C BD Battery Design LLC Advanced Thermal Modeling of Batteries - - PowerPoint PPT Presentation

Advanced Thermal Modeling of Batteries

C

BD

Battery Design LLC

Empirical Battery Models Empirical Battery Models

STAR Asia Conference December 2013

Overview

• Battery Design Process
• Use of Physics-based model for synthetic

data generation

• Parameters for RCR Table Model
• Evaluation of RCR Table Model Fits
• Conclusion

2

Overview of Battery Design Process with Empirical Model

3

Test cell Fit data to model Use model to simulate battery performance What procedure(s) to use? What temperatures? What model? How to validate fit? Can battery deliver required performance? Is heat‐transfer adequate to ensure

• max. temperature

not exceeded and uniform?

Selected Simulation Models in Battery Design Studio

Page 4 Page 4 (1) T. Fuller, M. Doyle, J. Newman, J.

• Electrochem. Soc. 141 (1994) 1-10

(2) Battery Design LLC, “BDS Documentation”

• M. Verbrugge and R. Conell, J.
• Electrochem. Soc. 149 (2002) A45-A53

Consumer Electronics, EV

NTG

HEV/PHEV Module/Pack

RCR

Cell Design

DISTNP

Simple, easy to create model, best for simple discharge/charge, thermal analysis Quick response for frequent charge/ discharge like HEV/PHEV Useful for design. Solves transport, kinetics, equations

4

(1) J. Newman and W. Tiedemann, J.

• Electrochem. Soc. Vol. 140, No. 7, July 1993 pp.

1961-1968. (2) H. Gu, J. Electrochem. Soc., Vol. 130 No. 7 1983 pp. 1459-1464. (3) U. S. Kim, Ch.B. Shin, C.-S. Kim, J. Power

• Src. 189 (2009) 841-846

DISTNP model allows prediction of cell performance based

• n materials

properties and cell design.

Battery Physics – DISTNP Model

5

• T. Fuller, M. Doyle, J. Newman, J. Electrochem. Soc., Vol. 141 (1994) 1‐10

DISTNP model can simulate aging

6

Impedance rise and capacity loss due to SEI growth.

4C rate 2C rate

Lithium‐Ion Battery Transport Solid‐ and Liquid‐Phase Gradients

• A. Nyman, T. G. Zavalis, R. Elger, M. Behm, G. Lindbergh “Analysis of the Polarization in a Li‐Ion Battery Cell by Numerical

Simulation” J. Electrochem. Soc., 157(11) A136‐A1246 (2010).

7

DISTNP model allows sources of voltage loss in cell to be quantitatively assigned to specific physical processes.

Electrolyte diffusion polarization

dx j x c F c RT j

L L c L L tot

  



 2 1

Kinetic Overpotential

 





L ave surf loc tot

dx E E aj j 1

Contact Resistance Losses

contact appl R

j 

Liquid-Phase Ohmic drop



L eff L tot

dx j j

2

1 

Solid-Phase Ohmic drop



L eff S tot

dx j j

2

1 

Solid Diffusion Polarization

 



 

L surf L S loc tot

dx E aj j 1  

Approach to parameterize RCR model in BDS

Generate synthetic data with DISTNP model Use BDS gap tool to parameterize RCR model Simulate HPPC test using RCR model Simulate Drive Cycle using RCR model Compare to synthetic data

Questions:

• For HPPC test, how sensitive are fitted

parameters/simulation results to:

• pulse duration? 2 s, 10 s, 30 s
• pulse currents? 5 C rate versus 10 C rate
• Temperature? 10, 20, 40, 50C
• How well does temperature interpolation work?

8

DISTNP Model Cell Used for Generation of Synthetic Data

Model uses two different particle sizes, realistic electrolyte properties, temperature dependent solid-phase diffusion and kinetics.

Generate Synthetic Data

Goal should be to emulate

good physical testing.

Testing should be done under controlled temperature and ideally cell temperature will be uniform.

• Thin, pouch cell is ideal for

characterizing electrochemical behavior.

Cell can be clamped with aluminum plates and placed vertically in environmental chamber to provide well- defined boundary conditions for heat transfer between cell and chamber.

10

Generate Synthetic Data (cont.)

Hybrid Pulse Power Characterization (HPPC) Test provides convenient, standardized method to characterize impedance and voltage relaxation of battery of state of charge from 90 to 10%.

http://www.uscar.org/guest/teams/11/U-S-DRIVE- Electrochemical-Energy-Storage-Tech-Team

11

Analysis of Voltage Losses in HPPC Test

Contact resistance, a purely ohmic loss, is major source of voltage loss. At 10C the voltage loss due to electrolyte diffusion polarization and activation overpotential are much larger than at 30C. At 30 the voltage loss is mainly due to contact resistance and electrolyte diffusion polarization.

12

Breakdow n of voltage losses in HPPC Discharge Pulse

Major source of voltage drop is contact resistance between electrode coating and current collector

13

RCR Table Model

Parameters can be entered as tables. Interpolation between values as a function of state of charge is done using Bezier splines. Linear interpolation is used to compute values at intermediate temperatures.

Model is easy to use and computationally efficient.

14

Gap tool in BDS provides automatic regression of USABC Equivalent Circuit Model (RCR type)

User simply selects data file with HPPC data and Gap tool automatically generates table of RCR parameters.

15

RCR Parameters can be entered into BDS RCRTable Model

One table for each temperature can be entered. Linear interpolation is used to obtain parameter values at intermediate temperatures.

16

Parameterized RCRTable Model

Entered parameters for 10, 20, 40 and 50 °C

17

Drive Cycle Simulation: US06 PHEV Charge Depleting

A Ah

18

Comparison of computation times

RCRTable model is ~ 104 times faster than DIST model

45.9 min 0.2 s

19

COMPARISON OF RCR MODEL TO SYNTHETIC DATA

20

Comparison of Fitted HPPC test to synthetic data

RMS error = 17 mV 10°C, 5 C rate, 10 s pulse

21

Drive Cycle Simulation at 10C: Voltage

RMS error = 12 mV 10°C, 5 C rate, 10 s pulse

22

Drive Cycle Simulation at 10C: Temperature

RMS error = 0.08 °C

10°C, 5 C rate, 10 s pulse

10 12 14 16 18

23

Drive Cycle Simulation at 30C: Voltage Parameters Interpolated from 20 and 40C

RMS error = 12.8 mV

30°C, 5 C rate, 10 s pulse

24

Drive Cycle Simulation at 30C: Temperature Parameters Interpolated from 20 and 40C

RMS error = 0.108 °C

30°C, 5 C rate, 10 s pulse

As time progresses differences in temperature accumulate.

30 32 34 36 25

Summary of Drive Cycle Simulations

* Interpolated

Comments

• Predictions are very good, 2 s give best overall voltage simulation
• Fits at 10C rate were comparable

T, °C 5C Rate - RMS Error milliVolts

Pulse Time, s

T, °C 5C Rate - RMS Error degrees C

Pulse Time, s 2 10 30 2 10 30

10 7.7 12.0 10 0.044 0.012 20 13.8 13.9 13.0 20 0.163 0.149 0.063 30 13.2 12.8 17.3 30 0.136 0.108 0.253 40 8.7 13.1 13.3 40 0.202 0.155 0.153 average 10.8 13.0 14.5 average 0.14 0.11 0.16

55 31.6 % SOC

26

Drive Cycle Simulation at 20C: Voltage

RMS error = 33 mV

20°C, 5 C rate, 10 s pulse

70 23.4% SOC

27

Drive Cycle Simulation at 20C : Temperature

RMS error = 0.64 °C

10°C, 5 C rate, 10 s pulse

70 23.4% SOC ~ 15°C temperature rise

20 25 30 35

28

Summary

• Physics-based model is useful for generating

synthetic data

• Realistic simulations can account for different

particle sizes, electrode formulations, separators, electrolytes, kinetics, aging

• RCR Table model is computationally efficient

and provides excellent prediction for data sets where voltage losses are mainly ohmic.

• Predicted voltage and temperature correspond

closely for parameters obtained over range of rates and temperatures

29