2-Li-Ion Battery Models for HEV SimulatorCHAMAILLARD

7/31/2019 2-Li-Ion Battery Models for HEV SimulatorCHAMAILLARD

1/22

25 / 11 / 2008Advances in Hybrid Powertrain

Li-ion battery models for HEV simulator

M. Debert, G. Colin, M. Mensler, Y. Chamaillard, L. Guzzella


2/22

2Advances in Hybrid Powertrain 25 / 11 / 2008 Li-ion battery models for HEV simulator

Contents

Introduction

Hybrid Electric Vehicle a solution to reduce fuel consumption

Simulation of HEV

Why Li-ion?

Battery model

Overview of literature

Three models tested and SOC calculation

RINT model

RCSERIES model

RCPARALLEL model

Thermal Aspect

Effect

Modelling Validation & results

Conclusion


3/22


Boost

Hybrid Electrical Vehicle is a solution to consume less fuel

How can a hybrid vehicle consume less fuel?

By Downsizing the internal combustion engine (with the same performances)

By turning-off the engine when the vehicle is stopped (Start & Stop) By optimizing the energy distribution between the sources

By recovering a part of the kinetic energy during deceleration Drawback of this technology

The additional cost

The additional weight

BrakingZEVEngine onlyrecharging

INTRODUCTION

EnergyManagement


4/22


INTRODUCTION

Simulation is useful for the Energy Management development

The energy management performances are tested and compared in simulation

Some of these Energy management uses Battery State Of Charge (SOC)

HEV simulator requirement

Good battery behaviour prediction

Reasonable time of simulation

Quick adaptation for different cells (constant innovation in cells technology)


5/22


INTRODUCTION

Why Li-ion?

Li-ion has high Energy density

Li-ion has high Specific Energy*

Li-ion has no memory effect

Li-ion has a low self-discharge rate

* i.e. : Gasoline has a Specific Energy of 13 kWH/kg


6/22


Batteries Models

Usually transformed into equivalent circuit Overview of battery modeling for HEV simulator & thermal aspect

L.Guzzella A.Sciarretta, Vehicle propulsion systems.

Quasistatic modeling

VH. Johnson, Battery performances in ADVISOR.

Overview of battery dynamics

A. Jossen, Fundamentals of battery dynamics.

Dynamic modelling

A. Capel Mathematical model for the representation of the electrical behaviour of a lithium cell.

E. Kuhn, C. Forgez, P. Lagonotte, G. Friedrich, Modelling Ni-mH battery using Cauer and Foster structure.

S. Buller, M. Thele, R. De Doncker, W. Karden, Impedance-based non-linear dynamic battrey modeling forautomotive applications.

Electrochemical models L.Guzzella A.Sciarretta, Vehicle propulsion systems.

Black box model

Thermal Aspect A. Pesaran, A Battery thermal models for hybrid vehicle simulations.

N. Sato Thermal behavior analysis of lithium-ion batteries for electric and hybrid vehicles.


7/22


BATTERIES MODELS

Three models tested

Battery State of Charge

RINT RCSERIES RCPARALLEL


8/22


BATTERIES MODELS

RINT model

OCV: non linear function of battery State of charge

DCR: non linear function of battery State of Charge which isdifferent in case of charge or discharge


9/22


BATTERIES MODELS

Open Circuit Voltage (OCV)

Polynomial fitting

On static points

Spline

On static points

Neural Network

Perceptron with one hidden layer

Sigmode for input layer

Linear for output layer

For confidential reasons battery cells voltage is normalisedwith respect to its maximum value (in %)


10/22


Batteries Models

RINT model results

FIT criterion (evaluation of the correlation between model prediction and battery behaviour

Quasi-static model

2.005 2.007 2.009 2.011 2.01380

80.5

81

81.5

82

82.5

83

time (s)

Ubat

(%)

DATA

RINT

70

72

74

76

78

80

82

84

Ubat

(%)

1.8 2 2.2 2.4 2.6 2.8 3 3.2 3.4

x 104

-505

Time (s)

error(%

)

RINT

DATA

FIT= 85% (identification)

FIT= 79,3% (validation)


11/22


Batteries Models

RCSERIES model


R: Constant which is different in case of charge or discharge

CSERIES: Constant which is different in case of charge or discharge


12/22


Batteries Models

RCSERIES models identification

RCSERIES results

Non physical behaviour (due to the saturation)

6 undetermined parameters : CSERIES(ch/disch), R (ch/disch), satmin, satmax

A simplex algorithm find the solution which minimises the criterion :

70

72

74

76

78

80

82

84

Ubat

(%)

1.8 2 2.2 2.4 2.6 2.8 3 3.2 3.4

x 104

-5

05

Time (s)

error(%)

RCseries

DATA

2.005 2.007 2.009 2.011 2.01380

80.5

81

81.5

82

82.5

83

time (s)

Ubat

(%)

DATA

RCSERIESFIT= 89.6%

(identification)

FIT= 89.5%(validation)


13/22


Batteries Models

RCPARALLEL models

R1,R2: Constant

CSERIES: Constant


Hz < Good behaviour < kHz


14/22


Batteries Models

RCPARALLEL Identification

Problem formulation

Normalisation

Ibat

, SOCare centred around 0 and

New equation

Continuous parametric

model of type output error

White noiseFirst order with one poleN order which increasesidentification performances


15/22


Batteries Models

GPMF

OE

SYS

LT

SYS

LT

Min


16/22


Batteries Models Output Error


17/22


RCPARALLEL Identification

Thanks to the continuous methods the analogy gives:

Discrete identifications (ARX, ARMAX, Box Jenkins) gives good results but theanalogy is not direct and the solution depends on the sampling period.

Batteries Models


18/22


Batteries Models

RCPARALLEL Results

Another experiments with this model and method

70

72

74

76

78

80

82

84

Ubat

(%)

1.8 2 2.2 2.4 2.6 2.8 3 3.2 3.4

x 104

-505

Time (s)

erro

r(%)

RCPARALLEL

DATA

2.005 2.007 2.009 2.011 2.01380

80.5

81

81.5

82

82.5

83

time (s)

Ubat

(%)

RCPARALLEL

DATA

FIT= 91.7%(identification)

FIT= 91.6%(validation)

2000 2050 2100 2150 2200 2250 230080

81

82

83

84

85

time (s)

Ubat

(%)

RCparalelle

Data


19/22

19Advances in Hybrid Powertrain 25 / 11 / 2008 Li-ion battery models for HEV simulator 19

Batteries Models

Model performances

Improvement with temperatureconsideration?

91.6%91.7%RCPARALLEL

89.5%89,6%RCSERIES

79.3%85%RINT

Validationdata

Identificationdata

2.532 2.534 2.536 2.538 2.54 2.542 2.544

x 104

78

78.5

79

79.5

80

80.5

81

time (s)

Ubat

(%)

RCparallel

Data

RCserie

Rint


20/22


THERMAL ASPECT

Temperature effect

cold condition

Increase the internal resistance

Decrease the Open Circuit Voltage Decrease the capacity

Temperature modelling

1.9 2 2.1 2.2 2.3

x 104Time (s)

Ubat

data 0C

data 25C

Joule effect + cells reaction (which can be neglected)

Integral of the difference between heat produced andheat exchange

Convection & conduction


21/22


THERMAL ASPECT

Temperature model validation

Implementation & Results

1.04 1.045 1.05 1.055

x 105

0

5

10

15

20

Time(s)

Temperature(

C)

Experiment

Simulation

The prediction error is low ~3C max

hand Cpare identified in a 4Cdischarge at 25C in initial condition

Temperature model validation is madewith the same test but at 0C in initialcondition

0

1

2

3

4

5

6

7

Error(%)

1.04 1.0405 1.041 1.0415 1.042 1.0425 1.043 1.0435

x 105

0

10

20

Time (s)Temperature

(C)

RCPARALLEL

with temperature

RCPARALLEL

without temperature


22/22


CONCLUSION

RCPARALLEL model

Improvement of the battery behaviour prediction for the HEV simulator

Most part of battery dynamic is represented

Fast identification with a continuous identification method

Temperature consideration

Cells are extrapolated to give battery pack

Improvements & future work

Development of specific strategy in cold condition

Model with the effect of mass transportation (all frequencies represented)

Heating Interaction between cells in the battery pack

Battery pack model with n-cells (effect of dispersion, State Of health) Battery management system consideration

Documents

2-Li-Ion Battery Models for HEV SimulatorCHAMAILLARD