Content uploaded by Haider S Najem
Author content
All content in this area was uploaded by Haider S Najem on Oct 15, 2021
Content may be subject to copyright.
AIP Conference Proceedings 2404, 080004 (2021); https://doi.org/10.1063/5.0068888 2404, 080004
© 2021 Author(s).
Development of genetic algorithm control
technique for powersplit hybrid electric
vehicle with application on UDDS and
BCCDC driving cycles
Cite as: AIP Conference Proceedings 2404, 080004 (2021); https://doi.org/10.1063/5.0068888
Published Online: 11 October 2021
Haider S. Najem, Basil S. Munahi and Abdulbaki K. Ali
Development of Genetic Algorithm Control Technique for
PowerSplit Hybrid Electric Vehicle with Application on
UDDS and BCCDC Driving Cycles
Haider S. Najem 1, a), Basil S. Munahi 1) and Abdulbaki K. Ali 2)
1Mechanical Engineering Department, College of Engineering, University of Basra, Iraq
2Ministry of Oil, Iraq
a) Corresponding author: haidersuhail0@gmail.com
Abstract. This paper focuses on the development of efficient modern control technique for the power split hybrid electric
vehicle PSHEV. This technique depends on a genetic algorithm which is used to obtain the optimum torque management
between internal combustion engine ICE and electric motor EM. Also, a modified torque strategy approach based on a
driving cycle behaviour is developed. The aim of this study is achieved by development of a PSHEV model in
Matlab/Simulink. This model consists of ICE, DC generator and DC motor. All power components connected through a
power split unit. In addition, a modified configuration of continuously variable transmission CVT is presented which is
used to push the engine works near its optimal performance torque. Finally, the controller is tested over two different
driving cycles including urban dynamometer driving schedule UDDS and Basrah city canter driving cycle BCCDC. The
simulation results show that the effectiveness of control system to keep to obtained minimum fuel consumption, maintain
the battery state of charge SOC within the target limit, also control the ICE’s operating points within the optimum
efficiency maps and exploiting of DC generator machine for capturing maximum braking power. The proposed control
system must guarantee the best vehicle performance, especially in fuel economy.
Keywords: PowerSplit Hybrid Electric Vehicle, Torque Strategy, Genetic Algorithm, Power Split Unit, Matlab/Simulink.
INTRODUCTION
The fuel economy and clean environment are two goals aimed at by many vehicles manufactures all over the
world Many different types and design of the vehicle powertrain have been developed and tried to meet these goals,
the most recent is the hybrid engine or electric vehicle (HEV). In the (HEV) electric power is added to the
conventional powertrain path supporting the power of the internal combustion engine. This concept helps to improve
the fuel economy by engine rightsizing, load levelling, smaller heat loss and regenerative braking. The reduced
engine power is compensated by an electrical machine (or machines). Compared with internal combustion engines,
electric machines provide torque more quickly, especially at low vehicle speeds. Therefore, launching performance
can be improved even with reduced overall rated power. Load levelling can also be achieved by adding the electrical
path, which enables the engine to operate more efficiently, independent from the road load [1]. Regenerative braking
allows the electric machine to capture part of the vehicle kinetic energy and recharge the battery when the vehicle is
decelerating. [2]. HEVs can be classified into different types, depending on the degree of hybridization and
Powertrain architecture [3]. The power split hybrid electric vehicle structure is adopted in this paper concept owing
to its complexity and its need for an intelligent controller due to the content of different power sources of different
nature [4]. In the power split hybrid, the architecture combines the features of both the Series Hybrid and Parallel
Hybrid, as shown in Figure 1. This architecture needs more mechanical and electrical links, an additional electric
machine and a power split unit. This requires more progress in control and manufacturing techniques. Some modern
(HEVs) prefer this type of systems. The power split can be implemented by using a reduction gear, (CVT), an
advanced planetary gear and clutches, [5]. (HEV) is unique because it has two or more possible energy sources that
can be used to drive the vehicle. The main goal of an energy management strategy, in (HEV), is to improve the
quantity and types of energy used to achieve the requirements of driving. The control strategy determines the control
logic that governs how to control the internal combustion engine and the secondary energy source to provide energy
for propulsion, [8 & 9].
2nd International Conference on Engineering & Science
AIP Conf. Proc. 2404, 0800041–08000411; https://doi.org/10.1063/5.0068888
Published by AIP Publishing. 9780735441361/$30.00
0800041
FIGURE 1. Power Split Hybrid Electric Vehicle Structure, [6,7]
This paper adopts the genetic algorithm approach to develop a modern control technique for PSHEVs
configuration. The genetic algorithm is used to obtain the optimum torque distribution between ICE and EM for each
driving cycles are used. The remainder of this work is organized as follows. section 2 explains the structure of the
PSHEV model, the controller based on genetic algorithm is presented in section 3, the simulation results and
dissection are described in section 4 and finally, the conclusions are presented in section 5.
PSHEV CONFIGURATION MODELLING
Model Overview
All (PSHEV) components are modelled using the blocks in the simscape library and other Matlab/Simulink tools.
The (PSHEV) structure is very complex in modelling because it contains more than one electric machine and
complex power units. The (PSHEV) success depends heavily on the design of the powersplit units. In this model,
the work focusses on the powersplit unit to achieve maximum utilizing of the various power sources. The model has
been verified by testing its performance presented by the fuel consumption within the city (UDDS) and (BCCDC)
driving cycles, As shown in Figure 2, the vehicle parameters are used in this paper and listed in Table 1.
Model Description
Figure 2 represents the complete model that consists of several subsystems, and each subsystem represents a
certain component. The green line represents a mechanical link and the blue line represents an electrical link. All
Components of the (PSHEV) used will be introduced briefly as follows:
FIGURE 2. PowerSplit HEV Complete Model in Matlab/Simulink
0800042
TABLE 1. Parameters of PSHEV Model [10, 11 & 12]
Vehicle Body
This model is a body of the vehicle with two axles in longitudinal movements. The vehicle body modelling takes
into account, the vehicle body mass, road incline and aerodynamic drag. In this study, only the rear wheels are driven
by the transmission.
DC Generator and DC Motor Model
The (PSHEV) model contains one (DC) generator and one (DC) motor. These two models have used the
performance map data and efficiency of (Toyota Prius_JPN 15kW DC) machine [13]. This (DC) machine block
operates as an electric motor only if the torque demand signal is positive and works as a generator if the torque
demand is negative.
Battery and DCDC Voltage Converter
The purpose of a battery in the (HEV) that represents one of the power sources when the vehicle is at
acceleration. Also, the battery stores all the regenerative power produced from the generator when the vehicle is in a
Components Design Parameter Value
Vehicle
Vehicle mass,ܯ௩(Kg)
1350
The coefficient of rolling resistance, ܥ
0.012
Air Density, ߩ (kg/݉ଷ)
1.2
Frontal Area, ܣሺ݉ଶ)
2.686
Aerodynamic Drag Coefficient, ܥௗ
0.417
Rolling radius of a wheel, (m) ݎ
௪
0.32
Moment of inertia of each wheel, (kg.݉ଶ) ܫ௪
0.949
Engine
Engine Maximum torque, (N.m)
80.9
Moment of inertia of ICE, (kg.݉ଶ) ܫா
0.18
Generator Maximum torque, (N.m)
305
Moment of inertia of motor, (kg.݉ଶ) ܫெ
0.02
Generator Maximum torque, (N.m)
305
Moment of inertia of the generator, (kg.݉ଶ) ܫீ
0.05
Battery
Capacity, (amperehour)
6.5
Nominal Battery cell voltage, (V)
1.2
Total Battery Cells,
240
Transmission
Final Gear Ratio,ܰி
4
Transmission Efficiency,ߟ௧௦௦௦
0.9
CVT
Gear Ratio limitation ݃ሺݐሻ
80.42
0800043
deceleration state. In this study, the (SOC) is limited between (0 to 1) and the starting value used in this work is
equal to (70%). In addition, a bidirectional (DCDC) converter is used, because the battery side works with (288 V)
and the motor and generator side work with (500V). The bidirectional (DCDC) converter has a transformation
efficiency equals to (97 %).
Internal Combustion Engine
In this hybrid model, the performance map of the (ICE) depends on Toyota Prius vehicle (ICE). The (ICE)
delivers a torque to propel the vehicle through a (CVT) gear and power split. Otherwise, it returns to an idle state and
delivers the idle torque directly to the generator to charge the battery. The (ICE) idle state is taken as (7 N.m) and
(1000 rpm).
PowerSplit Unit
The purpose of the power split in this hybrid model is to divide and distribute the power between the main power
components of the vehicle and the vehicle wheels. During acceleration, the motor and (ICE) deliver the required
power to the vehicle wheels through the power split device. Further, during deceleration, power is delivered from the
vehicle wheels to the generator, also through the power split device. In addition, the (ICE) delivers power to the
generator through the power split device. The power split in this hybrid model includes two parts, controlled (CVT)
and power management as shown in Figure 3. In this Figure ports, 2 and 4 are unidirectional inputs, while port 3 is a
unidirectional output, and port 1 is a bidirectional input or output. The input means the power flows into the power
split and the output means the power flows out of the power split.
Controlled CVT
The purpose of a (CVT) gearbox, in this model, is to push the (ICE) to work near the optimum performance
points. The effective range of gear ratio in this (CVT) is between (0.42 to 8). The feature in this (CVT) is a
continuously variable gear ratio. In this hybrid model, the (ICE) output shaft is connected to the base of the (CVT)
and the power split is connected to the follower of the (CVT). The following two relations are used.
ൌ ሺሻǤ (1)
ൌሺሻǤ
(2)
Where (t) is dynamic time, ሺሻ is (CVT) gear ratio, is the base rotating speed (rpm), is the follower
rotating speed (rpm), is the base torque (N.m) and is the follower torque (N.m), as shown in Figure 4.
FIGURE 3. PowerSplit Unit Representation.
0800044
FIGURE 4. Controlled (CVT).
GENETIC ALGORITHM CONTROLLER
The purpose of this controller is to determine the optimum torque distribution between the ICE and EM. Also, a
new approach of torque management between is developed called torque strategy. A genetic algorithm is used to
obtain the optimal torque strategy. The torque strategy represents the amount of torque delivered from ICE and EM
at acceleration state. The main objective of the optimal torque strategy is to minimize fuel consumption and maintain
battery SOC with the target limits. The torque strategy for a specific driving cycle not applicable to another driving
cycle, because the method introduced highly depend on the driving cycle behaviour. The genetic algorithm control
system is shown in Figure 5 in this Figure;
1. ICE optimum torque ratio and segments time are coming from optimization process are implemented in a separate
Matlab file.
2. Propel torque demand is due to vehicle acceleration mode
FIGURE 5. Genetic Algorithm Subsystem
Explanation of the Torque Strategy
The presented torque strategy depends on dividing the driving cycle into piece segments and this division
depends on the total travel time of the driving cycle (ݐ݀), the number of segments for each driving cycle is equal to
100. The time between segment (i) is calculated from the equation (3). The time between segment different from one
driving cycle to another.
i =ሺ୲ౚሻ
ଵ ( 3)
Here, (ݐ݀): total travel time for driving cycle (sec), i: time between segments (sec). For example, the total travel
time of UDDS driving cycle equal to 1370 sec. The time between segments,
i =ଵଷ
ଵ ≈ 14 sec (4)
In each segment, during acceleration, the ICE and EM work at a constant torque ratio and the summation torque
ratios must equal to 100%. This is called substrategy (q) and for all driving cycles, it’s called torque strategy (str).
For example: during a segment, the substrategy of the ICE and EM equal to (70%) and (30%), respectively, and the
0800045
torque required to propel the vehicle equal is to (100 N.m). In this case, the ICE and EM deliver torque equal to
(70N.m) and (30 N.m) respectively. Figure 6 explains the scheme of torque with substrategies. In this Figure the
symbols A and B represented the ICE and EM torque ratio in each substrategy.
FIGURE 6. Explanation Scheme for Torque Strategy and SubStrategy
Optimization Formulation
To find the optimum torque strategy, a multiobjective cost function is needed. This function represents a fitness
function in the genetic algorithm and to be used in path planning as a decision criterion. This function could be
calculated using equation (5).
ൌ כ
େ כ
୧୬ୟ୪ െ
ୗେ כ
୧୬ୟ୪ (5)
Here, ்ܹ is the time weighting, ܶ is the total travel time of driving cycle (hour), ܹ
ி : is the fuel consumption
weighting, ܨܥி is the fuel consumption (liter). ܹ
ௌை : battery SOC weighting and ܱܵܥி : final battery SOC
ratio. The method used to obtain the optimal torque strategy for each driving cycle is as follows: The optimal torque
strategy for ICE and EM is found, where EM torque is equal to (1ICE Torque).
In the first step, a twenty random torque strategy for ICE is generated. Each strategy in a matrix form and the
matrix size (1*100). The PSHEV Simulation with each strategy. After the simulation, the fuel consumption and final
battery SOC are obtained and cost function for each strategy is calculated.
In the second step in the optimization process formulation, the genetic algorithm is used . The twentytorque
strategy represents the initial population p (0) for fitness function and twenty cost function represent the initial scores
for the fitness function. After running the program, the result represents the candidate torque strategy for ICE.
Finally, after finding the candidate torque strategy which does not directly represent the optimal. Simulate the
model by using this strategy and calculate the cost function. If the fuel consumption for the candidate strategy is
minimum and SOC above 0.5 compared with twenty torque strategy. In this case, the optimal torque strategy for a
specific driving cycle is represented as the optimal torque strategy. If fuel consumption not minimum, the process
from the beginning step must be repeated and must generate a new twenty torque strategy. As shown in Figure 7.
FIGURE 7. Flowchart of Optimization
0800046
Genetic Algorithm Constraints
After calculating the (Initial Engine Torque Demand) and (Initial Motor Torque Demand), some constraints of
the genetic algorithm controller work should be achieved, the purpose of these constraints is to maintain the (SOC)
above a certain value and the (ICE) does not supply any torque if the vehicle speed is lower than a certain value due
to the (ICE) inefficiency during the lower vehicle speed. The constraints can be described as:
1. If the vehicle speed is lower than (30 km/h) and battery (SOC) is above (0.5), the vehicle works in pure electric
mode, and only the (EM) is supplying the total propelling torque.
2. If the battery (SOC) is below (0.5), the (ICE) is supplying the total propelling torque, regardless of the vehicle
speed.
3. If vehicle speed is greater than (30km/h) and the battery (SOC) is above (0.5), the resulting distribution from the
GA controller works. The inputs and outputs of the truth Table are shown in Figure 8.
FIGURE 8. Constraints Genetic Algorithm Subsystem
The (Final Motor Torque Demand), is the final signal delivered from the controller directly to the electric motor
representing the torque demand from the electric motor. The Basic Engine Torque Demand is the signal that is not
delivered directly to the (ICE) but transformed to a specific controller in the (ICE). In other words, the (ICE) needs
an additional work in torque management, leading to the optimum utilizing of the (ICE).
ICE Controller Subsystem
The purpose of this part is to run the (ICE) near its optimal torque curve which is shown in Figure 9. This curve
represents the optimal operating points for the engine used in this study in the case of minimum fuel consumption,
[25]. The (ICE) cannot always work near these points, but under certain conditions, it could work near these points
like a high torque demand. With lower torque demand, the control strategy will find the best possible operation
criteria for the (ICE)
FIGURE 9. ICE Work with Minimum Fuel Consumption [14]
According to this curve, the (ICE) torque is closer to (54 N.m.) for a high range of rotating speeds, this feature
can be used in the controller design and can be described as follows:
1.If the signal (Basic Engine Torque Demand) is greater than (54 N.m.), the (ICE) is supplying only (54 N.m.) and
the remaining torque is supplied from the changed gear ratio of the (CVT).
2.If the signal (Basic Engine Torque Demand) is lower than (54 N.m.), the (ICE) supplies the torque required, and
the (CVT) gear ratio is equal to (1).
3.If the signal (Basic Engine Torque Demand) is equal to zero. The (ICE) returns to the idle and delivers torque to
the generator to charge the battery leading to an improvement in the battery (SOC).
0800047
SIMULATION RESULTS
In the previous section, how to find an optimal control strategy was discussed in the definition of the energy
management problem. This section will show the simulation results of implementing a genetic algorithm controller,
done UDDS and BCCDC driving cycles are used to test the effectiveness of the controller of the vehicle. Basic
characteristics of typical driving cycles patterns are shown in Table 2.
TABLE 2. Basic Characteristics of The UDDS Driving Cycle [15,16]
Driving
cycle
Distance
(km)
Time (sec)
Maximum
velocity (km/h)
Mean velocity
(km/h)
UDDS
12.0
1369
91.2
31.5
BCCDC
6.273
1041
63.62
21.632
UDDS Driving Cycle Results
The PSHEV performance results are based on optimum torque strategy obtained by genetic algorithm. The
results showed, The ICE delivered torque through acceleration mode with ratio equal to an optimum ratio in Figure
10 and the electric motor ratio is equal to (1ICE Torque Ratio).
FIGURE 10. Optimum Torque Strategy of ICE
Figure 11 shows the following vehicle speed to the desired value. As shown in the Figure, the vehicle follows the
desired speed pattern exactly. This proves the efficiency of the PSHEV model, optimization technique, and the
overall controller presented in this work.
FIGURE 11. UDDS Driving Cycle Speed Pattern
Figure 12 show engine torque variation with time in the whole driving cycle, as shown in the Figure, the
maximum engine torque is equal to (54 N.m) this value is equal to optimum torque obtained from the performance
maps. Also, the ICE idle state at every torque equal to (7 N.m). In this state, delivered torque to charge the battery.
0800048
FIGURE 12. Engine Torque Variations with Time
Figure 13 plot the CVT transmission ratio change with driving cycle time, the CVT ratio varies with time
depending on ICE torque request from the controller to maintain the ICE near the optimum performance curve.
During the driving cycle duration, the number of times the values of the CVT transmission ratio equal to one most
driving cycle time due to the controller requested torque form engine less than (54 N.m) and the CVT transmission is
a function of torque requested. The SOC is the most important variable to be monitored in the design of the HEV.
The variation of SOC with the driving cycle time is shown in Figure 14. As shown in Figure, due to limitations and
constrain of the controller the SOC doesn’t fall below the design limit 0.5. Figure 15 shows the generator torque
variations with time. These values represent the regenerative power acquired during the deceleration mode which is
used to charge the battery. Also, from this Figure shows at every generator torque equal to (7 N.m) that mean used
ICE only to charge the battery. Figure 16 shows the motor torque variations with time. In this Figure, the values
shown represent the torque delivered from the motor to propel the vehicle all the torque values within the range of
performance maps of DC motor used
FIGURE 13. CVT Ratio Variations with Time
FIGURE 14. Battery SOC Variation with Time
0800049
FIGURE 15. Generator Torque Variations with Time
FIGURE 16. Motor Torque Variations with Time
Fuel Consumption Results
The results obtained vary according to the driving mode that is chosen. Comparisons of the results obtained with
those of the (ADVISOR) program show an improvement and reliability of the controller presented in this work over
that of the (ADVISOR). The final fuel consumption and end battery (SOC) for different driving modes, based on
(UDDS) and (BCCDC), are shown in Table 3.
TABLE 3. UDDS and BCCDC Final Results
Parameters
Present
Work
ADVISOR
Program
Fuzzy Controller
Normal Mode
UDDS
BCCDC
UDDS
BCCDC
UDDS
BCCDC
FC (L/100
km)
3.343 3.349
5.6000
6.8 3.506
3.618
SOC end
value
0.5272
0.6032
0.5005
0.532
0.5293
0.6171
The results obtained vary according to the optimum torque strategy developed. The comparison of the results
obtained with those of the ADVISOR program and fuzzy controller shows improvement and reliability of the
genetic algorithm controller presented in this work over that of the ADVISOR software and fuzzy controller.
CONCLUSION
A new control system for a power split hybrid electric vehicle is studied and developed in this paper. The most
important conclusions that can be made from this work are listed as follows.
1. A developed GA optimization technique is used to obtain the optimum torque strategy for UDDS and BCCDC
driving cycles. The GA technique gave more economic and clean driving conditions.
2. A Complete PSHEV model depending on the developed configuration and control strategies are presented in this
work. This model is tested under UDDS and BCCDC driving cycles. The results of tested show minimum fuel
08000410
consumption over that of the ADVISOR program. This saving was achieved due to forcing the engine to work within
the target area.
3. A modified configuration of CVT is developed which is used to push the engine works near its optimal
performance torque.
4. The use of GA as a controller in PSHEV gives more reliability as compared to ADVISOR. This approved when
the ADVISOR controller needs more modification and changing to transform from one mode to another but GA
controller needs only changing the weighting values of the cost function.
REFERENCES
1.
J. Liu and H. Peng, “Modeling and Control of a PowerSplit Hybrid Vehicle,” IEEE Transactions on
Control Systems Technology.16, pp.12421251 (2008).
2.
J. Nellen, B. Wolters, L. Netz, S. Geulen, and E. Ábrahám, “A Genetic Algorithm based Control Strategy
for the Energy Management Problem in PHEVs,”
EPiC Series in Computer Science. 36, pp. 196–
214
(2015).
3.
M. MontazeriGh and M. Mahmoodik, “Development a new power management strategy for power split
hybrid electric vehicles,” Transp. Res. Part D Transp. Environ.37, pp. 79–96 (2015).
4.
Z. C. Johanyak, “A simple fuzzy logicbased power control for a series hybrid electric vehicle,” in
Proceedings
 EMS 2015: UKSim
AMSS 9th IEEE European Modelling Symposium on Computer
Modelling and Simulation, pp. 207–212 (2016).
5.
J. Hong, L. Zhao, Y. Lei, B. Gao, “Architecture optimization of hybrid electric vehicles with future high
efficiency engine,” Energies 11, (2018).
6.
M. F. M. Sabri, K. A. Danapalasingam, and M. F. Rahmat, “A review on hybrid electric vehicles
architecture and energy management strategies,” Renew. Sustain. Energy Rev.3, pp. 1433–1442 (2016).
7.
K. V. Singh, H. O. Bansal, and D. Singh, “A comprehensive review on hybrid electric vehicles:
architectures and components,” J. Mod. Transp. 27, pp. 77–107 (2019).
8.
T. H. Bradley and A. A. Frank “Design, demonstration and sustainability impact assessments for plugin
hybrid electric vehicles,” Renewable & Sustainable Energy Reviews.13, pp.115128 (2009).
9.
Z. Chen, N. Guo, J. Shen, B. Liu, R. Xiao, and W. Yan, “Comparisons of Energy Management Methods for
a Parallel Plug

In Hybrid Electric Vehicle between the Convex Optimization and Dynamic Programming,”
Appl. Sci.8, (2018).
10.
G. Sovran And D. Blaser, “A Contribution to Understanding Automotive Fuel Economy and Its Limits,”
SAE International, (2003).
11.
Ngo, V. Hofman, T. Steinbuch, M. and Serrarens, “A. Optimal Control of the Gearshift Command for
Hybrid Electric Vehicles,” IEEE Trans. Vehicular Technology.61, pp 35313543 (2012).
12.
P. Zhang, Y. Chen, M. Lin, and B. Ma, “Optimum matching of electric vehicle powertrain,” Energy
Procedia .88, pp. 894–900(2016).
13.
A. Poursamad and M. Montazeri, “Design of geneticfuzzy control strategy for parallel hybrid electric
vehicles,” Control Engineering Practice.16, pp 861–873(2008).
14.
Basil S. M., “Development of an Intelligent Controller for a Hybrid Engine Vehicle,” PhD Thesis,
University of Basrah, (2013).
15.
H. S. Najem, Q. A. Jawad, A. K. Ali, and B. S. Munahi, “Developing a General Methodology and Construct
a Driving Cycle with an Economic Performance Evaluation for Basrah City,” International Journal of
Engineering & Technology. 7, pp. 939– 944 (2018).
16.
Y. Zou, R. Huang, X. Wu1, B. Zhang, Q. Zhang, N. Wang and T. Qin, “Modeling and energy management
strategy research of a powersplit hybrid electric vehicle,” Advances in Mechanical Engineering.12, (2020).
08000411