Research Article

Published: 2023-09-12
https://doi.org/10.20935/AcadEng6115

An adaptive power management tool for sizing a PV-

battery-hydrogen off-grid electrical energy system
Loris Pace1,*, Ayyoub Zouaghi1, Matthieu Beley1, Naoufel Haddour1

Academic Editor: Dimitrios A. Karras

Abstract
Off-grid electrical energy systems based on renewable energy sources (RESs) have become increasingly popular for their ability to
generate low-carbon electricity in remote areas without access to traditional power grids. These systems rely on the effective
management of RESs and storage solutions. Designing and sizing these systems can be a complex task, requiring careful consideration
of various parameters such as energy demand, solar irradiance, storage capacity of batteries and State of Charge, power management
of fuel cells, and hydrogen production and storage. This study presents an adaptive power management tool that facilitates the sizing
of energy equipment for standalone low-carbon microgrids. The proposed simulation tool, implemented in Matlab/Simulink, is based
on mathematical models for each energy unit and incorporates a specific power management strategy to determine the optimal size of
each component in the system. The effectiveness of the tool is illustrated through a case study involving a PV-battery-hydrogen energy
system designed to supply electricity to a standalone district. Results show that the developed tool can be a valuable aid for system
designers and planners for creating sustainable and reliable off-grid electrical energy systems, as well as for educational and learning
activities.

Keywords: solar panels, green hydrogen, fuel cell, battery storage, optimal sizing, standalone microgrid

Citation: Pace L, Zouaghi A, Beley M, Haddour N. An adaptive power management tool for sizing a PV-battery-hydrogen off-grid
electrical energy system. Academia Engineering 2023;1. https://doi.org/10.20935/AcadEng6115

1. Introduction
Within the context of climate change, depletion of natural battery and hydrogen storage. However, these articles do not
resources and energy crisis, low-carbon emitting off-grid cope with the sizing of different energy equipment and
electrical energy systems, and smart autonomous microgrids consequently no optimization procedures are applied. Several
have become increasingly popular in recent years [1–3]. The optimization algorithms are compared for the sizing of a PV-
development of such systems is enabled by the continuous battery-hydrogen microgrid in the study by Diab et al. [8].
improvement in the exploitation of renewable energy sources Nevertheless, the proposed sizing methodology is mostly driven
(RESs), such as solar or wind energy, and advancements in the by costs and the obtained results strongly depend on these
power system efficiency [4]. Although RESs show high capability economic constraints [9]. Similar approaches are used in the
in resolving actual environmental issues related to fossil energy notorious existing software such as HOMER Pro [10] or H2RES
exploitation, their intermittency requires efficient storage [11] to design such energy systems in regard to techno-economic
systems to ensure a steady supply of electricity. Previous work aspects [5, 12–14]. However, there are no real guidelines for the
has shown that hybrid storage systems based on batteries and user to set the optimal size of the energy equipment during
green hydrogen systems are the most efficient solutions to satisfy simulations and no obvious indications are given about the
short- and long-term electrical load requirements [5]. energy management process in these software solutions.

In autonomous low-carbon energy microgrids, the management To the best of the authors’ knowledge, there is still a lack of
of combined energy sources and storage systems led to a more methodology, independent of economic constraints and easily
complex design of the installed equipment. Batiyah et al. [6] adaptable, to provide the optimal size of the energy equipment in
present a Model Predictive Control (MPC) scheme based on the case of standalone microgrids comprising RESs and hybrid
mathematical models which is able to manage disturbances and battery-hydrogen storage systems. In this context, this study
storage systems in a standalone microgrid. The studied system aims to present an adaptive power management tool that
includes photovoltaic (PV) panels, wind turbines (WT), and achieves this goal. The main contributions of this work are based
battery storage. In the study by Cecilia et al. [7], the same MPC on an innovative power management strategy using state
method is applied to a power system that includes a hybrid machine principles and an iterative sizing optimization of the

1UnivLyon, Ecole Centrale de Lyon, INSA Lyon, Université Claude Bernard Lyon 1, CNRS, Ampère, UMR5005, 69130 Ecully, France.
*email: [email protected]

ACADEMIA ENGINEERING 2023, 1 1 of 12

https://www.academia.edu/journals/3/academia_engineering/about https://doi.org/10.20935/AcadEng6115

energy equipment from a power study conducted hour by hour, 2. Materials and methods
on a yearly basis. Moreover, proven mathematical models [3, 9]
and a detailed specification extraction methodology are used to 2.1. System overview
represent the energy units (sources and storage systems). Fur- The power management tool proposed in this work is designed in
thermore, since users have full access to each part of the system, Matlab/Simulink in order to supply electrical energy to a
the presented tool can be easily modified to integrate other energy standalone green-energy district. The only source of energy
units, as well as costs or environmental constraints. In this work, comes from a photovoltaic (PV) unit composed of several PV
the proposed sizing tool is developed in Matlab/Simulink. It is panels. The rated power of the PV group and the number of
applied to a fictive off-grid district, named New Era thereafter, panels must be determined. The electrical energy is stored in two
that is supplied by 100% of green electricity through a PV- distinct ways:
battery-hydrogen power system.
• Short-time storage by using a battery unit composed of
The study is structured as follows. First, the Section 2 is divided multiple battery cells. The number of cells and the rated
into five sub-sections. The standalone New Era district, to which power of the battery group must be determined.
the developed simulation tool is applied, is presented in Section
2.1. An overview of the studied PV-battery-hydrogen energy • Long-time storage by converting excess electricity into
system is also provided in this part. Section 2.2 presents the hydrogen. The unit consists of an electrolyzer group, a fuel
hourly electrical consumption profiles considered over a year of cell group, and a hydrogen tank. The parameters to be sized
operation. Subsequently, Section 2.3 details the proposed models include the rated power of the electrolyzer and fuel cell
for the different energy units considered in this work. The power group, as well as the required mass of hydrogen to be stored.
management strategy used to satisfy the electrical load at each The hydrogen is considered to be stored in gaseous form.
time step is presented in Section 2.4, while Section 2.5 describes To size the studied electrical energy system, the power flows must
the proposed sizing methodology to get the optimal values of the be known during a complete year of operation depending on the
equipment capacities in order to ensure a whole year operation PV production, the battery group State of Charge (SoC), and the
of the power system. The Section 3 highlights the obtained stored quantity of hydrogen. Therefore, a power management
simulation results for the specific case study. Finally, a discussion strategy based on a state machine principle is proposed in this
about the performance of the proposed sizing tool is engaged, and work. Figure 1 provides an overview of the energy system
the prospects are identified for future work. implemented in Matlab/Simulink.

Figure 1 • Synopsis of the implemented energy system.

The fictive New Era district is considered to be located in the city are also available around the district and could be used to place
of Ecully, France. It is composed of several buildings: photovoltaic panels.

• A residence including 40 apartments

2.2. Electrical consumption profiles
• An office center including 20 offices The electrical consumption profiles of New Era are built by
• Three different stores: a grocery, a bakery, and a pharmacy stocktaking all the electrical equipment, identifying their
electrical power and determining their operation time as shown
• An Electric Vehicle (EV) service station comprising four in Table 1. These data have been collected from the French
charging points and an office Agency of Environment and Management of Energy (ADEME)
Additionally, two parking areas are available for the residents [15] and the energy consultant Hello Watt [16], in order to give
and some areas are left available and could be used to install the most realistic characteristics to this fictive case study. Some
some parts of the energy system such as storage units. Figure 2 assumptions are made to simplify the profiles:
shows an overview of the studied New Era district. Vacant lands

ACADEMIA ENGINEERING 2023, 1 2 of 12

https://www.academia.edu/journals/3/academia_engineering/about https://doi.org/10.20935/AcadEng6115

Figure 2 • 3D model of the New Era district.

Table 1 • Electrical equipment power and operation time • Two seasons are considered: summer (Apr–Sept) and
winter (Oct–Mar)
Power Operation
Equipment Quantity
(kW) time • People are assumed to be at work during the day (8AM–
Office computer 20 0.2 p.u. 8 AM–6 PM 6PM) and at the apartment during the night (7PM–7AM)

Wi-Fi Box 43 0.025 p.u. Midnight–11 PM • Only LED lighting is considered

Laptop charger 40 0.1 p.u. 7 PM–8 PM • Dishwashers, washing machines, and clothe dryers are
assumed to operate during the night to level off the
6 AM–8 AM
Lighting 3,840 m² 0.0025/m² electrical consumption profiles
7 PM–Midnight
• 4 EVs are charged every hour during the day (7AM–8PM)
TV 40 0.1 p.u. 8 PM–Midnight
Figure 3a reveals the electrical consumption profiles of each
Oven 40 2 p.u. 6 PM–7 PM
building. It can be observed that the major part of the electrical
Industrial oven 2 2.5 p.u. 5 AM–8 AM power consumption takes place early in the morning and at the
Induction plate 40 2.5 p.u. 6 PM–7 PM
evening for the residence, while the electricity is mainly
consumed during the day by the office and the service station.
Fridge 40 0.03 p.u. Midnight–11 PM Figure 3b illustrates the total consumption profiles of New Era
Vacuum cleaner 40 0.4 p.u. 5 PM–6 PM during winter and summer seasons. A slight difference between
the winter and summer profiles originates from light usage
Dishwasher 40 1.5 p.u. Midnight–1 AM
during the day, mainly at the office. It should be noted that the
Washing machine 40 2.5 p.u. 1 AM–2 AM heating system is not considered in this work.
Clothes dryer 40 2.5 p.u. 2 AM–3 AM From Figure 3b, the maximum demand of electricity per day,
EV charger 4 7 p.u. 7 AM–8 PM Pload , is found to be equal to 168 kW. This value will be used
max
for the initialization of the proposed system sizing strategy
presented in Section 2.5.

Figure 3 • Electrical consumption profiles (a) by building and (b) total consumption of the district during winter and summer.

2.3. Energy units modeling the development of the power management tool and the sizing
This section aims to present in detail the implemented models for strategy, the models proposed here are simplified and several
the considered energy units which include the PV panels group, physical phenomena that could impact the power conversion are
the battery unit, and hydrogen storage. As this work focuses on neglected. For each energy unit, the limitations of the presented
models and some leads to overcome these limitations are given.

ACADEMIA ENGINEERING 2023, 1 3 of 12

https://www.academia.edu/journals/3/academia_engineering/about https://doi.org/10.20935/AcadEng6115

2.3.1. Photovoltaic unit irradiance data used in this work are obtained from Global Solar
The PV unit, comprising all the solar panels, aims to evaluate the Atlas [17] and are presented in Figure 4a. In one day, the solar
produced power, PPV , at every simulation time step, which is irradiance increases to reach its higher value at mid-day, then it
decreases. It must be noted that the four seasons are considered,
considered equal to one hour in this work. This parameter can be whereas the same solar irradiance has been considered for spring
adjusted regarding the energy consumption profiles. and fall for simplification.
The proposed model is inspired from the study by Zahboune et Knowing that the reference solar irradiance, Gref , given in
al. [9] but the energy variables are converted in power variables,
considering then that the power remains constant during one datasheet for solar panels is 1,000 W/m², the electrical power
time step. Thus, the input data of the model are solar irradiance, generated by the PV group, PPV , is given by equation 1. The block
G (W/m²), given every hour of the day for each season. The solar diagram of the PV unit is given in Figure 4b.

Figure 4 • PV unit (a) solar irradiance data for each season and (b) block diagram.

G methodology to build the battery unit model based on the

PPV = nPV Pr (1)
Gref technical documentation of the chosen cell [18]. Although the
presented model lays on the same approach proposed by
where Pr is the rated power of one solar panel at an irradiance of Zahboune et al. [9], it is not based on energy variables but
Gref and nPV is the number of solar panels considered. In this depends mainly on SoC and the power supplied or demanded to
the battery group. In the proposed battery unit model, all battery
case study, the selected solar panels are RSP200D from Renogy
cells are considered to be connected in parallel, meaning that the
©. The rated power at Gref = 1000 W/m² is Pr = 200 Wc. The
voltage of battery group is considered equal to the voltage of one
photovoltaic conversion efficiency is found to be 18.1%. The cell, Vcell , being around 12 V.
surface area of the PV panel is 1.1 m2.
According to Ultimatron JDG12-100 [18], the battery cell voltage,
The presented PV unit model assumes an electrical power
Vcell , is given as a function of the SoC following the expression in
production that only depends on solar irradiance data and the
number of panels considered. It does not take into account equation 2.
several factors that could impact the produced power such as PV
Vcell = V0 + k1SoC (2)
panel orientation, shadow, temperature, or aging effects. Also,
the probable use of power electronics systems such as Maximum where V0 = 12 V is the battery cell voltage at SoC = 0% and
Power Point Tracking (MPPT) converters and their associated
k1 = 1 V. In the formulae SoC is between 0 and 1 without unit.
efficiency are not considered here. However, the implementation
of this model in the Matlab/Simulink software only requires a few Once the battery voltage is known, the battery cell current Icell can
operator blocks and the user can easily modify the PV unit to
be obtained from the battery unit power, PB , using equation 3.
improve the model accuracy by considering some physical
parameters and possibly its own characterization data. This current is either positive when the battery cell is charging
( PB > 0), or negative when the battery cell is discharging ( PB < 0).
2.3.2. Battery unit
PB
The battery unit aims to evaluate the SoC of the battery group at I cell = (3)
ncellVcell
every simulation time step based on the SoC at the previous time
step and the present supplied/demanded electrical power. The Figure 5 gives the SoC variation, ∆SoC, according to the
battery group is composed of ncell battery cells, where ncell is a charging current per cell. This curve is obtained from data
parameter to optimize, which allow one to level off the SoC available in the technical documentation [18]. For simplicity, it is
variation by limiting the cell current. considered that the SoC variation is symmetrical in charging and
discharging modes. Therefore, the hysteresis phenomena in the
For this case study, the selected solar battery cell is the GEL
battery are not taken into account in this work.
Ultimatron JDG12-100 12 V 100 Ah [18]. The recommended
charging current is 10 A with a limit at 25 A. Here, we propose a

ACADEMIA ENGINEERING 2023, 1 4 of 12

https://www.academia.edu/journals/3/academia_engineering/about https://doi.org/10.20935/AcadEng6115

From Figure 5, the relation between SoC and Icell can be As previously mentioned, the design of the proposed model is
expressed as equation 4. fully based on the battery cell specifications. However, it presents
some simplifications and some factors are neglected. Apart from
( ) 2
SoC = sign I cell k2 I cell + k3 I cell (4) the hysteresis phenomenon, the temperature and aging effects,
for example, are not considered here. Similar to the PV unit
where k2 = 8.10−5 A−2 and k3 = 0.01 A−1. model, power electronics systems, such as the Battery Manage-
According to equations 2–4, the block diagram of the battery unit ment System (BMS), and their associated efficiency are not taken
is given in Figure 6. into account. Nevertheless, the user can easily modify the battery
unit model to improve its accuracy by considering some physical
parameters and possibly its own characterization data.

Figure 5 • State of charge variation vs. cell current: extracted data (the blue line); polynomial model (red dots).

Figure 6 • Proposed block diagram for the battery unit.

2.3.3. Hydrogen storage unit group, mH , is obtained using equation 5. By the same
2prod
The proposed hydrogen storage (HS) unit model uses the same
principle, the mass of hydrogen consumed by the fuel cell group,
approach as those presented in Chamandoust et al. [3], as it aims
mH , is given by equation 6.
to evaluate the quantity of stored hydrogen in grams, mH , at 2cons
2
every simulation time step. The hydrogen mass available is mH = EL k4 PEL (5)
evaluated based on the stored quantity at the previous time step, 2prod
the electrical power supplied to the electrolyzer group PEL , and
k4 PFC
the electrical power demanded by the fuel cells group PFC . In the mH = (6)
2cons FC
proposed sizing strategy detailed in Section 2.5, the stored
quantity of hydrogen given by the HS unit model will be used as where EL is the efficiency of the electrolyzer group and FC is
the main optimization criteria.
the efficiency of the fuel cell group. In this work, these efficiency
According to the Gibbs relation [19], 30 g of H2 corresponds to values are taken from El-Emam and Özcan [20] and Felseghi et
an electrical power of 1 kW during one hour (one time step). al. [21] considering the Proton Exchange Membrane (PEM)
Based on this value, the conversion factor k4 = 30 g/kW is electrolyzer and fuel cell groups.
considered. The mass of hydrogen produced by the electrolyzer

ACADEMIA ENGINEERING 2023, 1 5 of 12

https://www.academia.edu/journals/3/academia_engineering/about https://doi.org/10.20935/AcadEng6115

The block diagram of the HS unit is given in Figure 7. In the The presented HS unit model, inspired by Chamandoust et al.
model, the power generated by the electrolyzer group, PEL , is [3], does not take into account some factors that could impact HS
considered positive and the power demanded by the fuel cell capabilities such as temperature or pressure effects for example.
group, PFC , is considered negative. Nevertheless, similar to the other energy unit models, the user
can straightforwardly modify the HS unit model to improve its
accuracy by considering some physical parameters and possibly
its own characterization data.

Figure 7 • Hydrogen storage unit block diagram.

2.4. Power management unit battery group if the SoC is not at its maximal value, here, SoC <
This subsection presents the proposed power management 0.9 (State 1 or State 2); otherwise the electricity will be used to
strategy in order to optimize the sizing of the different energy produce hydrogen using the electrolyzer group (State 3). It
equipment. First, the basic principle of the method is exposed. should be noted that there are two impossible combinations,
Then, different ways to implement the Power Management Unit highlighted in red in Table 2, that would lead to SoC > 0.9 but
(PMU) in the Matlab/Simulink software are given. SoC < 0.2. Depending on the resulting state value (0, 1, 2 or 3),
∆P is either attributed to the battery group, to the electrolyzer
2.4.1. Methodology group, or to the fuel cell group as illustrated in Figure 8 which
gives the block diagram of the proposed PMU.
The PMU aims to evaluate at every time step the difference of
electrical power, P , between the power produced by the PV Table 2 • Proposed management strategy using a state machine
unit, PPV , and the power demanded by the district, Pload , as
State
given by equation 7. ∆P > 0 SoC > 0.2 SoC > 0.9 ∆P=
(sum)
P = PPV − Pload (7) 0 0 0 0 PFC

Then, depending on the sign of ∆P (i.e. an excess or deficit of 0 0 1 NA NA

electrical energy) and the SoC returned by the battery unit model, 0 1 0 1 PB
the power difference ∆P is sent either to the battery, or to the
0 1 1 2 PB
electrolyzer or to the fuel cell group. This strategy is implemented
using a state machine as summarized in Table 2. In the case of 1 0 0 1 PB
electrical energy deficit (∆P < 0), it is preferable to use the battery
1 0 1 NA NA
group (due to higher electrical energy conversion efficiency) if the
SoC is sufficient, here, SoC > 0.2 (State 1 or State 2); otherwise 1 1 0 2 PB
the fuel cell group will be used (State 0). In the case of an excess 1 1 1 3 PEL
of PV production (∆P > 0), it is chosen to preferably charge the

Figure 8 • Proposed PMU block diagram.

ACADEMIA ENGINEERING 2023, 1 6 of 12

https://www.academia.edu/journals/3/academia_engineering/about https://doi.org/10.20935/AcadEng6115

switch conditions. Then, a “Matlab Function” block which

2.4.2. Matlab/Simulink implementation
includes Matlab code is performed to specify the actions
There are many possible ways to implement the previously according to the state value (see Figure 8). This solution is
described PMU in the Matlab/Simulink software. Here, three robust and is almost instantaneous to simulate. However, this
different possibilities, tested by the authors, will be presented method requires the user to be familiar with the Matlab code.
and will give the same results. The first implementation, Another drawback is that the modification of the power
presented in Figure 9, is based on the diagram presented in management strategy will require the user to modify at least the
Figure 8. The state machine is carried out using “Switch” blocks Matlab code and potentially all the block architectures.
that commutate between 0 and 1 states according to defined

Figure 9 • PMU implementation in Matlab/Simulink using switches and Matlab code.

A second proposed implementation of the PMU is by using a 2.5. System sizing methodology
“Truth Table” block. This can be done by a direct implementation In this subsection, an iterative process to optimize the size of the
of Table 2 as shown in Figure 10. The “Condition Table” will energy equipment is suggested. In contrast to existing software
return actions defined in the “Action Table” according to the state solutions such as HOMER Pro or H2RES which are mainly
value. In comparison with the previous solution, this method driven by economics, the proposed sizing method is only driven
does not require the user to have specific knowledge in Matlab by energy requirements. This could be an attractive methodology
code. As the previously proposed implementation, the simulation especially in the case of standalone microgrids. The sizing results
time is not increased by this method. obtained with the proposed strategy are presented in the Section
Finally, the PMU can be implemented using the “Stateflow” 3. It should be noted that the following optimization strategy is
toolbox as shown in Figure 11 using a “Chart” block. This not unique and shall be compared in terms of performance to a
solution gives a graphic view of the different states and set of other solutions in future work.
transitions as well as an easily adaptive simulation block to the The sizing of the studied PV-battery-hydrogen system depends
user. Furthermore, when increasing the studied system on three parameters that must be optimized:
complexity, the state flow methodology will become the most
convenient implementation. Another advantage in the software • The number of PV panels, nPV
is that the user is able to observe the switching between states
while running the simulation, which could be useful for • The number of battery cells, ncell
debugging. However, the simulation time is notably lengthened
• The maximal quantity of hydrogen to store, mH
in most cases which constitutes a major drawback in the case of 2max
the implementation of the optimal sizing method that is
presented in the next subsection.

ACADEMIA ENGINEERING 2023, 1 7 of 12

https://www.academia.edu/journals/3/academia_engineering/about https://doi.org/10.20935/AcadEng6115

Figure 10 • PMU implementation in Matlab/Simulink using a truth table.

Figure 11 • PMU implementation in Matlab/Simulink using state flow with PB: the battery unit power, PEL: the electrolyzer group
power, and PFC: the fuel cell group power.

The iterative process starts by setting an initial value for nPV the end of the year ( mH
2
( 8,760 ) ), leading to a mass of hydrogen
according to the maximal load demand Pload determined variation coefficient,  as given by equation 8.
max
previously. Then, the ncell value is calculated according to the mH = mH
2 2
( 8,760 ) − mH2 ( 0 ) (8)
maximal PV production, PPV , to ensure that the maximal
max
If mH = 0, then the energy system is optimized. Else, nPV is
2
admissible cell current I cell is not exceeded. The simulation
max either increased or decreased and the algorithm process is
is launched and the algorithm is completed to compare the stored repeated. The stopping criterion ( mH = 0 ) can be adapted to
hydrogen quantity at the beginning of the year ( mH ( 0 ) ) and at 2
2
the user needs in terms of remaining hydrogen mass at the end
of the year. Figure 12 shows a schematic representation of the

ACADEMIA ENGINEERING 2023, 1 8 of 12

https://www.academia.edu/journals/3/academia_engineering/about https://doi.org/10.20935/AcadEng6115

proposed sizing method. In Figure 12, Gmax corresponds to the • Initial SoC: 0.5
ratio of the maximal solar irradiance in W/m² obtained during • Initial stored mass of hydrogen: 1,200 kg
the year and the reference value of 1,000 W/m².
The sizing methodology presented in Figure 12 is then applied.
The convergence of the proposed iterative process is highlighted
3. Results in Figure 13 which gives the mass of hydrogen difference
The simulations are conducted over one year starting by the between the beginning and end of one year, ΔmH , according to
winter season. It has been verified by authors that there is no 2

impact on the sizing results if the user starts the simulation by the number of PV panels, nPV . The optimal number of PV
another season than winter. However, careful consideration panels, leading to ΔmH = 0 , is 2,095 for the studied system.
must be taken with regard to initial values. If the simulation 2

starts with the winter season, because of low global irradiance, an Moreover, it has been observed that the initial value of nPV
important initial stored mass of hydrogen is needed for correct (1,050) is quite far from the optimal value (2,095) for this case
operation of the power system. Thus, in this work, the initial study which suggests that future work shall focus on this initial
values listed below are set: value formulae to improve the convergence time of the algorithm.

Figure 12 • Schematic representation of the proposed system sizing strategy.

Figure 13 • Convergence of the sizing method for the case study.

Figure 14 shows the yearly evolution of the amount of stored production is important and the battery group is well charged. In
hydrogen. From Figure 14, it can be seen that the stored hy- spring and fall, the stock of hydrogen still increases but with a
drogen quantity quickly decreases in winter because of the high clearly lower slope, which suggests that the balance between
energy required for the storage, and therefore a high hydrogen produced and demanded electricity is more effective during these
consumption by the fuel cells. In contrast, the stored hydrogen seasons. These results are reasonable and expected in the case of
quantity quickly increases during summer since the PV a standalone microgrid supplied by solar panels.

ACADEMIA ENGINEERING 2023, 1 9 of 12

https://www.academia.edu/journals/3/academia_engineering/about https://doi.org/10.20935/AcadEng6115

Figure 14 • Simulated stored mass of hydrogen over one year after optimization.

Figure 15 gives the yearly evolution of the power of the battery main results using the proposed sizing which are the rated power
group, the current of a single battery cell, the power of the of the PV unit, PPV , the size of the hydrogen tank, mH ,
max 2max
electrolyzer group, and the power of the fuel cell group. Figure 16
shows the electrical power evolution over a day in each energy unit the rated power of the battery unit, PB , the rated power of the
max
for the winter and summer seasons. These results are obtained for
a maximal PV unit power PPV of 335 kW. electrolyzer group, PEL , the rated power of the fuel cell group,
max
max
PFC , the number of PV panels, nPV , and the number of
The power profiles depend on the season, as highlighted in max

Figures 15 and 16. As expected, there is no power in the battery cells, ncell . The total area occupied by the solar panels is
electrolyzer group in winter because of the lack of PV production. estimated to be 2,305 m².
At the same season, the fuel cell power is high. A contrasting
behavior can be observed in summer where the electrolyzer Table 3 • Main simulation results
group power is high and the fuel cell group power is low. As
PPV mH PB PEL PFC nPV ncell
shown in Figure 16, the daily electrical power in the battery max 2max max max max
group has positive and negative values, meaning that the battery (kW) (kW) (kW) (kW)
(kg)
group is charging and discharging every day during the year. As
shown in Figure 15b, it should also be verified that the battery 335 1,200 250 230 110 2,095 1,085
cell current, I cell , does not exceed the maximal specified
charging or discharging current of 25 A. Table 3 summarizes the

Figure 15 • Simulation results over one year: (a) battery group power, (b) single battery cell current, (c) electrolyzer group power, and
(d) fuel cell group power.

ACADEMIA ENGINEERING 2023, 1 10 of 12

https://www.academia.edu/journals/3/academia_engineering/about https://doi.org/10.20935/AcadEng6115

Figure 16 • Daily simulated power profiles (a) in winter and (b) in summer.

4. Conclusion It should be noted that the management of thermal loads have

not been taken into account in this work. Indeed, various low-
This work has presented the development of a helpful fully carbon heating solutions could be considered such as electric
accessible and customable power system sizing tool for heaters, heat pump, hydrogen boiler, etc. The proposed power
standalone hydrogen-based microgrids which is not directly sizing tool can also be of great help to determine the best solution
driven by costs. Thus, the user can obtain optimal initial values for a specific application. This study will be a part of future work.
for equipment capacities in its application before performing Future work shall also focus on the implementation of economic
techno-economic analysis. constraints and its influence on the power system sizing. Adding
A straightforward state machine has been proposed to drive the other energy sources such as wind power should be considered as
power flows at every hour of the year depending on load demand, well. Finally, the integration of environmental criteria and life
PV production, battery group SoC, and hydrogen availability. cycle analysis with the sizing strategy should be considered.
Then, an optimal sizing strategy is proposed and implemented to
set the PV and battery storage capacities in order to have the
same amount of stored hydrogen at the beginning and the end of Acknowledgments
every year. Consequently, the sizing method only relies on a This research work has been carried out and tested within the
proper energy management. The tool has been applied to a case frame work of the innovative pedagogical activity “Weeks of
study, consisting of a PV-battery-hydrogen power system, to Engineering EXperience (WEEX)” at the engineering school
demonstrate its effectiveness. It provides a deep view in the “Ecole Centrale de Lyon”.
power exchange in the system hour by hour on a yearly basis,
which could be of great help for energy designers. Moreover, the
obtained results conform to the expectations for a standalone Funding
microgrid supplied by solar panels.
This work was supported by the French engineering school
Furthermore, the proposed simulation tool presents various “Ecole Centrale de Lyon”.
advantages:

• Comfortable implementation: the tool is based on mathe- Author contributions

matical models. In this work, it has been implemented in
the Matlab/Simulink software but other simulation envi- All authors contributed equally to this research work. L. Pace
ronments can be used. (LP), A. Zouaghi (AZ), and M. Beley (MB) work in the field of
electrical engineering while N. Haddour is specialized in
• Easy model improvement: the energy unit models electrochemistry. Conceptualization and methodology: LP, AZ,
presented in this study can be straightforwardly improved MB, and NH. Simulation results: LP. Formal analysis and
to better fit the user needs. For example, the temperature validation: LP, AZ, MB, and NH. Project administration: LP.
and pressure conditions or aging could be taken into Writing-original draft preparation: LP.
account. The efficiency of the associated power electronics
could be considered and added.

• Adaptability (flexibility): The power management and Conflict of interest

sizing strategy can be applied to other power architectures. The authors declare no conflict of interest.
For example, WT could have been integrated to the case
study based on the model presented in the study by
Zahboune et al. [9]. Moreover, the tool could be used to size Data availability statement
a hybrid H2-battery storage system for e-mobility applica-
Data supporting these findings are available within the article or
tions. In this specific last case, no energy source would be
upon request.
considered.

ACADEMIA ENGINEERING 2023, 1 11 of 12

https://www.academia.edu/journals/3/academia_engineering/about https://doi.org/10.20935/AcadEng6115

Institutional review board statement 5. Dawood F, Shafiullah G, Anda M. Stand-alone microgrid

with 100% renewable energy: a case study with hybrid solar
Not applicable. PV-battery-hydrogen. Sustainability. 2020;12(5):2047.

6. Batiyah S, Sharma R, Abdelwahed S, Zohrabi N. An MPC-

Informed consent statement based power management of standalone DC microgrid with
energy storage. Int J Electr Power Energy Syst. 2020;120:
Not applicable.
105949.

7. Cecilia A, Carroquino J, Roda V, Costa-Castelló R, Barreras

Sample availability F. Optimal energy management in a standalone microgrid,
The author(s) declare that no physical samples were used in the with photovoltaic generation, short-term storage, and
study. hydrogen production. Energies. 2020;20;13(6):1454.

8. Diab AAZ, El-Rifaie AM, Zaky MM, Tolba MA. Optimal

sizing of stand-alone microgrids based on recent metaheu-
Additional information ristic algorithms. Mathematics. 2022;10(1):140.
Received: 2023-04-28 9. Zahboune H, et al. Optimal hybrid renewable energy design
Accepted: 2023-07-27 in autonomous system using Modified Electric System
Published: 2023-09-12 Cascade Analysis and Homer software. Energy Convers
Manag. 2016;126:909–22.
Academia Engineering papers should be cited as Academia
Engineering 2023, ISSN pending, https://doi.org/10.20935/ 10. HOMER Pro microgrid software [Internet]. Colorado, USA:
AcadEng6115. The journal’s official abbreviation is Acad. Engg. UL Solutions; [cited 2023 Aug 8]. Available from: https://
www.homerenergy.com/products/pro/index.html

11. H2RES - Energy System Modelling Software [Internet].

Publisher’s note Zagreb, Croatia: Faculty of Mechanical Engineering and
Academia.edu stays neutral with regard to jurisdictional claims Naval Architecture; [cited 2023 Aug 8]. Available from:
in published maps and institutional affiliations. All claims https://h2res.org/
expressed in this article are solely those of the authors and do not 12. Khalil L, et al. Optimization and designing of hybrid power sys-
necessarily represent those of their affiliated organizations, or tem using HOMER pro. Mater Today: Proc. 2021;47:S110–5.
those of the publisher, the editors, and the reviewers. Any
product that may be evaluated in this article, or claim that may 13. Kalamaras E, et al. Techno-economic assessment of a hybrid
be made by its manufacturer, is not guaranteed or endorsed by off-grid DC system for combined heat and power generation
the publisher. in remote islands. Energy Procedia. 2019;158:6315–20.

14. Krajačić G, Duić N, Carvalho M da G. H2RES, Energy

planning tool for island energy systems – the case of the
Copyright Island of Mljet☆. Int J Hydrog Energy. 2009;34(16):
© 2023 copyright by the authors. This article is an open access 7015–26.
article distributed under the terms and conditions of the
15. ADEME - Agir pour la Transition [Internet]. [cited 2023 Aug
Creative Commons Attribution (CC BY) license (https://
9]. Available from: https://agirpourlatransition.ademe.fr/
creativecommons.org/licenses/by/4.0/).
16. Hello Watt [Internet]; [cited 2023 Aug 9]. Available from:
https://www.hellowatt.fr/
References
17. Global Solar Atlas [Internet]. [cited 2023 Apr 19]. Available
1. Siddaiah R, Saini RP. A review on planning, configurations, from: https://globalsolaratlas.info/map
modeling and optimization techniques of hybrid renewable
energy systems for off grid applications. Renew Sust Energ 18. Ultimatron JDG12-100 [Internet]. [cited 2023 Apr 19].
Rev. 2016;58:376–96. Available from: https://www.batteries44.com/index.php?
controller=attachment&id_attachment=222
2. Mbungu NT, et al. Control and estimation techniques
applied to smart microgrids: a review. Renew Sust Energ 19. Khotseng L. Fuel cell thermodynamics. In: Vizureanu P,
Rev. 2023;179:113251. editor. Thermodynamics and energy engineering [Internet].
IntechOpen; 2020 [cited 2023 Apr 19]. Available from:
3. Chamandoust H, Hashemi A, Bahramara S. Energy manage- https://www.intechopen.com/books/thermodynamics-
ment of a smart autonomous electrical grid with a hydrogen and-energy-engineering/fuel-cell-thermodynamics
storage system. Int J Hydrog Energy. 2021;46(34):
17608–26. 20. El-Emam RS, Özcan H. Comprehensive review on the
techno-economics of sustainable large-scale clean hydrogen
4. Karamov DN, Ilyushin PV, Suslov KV. Electrification of rural production. J Clean Prod. 2019;220:593–609.
remote areas using renewable energy sources: literature
review. Energies. 2022;15(16):5881. 21. Felseghi RA, Carcadea E, Raboaca MS, Trufin CN, Filote C.
Hydrogen fuel cell technology for the sustainable future of
stationary applications. Energies. 2019;12(23):4593.

ACADEMIA ENGINEERING 2023, 1 12 of 12