Sunteți pe pagina 1din 8

Doruk Öztürk

Control strategies for Plug-in Hybrid Electric Vehicles

I. Introduction
In the past two decades hybrid electric drivetrains have been introduced to the
commercial vehicles in order to reduce the consumption of fossil fuels [1], and thus to reduce the
emission of greenhouse and other harmful gases [2]. Given the fact that the hybrid drivetrains
always consist of two or more different types of energy sources, there are many variants of drive
configurations, from simple, such as serial and parallel to more complex serial-parallel structures
[3]. Depending on the capacity and capabilities of charging the battery, the hybrid drive vehicle
can be divided into several categories [4], which are, hybrid electric vehicles (HEV), plug-in
hybrid electric vehicles (PHEV) and range extended electric vehicles (REEV). In HEV battery
capacity is generally lower and battery cannot be recharged from the grid, so, the operation of
these type of vehicles is mostly related to charge sustaining (CS) [5]. PHEV and REEV contain
battery with a larger capacity and the possibility of charging the battery from the grid. Therefore,
such vehicles can operate in either charge depletion (CD) and charge sustaining (CS) mode [5],
but also in combined mode in which the battery is discharged gradually, which further increases
the overall efficiency of the vehicle [6]. In this paper main focus will be put on the PHEV.
State of the art is such that PHEV drivetrain systems are generally based on a serial-
parallel configuration, consisting of electrochemical batteries, internal combustion engine, two
electric motors, one or more planetary gear having function of distribution of mechanical power,
and possibly couplings that determine mode of operation [7]. Such drivetrain concept is known
as electrical transmission with continuously-variable gear ratio (e-CVT) [8], given that, it can
provide optimal speed internal combustion engine independently of vehicle speed. The complex
structure of the hybrid electric drivetrain [1] raises many questions regarding the selection of the
drive structure, sizing of components and managing the flow of power between the components
of the drive [7]. In order to achieve the minimum fuel consumption of the PHEV it is necessary
to implement the optimization of control variables, with the selection of optimization methods
that guaranty the global optimality for a given problem while being computationally efficient [7].
The problem of optimization is to find the optimal time response of control variables,
such as engine speed and torque of individual machines, and to find the optimal drive mode

1
which minimizes the fuel consumption, i.e. maximizes the efficiency of vehicle drive. In general,
there exist two types of strategies for control of PHEV, namely, rule-based control strategy and
optimization based control strategy [9]. The rule based control is usually computationally
efficient, however such control strategy does not guarantee global optimal solution (i.e.
minimum fuel consumption). On the other hand, optimization-based controllers can guaranty the
desired global optimality, however they are not computationally efficient which result in control
system that is either slow or complex and expensive [9].
In this paper, a solution is proposed that can increase the computational efficiency of
optimization-based controllers and hence eliminate their biggest disadvantage. The proposed
solution is based on the learning and data driven controller that migrates the majority of
cumbersome computation from an on-board computer to cloud based virtual machine that can
use parallel computing to compute desired variables quickly and efficiently. The proposed
solution is relaying on the advances in connectivity solutions (e.g 5G) and cloud technology.

II. Trends

Currently there exist two control strategies for control of PHEV, namely rule-based
controllers and optimization-based controllers which are further divided into two subcategories.
The rule based controllers are divided into deterministic and fuzzy type controllers. Optimization
based controllers are divided into acasual and casual type controllers. The control strategy
classification tree is demonstrated in the Figure 1 [9].

Figure 1. – PHEV control strategy tree [9].


a. Rule-based control strategy

2
A rule-based control strategy has a goal to operate PHEV at the most efficient point. This
is usually done by utilizing the present electric machine and diesel engine at their highest
efficiency. This is achieved by utilizing the maximum all-electric range (electric machine run
most of the time) while taking advantage of regenerative breaking of the vehicle [9]. In order to
realize the rule based control strategy a predefined set of rules are determined based on the
desired outputs without any knowledge about the actual trip taken by the vehicle. The transition
between the vehicle modes (CS, CD, all-electric, all-diesel) depend on the predefined conditions
such as speed, acceleration, power requirement and battery state of charge. The transition and
operation rules are determined using deterministic rule-based methods and fuzzy logic rule-based
methods [9].
In deterministic rule-based method the controller operates on the set of rules that defined
before an actual drive happens. Logical state machine method for achieving such control was
proposed by authors in [10]. In this work, authors propose a hybrid control architecture
consisting of a logical state machine with corresponding dynamic control outputs for each state.
In this controller design, dynamic control is only incorporated into the vehicle system controller
to achieve smooth transition between states and dynamic control within each subsystem is
handled by each individual subsystem controller. The optimization on the other hand is handled
by the logic of the state machine [10]. In [11] a rule based strategy is proposed where the
maximum power is drawn from battery via motor to drive the vehicle and the rest of power is
provided by the diesel engine. Additionally the operating torque of the diesel engine is controlled
to provide battery charging and vehicle driving power as function of vehicle speed. The diesel
engine is turned on and off by predefined set of rules [11].
In fuzzy rule-based method the fuzzy logic controller is utilized to act on the predefined
set of fuzzy rules and optimize the efficiency of the vehicle for specific type of drive. Authors in
[12] demonstrate the fuzzy logic controller that has improved fuel economy when compared to
simple deterministic rule based controllers. Another type of fuzzy controller is presented in [13]
that can autonomously select operating point of the engines so as to have least impact on the fuel
economy.

3
b. Optimization-based control strategy

In order to cope with the stated disadvantages of the rule based controllers, optimization
based controllers have been proposed in literature. Optimization based controllers theoretically
have ability to provide global optimum operation point of PHEV. Optimization based controllers
are divided into two sub-categories, optimization controllers based on the historical data
(acasual) and optimization controllers based on the real-time data (casual) [9]. In either type,
system optimization is the result of learning and adaptation to the conditions within the given
rules and constraints. Optimization based controllers also have ability to do multi-objective
optimization and can achieve for example maximum range with minimum emission in the same
time [9]. Study of such controllers conducted in [14], shows that the optimal pathway is based on
the maximizing the charge depletion condition (CD). System optimized in CD mode rather than
CD-CS mode or electrical vehicle (EV) and CS mode will provide the best fuel economy [14].
This scenario is shown in Figure 2. where SOC denotes the state of charge of the battery [14].
In acausal optimization based controllers the optimization is done offline and then
controller implemented in PHEV. In this type of controller system depends on future values with
the possibility of depending on past and present variables. Most commonly used method in
acausal optimization based controllers is linear programming [15]. Particular controller in [15]
defines the cost function using two decision variables. Such controller is easy to implement but
has disadvantage of not being general enough to be applied to all types drive cycles. Authors in
[6] improve this controller by isolating control patterns to those that are drive cycle dependent
and to those that are not and improve the optimization performance in such way.

Figure 2. – SOC profile [14].

4
Authors in [16] use the genetic algorithm to find approximate solution to complex
nonlinear optimization and implement a cost function that minimizes fuel consumption. Some
other methods used for construction of acausal optimization based controllers are particle swarm
optimization [17], simulated annealing, DIRECT global optimization algorithm [18], two-scale
DP [19], gas-kinetic traffic flow model [20], and game theory [21].
On the other hand, causal optimization based controller are real-time optimization
controller that eliminate the aforementioned disadvantage of acausal controllers. These
controllers have ability to optimize the cost function in real time and are able to provide optimal
operation of PHEV. Various real time optimization control techniques were proposed in
literature. One example is a dual point closed loop form of control for PHEV that is able to find
optimum solution based on the historic data of drive cycle and traffic reports [22]. Such
stochastic dual point optimization is not real-time solution in its core, however it has been
modified for real time use [23]. Another example of causal controllers are adaptive controllers
that understand the average behavior of the driver and self-optimize in real time [12]. Artificial
neural networks (ANNs) are also used as causal controller with the aim of minimization of fuel
consumption regardless of driving pattern [9].

III. Limitations

Control methods described in the previous section have certain limitations that should be
mentioned. Deterministic and fuzzy rule based controllers have disadvantages of not being able
to adapt to real world case scenarios other than those predetermined at the time of vehicle
production. Although these type of controllers can provide relatively robust control algorithms,
they are being considered as inefficient because of variety of actual drive cycles that cannot be
predetermined. This makes the rule based controllers only locally optimal and they cannot
provide global optimality.
Acausal optimization based controllers can be computationally efficient but have
disadvantage of not being able to fully adapt to the real world drive cycle, hence have low degree
of optimality [9].
Among all of the analyzed controllers, artificial neural network (ANN) based controllers
have proved to be the best optimization method for PHEV control until now. These controllers
are designed such that they can learn from the experience for different drive cycles. These

5
controllers, ideally, also have ability to re-train themselves with the most recent drive cycle data.
However, ANN based controllers require enormous on-board computational power. Re-training
of an ANN controller for rather complex models could take up to days of time even for very fast
computers. Such computational power cannot be provided in the means of on-board computers
as it would result in high overall cost and design complexity of a PHEV.
A method for further increase of computational efficiency of ANN based causal
optimization controller is proposed in this paper that makes use of cloud and parallel computing.
In the next subsection further information about ANN and proposed method will be given.

IV. Solution

ANNs have many advantages over all other optimization based controllers [9]; The
controller is drive cycle and used independent; The more detailed the training data the more
robust controller will be; The training set can be updated as required. Success of ANN based
controller is based on the ability of updating of the neural network in the situation when the
driving cycle changes. This is also known as re-training of ANN. Such process can be done with
computational power that is not present in the on-board computer.
The proposed solution eliminates the need for large on-board computational power and
mitigates the computational burden to the cloud. A proposed controller can be best explained
through a following scenario. The controller is implemented using ANN and provides the
optimal solution for the given training data. During the operation of the PHEV, an operational
data is gathered over a significant period of time. In this scenario, a PHEV must also incorporate
a module that allows internet connectivity (Tesla S incorporates global internet connectivity).
The collected operational data would be then transferred to a cloud located virtual machine that
uses the data to retrain ANN in relatively short time due to the parallel computing ability.
Retrained model is then transferred back to the vehicle and installed as current model. In this
way the model could become very adaptive to the changes in driving cycles and user habits.
Such control strategy would ultimately result in fully customized PHEV control. It is important
to mention that implementation of such controller is more than possible with recent advances in
internet connectivity and cloud computing and such controller solves the drawbacks of ANN
based casual optimization based controller.

6
V. Conclusion

As seen throughout the literature review, the design of control strategy for PHEV can be
done in two ways. First method is rule-based and second one is optimization based. Rule based
controllers have advantage of being computationally efficient and simple but do not guaranty the
global optimal solution and cannot adopt to changing real world drive scenarios. On the other
hand the optimization based controllers, ANN based in particular, can provide global optimal
solution, however suffers from computational inefficiency and requires powerful on-board
computational system that is often too complex and expensive to be implemented in the car.
The proposed controller mitigates the disadvantages of ANN optimization based
controller by utilizing internet connectivity and cloud computing. Basically the proposed
controller is structurally the same as ANN based controller but, it additionally brings the
computation power of cloud in the loop. The operational data is only recorded by the on-board
computer and sent to cloud. All of the computation intensive calculations are done on the cloud
and updated controller is returned and installed to vehicle.

7
REFERENCES

1. A.E. Fuhs, Hybrid Vehicles and the Future of Personal Transportation. Taylor & Francis Group, LLC, ISBN-13:
978-1-4200-7534-2, 2009.
2. R.T. Doucette,M.D. McCulloch. Modeling the prospects of plug-in hybrid electric vehicles to reduce CO2
emissions. Applied Energy, vol.88, no. 2,pp.2315-2323, 2011.
3. Y. Gao, M. Ehsani, J.M. Miller. Hybrid Electric Vehicle: Overview and state of the Art. Proc. of IEEE Int. Symp. On
Industrial Electronics, pp. 307-315, Dubrovnik, Croatia, 2005.
4.
5. W. Liu. Introduction to Hybrid Vehicle System Modeling and Control. John Wiley & Sons, Inc., Hoboken, NY, USA,
2013
6. A. Rousseau, S. Pagerit, D. Gao. Plug-in hybrid electric vehicle control strategy parameter optimization. The 23rd
International Electric Vehicle Symposium (EVS23), Anaheim, CA, 2007.
7. M. Duoba, R. Carlson, J. Wu. Test procedures and benchmarking: Blended-type and EV-capable plug-in hybrid
electric vehicles. 23rd International Electric Vehicle Symposium (EVS23), Anaheim, CA, 2007.
8. L. Guzzella, A. Sciarretta. Vehicle Propulsion Systems - Introduction to Modeling and Optimization. 2nd ed., Berlin
Heidelberg: Springer-Verlag, 2007.
9. J.M. Miller. Hybrid Electric Vehicle Propulsion System Architectures of the e-CVT Type. IEEE Transaction on Power
Electronic, vol.21,no.3,pp.756-767, 2006.
10. S.G. Wirasingha, A. Emadi. Classification and review of control strategies for plug-in hybrid electric vehicles. IEEE
Transactions on vehicular technology. Vol.60, no. 1, pp.111-122,2011.
11. A. M. Phillips, M. Jankovic, and K. Bailey, Vehicle system controller design for a hybrid electric vehicle. Proc.
IEEE Int. Conf. Control Appl., Anchorage, AK, Sep. 2000, pp. 297–302.
12. H. Banvait, S. Sohel, and Y. Chen, A rule-based energy management strategy for plug-in hybrid electric vehicle
(PHEV). Proc. Amer. Control Conf., St. Louis, MO, Jun. 2009, pp. 3938–3943.
13. H. Lee, E. Koo, S. Sul, and J. Kim, Torque control strategy for a parallelhybrid vehicle using fuzzy logic. Conf. Rec.
IEEE IAS Annu. Meeting, St. Louis, MO, Oct. 1998, pp. 1715–1720.
14. V. H. Johnson, K. B. Wipke, and D. J. Rausen, HEV control strategy for realtime optimization off fuel economy and
emissions. Proc. Future Car Congr., Crystal City, VA, Apr. 2000.
15. M. Duoba, Evaluating PHEV technology using component HIL, subsystem, and chassis dynamometer testing:
Methods and results. Proc. SAE Hybrid Vehicle Technol. Symp., San Diego, CA, Feb. 2007.
16. S. Delpar, J. Lauber, T. M. Guerra, and J. Rimaux. Control of a parallel hybrid powertrain: Optimal control. IEEE
Trans. Veh. Technol., vol. 53, no. 3, pp. 872–881, 2004.
17. A. Piccolo, L. Ippolito, V. Galdi, and A. Vaccaro. Optimization of energy flow management in hybrid electric
vehicles via genetic algorithms. Proc. IEEE/ASME Int. Conf. Adv. Intell. Mechatronics, Corno, Italy, Jul. 2001, pp.
434–439.
18. H. Banvait, X. Lin, S. Sohel, and Y. Chen. Plug-in hybrid electric vehicle energy management system using particle
swarm optimization. Proc. EVS-24, Stavanger, Norway, May 2009.
19. D. R. Jones. DIRECT global optimization algorithm. in Encyclopedia of Optimization. Norwell, MA: Kluwer, 2001.
20. Q. Gong and Y. Li. Trip based power management of plug-in hybrid electric vehicle with two-scale dynamic
programming. Proc. IEEE Vehicle Power Propulsion Conf., Arlington, TX, Sep. 2007, pp. 12–19.
21. Q. Gong and Y. Li. Trip based optimal power management of plug-in hybrid electric vehicles using gas-kinetic
traffic flow model. Proc. Amer. Control Conf., Seattle, WA, Jun. 2008, pp. 3225–3230.
22. M. J. Gielniak and Z. J. Shen. Power management strategy based on game theory for fuel cell hybrid electric
vehicles. Proc. 60th IEEE Veh. Technol. Conf., Los Angeles, CA, Sep. 2004, pp. 4422–4426.
23. M. O’Keefe and T. Markel. Dynamic programming applied to investigate energy management strategies for a plug-
in HEV. Nat. Renew. Energy Lab., Golden, CO, Rep. NREL/CP-540-40376, 2006.
24. S. J. Moura, H. K. Fathy, D. S. Callaway, and J. L. Stein. A stochastic optimal control approach for power
management in plug-in hybrid electric vehicles. Proc. ASME Dyn. Syst. Control Conf., Ann Arbor, MI, Oct. 2008.

S-ar putea să vă placă și