Shengzhe 2017

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 4

Fuzzy Logic-Based Control Strategy for a

Battery/Supercapacitor Hybrid Energy Storage


System in Electric Vehicles
Zhou Shengzhe1, Wang Kai1*, Xue Wen1
(1. School of Automation and Electrical Engineering
Qingdao University 266071
Qingdao, China
* Email: 15863060145@163.com)

Abstract—In order to optimize the operation status of hybrid The most suitable system for the fuzzy logic strategy is the
energy storage system in electric vehicles, a novel fuzzy logic system with complex nonlinearity, and the robust real-time
control strategy is proposed. This strategy adopts Kalman performance brings it increasingly wide application. In ref.
filtering algorithm to estimates state of charge (SOC) and state of [11], the fuzzy logic algorithm was applied to control a
power (SOP), which can calculate the optimum power and supercapacitor/battery HESS with management methodology
alleviate the errors of SOC effectively. Besides, the framework of simultaneously. In ref. [12, 13], the fuzzy logic algorithm has
the strategy and the model of both battery and supercapacitor been adopted into the control strategies of HESS which were
are introduced. The proposed control strategy will help to deal composed of fuel cells, supercapacitor and battery. Compared
with energy storage system problem of relatively short lifespan
with other methods, the fuzzy control avoided the modeling of
and poor limited range in electric vehicles. So stability and
complex systems but the errors were much less. In ref. [14],
efficiency of hybrid energy storage system (HESS) improve
greatly
the fuzzy logic algorithm was taken into the control of a power
plant HESS and the algorithm was more simple than the others
Keywords—Hybrid energy storage system; Energy management with the same result. In this paper, a novel fuzzy logic control
system; Fuzzy Logic-Based Control; Kalman filtering algorithm; strategy is proposed. The framework of the strategy and the
Eectric vehicles model of both battery and supercapacitor are briefly
introduced. On the base of Kalman filtering algorithm, the
I. INTRODUCTION SOC and SOP are estimated. The frame of the fuzzy controller
and fuzzy rules of controller are discussed.
Electric vehicles (EVs) technology becomes increasingly
popular as a feasible solution to face the energy crisis and
II. THE FRAMEWORK OF PROPOSED CONTROL STRATEGY
reduce green-house gas emissions of vehicles. It can be seen
as a strong step toward sustainable transportation [1-3]. There are four main parts of the control strategy of EMS:
However, compared with traditional vehicles, EVs have not the modeling of supercapacitor and battery, the parameter
been widely accepted and the market share is negligible. The identification of two models, the state estimation and the fuzzy
main limiting factor is that the energy storage system has a logic control. The implementation flowchart is as followed:
relatively short lifespan and poor limited range. In EVs, the
main energy storage devices are supercapacitor and battery.
HESS composed of supercapacitor and battery has been
developed in recent years. In HESS, the battery can be used to
supply power at light loads, and the supercapacitor is
employed to cover the power demand of acceleration as well
as meet the regenerative braking energy requirements [4]. An
appropriate energy management system (EMS) can increase
the efficiency of HESS. EMS can not only conduct the
distribution of power in the energy storage devices, but also
extend the lifespan of HESS with the rational current control.
SOCc
Therefore EMS is the core of HESS. Many control strategies Preq

of EMS have been developed [5-7], and the fuzzy logic SOCb
algorithm has been one of the most popular methods to Pbat SOCb

conduct the control strategy because there is no need to SOP

establish the specific model for complex and nonlinear


systems [8-10].
Popt

‹,((( 
Fig. 1. The implementation flowchart of control strategy
­ χ 0,k −1 = ~xk −1 ; χ i ,k −1 = ~
xk −1 + Pk −1 ( N + λ ) i , i = 1,..., L
°
~
° χ i ,k −1 = xk −1 − Pk −1 ( N + λ ) i , i = L + 1,...,2 L (1)
III. THE MODELING OF THE BATTERY AND SUPERCAPACITOR ° (m)
°W0 = λ /( L + λ );W0( c ) = λ /( L + λ ) + (1 − α 2 + β )
Fig. 2 shows the compound model of supercapacitor. Its ® ( m)
°Wi = Wi = 1 /[ 2( L + λ )], i = 1,...,2 L
(c)
details are as follows. ° ~ ~ ~
° χ k −1 = [ xk −1 , xk −1 + Pk −1 ( N + λ ) , xk −1 − Pk −1 ( N + λ ) ]
I °λ = α 2 ( L + κ ) − L; α = 1e − 3; κ = 3 − n; β = 2
¯
Rs
Į is used to set the distances between sigma points and
average point, and it usually is a small positive number.
Cd
U
Substitute the set into the state equation to gain the
Rp function value point set:
Rf Cf χ kx|k −1 = F ( χ ki −1 ), i = 0,...,2 L (2)

Project the state ahead:


2L
~
xk− = ¦ Wi ( m ) χ ix, k |k −1 (3)
Fig. 2. The compound model of supercapacitor
i =0

Cf and Rf constitute the positive electrode model; Cd is the


negative electrode model; Rs is the equivalent series resistance; Project the error covariance ahead:
Rp is the equivalent parallel resistance. 2L
Px− = ¦Wi ( c ) ( χ ix,k |k −1 − ~
xk− )( χ ix,k |k −1 − ~
xk− )T (4)
Fig. 3 shows the second-order RC model of battery. Its i =0
details are as follows.
Rp Rl Compute the output estimate and its covariance:
Re I
Ip Il yk |k −1 = H ( χ kx|k −1 , χ kn|k −1 ) (5)

U oc Cp Cl U
2L
Up Ul ~
yk− = ¦ Wi ( m ) yi ,k |k −1 (6)
i =0

Fig. 3. The second-order RC model of battery


­ 2L
~− ~− T
° P~yk ~yk = ¦ Wi ( yi.k|k −1 − yk )( yi.k|k −1 − yk )
(c)
Uoc is the open circuit voltage of battery; I is the load ° i =0
(7)
current (positive for discharge, negative for charge); U is the ® 2L
° P~ ~ = W ( c ) ( y ~− ~− T
terminal voltage; Re is the Ohmic resistance; Rp and Rl are the °¯ xk yk ¦ i i .k |k −1 − xk )( yi . k |k −1 − y k )
polarization resistance; Cp and Cl are the polarization i =0

capacitance.
In this way, UKF algorithm can be conducted effectively
and SOC as a variable in the state vector xk can be estimated
IV. THE STATE ESTIMATION
realtimely.
The state estimation of SOC and SOP will be conducted in
the joint state estimator. The core structure is the SOC B. The SOP estimation of baterry
estimator, and SOP will be calculated based on the SOC
SOP can supply the information of the battery peak power
values.
capability, which can avoid the battery discharging at high
currents. Due to the complexity and computation amount of the
A. The SOC estimation with Kalman filtring algorithm second-order RC model for the SOP estimation, the first-order
There are various Kalman filtering algorithms, and the RC model is more appropriate for EMS. As shown in Fig. 1,
extended Kalman filtering algorithm (EKF) and unscented the two RC components can be replaced by an equivalent RC
Kalman filtering algorithm (UKF) are two main algorithms. component. The corresponding parameter identification can be
UKF possesses better approximation capability of nonlinear carried out according to the methods mentioned above.
systems than EKF. So in this paper, UKF has been employed.
SOC can be set as a variable in the state equations of
battery and supercapacitor. According to the principle of UKF,
the sigma point set should be determined:


Rsum A. The frame of the fuzzy controller
preq
Csum SOCb Fuzzy pbat
A Fuzzy Popt
SOCc B
Re I
SOP

U oc U Fig. 5. The frame of the fuzzy controller

There are three input variables of fuzzy logic controller A:


Preq, SOCb and SOCc. One output variable: Pbat. Pbat is the
Fig. 4. The equivalent first-order RC model of battery power which should be provided by battery and it is calculated
based on the comprehensive consideration of Preq, SOCb and
For the sake of brevity, the detailed derivation process is SOCc.
omitted here and presented in ref. [15].

Pdis − max ≈ min{Pmax , (U oc (Z k + L ) − U sum ,k (exp(


− Δt
)) L ...
B. The fuzzy rules of controller A
τ
(8)
When the vehicle accelerates, the request power of motor
− Δt L −1 − Δt will be high. The battery and supercapacitor should provide
... − I dis −max ( Re + Rsum (1 − exp( )) ¦ (exp( )) L−1− j )) I dis −max }
τ j =0 τ power to the vehicle, and the supercapacitor should respond to
more power demand in order to keep the battery discharge at
U oc ( Z k ) − U sum,k (exp(
− Δt
)) L − U t ,min
(9) relatively low currents. Moreover, the supercapacitor should
I dis −max = τ respond for quick variation of load and provide transient
η i LΔt dU oc ( z ) − Δt L −1 − Δt power peaks. When the vehicle drives at a constant speed and
+ Rsum (1 − exp( )) Σ (exp( )) L−1− j + Re
Ca dz Zk
τ j =0 τ the required power is stable, the power distribution principle
will be more complex and depend on SOCb and SOCc.
In the equation, Pdis-max is the maximum power that the
The control strategy is as followed:
battery can provide at k th sampling time. Pmax is the rated
maximum power; ǻt is the sampling time; Idis-max is the
maximum current that the battery can discharge; zk is the SOC TABLE I. THE FUZZY RULES WHEN THE REQUIED POWER IS STABLE
at k th sampling time; ηi is the Coulombic efficiency; Ca is the
SOCc
rated capacity; L is the number of ǻt, which means that the High Medium Low
SOCb
battery can provide the maximum power for Lhǻt.
Battery: d Battery: d
High Battery: d
Then, SOP can be set as Pdis-max: SC: d SC: b
Battery: d Battery: d
Medium Battery: d
SOP = Pdis−max (10) SC: d SC: d
Battery: b Battery: b
Low SC: d
SC: d SC: d

V. THE FUZZY CONTROL STRATEGY DESIGNING


In this paper, the refined speed-based strategy is employed. In the table, ‘d’ means discharge, and ‘b’ means be
The vehicle speed has been adopted as the reference value to charged.
conduct the energy distribution control. In the consideration of When SOCb is high, the battery will always provide power
the relationship between vehicle speed and power demand, the to the motor, and the supercapacitor will discharge only if
vehicle power is employed in the controller. Because the SOCc is high too. If SOCc is low, the supercapacitor will be
power demand is simpler for analysis and calculation as well charged by the battery.
as can avoid the errors in the conversion of vehicle speed.
When SOCb is medium, the battery will also provide
The EMS should determine the split of power between the power, and the supercapacitor will not discharge only if SOCc
battery and supercapacitor to meet the power requirement of is low.
motor, as well as maintain the SOC of battery and
supercapacitor within rational range. The power distribution When SOCb is low, the supercapacitor will provide power
between the battery and supercapacitor is determined by to the load and charge the battery only if SOCc is not low.
taking account of the required power of the load and the SOC Because the low SOCb will bring negative impact on the
of battery and supercapacitor in the first fuzzy controller. The battery life, the energy in the supercapacitor should be
output of this fuzzy controller is the power which should be transferred into the battery.
supplied by battery and then it is taken as the input of the When the vehicle decelerates, the request power is low and
second fuzzy controller with SOP. These two variables will the regenerative energy will flow into the HESS. At that time,
estimate the optimal power distribution of battery which can the supercapacitor should be the first choice to store the
minimize the impact on battery life. energy no matter what SOCc is. When the supercapacitor is


fully charged, the rest of regenerative energy will flow into the REFERENCES
battery. [1] Wang K, Zhang L, Ji B, et al, “The thermal analysis on the stackable
supercapacitor,” Energy, 2013, 59, pp. 440-444.
C. The fuzzy rules of controller B [2] Li Z, Zhu C, Jiang J, et al. “A 3-kW Wireless Power Transfer System for
Sightseeing Car Supercapacitor Charge,” IEEE Transactions on Power
There are two input variables of fuzzy logic controller B: Electronics, 2017, 32(5), pp. 3301-3316.
Pbat and SOP, and one output variable: Popt. Popt is the power [3] Yoichi Hori, “Future vehicle society based on electric motor, capacitor
which will be provided by battery and it is calculated on the and wireless power supply,” in International Power Electronics
base of the comprehensive consideration of Pbat and SOP. Cnference, 2010, pp. 2930-2934.
[4] K. Wang, L. W. Li, H. X. Yin, and T. Z. Zhang, “Thermal modelling
The control strategy is as followed: analysis of spiral wound supercapacitor under constant-current cycling,”
Plos One, 2015, 10, e0138672.
TABLE II. THE FUZZY RULES OF THE CONTROLER B [5] J.M.A Curti, X. Huang, R. Minaki, and Y. Hori, “A Simplified Power
Management Strategy for a Supercapacitor/Battery Hybrid Energy
Pbat Storage System Using the Half-Controlled Converter,” in IECON 2012-
High Medium Low 38th Annual Conference on IEEE Industrial Electronics Society, 2012,
SOP pp. 4006-4011.
High Pbat Pbat Pbat [6] Lai C M, Lin Y J, Hsieh M H, et al. “A newly-designed multiport
bidirectional power converter with battery/supercapacitor for hybrid
Medium SOP Pbat Pbat
electric/fuel-cell vehicle system,” Transportation Electrification Asia-
Low SOP SOP Pbat Pacific (ITEC Asia-Pacific), 2016 IEEE Conference and Expo. IEEE,
2016, pp. 163-166.
[7] Kiyotaka Kawashima, Toshiyuku Uchida, and Yoichi Hori,
When SOP is high, the battery possesses the capability to “Development of a novel ultracapacitor electric vehicle and methods to
cope with voltage variation,” in Vehicle Power and Propulsion
cover the power demand, the output is Pbat. Conference, 2009, pp. 724-729.
When SOP is medium, the battery can’t provide the power [8] K. Wang, L. W. Li, T. Z. Zhang, and Z. F. Liu, “Nitrogen-doped
of high Pbat, and the output is SOP. In other situations, the graphene for supercapacitor with long-term electrochemical stability,”
Energy, 2014, 70, pp. 612–617
output is Pbat.
[9] Sibo Wang, Zhiping Qi and Tongzhen Wei, “Fuzzy Logic Energy
When SOP and Pbat are low, the output is Pbat. Otherwise, Management Strategy for Supercapacitor-based Energy Saving System
the output is SOP. for Variable-speed Motor Drives,” in Electrical Machines and Systems,
2008, pp. 1473-1478.
If the output is SOP, the difference between Pbat and SOP [10] Xu L, Li J, Hua J, Li X, and Ouyang M, “Optimal vehicle control
should be provided by the supercapacitor. However, if the strategy of a fuel cell/battery hybrid city bus,” J. Hydrogen Energy, 2009,
supercapacitor cannot completely cover the difference, the rest 34(17), pp. 23-33.
will be supplied by the battery compulsively, and the EMS [11] L. Rosario and P. C. K. Luk, “Applying Management Methodology to
Electric Vehicles with Multiple Energy Storage Systems,” Proceedings
will alarm. of the Sixth International Conference on Machine Learning and
Cybernetics, 2007, pp. 4223-4230.
VI. CONCLUSION [12] Gao Dawei, Jin Zhenhua, and Lu Qingchun, “Energy management
strategy based on fuzzy logic for a fuel cell hybrid bus,” J. Power
This paper proposes a novel fuzzy logic control strategy. Sources, vol. 185, 2008, pp. 311-317.
The SOC and SOP are estimated on the base of Kalman [13] O. Erdinc, B. Vural, and M. Uzunoglu, “A wavelet-fuzzy logic based
filtering algorithm and then employed into the fuzzy logic energy management strategy for a fuel cell/battery/ultra-capacitor hybrid
algorithm to calculate the optimum power which should be vehicular power system,” J. Power Sources, vol. 194, 2009, pp. 369-380.
provided by the battery. Due to the adoption of Kalman [14] P. Thounthong, S. Sikkabut, A. Luksanasakul, P. Koseeyaporn, and P.
filtering algorithm, the errors of SOC are effectively Sethakul, “Fuzzy logic based DC bus voltage control of a stand alone
photovoltaic/fuel cell/supercapacitor power plant,” in Proc. 11th EEEIC,
alleviated. Furthermore, SOP can supply the information of Venice, Italy, 2012, pp. 725-730.
the battery peak power capability, which can avoid the battery
[15] Rui Xiong, Hongwen He, Fengchun Sun, Xinlei Liu, and Zhentong Liu,
discharging at high currents. The proposed control strategy “Model-based State of Charge and Peak Power Capability Joint
will guarantee the rational distribution of power between the Estimation of Lithium-ion Battery in Plug-in Hybrid Electric Vehicles,”
supercapacitor and battery, which can improve the stability J. Power Sources, vol. 229, 2013, pp. 159-169.
and efficiency of HESS.



You might also like