COE Control Seminar-11/01-2002- 2

arc POWER MANAGEMENT STRATEGY FOR A PARALLEL HYBRID ELECTRIC TRUCK Introduction and Motivation

Simulation model and preliminary rule-based control

Dynamic programming techniques

Improved rule-based control law

Simulation results

Dyno test results

Conclusions Chan-Chiao (Joe) Lin, Huei Peng, Jessy W. Grizzle The University of Michigan

arc What is a Hybrid Vehicle?

• A HV has at least two sources of motive power

- Prime: IC engines, fuel cells, gas turbines

- Secondary: Batteries, flywheels, ultra- capacitors, hydraulics

• The most popular type: ICE + battery/motor (HEV)

• Bottom-line: “Better” than any single power source.

COE Control Seminar-11/01-2002- 2

1 Potential Steps Towards Breakthrough arc in Fuel Efficiency

90 HEV

DI ∆ 10mpg 80 80 ∆ 15mpg 70 MT

Small engine ∆ 10mpg Aero-roll 50 Light-weight ∆ 8mpg

40 ∆ 4mpg ∆ 6mpg 30 27

0 EstimatedMetro-Highway Fuel Economy (mpg) Description: Current Taurus Ultralite Aero/Roll Conv. P/T Match Opt. P/T Match Adv. P/T Adv. Hybrid P/T Weight (lbs): 3200 2000 2000 2000 2000 2000 TBD Powertrain: 3.0 L, 2-valve Same Same 1.4 L, A/T 1.4 L, Auto M/T 1.2L CIDI, Auto M/T Same Source: Mike Schwarz, Ford Motor Company COE Control Seminar-11/01-2002- 3

HEV Configurations arc Ê Series Î Simpler mechanical transmissions Î Optimum engine efficiency and low emission Î Both ICE and Electric drive rated to the maximum power requirement (main disadvantage) Î lower overall efficiency (main disadvantage)

Ê Parallel (e.g. Chrysler MYBRID ESX2) Î More flexibility—sizing and control design more complicated Î Lower power/weight/cost, and better efficiency.

Ê Series-parallel combination (e.g. Toyota Prius)

COE Control Seminar-11/01-2002- 4

2 arc Growing Market (Toyota)

1997 1998 1999 2000 2001 2002/ 1997- Period 1-3 2002/3 Model

Prius 323 17,653 15,243 19,011 29,459 7,402 89,091

Estima - - - - 5,886 5,840 11,726 Hybrid minivan Crown - - - - 1,574 520 2,094 mild hybrid system

Coaster 9 3 12 15 9 8 56 hybrid (bus)

Total by 332 17,656 15,255 19,026 36,928 13,770 102,967 year/month

Cumulative 332 17,988 33,243 52,269 89,197 102,967 total

http://www.toyota.com/about/news/environment/2002/04/22-1-hybrid.html

COE Control Seminar-11/01-2002- 5

Advanced Powertrain arc

50 FUEL CELL SYSTEM USING H2 (GM Data)

40 HSDI DIESEL

30 Average Car Pass. Power G-DI ENGINE

10 PERCENT THERMAL EFFICIENCY Bus/Truck Average Average Bus/Truck Power

0 0 20406080100120 PERCENT LOAD Source: Ricardo Why hybrid vehicles? - load leveling - regenerative braking

COE Control Seminar-11/01-2002- 6

3 Transportation Sector Contribution to Oil arc Gap

Stacy C. Davis, ORNL, TRANSPORTATION ENERGY DATA BOOK: EDITION 21

COE Control Seminar-11/01-2002- 7

arc Ground Vehicle Energy Use

12 Actual Projected

10 Class 3-8 Trucks

8 Class 1-2 Trucks (Pickups, Vans, SUVs) 6 (90% 2a, 10% 2b)

4 Automobiles 2

0 Energy Use - Million Barrels per Day 1970 1980 1990 2000 2010 2020 source: EIA Annual Energy Outlook, 1998 Federal Highway Administration, Highway Statistics

COE Control Seminar-11/01-2002- 8

4 arc Emission

Ê Engine out vs. tail-pipe

Ê Temperature effect?

This presentation: Engine out, hot engine only V. Johnson, NREL COE Control Seminar-11/01-2002- 9

arc Nox Emission

Stacy C. Davis, ORNL, TRANSPORTATION ENERGY DATA BOOK: EDITION 21 COE Control Seminar-11/01-2002- 10

5 arc PM Emission

Stacy C. Davis, ORNL, TRANSPORTATION ENERGY DATA BOOK: EDITION 21

COE Control Seminar-11/01-2002- 11

arc Hybrid Vehicle Control Problems Main-loop Servo-loop Performance Fuel economy/ NVH/driveability/fuel target emission economy/emission “Control signal” High level Low level (throttle, (Power) current, duty cycle) Horizon/dynamics Long/slow Short/fast (detailed) (simple) To satisfy Government/ Consumers Green consumers Scenarios Driving cycles Test matrix Algorithm Optimal, Robust, adaptive, intelligent ad-hoc

COE Control Seminar-11/01-2002- 12

6 Power Management Problem for Parallel HEV arc

• Goal: To develop a control strategy to minimize fuel consumption (and emission) while meeting the power demand from the driver

• Key problem: determine the total amount of power to be generated, and its split between the two power sources

• Maintain adequate battery energy (charge sustaining)

RegenerativeRechargeMotorEngineHybrid Mode ModeMode: Mode Braking: When? : WhenWhen?: When Mode and and how the much split? power?

Engine Transmission

Eng. Command Driver Control Gear Command Module Motor Command

Battery Electric Motor

COE Control Seminar-11/01-2002- 13

Preliminary Rule-Based Control Strategy arc

Power assist Desired vehicle Actual vehicle 0.65 DriverFuel Consumption (kg/kW hr) speed speed Ê Low power 500

450 0.6 4 .2 Æ Motor only 0.26 0 27 the battery is charged by the 400 0. SOC 5 2 2 1 0. 23 . 2 0.55 0 . 350 0

Power request0 engine at a constant “recharge . Ê Medium power 2 300 1 Battery SOC 4 0.5 0 200power”400 600 level800 1000until1200 SOC1400 hits1600 1800the2000 250 6 Æ Use engine 1 .2 0.2 0 Time (sec) 200 2

Engine Torque (Nm) Engine Torque max boundary Charge 150 0.23 0.24 0.25 Ê High power 100 0.26 Sustaining .27 0 Pmot = −Pch , Peng =ÆPreqBoth+ Pch engine Strategy 50 Braking 800 1000 1200 1400 1600 1800 2000 2200 2400 & motor Engine Speed (rpm) Control

Motor only Power Split Recharge Control Control • Regenerative braking • Friction brake • Motor only mode • Engine provides • Engine only mode additional power to • Hybrid mode charge the battery

COE Control Seminar-11/01-2002- 14

7 Problems of Existing (Main-loop) Control arc Approach

Ê Based on simple concepts (e.g., load leveling) and mostly static maps.

Ê Tuning is done through trial-and-error.

Ê Don’t know the full potential (is 20% improvement good?)

Ê Not re-useable (fuel economy vs. fuel economy plus emission).

COE Control Seminar-11/01-2002- 15

arc 2003 Fuel Economy Champions

Honda Civic Honda Insight Toyota Prius (Manual) (Automatic) (Automatic)

Gas Mileage 46/51 57/56 52/45 (City/Highway)

http://www.fueleconomy.gov/feg/feg2000.htm

COE Control Seminar-11/01-2002- 16

8 arc Outline Ê Introduction and Motivation

Ê Simulation model and preliminary rule-based control results

Ê Dynamic programming techniques

Ê Improved rule-based control law

Ê Simulation results

Ê Dyno test results

Ê Conclusions

COE Control Seminar-11/01-2002- 17

Baseline Truck: Navistar International arc 4700 Series Engine Vehicle/Driveline Ê V8 DI Diesel Ê Total Mass: 7258 Kg Ê Turbocharged, Intercooled Ê Wheelbase: 3.7 m Ê 7.3 liters Ê CG Location: 2.2 m from front Ê Bore: 10.44 cm Ê Frontal Area: 5 m2 Ê Stroke: 10.62 cm Ê Air Drag Coefficient (CD): 0.8 Ê Compression Ratio: 17.4 Ê 4 Speed Automatic Transmission Ê Rated Power: 210 HP@2400 rpm Ê Rear Wheel Drive - 4x2

From www.navistar.com COE Control Seminar-11/01-2002- 18

9 arc Schematic of the Hybrid-Electric Truck

Vehicle Dynamics

Traction Force Engine Exhaust DS Gas Drivetrain Ê Parallel hybrid T EM PS ICM TC Trns D Motor Ê Post-transmission type C IM

Air Inter cooler DS

Power Control Module Battery

COE Control Seminar-11/01-2002- 19

arc Hybrid Truck Parameters

DI Diesel Engine V6, 5.475L, 157HP/2400rpm DC Motor 49kW Capacity: 18Ah Module number: 25 Lead-acid Battery Energy density: 34 (Wh/kg) Power density: 350 (W/kg) Automatic 4 speed, GR: Transmission 3.45/2.24/1.41/1.0 Vehicle Curb weight: 7504 kg

COE Control Seminar-11/01-2002- 20

10 arc HE-VESIM in Simulink Hybrid Electric Vehicle Engine SIMulation

Load Input Data

T pump w eng w eng T pump Eng cmd T motor T shaft DIESEL ENGINE Gear Load Output Variables w shaft w motor cyc_mph clutch cmd w trans DRIVELINE Dring Cycle

Motor cmd Current T wheel HEV Current soc w motor T motor w wheel DRIVER Controller ELECTRIC MOTOR Brake BATTERY Slope v veh VEHICLE DYNAMICS 0 slope

Double Click Double Click to Close All to Plot Result

Assanis, D.N. et al. “Validation and Use of SIMULINK Integrated, High Fidelity, Engine-In-Vehicle Simulation of the International Class VI Truck,” SAE Paper No. 2000-01-0288 Lin, C.C., Filipi, Z.S., Wang, Y., Louca, L.S., Peng, H., Assanis, D.N., and Stein, J.L., “Integrated, Feed-Forward Hybrid Electric Vehicle Simulation in SIMULINK and its Use for Power Management Studies”, SAE Paper No. 2001-01-1334

COE Control Seminar-11/01-2002- 21

arc Preliminary Rule-Based Control

Ê Braking rule Fuel Consumption (kg/kW hr) Power 500

4 450 .2 0.26 0 assist 27 400 0. 5 2 2 1 0. 23 . 2 0 . 350 0

. 2 300 1 Ê Power split rule 4

250 6 1 .2 0.2 0 200 2 Engine Torque (Nm) Torque Engine

150 0.23 0.24 0.25 100 0.26 0.27 Ê Recharge rule 50 Motor 800 1000 1200 1400 1600 1800 2000 2200 2400 onlyEngine Speed (rpm)

FE NOx (g/mi) PM (g/mi) (mi/gal) Conventional Truck 10.343 5.3466 0.5080 Hybrid Truck (Preliminary Rule-Base) 13.159 5.7395 0.4576

COE Control Seminar-11/01-2002- 22

11 arc Outline Ê Introduction and Motivation

Ê Simulation model and preliminary rule-based control results

Ê Dynamic programming technique

Ê Improved rule-based control law

Ê Simulation results

Ê Dyno test results

Ê Conclusions

COE Control Seminar-11/01-2002- 23

Dynamic Optimization Based Algorithm arc

Objective • To develop an optimal control policy that minimizes the fuel consumption over a giving driving cycle Advantage • Explore the full potential of hybridization • Benchmark for a better control strategy Methodology • A sequence of decisions applied to a dynamic system to minimize the total fuel costs --> Multistage Decision Process • Optimal control problem u(0) u(1) u(N −1) x(1) x(2) x(N −1) x(0) f (0) f (1) f (N-1) x(N)

L(0) L(1) L( N −1)

COE Control Seminar-11/01-2002- 24

12 Problem Formulation arc

• Discrete-time Nonlinear System x(k +1) = f (x(k),u(k)) • Control Inputs Fuel injection rate and gear shift command to the transmission

• Constraints ωωω≤≤()k eee__min max TkTkTkωω()≤≤ () () ee__min() e e max () e

Tmm_min()ωω(), k SOC () k≤≤ T m () k T m _max () m (), k SOC () k

SOCmin≤≤ SOC() k SOC max • Minimize the Cost Function N −1  JLxkukGxN=+∑ ()(),() (( )) k=0 N −1 2 =∑ fuel ( k )+⋅µυα NOx ( k ) +⋅ PM ( k ) + ( SOC ( N ) − SOC f ) k=0 COE Control Seminar-11/01-2002- 25

Dynamic Programming (DP) Algorithm arc

Ê State and control values are quantized into finite grids

x x J Ê Solve recursive equation backwards in time k +1 Jˆ* L k +1 Step N-1: k * JxNN −1( (−= 1)) min[ LxNuNGxN ( ( − 1), ( −+ 1)) ( ( ))] uN(1)− Step k, for 0 ≤ k < N −1

**k k +1 Jxkkk( ( ))=++ min LxkukJ ( ( ), ( ))+1 ( xk ( 1)) uk() * Interpolation to find J k +1 Ê Obtain an optimal, time-varing, state-feedback control policy

uxkk*((),) Æ Stored in a table for each of the quantized states and time stages

Ê DP creates a family of optimal paths for all possible initial conditions

COE Control Seminar-11/01-2002- 26

13 Procedure of Dynamic Optimization arc

Dynamic Optimization Process

Driving Cycle

Reduced order Dynamic Programming HEV model

Optimal Control Law, u(x(k),k)

Simulation (Complete HE-VESIM )

Fuel Economy, Vehicle Response

COE Control Seminar-11/01-2002- 27

EPA Urban Dynamometer Driving Schedule arc for Heavy-Duty Vehicles (UDDSHDV) Cycle

UDDSHDV 40 Actual 20 0

Veh Spd (MPH) Spd Veh 0 100 200 300 400 500 600 700 800 900 1000 0.58

Initial SOC 0.56 SOC 0.54 Final SOC 0 100 200 300 400 500 600 700 800 900 1000 100

50 Eng Pwr Eng 0 0 100 200 300 400 500 600 700 800 900 1000 40 20 0

Mot Pwr Mot -20 0 100 200 300 400 500 600 700 800 900 1000 Time (sec)

COE Control Seminar-11/01-2002- 28

14 arc Fuel Economy Comparison

FE (mi/gal) NOx (g/mi) PM (g/mi) Conventional 10.343 5.3466 0.5080 Truck Hybrid Truck (Preliminary 13.159 5.7395 0.4576 Rule-Base) DP µ ==0,υ 0 13.705 5.627 0.446

N−1 2 J=()()()(())∑  fuel k+⋅µυα NOx k +⋅ PM k + SOC N − SOC f k=0

COE Control Seminar-11/01-2002- 29

arc Fuel Economy and Emission Results

N−1 2 J=()()()(())∑  fuel k+⋅µυα NOx k +⋅ PM k + SOC N − SOC f k=0 µ ∈{0,5,10,20,40} υ ∈{}0,100,200,400,600,800

0.49 5.6 µ=[0, 5 10, 20, 40], ν=0 µ=10 µ=0 µ=10, ν=[0, 100, 200, 400] ν=400 ν=0 µ=40 µ=20 µ=20, ν=[0, 200, 400, 600] ν=0 ν=0 µ=40, ν=[0, 400, 600, 800, 1000] 5.2 µ=20 0.46 ν=600 µ=40 µ=5 ν=1000 µ=5 ν=0 4.8 ν=0 µ=0 0.43 ν=0 µ=10 4.4 ν=0 NOx Emissions (g/mi) Emissions NOx PM Emissions(g/mi) 0.4 µ=[0, 5 10, 20, 40], ν=0 4 µ=10 µ=40 µ=10, ν=[0, 100, 200, 400] ν=400 =40 ν=0 µ=20, ν=[0, 200, 400, 600] µ µ=40, ν=[0, 400, 600, 800, 1000] ν=1000 3.6 0.37 12.8 12.9 13 13.1 13.2 13.3 13.4 13.5 13.6 13.7 3.8 4 4.2 4.4 4.6 4.8 5 5.2 5.4 5.6 5.8 Fuel Economy (mi/gal) NOx Emissions (g/mi)

COE Control Seminar-11/01-2002- 30

15 arc Outline Ê Introduction and Motivation

Ê Simulation model and preliminary rule-based control results

Ê Dynamic programming technique

Ê Improved rule-based control law

Ê Simulation results

Ê Dyno test results

Ê Conclusions

COE Control Seminar-11/01-2002- 31

Design Procedure by Dynamic Optimization arc

Dynamic Optimization Process

Preliminary Driving Cycle Rule-based Controller Dynamic Programming

Optimal Control Law, u(x(k),k)

Improved Simulation Rule-based (Complete HE-VESIM ) Controller Fuel Economy, Vehicle Response

COE Control Seminar-11/01-2002- 32

16 arc Improved Rule-Based Control

Ê Braking rule

Ê Power split rule

Ê Recharge rule

Ê Transmission shift strategy

COE Control Seminar-11/01-2002- 33

arc Improved Transmission Shift

The original optimal gear trajectory has frequent shifting which is undesirable.

N −1 JfuelkNOxkPMk=+⋅+⋅∑()() 40 () 800 ()++−β ⋅ gxx(1)kg(k) k =0 62 +⋅ 510 ⋅ (SOC ( N) − SOC f )

5 4 β=0 3

Gear 2 1 0 0 100 200 300 400 500 600 700 800 900 1000 5 4 β=0.5 3

Gear 2 1 0 0 100 200 300 400 500 600 700 800 900 1000 5 4 β=1.5 3

Gear 2 1 0 0 100 200 300 400 500 600 700 800 900 1000 Time (sec) COE Control Seminar-11/01-2002- 34

17 arc Improved Transmission Shift

100 1st gear 90 2nd gear 3rd gear 80 4th gear

30 Engine Power Demand(kW) 20

0 0 20 40 60 80 100 120 140 160 180 200 220 Transmission Speed (rad/s)

COE Control Seminar-11/01-2002- 35

arc Improved Power Split Rule PSR = 0 (Motor-only mode) P PSR = eng 0<1 (Recharging mode) 4

3.5 PSR > 1 Approximated optimal PSR curve 3 Optimal operating points P = PSR× P 2.5 eng dem

Pmot = Pdem − Peng 2 PSR = 1

1.5 Power Split Ratio (PSR) Ratio Split Power PSR < 1

0.5 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Power Demand / Trans Speed (kN-m)

COE Control Seminar-11/01-2002- 36

18 Improved Recharge Rule arc

0.77 Rule_d4

PSR = (Eng Pwr) / (Eng Pwr+Mot Pwr) 4 0.72 Rule_d3 3.5

PSR = (Eng Pwr) / (Eng Pwr+Mot Pwr) 2.5 4.5 0.67 Rule_d2 PSR 4 2

3.5 1.5

3 1 0.62 2.5 Rule_d1 0.5 PSR 0 20 40 60 80 100 120 140 2 Power Demand

1.5

1 0.57 Rule_0

0.5 0 20 40 60 80 100 120 140 PSR = (Eng Pwr) / (Eng Pwr+Mot Pwr) Power Demand 5 Rule_c1 0.52 4.5 4

3.5

3 Rule_c2

0.47 PSR 2.5

1.5

Rule_c3 1

0.42 0.5 0 20 40 60 80 100 120 140 Power Demand Rule_c4 0.37

COE Control Seminar-11/01-2002- 37

arc Outline Ê Introduction and Motivation

Ê Simulation model and preliminary rule-based control results

Ê Dynamic programming technique

Ê Improved rule-based control law

Ê Simulation results

Ê Dyno test results

Ê Conclusions

COE Control Seminar-11/01-2002- 38

19 arc UDDSHDV Cycle Results

FE NOx PM Performance (mi/gal) (g/mi) (g/mi) Measure *

Baseline Rule- 13.159 5.7395 0.4576 840.63 Based New Rule-Based 12.8738 4.8355 0.4292 787.0965 DP (FE & Emis) 13.237 4.6422 0.3992 739.56

Performance Measure: fuel+ 40⋅+⋅ NOx 800 PM

Cycle beating?

COE Control Seminar-11/01-2002- 39

arc NOx PM Weighted Cost mpg (g/mi) (g/mi) (Fuel+40*NOx+600*PM) Pre Rule- 16.18 3.87 0.332 621.2 Based WVUCITY New Rule- 15.36 2.41 0.219 480.7 (-22.6%) (RDP 1) Based DP 16.63 2.04 0.161 403.6 (-35.0%) Pre Rule- 15.31 4.43 0.355 671.225 Based WVUSUB New Rule- 14.58 2.927 0.296 574.63 (-14.4%) (RDP 2) Based DP 15.4 2.78 0.259 526.7 (-21.5%) Pre Rule- 12.84 7.28 0.509 948.83 Based WVUINTER New Rule- 12.72 6.27 0.488 894 (-5.8%) (RDP 3) Based DP 12.97 6.17 0.44 847.7 (-10.7%) Pre Rule- 13.159 5.74 0.4576 840.63 Based UDDSHDV New Rule- 12.874 4.83 0.429 787.1 (-6.4%) Based DP 13.237 4.64 0.399 739.56 (-12.0%)

COE Control Seminar-11/01-2002- 40

20 arc Outline Ê Introduction and Motivation

Ê Simulation model and preliminary rule-based control results

Ê Dynamic programming technique

Ê Improved rule-based control law

Ê Simulation results

Ê Dyno test results (Fuel Economy Only)

Ê Conclusions

COE Control Seminar-11/01-2002- 41

arc FedEx HEV Truck Proposal

Ê FedEx and the Alliance for Environmental Innovation proposed to develop an environmentally progressive truck that would reduce emission by 90% and improve fuel efficiency by 50%, while work as well as FedEx’s current (1999 W700) white delivery trucks and cost about the same over the vehicle’s lifetime.

Ê Winning design(s) that meet targets will get FedEx purchase (10-50 pre-production by mid 2003, production vehicles by 2004).

COE Control Seminar-11/01-2002- 42

21 Current FedEx Pickup/Delivery Vehicle arc Specification

• Year 1999 baseline vehicle (W700 series) • GVWR: 16,000 lbs • Cargo capacity: approximately 670 cubic feet, 6000 lbs • Cummins 138 kW 6-cylinder diesel engine • Allison 4-speed automatic transmission (AT)

(http://www.environmentaldefense.org/alliance)

COE Control Seminar-11/01-2002- 43

arc Eaton Prototype Hybrid Electric Truck • Same chassis and body • 44 kW AC motor • 125 kW 4-cylinder diesel engine • Lithium-Ion battery • 6-speed automated manual •PM trap transmission (AMT) Supervisory Controller

Engine Mot/Battery Trans controller controller controller

EngineEngine ClutchClutch MotorMotor TransmissionTransmission VehicleVehicle

PMPM Trap Trap BatteryBattery

COE Control Seminar-11/01-2002- 44

22 Test of Prototype Vehicle arc

• Modified FUDS cycle • Tested at Southwest Research Institute (SwRI) • Chassis dynamometer testing • EPA standard procedure to determine fuel economy and exhaust emissions

COE Control Seminar-11/01-2002- 45

arc Fuel Economy and Emission Results

• Hybrid Prototype #1: baseline prototype hybrid truck • Hybrid Prototype #2: redesign control strategy (Our Dynamic Programming Design Technique) and add PM trap

COE Control Seminar-11/01-2002- 46

23 arc Fuel Saving

Ê Rough Calculation

50000 miles/yr/truck 50000 trucks in the US Original: 10miles/gallon

15,000,000 gallons/yr saved by Control algorithm improvement

COE Control Seminar-11/01-2002- 47

arc

http://www.sierralegal.org/reports/climtorfinal.pdf

COE Control Seminar-11/01-2002- 48

24 arc Conclusions ÊDesigning the power management strategy for HEV by learning from the Dynamic Programming (DP) results has the clear advantage of being near-optimal, accommodates multiple objectives, and systematic.

ÊImproved rule-based control strategy can be developed by analyzing the DP results Î Significant reduction in NOx and PM emissions can be achieved at the price of a small increase in fuel consumption

ÊThe new control strategy was found to be robust on different cycles

COE Control Seminar-11/01-2002- 49