Development of Energy Management System Based On A Rule-Based Power Distribution Strategy For Hybrid Power Sources
Development of Energy Management System Based On A Rule-Based Power Distribution Strategy For Hybrid Power Sources
Development of Energy Management System Based On A Rule-Based Power Distribution Strategy For Hybrid Power Sources
Energy
journal homepage: www.elsevier.com/locate/energy
a r t i c l e i n f o a b s t r a c t
Article history: The proton exchange membrane fuel cell has become a good candidate for the power source of future
Received 17 November 2018 green transportations due to its excellent performance of high efficiency and cleanness. To overcome the
Received in revised form slow dynamic characteristic, the fuel cells are always grouped with other energy storage devices such as
24 March 2019
lithium-ion batteries and supercapacitors. The development of energy management system and design
Accepted 26 March 2019
of the power distribution strategy are critical issues for the hybrid power source system. In previous
Available online 29 March 2019
work, few studies have comprehensively considered the criteria of power capabilities of the batteries and
supercapacitors which are intimately correlated with the power requirements of accelerating, gradient
Keywords:
Hybrid power source system
climbing and regenerative braking. The development of power distribution strategy considering the
Energy management system constraints of the power capability is urgent and critical to the safety and longevity of the hybrid energy
Power distribution strategy storage system. In this paper, a distributed energy management system is developed for the hybrid
Power capability prediction power source system based on a rule-based power distribution strategy. The presented power distri-
bution strategy has comprehensively considered the criterias of the demand power, the remaining ca-
pacity and power capability of the hybrid power source system. Moreover, the Bayes Monte Carlo
approach is employed for co-estimation of the remaining capacity and power capability of the batteries
and supercapacitors. Compared with the existing rule-based strategy without considering the re-
strictions of the power capability, the presented strategy has better rationality in terms of fuel economy
and dynamic property. The presented energy management strategy can extend the lifespan and improve
the economy of the hybrid energy storage system by employing the charge and discharge limits of power
capability and residual capacity.
© 2019 Elsevier Ltd. All rights reserved.
1. Introduction has become a good candidate for the power source of future green
transportations [2].
1.1. Background and motivation Compared with the battery-powered electric vehicles (BEVs),
the fuel cell vehicles (FCVs) have the superiorities in energy effi-
As one of the most promising technologies for clean and effi- ciency, endurance mileage, charging speed and climate tolerance
cient power generation, the fuel cell has received much attention [3]. However, a FCV with a fuel cell as its only power supply has
during the last couple of decades or so. The fuel cell is defined as an several disadvantages. For instance, the fuel cell provides slow
electrochemical device which converts the chemical energy of the start-up and power response and has relatively lower efficiency at
fuel directly into electrical energy [1]. Compared with other types low and high output power conditions. In addition, because of its
of fuel cells such as the solid oxide fuel cells, the alkaline fuel cell, unidirectional energy flow characteristic, it is incapable of energy
the phosphoric acid fuel cell, etc., the proton exchange membrane saving from regenerative braking. Moreover, its lifespan is
fuel cell has the advantages of high power density, low operating impacted dramatically when the demand power changes rapidly
temperature, good dynamic characteristics, etc. Owing to these [4]. To overcome the slow dynamic characteristic and other draw-
excellent performance, the proton exchange membrane fuel cell backs of the fuel cells, the fuel cells are always grouped with other
energy storage devices such as the battery, supercapacitor, or a
combination of them both [5]. In general, the battery have a higher
* Corresponding author. specific energy density which can provide extra power for a longer
E-mail address: chenzh@ustc.edu.cn (Z. Chen). period of time. The supercapacitor have a higher specific power
https://doi.org/10.1016/j.energy.2019.03.155
0360-5442/© 2019 Elsevier Ltd. All rights reserved.
1056 Y. Wang et al. / Energy 175 (2019) 1055e1066
density and longer lifetime [6]. The problems faced by alternative power management of hybrid electric vehicles which can identify
single battery with low power density or supercapacitor with low the high-frequency transient of the load power, and allocating
energy density can be well solved by the hybrid energy storage power components with different frequency contents. Xu et al. [13]
system (HESS) for the hybrid electric vehicles (HEVs) [7]. The hy- presented an optimal energy management strategy based on the
bridization can take the advantages of both the batteries and pontryagin's minimal principle and dynamic programming. The
supercapacitors, and overcome the drawbacks of the pure fuel cell strategy is compared in simulating models and the fuel economy is
system [8]. In order to meet the vehicle requirements such as the analyzed. Vahidi et al. [14] formulated the distribution of current
demand power, safety and reliability, the development of energy demand between the fuel cell and the bank of supercapacitors in a
management system and design of the power distribution strategy model predictive control framework. Lopez et al. [15] applied a
are critical techniques. rule-based power management system where a low pass filter
splits the power between the different power sources. The benefits
1.2. Literature review of the energy management with respect to the oxygen starvation
prevention and reduction of startups and shutdowns are analyzed.
To fulfill the power balance between the load power and the Xiong et al. [16] compared three typical power management stra-
power sources including the fuel cells, the batteries and super- tegies, which are based on rules, dynamic programming algorithm,
capacitors, the energy management system plays an important role and real-time reinforcement learning algorithm. A hardware-in-
in the hybrid system. A variety of energy management strategies loop (HIL) simulation test bench has been established to verify
have been put forward to solve the power distribution issues in the the performance of the three power management strategies. Zhou
hybrid power sources system. For instance, Corbo et al. [9] et al. put forward a unified dynamic programming model and its
employed a load-following control strategy for a small-size electric solution framework for the hybrid electric vehicles [17] and the fuel
propulsion system in order to regulate the power flows between cell electric vehicles [18]. Segura et al. [19] proposed a hybrid
the fuel cells, the energy storage systems and drive loads. Li et al. control based power management strategy. The hybrid control
[10] presented a fuzzy logic control strategy for the hybrid power concept joining constant and variable control techniques was
sources system. The results indicate that the presented method has studied on different topologies of fuel cell based systems. Yuan et al.
better rationality and validity in terms of dynamic property and fuel [20] proposed a novel hierarchical reinforcement learning strategy
economy. Furthermore, to reduce the damage of rapid changes of to approximate global optimization. The long-term and short-term
large current, a power distribution strategy based on wavelet speeds are predicted by k-Nearest Neighbor (KNN) and a model
transform and fuzzy logic control is presented in Ref. [11]. Zhang averaging method. The results show that the proposed strategy can
et al. [12] presented a wavelet transform based strategy for the extend the fuel cell service life and has lower energy consumption
Y. Wang et al. / Energy 175 (2019) 1055e1066 1057
compared with the widely used rule-based energy management management systems. The remaining capacity and power capa-
strategy. Du et al. [21] proposed an optimal strategy for minimizing bility of the hybrid systems are intensively dependent on the
the battery degradation and the life cycle economic cost of a battery operation and charging patterns and intimately correlated with the
and supercapacitor hybrid energy storage system. A battery power requirements of accelerating, gradient climbing and regen-
degradation model was adopted to assess the battery degradation erative braking. The real-time predictions of the remaining capacity
cost. A two-dimensional Pontryagins Minimum Principle (PMP) and power capability are complicated because of the sensor noise
algorithm was proposed to derive the optimal strategy to manage and model nonlinearity [32e34]. Moreover, the relationship be-
the hybrid system. Hu et al. [22] proposed a multi-objective power tween the power capability and remaining capacity are complex
allocation strategy for the hybrid fuel cell vehicles by taking the and coupled [35e37]. The development of power distribution
battery size into consideration. The battery capacity and fuel cell strategy considering the constraints of the power capability is ur-
service life for the system life cycle cost are optimized. Wang et al. gent and critical to the safety and longevity of the hybrid energy
[23] proposed a fuzzy-logic power management strategy for an storage system.
active parallel battery and supercapacitor hybrid energy storage
system based on Markov random prediction. Zhang et al. [24] 1.3. Original contributions
developed a model predictive control for power management in a
plug-in hybrid electric vehicle. The robustness of the proposed To the best of our knowledge, there are few papers focusing on
approach was verified by three typical driving cycles. The results the power capability as well as the remaining capacity when
showed that the proposed model predictive control strategy can developing the power allocation strategy. Therefore, this study
promote fuel economy compared with the original control strategy. combined with our previous work in system modeling and state
Yu et al. [25] proposed an active power flow control strategy based estimation attempts to develop a distributed energy management
on optimal control theory in order to meet the demand of different system for the hybrid power source system and provides some new
vehicle loads while optimizing the total energy cost, battery life and insights in terms of the control strategy of power allocation. The
other possible objectives. The results indicated that the total energy following original contributions make this paper different from its
was well saved over long driving cycles, and the fuel cell and bat- relevant ones. First, the equivalent circuit model frameworks of the
teries were kept operating in a healthy way. Odeim et al. [26] lithium-ion battery, supercapacitor and fuel cell hybrid power
compared both the off-line and real-time power management source system are established, and a novel distributed energy
strategies for the fuel cell/battery/supercapacitor hybrid vehicular management system is developed based on the controller area
system. The off-line dynamic programming and Pontryagin's min- network. Second, a rule-based control strategy is proposed for the
imum principle optimization algorithms were compared with the distributed energy management system which has comprehen-
real-time multi-objective genetic algorithm. The real-time strategy sively considered the criterias of the demand power, the remaining
was found to consume more hydrogen but dramatically improved capacity and the power capability of the energy storage devices.
the system durability. Zhang et al. [27] compared the battery state- The presented rule-based control strategy is easy to implement and
of-charge balanced strategy and the dynamic programming strat- has less calculation time compared with other real-time control
egy for the power allocation of a fuel cell and battery hybrid system. strategies. Third, the Bayes Monte Carlo method is employed for co-
Moreover, a simplified form of dynamic programming strategy was estimation of the remaining capacity and power capability of the
deduced for accelerating the calculation, which has equivalent batteries and supercapacitors based on the presented model
hydrogen consumption as the battery state-of-charge balanced frameworks in order to overcome the initial bias and noises. Finally
strategy but ensures the durability of the fuel cell. experimental and simulated studies are conducted to verify the
In recent years, simulation technology is cultivated in the fields performance of the proposed method. The results indicate that
of control, system engineering, computer, etc., and plays an compared with the existing rule-based strategy without consid-
important role in the fields of information, materials and energy. ering the criteria of the power capability, the presented strategy has
Bubna et al. [28] presented simulation studies on the benefits of better rationality in terms of fuel economy and dynamic property.
adding supercapacitors to a fuel cell and battery hybrid transit bus The presented energy management strategy can extend the life-
operating on two standardized driving schedules. Hannan et al. [29] span and improve the economy of the hybrid energy storage system
presented a feedback control algorithm of an energy management by employing the charge and discharge limits of power capability
system for a battery, fuel cell and supercapacitor hybrid system, the and residual capacity.
algorithm is simulated using Matlab/Simulink. Ettihir et al. [30]
proposed an adaptive optimal power distribution strategy for the 1.4. Outline of the paper
fuel cell and battery hybrid system. The online optimization based
on the PMP was established to minimize the hydrogen consump- The remainder of the paper proceeds as follows. The configu-
tion. The effectiveness of the optimal power distribution strategy is ration and model frameworks of the lithium-ion battery, the
verified by simulations on two fuel cell systems with different supercapacitor and fuel cell hybrid power source system are
levels of degradation. introduced in Section 2. The energy management system and the
The power management strategies both real-time and off-line rule-based power distribution strategy are introduced in Section 3.
can be generally classified into the following categories: dynamic The experiments and verifications are discussed in Section 4.
programming, fuzzy logic control, model predictive control, rule- Finally, the conclusions are given in Section 5.
based, and the optimization principles based on equivalent fuel
consumption or degradation cost minimization strategies. These 2. Configuration and model framework of the hybrid power
strategies can deal with the power distribution problems with their sources
own properties and a comparative study of these strategies can be
found in Ref. [31]. To extend the battery life and ensure the safety 2.1. Configuration of the hybrid power source system
operation of the hybrid energy storage system, it is crucial to
enhance the control constraints of the energy devices such as the The drive structure of the lithium-ion battery, supercapacitor
batteries and supercapacitors. However, the power capability still and fuel cell hybrid power source system is presented in Fig. 1.
lacks consideration in the evaluation and design of energy In the presented structure, the fuel cell stack, the lithium-ion
1058 Y. Wang et al. / Energy 175 (2019) 1055e1066
Unidirectional
DC/DC Convertor
Fuel Cell System
Bidirectional
DC/DC Convertor
Lithium-ion Invertor Motor
Batteries DC/AC Drive
System
Bidirectional
DC/DC Convertor
Supercapacitors
There are three energy models used in the hybrid power source
system: the lithium-ion batteries, the supercapacitors and the fuel 2.2.2. Supercapacitor pack model
cell stack as shown in Fig. 2. The details are explained as follows. The schematic diagram of the equivalent circuit model of the
supercapacitor pack can be found in Fig. 2 (b). The electrical
behavior of the cell model of the supercapacitor can be described as
follows:
2.2.1. Lithium-ion battery pack model
The schematic diagram of the equivalent circuit model of the V_ m;l ¼ ic Cm;l (4)
lithium-ion battery pack is shown in Fig. 2 (a). The electrical
behavior of the cell model can be described as follows:
Y. Wang et al. / Energy 175 (2019) 1055e1066 1059
(a)
Rp,1 Rp,2 Rp,N
ib
...
Vocv,1 Ro,1 Vocv,2 Ro,2 Vocv,N Ro,N
Cp,1 Cp,2 Cp,N
(b)
ic
...
Rc,1 Cm,1 Rc,2 Cm,2 Rc,L Cm,L
Fig. 2. Equivalent circuit model framework of the hybrid power source system: (a) lithium-ion battery pack equivalent circuit model. (b) Supercapacitor pack equivalent circuit
model. (c) Fuel cell stack equivalent circuit model.
ifc c3
stack voltage model, the open-circuit voltage and three main types Vfc ¼ E v0 va 1 ec1 ifc ifc Rohm ifc c2 (12)
of losses in the fuel cell including the activation loss, the ohmic loss, imax
and the concentration loss should be considered [38]. Therefore, Based on the above analysis as well as considering the dynamic
the voltage supplied by a fuel cell can be written as: behavior of the fuel cell, the equivalent circuit model shown in
Fig. 2 (c) is presented to approximate the fuel cell voltage. Using Eq.
Vfc ¼ E Vact Vohm Vconc (7)
(12), we define the activation and concentration resistance as:
where Vfc represents the fuel cell voltage, E represents the open- 1h i
circuit voltage of the fuel cell which is calculated from the energy Ract ¼ v þ va 1 ec1 ifc (13)
ifc 0
balance between chemical energy in the reactants and electrical
energy, Vact represents the activation voltage loss, Vohm represents
c3
the ohmic voltage loss, and Vconc represents the concentration ifc
Rconc ¼ c2 (14)
voltage loss. imax
For the polymer electrolyte membrane fuel cell, the open-circuit
The electrical behavior of the fuel cell model can be described as
voltage E can be expressed by the following equation:
follows:
E ¼ 1:229 0:85 103 Tfc Tref þ 4:3085 dVfc;m Vfc;m Vfc;m init
Cfc;m þ ¼ ist (15)
1 dt Ract;m þ Rconc;m
105 Tfc ln pH2 þ ln pO2 (8)
2
Vcell;m ¼ Em Vfc;m ist Rohm;m (16)
where Tfc is the fuel cell temperature expressed in Kelvin, Tref is the
reference temperature (Tref ¼ 298.15 K), pH2 and pO2 are the where m denotes the cell number of the fuel cell stack (m ¼ 1, 2, 3,
hydrogen and oxygen pressure expressed in atm. …, M), Ract,m and Rconc,m represent the activation resistance and
1060 Y. Wang et al. / Energy 175 (2019) 1055e1066
concentration resistance of cell m, Cfc,m represents the equivalent supercapacitors and the fuel cell stack, the energy management
capacitance of cell m, Vfc,m represents the voltage of the equivalent system and the rule-based power distribution strategy are intro-
capacitance of cell m, Vfc,m_init represents the initial voltage of the duced in this section.
equivalent capacitance of cell m, Vcell,m represents the terminal
voltage of fuel cell m, Em represents the open-circuit voltage of fuel
3.1. Energy management system
cell m, Rohm,m represent the ohmic resistance of cell m.
Therefore the stack voltage of the fuel cell can be described as
The energy management system of the hybrid power source
follows:
system consists of a host controller, a battery management unit, a
M supercapacitor management unit and a fuel cell management unit
X
M X
Vstack ¼ Vcell;m ¼ Em Vfc;m ifc Rohm;m (17) as shown in Fig. 3. The information transmission is based on the
m¼1 m¼1 controller area network. The host controller is developed to
communicate with the motor drive system and the lithium-ion
where Vstack represents the stack voltage of the fuel cell. battery, supercapacitor and fuel cell management units in order
To further determine the hydrogen and oxygen pressure in Eq. to obtain the vehicle demand power, and real-time status of the
(8), the anode flow model, and the cathode flow model are devel- power sources. In addition, the power distribution strategy is in-
oped as follows: tegrated into the host controller. Based on the power distribution
strategy, the required power of the lithium-ion batteries, the
dpH2 RH2 Tfc in
supercapacitors and the fuel cell system can be calculated
¼ q qrH2 qout (18)
dt Volanode H2 H2
automatically.
The battery and supercapacitor management units in this sys-
dpO2 RO2 Tfc in tem are used to measure and protect the lithium-ion battery and
¼ qO2 qrO2 qout
O2 (19) supercapacitor systems. On the one hand, the management units
dt Volcathode
estimate the key states of the batteries and supercapacitors such as
where RH2 represents the hydrogen gas constant, Volanode repre- the state-of-charge (SOC), state-of-voltage (SOV), maximum
sents the volume of anode, qin charge/discharge power capability, etc. These key states are requi-
H2 represents the hydrogen input flow,
site input parameters for the rule-based power distribution strat-
qrH2 represents the hydrogen reaction flow, qout
H2 represents the egy. On the other hand, they control the bidirectional DC/DC
hydrogen output flow, RO2 represents the oxygen gas constant, converters based on the commands of the host controller.
Volcathode represents the volume of cathode, qin
O2 represents the The fuel cell management unit is used to manage the fuel cell
oxygen input flow, qrO2 represents the oxygen reaction flow, qout
O2
stack and the balance of plant (BOP) system. The BOP system
represents the oxygen output flow. mainly consists of a hydrogen supply subsystem, an oxygen supply
subsystem, a water recycling subsystem and a cooling and heating
subsystem. The hydrogen supply subsystem is composed of a
3. Energy management system and power distribution hydrogen gas tank, a pressure reducing valve and a humidifier. The
strategy oxygen supply subsystem is composed of an air compressor, a
cooler and a humidifier. By changing the flow of hydrogen and
To fulfill the power balance between the demand power and the oxygen, the power output of the fuel cell system can meet the
power sources including the lithium-ion batteries, the required power calculated by the host controller. The main task of
Pb Psc
SOCb Pfc
SOPb
SOVsc
SOPsc
Pfc,max
Host Controller Battery Management Unit Supercapacitor Management Unit Fuel Cell Management Unit
Pm V I T V I T V I T p q s
controller. Therefore the overall weight of all particles and the prediction
result can be derived as:
3.2. Model-based remaining capacity and power capability
prediction ! XNs !i
Xk ¼ X ,uik
i¼1 k
(24)
The remaining capacity and power capability of the lithium-ion
For the supercapacitors, the relationship between the remaining
batteries and supercapacitors are key restrictions for the power
capacity and terminal voltage is more linear, and the SOV is used to
distribution strategy. In this section, the model based remaining
remind the remaining capacity which can be deduced as follows:
capacity and power capability prediction methods are introduced.
.
3.2.1. Remaining capacity prediction SOV ¼ Vpackc Vpackc;min Vpackc;max Vpackc;min (25)
For the lithium-ion batteries, the relationship between the
remaining capacity and voltage is nonlinear, therefore the SOC is where Vpackc represents the terminal voltage of the supercapacitor
used to remind the remaining capacity which serves a similar pack, Vpackc,max and Vpackc,min represent the maximum and mini-
function to the petrol gauge in the conventional gasoline car. The mum cut-off voltage of the supercapacitor pack.
SOC is defined as the percentage of the remaining capacity to the
maximum available capacity, and can be calculated by using the
ampere-hour counting method which is simple and easy to realize.
The definition of SOC and calculation of the ampere-hour counting 3.2.2. Power capability prediction
method are as follows: The power capability of the batteries is affected by many factors
such as the charge and discharge current, the terminal voltage, the
ð
t1
remaining capacity, etc. The continuous charge and discharge po-
SOCt1 ¼ Crem =Cmax ¼ SOCt0 þ h ib dt=Cmax (20) wer capability for the lithium-ion batteries can be derived as:
t0
8
< P chg ¼ max Pmin ; Vpackb I chg
where SOCt0 and SOCt1 represent the battery SOC at time t0 and t1 SOPb ¼
min
min
(26)
respectively, ib represents the battery current, Crem and Cmax : P dchg ¼ min Pmax ; V dchg
max packb I max
represent the remaining capaity and maximum available capacity,
and h represents the coulombic efficiency.
where P chg and P dchg
max represent the minimum power capability for
The ampere-hour counting method is definitely a simple min
approach to calculate the SOC when the SOC at time t0 is certain charging and maximum power capability for discharging (define
and the current ib is measured accurately. Unfortunately, inevitable negative for charging and positive for discharging), Pmin and Pmax
sensor noises provide accumulated error by using this method. represent the cut-off power for charging and discharging, I chg
min
and
Moreover, the initial SOC value of the lithium-ion battery is difficult I dchg
max represent the minimum charging current and maximum dis-
to determine at the flat voltage platform area. This phenomenon is charging current which are affected by the designed current limits,
even more significant for the lithium iron phosphate batteries. To the terminal voltage and SOC.
handle the above-mentioned problems, the Bayes Monte Carlo chg
The minimum charging current I min and maximum discharging
approach is applied in the SOC estimation because of its superiority
in solving nonlinear and non-Gaussian problems. The key of the current I dchg
max can be calculated as:
Bayes Monte Carlo approach is to approximate the probability 8
density function by a set of random samples with associated < Ichg ¼ max I chg;des ; I chg;volt ; I chg;SOC
min min min min
weights: (27)
: Idchg ¼ min I dchg;des ; I dchg;volt ; I dchg;SOC
max max max max
! XNs ! !i
P X k Vb;k z i¼1 uik ,d X k X k (21)
where I chg;des
min
and I dchg;des
max represent the designed current limits,
!i T chg;volt
I min
dchg;volt
and I max represent the current considering the influence
where X k ¼ ½SOC ik ; V ip;k represents the set of random particles
! of terminal voltage, I chg;SOC and I dchg;SOC represent the current
drawn from Pð X k Vb;k Þ, Ns represents the size of the random par- min max
considering the influence of SOC.
ticles, uik represents the importance weight associated with each
!i 8
particle X k .
> chg;volt Vocv;kþL Vp;k eLDt =Rp Cp Vt;max
>
> Imin ¼
The updating law of the importance weights can be derived as: >
> XL1 L1j
>
0
1
>
>
> Ro þ Rp 1 eDt=Rp Cp eDt=Rp Cp
>
<
!i Vb;k Vb j¼0
B 1 b;k C
uik ¼ uik1 P Vb;k X k ¼ uik1 pffiffiffiffiffiffiffiffiffiffiffi exp B
1 C >
@2 s2 A >
>
> Vocv;kþL Vp;k eLDt =Rp Cp Vt;min
2ps2 >
> I dchg;volt
¼
>
> max XL1 L1j
>
>
: Ro þ Rp 1 eDt=Rp Cp eDt =Rp Cp
(22)
j¼0
b (28)
where Vb;k and V b;k are the measured and model output values.
Then the overall weight of all particles can be normalized as:
1062 Y. Wang et al. / Energy 175 (2019) 1055e1066
sc
No {
SOC > UTb { Pb = 0 |Pm| > |SOPb| Pb = -|SOPb|
Model-based Pm > 0 ? SOV > UTsc Psc = 0 {
SOC < UTb Psc = -|Pm-|Pb||
SOC/SOV SOC > UTb {
{ Pb = 0
|Pm| < |SOPb|
{ SOV < UTsc { Pb = -|Pm|
Estimation Yes SOV < UTsc Psc = -|Pm| Psc = 0
Eqs.(20), (25)
Discharging mode: Pfc = Pfc,max Pfc = Pfc,max
SOPb/SOPsc
A. Pfc,max > Pm SOC < LTb { Pb = Pm - |Pfc,max| SOC < LTb { Pb = -|Pm - |Pfc,max||
{ {
SOV < LTsc Psc = 0 SOV > LTsc Psc = 0
SOC > LTb Pfc = Pfc,max SOC > LTb Pfc = Pm
SOP Prediction {
SOV < LTsc { Pb = 0 {
SOV > LTsc { Pb = 0
Eqs. (26), (30) Psc = -|Pm - |Pfc,max|| Psc = 0
B. Pfc,max < Pm Pfc = Pfc,max Pfc = Pfc,max
SOC < LTb SOC < LTb
{ { Pb = 0 { { Pb = 0
SOV < LTsc SOV > LTsc
Psc = 0 Psc = Pm - Pfc,max
Startup Mode
Pfc = Pfc,max Pfc = Pfc,max
Braking Mode {
SOC > LTb { SOC > LTb { Pb = Pm - Pfc,max - SOPsc
SOV < LTsc { Pb = Pm - Pfc,max SOV > LTsc
Psc = 0 Psc = SOPsc
Discharging Mode
4. Experiment and simulation verification hybrid topology. The designed power distribution strategy can
satisfy the power requirements for the battery, supercapacitor and
In order to compare the dynamic property of the hybrid power fuel cell hybrid system except for the case when both the SOC of the
source system, the power distribution strategies are designed for battery and SOV of the supercapacitor are under their lower
the battery and fuel cell hybrid system and the battery, super- thresholds as shown in Fig. 5 (b). In addition, the supercapacitors
capacitor and fuel cell hybrid system. The rated voltage of the tested can reduce the burden of the battery pack, and avoid batteries
lithium-ion battery cell is 4.2 V and the rated capacity is 2.2 Ah. The discharging in the large current. The power fluctuations of the fuel
rated voltage of the tested supercapacitor is 2.7 V and the rated cells in the lithium-ion battery, supercapacitor and fuel cell system
capacity is 3000F. The model parameters of the lithium-ion batte- are much smaller than the system without supercapacitor. Since
ries and the supercapacitors are identified by experimental datasets the lifespans of the fuel cells and batteries are influenced by the
(at room temperature 25 C) using recursive least-squares method. cycle times and power fluctuations, the topology with super-
The standard-state reversible voltage of the fuel cell is 1.229 V. The capacitor has a longer service time. The supercapacitor in the
parameters of the hybrid power source system is shown in Table 1. lithium-ion battery, supercapacitor and fuel cell hybrid system can
The simulations of the energy management strategies for the bring its superiorities into full play and prolong the lifespans of the
hybrid system are conducted under different driving cycles using lithium-ion batteries and fuel cells by less charging and
Matlab/Simulink. discharging.
The results of power tracking with different hybrid topologies To further illuminate the fuel economy and dynamic property of
are compared in Fig. 5. For the lithium-ion battery and fuel cell the presented power distribution strategy with power capability
hybrid topology system, with regards to the power curves of the prediction, the power distribution strategy considering SOP esti-
demand power and real power, the real power cannot match the mation are contrasted with the strategy without considering SOP
demand power in some cases as shown in Fig. 5 (a). This is because estimation under two typical driving cycles, the Urban Dyna-
when the SOC of the lithium-ion battery is under its lower mometer Driving Schedule (UDDS) cycle and the Federal Urban
threshold, the fuel cell can not response the quick power change Driving Schedule (FUDS) cycle. The hydrogen consumptions of the
without the supercapacitors. In addition, on account of high power strategy considering SOP estimation and the strategy without
density, the supercapacitor can help absorb energy when the considering SOP estimation under the UDDS cycle (with 10000
vehicle is in braking modes. Therefore the power tracking results of time steps) are 1.909 kg and 2.083 kg, respectively. The strategy
the lithium-ion battery, supercapacitor and fuel cell hybrid topol- considering SOP estimation can save 8.35% hydrogen consumption
ogy are better than that of the lithium-ion battery and fuel cell compared with the strategy without considering SOP estimation.
The power distribution and remaining capacity changes of the
batteries and supercapacitors under UDDS cycle are shown in Fig. 6.
Table 1 From Fig. 6 (a) and (b), we can see that the power fluctuations of
Parameters of the hybrid power source system (at room temperature 25 C).
strategy considering SOP estimation are less than the strategy
Lithium-ion battery Parameters Values without considering SOP estimation. The presented power distri-
Ohmic internal resistance 40.2 mU bution strategy with power capability prediction has the potential
Polarization resistance 41.9 mU
Polarization capacitance 1397.2 F
to prolong the lifespans of the lithium-ion batteries and super-
Supercapacitor Parameters Values capacitors by less high power shocks. Moreover, from Fig. 6 (c) and
Equivalent serial resistance 3.085 mU (d), we can see that the charge and discharge cycle numbers of the
Equivalent serial capacitance 2963.2 F lithium-ion batteries and supercapacitors considering SOP esti-
Fuel cell Parameters Values
mation are smaller than that without considering SOP estimation.
Maximum current density 2.2 A/cm2
Charge transfer coefficient 0.5 This means the energy storage devices can have longer lifespans by
Volume of anode 0.005 m3 the power constraint strategy. In addition, because of the energy
Volume of cathode 0.01 m3 loss on the DC/DC converter, less use of the batteries can reduce
Hydrogen gas constant 4124.36 energy loss and improve the fuel economy.
Oxygen gas constant 259.81
Ideal gas constant 8.31
The hydrogen consumptions of the strategy considering SOP
Faraday constant 96485.33 C/mol estimation and the strategy without considering SOP estimation
Reference temperature 298.15 K under the FUDS cycle (with 10000 time steps) are 1.272 kg and
104 104
D 8 E 8
Demand power Demand power
Real power Real power
6 6
3RZHU:
3RZHU:
4 4
2 2
0 0
-2 -2
0 2000 4000 6000 8000 10000 0 2000 4000 6000 8000 10000
7LPH6WHS 7LPH6WHS
Fig. 5. Power tracking with different topologies: (a) Lithium-ion battery and fuel cell hybrid topology. (b) Lithium-ion battery, supercapacitor and fuel cell hybrid topology.
1064 Y. Wang et al. / Energy 175 (2019) 1055e1066
104 104
D E 5
4 Without considering SOP Without considering SOP
Considering SOP 4 Considering SOP
3
3
2
3RZHU:
3RZHU:
2
1
1
0 0
-1 -1
-2 -2
0 2000 4000 6000 8000 10000 0 2000 4000 6000 8000 10000
7LPH6WHS 7LPH6WHS
F 1 G 1
0.8
0.8
0.6
62&
629
0.6
0.4
0.4
0.2
Without considering SOP Without considering SOP
Considering SOP Considering SOP
0.2 0
0 2000 4000 6000 8000 10000 0 2000 4000 6000 8000 10000
7LPH6WHS 7LPH6WHS
Fig. 6. Power distribution and remaining capacity change under UDDS: (a) Power of lithium-ion batteries. (b) Power of supercapacitors. (c) SOC of lithium-ion batteries. (d) SOV of
supercapacitors.
1.365 kg, respectively. The presented strategy can save 6.81% capacity changes of the batteries and supercapacitors under the
hydrogen consumption compared with the strategy without FUDS cycle are shown in Fig. 7. The cycle numbers of the batteries
considering SOP estimation. The power distribution and remaining and supercapacitors considering SOP estimation are smaller than
104 104
D 4 E
Without considering SOP Without considering SOP
3 Considering SOP 3 Considering SOP
2 2
3RZHU:
3RZHU:
1 1
0 0
-1 -1
-2 -2
0 2000 4000 6000 8000 10000 0 2000 4000 6000 8000 10000
7LPH6WHS 7LPH6WHS
F 1 G 1
0.8
0.8
0.6
629
62&
0.6
0.4
0.4
0.2
Without considering SOP Without considering SOP
Considering SOP Considering SOP
0.2 0
0 2000 4000 6000 8000 10000 0 2000 4000 6000 8000 10000
7LPH6WHS 7LPH6WHS
Fig. 7. Power distribution and remaining capacity change under FUDS: (a) Power of lithium-ion batteries. (b) Power of supercapacitors. (c) SOC of lithium-ion batteries. (d) SOV of
supercapacitors.
Y. Wang et al. / Energy 175 (2019) 1055e1066 1065
Table 2
Performance Comparison of the hybrid system under different driving cycles.
UDDS Battery fluctuations (kW) FC fluctuations (kW) Start-stop times of fuel cell
that without considering SOP estimation. Since the load power of models. Moreover, the Bayes Monte Carlo approach is applied for
the FUDS driving cycle is lower than the UDDS cycle, the power the prediction of the remaining capacity and power capability of
capability prediction algorithm do not limit the use of batteries for the electrical energy storage devices in order to overcome the un-
most cases. It indicates that the presented power distribution certainty of the initial values and noises. Experimental results
strategy with power capability prediction is more effective for the indicate that compared with the strategy without considering the
improvement of fuel economy and dynamic property especially in criteria of power capability assessment, the presented strategy has
high power demand conditions. better rationality in terms of fuel economy and dynamic property.
For more quantitative analysis, the standard deviations of the The configuration and control of the fuel cell, battery and
battery power and fuel cell power are calculated to evaluate the supercapacitor hybrid energy storage system have been studied
power fluctuations of the battery and fuel cell systems. The nu- intensively in recent years. In this work, we discussed the solution
merical results are shown in Table 2. The results indicate that the of the development of the distributed energy management system
power fluctuations of the lithium-ion batteries and fuel cell system and the rule-based power distribution strategy. Our future work
with strategy considering SOP are less than those without consid- will focus on the multi-objective optimization of the fuel cell,
ering SOP both in the UDDS and FUDS driving cycles. In the UDDS battery and supercapacitor hybrid energy storage system, including
driving cycle, compared with the strategy without considering SOP, the optimal configuration and sizing, the development of life
the proposed strategy has 60.1% and 7.3% reductions of standard extending strategy of the fuel cell and battery, and the real-time
deviations for the battery and fuel cell system respectively. In the optimal power splitting strategy.
FUDS the reductions for the battery and fuel cell are 40.0% and 5.4%
respectively. Moreover compared with the strategy without Acknowledgements
considering SOP, the proposed method can evidently reduce fuel
cell start-stop times by 11% on average of the two driving cycles, This work is supported partly by the National Natural Science
which can suppress the aging of the fuel cell system. Fund of China (Grant No. 61803359), partly by the CPSF-CAS Joint
In conclude, the presented energy management strategy can Foundation for Excellent Postdoctoral Fellow (Grant No.
extend the lifespan and improve the economy of the hybrid energy 2017LH007), and partly by China Postdoctoral Science Foundation
storage system by employing the charge and discharge limits of Funded Project (Grant No. 2017M622019).
power capability and residual capacity. On the one hand, the bat-
teries and supercapacitors can avoid the damage caused by over-
References
charge and overdischarge by the constraints of power and
remaining capacity. The power fluctuations of the fuel cell system [1] Sharaf OZ, Orhan MF. An overview of fuel cell technology: fundamentals and
and batteries are reduced by using supercapacitors. On the other applications. Renew Sustain Energy Rev 2014;32:810e53.
hand, the batteries and supercapacitors have more redundancy by [2] Peighambardoust SJ, Rowshanzamir S, Amjadi M. Review of the proton ex-
change membranes for fuel cell applications. Int J Hydrogen Energy
setting the thresholds to cope with instantaneous power output
2010;35(17):9349e84.
and absorption of regenerative braking energy which means more [3] Wang Y, Chen KS, Mishler J, et al. A review of polymer electrolyte membrane
energy can be utilized and stored by the energy storage system. The fuel cells: technology, applications, and needs on fundamental research. Appl
Energy 2011;88(4):981e1007.
usage of the batteries and supercapacitors have been improved and
[4] Xiong R, Li L, Tian J. Towards a smarter battery management system: a critical
the fuel cell system has more stable power outputs. The hydrogen review on battery state of health monitoring methods. J Power Sources
consumption of the hybrid system has reduced significantly in 2018;405:18e29.
different driving cycles. [5] Gao W. Performance comparison of a fuel cell-battery hybrid powertrain and
a fuel cell-ultracapacitor hybrid powertrain. IEEE Trans Veh Technol
2005;54(3):846e55.
5. Conclusions [6] Wang Y, Zhang X, Liu C, et al. Multi-timescale power and energy assessment of
lithium-ion battery and supercapacitor hybrid system using extended Kalman
filter. J Power Sources 2018;389:93e105.
The hybridization of fuel cells with other electrical energy [7] Xiong R, Chen H, Wang C, et al. Towards a smarter hybrid energy storage
storage systems such as batteries and supercapacitors provides a system based on battery and ultracapacitor-A critical review on topology and
energy management. J Clean Prod 2018;202:1228e40.
good solution for the improvement of dynamic property and life- [8] Wang Y, Liu C, Pan R, et al. Modeling and state-of-charge prediction of
span of the fuel cells. The development of energy management lithium-ion battery and ultracapacitor hybrids with a co-estimator. Energy
system and power distribution strategy are critical issues for the 2017;121:739e50.
[9] Corbo P, Corcione FE, Migliardini F, et al. Experimental assessment of energy-
hybrid power source system. In order to meet the vehicle re- management strategies in fuel-cell propulsion systems. J Power Sources
quirements such as the demand power, safety and reliability, this 2006;157(2):799e808.
paper attempts to build a distributed energy management system [10] Li Q, Chen W, Li Y, et al. Energy management strategy for fuel cell/battery/
ultracapacitor hybrid vehicle based on fuzzy logic. Int J Electr Power Energy
for the hybrid power source system based on a rule-based power
Syst 2012;43(1):514e25.
splitting strategy which has fully considered the criteria of the [11] Li Q, Chen W, Liu Z, et al. Development of energy management system based
demand power, the remaining capacity and power capability of the on a power sharing strategy for a fuel cell-battery-supercapacitor hybrid
electrical energy storage systems. The equivalent circuit model tramway. J Power Sources 2015;279:267e80.
[12] Zhang X, Mi CC, Masrur A, et al. Wavelet-transform-based power manage-
frameworks of the lithium-ion battery pack, the supercapacitor ment of hybrid vehicles with multiple on-board energy sources including fuel
pack and the fuel cell stack are developed based on their cell cell, battery and ultracapacitor. J Power Sources 2008;185(2):1533e43.
1066 Y. Wang et al. / Energy 175 (2019) 1055e1066