Plug‐in Hybrid Powertrain Configuration C i U i Gl b l O i i iComparison Using Global Optimization

SAE 2009‐01‐1334

April 2009

Dominik Karbowski, Karl‐Felix Freiherr von Pechmann

Sylvain Pagerit, Jason Kwon, Aymeric RousseauArgonne National Laboratory

Sponsored by Lee Slezakp y

This presentation does not contain any proprietary, confidential, or otherwise restricted information

PHEVs Come in Various Powertrain Configurations: How to Compare Them?Configurations: How to Compare Them? Series, Parallel, Power‐split, etc.

Prototypes exist but built upon different requirements Prototypes exist, but built upon different requirements, sizes, electric ranges, etc.

PHEV control allows for much freedom in the power split

Simulation tools allows comparison on common grounds: Forward‐looking, such as Argonne’s PSAT

• Mimics real‐world chain of decision includes transient drivability• Mimics real‐world chain of decision, includes transient, drivability features 

• Rule‐based controls need to be tuned

Global Optimization/Dynamic Programming Global Optimization/Dynamic Programming• Backward‐looking and cycle‐dependent approach makes this a theoretical tool

• Control is an output, each vehicle operates “at its best”p , p

Global optimization is ideal tool for “fair” comparison

Study ProcessDefine Vehicle Requirements Industry standards

Size VehiclesImplement  + Automated

Sizing

Optimally Si l t

optimization code for various configurations

Sizing

tSOCtP nm ,

0X

Simulate Vehicles

nm

tJtSOCtP 11

tU0 tU M

...

Different cycles,Different distances

Analyze data

tJtSOCtP '0,000 ,1',1 tJtSOCtP MMMM ',' ,1,1

TX

...

Outline

Vehicle Sizing

Global Optimization Algorithm

Simulation Results

Requirements and AssumptionsType Requirement Comments

All‐electric Range 10 mi (16 km) on the UDDS 90% to 30% SOC

Grade  6%@65 mph (105 km/h) @GVWR, w/o the battery

0‐60 mph (97 km/h) time 9.3s @ 30% SOC

Maximum speed >100 mph (160 km/h)  

Parameter Value

Glider mass 1142 kg

Drag coefficient 0.28

Iterative automated routine using PSAT:

Parallel Split Series

EV UDDS EM ESS EM ESS EM ESSFrontal area 2.3 m2

Rolling resistance 0.009

Wheel Radius 0.332 m

EV on UDDS (power)

EM, ESS EM,ESS EM,ESS

AER (energy) ESS ESS ESS

0‐60 time  ICE ICE,EM2, EM, ICE,

Comparable to Toyota Camry, Chevrolet Malibu

(power) EM EM2

Grade (power) ICE ICE,EM2,EM

ICE,EM2

EM: Electric Machine - ESS: Energy Storage SystemEM: Electric Machine ESS: Energy Storage System

Sizing ResultsParallel Split Series

Parallel, ICE power (kW), 73.9

Split, ICE power (kW), 90.1

Split, EM1 power (kW), 78.1Series ICE power (kW) 72 6

Series, EM1 power (kW), 100.2

Series EM2 power (kW) 72 6

Parallel Split Series

Parallel, EM1 power (kW), 42.1

Parallel, ESS power (kW), 49.9Split, EM2 power (kW), 55.7

Split, ESS power (kW), 50.5

Split, ESS capacity (Ah), 18.4

Series, ICE power (kW), 72.6 Series, EM2 power (kW), 72.6

Series, ESS power (kW), 52.7

Parallel, EM2 power (kW), 0

Parallel, ESS capacity (Ah), 19.5

Split, ESS capacity (Ah), 18.4

Series, ESS capacity (Ah), 17.8

Bigger ICE and EM1 for the Power‐Split to meet the 0‐60 time: ICE cannot operate at maximum power at low vehicle speeds

Big EM1 on the Series, because it is the sole source of movement Big EM1 on the Series, because it is the sole source of movement

Battery in the Parallel config. has higher capacity because of lower EV efficiency

Difference in mass is less than 50 kg (~1800 kg) Difference in mass is less than 50 kg ( 1800 kg)

Outline

Vehicle Sizing

Global Optimization Algorithm

Simulation Results

Global Optimization Uses a Backward Model, Based on Look-up TablesBased on Look-up Tables

ICE ESSICE GEN

ICE ACCESS

Thermal Elec. Mech.

ACC

MG

ACC ESSMG2MG1

TC

GEARBOX

MG

TC

TC

PLANETARY

DRIVELINEDRIVELINEDRIVELINE

Parallel Pre-Transmission

Power-Split SeriesTransmission

Engine (ICE), electric machines (MG) : input power (thermal or electric) is function of output torque and speed

Battery voltage and resistance is function of SOC Battery voltage and resistance is function of SOC

Optimization Problem State: Battery State‐of‐charge (SOC)

Command u: gear and engine torque (Parallel)– gear and engine torque (Parallel)

– generator power (Series)

– engine power (Power‐split)

Link between state and command Link between state and command:

Initial and final conditions: initial SOC between 90 and 30% , final SOC=30%, 

Cost function to minimize:

Penalty function: avoid excessive gear changes and engine starts/stops

Constraints:– Vehicle follows drive trace

– Keep ICE/EM speed/torque and battery current within limits

– Series: ICE torque and speed s.a. gen‐set operates optimally for a given power

– Power‐split: ICE torque and speed s.a. ICE operates optimally for a given powerp q p p p y g p

Method: application of Bellman’s principle

Outline

Vehicle Sizing

Global Optimization Algorithm

Simulation Results

Example of Simulation Output: 2nd Hill of the UDDS4

333

x 104

1

2

1

2

1

2

-1

0

-1

0

-1

0

Parallel SplitSeries

150 200 250 300-2

time (seconds)

150 200 250 300

-2

time (seconds)

150 200 250 300-2

time (seconds)

Time (s) Time (s) Time (s)( )( )ICE Power [W]

ESS Power [W]

Veh. Speed [m/s] x1000

Same initial SOC (60%)

Similar behavior for each configuration

( ) ( ) ( )

p

“Macroscopic” Analysis of Results for Different SOC YDifferent SOC

SOC.9

Y

1 vehicle1 cycle1 distance

calculation.8

.6

1 run .4

.3

0.60 0.1 0.3 0.5SOC

Y

time Algorithm outputs results for various Initial SOC

It is therefore possible to look at a macroscopic

Plug-to-Wheel Wh/km

It is therefore possible to look at a macroscopic result Y (e.g. RMS current, average power, etc.) for different electric consumption values

Plug to Wheel Wh/km

In Series Configuration, ICE Starts at High P D dHigher Power Demand

4x 10

4 Power at the Wheels Above Which ICE Is ON 95% of the Time

3

4

parallelseries

Parameter similar to “Engine ON” threshold used in rule‐based controls

2

3

er (W

)

se essplit Linear trend for each 

configuration (except at near EV‐operation points)

f

1

2

Pow

e Similar values for power‐split and parallel configuration

Series power at wheels is higher => ICE is ON less often and at

0 50 100 1500

=> ICE is ON less often and at higher loadsUDDS x1

0 50 100 150Electric energy consumption (Wh/km)

In Series, ICE Also More Likely To Charge Th B ttThe Battery

ng

Ratio of ICE Energy Used to Charge the Battery0.15

or C

harg

in parallelseries

In the parallel configuration, charging is more costly: ICE energy is converted twice

the Battery

0.1

rgy

used

fo split energy is converted twice before getting to the wheels, while there is a direct path

Double conversion also occurs 

0.05

f IC

E E

ner

in Series, but no direct path existsUDDS x1

0 50 100 1500

Electric energy consumption (Wh/km)

Rat

io o

f

Electric energy consumption (Wh/km)

In Series Config., ICE is Used at Higher Power for Higher Efficiency Higher Efficiency

Average Engine Power (when ON)

Different trends:1.5

2x 104

Different trends:– Series = independent of Wh/km

– Parallel = higher for higher Wh/km

Series: ICE power is much higher 

1

Pow

er (W

)

parallel p g

0 50 100 1500

0.5

Electric energy consumption (Wh/km)

seriessplit

0.37

0.38 parallelseries

UDDS x1

Electric energy consumption (Wh/km)

0.35

0.36

0.37

Effi

cien

cy

seriessplit

Overall Engine Efficiency

Trends similar to average power

S i i t l t i l ffi i

0 50 100 1500.32

0.33

0.34E

Series = engine operates close to maximal efficiency (36%)

Parallel = much lower efficiency in CS mode

Power Split = in betweenUDDS x1

0 50 100 150Electric energy consumption (Wh/km)

Energy Consumption : On The UDDS, Power Split Seems to Be the Best CompromiseSeems to Be the Best Compromise

Charge‐Sustaining: Series 5

m)

parallel

config. uses most fuel, due to poor electric transmission efficiency

Parallel uses the least fuel3

4

n (L

/100

km

parallelseriessplit

Parallel uses the least fuel, as engine‐to‐wheel path is more efficient

In electric‐only, parallel is 2

3

nsum

ptio

n

y, ppenalized by the transmission efficiency

Power‐Split performs well i b h i i0

1

Fuel

Con

in both situations0 50 100 150

0

Electric energy consumption (Wh/km)

Similar Hierarchy When Using Other Cycles I h t i i d th diff b t th i fi ti In charge‐sustaining mode, the difference between the series configuration 

and the two other ones is more important on more energy‐intensive cycles (HWY) or energy‐aggressive (LA92)

I EV ti th d t f th i i l i ifi t LA92

5

6

In near‐EV operation, the advantage of the series is less significant on LA92 and HWY than on the UDDS

LA92Parallel Series

4

5

on (L

/100

km)

LA92

HWY

UDDS32 km

∞ km

SeriesPower‐Split

Green = HWYBlue = LA92Red = UDDS

80

100

120

(km

/h)

HWY

LA92

2

3ue

l Con

sum

ptio

LA92

HWY

16 km

∞ km

(charge‐sustaining)

20

40

60

80

Veh

icle

Spe

ed (

0 50 100 150 2000

1

Fu

LA92

UDDS

HWYUDDS12 km

24 km

0 500 1000 15000

Time (s)

UDDS

Electrical Energy Consumption (Wh/km)

Conclusion

Comparing different powertrains requires a common set of requirements

An automated routine using PSAT allows to quickly size vehicles

Global optimization can be used to ensure “fair” comparison, because control is optimal

Series is potentiallymore efficient on the “electric‐only” side, parallel (pre‐tx) is potentiallymore efficient on the “charge‐sustaining” side, power‐split is a good compromise

Global optimization is a theoretical tool, and in real‐world optimal control is harder to get because:– it requires the knowledge of the cycle beforehand

– real‐world must also take into account drivability, safety, etc.

However it provides the theoretical boundaries, as well as control patterns

Future work will use patterns obtained from global optimization to implement adaptive rule‐based controls.