Professional Documents
Culture Documents
Application of Genetic Algorithm For Optimization of Control Strategy in Parallel Hybrid Electric Vehicles
Application of Genetic Algorithm For Optimization of Control Strategy in Parallel Hybrid Electric Vehicles
Abstract
This paper describes the application of the genetic algorithm for the optimization of the control
parameters in parallel hybrid electric vehicles (HEV). The HEV control strategy is the algorithm
according to which energy is produced, used, and saved. Therefore, optimal management of the
energy components is a key element for the success of a HEV. In this study, based on an electric assist
control strategy (EACS), the tness function is dened so as to minimize the vehicle engine fuel
consumption (FC) and emissions. The driving performance requirements are then considered as
constraints. In addition, in order to reduce the number of the decision variables, a new approach is
used for the battery control parameters. Finally, the optimization process is performed over three
different driving cycles including ECE-EUDC, FTP and TEH-CAR. The results from the computer
simulation show the effectiveness of the approach and reduction in FC and emissions while ensuring
that the vehicle performance is not sacriced.
r 2006 The Franklin Institute. Published by Elsevier Ltd. All rights reserved.
Keywords: Genetic algorithm; Optimization; Hybrid electric vehicle; Fuel consumption; Emissions
1. Introduction
0016-0032/$30.00 r 2006 The Franklin Institute. Published by Elsevier Ltd. All rights reserved.
doi:10.1016/j.jfranklin.2006.02.015
ARTICLE IN PRESS
M. Montazeri-Gh et al. / Journal of the Franklin Institute 343 (2006) 420435 421
such as acceleration and starting gradeability [1]. Oversizing the engine moves the cruising
operating point away from the optimal operation [2]. Moreover, an engine cannot be
optimized for all the speed and load ranges under which it must operate.
Hybrid electron vehicle (HEV) is projected as one of the solutions to the worlds need for
cleaner and more fuel-efcient vehicles. HEVs encompass two energy converters to
generate the power required to drive the vehicle and to operate on-board accessories. Most
typically, the architecture of these vehicles includes an IC engine with an associated fuel
tank and an electric machine with its associated energy storage system (battery).
For a HEV to be as efcient as it is possible, proper management of its energy elements
is required. This task is performed by HEV control strategy [310]. In other words, the
control strategy is the algorithm according to which energy is produced, used, and saved.
Control strategy for HEV is usually aimed at several simultaneous objectives. The main
goal is the minimization of the vehicle fuel consumption (FC) while also attempting to
minimize engine emissions and maintaining driving performance. Due to the complex
nature of HEV, a control strategy based on the engineering intuition frequently fail to
achieve satisfactory overall system efciency, and therefore an optimization algorithm
must be used. An important characteristic of this optimization problem is that the
powertrain system characteristics are highly nonlinear and non-continuous that may have
a large number of local optimums. Thus, the gradient-based optimization methods may
not converge to a global solution, and therefore a non-gradient based evolutionary
optimization method is proposed.
In this paper, application of genetic algorithm (GA) is described for optimization of the
control strategy in parallel HEV. The objective of the optimization is dened to minimize
the engine FC and emissions within standard criteria. Vehicle performance requirements
are also dened as constraints. However, as GA is not directly applicable to constrained
optimization problems, the constraints are handled by using penalty functions. In
addition, based on an electric assist control strategy (EACS), a new approach is used for
battery control parameters in order to reduce the number of the decision variables. The
optimization process is then performed for three different driving cycles including ECE-
EUDC (European cycle), FTP (US unied cycle) and TEH-CAR. TEH-CAR is a driving
cycle that is developed based on the experimental data collected from the real trafc
condition. The simulation results are nally obtained to investigate the effectiveness of the
approach and the effect of driving cycle on the optimization of HEV control strategy.
2. HEV congurations
There are generally two accepted basic congurations for HEVs including series and
parallel. A dual or multi-mode type is also considered as a third type that combines the
features of both the series and parallel hybrids [11].
The series HEV conguration incorporates a fuel converter (IC Engine), a generator,
battery, and an electric motor as shown in Fig. 1. In this case, the fuel converter does not
drive the vehicle shaft directly. Instead, it converts the mechanical power into the electrical
energy using a generator. The electric energy is also saved in the energy storage system (i.e.
battery). In this conguration, the torque required to drive the vehicle is supplied by the
electric motor. However, in parallel HEV, both electric motor and IC engine may deliver
power to the vehicle wheels as shown in Fig. 2. The electric motor may also be used as a
generator to charge the battery by either the regenerative braking or absorbing the excess
ARTICLE IN PRESS
422 M. Montazeri-Gh et al. / Journal of the Franklin Institute 343 (2006) 420435
Fuel Engine
Fuel Engine
Transmission Wheels
Motor/
Battery
Generator
power from the engine when its output is greater than that required to drive the wheels.
One of the advantages of the parallel HEV over the series type is that the parallel HEV
requires a smaller engine and a smaller electric motor to provide the same performance.
This feature makes the parallel HEV more suitable for passenger cars where the series
conguration is usually used for heavy duty vehicles. In the combined series-parallel
hybrid, the conguration involves an additional mechanical link compared with the series
hybrid and also an additional generator compared with the parallel hybrid that make the
series-parallel HEV a relatively more complicated and costly version.
Table 1
Parameters of electric assist control strategy
Parameter Description
(VL). Furthermore, if the required torque is less than a cutoff torque (Toff Tmax) that is
referred to as off-torque fraction, the engine will also turn off.
Fig. 4 illustrates the case when the battery SOC is lower than its low limit (LSOC). In this
case, an additional torque (Tch) is required from the engine to charge the battery. This
additional charging torque is proportional to the difference between SOC and the average
of LSOC and HSOC. Engine charging torque is only requested when the engine is on. The
engine torque is prevented from being below a certain fraction of the maximum engine
ARTICLE IN PRESS
424 M. Montazeri-Gh et al. / Journal of the Franklin Institute 343 (2006) 420435
torque (Tmin Tmax) that is referred to as min-torque fraction. This strategy is designed
to prevent the engine from operating at an inefcient low torque condition.
The control strategy parameters highly affect the car FC, emissions, battery charge and
vehicle performance. Therefore, an optimization problem is formulated and solved in this
study to achieve the optimal control parameters.
Two different approaches may be considered for handling such optimization problems
[13,14] as follows:
1. weighted aggregation of all the targets into a single objective function to be optimized,
2. optimization of one of the targets by itself and imposition on the other targets of
appropriate constraints to be satised.
Low Low
FC HC Low
Low NO x
CO
where x is a solution to the problem including a vector of the control strategy para-
meters within the solution space X. X denes the lower and upper bounds of the
parameters. Moreover, J(x) is the objective function and each inequality hi(x)p0
represents one of the nonlinear constraints discussed above. Finally, ncon is the number
of constraints.
GA, rst formulated by Holland [16], is the probabilistic global search and optimization
method that mimics the metaphor of natural biological evolution. GA operates on a
population of individuals (potential solutions), each of which is an encoded string
(chromosome), containing the decision variables (genes).
The structure of a GA is composed by an iterative procedure through the following ve
main steps [17,18]:
1 X
ncon
F x ai Pi x, (3)
Jx i1
where F(x) is the tness function, J(x) is the objective function, Pi(x) is the penalty function
related to the ith constraint and ai is a positive constant value that determines the degree to
which the ith constraint is penalized, normally called penalty factor. These factors are
treated as constant here and their values are obtained by trial and error.
The denition and the number of the decision variables are critical for the optimization
process, especially from simulation run-time point of view. In this study, a new approach is
used for the denition of the battery control parameters in order to reduce the number of
the control parameters. In this approach, the EACS variables, described in Table 1, are
rstly considered as the decision variables. The battery control parameters are then
reduced by the denition of a target value, TSOC. The battery SOC target value is dened
using the battery specications as shown in Fig. 6. This target value is selected to be as
close as possible to the minimum charge resistance. Once the target value is determined,
both LSOC and HSOC are determined by only one variable, i.e. DSOC through the
optimization process as follows:
50
Discharge
45 Charge
40
Resistance (Ohms)
35
30
25
20
SOC SOC
15
10
0 0.2 0.4 0.6 0.8 1
SOC
Using this approach, the dimension of the solution space is reduced to ve. Therefore,
the following ve variables are coded in a chromosome using a binary coding scheme
consisting of the following genes:
x DSOC ; V L ; T ch ; T off ; T min . (5)
To start the algorithm, an initial population of individuals (chromosomes) is dened.
A tness value is associated with each individual, expressing the performance of the related
solution with respect to a xed objective function to be minimized.
Reproduction is the process of generating a new population from the current
population. Selection is the mechanism for selecting the individuals with high tness over
low-tted ones to produce the new individuals for the next population. The variant used
here is the roulette wheel method in which the probability to choose a certain individual is
proportional to its tness:
f p
Probpi is selected Pn i .
k1 f pk
Crossover is the method of merging the genetic information of two individuals (parents) to
produce the new individuals (children). In the simplest case, this process is realized by
cutting two chromosomes at a randomly chosen position and swapping the two tales, as is
visualized below:
Parents
Children
100 11110 Crossover
! 100 10010
101 10010 101 11110:
6. Driving cycles
Driving cycles are dened as the test cycles used to standardize the evaluation of
the vehicles fuel economy and emissions. Driving cycles are speedtime sequences
that represent the trafc conditions and driving behavior in a specic area. Driving pat-
terns may vary from city to city and from area to area. Therefore, the use of an available
driving cycle obtained for the certain cities or countries are not necessarily applicable for
other cities.
ARTICLE IN PRESS
428 M. Montazeri-Gh et al. / Journal of the Franklin Institute 343 (2006) 420435
100
(a)
50
0
0 500 1000 1500 2000 2500
150
(b)
speed (km/h)
100
50
0
0 500 1000 1500
100
(c)
50
0
0 500 1000 1500 2000
time (sec)
Table 2
Driving cycle characteristic parameters
In this optimization study, in order to evaluate the tness function, three driving cycles
have been used. The most important cycles are shown in Figs. 7a and b including the
currently used cycles in United States (FTP cycle) and European community (ECE-EUDC
cycle). A recent driving cycle is also developed for the city of Tehran based on the
experimental data collected from the real trafc conditions. This driving cycle is named
TEH-CAR and is shown in Fig. 7c [12].
A driving cycle consists of a mixture of driving modes including idle, cruise, acceleration
and deceleration. The maximum, minimum and average speeds are also considered as the
cycle characteristics. Table 2 compares the parameters of these driving cycles. Signicant
variations may be expected depending on the type of the driving cycle.
The advanced vehicle simulator (ADVISOR) [22] is used for simulation study.
ADVISOR employs a combined forward/backward facing approach for the vehicle
performance simulation. The simulation parameters as well as the vehicle components
model have been set for a parallel HEV with the following specications:
The size of the engine and electric motor and the number of battery modules are obtained
based on a previous study through a scaling technique using a GA approach [23].
Table 3
Step size and range of variation for design variables
The step size and the range of the variations for each decision variable are determined
based on the desired performance and characteristics of components, as illustrated in
Table 3.
In order to compromise between FC and emissions, all weights in objective function can
be set equal. This version of weights leads to a solution for the multi-objective optimization
that is called tradeoff solution in this study.
TEH-CAR
1.498
1.496
1.494
1.492
0 10 20 30 40
FTP
1.358
Value of objective
1.356
1.354
1.352
0 10 20 30 40
ECE-EUDC
1.71
1.705
1.7
1.695
0 10 20 30 40
Generation number
Table 4
Parameters obtained by GA for three cycles
Table 5
Fuel consumption and emissions for three cycles
2.5
Optimal Default
2
Value of objective
1.5
0.5
0
TEH-CAR FTP ECE-EUDC
Table 6
Reduction for TEH-CAR cycle
Table 7
Denition of weights for optimization
w1 w2 w3 w4
Tradeoff 1 1 1 1
FC-targeted 1 0 0 0
HC-targeted 0 1 0 0
CO-targeted 0 0 1 0
NOx targeted 0 0 0 1
Table 8
Optimal parameters for FC-targeted optimization
compares the objective values for two vehicles over the three driving cycles. In addition,
the reduction in the FC and emissions for the optimal HEV over Tehran driving cycle is
illustrated in Table 6.
Another feature of this optimization problem is the effect of weighting factors on the
optimal parameters. For instance, when the main objective is the minimization of the
vehicle FC, the weight of FC in Eq. (1) is set to 1 and the weights of emissions will be set to
zero. This case is called FC-targeted optimization problem. More versions of weights are
also dened as HC-targeted, CO-targeted and NOx-targeted. The denition of weights for
different versions is illustrated in Table 7.
In order to study the sensitivity of the optimal control parameters due to the weight
selection, the optimal parameters obtained over the Tehran driving cycle for the FC-
targeted and HC-targeted optimizations are compared with those of tradeoff solution in
Table 8. As it is clear, the optimal parameters obtained for FC-targeted and HC-targeted
optimizations are different from the optimal parameters obtained for tradeoff solution.
These results show that although LSOC and HSOC are not very sensitive to weight
denition, the other parameters are changed signicantly due to the weight selection and
objective denition.
ARTICLE IN PRESS
M. Montazeri-Gh et al. / Journal of the Franklin Institute 343 (2006) 420435 433
Table 9
Fuel consumption and emissions for different weights in objective function
FC HC CO NOx Obj.
1.2
1
Normalized value
0.8
0.6
0.4
0.2
0
FC HC CO NOx
Fig. 10. Normalized values of FC and emissions for optimization with different weights.
8. Conclusion
Application of genetic algorithm is described for the optimization of the control strategy
in parallel HEV. The optimization problem is formulated for an electric assist control
strategy (EACS) in order to meet the minimum fuel consumption and emissions while
maintaining the vehicle performance requirements. A novel approach is also employed for
battery control parameters in order to reduce the number of the EACS decision variables.
Finally, the optimization is performed over three driving cycles, and the effect of driving
pattern on the optimization of HEV control strategy is investigated. The optimization is
also performed for different objective functions. The results show the impact of the driving
pattern on the HEV optimal control parameters. Moreover, the simulation results reveal
that the optimal parameters are not identical for different objective functions, implying
that the minimization of the fuel consumption is not necessarily associated with the
optimization of emissions.
References
[1] T. Moore, Tools and strategies for hybrid electric drivesystem optimization, SAE Paper 961660, 1996.
[2] Y. Gao, K. Rahman, M. Ehsani, Parametric design of the drive train of an electrically peaking hybrid
(ELPH) vehicle, SAE Paper 970294, 1997.
[3] J. Van Mierlo, Views on hybrid drivetran power management, Proceedings of the 17th International
ElectricVehicle Symposium, CD-ROM, 2000.
[4] B.M. Baumann, et al., Mechatronic design and control of hybrid electric vehicles, IEEE/ASME Trans.
Mechatronics 5 (1) (2000) 5872.
[5] S.D. Farrall, R.P. Jones, Energy management in an automotive electric/heat engine hybrid powertrain
using fuzzy decision making, Proceedings of the 1993 International Symposium on Intelligent Control,
Chicago, IL.
[6] C. Kim, E. NamGoong, S. Lee, Fuel economy optimization for parallel hybrid vehicles with CVT, SAE
Paper No. 1999-01-1148.
[7] G. Paganelli, et al., A general formulation for the instantaneous control of the power split in charge-
sustaining hybrid electric vehicles, Proceedings of Fifth International Symposium on Advanced Vehicle
Control, Ann Arbor, MI, 2000.
[8] V.H. Johnson, K.B. Wipke, D.J. Rausen, HEV control strategy for real-time optimization of fuel economy
and emissions, Proceedings of the Future Car Congress, SAE Paper No. 2000-01-1543, April 2000.
[9] A. Brahma, et al., Dynamic optimization of mechanical electrical power ow in parallel hybrid electric
vehicles, Proceedings of the Fifth International Symposium on Advanced Vehicle Control, Ann Arbor, MI,
2000.
[10] U. Zoelch, D. Scroeder, Dynamic optimization method for design and rating of the components of a hybrid
vehicle, Int. J. Vehicle Des. 19 (1) (1998) 113.
[11] K.T. Chau, Y.S. Wong, Overview of power management in hybrid electric vehicles, Energy Convers.
Manage. 43 (2002) 19531968.
[12] M. Montazeri-Gh, M. Naghizadeh, Development of car drive cycle for simulation of emissions and fuel
economy, 15th European Simulation Symposium Proceedings, Delft, Netherlands, October 2003.
[13] Y. Sawaragi, H. Nakayama, T. Tanino, Theory of Multiobjective Optimization, Academic Press, Orlando,
FL, 1985.
[14] S. Martorell, S. Carlos, A. Sanchez, V. Serradell, Constrained optimization of test intervals using a steady-
state genetic algorithm, Reliab. Eng. Syst. Saf. 67 (2000) 215232.
[15] T.C. Moore, A.B. Lovins, Vehicle design strategies to meet and exceed PNGV goals, Electric and Hybrid
VehiclesImplementation of Technology SAE Special Publication SP-1105, 1995, pp. 79121.
[16] J.H. Holland, Adaptation in Natural and Articial System, University of Michigan Press, Ann Arbor, MI,
1975.
[17] C. Houck, J. Joines, M.G. Kay, A genetic algorithm for function optimization: a Matlab implementation,
NCSU-IE TR 9509, 1995.
ARTICLE IN PRESS
M. Montazeri-Gh et al. / Journal of the Franklin Institute 343 (2006) 420435 435
[18] K.F. Man, K.S. Tang, S. Kwong, Genetic algorithms: concepts and applications, IEEE Trans. Ind. Electron.
43 (5) (1996).
[19] J. Richardson, M. Palmer, G. Liepins, M. Hilliard, Some guidelines for genetic algorithms with penalty
functions, Proceedings of the Third International Conference on Genetic Algorithms, 1989.
[20] Z. Michalewicz, Genetic Algorithms+Data Structures Evolution Programs, AI Series, Springer,
New York, 1992.
[21] J. Joines, C. Houk, On the use of non-stationary penalty functions to solve constrained optimization
problems with genetic algorithms, IEEE International Symposium on Evolutionary Computation, Orlando,
FL, 1994, pp. 579584.
[22] T. Markel, A. Brooker, T. Hendricks, V. Johnson, K. Kelly, B. Kramer, M. OKeefe, S. Sprik, K. Wipke,
ADVISOR: a systems analysis tool for advanced vehicle modeling, J. Power Sources 110 (2002) 255266.
[23] M. Montazeri-Gh, A. Poursamad, Application of genetic algorithm for optimal design of hybrid electric
vehicles, Proceedings of the 13th International Conference of Iranian Society of Mechanical Engineers,
ISME2005, May 2005, Iran.