Implementation in COMSOL
15.02.2013 [email protected]
( ) ( ) TP QT
tTC +∇∇=
∂∂ λρ
Hochschule Offenburg University of Applied Sciences
Coupling the scales
• 6 representative points chosen for coupling of macroscale with microcale
15.02.2013 [email protected]
Li+e–
Microscale
Macroscale
3D thermal model
Electrochemistry & heat model
T TQScale interface
Hochschule Offenburg University of Applied Sciences
Exemplary simulations
• Full 3D simulation compared to 1D simulation
• Here: Surface temperature vs. time
• Nominal operation, discharge in 1 h (1C rate)
• Complex temperature behavior
15.02.2013 [email protected]
1D simulation
3D simulation
Hochschule Offenburg University of Applied Sciences
Li+e–
Microscale
Macroscale
3D thermal model
Electrochemistry & heat model
T TQ
Electro-chemical and micro-structural parameters
Thermal and macro-structural parameters
Trigger and runaway simulation
Monte Carlo
Stochastic para-meter variation
Scale interface
Experiments
Trigger of extreme events
Model validation
Model validation
Stochastic modeling and analysis
15.02.2013 [email protected]
Hochschule Offenburg University of Applied Sciences
Uncertainty and model error
● Parameter uncertainty
– Incomplete knowledge about model parameters
– For a “perfect” model: still uncertain results
● Model error
– Model per definition ≠ reality
– Structural error (wrong equations, wrong numerical implementation, …)
• Problem: Uncertain model → prediction and reality deviate increasingly with time
• Measurement errors
Concept: Measurement updates
15.02.2013 [email protected]
Hochschule Offenburg University of Applied Sciences
𝑝 𝑥0,…,𝑡 𝑦0,…,𝑡 𝑝(𝑥0,…,𝑡|y0,…,t−1)
Complete sequential procedure (simplified):
Update of uncertain model predictions with measurements via Bayes‘ theorem:
System model: 𝑥𝑡 = 𝑓(𝑥𝑡−1,𝜇𝑡)
Measurement model: 𝑦𝑡 = 𝑔(𝑥𝑡, 𝜈𝑡)
𝑥𝑡 … model state at time t 𝑦𝑡 … measurement at time t 𝜇𝑡 …. model error 𝜈𝑡 … measurement error
𝑝(𝑥0,…,𝑡−1|𝑦0,…,𝑡−1)
𝑝 𝑥𝑡 𝑦0, … ,𝑦𝑡 =𝑝 𝑦0, … ,𝑦𝑡 𝑥𝑡 ⋅ 𝑝(𝑥𝑡)
𝑝 𝑦0, … ,𝑦𝑡
prediction
information loss
information gain
update
Solution of model equations with a particle filter: - Continuous probability density is discretized by particles (individual model runs) - Measurement update via reweighting of the particles
𝑡 = 𝑡 + 1
15.02.2013 [email protected]
Approach: Bayesian filtering
Hochschule Offenburg University of Applied Sciences
Explanatory power of measurements
• How much does a measurement tell about a quantity of interest?
• Here: Prediction of cell core temperature as function of surface temperature for a number of simulation runs under uncertainty
15.02.2013 [email protected]
Hochschule Offenburg University of Applied Sciences
Li+e–
Microscale
Macroscale
3D thermal model
Electrochemistry & heat model
T TQ
Electro-chemical and micro-structural parameters
Thermal and macro-structural parameters
Trigger and runaway simulation
Monte Carlo
Stochastic para-meter variation
Scale interface
Model validation
Model validation
Experiments
Trigger of extreme events
Thermal runaway experiments
15.02.2013 [email protected]
Hochschule Offenburg University of Applied Sciences
Investigated cells
• Three different commercial cell types investigated.
• Influence of cell chemistry and design on runaway propensity?
15.02.2013 [email protected]
Hochschule Offenburg University of Applied Sciences
Types of cell tests
• General characterisation (weight, volume, capacity, power and energy densities)
• Nominal operation characteristics as function of temperature and charge/discharge rate
• Abuse experiments – External heating – Short circuit – Nail penetration – Overcharge
15.02.2013 [email protected]
Hochschule Offenburg University of Applied Sciences
Exemplary results: Temperature during cycling
• Here: Sony cell
15.02.2013 [email protected]
Hochschule Offenburg University of Applied Sciences
Nail penetration tests
15.02.2013 [email protected]
Hochschule Offenburg University of Applied Sciences
Summary
• Thermal runaway is an outstanding example for an extreme event in a complex technical system
• Ongoing project using combined modeling and experimental approach
• Microscale model of heat sources and runaway chemistry
• Macroscale model of heat transport
• Stochastic model of model and measurement uncertainty
• Comprehensive experiments using three different battery types
15.02.2013 [email protected]
Thank you for your attention!
Thermal runaway of lithium batteries
15.02.2013
Wolfgang G. Bessler, Hochschule Offenburg
Nanako Tanaka, Deutsches Zentrum für Luft- und Raumfahrt, Stuttgart
Michael Danzer, Harry Döring, Zentrum für Sonnenenergie- und Wasserstoffforschung, Ulm
Julian Mehne, Felix Bode, Wolfgang Nowak Universität Stuttgart
Project funded by VolkswagenStifung, 01.08.2011-31.07.2014
Hochschule Offenburg University of Applied Sciences
15.02.2013 [email protected]
Example simulation runs: Surface temperature update