1

Drive Cycle Development and Real-world data in the United States

Edward NamUS Environmental Protection Agency

WLTP meeting, Geneva, SwitzerlandJanuary 15, 2009

Working paper No. WLTP-02-17GRPE Informal Group on WLTP, 2nd Meeting January 2009, Agenda Item 6.

2

Outline

• Drive cycle development history in US– FTP & off-cycle – Inventory cycles– Other cycle development

• SAFD vs Vehicle Specific Power• New driving activity data• US EPA 5-cycle fuel economy labeling

(drive cycle weighting)• Next steps

Apologies for the American units!

3

Why drive cycles are important• Serve as a standardized measurement stick for

emissions and fuel economy• Can compare across vehicles (benchmark)• Manufacturers design vehicles to meet standards

set by cycles and test procedures• Serves as proxy for “average” or typical driving• Emissions standards are strongly dependent on

the cycle and test procedure• Drive cycles also change over time, with

infrastructure, policy (speed limits), and technology (power:weight)

4

FTP cycle developmentUS Federal Test Procedure

• LA4 (City, UDDS) Developed in the late 1960’s to describe typical driving (acquired in Los Angeles)

• The highway HFET cycle was developed to describe a typical rural route (acquired outside Ann Arbor, MI in the 1970s)

• These 2 cycles used to describe fuel economy in the US

• US06 (aggressive) and SC03 (A/C) cycles developed based on “3 cities data” of instrumented vehicles from Baltimore, Spokane, Atlanta. – REP05, REM01, SC03 were developed to cover the full

range of driving, but were simplified to US06 and SC03, while FTP remained in place

– These cycles are extreme to prevent “cycle beaters”

5

Inventory Cycles - California• Implemented new base cycle

– LA92 – determined from driving in LA in early 1990’s (40kph)

– More representative of 1990’s driving• Facility Cycles (speed correction factors)

– 12 unified correction cycles (UCC) mainly from chase car data in LA

– From 4 to 95kph– Chosen by mean speed, speed-acceleration

frequency distribution, positive kinetic energy (PKE), load, maximum acceleration, maximum deceleration, percent idle, percent acceleration, distance, etc

– These cycles are used to correct the base emission factor from LA92 to other speeds

6

CARB LA92 &

CO2 speed correction

factor

0.0

0.5

1.0

1.5

2.0

2.5

3.0

3.5

4.0

4.5

5.0

0 5 10 15 20 25 30 35 40 45 50 55 60 65

Speed (mph)

Cor

rect

ion

Fact

or

CO2 CARB CO2 FI CO2 TBI

0.0

10.0

20.0

30.0

40.0

50.0

60.0

70.0

0 200 400 600 800 1000 1200 1400time

spee

d (m

ph)

Bag 1Bag 2

7

US EPA Inventory Cycles

• USEPA - Facility cycles (1997)– Developed by Sierra Research from 3 cities

chase car data (Baltimore, Spokane, LA)– 11 cycles based on roadway type and

congestion level (+ramp)– Each cycle lasts ~10 minutes– Matched second-by-second segments of

chase car data by comparing SAFDs (speed acceleration frequency distribution)

– Can find on EPA website under MOBILE6 technical support documentation

8

Speed/Acceleration Frequency Distribution (SAFD)

_ 0_ 1 0

_ 2 0

_ 3 0

_ 4 0

_ 5 0

_ 6 0

_ 7 0

_ 8 0- 6 - 5 - 4 - 3 - 2 - 1 0 1 2 3 4 5 6

0

0 .0 1

0 .0 2

0 .0 3

0 .0 4

0 .0 5

0 .0 6

f r e q u e n c y

s p e e d

a c c e l

9

EPA & Sierra cycle specifications• time in acceleration/deceleration • time at cruise • time at idle • maximum speed • average speed • average or predominant speed during cruise • maximum acceleration/deceleration rate • maximum power • length (time and miles) • stops per mile • average positive kinetic energy (PKE) change per mile

and specific power• distributions of speed and acceleration

10

USEPA Cycle development (cont’d)

• Microtrips were chosen based on how well they matched the specifications

• Cycle choice criteria – lowest sums of differences on SAFD– matching real world power (2va) – Segments shortened or lengthened

• Cycles used for (emissions) “speed correction factors” and for speed dependent transportation planning in MOBILE6 (emissions inventory model)

11

UC-Riverside CE-CERT International Vehicle Emission Model

• data collection and cycle development for rapid emissions inventory estimation– Almaty, Kazakhstan – Beijing, China – Lima, Peru – Mexico City, Mexico – Nairobi, Kenya – Pune, India – Santiago, Chile – Sao Paulo, Brazil – Shanghai, China

http://www.issrc.org/ive/

12

Vehicle Specific Power: an alternate metric for cycle manipulation

• MOVES is EPA inventory model replacing MOBILE• MOVES is a modal model based on VSP activity• Cycle metric should be based on a more physically causal

variable for emissions formation: e.g. • Road load (tractive) Power • P ~ Av + Bv2 + Cv3 + Mva*

– A,B,C are vehicle target coastdown coefficients– * including road grade

• VSP = P/M (M is mass of vehicle)– Divided by mass since emissions (measured in g/km) is largely

independent of vehicle mass• With VSP distributions and proper modal data, emissions can

be converted from one drive cycle to another• This approach has been validated by a number of studies

13

VSP Distribution in MOVESDistribution of T im e by M ode

0%

10%

20%

30%

Brak ingIdl ing <0 0-3 3-6 6-9

9-12 >1 2 <0 0-3 3-6 6-99-12

12-1818-24

24-30 >3 0 <66-12

12-1818-24

24-30 >3 0

VSP/Sp e e d Bin

Perc

ent o

f Tim

e

Rura l Restricted Urban Unrestricted

0-25 mph0-40 kph

25-50 mph40-81 kph

>50 mph, >80kph

kW/tonne

14

Cycle Development requires data

• Much second by second data has already been collected

• But these were not collected with harmonization in mind, so data is scattered and inconsistent

• Require new and more rigorous analysis methods

15

Real world driving data since 3 cities & LA92

• Chase car:– Los Angeles 2000

• Instrumented Vehicle:– US EPA Ann Arbor shootout data 2001 – Kansas City data 2004-2005– Atlanta (Georgia Tech) 2001-2004

• 1600+ vehicles, 800+ households, GPS, accelerometer, OBD

• Vehicles instrumented for 2-3 years• Most comprehensive activity data in existence• Data excellent source for start activity as well• Due to privacy concerns, data has a finite lifetime

16

Los Angeles ComparisonSpeed

More time is spent at high speeds in 2000 than in 1992.Speed limits were increased during this period

00.020.040.060.080.1

0.120.140.16

0 5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90

Speed (miles per hour)

Frac

tion

of A

ctiv

ity Los Angeles 1992Los Angeles 2000

17

Los Angeles ComparisonAcceleration

More time is spent at high acceleration in 2000 than in 1992.

0.00

0.02

0.04

0.06

0.08

0.10

0.12

>3 mph/s

Acceleration (miles per hour per second)

Frac

tion

of A

ll Acc

eler

atio

ns

Los Angeles 1992

Los Angeles 2000

18

1-25 25-50 50 +30 + 16 30 4027-3024-27 29 3921-2418-2115-1812-15 279-12 15 256-9 14 24 353-6 13 230-3 12 22 33< 0 11 21

VSP

Cla

ss (k

W/to

nne)

Speed Class (mph)

28 38

37

Operating Modes for MOVES VSP bins

For coast and cruise,

13 modes retainedfrom MOVES2004, plus

8 modes added forMOVES2006 formerly bins26 and 36

PLUSOne mode each for

idle, and decel/braking

--------------------------------Gives a total of23 opModes

19

Los Angeles in 1992 compared to 2000Los Angeles then and now

0

0.02

0.04

0.06

0.08

0.1

0.12

0.14

0.16

0.18

0 1 11 12 13 14 15 16 21 22 23 24 25 26 33 35 36

VSP bin

Frac

tion

of a

ctiv

ity

Los Angeles 1992 (no idle)Los Angeles 2000 (no idle)

<25 mph 25-50 mph >50mph

20

Other Factors:California (2000) Urban vs Rural

0

0.05

0.1

0.15

0.2

0.25

0.3

0.35

0 1 11 12 13 14 15 16 21 22 23 24 25 26 33 35 36

VSP Bin

Freq

uenc

y of

Act

ivity Urban California

Rural California

<25 mph 25-50 mph >50mph

Rural driving is faster than urban driving

21

SouthEast Michigan Shootout data

• 15 PEMS instrumented vehicles.• Driven by US EPA or SENSORS

employees in calendar year 2001.• Note: Drivers were not randomly chosen.• Avg. distance driven per vehicle : 52 miles• Average speed : 31.2 mph

22

Kansas City (Round 1.5 Only)• US EPA study 2004-2005• Instrumented vehicles (not chase car) from

random population of newer vehicle owners• PEMS (Portable Emissions Measurement

System) equipped to measure emissions and activity.

• Drivers measured for 15,000-30,000 seconds (battery lifetime)

• Measured conventional as well as hybrid vehicles• Avg. distance driven per vehicle: 41 miles• Average speed: 30.1 mph

23

Speed acceleration frequency distributions

• Based on a 2-dimensional matrix of speed and accelerations

24

Recent Activity Surveys

Recent surveys have the more high speed, high power driving.

LA driving is the most aggressive, & may be comparable to US06

Activity comparison of MOVES, SE Michigan, Kansas City, and Los Angeles

0

0.02

0.04

0.06

0.08

0.1

0.12

0.14

0.16

0.18

0 1 11 12 13 14 15 16 21 22 23 24 25 26 33 35 36

VSP bin

Freq

uenc

y of

act

ivity

MOVES no idleSE Michigan no idleKC no idleLA 2000 (no idle)

<25 mph 25-50 mph >50mph

25

Conventional vs Hybrid Activity

0

0.02

0.04

0.06

0.08

0.1

0.12

0.14

0.16

0.18

0.2

0 1 11 12 13 14 15 16 17 18 19 21 22 23 24 25 26 27 28 29 33 35 36 37 38 39

VSP Bin

Freq

Total ActivityNon-HybridHybrid

<25 mph 25–50 mph >50 mph

Hybrid driving is slightly less aggressive than conventional vehicles

26

5-Cycle fuel economy labeling• Previously, label fuel economy was based on a 2

cycle number: city and highway– Real-world fuel economy effects were captured with a

20% correction factor• Thesis: Real-world driving can be described by a

linear combination of existing drive cycles• If weighted properly the activity and fuel economy

can be captured better than a fixed correction factor • The key is to use cycles with a broad enough range

to capture the VSP profile– US driving activity is mainly represented by a linear

weighting of 3 cycles: FTP(city), HFET, and US06

27

VSP distribution of FTP, HWY & US06VSP frequency for City, Highway, and US06 Cycles

0

5

10

15

20

25

30

35

0 1 11 12 13 14 15 16 21 22 23 24 25 26 33 35 36 37 38 39VSP bin

Freq

% (b

y tim

e)

UDDS CityHighwayUS06

<25mph_ 25-50 mph >50mph

28

Other relevant 5 cycle issues• Through a combination of FTP (city), highway,

US06, cold FTP, and hot SC03 (air conditioning) real-world driving was bracketed

• Result: fuel economy label became more representative of what vehicle owners would truly get throughout a year of driving

• Potential lessons for harmonization:– VSP weighting methodology can be a powerful tool– May be able to represent different regions through a

combination (and proper weighting) of drive schedules

29

Comparison of FTP, NEDC, JC08

• The NEDC and JC08 have similar VSP profiles – in hot start, should have similar emissions

• Main difference likely due to sequence following starts

0

0.05

0.1

0.15

0.2

0.25

0.3

0.35

0.4

0.45

0.5

0 1 11 12 13 14 15 16 21 22 23 24 25 27 28 29 30 33 35 37 38 39 40

VSP bin

Freq

uenc

y

FTPNEDCJC08

low speed medium speed high speed

30

Conclusions• US has a long history of drive cycle

development• Much new data exists in the US for a

harmonization project (with more coming)• Driving has been getting faster and more

aggressive with each passing decade• VSP activity is a powerful (yet simple) tool

to characterize and combine drive cycles and to compare driving from different regions

31

Issues for future consideration

• Definition of “city” vs “highway” (urban vsextra-urban)– City can include high speed driving and

highways can be congested stop and go• Starts/km variability• Shift schedules• Scaling drive cycle for small engine “cars”

32

Next steps• We should analyze the existing data especially

the Atlanta data before it is destroyed• Should agree on QA/QC procedures for data• As well as a methodology for cycle generation

and comparing “representativeness”• The US EPA is going to start collecting second

by second activity and emissions data from real-world operation from a variety of cities

33

Appendix

34

Data Limitations• OBD data is noisy when used for accelerations• GPS is noisier than OBD• Should make an effort to determine best filtering

mechanism compared to directly measured accelerations – Accelerometer, 5th wheel, etc.

• Do accelerations matter?– Certification tests have x% eror band for drivers to

follow cycles– But Robot drivers and computer simulations can

follow cycles exactly, so the exact cycle trace can matter