Content uploaded by Clemens Deutsch
Author content
All content in this area was uploaded by Clemens Deutsch on Feb 14, 2022
Content may be subject to copyright.
Energy Conversion and Management: X 14 (2022) 100193
Available online 9 February 2022
2590-1745/© 2022 The Author(s). Published by Elsevier Ltd. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).
Evaluation of energy management strategies for fuel cell/battery-powered
underwater vehicles against eld trial data
Clemens Deutsch
a
,
*
, Ariel Chiche
b
, Sriharsha Bhat
a
, Carina Lagergren
b
, G¨
oran Lindbergh
b
,
Jakob Kuttenkeuler
a
a
KTH Royal Institute of Technology, Teknikringen 8, 114 28 Stockholm, Sweden
b
KTH Royal Institute of Technology, Teknikringen 42, 114 28 Stockholm, Sweden
ARTICLE INFO
Keywords:
Autonomous underwater vehicle (AUV)
Fuel cell
Hybrid
Energy management strategies
ABSTRACT
This study combines high-delity simulation models with experimental power consumption data to evaluate the
performance of Energy Management Strategies (EMS) for fuel cell/battery hybrid Autonomous Underwater
Vehicles (AUV). The performance criteria are energy efciency, power reliability and system degradation. The
lack of standardized drive cycles is met by the cost-efcient solution of synthesizing power proles from sampled
AUV eld trial data. Three power proles are used to evaluate nite-state machine, fuzzy logic and two
optimization-based EMS. The results reveal that there is a trade-off between the objectives. The rigidity of the
EMS determines its load-following behavior and consequently the performance regarding the objectives. Rule-
based methods are particularly suitable to design energy-efcient operations, whereas optimization-based
methods can easily be tuned to provide power reliability through load-following behavior. Both classes of
EMS can be best-choice methods for different types of missions.
Introduction
Autonomous Underwater Vehicles (AUVs) are robotic sensor plat-
forms that are used in industrial, civil and military applications [64].
AUVs perform missions which include tasks such as bathymetric map-
ping, oceanographic studies [37], marine archaeology [44] or homeland
security and surveillance [17]. Typically, the vehicle’s payload sensor
suite comprises a wide range of acoustic sensors (e.g. sidescan sonar,
multibeam echo sounder, subbottom proler, acoustic Doppler current
proler) and oceanographic instruments such as CTDs (for the mea-
surement of conductivity, temperature and depth), uorometers, dis-
solved oxygen sensors or turbidity sensors [61].
In recent years, there has been a signicant interest in both the
operation of long-range AUVs [34,14,45,23] and the operation of AUVs
in extreme environments, e.g. under the Antarctic ice shelf
[25,35,63,29]. The high-risk nature of such operations leads to an in-
crease in requirements regarding AUV performance, especially vehicle
endurance and power reliability.
Today, most AUVs are powered by electrochemical batteries (pri-
mary and secondary) [6,15,53,21] whereas only few vehicles make use
of alternative energy systems such as wave energy converters [56,57]
saltwater batteries [19,20], nuclear technology [49,33], or fuel cells, in
particular proton-exchange membrane fuel cells (PEMFCs)
[24,36,5,18,22,10,60]. Fuel cell systems utilize the electrochemical re-
action of hydrogen and oxygen to provide electric power. In order to
operate these systems efciently, a controlled supply of the reactants is
necessary. In case of air-independent operation (e.g. underwater or in
space), external storage of both hydrogen and oxygen is required. There
are different storage solutions available, the most common being com-
pressed gas storage [1,43].
Commercial state-of-the-art AUVs have an endurance of >70 h [60].
Fuel cell/battery hybrid systems have been identied as being capable of
further increasing vehicle endurance [23,36,22,42,8]. These hybrid
systems combine the strengths of both power sources, i.e. high energy
density of fuel cells and power exibility of electrochemical batteries. As
the primary power source, the fuel cell system is used to supply the base
power while the battery system supplies the transient peak power de-
mands. During phases of lower power demand, the battery system is
charged by the surplus power from the fuel cell. The power split between
the power sources is determined by Energy Management Strategies
(EMS) [13,28]. These control strategies are designed specically to
provide energy-efcient operation, power reliability or reduction of
* Corresponding author.
E-mail address: clemensd@kth.se (C. Deutsch).
Contents lists available at ScienceDirect
Energy Conversion and Management: X
journal homepage: www.sciencedirect.com/journal/energy-conversion-and-management-x
https://doi.org/10.1016/j.ecmx.2022.100193
Energy Conversion and Management: X 14 (2022) 100193
2
system degradation [41].
In the automotive industry, fuel cell hybrid electric vehicles
(FCHEVs) have been researched for over a decade as an environmentally
friendly alternative to internal combustion engine (ICE) cars [48,54]. As
a result, there has been extensive research on EMS for automobiles
[13,40,27,54]. For automobiles, standard drive cycles [46,12] such as
the New European Drive Cycle (NEDC, 1997) are used for the evaluation
of EMS. Unfortunately, results from automotive research are not directly
transferable to underwater vehicles since load patterns and drive cycles
for AUVs are inherently different. Therefore, it is yet to be investigated
how standard EMS perform on realistic AUV drive cycles.
Sea trials with a Hugin 3000 AUV have been performed to gather
power consumption data during different maneuvers and modes of
operation. In this study, these data are used not only to directly evaluate
EMS performance, but also to generate further load proles consisting of
only specic maneuvers (e.g. lawn-mower patterns). These realistic load
proles are combined with a high-delity fuel cell/battery hybrid
simulation model that has been developed in Simulink®. The combi-
nation of a high-delity model with real power consumption data dis-
tinguishes the present study from our previous work [11] and marks a
novelty in the eld of AUV research.
The EMS performance is evaluated on the basis of the three key
parameters: Energy efciency (operation at maximum efciency),
power reliability (preservation of battery state of charge), and power
system degradation (reduction of fuel cell transients and critical battery
cycling).
Methods
A thorough analysis of the performance of EMS for underwater ve-
hicles requires high-delity electrical models of the power system
components, a representative selection of state-of-the-art EMS, and
evaluation against validated power consumption data from eld ex-
periments. In the following, the acquisition of power proles, modelling
of the hybrid fuel cell/battery system, and the implementation of the
considered Energy Management Strategies are described.
Experimental power data acquisition
The University of Gothenburg’s Hugin 3000 AUV (Fig. 1) is a battery-
powered, 7.5 m long and 1850 kg heavy AUV designed and manufac-
tured by Kongsberg, Norway. It is frequently used in the offshore oil &
gas industry and in the defense sector [9,16], but is also operated by
several research institutes [31,62]. The AUV was used to sample power
consumption data during eld trials on the Swedish west coast.
Providing a reference load prole, the tests were designed to provide
eld data on vehicle and power dynamics, which could subsequently be
used to derive new load proles from the recorded segments. The tests
comprised the performance of several of the following maneuvers at
different speeds:
•Lawnmower patterns
•Turning circles
•Zigzag maneuvers
•Diving sequences
Throughout the mission, the sensor suite (incl. CTD, multibeam
echosounder, sidescan sonar and subbottom proler) was fully opera-
tional. The two-dimensional trajectory of the eld trials is shown in
Fig. 2 for reference and the resulting electric load on the vehicle’s 52 V
Fig. 1. The Hugin 3000 AUV at the Kristineberg Marine Research Station at Gullmarsfjorden, Sweden. (Photo: Malin M¨
ork).
Fig. 2. Trajectory of the eld trials performed with the Hugin 3000 AUV at
Stora Born¨
o, Sweden, in June 2019.
C. Deutsch et al.
Energy Conversion and Management: X 14 (2022) 100193
3
DC bus – with labels indicating maneuvers and transit phases - is shown
in Fig. 3a.
In addition to serving as a reference load prole for the EMS per-
formance evaluation, the sampled power data is also used to generate
two further load proles. For this purpose, load segments are extracted
and joined systematically to generate the new load proles. These new
proles retain their natural dynamic characteristics and at the same time
can resemble standard missions. In this study, the Energy Management
Strategies are evaluated against the following three load proles:
Fig. 3. The reference load prole and load proles generated from bathymetric mapping maneuvers. (a) Mission A (reference prole), (b) Mission B (constant base
power), (c) Mission C (varying power levels).
C. Deutsch et al.
Energy Conversion and Management: X 14 (2022) 100193
4
1. Mission A is the base prole (Fig. 3a). It represents the potentially
most challenging load prole. A corresponding real-world mission is
characterized by high grades of autonomy such as tracking of mobile
targets.
2. Mission B represents a standard mission that includes e.g. homo-
geneous mapping of a safe area. It consists of a series of lawnmower
and square patterns at constant speed (Fig. 3b).
3. Mission C also represents bathymetric mapping missions. In contrast
to the second prole, this load prole contains mapping at different
speeds (Fig. 3c). A variation in speed can be required for increased
mapping resolution or navigating in unknown terrain.
Power system model in Simulink®
The load proles are used to simulate the instantaneous electric load
in the fuel cell/battery hybrid system model. The simulation model is
developed in Simulink®, enabling accurate modelling of the system’s
dynamic behavior. These dynamics are not only caused by the electro-
chemical reactions in both the fuel cells and the lithium-ion (Li-ion)
batteries, but also by the low-level controllers that regulate the re-
actants’ ow rates and the DC-DC power converters. The model is
largely based on an example for More Electric Aircrafts (MEA) [32]
provided by MathWorks®. The example model, which has previously
been used by other research groups for similar purposes [2,3], is
modied to model our research subject: a hybrid power system con-
sisting a fuel cell and a battery system, connected to a DC bus with a
variable DC load.
In this model, the fuel cell stack is modeled as a proton-exchange
membrane fuel cell (PEMFC) using a generic model for fuel cells
[51,50]. This stack model is readily available as a simulation block in
Simulink® and already contains parameters for a 1.26 (24 V) PEMFC,
fullling the power system sizing requirements for the load proles used
in this study (based on an estimate following the sizing methodology
presented in [7,8]). The fuel cell parameters are given in Table 1.
Since the load proles have been sampled with a battery-powered
AUV, they do not cover the additional power consumed by the fuel
cell system’s auxiliary components (balance of plant, BoP) [26].
Therefore, an additional parasitic load is superimposed on the DC load to
account for BoP efciency. Following the approach by [55], the parasitic
power consumption Pp is modelled as a linear function of the fuel cell
power density Vi as
Pp=a3+a4⋅Vi,(1)
where V is the cell voltage, i is the current density, and a3 and a4 are
constants. The values for a3 and a4 are taken as 0.6 W cm
−2
and 0.05,
respectively [52].
The battery is modeled as a lithium-ion battery using the generic
battery block of the Simscape Electrical Specialized Power Systems li-
brary [58,65,47,39]. The battery block parameters are given in Table 2.
Both power sources are connected to the DC bus via DC power
converters. The nominal bus voltage is 52 V. Since the fuel cell provides
only lower voltage, a boost converter is required to increase the voltage
to nominal bus voltage. The battery requires both a boost and a buck
converter to account for charging and discharging conditions. At this
point, the same DC-DC power converter topology is used as in the pro-
vided example [32]. The power converter parameters used in this study
are given in Table 3.
Fig. 4 shows the Simulink® model with the two power sources and
the load connected to the DC-DC bus.
Energy management strategies
In hybrid power systems, Energy Management Strategies determine
the power split between the power sources. These strategies can be
implemented in different ways, e.g. in terms of heuristics or as optimi-
zation problems. In this study, two rule-based and two optimization-
based EMS are investigated.
The EMS inputs describe the current state of the system; input pa-
rameters are the electric load, fuel cell power and battery state of charge
(SOC). The output from the EMS is a setpoint for the fuel cell power PFC.
The battery power PB is equal to the remaining power demand since
Preq(t) = PFC (t) + PB(t),(2)
where Preq is the power demand. The fuel cell power setpoint is bounded
by a minimum (PFCmin =378 kW) and a maximum (PFCmax =2000 kW)
value.
The EMS aims at maximizing the following objectives:
1. Energy efciency (operation of the system at its maximum efciency
point, MEP)
2. Power reliability (preservation of SOC)
3. System durability (avoiding system component degradation)
The latter two objectives are commonly implemented as soft con-
strains or operating limits. In this study, the degradation of the fuel cell
stack is minimized by limiting the power rate to one-sixtieth of the
nominal fuel cell power PFCnom [38,59]:
−PFCnom
60s⩽
∂
PFC
∂
tmax
⩽PFCnom
60s.
(3)
In a similar manner, battery degradation is reduced by cycling between
35
Table 1
Block Parameters: Fuel Cell Stack.
Parameters Value
Preset model PEMFC – 1.26 kW – 24 Vdc
Model detail level Detailed
Signal variations Fuel ow rate
Air ow rate
Response time (s) 1
Peak O2 utilization (%) 60
Voltage undershoot (V) 2
@ peak O2 utilization
Table 2
Block Parameters: Battery.
Parameters Value
Type Lithium-ion
Simulate temperature effects No
Simulate aging effects No
Nominal voltage (V) 48
Rated capacity (Ah) 52
Initial state-of-charge (%) 65
Battery response time (s) 20
Table 3
DC-DC Power Converter Parameters.
Parameters Value
Inductance (L) 500×10−6H
Capacitance (C) 5000×10−6F
Resistance (R) 0.01Ω
Load capacitance 15.6F
FC converter efciency
- at full load 85%
- at partial load 90%
Battery converter efciency
- at full load 80%
- at partial load 88%
C. Deutsch et al.
Energy Conversion and Management: X 14 (2022) 100193
5
SOCmin =0.35,(4)
SOCmax =0.65.(5)
The lower limit of 35% is also assumed to be sufciently high to guar-
antee power reliability, providing enough charge to meet sudden and
long-lasting power surges. The following four EMS are implemented in
the simulation model:
Rule-based EMS – deterministic
The deterministic EMS is a nite-state machine (SMEMS) and
therefore the simplest of the used EMS. The fuel cell power setpoint is
determined from the battery SOC and the corresponding battery power
can be determined from (2). The heuristics are embedded in a look-up
table (Table 4).
Rule-based EMS – fuzzy logic
The fuzzy logic EMS (FZYEMS), based on a Mamdani fuzzy inference
system [30], is designed similarly to the deterministic EMS. Applying
fuzzy logic has the advantage of avoiding sudden case switching,
resulting in smoother transitions between operating states. This is due to
the innite-valued logic native to fuzzy logic. Like the deterministic
EMS, the fuzzy logic controller uses the battery SOC as the input. The
resulting transfer function is shown in Fig. 5.
Optimization-based EMS
Optimization-based EMS minimize a cost function that drives the
system to be operated at an optimal point while satisfying specic
constraints. It is important to note that the optimum is specic to the cost
function. In general, cost functions use weights to penalize the use of
either of the power systems when the current state of the system is close
to an unfavourable operating point. Example: When battery SOC ap-
proaches a lower limit, usage of battery must be heavily penalized.
The rst optimization-based EMS (OEMS1) poses a linear optimisa-
tion problem with constraints (linear program):
Fig. 4. The fuel cell/battery hybrid system model in Simulink®.
Table 4
A deterministic Energy Management Strategy (EMS) based on battery state-of-
charge.
SOC⩾SOCmax SOCmin <SOC <SOCmax SOC⩽SOCmin
PFCmin PFCnom PFCmax
Fig. 5. Single input–single output transfer functions (fuzzy logic inference and
deterministic state machine).
C. Deutsch et al.
Energy Conversion and Management: X 14 (2022) 100193
6
min
PFC,PB
ω
1PFC +
ω
2PB
s.tPreq =PFC +PB
PFCmin⩽PFC ⩽PFCmax
−PFCnom
60s⩽
∂
PFC
∂
tmax
⩽PFCnom
60s
(6)
The coefcients
ω
1 and
ω
2 are the penalizing weights for fuel cell
power and battery power, respectively, and stem from the work of Wang
et al. [59]. These weights are given as
ω
1=1+2||P∗FC −P∗FCMEP||
P∗FCmax −P∗FCmin 2
,P∗FCmin⩽P∗FC ⩽P∗FCmax
1+2||P∗FC −P∗FCMEP||
P∗FCmax −P∗FCmin 4
,P∗FCmax <P∗FC <P∗FCmin
,(7)
ω
2=1−2SOC −SOCini
SOCmax −SOCmin2
,SOCmin⩽SOC⩽SOCmax
1−2SOC −SOCini
SOCmax −SOCmin4
,SOCmax <SOC <SOCmin
,(8)
where the superscripted asterisk x* denotes normalization by the nom-
inal fuel cell power PFCnom and ||x|| denotes the L2 norm.
The second optimization-based EMS (OEMS2) has previously been
published as part of our previous work [11] and uses a non-linear cost
function based on the idea of minimizing the distance to desired states, i.
e. the distance of the fuel cell power output to the nominal operating
point and the distance of the battery SOC to a desired SOC, denoted by
SOCdes. In order to implement SOCdes into the cost function, an estimator
for the future battery SOC based on the instantaneous battery power is
required. Such an estimator can utilize a feasible nite-time horizon Δt,
assuming constant battery power. In discrete time, this estimator is
expressed as
SOC(k+1) = SOC(k) − Δt*
Eb
PB(9)
A feasible nite-time horizon for this study is Δt=300s, but this
may vary from application to application. The minimization problem
then becomes
min
PFC,PB
K1P†
FCnom −P†
FC2+K2SOCdes −SOC +Δt
EB
PB2
s.tPreq =PFC +PB
PFCmin⩽PFC ⩽PFCmax
−PFCnom
60s⩽
∂
PFC
∂
tmax
⩽PFCnom
60s
(10)
where the superscripted dagger x†denotes normalization by the
maximum fuel cell power and K1,K2 are gains similar to the weights in
the previous optimization problem. These gains need to be chosen such
that the optimization problem is scaled to suitable dimensions for the
solver. A feasible choice for the gains is represented by K1=10,K2=
1000. Similarly to the role of gains in feedback controllers, proper
tuning of gains can improve the controller’s performance.
In Simulink®, the nonlinear programming solver fmincon with the
sequential quadratic programming (sqp) algorithm is used to solve the
optimisation problems.
Results
In order to evaluate their performance, the EMS are tested against
the load proles from Sec. 2.1 using the simulation model presented in
Sec. 2.2. The determined power splits are inherently different, depend-
ing on the chosen EMS. The achievement of possibly conicting
objectives is generally a trade-off. In the following, the results are
summarized for each of the missions. The power splits and the battery
SOC are presented in Fig. 6, Fig. 7, Fig. 8. Additionally, key gures on
hydrogen consumption, battery SOC, FC transients and C-rate are also
summarized in Table 5.
Fig. 6. Power splits and SOC for Mission A. (a) Deterministic rule-based EMS,
(b) Fuzzy logic EMS, (c) Optimization-based 1, (d) Optimization-based 2, (d)
Battery state of charge.
C. Deutsch et al.
Energy Conversion and Management: X 14 (2022) 100193
7
Mission A
As expected, the two classes of EMS exhibit inherently different
control patterns. The two rule-based strategies (state machine and fuzzy
logic) operate the fuel cell in a similar manner. The fuel cell is operated
at nominal power rating for approximately the rst 6.5 h before the
battery SOC approaches the lower limit and the fuel cell power is
increased to maximum output (Fig. 6). Due its many-valued logic, the
fuzzy logic EMS results in a smoother transition between nominal and
maximum operating point.
The two optimization-based methods react more dynamically to
changes in the load pattern. The OEMS1 controller initially operates the
fuel cell at minimum power output (Fig. 6c) until the battery SOC drops
to the lower limit (Fig. 6e). From this point onward, the fuel cell is
Fig. 7. Power splits and SOC for Mission B. (a) Deterministic rule-based EMS,
(b) Fuzzy logic EMS, (c) Optimization-based 1, (d) Optimization-based 2, (d)
Battery state of charge.
Fig. 8. Power splits and SOC for Mission C. (a) Deterministic rule-based EMS,
(b) Fuzzy logic EMS, (c) Optimization-based 1, (d) Optimization-based 2, (d)
Battery state of charge.
C. Deutsch et al.
Energy Conversion and Management: X 14 (2022) 100193
8
operated near its maximum operating point. While the FC power uc-
tuates in this high power region, the battery SOC is maintained around
50%. The OEMS2 controller appears most sensitive to changes in battery
SOC and reacts to the load variation from the beginning of the mission
and therefore produces the most dynamic FC operation (Fig. 6d). For
most of this operation, the FC output power is regulated towards
maximum power output. Only during the AUV launch and recovery
phases, during which a few hundred watts of electrical power are drawn
from the batteries, is the FC power output regulated downwards.
On closer inspection of the evolution of battery SOC, it can be seen
that the OEMS1-operation suffers signicant losses of SOC during the
rst 2 h of operation (due to minimum-power operation of the FC),
whereas the OEMS2-operation effectively maintains its SOC. This rather
heavy use of the battery can also be seen from a higher C-rate (Table 5).
From 3 h of operation until the end, both optimization-based methods
show nearly identical battery charge and discharge characteristics, but
due to the initial losses OEMS1 nishes the mission at 54% SOC
(compared to 63.2% for OEMS2). As compared to the state machine
EMS, both optimization-based EMS consume signicantly more
hydrogen (≈20%, Table 5).
Mission B
This 12 h long mission is rather power-demanding with an average
peak power of about 2000 kW. Considering power conversion losses, the
expected maximum power output from the FC is only approximately
1800 kW. Therefore, preservation of battery SOC is not possible
throughout the mission and all four EMS behave as expected.
After a certain period of time, the EMS is triggered to switch to
maximum-power mode. This is the case after ca. 3 h–3.5 h for the rule-
based EMS (Fig. 7a and Fig. 7b) and slightly earlier at about 2 h for the
optimization-based EMS (Fig. 7c and Fig. 7d). The battery state of charge
develops accordingly (Fig. 7e). Since all EMS operate at maximum fuel
cell power for most of the mission, the battery SOC decreases equally
during these periods (between 4 h and 12 h). However, signicant dif-
ferences can be seen in the rst 3 h of the mission. As in Mission A, the
OEMS1 starts with operation at minimum fuel cell power, whereas
OEMS2 starts at nominal fuel cell power. As a result, the SOC decreases
rather steeply in the OEMS1-case. This operation of the FC at minimum
power output is heavily penalized towards the end of the mission, when
the battery SOC falls to 0% and the mission cannot be completed. In such
a case, the fuel cell responds in an erratic manner, which in reality is not
feasible. In direct contrast to this behavior, the OEMS2 manages to
preserve a relatively high SOC at the mission beginning and therefore
nishes with the highest state of charge.
The overall energy consumption varies only marginally across the
different EMS (Table 5). In principle, both rule-based EMS behave
identically and are equally energy-efcient. The OEMS2-controller
consumes 5.3% more hydrogen, of which a fair amount is converted
into electrical energy and stored in the battery, enabling a higher nal
SOC at 30.8%.
Mission C
This third mission is characterized by semi-cyclic variation of high
power and medium power phases, allowing for alternating periods of
battery charging and discharging.
While both rule-based EMS behave largely as expected, it should be
noted that the two EMS operate near the lower SOC limit at the 6 h and
16 h marks (Fig. 8a and Fig. 8b). This near-limit operation leads to the
state machine switching between operating states with strong subse-
quent oscillations in the FC power output. These oscillations lead to an
increase in FC transients (c.f. Table 5). The fuzzy logic controller does
not produce this unwanted and potentially degrading behavior.
The optimization-based controllers exhibit strong load-following
behavior (Fig. 8c and Fig. 8d), where - as before - the largest differ-
ences can be seen during the initial 3 h of operation. During this period,
the OEMS1 controller again operates the FC at minimum power output,
essentially sacricing battery SOC. After this period, the fuel cell power
output varies with the load and the battery SOC develops nearly
identically.
Both optimization-based EMS end the mission with relatively high
battery SOC (55.5% and 63.1%, see Table 5). Analogously, both the total
energy and hydrogen consumption are increased by ≈5%, compared to
the rule-based methods. The recorded sum of absolute transients shows
quasi-negligible numbers for the fuzzy logic and OEMS2 controllers.
However, for the other two controllers, these gures exceed 100 kW
(state machine EMS) and even 300 kW (OEMS1).
Discussion
Methodology and methods
The simulation environment Simulink® has been used for the
modelling of a fuel cell/battery hybrid power system and the perfor-
mance evaluation of Energy Management Strategies. The use of such
simulation environments offers several advantages. There is a contin-
uous development and validation of generic models for a variety of
power system components. Despite being generic, these models are
often detailed enough to capture all dynamics of the system, including
fuel cell, battery and power converter dynamics. These dynamics can
typically not be reproduced by simpler models. The availability of
component models allows for the quick building of accurate high-
delity power system models. However, the simulation time can
Table 5
Resulting key gures from comparison of different Energy Management Strategies (EMS).
EMS Energy consumption H2 consumption SOC Sum of abs. C-rate
Final Minimum Transients
Mission A SMEMS 15.47 kWh 0.430 kg 41.1% 31.6% 6.2 kW 0.33h
−1
FZYEMS 15.57 kWh +0.66% 0.435 kg +1.16% 42.6% 32.6% 4.3 kW 0.33h
−1
OEMS1 18.09 kWh +16.97% 0.526 kg +22.33% 53.4% 37.8% 121.4 kW 0.52h
−1
OEMS2 17.21 kWh +11.28% 0.514 kg +19.54% 63.2% 52.2% 8.2 kW 0.47h
−1
Mission B SMEMS 35.99 kWh 1.006 kg 13.8% 1.2% 691.6 kW 0.33h
−1
FZYEMS 35.99 kWh +0.02% 1.007 kg +0.1% 14.6% 2.0% 688.0 kW 0.33h
−1
OEMS1 Mission not completed 0.0% 0.0% 741.4 kW 2.00h
−1
OEMS2 36.94 kWh +2.66% 1.059 kg +5.3% 30.8% 18.2% 559.9 kW 0.47h
−1
Mission C SMEMS 53.73 kWh 1.575 kg 39.1% 22.5% 106.4 kW 0.33h
−1
FZYEMS 54.01 kWh +0.52% 1.586 kg +0.7% 41.0% 23.9% 8.8 kW 0.33h
−1
OEMS1 57.09 kWh +6.24% 1.699 kg +7.9% 55.5% 34.4% 328.4 kW 0.52h
−1
OEMS2 56.00 kWh +4.21% 1.677 kg +6.5% 63.1% 45.8% 13.1 kW 0.47h
−1
C. Deutsch et al.
Energy Conversion and Management: X 14 (2022) 100193
9
exceed several hours, depending on the mission length and the time step.
This is a major drawback and requires careful consideration of time and
required level of detail.
The design of effective Energy Management Strategies is a funda-
mental part of the development of hybrid power systems. The choice of
the most suitable EMS for a particular application is a non-trivial task
and depends on the drive cycle. Since AUV operations are expensive,
power consumption data from eld trials has been used to synthesize
additional power proles that resemble typical AUV missions such as
bathymetric mapping. This approach has been found to be a reasonable
and cost-efcient way of evaluating EMS against a variety of missions,
without having to model the actual vehicle dynamics or to perform
additional tests.
In order to allow for comparability, the same parameters have been
used to design rule-based and optimization-based EMS. Both types of
strategies are capable of fullling at least one objective to a certain
degree. However, energy efciency, preservation of battery SOC and
also minimization of fuel cell transients are at least partially competing
objectives.
Objectives
System efciency is maximized by operating the system at its
maximum efciency point. The degree to which this objective is ach-
ieved depends mostly on the rigidity of the chosen strategy. Rule-based
strategies with few nite states can most easily achieve energy-
efcient operation. In order for the studied optimization-based EMS to
operate more energy-efciently, the respective weights in the cost
functions need to be modied, which would worsen their ability to
preserve battery SOC.
Battery SOC is preserved by allowing the system to exhibit dynamic
behavior. More dynamic behavior comes at the cost of higher hydrogen
consumption. Rule-based methods like presented in this study lack the
ability to preserve battery SOC during load-varying missions. This can
be a serious problem since a critically low battery SOC can remove any
operational margins from the system and therefore poses a potential
threat to mission success and even vehicle survival, e.g. in the case of
emergency homing while operating in hazardous environments. In order
for rule-based methods to improve their ability of preserving battery
SOC, rules for higher fuel cell power usage or more nite states need to
be introduced. In any case, this would compromise on energy efciency
and increase EMS complexity.
Two particular cases of high fuel cell transients due to bang-bang
control need to be mentioned. Firstly, the operation of nite-state ma-
chine EMS near rule-switching points can lead to unwanted FC tran-
sients (c.f. Mission C, Table 5). A potential solution for this problem are
fuzzy logic controllers, which introduce essentially innite states due to
their many-valued logic. Secondly, optimization-based methods with
linear cost functions (OEMS1) only have two possible setpoints, i.e.
either minimum or maximum fuel cell power. In certain cases, this can
also lead to a signicant increase in FC transients (c.f. Table 5).
Recommendations and further investigation
The ndings indicate that there are best-choice EMS for different
mission types. Fuzzy logic EMS are particularly suitable for missions that
are characterized by low risk and predictable vehicle behavior (standard
missions), where hydrogen consumption and FC degradation should be
prioritized. More complex optimization-based methods, such as the
OEMS2, can be recommended for missions with high grades of auton-
omy and uncertainty. By prioritizing power reliability, chances for
mission success and vehicle survival can be maximized.
Given the number of model and rule parameters, cautious system and
controller design is required. It is important to follow appropriate sys-
tem sizing guidelines in order to avoid under- or oversizing of the power
sources [8]. Furthermore, good rule design is key to good EMS
performance. The permissible SOC-interval affects the EMS behavior
and is chosen such that battery capacity can be utilized to the best extent
possible, while avoiding degrading battery cycling [4]. Simulations with
narrower SOC-limits (50%to 70%) have generally shown better load-
following behavior and increased hydrogen consumption, not affecting
this study’s conclusions.
In order to further investigate the performance of different EMS on
fuel cell/battery hybrid AUVs a more thorough analysis of drive cycles
and load prole characteristics will be benecial. Better understanding
of power consumption on AUVs would help in designing the best suit-
able EMS for different use cases. Additionally, future studies need to
combine high-delity models with state-of-the-art strategies from the
elds of machine learning and model predictive control.
Additionally, considering fault tolerance will be an interesting di-
rection for future research. For example, adaptive control strategies for
cases of battery failure are an important topic. In such a scenario (c.f.
Fig. 7c in Sec. 3.2), the fuel cell will increase its power output until the
power demand is met or the maximum power output is reached. In the
meantime, functionality of electrical systems could be impaired.
Simultaneous mission re-planning and corrective emergency actions
would be necessary to facilitate a safe return to base. Such emergency
maneuvers can include disabling of payload sensors, with only essential
sensors remaining active.
Conclusion
In this study, it is shown how Energy Management Strategies for fuel
cell/battery hybrid power systems can be evaluated using a combination
of high-delity simulation environments and experimental power con-
sumption data. The power consumption data has been sampled during
eld trials with a Hugin 3000 AUV. The sampled power prole is used as
a base prole for the evaluation of the EMS. By extracting and appending
selected segments, additional realistic power proles have been gener-
ated. Since there are no standardized drive cycles for underwater vehi-
cles, this is considered a valuable approach.
In the evaluation of the EMS, two rule-based and two optimization-
based methods have been used. The simulation has shown that while
rule-based methods can easily be tuned to enable the most energy ef-
cient operation of the AUV, their inherent weakness is the rather poor
capability of maintaining battery SOC throughout varying load condi-
tions - an ability that can only be improved by increasing complexity of
the underlying rule sets. If simple, rule-based methods are preferred,
fuzzy logic provides a suitable framework. Optimization-based methods,
in particular those based on quadratic (or nonlinear) cost functions, can
yield outstanding load-following behavior and performance with respect
to the preservation of battery SOC. Preserving battery SOC is achieved
by dynamic operation of the fuel cell, essentially covering its full oper-
ational range. This leads to a trade-off between maintaining SOC and
minimizing hydrogen consumption. Potentially degrading FC transients
can occur as a result of bang-bang control.
Based on these ndings, it is concluded that EMS complexity needs to
increase with mission-associated risks and level of autonomy. Standard-
type missions in relatively safe waters benet from energy-efcient
fuzzy logic control, whereas high-autonomy, high-risk missions benet
from power-reliable control based on non-linear optimization.
CRediT authorship contribution statement
Clemens Deutsch: Conceptualization, Methodology, Software,
Formal analysis, Investigation, Data curation, Writing - original draft.
Ariel Chiche: Conceptualization, Methodology, Software, Formal
analysis, Investigation, Writing - original draft. Sriharsha Bhat:
Conceptualization, Methodology, Software, Formal analysis, Investiga-
tion, Writing - review & editing. Carina Lagergren: Resources, Writing -
review & editing, Supervision, Funding acquisition. G¨
oran Lindbergh:
Resources, Writing - review & editing, Supervision, Funding acquisition.
C. Deutsch et al.
Energy Conversion and Management: X 14 (2022) 100193
10
Jakob Kuttenkeuler: Resources, Writing - review & editing, Supervi-
sion, Funding acquisition.
Declaration of Competing Interest
The authors declare that they have no known competing nancial
interests or personal relationships that could have appeared to inuence
the work reported in this paper.
Acknowledgment
This work was supported by Swedish Foundation for Strategic
Research (SSF) through the Swedish Maritime Robotic Centre (SMaRC
(IRC 15–0046)). The authors thank Prof. Anna Wa˚hlin and the Univer-
sity of Gothenburg for the opportunity to perform eld trials with the
Hugin AUV at the Born¨
o eld station.
References
[1] Abdalla, A.M., Hossain, S., Nisndy, O.B., Azad, A.T., Dawood, M., Azad, A.K.,
2018. Hydrogen production, storage, transportation and key challenges with
applications: A review. doi:10.1016/j.enconman.2018.03.088.
[2] Albarghot MM, Iqbal MT, Pope K, Rolland L. Sizing and Dynamic Modeling of a
Power System for the MUN Explorer Autonomous Underwater Vehicle Using a Fuel
Cell and Batteries. J Energy 2019. https://doi.org/10.1155/2019/4531497.
[3] Albarghot MM, Iqbal MT, Pope K, Rolland L. Dynamic modeling and simulation of
the MUN Explorer autonomous underwater vehicle with a fuel cell system. AIMS
Electron Electr Eng 2020;4:114–31. https://doi.org/10.3934/
ElectrEng.2020.1.114.
[4] Benavente-Araoz F, Varini M, Lundblad A, Cabrera S, Lindbergh G. Effect of Partial
Cycling of NCA/Graphite Cylindrical Cells in Different SOC Intervals.
J Electrochem Soc 2020;167:040529. https://doi.org/10.1149/1945-7111/
ab78fd.
[5] Borgogna, G., Lamberti, T., Massardo, A.F., 2015. Innovative power system for
autonomous underwater vehicle, in: MTS/IEEE OCEANS 2015 – Genova:
Discovering Sustainable Ocean Energy for a New World. doi:10.1109/OCEANS-
Genova.2015.7271339.
[6] Bradley A, Feezor M, Singh H, Yates Sorrell F. Power systems for autonomous
underwater vehicles. IEEE J Oceanic Eng 2001;26:526–38. https://doi.org/
10.1109/48.972089. url: http://ieeexplore.ieee.org/document/972089/.
[7] Chiche, A., Lagergren, C., Lindbergh, G., Stenius, I., 2018. Sizing the energy system
on long-range AUV, in: AUV 2018–2018 IEEE/OES Autonomous Underwater
Vehicle Workshop, Proceedings. doi:10.1109/AUV.2018.8729812.
[8] Chiche A, Lindbergh G, Stenius I, Lagergren C. A Strategy for Sizing and Optimizing
the Energy System on Long-Range AUVs. IEEE J Oceanic Eng 2021. https://doi.
org/10.1109/JOE.2021.3062047.
[9] Church RA, Warren DJ. Autonomous underwater vehicles: The latest tool for
archaeological investigations. Marine Technol Soc J 2002;36:45–50. https://doi.
org/10.4031/002533202787913431.
[10] D’Amore-Domenech R, Raso MA, Villalba-Herreros A, Santiago ´
O, Navarro E,
Leo TJ. Autonomous underwater vehicles powered by fuel cells: Design guidelines.
Ocean Eng 2018. https://doi.org/10.1016/j.oceaneng.2018.01.117.
[11] Deutsch, C., Chiche, A., Bhat, S., Lagergren, C., Lindbergh, G., Kuttenkeuler, J.,
2020. Energy Management Strategies for Fuel Cell-Battery Hybrid AUVs, in: 2020
IEEE/OES Autonomous Underwater Vehicles Symposium (AUV), IEEE, St Johns,
NL, Canada. url: https://ieeexplore.ieee.org/document/9267932, doi:10.1109/
AUV50043.2020.9267932.
[12] Divakarla KP, Emadi A, Razavi SN. Journey Mapping-A New Approach for Dening
Automotive Drive Cycles. IEEE Trans. Ind. Appl. 2016;52. https://doi.org/
10.1109/TIA.2016.2593694.
[13] Erdinc O, Uzunoglu M. Recent trends in PEM fuel cell-powered hybrid systems:
Investigation of application areas, design architectures and energy management
approaches. Renew. Sustain. Energy Rev. 2010;14:2874–84. https://doi.org/
10.1016/j.rser.2010.07.060.
[14] Furlong, M.E., Paxton, D., Stevenson, P., Pebody, M., McPhail, S.D., Perrett, J.,
2012. Autosub Long Range: A long range deep diving AUV for ocean monitoring,
in: 2012 IEEE/OES Autonomous Underwater Vehicles, AUV 2012, pp. 1–7. doi:
10.1109/AUV.2012.6380737.
[15] Grifths, G., Jamieson, J., Mitchell, S., Rutherford, K., 2004. Energy storage for
long endurance AUVs, in: Advances in Technology for underwater Vehicles, pp.
8–16. url: http://eprints.soton.ac.uk/15783/.
[16] Hagen, P.E., Størkersen, N., Sten, G., Marthinsen, B.E., Vestgård, K., 2005. Military
operations with HUGIN AUVs: Lessons learned and the way ahead, in: Oceans 2005
– Europe. doi:10.1109/OCEANSE.2005.1513160.
[17] Hagen PE, Størkersen N, Vestgård K. HUGIN - use of UUV technology in marine
applications. MTS/IEEE Oceans 1999;2:967–72.
[18] Han IS, Kho BK, Cho S. Development of a polymer electrolyte membrane fuel cell
stack for an underwater vehicle. J Power Sources 2016. https://doi.org/10.1016/j.
jpowsour.2015.11.049.
[19] Hasvold Ø, Johansen KH, Mollestad O, Forseth S, Størkersen N. The alkaline
aluminium/hydrogen peroxide power source in the Hugin II unmanned
underwater vehicle. J Power Sources 1999;80:254–60. https://doi.org/10.1016/
S0378-7753(98)00266-3.
[20] Hasvold Ø, Lian T, Haakaas E, Størkersen N, Perelman O, Cordier S. CLIPPER: A
long-range, autonomous underwater vehicle using magnesium fuel and oxygen
from the sea. J Power Sources 2004;136:232–9. https://doi.org/10.1016/j.
jpowsour.2004.03.023.
[21] Hasvold Ø, Størkersen NJ, Forseth S, Lian T. Power sources for autonomous
underwater vehicles. J Power Sources 2006;162:935–42. https://doi.org/10.1016/
j.jpowsour.2005.07.021.
[22] Hyakudome, T., Nakatani, T., Yoshida, H., Tani, T., Ito, H., Sugihara, K., 2016.
Development of fuel cell system for long cruising lange autonomous underwater
vehicle, in: Autonomous Underwater Vehicles 2016, AUV 2016. doi:10.1109/
AUV.2016.7778666.
[23] Hyakudome, T., Tani, T., Ito, H., Sugihara, K., 2018. Concept of long cruising range
autonomous underwater vehicle powered by fuel cell, in: 2018 OCEANS – MTS/
IEEE Kobe Techno-Oceans, OCEANS – Kobe 2018. doi:10.1109/
OCEANSKOBE.2018.8559227.
[24] Hyakudome, T., Yoshida, H., Nakatani, T., Ohta, Y., Tani, T., Sugihara, K., Moriga,
T., Iwamoto, T., Kawaharazaki, Y., Oda, T., Fujita, Y., 2013. Development of fuel
cell system for underwater power source, in: OCEANS 2013 MTS/IEEE Bergen: The
Challenges of the Northern Dimension. doi:10.1109/OCEANS-
Bergen.2013.6608025.
[25] King, P., Williams, G., Coleman, R., Zurcher, K., Bowden-Floyd, I., Ronan, A.,
Kaminski, C., Laframboise, J.M., McPhail, S., Wilkinson, J., Bowen, A., Dutrieux,
P., Bose, N., Wahlin, A., Andersson, J., Boxall, P., Sherlock, M., Maki, T., 2019.
Deploying an AUV beneath the Sørsdal Ice Shelf: Recommendations from an
expert-panel workshop, in: 2018 IEEE/OES Autonomous Underwater Vehicle
Workshop (AUV), Porto. pp. 1–6. doi:10.1109/auv.2018.8729786.
[26] Larminie, J., Dicks, A., 2013. Fuel cell systems explained: Second edition. John
Wiley & Sons Ltd. doi:10.1002/9781118878330.
[27] Li G, Zhang J, He H. Battery SOC constraint comparison for predictive energy
management of plug-in hybrid electric bus. Appl Energy 2017;194:578–87.
https://doi.org/10.1016/j.apenergy.2016.09.071.
[28] Li H, Ravey A, N’Diaye A, Djerdir A. A review of energy management strategy for
fuel cell hybrid electric vehicle. In: 2017 IEEE Vehicle Power and Propulsion
Conference, VPPC 2017 – Proceedings; 2018. p. 1–6. https://doi.org/10.1109/
VPPC.2017.8330970.
[29] Loh TY, Brito MP, Bose N, Xu J, Tenekedjiev K. A Fuzzy-Based Risk Assessment
Framework for Autonomous Underwater Vehicle Under-Ice Missions. Risk Anal
2019;39:2744–65. https://doi.org/10.1111/risa.13376.
[30] Mamdani EH, Assilian S. An experiment in linguistic synthesis with a fuzzy logic
controller. Int J Man Mach Stud 1975;7. https://doi.org/10.1016/S0020-7373(75)
80002-2.
[31] Marthiniussen, R., Vestgård, K., Klepaker, R.A., Størkersen, N., 2004. HUGIN-AUV
concept and operational experiences to date, in: Ocean ’04 – MTS/IEEE Techno-
Ocean ’04: Bridges across the Oceans – Conference Proceedings, pp. 846–850. doi:
10.1109/oceans.2004.1405571.
[32] MathWorks, Energy Management Systems for a Hybrid Electric Source
(Application for a More Electric Aircraft). url: www.mathworks.com/help/phys
mod/sps/ug/energy-management-systems-for-a-hybrid-electric-source-applicatio
n-for-a-more-electric-aircraft.html.
[33] McClure, P.R., Reid, R.S., Poston, D.I., Dasari, V.R., 2016. Reactor Power for Large
Displacement Autonomous Underwater Vehicles. Technical Report. Los Alamos
National Laboratory (LANL). Los Alamos, NM (United States). doi:10.2172/
1312635.
[34] McPhail S. Autosub6000: A Deep Diving Long Range AUV. J Bionic Eng 2009;6:
55–62. https://doi.org/10.1016/S1672-6529(08)60095-5.
[35] McPhail, S., Templeton, R., Pebody, M., Roper, D., Morrison, R., 2019. Autosub
Long Range AUV Missions Under the Filchner and Ronne Ice Shelves in the
Weddell Sea, Antarctica – an Engineering Perspective, in: OCEANS 2019 MTS/IEEE
Marseille, Marseille.
[36] Mendez A, Leo TJ, Herreros MA. Current state of technology of fuel cell power
systems for autonomous underwater vehicles. Energies 2014;7:4676–93. https://
doi.org/10.3390/en7074676.
[37] Millard, N.W., Grifths, G., Finegan, G., McPhail, S.D., Meldrum, D.T., Pebody, M.,
Perrett, J.R., Stevenson, P., Webb, a.T., 1998. Versatile Autonomous
Submersibles–the realising and testing of a practical vehicle. Underwater Technol:
Int J Soc Underwater 23, 7–17. doi:10.3723/175605498783259894.
[38] Njoya Motapon S, Dessaint LA, Al-Haddad K. A comparative study of energy
management schemes for a fuel-cell hybrid emergency power system of more-
electric aircraft. IEEE Trans Industr Electron 2014;61:1320–34. https://doi.org/
10.1109/TIE.2013.2257152.
[39] Omar N, Monem MA, Firouz Y, Salminen J, Smekens J, Hegazy O, Gaulous H,
Mulder G, Van den Bossche P, Coosemans T, Van Mierlo J. Lithium iron phosphate
based battery – Assessment of the aging parameters and development of cycle life
model. Appl Energy 2014;113:1575–85. https://doi.org/10.1016/j.
apenergy.2013.09.003.
[40] Panday, A., Bansal, H.O., 2014. A review of optimal energy management strategies
for hybrid electric vehicle. doi:10.1155/2014/160510.
[41] Pei, P., Chen, H., 2014. Main factors affecting the lifetime of Proton Exchange
Membrane fuel cells in vehicle applications: A review. doi:10.1016/j.
apenergy.2014.03.048.
C. Deutsch et al.
Energy Conversion and Management: X 14 (2022) 100193
11
[42] Raugel, E., Rigaud, V., Lakeman, C., 2010. Sea experiment of a survey AUV
powered by a fuel cell system, in: 2010 IEEE/OES Autonomous Underwater
Vehicles, AUV 2010. doi:10.1109/AUV.2010.5779676.
[43] Rivard, E., Trudeau, M., Zaghib, K., 2019. Hydrogen storage for mobility: A review.
doi:10.3390/ma12121973.
[44] Roman C, Mather R. Autonomous underwater vehicles as tools for deep-
submergence archaeology. In: Proceedings of the Institution of Mechanical
Engineers Part M: Journal of Engineering for the Maritime Environment 224; 2010.
https://doi.org/10.1243/14750902JEME202.
[45] Roper, D.T., Phillips, A.B., Harris, C.A., Salavasidis, G., Pebody, M., Templeton, R.,
Amma, S.V.S., Smart, M., McPhail, S., 2017. Autosub long range 1500: An ultra-
endurance AUV with 6000 Km range, in: OCEANS 2017 - Aberdeen, pp. 1–5. doi:
10.1109/OCEANSE.2017.8084928.
[46] Samuel S, Austin L, Morrey D. Automotive test drive cycles for emission
measurement and real-world emission levels – A review. Proc Inst Mech Eng, Part
D: J Automobile Eng 2002:216. https://doi.org/10.1243/095440702760178587.
[47] Saw LH, Somasundaram K, Ye Y, Tay AA. Electro-thermal analysis of Lithium Iron
Phosphate battery for electric vehicles. J Power Sources 2014;249:231–8. https://
doi.org/10.1016/j.jpowsour.2013.10.052.
[48] Sharaf, O.Z., Orhan, M.F., 2014. An overview of fuel cell technology: Fundamentals
and applications. doi:10.1016/j.rser.2014.01.012.
[49] Singh, A.S., Akram, M.W., Kumar, P., Sanjay, S., 2019. Analysis of power pack for
Autonomous-Underwater-Vehicle: Potential of super-critical CO2 brayton cycle, in:
SAE Technical Papers. doi:10.4271/2019-01-1260.
[50] Souleman NM, Tremblay O, Dessaint LA. A generic fuel cell model for the
simulation of fuel cell power systems. In: 2009 IEEE Power and Energy Society
General Meeting, PES; 2009. ’09.. https://doi.org/10.1109/PES.2009.5275853.
[51] Souleman, N.M., Tremblay, O., Dessaint, L.A., 2009b. A generic fuel cell model for
the simulation of fuel cell vehicles, in: 5th IEEE Vehicle Power and Propulsion
Conference, VPPC ’09, pp. 1722–1729. doi:10.1109/VPPC.2009.5289692.
[52] St-Pierre J. Fuel Cells – Proton-Exchange Membrane Fuel Cells — Overview
Performance and Operational Conditions. Encycl Electrochem Power Sources
2009:901–11. https://doi.org/10.1016/B978-044452745-5.00233-1.
[53] Størkersen, N., Hasvold, Ø., 2004. Power Sources for AUVs, in: Science and Defence
Conference.
[54] Sulaiman, N., Hannan, M.A., Mohamed, A., Ker, P.J., Majlan, E.H., Wan Daud, W.
R., 2018. Optimization of energy management system for fuel-cell hybrid electric
vehicles: Issues and recommendations. doi:10.1016/j.apenergy.2018.07.087.
[55] Thorstensen B. Parametric study of fuel cell system efciency under full and part
load operation. J Power Sources 2001;92:9–16. https://doi.org/10.1016/S0378-
7753(00)00497-3.
[56] Townsend, N., Shenoi, A., 2013. Recharging Autonomous Underwater Vehicles
from Ambient Wave Induced Motions. Oceans 2013 (IEEE), 2159–2168. doi:
10.23919/OCEANS.2013.6741051.
[57] Townsend NC. Self-powered autonomous underwater vehicles: results from a
gyroscopic energy scavenging prototype. IET Renew Power Gener 2016;10:
1078–86. https://doi.org/10.1049/iet-rpg.2015.0210. url: http://digital-library.
theiet.org/content/journals/10.1049/iet-rpg.2015.0210.
[58] Tremblay O, Dessaint LA. Experimental validation of a battery dynamic model for
EV applications. World Electric Vehicle J 2009;3:289–98. https://doi.org/
10.3390/wevj3020289.
[59] Wang T, Li Q, Wang X, Qiu Y, Liu M, Meng X, Li J, Chen W. An optimized energy
management strategy for fuel cell hybrid power system based on maximum
efciency range identication. J Power Sources 2020;445. https://doi.org/
10.1016/j.jpowsour.2019.227333.
[60] Weydahl H, Gilljam M, Lian T, Johannessen TC, Holm SI, Hasvold JØ. Fuel cell
systems for long-endurance autonomous underwater vehicles – challenges and
benets. Int J Hydrogen Energy 2020. https://doi.org/10.1016/j.
ijhydene.2019.05.035.
[61] Whitt, C., Pearlman, J., Polagye, B., Caimi, F., Muller-Karger, F., Copping, A.,
Spence, H., Madhusudhana, S., Kirkwood, W., Grosjean, L., Fiaz, B.M., Singh, S.,
Singh, S., Manalang, D., Gupta, A.S., Maguer, A., Buck, J.J., Marouchos, A.,
Atmanand, M.A., Venkatesan, R., Narayanaswamy, V., Testor, P., Douglas, E., de
Halleux, S., Khalsa, S.J., 2020. Future Vision for Autonomous Ocean Observations.
doi:10.3389/fmars.2020.00697.
[62] Wynn RB, Huvenne VA, Le Bas TP, Murton BJ, Connelly DP, Bett BJ, Ruhl HA,
Morris KJ, Peakall J, Parsons DR, Sumner EJ, Darby SE, Dorrell RM, Hunt JE.
Autonomous Underwater Vehicles (AUVs): Their past, present and future
contributions to the advancement of marine geoscience. Mar Geol 2014;352:
451–68. https://doi.org/10.1016/j.margeo.2014.03.012.
[63] Yamagata, H., Maki, T., Yoshida, H., Nogi, Y., 2019. Hardware Design of Variable
and Compact AUV ’MONACA’ for Under-Ice Survey of Antarctica, in: 2019 IEEE
International Underwater Technology Symposium, UT 2019 – Proceedings. doi:
10.1109/UT.2019.8734395.
[64] Yuh J, Marani G, Blidberg DR. Applications of marine robotic vehicles. Intel Serv
Robot 2011;4:221–31. https://doi.org/10.1007/s11370-011-0096-5.
[65] Zhu C, Li X, Song L, Xiang L. Development of a theoretically based thermal model
for lithium ion battery pack. J Power Sources 2013;223:155–64. https://doi.org/
10.1016/j.jpowsour.2012.09.035.
C. Deutsch et al.
