Energy Management Of A Series Parallel Hybrid Electric Vehicle UsingFuzzy Controller

Reza Ghorbani, Eric Bibeau, Paul Zanetel and Athanassios KarlisDepartment of Mechanical and Manufacturing EngineeringUniversity of Manitoba, Winnipeg, MB, Canada R3T 5V6

Email: umghorba@cc.umanitoba.ca, ericbibeau@umanitoba.capaul zanetel@umanitoba.ca, akarlis@ee.duth.gr

Abstract— A series parallel hybrid electric vehicle (HEV)with a conventional internal combustion engine (ICE) and apower split device have been designed using REVS (RenewableEnergy Vehicle Simulator) simulation package. REVS has beendeveloped at University of Manitoba to assist in detail studiesof different configurations of hybrid electric vehicles (HEV)through visual programming by creating components as hierar-chical subsystems. REVS was developed in the Matlab/Simulinkgraphical simulation language as well as IDEAS package andis portable to most computer platforms. A Fuzzy Controllerhas been developed for energy management strategy of aseries parallel HEV. The output power of ICE is directlycommanded by driver’s acceleration pedal. The electric motorpower assignment is implemented using the fuzzy controllerunder a Rule-base frame. The rules are defined based on thedriver’s acceleration command and the status of the SOC (stateof charge) of the battery. The rules ensure that the electricmotor and the battery operate at high efficiency wheneverpossible. Simulation results of the energy management systemand dynamic responses of the system are discussed for proposedvehicle.

I. I NTRODUCTION

Transportation is almost exclusively based on the use ofnon-renewable fossil fuels. Electricity use for transportationhas limited applications because of battery storage rangeissues, although many recent successful demonstrations ofelectrical vehicles have been achieved. Renewable biofuelssuch as biodiesel and bioethanol are only a small percentageof the overall energy sources for mobility. With productionof oil predicted to decline, the number of transportationvehicles continuing to increase globally, and the realizationthat we live in a carbon constrained world, a transformationof the transportation sector is inevitable. Next generationsof transportation vehicles will not rely exclusively on theuse of fossil fuels burnt in an internal combustion engine.Furthermore, the hydrogen fuel cell proposition is not asattractive as first believed, as no gain is possible when thehydrogen is derived from electricity or fossil fuels. It istherefore important to have the ability to simulate differenttransportation vehicles configurations to enable the optimiza-tion of available renewable energy resources and minimizegreenhouse gases. Our ability to predict and investigatevarious transportation pathways can contribute effectively todecrease our reliance on fossil fuels.Hybrid vehicles offer the promise of higher energy efficiencyand reduced emissions when compared with conventional

automobiles, but they can also be designed to overcome therange limitations inherent in a purely electric automobilebyutilizing two distinct energy sources for propulsion. Withhybrid vehicles, energy is stored as a petroleum fuel and inan electrical storage device, such as a battery pack, and isconverted to mechanical energy by an internal combustionengine (ICE) and electric motor, respectively. The electricmotor is used to improve energy efficiency and vehicleemissions while the ICE provides extended range capability.Though many different arrangements of power sources andconverters are possible in a hybrid power plant, the twogenerally accepted classifications are series and parallel[1].Nowadays, researchers focus on understanding the dynamicsof the hybrid vehicles by developing the simulators [2], [3].The results can be used to optimize the design cycle of hybridvehicles by testing configurations and energy managementstrategies before prototype construction begins. Power flowmanagement, optimization of the fuel economy and reducingthe emissions using intelligent control systems are part ofthe current research [4]–[17]. Practical and experimentalverification of the vehicle simulators is an important partof ongoing researches [18]–[20].The University of Manitoba, Canada, in cooperation withDemocritus University of Thrace, Greece, are developing aRenewable Energy Vehicle Simulator (REVS) that enablesto simulate renewable energy vehicles using combination ofpropulsion system and fuels by adapting library modules tosuit particular applications. The simulation software predictsthe energy use of the vehicle, taking into account the dutycycle and driver habits. Library modules have been developedto simulate the PHEV architecture as this platform offersenergy scenario for cars and buses allowing combinationof energy sources that include renewable electricity andrenewable biofuels. REVS was developed to address the nextgenerations of vehicles which will not rely exclusively on theuse of fossil fuels burnt in an internal combustion engine.The modeling of the internal combustion engines is carriedout by IDEAS. The modeling of the transmission system,dynamics of the vehicle, electrical motor and power driversis done using Matlab and Simulink. REVS simulates differentvehicles configurations to enable the optimization of avail-able renewable energy resources and minimize greenhousegases. This paper discusses the methodology for designingsystem level vehicles using the REVS package. A series

Fig. 1. General schematic of the series parallel HEV.

parallel HEV with a conventional ICE and power splittransmission system have been designed using the simu-lation package. A fuzzy controller has been developed tosimulate the driver and to command the acceleration andbrake pedal. Another fuzzy controller has been employedto manage the power flow in hybrid vehicle. The energymanagement strategy has been applied through a rule-basedfuzzy approach. Several rules are used to determine, basedon the power demand value, how much power to get fromelectric motor if the power demand is positive. Regenerativebraking is activated inherently by power split device in seriesparallel HEV configuration. Other rules are used to cause thebattery to operate within an efficient range of state of chargesand to ensure that some limits are not exceeded. These limitsrepresent the maximum allowable charging and dischargingpowers of the battery, in addition to the maximum allowableregenerated power. Finally, the simulation results of thevehicle are presented for high and low capacity battery invehicle acceleration situation as well as for an urban drivingschedule.

II. REVS: MODULES AND DESIGN METHODOLOGY

REVS has been developed in Matlab/Simulink environ-ment as well as IDEAS [21], [22]. Mainly, the modules andcomponents related to drive trains, dynamics modeling andcontrol are developed in Simulink and the fluid and heattransfer systems are carried out by IDEAS. The commu-nication module transfers the data between Simulink andIDEAS at each time step. A user can select the componentsof the vehicle from the libraries and create a specific vehicleconfiguration. The vehicle can be constructed graphically by

connecting the main component blocks (environment, drivecycle, controller, engine, motor/generator, transmission, bat-teries, vehicle dynamics and renewable energy resources) us-ing the Simulink visual programming methodology throughthe connection of the appropriate input and output ports. Onthe other hand, user can set the heat/fluid system components(engine chemical reactions, fuel, solar, fuel cells) usingtheIDEAS. Energy flow and electrical signals are the mainelements transferred between library modules. REVS imple-ments three kinds of controls: direct control, vehicle systemlevel control and component level control. Direct controlgoverns the flow of information from block to block in themodel. One block can control another block through outputconnectors; the same block can be controlled by anotherblock through input connectors. Signal and energy flow fromblock to block in the model create a direct control network.Results such as engine, motor and vehicle speeds, torque,power and emissions are displayed using the graphicalplotting tools that can consider transient responses. Vehiclecharacteristics such as size and weight, gear ratios, drag andfriction coefficients, inertias and the environmental situationscan be changed in an excel worksheet file to specify thedrive train. A controller block is designed with conventionaland fuzzy logic controller blocks which create the signalsrequired to control the individual system-level components.REVS has been designed to be flexible in adding on ofmore Matlab/Simulink toolboxes for optimization purposesand virtual reality interfaces.

Fig. 2. Model of series parallel HEV in REVS.

III. D ESIGN OF THESERIESPARALLEL HYBRID

ELECTRIC VEHICLE

In this section, the design and analysis of the model ofToyota Prius as a series parallel HEV drive train usingthe REVS is discussed. The Prius’ components such asICE, motor, battery, and vehicle dynamics models weredefined based on vehicle’s available information. The modelof the Power Split Device (PSD) and battery is explainedin paper by Liu et al. [23]. A description is given of theperformance specifications, the control strategies and powerplant developed for the vehicle design. A fuzzy controlleris designed to manage the output power of the electricmotor based on accelerating pedal and State of the Charge(SOC) of the battery. Another fuzzy controller parallel witha first order system has been employed to model the driverresponse to the vehicle velocity error. Simulation studiesareperformed for Prius using two different vehicle velocity drivecycle. Various performance parameters of the vehicle, such

as vehicle velocity, SOC and generated power by ICE andEM, during the simulation studies are graphically presentedin this paper.

A. Drive Train Design

In a typical series parallel drive train design, consistingof an ICE, an electric motor, a generator and a powersplit device (PSD), either the ICE or the electric motor canbe considered the primary energy source depending on thevehicle design and energy management strategy. The PSDdivides the output torque of the ICE, with a fixed torque ratio,into the wheels and generator. The output power of the ICEcan be divided into an infinite ratio between the wheels andgenerator. This configuration is designed so that the ICE andelectric motor are both responsible for propulsion or each isthe prime mover at a certain time in the drive cycle. Alsopart of the power of the ICE transfers to the wheels whilethe other is used to recharge an energy accumulator, usuallya battery pack. The general schematic of the series parallel

configuration in REVS is shown in Figure 1.Series parallel HEV consists of different elements withvarious configurations that make the vehicle modeling morecomplex by providing different number of choices and theireffect on a vehicles performance for a special mission.The modeling of Prius drive train is shown in detailsin Fig. 2. The ICE model was designed based on Priustorque/power/velocity data and threshold using lookup table.The permanent magnet asynchronous AC motor of the Priusmodel is also modeled based on available data using lookuptable by considering the motor power threshold. Capacityand number of cells of the battery assumed as an initialinput parameter in the simulation. State of the charge ofthe battery and current load determine the DC bus voltagebased on battery model [23]. Regenerative braking is inher-ently performed through PSD and generator whenever thedecreasing velocity of the vehicle is demanded by driver. Afuzzy controller manipulates the power contributions of theelectric motor that is explained in detain in the followingsection.

Fig. 3. Schematic of the fuzzy logic power controller with membershipfunctions.

IV. ENERGY MANAGEMENT

The energy management strategy to control the powerflow of the vehicle is described in this section. Followingcriteria have should be considered in developing the energymanagement block: 1- The driver inputs (from brake andaccelerating pedals) is consistent with conventional vehicle(driving the series parallel HEV should not ”feel” differentfrom driving a conventional vehicle). 2- The state of chargeof the battery is sufficient at all times. The power controllerdetermines the power needed to drive the wheels and chargethe batteries. It also commands the power required fromelectric motor. The batteries can be charged at the sametime of power assigned to the electric motor. The ICE canprovide the power for both charging the batteries and drivingthe wheels using PSD. The next section discusses the power

Fig. 4. 3D graph of the fuzzy controller rules for power controller; SOCand Acceleration pedal are inputs and Scaling factor is output.

controller that implements the energy management strategyand uses fuzzy logic to compute the power flow.

A. Power Controller

Figure 2 presents a simplified block diagram of the ve-hicle model. As shown in Fig. 2, a fuzzy logic controllerdetermines the output power of the EM with regard to theinputs of accelerator pedal and the SOC of the battery. Theacceleration pedal signal is normalized to a value betweenzero and one (zero: pedal is not pressed, one: pedal fullypressed). The normalized braking pedal signal is directlyconnected to the vehicle dynamic block to subtract brakingforces from wheel forces. The output of the power controllerblock is the scaling factor of the EM power which is normal-ized between zero and one. The scaling factor is multipliedby the maximum available power of the EM in EM block.On the other hand, the normalized value of the acceleration

Fig. 5. Schematic of the fuzzy logic used to model the driver withmembership functions.

Fig. 6. Graph of the driver fuzzy controller rules; Velocityerror is outputand Acceleration pedal is input.

pedal multiplies by maximum available power of the ICE.Finally, the total power of the vehicle is the ICE+EM power.By this way, the driver can command the complete range ofavailable power at all times. The maximum available EMand ICE power depends on their speed and temperature,and is computed using a 2D look-up table with speed andtemperature as inputs of the EM and ICE blocks.The EM scaling factor computed through fuzzy logic con-troller is close to zero when the SOC of the battery is too low.In that case the EM is not used to drive the wheels, in order toprevent battery damage. When the SOC is high enough, thescaling factor equals one. User can change the membershipfunctions of the input and output signals. With respect tothe SOC limitations, the scaling factor is proportional to theacceleration pedal. The membership functions of the powercontrol unit are illustrated in Fig. 3. To illustrate the fuzzylogic rules, Fig. 4 shows the scaling factor as a function ofacceleration pedal and SOC. As it is shown in Fig. 4, thescaling factor is zero when the SOC is below 0.8.

B. Velocity Tracking Controller

A combination of a low pass filter and a fuzzy controller isassumed to model the driver for tracking the desired velocity.The vehicle velocity error is assumed as the input of the lowpass filter of the form 0.1

0.05S+1. On the other hand, a fuzzy

controller, parallel with the first order system, commandsthe acceleration pedal. The membership functions of thedriver fuzzy controller have been shown in Fig. 5. When thevelocity of the vehicle is lower than desired one, the driverfuzzy controller sets a positive value for acceleration pedaland the more velocity error the more acceleration pedal. Thefunctionality of the rules of the driver fuzzy controller hasbeen shown in Fig. 6.

V. SIMULATION RESULTS

The vehicle has been simulated with REVS by velocitycommands. The parameters of the vehicle are listed in Table1. Two different set of desired vehicle velocity applied onthe model to examine the response of the system. The first

0 2 4 6 8 10 12 14 16 18 200

50

100

Veh

icle

spe

ed [k

ph]

0 2 4 6 8 10 12 14 16 18 20

0.70.80.9

SO

C

0 2 4 6 8 10 12 14 16 18 200

0.5

1

Sca

ling

fact

or

0 2 4 6 8 10 12 14 16 18 200

0.5

1

Acc

eler

atio

n pe

dal

0 2 4 6 8 10 12 14 16 18 20−20

0

20

EM

Pow

er0 2 4 6 8 10 12 14 16 18 20

−50

0

50

ICE

pow

er

0 2 4 6 8 10 12 14 16 18 200

20

Gen

erat

or p

ower

Time [sec]

Fig. 7. Cas1: Results of the simulation for an acceleration where batterycapacity =2[KWh].

and second cases examine the vehicle on acceleration whilethe battery capacity equals to2[KWh] and 25[KWh] thatresults are shown in Fig. 7 and 8 respectively. The third caseis a random desired vehicle velocity that includes the sharpaccelerations and decelerations by high battery capacity of25[KWh]. The results of the third case are illustrated inFig. 9. By specified parameters of the system, the desiredvehicle velocity shown by solid line, actual vehicle velocityshown by dashed line, SOC, scaling factor, accelerationpedal, electric motor power in KW, ICE power in KW andgenerator power in KW are shown in Fig. 7–9. As shownin Fig. 7-8 for the first and second cases, the vehicle canreasonably track the desired velocity. It has also shown thatthe SOC is kept higher than 0.8. In addition, ICE providesthe main power to the vehicle during the cruise period,after 16 second, that is highly desirable. In first case, theacceleration period is longer than the second case causedby lower battery capacity and respectively lower amount ofelectric motor power has been used during first case. Theresults of the third case in Fig. 9 show that the battery wasoperated at a relatively high SOC (between 0.89 and themaximum 0.96) for the whole period of driving cycle. It alsodemonstrates that the controller is well defined to minimizethe velocity error. It can be seen that the SOC increasesduring deceleration period that is also one of the important

0 2 4 6 8 10 12 14 16 18 200

50

100V

ehic

le s

peed

[kph

]

0 2 4 6 8 10 12 14 16 18 200.85

0.9

0.95

SO

C

0 2 4 6 8 10 12 14 16 18 200

0.5

1

Sca

ling

fact

or

0 2 4 6 8 10 12 14 16 18 200

0.5

1

Acc

eler

atio

n pe

dal

0 2 4 6 8 10 12 14 16 18 20−100

0

100

EM

Pow

er

0 2 4 6 8 10 12 14 16 18 20−50

0

50

ICE

pow

er

0 2 4 6 8 10 12 14 16 18 200

10

20

Gen

erat

or p

ower

Time [sec]

Fig. 8. Cas2: Results of the simulation for an acceleration where batterycapacity =25[KWh].

Definition ValuesCurb weight 1600Kg

Max ICE power 57 kW 5,000 r.p.m.MAX ICE torque 115 Nm 4,200 r.p.m

EM power 50 kW 1200 - 1540 r.p.m.Maximum voltage 500 V

Maximum EM torque 400 N.m 0 - 1200 r.p.m.

TABLE I

PARAMETERS USED IN PAPER.

factors in hybrid electric vehicle designs to regenerate thepower. Fig. 9 shows that in cruise velocity, period of 240–300 sec, ICE is the major source of energy.

VI. CONCLUSIONS

A renewable energy vehicle simulator REVS, for mod-eling, simulation, and analysis of a drive train developedat University of Manitoba using Matlab/Simulink/IDEAShas been presented in this paper. The goal of REVS is tostudy issues related to plug in hybrid electric vehicle designsuch as dynamics, energy management, fuel economy. REVSprovides new libraries to model different vehicle configu-rations in addition to Matlab/Simulink standard toolboxes.The design of a series parallel electric vehicle, Toyota Prius,is presented. Fuzzy controller has been used to control thepower flow as well as to track the vehicle velocity. The

0 50 100 150 200 250 3000

50

100

Veh

icle

spe

ed [k

ph]

0 50 100 150 200 250 3000.8

0.9

1

SO

C

0 50 100 150 200 250 3000

0.5

1

Sca

ling

fact

or

0 50 100 150 200 250 3000

0.5

1

Acc

eler

atio

n pe

dal

0 50 100 150 200 250 300

0

20

40

EM

Pow

er0 50 100 150 200 250 300

−100

102030

ICE

pow

er

0 50 100 150 200 250 3000

5

10

15

Gen

erat

or p

ower

Time [sec]

Fig. 9. Results of the simulation for a desired vehicle velocity in citywhere battery capacity =25[KWh].

results of velocity tracking performance, SOC, accelerationpedal commanded by fuzzy controller and electric motorcommand are illustrated to show the flexibility of the REVSfor studying various issues related to electric and hybrid EVdesign.

VII. A CKNOWLEDGMENT

This project was supported by Mathematics of InformationTechnology and Complex systems (MITACS) and ManitobaHydro through the MITACS internship program.

REFERENCES

[1] A. Kalberlah, Electric hybrid drive systems for passenger cars andtaxis. SAE Tech. Rep. 910247, 1991.

[2] H. Ed Bargar, Jingbo Li, Douglas J. Goering, and Jonah H. Lee,Modeling and Verification of Hybrid Electric HMMWV Performance.IECON’03 vol. 1, pp. 939- 944.

[3] S. R. Cikanek, K. E. Bailey and B. K. Powell,Parallel Hybrid ElectricVehicle Dynamic Model and Powertrain Control.Proceedings of theAmerican Control Conference Albuquerque, New Mexico June 1997.

[4] L. Ippoliot, V. Loia and P. Siano,Extended Fuzzy C-Means andGenetic Algorithms to Optimize Power Flow Management in HybridElectric Vehicles.Fuzzy Optimization and Decision Making, 2, 359374, 2003.

[5] Sheng-Chung Tzeng , K. David Huang , Chia-Chang Chen,Optimiza-tion of the dual energy-integration mechanism in a parallel-type hybridvehicle.Applied Energy 80 (2005) 225245.

[6] Naim A. Kheir, Mutasim A. Salman, Niels J. Schouten,Emissionsand fuel economy trade-off for hybrid vehicles using fuzzy logic.Mathematics and Computers in Simulation 66 (2004) 155172.

[7] P. Bowles, H. Peng, X. Zhang,Energy Management in a ParallelHybrid Electric Vehicle With a Continuously Variable Transmission.Proceedings of the American Control Conference Chicago, IllinoisJune 2000.

[8] Jong-Seob Won, R. Langari and M. Ehsan,An Energy Managementand Charge Sustaining Strategy for a Parallel Hybrid Vehicle WithCVT. IEEE TRANSACTIONS ON CONTROL SYSTEMS TECH-NOLOGY, VOL. 13, NO. 2, MARCH 2005.

[9] Hyeoun-Dong Lee, Seung-Ki Sul, Han-Sang Cho, and Jang-Moo Lee,Advanced gear–shifting and clutching strategies for a parallel hybridvehicle.IEEE Industry Applications Magazine, November/December2000.

[10] B. Bates,On the road with a Ford HEV.IEEE Spectrum, pp. 2225,July 1995.

[11] M. Hayashida et al.,Study on series hybrid electric commuter-carconcept.SAE J. SP-1243, Paper 970197, Feb. 1997.

[12] A. Nikopoulos, H. Hong, and T. Krepec,Energy consumption studyfor a hybrid electric vehicle.SAE J. SP-1243, Paper 970198, Feb.1997.

[13] R. D. Senger,Validation of ADVISOR as a simulation tool for a serieshybrid electric vehicle using the Virginia Tech future car Lumina.Masters thesis, Virginia Tech Univ., Blacksburg, 1997.

[14] D. Hermance and S. Sasaki,Hybrid electric vehicles take the streets.IEEE Spectrum, pp. 4852, Nov. 1998.

[15] Y. Takehisa, S. Shoichi, and A. Tetsuya,Toyota hybrid system: Itsconcept and technologiesin Proc. FISITA World Automotive Conf.,Paris, France, Sept. 1998.

[16] P. A. Abthoff and J. S. Kramer,The Merecedes-Benz C-class serieshybrid. in Proc. Electric Vehicle Symp. 14, Orlando, FL, Dec. 1997.

[17] L. J. Oswald and G. D. Skellenger,The GM/DOE hybrid vehiclepropulsion systems program: A status report.in Proc. Electric VehicleSymp. 15, Orlando, FL, Dec. 1997.

[18] M. Alexander, et al.,A Mid-Sized Sedan Designed for High FuelEconomy and Low Emissions: The UC Davis Future Car.TechnicalReport, 1998.

[19] S. Sasaki,Toyota’s Newly Developed Hybrid Powertrain.in 10thInternational Symposium on Power Semiconductor Devices and ICs.1998. Kyoto, Japan.

[20] C. Bourne, P. Faithfull and C. Quigley,Implementing Control of aParallel Hybrid Vehicle.International Journal of Vehicle Design, 1996.17(5-6): p. 649-62.

[21] M. McGarry and M. Watson,Dynamic simulation: a maturing technol-ogy generating real benefits.Ideas Simulation, available at www.ideas-simulation.com.

[22] G. A. Pugh, , 2001.,Extending Simulation.’ Indiana University-PurdueUniversity.

[23] Jinming Liu Huei Peng Filipi, Z.Modeling and Control Analysisof Toyota Hybrid System.International Conference on AdvancedIntelligent Mechatronics. Proceedings, 2005 IEEE/ASME pp. 134-139.