Professional Documents
Culture Documents
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
Handling Editor: X Ou This paper proposes three online real-time energy management strategies for fuel cell hybrid electric vehicles
aimed at minimizing the total running cost over the vehicle’s lifecycle. The cost comprises hydrogen con
Keywords: sumption, battery electricity consumption, and the degradation of fuel cell and battery systems. The first strategy
Fuel cell hybrid electric vehicles is a rule-based approach that controls fuel cell power at several fixed levels, with the battery state-of-charge
Real-time energy management strategy
serving as the shifting condition between these levels. The second strategy is an optimization-based approach
Cost minimization
that determines the fuel cell power by minimizing the instantaneous running cost. Two coefficients, dependent
Fuel cell degradation
on the battery state-of-charge, are employed to adjust the cost weights of hydrogen consumption, battery
electricity consumption, and fuel cell system degradation. The third strategy combines rules with optimization,
incorporating the rules as constraints into the cost optimization model. To account for practical driving condi
tions throughout the vehicle’s lifecycle, short distance driving cycles, long distance driving cycles, and commuter
driving cycles are separately investigated. The rule and optimization combined energy management strategy
outperforms the others in all driving conditions, benefiting from fewer fuel cell system start-stop cycles and
increased opportunities for operating in high-efficiency regions.
1. Introduction hydrogen consumption, but also the degradation cost of fuel cell and
battery. Due to the vulnerability and high expense of fuel cell and bat
Fuel cells can be divided into Polymer electrolyte membrane fuel tery systems, improper loadings would shorten their lifespans and thus
cells (PEMFC), Solid Oxide Fuel Cell (SOFC), Alkaline Fuel Cell (AFC), increase the vehicle’s maintenance costs, which are the degradation cost
Molten Carbonate Fuel Cell (MCFC), and Phosphoric Acid Fuel Cell of fuel cell and battery. Fuel cell degradation costs, as indicated in
(PAFC). Among these, PEMFCs have received the most attention in Ref. [3], account for a considerable proportion of the overall operating
research and development due to their operation at relatively low cost. The degradation of a fuel cell system is primarily influenced by
temperatures (− 40 ◦ C–120 ◦ C), making them suitable for portable, various operating conditions, such as start-stop cycles, high and low
transportation, and stationary power applications [1]. Fuel cell hybrid loadings, and load shifts [4]. While effective control strategies for
electric vehicles (FCHEVs) offer the advantages of zero pollution and gas polymer electrolyte membrane fuel cell systems (PEMFCS), such as air
emissions like electric vehicles (EVs), while also providing longer supply control, are crucial for ensuring reliability and efficiency [5,6],
driving ranges and shorter refueling times. However, FCHEVs face the an Energy Management Strategy (EMS) for FCHEVs can help reduce
drawback of high costs, primarily due to operating expenses related to hydrogen consumption costs and mitigate power source system degra
onboard fuel cell and battery systems, which constitute a significant dation by appropriately allocating power between the fuel cell system
portion of the total ownership costs [2]. The running cost is not only the and battery.
☆
This paper is submitted for possible publication in Energy. It has not been previously published, is not currently submitted for review to any other journals, and
will not be submitted elsewhere during the peer review.
* Corresponding author.
** Corresponding author.
E-mail addresses: zhangyq@hust.edu.cn (Y. Zhang), ruanjiageng@gmail.com (J. Ruan).
https://doi.org/10.1016/j.energy.2023.129442
Received 7 April 2023; Received in revised form 28 September 2023; Accepted 21 October 2023
Available online 21 October 2023
0360-5442/© 2023 Elsevier Ltd. All rights reserved.
J. Wu et al. Energy 285 (2023) 129442
The EMS allocates the output power between power sources to of charge (SOC). Prediction-based control strategies heavily rely on the
reduce the cost and satisfy the power demand. The power following accuracy of speed prediction, posing challenges for practical imple
strategy and thermostat strategy [7–9] are two typical deterministic mentation. On the other hand, battery SOC feedback control strategies
rule-based control methods. For the power following EMS (PFEMS), the do not require future driving condition information but require further
required power is mainly provided by a fuel cell system, with any investigation to ensure stability and optimality across different driving
insufficiency covered by battery [10]. PFEMS usually results in high conditions.
hydrogen consumption because the fuel cell system generally operates One limitation in existing literature is the use of short-distance
outside the high-efficiency region. Additionally, PFEMS causes signifi driving cycles (typically under 30 km [37]) to develop and test EMS
cant load changes, which accelerate the degradation of fuel cell stack. strategies. This narrow focus on short distances leads to underutilization
The thermostat EMS (THEMS) controls the fuel cell to operate at a of battery capacity, increasing hydrogen consumption and fuel cell
constant power point, and the power of fuel cell remains unchanged degradation. Additionally, frequent start-stop cycles during
unless the battery state-of-charge (SOC) approaches its allowed bounds short-distance driving contribute to fuel cell degradation. To effectively
[7]. The determination of the constant power is quite important, but the reduce running costs throughout the entire lifespan of FCHEVs, it is
optimal constant power depends on the driving cycle. The control rules essential to consider more practical driving conditions, including
generally do not change with driving cycles, leading to a poor vehicle commuter driving for family vehicles and long-distance driving for
energy-saving rate [11]. commercial vehicles such as taxis, buses, coaches, and trucks. This study
The optimization-based methods formulate the EMS as an optimi combines rule-based EMS and optimization-based EMS to create an
zation problem, where the objective is the energy consumption or online real-time EMS for minimizing the total running cost of FCHEVs.
combining with the degradation cost of fuel cell and battery [10]. Dy The proposed EMS integrates SOC-based rules into an instantaneous
namic programming (DP) can provide the global optimal solution of optimization model that considers hydrogen consumption, battery
EMS when the driving cycle is known [12,13]. DP breaks down the consumption, and the degradation of the fuel cell system and battery. It
decision process into stages and solves multiple sub-problems to find the effectively utilizes battery capacity, applies to practical FCHEV sce
optimal solutions. It works recursively from the end state to the initial narios, and reduces running costs throughout the vehicle’s lifespan. The
state, resulting in a set of optimal solutions for each sub-problem that EMS demonstrates high computational efficiency and adaptability,
represents the global optimal solution [14]. However, DP is computa without the need for future vehicle speed predictions.
tionally intensive and is often used as a benchmark for other
optimization-based approaches. To improve the computational effi 2. Energy consumption model of FCHEV
ciency, some local optimization strategies have been proposed to
approximate the globally optimal solution, such as equivalent con This section focuses on modeling the energy consumption of
sumption minimization strategy (ECMS) [15,16], model predict control FCHEVs. It begins by introducing the powertrain configuration exam
(MPC)-based strategies [17–19], the Pontryagin’s minimum principle ined in the paper. Next, the modeling of key powertrain components,
(PMP)-based strategies [20–23], and others [24–26]. such as the motor, battery, and PEMFC system, is discussed. Lastly, an
The ECMS aims to minimize the combined hydrogen and battery equivalent hydrogen consumption model is developed, which involves
consumption in FCHEVs. It uses an Equivalent Factor (EF) to represent proposing the equivalent efficiency of hydrogen energy transformation
the overall fuel economy by unifying hydrogen and electric energy to battery energy.
consumption. The EF greatly impacts the auxiliary energy source’s
hydrogen consumption and the total hydrogen consumption over time.
Many researchers have focused on developing methods to determine the 2.1. Powertrain configuration of FCHEV
EF [14]. Offline optimization can provide EF values, but they may not
guarantee optimal fuel economy in different operating conditions [27]. The FCHEV is considered as a midsize sedan, and its powertrain
To address this, adaptive ECMS approaches adjust the EF in real-time configuration is depicted in Fig. 1. The powertrain components include a
[15,28–30]. The External Energy Maximum Strategy (EEMS) [31] and hydrogen tank, PEMFC system, DC/DC converter, battery, DC bus, DC/
its improved algorithm, OFA/DP-based EEMS [32], aim to maximize AC inverter, driving motor, and transmission system [3]. The DC/DC
external energy while minimizing hydrogen consumption by increasing converter is connected to the output of the PEMFC system to regulate
battery power demand. These methods use empirical formulas to obtain fuel cell current to the DC bus, while the battery is directly connected to
the cost function without calculating battery equivalent consumption. the DC bus. The equation representing the driving force of the vehicle at
However, they do not consider the degradation cost of power sources, a given speed and acceleration is expressed as Eq. (1):
focusing solely on the cost of hydrogen and battery consumption. 1
The MPC method predicts future vehicular power requests within a F(v, v̇) = mV g sin φ + ρAV Cd v2 + mV gCt cos φ + mV v̇ (1)
2
specific time horizon and optimizes an objective function over that
period. It utilizes a genetic algorithm optimized back-propagation neu where F is the driving force, v and v̇ are the vehicle speed and acceler
ral network predictor to estimate speed profiles. MPC is then employed ation, respectively. The right-hand side of the equation consists of four
to determine the optimal power allocation decisions for a FCHEV at each terms that correspond to the road incline resistance, aerodynamic drag,
receding horizon [33]. Hu et al. [34] develop a model predictive control tire rolling resistance, and vehicle acceleration. the mass of vehicle is
framework to minimize the overall operating cost of a hybrid electric denoted by mV, while g presents the gravity acceleration. φ denotes the
bus equipped with a fuel cell and a battery. This includes costs related to road incline angle, ρ represents the air density, AV is the windward area
hydrogen consumption as well as degradation of the fuel cell and bat of vehicle, Cd denotes the drag coefficient, and Ct represents the tire
tery. The framework assumes that the bus’s velocity trajectory within rolling friction coefficient. The parameter values can be found in
the prediction horizon is precisely known. Building upon this frame Table 1. Assuming the transmission efficiency is ηt, the required me
work, Zhou et al. [3] propose a real-time energy management strategy to chanical power of vehicle can be expressed by
reduce the vehicle’s operating cost using the MPC method. They utilize
F(v, v̇)v
an online-learning enhanced Markov chain to predict the vehicle speed, Pd = (2)
ηt
making their approach more practical and applicable. Similar works can
also be found in Refs. [35,36]. It can be found that the adaptive EMS can Here, we assume that the brake force is full provided by motor, i.e.
be summarized as two types: power allocating strategies adjusted based the braking energy is fully recycled by the motor if the required braking
on predicted vehicle power demand or through feedback of battery state torque is lower than the motor torque.
2
J. Wu et al. Energy 285 (2023) 129442
3
J. Wu et al. Energy 285 (2023) 129442
2.3. Equivalent hydrogen consumption model equivalent hydrogen consumption can be calculated using the following
equation:
Since both the PEMFCS and battery contribute to powering the ∫t ∫t
Pfc (t) Pb (t)
vehicle, it is important to compute the equivalent hydrogen consump Heq = dt + dt (7)
0 ηfc (t)LHVH 0 ηH− b (t)LHVH
tion of the battery energy in order to accurately describe the energy
consumption of the powertrain using the same evaluation criteria. The
4
J. Wu et al. Energy 285 (2023) 129442
where the first term on the right side is the actual hydrogen consumption 3. Cost model of FCHEV powertrain
of PEMFCS, while the second term is the additional hydrogen con
sumption induced by battery. The LHVH denotes the lower heating value In addition to the equivalent hydrogen consumption, the overall
of hydrogen, which is 120000 J/g. The symbols ηH-b denotes the running cost of the FCHEV powertrain includes the degradation cost of
equivalent battery efficiency, which reflects the conversion of hydrogen the PEMFCS and battery. Inappropriate operating conditions, such as
energy to battery electrical energy and is influenced by the power flow frequent start-stop cycles, high loadings, low loadings, load transitions
within the powertrain. The expression for ηH-b will be derived based on of the PEMFCS, and excessive current flow through the battery, can
the power flow of the powertrain, as shown in Fig. 5. accelerate the degradation of both components. To address this, a
The power from PEMFCS flows to the battery and/or motor, while degradation model for the PEMFCS and battery is established. This
the power flows between battery and motor bidirectionally, and the degradation model is then integrated with the equivalent hydrogen
detailed power flow can be divided into three cases. In the first case, consumption model to develop a comprehensive running cost model for
when the vehicle required power is greater than the power of PEMFCS, i. the FCHEV powertrain.
e. Pm > Pfc ηD/D ηD/A > 0, both PEMFCS and battery supply power to
motor simultaneously. In the second case, when the vehicle required 3.1. Degradation model of PEMFCS
power is greater than zero but less than the power output of PEMFCS, i.e.
Pfc ηD/D ηD/A > Pm > 0, the PEMFCS provides power to both the battery The main factors accelerating the degradation of PEMFCS are the
and the motor simultaneously. In the last case, when the vehicle start-stop cycling, high loadings, low loadings, and load transitions [43].
required power is equal to or less than zero while the power output of The start-stop cycling has a dominated negative impact on the duration
PEMFC is equal to or greater than zero, i.e. Pfc > 0 > Pm , both PEMFCS of FCHEV. Additionally, the PEMFCS also degrades when it works in the
and motor supply power to charge battery. Consequently, the equivalent high efficiency region, at a slower rate compared to low and high
power flows from hydrogen to the battery (PH-b) and the motor (PH-m) loadings [36]. The voltage degradation of a single fuel cell can be
can be expressed by following equations. determined by following equation, assuming that all degradation factors
are independent of each other [36], i.e.
⎧ ⎧
⎪ − Pb ⎪ Pb Pfc
⎪
⎪ , P > Pfc ηD/D ηD/A > 0 ⎪
⎪ + , P > Pfc ηD/D ηD/A > 0
⎪ ηH− b m
⎪ ⎪ ηH− b ηfc m
⎪
⎪
⎪ ⎪
⎪
⎪
⎨ Pfc ⎪
⎨
Pm Pm
PH b = η − η η η , Pfc ηD/D ηD/A ≥ Pm ≥ 0 , PH m= , Pfc ηD/D ηD/A ≥ Pm ≥ 0 (8)
⎪
⎪
⎪ fc fc D/D D/A ⎪ fc D/D ηD/A
⎪
⎪
η η
⎪
⎪ ⎪
⎪
⎪
⎪ P fc P m ⎪
⎪ Pm
⎪
⎩ − , Pfc > 0 > Pm ⎪
⎩ , Pfc > 0 > Pm
ηfc ηH− m ηH− m
where ηH-b and ηH-m denote the equivalent energy efficiency of hydrogen ΔUfc = αon− off Ncycle + αhigh Thigh + αlow Tlow + αhigh eff Thigh eff + αshift Nshift
transforming to the battery and the motor, and they are defined as the (12)
ratio between the received energy of the component and the consumed
energy of hydrogen by PEMFCS. Therefore, they are expressed by where ΔUfc is the voltage degradation in μV, αon-off denotes the cell
∫t ∫t voltage drop rate per on-off cycling, αhigh is the cell voltage degradation
Eb0 + 0 − Pb dt Pm dt
ηH− b = ∫t , ηH− m = ∫ t 0 (9) rate per hour at high loadings which are defined as when the power of
Eb1 + 0 PH− b dt P
0 H− m
dt PEMFC is higher than 80 % of its maximum power, αlow is the cell voltage
degradation rate per hour at low loadings which are defined as when the
where the Eb0 denotes the initial available energy of battery, and Eb1 is power of PEMFCS is lower than 20 % of its maximum power, αhigh_eff is
the maximum available energy of battery, expressed as the cell voltage degradation rate per hour at the high efficiency region
Eb0 = SOC0 Uoc,0 Qmax , Eb1 = Uoc,1 Qmax (10) that is between 20 % and 80 % maximum power of PEMFCS, and αshift
denotes the cell voltage degrading rate of load transitions. Ncycle, Thigh,
where Uoc,0 and Uoc,1 denote the battery voltage when battery SOC in Tlow, and Nshift denote the number of start-stop cycles, the duration of
initial value and the highest value, respectively. Substituting Eq. (8) into high loadings and low loadings in hour, and the amount of load shifts in
Eq. (9), the ηH-b(t) and ηH-m(t) can be calculated iteratively, but it is kW. The degrading rates are provided in Table 2 [36]. Typically, 10 %
computationally expensive. To reduce computational costs, the upper voltage drop of a fuel cell at rated current density is deemed as the
limit of the integral in Eq. (9) is changed to the previous time (t− ), which failure threshold [43]. Using ΔUEOL,fc denotes the voltage drop until the
has minimal impact on accuracy but eliminates the need for iteration. In end of life, the state-of-health (SOH) of PEMFCS can be expressed by
this case, the upper limit of the integral in Eq. (9) becomes (t− ), and all
ΔUfc
the integrands between the lower and upper limits of the integral have SOHfc = 1 − (13)
ΔUEOL,fc
been determined at the previous time. As a result, the right side of Eq. (9)
no longer includes ηH-b(t) and ηH-m(t), effectively eliminating the need for Based on the Ref. [44], ΔUEOL,fc is set as 60000 μV in this paper.
iteration. Consequently, the equivalent energy efficiency of the battery
and motor can be computed using the following equations.
∫ t− ∫ t−
Eb0 + 0 − Pb dt Pm dt Table 2
ηH− b = ∫ t− , ηH− m = ∫ t− 0 (11)
Eb1 + 0 PH− b dt PH− m dt Parameters of the PEMFCS degradation model.
0
αon-off (μV/cycle) αhigh (μV/h) αlow (μV/h) αhigh_eff (μV/h) αshift (μV/kW)
13.79 11.67 9.42 4.881 0.0423
5
J. Wu et al. Energy 285 (2023) 129442
6
J. Wu et al. Energy 285 (2023) 129442
Table 4
Shift points of battery SOC in REMS.
Parameters SOCL1 SOCL2 SOCL3 SOCL4 SOCH1 SOCH2 SOCH3 SOCH4
7
J. Wu et al. Energy 285 (2023) 129442
where a is the slope of the linear function. At the same time, both two γH Pfc γ H Pb βH βL γfc Pfc,max ΔUfc γ Eb |icell |
coefficients are limited to change between 0.01 and 1. Considering the min ΔCeq = + + + b
Pfc βH ηfc LHVH βL ηH− b LHVH ΔUEOL,fc 2NEOL Qcell
lower and upper bounds of SOC are set as SOCL = 0.2, and SOCH = 0.8,
s.t. if SOC ≥ 0.8, then Pfc = 0 kW
we set a = 5 in this paper, then the plot of the two coefficients is shown
as Fig. 8. if 0.7 ≤ SOC < 0.8, then Pfc = 0 kW or 6 kW;
It can be found that when the battery SOC closes to the allowed upper if 0.2 < SOC < 0.7, then Pfc = 0 kW, or 6 kW ≤ Pfc ≤ 12 kW;
bound 0.8, βH become smaller, increasing the cost of PEMFCS and
indicating that the battery is expected to provide more power. if SOC ≤ 0.2, then 12 kW ≤ Pfc ≤ 30 kW;
Conversely, βL increases the cost of the battery when the SOC ap βL = a(SOC − SOCL ), βH = a(SOCH − SOC)
proaches the lower bound of 0.2, encouraging the PEMFCS to provide (24)
more power. When the battery SOC is between 0.4 and 0.6, the two
adjusting coefficients remain at 1, indicating that they do not affect the Compared to the optimization model of OEMS shown in Eq. (22) the
cost weight within this interval. During each time step, the degradation optimization model of ROCEMS can be thought as the original optimi
cost of the PEMFCS caused by changes in load is typically larger than the zation model of OEMS with more constraints, and the constraints are the
other terms in the optimization model. This leads the model to strive for battery SOC-based rules. The same enumerating method is used to solve
a stable power output from the PEMFCS. However, to maintain the the optimization model, which can satisfy the requirement of real-time
battery SOC within the allowed range, the power output of the PEMFCS control.
needs to be adjusted. Consequently, the third term in Eq. (19) is also
multiplied by the two coefficients (βH and βL), which weakens its impact 5. Results and discussion
on load changes and effectively manages the battery SOC. In summary,
the two adjusting coefficients are utilized to control the battery SOC, and This section presents the performance of the proposed three EMSs
they only come into play when the battery SOC is close to the allowed under different driving conditions, which are categorized into short
bounds. distance driving cycles, long distance driving cycles consisting of mul
Due to the PEMFCS is low efficient when its output power is lower tiple repeated driving cycles, and commuter driving cycles based on
than the 20 % of its maximum power, and it also has high degrading rate standard driving cycles with several hours of stops. For the short dis
in this region, so the PEMFCS should avoid to work in this region. tance driving cycles, the DP method is directly utilized to generate the
Finally, the REMS can be expressed by following optimization model. reference solution. However, for the long distance driving cycles and
commuter driving cycles, the computational cost of applying DP directly
min ΔCeq =
γH Pfc
+
γ H Pb
+
βH βL γ fc Pfc,max ΔUfc γ Eb |icell |
+ b to solve such lengthy driving cycles is prohibitive. Hence, the DP method
Pfc βH ηfc LHVH βL ηH− b LHVH ΔUEOL,fc 2NEOL Qcell is employed to determine the optimal solution for a single driving cycle,
s.t. Pfc = 0 kW, or 6kW ≤ Pfc ≤ Pfc, max ; which is then repeated for each driving cycle in the sequence.
8
J. Wu et al. Energy 285 (2023) 129442
battery SOC profile of ROCEMS resembles that of REMS, but the former allows the PEMFCS to operate at 30 kW, which is a region of low effi
results in a lower minimum battery SOC value before returning to 0.5 at ciency and high degradation. On the other hand, the ROCEMS achieves
the end. In the case of DP-ROCEMS, the power profile of the PEMFCS is the lowest equivalent cost among the three proposed strategies by
similar to that of DP-REMS, but DP-ROCEMS shuts down the PEMFCS maximizing the PEMFCS’s operating time in the highest efficiency
earlier, aiming to achieve a final battery SOC of 0.5. range. The DP-REMS and DP-ROCEMS show slightly better performance
In terms of running cost, the OEMS stands out with significantly than the REMS and ROCEMS, respectively, but the difference is
higher costs compared to the other strategies. This is because the OEMS negligible.
9
J. Wu et al. Energy 285 (2023) 129442
Fig. 10. Speed and power profile for the integrated long driving cycle.
10
J. Wu et al. Energy 285 (2023) 129442
However, the single driving cycle of DP took over 17 h on a workstation 5.3. Commuter driving cycles
with a 2.5 GHz CPU and 128 GB RAM using the same software. This
highlights the impracticality of using DP directly to design the allocating To simulate the commuter driving cycles, we assume that the user
strategy for the entire 10 repeated driving cycles. drives the vehicle from home to the office at 7 a.m. and returns home at
6 p.m. on working days. Each single trip is represented by a standard
driving cycle, with the NEDC driving cycle being used as an example
11
J. Wu et al. Energy 285 (2023) 129442
Table 6 without the PEMFCS starting, the vehicle relies solely on the battery,
Detailed running cost for the long distance driving cycles. resulting in a rapid decrease in battery SOC. Conversely, when the
FC-C B–C FC-D B-D Ceq PEMFCS starts during a trip, the battery SOC increases rapidly.
(USD) (USD) (USD) (USD) (USD) The DP is employed to generate the reference solution using two
REMS 27.82 − 0.65 2.99 3.05 33.21 different strategies. The first strategy, termed DP-S, produces the
OEMS 29.16 − 0.68 4.35 3.31 36.13 reference solution for a single trip using DP and then repeats this solu
ROCEMS 27.68 − 0.49 2.97 3.06 33.21 tion for each trip. The second strategy, DP-ROCEMS, utilizes DP to
DP- 27.89 − 0.51 2.97 3.00 33.35 produce the solution only for the trip in which the PEMFCS starts under
ROCEMS
the control of ROCEMS. Hence, the PEMFCS operation in DP-ROCEMS
aligns with that of ROCEMS. Since the PEMFCS starts in the 5th trip
here. The simulation duration is set as one week, consisting of 5 working for all EMSs, the output power of the PEMFCS and battery SOC for this
days. During this time, there are 10 standard driving cycles with several trip are plotted in Fig. 13.
hours of downtime between each consecutive driving cycle. In the simulation results, it is evident that for DP-S and DP-ROCEMS,
The simulation results of commuter NEDC driving cycles for a week the PEMFCS starts at the beginning of the trip. However, DP-S operates
are presented in Fig. 12, with the x-axis representing the days. It can be the PEMFCS for a shorter period compared to DP-ROCEMS. This is
observed that not every trip requires the PEMFCS to start. In the case of because DP-S aims to achieve a final battery SOC equal to the initial
REMS, the PEMFCS starts 5 times, while for OEMS and ROCEMS, it only value of 0.5, whereas DP-ROCEMS allows the battery SOC to increase
starts 3 times. This means that REMS activates the PEMFCS every 2 trips, from 0.27 to 0.72, requiring a longer operation time for the PEMFCS.
whereas OEMS and ROCEMS activate it every 3 trips. During the trips Regarding the proposed EMS strategies, they start the PEMFCS during
12
J. Wu et al. Energy 285 (2023) 129442
Fig. 13. Speed profile and PEMFCS power during a single HWFET trip.
terms of running cost, with a total cost of 8.77 USDs. The PEMFCS
Table 7
degradation cost is noticeably reduced compared to the previous two
Detailed running cost for the commuter driving cycles.
EMSs, as the PEMFCS is only started three times. However, the equiv
FC-C B–C FC-D B-D Ceq alent hydrogen consumption cost is the highest at 5.86 USDs due to the
(USD) (USD) (USD) (USD) (USD)
significant amount of low-efficiency operating points for the PEMFCS.
REMS 5.51 − 0.18 3.22 1.05 9.60 Fourthly, ROCEMS further reduces the total running cost to 7.91 USDs
OEMS 5.45 0.41 2.10 0.81 8.77
by ensuring that the PEMFCS operates in the high-efficiency region for a
ROCEMS 4.74 0.35 1.96 0.86 7.91
DP-S 5.08 0.00 5.96 0.75 11.80 longer duration, effectively reducing both the equivalent hydrogen
DP- 4.53 0.52 1.90 0.69 7.65 consumption cost and PEMFCS degradation cost. Finally, DP-ROCEMS
ROCEMS achieves the lowest running cost (7.65 USDs), with slightly lower
degradation costs for both the PEMFCS and battery. However, the
improvement compared to ROCEMS is not significant. It’s important to
the latter stage of the trip and continue its operation until the battery
note that DP-ROCEMS utilizes information generated by ROCEMS,
SOC reaches the upper limit of 0.8. In this stage, the OEMS predomi
particularly regarding which trips require the PEMFCS to start.
nantly operates the PEMFCS at its maximum power, allowing for faster
ROCEMS, on the other hand, does not rely on any predicted information
battery charging and earlier shutdown. The PEMFCS in the ROCEMS
for individual trips, making it highly applicable. Overall, the number of
works at the highest efficiency point of 12 kW for most of the time,
start-stop cycles for the PEMFCS has the greatest impact on the total
leading to a reduced hydrogen consumption cost. In contrast, the REMS
running cost for commuter driving cycles. Therefore, reducing the start-
keeps the PEMFCS working at 6 kW, resulting in a longer charging time
stop cycles of the PEMFCS becomes the key factor in minimizing the
for the battery.
total running cost for such driving scenarios.
Table 7 provides a detailed breakdown of the running costs for each
Based on the results obtained from the analysis of short distance
EMS. From the analysis of Fig. 12 and Table 7, several observations can
driving cycles, long distance driving cycles, and commuter driving cy
be made.
cles, we have summarized the performance of the proposed three EMSs
Firstly, DP-S has the highest running cost, primarily due to the
in Table 8. The table indicates that the ROCEMS demonstrates the best
frequent start-stop cycles of the PEMFCS, as it needs to be started in each
adaptation to different driving conditions among the three online EMSs.
trip. Although DP-S has a low equivalent hydrogen consumption cost
(5.08 USDs), it results in the highest fuel cell degradation cost (5.96
USDs). Secondly, REMS has the second-highest running cost, with
Table 8
relatively high costs in each term. This includes a significant equivalent Comparison of the proposed EMSs.
hydrogen consumption cost (5.33 USDs), PEMFCS degradation cost
Short distance Long distance Commuter Comprehensive
(3.22 USDs), and battery degradation cost (1.05 USDs). These costs are
mainly influenced by the higher number of start-stop cycles of the REMS ** *** * **
PEMFCS and increased battery usage. Thirdly, OEMS ranks third in OEMS * * ** *
ROCEMS *** *** *** ***
13
J. Wu et al. Energy 285 (2023) 129442
The REMS shows better performance for continuous driving cycles References
compared to the OEMS, while the OEMS performs better for commuter
driving cycles. The ROCEMS effectively reduces the total running cost [1] Wang Y, Pang Y, Xu H, Martinez A, Chen KS. PEM Fuel cell and electrolysis cell
technologies and hydrogen infrastructure development – a review. Energy Environ
for the entire life cycle of FCHEVs and achieves performance that closely Sci 2022;15(6):2288–328.
resembles the offline DP-based EMS. [2] Lü X, Wu Y, Lian J, Zhang Y, Chen C, Wang P, et al. Energy management of hybrid
electric vehicles: a review of energy optimization of fuel cell hybrid power system
based on genetic algorithm. Energy Convers Manag 2020;205:112474.
6. Conclusions [3] Zhou Y, Ravey A, Péra M-C. Real-time cost-minimization power-allocating strategy
via model predictive control for fuel cell hybrid electric vehicles. Energy Convers
This paper proposes three online real-time EMSs to minimize the Manag 2021;229:113721.
[4] Fletcher T, Thring R, Watkinson M. An Energy Management Strategy to
total running cost of FCHEVs by considering the degradation of PEMFCS concurrently optimise fuel consumption & PEM fuel cell lifetime in a hybrid
and battery. REMS uses a gear shift strategy based on battery SOC to vehicle. Int J Hydrogen Energy 2016;41(46):21503–15.
reduce costs. OEMS optimizes instantaneous costs by adjusting weight [5] Chen H, Liu Z, Ye X, Yi L, Xu S, Zhang T. Air flow and pressure optimization for air
supply in proton exchange membrane fuel cell system. Energy 2022;238:121949.
ings with battery SOC-based coefficients. ROCEMS combines features of
[6] Liu Z, Chen H, Peng L, Ye X, Xu S, Zhang T. Feedforward-decoupled closed-loop
REMS and OEMS to prioritize objectives. The proposed EMSs were fuzzy proportion-integral-derivative control of air supply system of proton
evaluated under different driving conditions: short distance driving exchange membrane fuel cell. Energy 2022;240:122490.
cycles, long distance driving cycles, and commuter driving cycles. [7] Song K, Li F, Hu X, He L, Niu W, Lu S, et al. Multi-mode energy management
strategy for fuel cell electric vehicles based on driving pattern identification using
For short and long distance driving cycles, REMS and ROCEMS learning vector quantization neural network algorithm. J Power Sources 2018;389:
performed similarly to the offline dynamic programming (DP) solution 230–9.
in terms of running cost. However, OEMS had significantly higher [8] Gao J, Sun F, He H, Zhu G, Strangas E. A comparative study of supervisory control
strategies for a series hybrid electric vehicle. In: 2009 Asia-Pacific power and
running costs due to operating the PEMFCS in the low efficiency region energy Engineering Conference. Wuhan; 2009. China.
for extended periods and generating more start-stop cycles. In the case of [9] Li D, Feng D. Thermostatic control for series hydraulic hybrid vehicle (SHHV)
commuter driving cycles, REMS had the highest running cost, while energy management. Adv Mater Res 2012;512:2676–81.
[10] Song K, Chen H, Wen P, Zhang T, Zhang B, Zhang T. A comprehensive evaluation
ROCEMS had the lowest running cost among the three strategies. REMS framework to evaluate energy management strategies of fuel cell electric vehicles.
had more start-stop cycles of the PEMFCS, leading to higher PEMFCS Electrochim Acta 2018;292:960–73.
degradation costs compared to the other strategies. To minimize running [11] Guo J, He H, Li J, Liu Q. Real-time energy management of fuel cell hybrid electric
buses: fuel cell engines friendly intersection speed planning. Energy 2021;226:
costs in continuous driving cycles, improving the efficiency of PEMFCS 120440.
working points is crucial. For commuter driving cycles, reducing start- [12] Zhou W, Yang L, Cai Y, Ying T. Dynamic programming for new energy vehicles
stop cycles of the PEMFCS is key. ROCEMS outperformed the other based on their work modes Part II: fuel cell electric vehicles. J Power Sources 2018;
407:92–104.
strategies by improving the average working efficiency of the PEMFCS
[13] Ansarey M, Shariat Panahi M, Ziaratic H, Mahjoob M. Optimal energy management
and reducing start-stop cycles for different driving conditions, making it in a dual-storage fuel-cell hybrid vehicle using multi-dimensional dynamic
the best-performing strategy overall. programming. J Power Sources 2014;250:359–71.
The conclusions of the paper suggest that reducing start-stop cycles [14] Zhao X, Wang L, Zhou Y, Pan B, Wang R, Wang L, et al. Energy management
strategies for fuel cell hybrid electric vehicles: classification, comparison, and
of PEMFCS and utilizing battery capacity effectively can help minimize outlook. Energy Convers Manag 2022;270:116179.
running costs. However, these findings are based on specific power sizes [15] Zeng T, Zhang C, Zhang Y, Deng C, Hao D, Zhu Z, et al. Optimization-oriented
of PEMFCS and batteries used in the study. Changes in power sizes, adaptive equivalent consumption minimization strategy based on short-term
demand power prediction for fuel cell hybrid vehicle. Energy 2021;227:120305.
battery capacity, and economic factors may require reevaluation and [16] Skugor B, Deur J, Cipek M, Pavkovi D. Design of a power-split hybrid electric
adjustment of optimal control parameters. Therefore, while the con vehicle control system utilizing a rule-based controller and an equivalent
clusions are valuable, they should be applied with consideration of consumption minimization strategy. Proc Inst Mech Eng - Part D J Automob Eng
2014;228(6):631–48.
varying system parameters and economic conditions. [17] Zhou Y, Li H, Ravey A, Péra M-C. An integrated predictive energy management for
light-duty range-extended plug-in fuel cell electric vehicle. J Power Sources 2020;
Credit author statement 451:227780.
[18] Hu D, Wang Y, Li J, Yang Q, Wang J. Investigation of optimal operating
temperature for the PEMFC and its tracking control for energy saving in vehicle
Jinglai Wu: Conceptualization, Methodology, Investigation, Soft applications. Energy Convers Manag 2021;249:114842.
ware, Writing. Yunqing Zhang: Conceptualization, review & editing. [19] Zhou Y, Ravey A, Péra M-C. Multi-mode predictive energy management for fuel
cell hybrid electric vehicles using Markov driving pattern recognizer. Appl Energy
Zhaowen Liang: Investigation, Visualization. Jiageng Ruan: Software,
2020;258:114057.
Validation, review & editing. Kai Liu: Data Curation, Visualization. The [20] Hemi H, Ghouili J, Cheriti A. Combination of Markov chain and optimal control
entire work is executed with the guidance of Jinglai Wu. solved by Pontryagin’s Minimum Principle for a fuel cell/supercapacitor vehicle.
Energy Convers Manag 2015;91:387–93.
[21] Odeim F, Roes J, Heinzel A. Power management optimization of a fuel cell/
Declaration of competing interest battery/supercapacitor hybrid system for Transit bus applications. IEEE Trans Veh
Technol 2016;65(7):5783–8.
The authors declare that they have no known competing financial [22] Song K, Wang X, Li F, Sorrentino M, Zheng B. Pontryagin’s minimum principle-
based real-time energy management strategy for fuel cell hybrid electric vehicle
interests or personal relationships that could have appeared to influence considering both fuel economy and power source durability. Energy 2020;205:
the work reported in this paper. 118064.
[23] Ou K, Yuan W-W, Kim Y-B. Development of optimal energy management for a
residential fuel cell hybrid power system with heat recovery. Energy 2021;219:
Data availability 119499.
[24] Hu X, Murgovski N, Johannesson L, Egardt B. Optimal Dimensioning and Power
Data will be made available on request. Management of a Fuel Cell/Battery Hybrid Bus via Convex Programming IEEE/
ASME TRANSACTIONS ON MECHATRONICS 2015;20.
[25] Elbert P, Widmer M, Gisler HJ, Onder C. Stochastic dynamic programming for the
Acknowledgement enregy management of a serial hybrid electric bus. Int J Veh Des 2015;65(7):
5783–8.
[26] Caux S, Gaoua Y, Lopez P. A combinatorial optimisation approach to energy
This research is supported in part by National Science Foundation of
management strategy for a hybrid fuel cell vehicle. Energy 2017;133:219–30.
China (12272142), and Fundamental Research Funds for the Central [27] Hong Z, Li Q, Han Y, Shang W, Zhu Y, Chen W. An energy management strategy
Universities (2172021XXJS048). based on dynamic power factor for fuel cell/battery hybrid locomotive. Int J
Hydrogen Energy 2018;43(6):3261–72.
[28] Gao H, Wang Z, Yin S, Lu J, Guo Z, Ma W. Adaptive real-time optimal energy
management strategy based on equivalent factors optimization for hybrid fuel cell
system. Int J Hydrogen Energy 2021;46(5):4329–38.
14
J. Wu et al. Energy 285 (2023) 129442
[29] Li H, Ravey A, N’Diaye A, Djerdir A. Online adaptive equivalent consumption [38] ADVISOR. Advanced Vehicle Simulator. http://adv-vehicle-sim.sourceforge.net/.
minimization strategy for fuel cell hybrid electric vehicle considering power [39] Ettihir K, Boulon L, Agbossou K. Optimization-based energy management strategy
sources degradation. Energy Convers Manag 2019;192:133–49. for a fuel cell/battery hybrid power system. Appl Energy 2016;163:142–53.
[30] Wu J, Zhang N, Tan D, Chang J, Shi W. A robust online energy management [40] David Huang K, Tzeng S-C. A new parallel-type hybrid electric-vehicle. Appl
strategy for fuel cell/battery hybrid electric vehicles. Int J Hydrogen Energy 2020; Energy 2004;79(1):51–64.
45(27):14093–107. [41] Windarko NA. SOC estimation based on OCV for NiMH batteries using an improved
[31] Motapon SN, Dessaint L-A, Al-Haddad K. A robust H2-consumption-minimization- takacs model. Journal of Power Electronics 2010;10(2):181–6.
based energy management strategy for a fuel cell hybrid emergency power system [42] Bin Y, Li Y, Feng N. Nonlinear dynamic battery model with boundary and scanning
of more electric aircraft. IEEE Trans Ind Electron 2014;61(11):6148–56. hysteresis. Conference Nonlinear dynamic battery model with boundary and
[32] Quan R, Li Z, Liu P, Li Y, Chang Y, Yan H. Minimum hydrogen consumption-based scanning hysteresis, Hollywood, California, USA vol. DSCC2009-2745. p. 245-251.
energy management strategy for hybrid fuel cell unmanned aerial vehicles using [43] Chen H, Pei P, Song M. Lifetime prediction and the economic lifetime of Proton
direction prediction optimal foraging algorithm. Fuel Cell 2023;23(2):221–36. Exchange Membrane fuel cells. Appl Energy 2015;142:154–63.
[33] Liu Y, Li J, Chen Z, Qin D, Zhang Y. Research on a multi-objective hierarchical [44] Hua Z, Zheng Z, Péra M-C, Gao F. Remaining useful life prediction of PEMFC
prediction energy management strategy for range extended fuel cell vehicles. systems based on the multi-input echo state network. Appl Energy 2020;265:
J Power Sources 2019;429:55–66. 114791.
[34] Hu X, Zou C, Tang X, Liu T, Hu L. Cost-optimal energy management of hybrid [45] Bloom I, Cole BW, Sohn JJ, Jones SA, Polzin EG, Battaglia VS, et al. An accelerated
electric vehicles using fuel cell/battery health-aware predictive control. IEEE Trans calendar and cycle life study of Li-ion cells. J Power Sources 2001;101(2):238–47.
Power Electron 2020;35(1):382–92. [46] Wang J, Liu P, Hicks-Garner J, Sherman E, Soukiazian S, Verbrugge M, et al. Cycle-
[35] Quan S, Wang Y-X, Xiao X, He H, Sun F. Real-time energy management for fuel cell life model for graphite-LiFePO4 cells. J Power Sources 2011;196(8):3942–8.
electric vehicle using speed prediction-based model predictive control considering [47] Ebbesen S, Elbert P, Guzzella L. Battery state-of-health perceptive energy
performance degradation. Appl Energy 2021;304:117845. management for hybrid electric vehicles. IEEE Trans Veh Technol 2012;61(7):
[36] He H, Jia C, Li J. A new cost-minimizing power-allocating strategy for the hybrid 2893–900.
electric bus with fuel cell/battery health-aware control. Int J Hydrogen Energy [48] Global EV Outlook 2023. Catching up with climate ambitions. International Energy
2022;47(52):22147–64. Agency; 2023.
[37] https://www.car-engineer.com/the-different-driving-cycles/.
15