eTransportation 14 (2022) 100198

Contents lists available at ScienceDirect

eTransportation
journal homepage: www.journals.elsevier.com/etransportation

Energy management of a multi-battery system for renewable-based high

power EV charging
Jan Engelhardt, Jan Martin Zepter ∗, Tatiana Gabderakhmanova, Mattia Marinelli
Department of Wind and Energy Systems, Technical University of Denmark (DTU), Frederiksborgvej 399, 4000 Roskilde, Denmark

ARTICLE INFO ABSTRACT

Keywords: Hybrid fast charging stations with battery storage and local renewable generation can facilitate low-carbon
Battery energy storage system electric vehicle (EV) charging, while reducing the stress on the distribution grid. This paper proposes an energy
DC microgrid management system (EMS) for a novel multi-battery design that directly connects its strings to other DC
Electric vehicles
components through a busbar matrix without the need for interfacing power converters. The EMS has two
Energy management
primary control tasks: (i) allocating strings to other DC microgrid components, in this case a photovoltaic
Fast charging
(PV) system, two EV fast chargers, and a grid-tie inverter, and (ii) managing the energy exchange with the
local distribution grid. For the grid exchange, a basic droop control is compared with an enhanced control that
introduces a deadband with dynamic limits, which are adjusted based on PV energy forecasts. The performance
of the EMS is extensively assessed by Monte Carlo simulations generating a vast number of random scenarios
based on empirical EV fast charging data. The tests are performed for various utilisation levels of the chargers
(EVs/day) and actual PV measurements from different seasons in Denmark. The numerical analyses show that
the enhanced control increases self-sufficiency by reducing grid exchange, and decreases the number of battery
cycles. For a utilisation level of 20 EVs/day in summer, the enhanced control achieves a self-sufficiency of
87.3% compared with 73.8% in the base control. In winter, with significantly lower PV production, these
values drop to 9.3% and 9.1%, respectively. The corresponding annual battery cycles are estimated at 399 for
the enhanced and 456 for the base control. However, the enhanced control operates the battery closer to its
charge limits, which may accelerate calendar ageing.

1. Introduction commonly used as a metric to assess what percentage of the consump-

tion is covered by local generation [11]. Conventionally, the units in
The deployment of fast charging stations is facilitating the wide- a hybrid system are connected via power converters to a common
spread adoption of electric vehicles (EVs) as they address typical con- bus [12,13]. The converters are necessary to convert the voltage levels
cerns of potential customers such as range anxiety and long charging of the individual units to the bus voltage, and to actively control the
times [1,2]. At the same time, the electrification of the transport power flow [14,15].
sector challenges the operation of electric grids as EVs increase both Several configurations of this conventional design have been pre-
the overall energy consumption and the peak load demand [3,4]. In sented in the literature. For instance, Vermaak et al. [16] propose a
particular, fast charging stations with more than 50 kW can stress hybrid system comprising photovoltaic (PV) and wind installations, as
distribution grids to their limits and are only feasible where these are
well as a stationary battery energy storage system (BESS) for charging
sufficiently robust [5,6]. Reinforcing existing grid infrastructure entails
electric vehicles in areas without grid connection. Cunha et al. [17] fo-
high investment costs, and the associated long-term planning may delay
cus their investigation on vanadium redox flow batteries for providing
the necessary installation of fast chargers [7].
peak shaving capability in fast charging stations. The value of energy
In the past years, there has been an increasing interest in equipping
storage in a fast charging station for electric buses was highlighted
fast chargers with stationary battery systems that serve as a buffer
during high power charging [8]. The combination of EV chargers, bat- by Ding et al. [18], showing that storage contributed to a total cost
teries, and renewable energy sources (RES) in a hybrid system further reduction of 22.85%. Hafez et al. [19] dimensioned a microgrid with
allows to facilitate the local usage of renewable energy and make EV integrated chargers, renewable energy sources, and grid connection to
chargers to a certain degree self-sufficient [9,10]. Self-sufficiency is reduce life cycle costs for an assumed daily number of 20 EVs. The fixed

∗ Corresponding author.
E-mail addresses: [email protected] (J. Engelhardt), [email protected] (J.M. Zepter), [email protected] (T. Gabderakhmanova), [email protected] (M. Marinelli).

https://doi.org/10.1016/j.etran.2022.100198
Received 18 March 2022; Received in revised form 4 August 2022; Accepted 4 August 2022
Available online 11 August 2022
2590-1168/© 2022 The Author(s). Published by Elsevier B.V. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).
J. Engelhardt et al. eTransportation 14 (2022) 100198

strings and the other components: a PV system, two EV fast chargers,

Nomenclature and a grid-tie inverter [32]. To fulfil EV charging with a high share
𝛼EVs Success rate of EV charging events of renewable energy, an advanced energy management system has
th
𝐶str,out Thermal capacitance of a string to be designed to incorporate the (partly contradicting) requirements
of all system components. While energy management systems for the
𝐸cap,str Energy capacity of a string
conventional setup with a single bus have been previously presented in
𝐸grid,ex Total energy exported to the grid
above-mentioned literature, these cannot be applied to the new system
𝐸grid,im Total energy imported to the grid
due to its fundamentally different layout. Since the components are not
𝐸PV,fc Forecasted PV energy connected to the same bus but can be individually coupled through a
𝑛cyc,str Number of full cycles of a string busbar matrix, the energy management has the key task to allocate each
𝑁EVs,ch Number of successful EV charging events battery string to only one other system component at a time. For the
𝑁EVs Number of EVs arriving at the chargers prototype system at hand, no energy management strategies have to
𝑛str Number of strings date been developed or explored in the literature.
𝑃aux Auxiliary power Hence, the present work breaks fresh ground by proposing a control
𝑃grid,ex Export power at point of grid connection approach for this novel system design. The contributions of this paper
are summarised as follows:
𝑃grid,im Import power at point of grid connection
𝑃grid Power at point of grid connection 1. An electro-thermal model of the hybrid charging system is in-
𝑃inv,set Proposed inverter power setpoint troduced as a framework for developing and testing energy
𝑃inv Inverter power management strategies.
𝑃r,grid Rated grid power 2. A heuristic energy management system is proposed with two
𝑃r,inv Rated inverter power primary control tasks: First, allocating the battery strings to
other system components, and second, managing the energy
𝑃r,PV Rated PV power
exchange with the local distribution grid.
𝑃str Power of a string
3. The performance of two control strategies is extensively assessed
𝑄th Thermal input of a string
str,in by Monte Carlo simulations generating a vast number of random
𝑄thstr,out Thermal output of a string scenarios based on empirical EV charging behaviour. We suggest
𝑆𝑂𝐸DB,low Lower limit of state-of-energy deadband this method as a way to objectively evaluate designs and control
𝑆𝑂𝐸DB,up Upper limit of state-of-energy deadband algorithms for charging infrastructure.
𝑆𝑂𝐸ref Reference state-of-energy of the battery
The remainder of the paper is structured as follows: Section 2
𝑆𝑂𝐸str,init Initial state-of-energy of a string provides an overview of the system under investigation. Section 3
𝑆𝑂𝐸str State-of-energy of a string presents the model, the applied energy management strategies, as well
𝑡0 Start time of simulation as the case study. Section 4 summarises and discusses the numerical
𝑡end End time of simulation results, followed by concluding remarks in Section 5.
𝑡PV,fc Time horizon for the rolling PV forecast
𝑇str Temperature of a string 2. System overview

Fig. 1 provides a high-level overview of the hybrid system, com-

prising the BESS, two 175 kW EV fast chargers, a 61 kWp PV system,
charging demand profile was based on logged drive cycles of a 2013 and a 66 kW grid-tied inverter. The battery, as a power and energy
Chevrolet Volt with 16 kWh battery capacity. However, for studying buffer, enables full usage of the power capability of both the PV system
the impact of EV charging on the power system, the uncertainty and and EV chargers, despite the grid capacity being limited to only 43 kW.
diversity of charging profiles should be considered [20]. Domínguez- The battery system is hence able to facilitate high power EV charging
Navarro et al. [21] address this point by applying a genetic algorithm at two outlets up to 175 kW at a significantly lower grid rating. The
to optimise the installation and operation of a fast charging station for grid connection point is formed by the AC side of the inverter and
uncertain EV demand. the auxiliary (Aux). The auxiliary includes the supply of all electronic
Generally, the performance of hybrid systems depends on the en- devices, as well as the thermal management system comprising a
ergy management system (EMS), which is responsible for planning, heating, ventilation, and air conditioning system. The energy storage
monitoring, and controlling the power flow between different units, consists of 𝑛str = 3 battery strings of 104 kWh each, which can be
as well as the energy level of storage systems. Energy management independently allocated through a busbar matrix to the other power
strategies can consist of basic algorithms where the battery is recharged components of the system. The battery strings, the inverter, the busbar
whenever excess power is available [16] or target specific objectives matrix, and all auxiliary systems are placed inside a standard 10 ft.
while meeting the EV demand, such as minimising energy costs [22– container that isolates these units from the outdoor environment. Each
24], reducing the grid impact [25–27], as well as decreasing losses on of the strings is equipped with a string controller (SC), monitoring the
system level [28] or in the grid [29]. cell states and managing the cell configuration to meet the voltage
Within the H2020 InsulaE project we explore the benefits of a novel requirements of the connected component. Relevant system states, such
fast charging station design for decarbonising the transport sector. A as the state of energy (SOE) of the string, are transmitted to the battery
prototype of this design has been recently installed on the Danish island management system (BMS). The BMS is the central control unit of the
of Bornholm and is currently accessible for residents and tourists. The BESS responsible for communicating with other units in the container.
core of the prototype is formed by an innovative battery system with It further serves as a protection system and disconnects strings if they
reconfigurable topology, providing the intrinsic capability to adapt its are operated outside their safe operating area.
voltage level during operation [30,31]. The battery consists of subsys- The task of managing the stored energy to fulfil the needs of all
tems (strings) that can be directly coupled with other DC components system components is taken by the EMS. On the one hand, the EMS
through a busbar matrix without the need for any power converters must ensure that the charge levels of the strings are high enough
or a common DC bus. This makes the BESS the central entity of the to allow also for consecutive charging of EVs. On the other, it must
hybrid system, responsible for controlling the power flow between its take action if the SOE becomes so high that PV energy could not

2
J. Engelhardt et al. eTransportation 14 (2022) 100198

Fig. 1. Overview of the hybrid system comprising a multi-battery storage system, a grid-tie inverter, a PV installation, and two EV fast chargers.

be effectively captured. To achieve this, the EMS has two primary linear power dependency of 2% of the inverter power is caused by
control tasks, firstly allocating the battery strings to respective power forward voltage drops and linear current dependent switching losses.
components, and secondly managing the energy exchange with the A thermal model is used to estimate the temperatures of the in-
distribution grid. The EMS communicates its decisions to the BMS, dividual components in the battery system based on their thermal
which updates the inverter setpoint and the connection states of the characteristics, heat loads, and the effects of the cooling system (venti-
strings. lation system with heat pump). The temperature of each battery string
is calculated as
3. Methodology 𝑄th
str,in
− 𝑄th
str,out
𝑇str = 𝑑𝑡, (2)
∫ th
𝐶str
The modelling of the hybrid system comprises three layers. First,
the physical model of the power and energy flows in the DC microgrid where 𝑄th str,in
is the total thermal input, 𝑄th
str,out the thermal output, and
is presented. Second, the rationale of the energy management system th
𝐶str (1.232 MJ∕K) the thermal capacitance of the string. Thermal inputs
on how to calculate the inverter setpoint and the string allocation is include power losses during charging and discharging, string auxiliary
introduced. Third, the Monte Carlo simulation structure and associated consumption, and heat transfer from neighbouring components with
data are presented for the case study. higher temperature. Thermal outputs include thermal energy removed
by the cooling fans and heat transfer to neighbouring components with
3.1. Physical multi-battery model lower temperature. The same principle is applied to the control rack
comprising the inverter and the container auxiliaries (controllers, cool-
The system shown in Fig. 1 is modelled on a power and energy ing system) with a thermal capacitance of 79.8 kJ∕K. Heat transfer also
level in Matlab Simulink to create a framework for developing and test- occurs through the container walls to the outside environment, with the
ing energy management strategies. The sign convention of the power rate depending on the temperature difference between the inside and
flows is defined so that consumption (EV charging, grid export, losses, outside of the container (ambient temperature). Furthermore, a heat
auxiliary consumption) is represented by positive power values, while pump is used to actively cool the air of the ventilation system, thus
energy intake (PV generation, grid import) is represented by negative removing heat from the container. A detailed description of the thermal
power values. Hence, the SOE of a string is calculated as properties of the system can be found in [34]. The power consumption
of all auxiliaries including the thermal management system affects the
1
𝑆𝑂𝐸str = − 𝑃 (𝑡) 𝑑𝑡 + 𝑆𝑂𝐸str,init , (1) power that can be exported/imported through the inverter, as the grid
𝐸cap,str ∫ str
rating must not be exceeded. The power at the grid connection point
where 𝑃str (𝑡) is the string power, 𝐸cap,str the nominal energy capacity 𝑃grid (𝑡) is given as
of the string, and 𝑆𝑂𝐸str,init the initial charge level at the beginning
of the simulation. The strings are operated between SOE limits of 10% 𝑃grid (𝑡) = 𝑃inv (𝑡) − 𝑃aux (𝑡), (3)
and 90% to avoid under- and overcharging and stay within their linear where 𝑃inv (𝑡) is the inverter and 𝑃aux (𝑡) the auxiliary power. The power
voltage region. When crossing these limits, the BMS disconnects the of the PV system and the two EV chargers are considered as time series
string from the currently assigned component, regardless of what the inputs to the model, which will be further addressed in Section 3.3.
EMS requests.
During charging and discharging of the battery strings power losses 3.2. Energy management system
occur that decrease the efficiency of the system. The reconfigurable
strings were characterised in detail in a previous study [31], which The EMS is responsible for controlling the energy exchange with
allowed to derive the efficiency characteristic as a function of the string the grid and for allocating the three strings to the different power
power. The string efficiency shows a linear decrease with increasing components (EVs, PV, inverter) of the hybrid system. The following
power for both charging and discharging, with 100% at 0 kW, and section proposes heuristic control concepts for each of these tasks.
90% at ± 120 kW. This trend is mainly driven by the Joule losses in
cells and semiconductors, which increase with power. The inverter was 3.2.1. Inverter setpoint
tested as part of the TOPChargE project [33]. The inverter efficiency is The inverter is used to actively control the battery SOE by exporting
modelled with two loss components. Constant losses of 100 W account or importing energy to or from the grid. We compare two different
for consumption of inverter auxiliary systems, such as controllers, strategies used for the inverter setpoint, which are in the following
measuring electronics, and drive circuits. Moreover, a loss fraction with denoted as ‘‘base’’ and ‘‘enhanced’’. In the base strategy, the inverter

3
J. Engelhardt et al. eTransportation 14 (2022) 100198

Fig. 2. Basic droop control for inverter power based on battery SOE. The enhanced
control introduces dynamic deadband limits depending on PV forecasts.

setpoint is determined by a symmetric droop control based on the

overall battery SOE, which is calculated as the average string SOE. As
shown in Fig. 2, the reference SOE of the droop is 50%, and minimum
and maximum grid power are reached at 30% and 70%, respectively.
The ‘‘enhanced’’ control is based on the same droop characteristic, but
introduces flexible upper and lower deadband limits that avoid grid
exchange around the reference SOE. These limits are adapted dynami-
cally according to a one-hour (𝑡PV,fc = 1 h) rolling energy forecast of the
PV production, 𝐸PV,fc . The upper SOE deadband limit aims at initiating
a battery discharge if the forthcoming PV production will bring the
battery to its upper charge limit, and is calculated as
𝐸PV,fc 𝑃r,PV
𝑆𝑂𝐸DB,up = 90% − ⋅ ∈ [50%, 90%], (4)
𝑛str ⋅ 𝐸cap,str 𝑃r,grid
𝑃r,PV
where the factor 𝑃r,grid
considers the different power ratings of the PV
system and the grid. The lower SOE deadband limit aims at avoiding
energy import from the grid in case of forthcoming PV production, and
is calculated as Fig. 3. Flow chart for the allocation of battery strings to the other system components.
40% − 30%
𝑆𝑂𝐸DB,low = 40% − ⋅ 𝐸PV,fc ∈ [30%, 40%]. (5)
𝑃r,grid ⋅ 𝑡PV,fc
The lower deadband limit has a minimum value of 30% to retain a 3.3. Case study
sufficient energy content for EV charging, and a maximum value of
40% to avoid unnecessary recharging when being in idle mode (no PV, This section presents the case study for investigating the perfor-
EVs). The proposed inverter power is communicated from the EMS to mance of the two proposed heuristic control strategies under different
the BMS. When updating the inverter setpoint, the BMS considers the model configurations. Specifically, the simulation structure, the con-
current auxiliary consumption to fulfil grid constraints (3). struction of EV fast charging profiles, and the origin of PV profiles are
introduced.
3.2.2. String allocation
The technical constraints of the system must be taken into account Simulation structure. The simulation model of the DC microgrid has
when designing an algorithm for the allocation of strings and units. been constructed in Matlab Simulink and is run in a simulation time
Technically the system can only allow for the connection of EVs, PV and step of one minute. The model simulates for time periods of 14 days
the inverter to battery stings, and they cannot be connected directly to in June, September, and November to account for the impact of sea-
each other. This means that e.g. PV power cannot be directly exported sonality of the PV production and different ambient temperatures.
to the grid, but has to be stored in a string first. This has to be consid- Furthermore, different frequencies of EV charging are investigated,
ered in the decision making, since PV power can only be utilised if the ranging from 1 to 30 EVs on average per day. Fig. 4 provides an
connected string has chargeable capacity. Besides technical constraints, overview of the simulation structure.
additional considerations are needed: PV production must be always For assessing the sensitivity in key performance indicators of the
harvested and incoming EVs must always be charged. To ensure a two proposed heuristic control strategies, Monte Carlo simulations are
successful EV charging session, the string serving the EV must not be performed to capture the uncertainty of EV charging conditional to
disconnected before charge completion. Fig. 3 describes the procedure
a certain probability distribution. This is done to extensively test the
for allocating the strings, considering the aforementioned constraints. A
energy management strategies under random conditions, and not with
re-calculation of the string allocation is initiated by one of the following
customised EV charging profiles that might favour the proposed con-
triggers:
trol. For the creation of randomised charging patterns, three dimensions
1. An EV connects or disconnects at one of the chargers. of input variations are considered: for each control strategy, each of
2. The PV system starts/stops producing (e.g. morning/evening or the three seasons, and each utilisation level of EV charging, 10,000
due to maintenance). simulation runs are performed for a two-week time horizon. To this
3. Timer of 10 min that checks if a string has an SOE of > 80% or end, the Monte Carlo simulations allow for testing the performance
< 20%. of the proposed control with a vast number of possible scenarios,
4. One of the strings reaches its operational limits. including also rare and extreme EV charging events. This approach
substantiates conclusions based on results with statistical significance.

4
J. Engelhardt et al. eTransportation 14 (2022) 100198

overlap. The queue management algorithm employed in the simulations

allocates arriving EVs to the two fast-chargers depending on their
availability. In case both chargers are occupied, arriving EVs queue to
be recharged. When one of the two chargers becomes available, the
algorithm considers a one-minute interval for park-out of the previous,
as well as park-in, connection, and authentication of the next EV.

PV profiles. The PV installation of the hybrid system is composed of 10

strings with 23 PV modules. The modules are of the kind Trina TSM-
265 with a rated peak power of 265 W. Since the installation of the
hybrid system in June 2021, real PV power measurements for June and
September are obtained at two-second time intervals. Subsequently,
these profiles were downsampled to a one-minute resolution in order to
match the time steps of the simulation. For the analysis in November,
data from the PV system are only available in hourly averages.
The ambient temperature progressions for the examined weeks have
Fig. 4. Overview of the simulation structure. For each control strategy, season, and been taken from the website of the Danish Meteorological Institute in
average number of EVs/day, the EMS is tested with 10,000 Monte Carlo runs that
an hourly resolution.
capture the uncertainty of EV charging. For each run, the charging demand at the two
fast-chargers is randomly generated using probability distributions for the charging Performance metrics. To achieve a quantitative assessment of the EMS
behaviour.
performance, a group of metrics was defined:

1. Success rate of EV charging: The success rate 𝛼EVs quantifies

EV fast charging profiles. The EV fast charging profiles are created based what percentage of the EVs achieved their charging goal at the
on probability distributions of the arrival time, charging duration, and fast-chargers and is calculated as
charged energy (Fig. 5). Data behind these distributions were collected 𝑁EVs,ch
from EV fast charging stations in the Netherlands with charging powers 𝛼EVs = , (6)
𝑁EVs
of up to 175 kW [35,36]. The authors processed and analysed data from
MultiTankCard, a service provider for mobility in the Netherlands. The where 𝑁EVs is the total number of EVs that arrived at the charger
published distributions are based on over 1 million charging sessions in the respective simulation run, and 𝑁EVs,ch is the number of
in 2019, providing a strong statistical representation of the charging EVs that achieved their charging goal. The charging goal of an
behaviour at fast charging stations. Based on the hourly data points EV is considered as achieved if more than 99% of the requested
provided by the above-mentioned publications, the present study used energy was charged.
spline fits to obtain the continuous probability distributions depicted 2. Number of cycles & average charge level: These quantities are
in Fig. 5. By drawing from these distributions, it is possible to derive important indicators for how the battery strings are used and
random average charging power profiles with different arrival and influence battery degradation. The number of cycles of a string
connection times. In doing so, no dependency of the charging duration 𝑛cyc,str is calculated as
and the charged energy is assumed. For those cases where a charg- 𝑡
∫𝑡 end |𝑃str (𝑡)|𝑑𝑡
ing power above 175 kW emerges, a new sample is taken from the 𝑛cyc,str = 0
, (7)
distributions. For each of the 10,000 Monte Carlo runs, the depicted 2 ⋅ 𝐸cap,str
probability distributions are employed to generate random EV power where 𝑡0 and 𝑡end are start and end time of the simulation,
profiles for the two fast-chargers (Fig. 4). With only a limited number respectively. Correspondingly, the number of battery cycles is
of chargers available, queue management is important when randomly defined as the mean of all string cycles. The average string
drawing from probability distributions as obtained charging times may charge level 𝑆𝑂𝐸 str is given as
𝑡
∫𝑡 end 𝑆𝑂𝐸str (𝑡)𝑑𝑡
0
𝑆𝑂𝐸 str = , (8)
𝑡end − 𝑡0
while the overall battery SOE is determined as the mean of all
string SOEs.
3. Grid exchange: The total amount of energy exported to and
imported from the grid indicates to what degree the hybrid
system interacts with the primary grid. The exported energy
𝐸grid,ex is calculated as
𝑡end
𝐸grid,ex = 𝑃grid,ex (𝑡) 𝑑𝑡, (9)
∫𝑡0
where 𝑃grid,ex is the power at the grid connection point when
exporting. The imported energy 𝐸grid,im is calculated as
𝑡end
𝐸grid,im = 𝑃grid,im (𝑡) 𝑑𝑡, (10)
∫ 𝑡0
where 𝑃grid,im is the power at the grid connection point when
importing.
4. Self-sufficiency: This metric defines what percentage of the
energy delivered to the EVs was provided by the local PV system.
Fig. 5. Probability distributions for arrival time, charging duration, and charged energy The calculation is based on [37], where the applied equations
at fast charging station up to 175 kW, based on data presented in [35,36]. are derived and explained in detail.

5
J. Engelhardt et al. eTransportation 14 (2022) 100198

These metrics are applied in each of the Monte Carlo runs, allowing for
an assessment of the proposed EMS in different scenarios with varying
levels of EV charging and PV energy.

4. Results and discussion

This section presents and discusses numerical results that were

obtained from the performed Monte Carlo simulations. In particular,
it presents the functioning of the string allocation, the EV charging
satisfaction, the battery usage, and key indicators of the system per-
formance. The section concludes with directions for further research.

4.1. String allocation

Fig. 6 demonstrates the string allocation during one of the Monte

Carlo runs with the base control for an exemplary time window on the
8th June 2021. The plot shows the power profiles of PV and inverter in
the top subplot, of the two EV chargers in the second subplot, as well
as the SOE and temperature progression of the battery strings in the
third and fourth subplot, respectively. During the given time window,
four EVs are charging with average power values of up to 134 kW. In
contrast, the inverter power varies only between −5.6 kW (import) and
24.3 kW (export), underlining the buffering effect of the BESS towards
the grid.
The figure further exemplifies the functioning of the algorithm
presented in Fig. 3, showcasing a set of exemplary triggers for the Fig. 6. Excerpt of the string allocation for the morning of 8th June 2021.
re-allocation of the battery strings. At time ⃝ 1 an EV connects to
Charger 1 requesting an average power of 72 kW for 20 min. The
string with highest SOE (String 1) is consequently assigned to that 4.2. EV charging
charger, while the string with lowest SOE (String 2) connects to the
PV, and the remaining string (String 3) links to the inverter. The Fig. 7 visualises the EV charging success rate as a result of the Monte
inverter setpoint is determined based on the mean SOE of the battery Carlo simulations comparing the two strategies with a varying number
strings, as depicted in Fig. 2. Since the mean SOE is 54% at time of EV charging instances ranging from 1 to 30 EVs on average per day.
⃝,
1 String 3 is being discharged from the inverter with 8.6 kW. At A single EV charging session is considered as successful if the charger
time ⃝ 2 the EV disconnects, which causes a re-allocation of the strings provided more than 99% of the requested energy. Although this is a
due to changed SOE levels. The string most responsible for mean SOE strict criterion for defining charging success, it gives an indication of
deviation (String 1) is now connected to the inverter, while the string the lower bound of how well charging requests are fulfilled. The EV
with the lowest SOE (String 3) is assigned to the PV system. At time success rate states what percentage of all EV sessions within the same
⃝,
3 String 3 reaches a SOE level of above 80%, since no EV connected
simulation run were successful. For instance, if 390 out of 420 EV
in the meantime. Hence, the control reacts in order to balance the sessions in a two-week simulation (30 EVs/day) were successful, the
EV success rate of this simulation is 92.9%.
SOEs of the battery strings. String 3, previously connected to the PV,
A Monte Carlo run with 10,000 individual simulations generates a
is consequently assigned to the inverter, and String 1 connects to the
statistical distribution of the EV success rate, summarised in the box
PV. The inverter discharges String 3 at around 20 kW, while the PV is
plots in Fig. 7. Specifically, each box-whisker combination represents
charging String 1 at power values around 45 kW. Time ⃝ 4 serves as an
the statistical distribution of one Monte Carlo run. The box indicates the
example of an EV connecting at Charger 2, because Charger 1 is already
interquartile range (25% to 75%) of the data, while the whiskers are
taken by another EV. Hence, at that time, String 3 is not available. As
set to 1.5 times this range. Outliers that are outside the whisker range
a consequence, String 1, as the remaining string with the highest SOE,
are indicated by single dots. The lines connecting the individual box
switches its connection from the inverter to Charger 2, while String 2
plots show the progression of the mean EV success rate as a function
remains connected to the PV system. With two EVs connected during
of the number of EVs/day. All in all, the data visualised in Fig. 7
daylight hours, the inverter has no option of being connected due to the show the output of 420,000 simulations (3 seasons, 7 EV utilisation
premise that PV production must always be used. The EV connecting to levels, 2 control strategies, 10,000 runs per scenario), which allows a
Charger 2 at time ⃝ 4 has a high average power demand of 134 kW for
statistically robust interpretation of the results.
16 min, representing a total charged energy of 36 kWh. Considering In June, the enhanced control enables EV charging with a high
the probability distributions displayed in Fig. 5, this charging event success rate that decreases to 94.20% if 30 EVs use the chargers each
ranks among the more intense cases. Thus, the SOE level of String 1 day. In comparison, the base control reaches an average success rate of
decreases during the charging process from 58% to 19%. The thermal 92.45% for the same number of EVs. In September, the enhanced con-
management continuously cools the strings and removes heat caused trol performs better until the threshold of 15 EVs/day, while the base
by Joule losses. During the high power EV charging between ⃝ 4 and ⃝ 5 control achieves a higher success rate beyond that point. In November,
the temperature of String 1 increases and reaches peak values around the base control outperforms the enhanced control, although both
42 ◦ C. After the EV disconnects from Charger 2 at ⃝,5 the temperature strategies have lower success rates than during the other months.
decays exponentially, while the string is being recharged by the PV The progression of the success rates can be explained by the average
system at around 45 kW. battery SOE for the different scenarios, as compared in Fig. 8. With a

6
J. Engelhardt et al. eTransportation 14 (2022) 100198

Fig. 7. EV charging success rate for the base and the enhanced control for varying EV charging frequencies and different seasons.

higher charge level, consecutive EV charging sessions are more likely the results, the two control strategies show opposing relations for cycle
to be provided with the requested energy. Hence, when comparing the and calendar ageing.
average battery SOEs of the two strategies, they show the same relation It should be noted that some studies show that average charge levels
as the EV success rate. Moreover, the premise of keeping strings and of above 70% accelerate calendar ageing significantly [39]. Hence, the
EV connected will always lead to a failed (not completed) charging enhanced control could be adjusted to decrease the upper deadband
session if the requested energy is larger than the energy content of limit in periods with little EV charging. However, long-term studies
the respective string — regardless of the overall SOE of the BESS. are necessary to confirm the impact of the charge levels on battery
Hence, methods should be reviewed that allow pausing the power degradation. During operation, the battery health can be assessed by
transfer without terminating the entire charging process, in order to the BMS, e.g. through incremental capacity analyses and weighted
swap the discharged string. This is expected to improve the success Kalman filters [40]. Yet, basic estimations for the expected cycle ageing
rate significantly. It should be noted that in the performed simulations can be based on the results shown in Fig. 8. Assuming the highest
a charging process is considered as unsuccessful if less than 99% of EV penetration of 30 EVs/day in all months, the average number of
the requested energy was delivered. Customers of fast charging stations cycles in a year is 566 for the base, and 516 for the enhanced control.
might define charging success less strict and hence be content with For 20 EVs/day, the annual cycles are 456 for the base, and 399 for
slightly lower delivered energy than requested. the enhanced control. The cycle life eventually depends on the cell
chemistry [41], and can reach up to 6000–8000 cycles in case of
lithium iron phosphate (LFP) cells [42]. Hence, a cycle life of 10 years
4.3. Battery usage
could be achieved even in case of high EV utilisation levels.

Capacity degradation of lithium-ion batteries is commonly deter- 4.4. System performance

mined by two effects: cycle and calendar ageing. While cycle ageing
occurs when a battery is charged and discharged, calendar ageing also To assess the buffering effect of the battery storage, the energy flows
takes place when a battery is unused, and is affected by battery SOE to and from the hybrid system can be compared. Fig. 9 illustrates daily
and temperature [38]. Hence, the number of full cycles and the average average values for grid import and export, PV energy, and EV charging
battery SOE are important indicators for the battery usage and as such demand, as a result of the performed Monte Carlo simulations. The
a suitable metric for evaluating EMS performance. As seen in Fig. 8, dashed grey line represents the potential PV energy in the respective
the base control entails higher numbers of cycles during all months, seasons, while the solid grey line depicts the increase in EV charging
compared with the enhanced control. In contrast, the enhanced control demand with an increasing number of EVs/day.
allows a higher battery SOE in summer and a lower battery SOE in For June and September, the enhanced control entails significantly
winter due to the flexible deadband limits. The battery SOE is defined less grid exchange, i.e. import and export, due to the incorporated
as the mean of the individual string SOEs given by Eq. (8). As seen from deadband that requires less immediate control action. For 30 EVs/day

Fig. 8. Impact of the base and the enhanced control on number of cycles and mean battery SOE for varying EV charging frequencies and different seasons.

7
J. Engelhardt et al. eTransportation 14 (2022) 100198

Fig. 9. Daily average values for grid import and export for the base and the enhanced control, as well as PV energy, and EV charging demand for varying EV charging frequencies
and different seasons.

Fig. 10. Self-sufficiency for the base and the enhanced control for varying EV charging frequencies and different seasons.

in June, the import and export from and to the grid can be decreased importing and exporting energy introduces economic considerations
by 65.0% and 31.3%, respectively, compared with the base control. In with the aim of reducing EV charging costs. Utilising historical EV
November, the PV production level is insignificant. Hence, the import charging behaviour on the demonstration site allows for an improved
in both the base and enhanced control strategies aligns increasingly management of the battery SOE to increase the EV charging success
with the requested EV energy. The lower EV charging success rate for rate. In this regard, a distinction between weekdays and weekends may
high numbers of EVs/day in November is also reflected in decreasing provide a valuable extension of the present work. For instance, Simolin
grid import in both control strategies. The premise to always utilise PV et al. [43] find differences in the use of chargers in terms of arrival
energy when available prevents the use of the inverter to import energy times and plug-in duration between weekdays and weekends for two
from the grid, in case two EVs charge simultaneously. Hence, in months shopping centres in Finland. However, these differences highly depend
with low PV potential this might cause a depletion of the battery and, on the location of the charging stations. Similarly, seasonality effects
consequently, a lower success rate of EV charging. can be considered to better reflect charging demand variations through-
Based on grid exchange, PV production, and EV consumption, the out the year [44]. Furthermore, the battery SOE can be included in
self-sufficiency of the hybrid system can be estimated. It is defined the decision making to reduce calendar ageing. The strict constraint of
as the percentage of EV consumption that was covered by the local
always utilising local PV production when available could be relaxed
PV production, taking into account the power losses of strings and
for achieving higher charging success rates by prioritising grid imports
inverter, as well as the auxiliary consumption. Fig. 10 shows the
when PV production is low. Conversely, varying conditions in the pri-
distributions of obtained self-sufficiency values for different average
mary grid such as dynamic power constraints and voltage fluctuations
numbers of EVs per day. For both control strategies, the self-sufficiency
may occur in weak distribution grids, and could be considered in the
generally decreases with an increasing number of EVs per day. The
control of the grid-tie inverter setpoint. Formulating these tasks as a
enhanced control strategy, incorporating a dynamic deadband based
multi-objective optimisation problem is one possible way to capture the
on the rolling PV forecast, obtains higher self-sufficiency values for
the months of June and September. In June, even with an average of mentioned points. In this regard, the presented heuristic control can
30 EV charging instances per day, the self-sufficiency ends for most serve as a benchmark to assess the performance of the optimisation.
of the simulation runs above 70% in case of the enhanced control. Besides improvements in the EMS, future studies could quantify
In November, due to low PV production, the two control strategies the technological advantages and disadvantages of the reconfigurable
equalise in terms of self-sufficiency as both are equivalently reliant on string design. The general absence of DC–DC converters might reduce
grid import to fulfil EV charging. the overall system costs and power losses compared to conventional sys-
tem designs. Furthermore, the reconfigurable topology could increase
4.5. Future considerations safety and reliability of the battery, as individual cells can be bypassed
during operation which helps to reduce unbalanced cell states such as
The presented EMS can be further improved by considering ad- charge level and temperature. On the other hand, the reconfigurable
ditional inputs in the decision making. Including energy prices for string architecture requires the BMS to perform complex control tasks,

8
J. Engelhardt et al. eTransportation 14 (2022) 100198

which also increase the computational effort. The demonstration ac- Acknowledgements
tivities on the Danish island of Bornholm offer the platform to test
the components in field, and are expected to provide more insights on This work has received funding from the H2020 Insulae project,
advantages and disadvantages of the battery design. Denmark under the Grant Agreement No. 824433, and from the IFD
The presented work may also be extended to other applications funded TOPChargE project, Denmark under the Grant Agreement No.
where hybrid charging stations based on this novel multi-battery sys- 9090-00035A.
tem can provide technical and economic benefits. These applications
comprise e.g. the charging of electric bus fleets [45], air mobility
References
entities [46] or electric vessels [47] that require high energy deliv-
ery in a short amount of time, potentially at remote locations. As a
[1] Brand C, Cluzel C, Anable J. Modeling the uptake of plug-in vehicles in a
complement to long-term grid upgrades, these scalable fast charging heterogeneous car market using a consumer segmentation approach. Transp Res
stations could be dimensioned for specific locations such as highways Part A: Policy Pract 2017;97:121–36. http://dx.doi.org/10.1016/j.tra.2017.01.
or city centres by optimising the number of strings, the cell rating and 017.
the inverter power. Multi-battery systems could hence contribute to [2] Zhang Q, Li H, Zhu L, Campana PE, Lu H, Wallin F, et al. Factors influencing the
economics of public charging infrastructures for EV – a review. Renew Sustain
facilitating sufficient charging infrastructure for a rapid electrification
Energy Rev 2018;94:500–9. http://dx.doi.org/10.1016/j.rser.2018.06.022.
of the transport sector.
[3] Yilmaz M, Krein PT. Review of the impact of vehicle-to-grid technologies
on distribution systems and utility interfaces. IEEE Trans Power Electron
5. Conclusion 2013;28(12):5673–89. http://dx.doi.org/10.1109/TPEL.2012.2227500.
[4] Calearo L, Thingvad A, Suzuki K, Marinelli M. Grid loading due to EV charging
Hybrid fast charging stations with incorporated battery storage and profiles based on pseudo-real driving pattern and user behavior. IEEE Trans
local renewable generation units can foster the decarbonisation of the Transp Electr 2019;5(3):683–94. http://dx.doi.org/10.1109/TTE.2019.2921854.
[5] Mauri G, Valsecchi A. Fast charging stations for electric vehicle: The impact
transport sector, while reducing the grid impact on a distribution level.
on the mv distribution grids of the milan metropolitan area. In: 2012 IEEE
The present study focuses its attention on a novel system architecture International Energy Conference and Exhibition. IEEE; 2012, p. 1055–9. http:
comprising a multi-string battery system with reconfigurable topology //dx.doi.org/10.1109/EnergyCon.2012.6347725.
that allows the direct coupling to other DC components through a [6] Dharmakeerthi CH, Mithulananthan N, Saha TK. Impact of electric vehicle fast
busbar matrix. This paper breaks fresh ground by proposing an energy charging on power system voltage stability. Int J Electr Power Energy Syst
management system for this prototype battery storage interfacing a 2014;57:241–9. http://dx.doi.org/10.1016/j.ijepes.2013.12.005.
[7] Unterluggauer T, Rich J, Andersen PB, Hashemi S. Electric vehicle charging
PV installation, a low-rated distribution grid connection, and two high
infrastructure planning for integrated transportation and power distribution net-
power EV fast-chargers. works: A review. ETransportation 2022;12:100163. http://dx.doi.org/10.1016/j.
The key task of the energy management system is to allocate each of etran.2022.100163.
the three battery strings to only one of the other four components at a [8] Mahfouz MM, Iravani MR. Grid-integration of battery-enabled DC fast charging
time. For the energy exchange with the grid, a base control is compared station for electric vehicles. IEEE Trans Energy Convers 2020;35(1):375–85.
with an enhanced control that includes PV forecasts in the decision http://dx.doi.org/10.1109/TEC.2019.2945293.
[9] Simpson M, Markel T. Plug-in electric vehicle fast charge station operational
making. The performance of these proposed strategies is assessed by
analysis with integrated renewables. In: International Battery, Hybrid and Fuel
Monte Carlo simulations with a vast number of EV charging scenarios Cell Electric Vehicle Symposium. May 6 – 9, California, Los Angeles; 2012.
that allow a statistically robust interpretation of the results. For an [10] Liu G, Xue Y, Chinthavali MS, Tomsovic K. Optimal sizing of PV and energy
average number of 20 EVs per day in summer, base and enhanced storage in an electric vehicle extreme fast charging station. In: 2020 IEEE Power
control fulfil the energy requests of the EVs in 97.5% and 98.8% of & Energy Society Innovative Smart Grid Technologies Conference. IEEE; 2020,
the charging instances, respectively. Yet, the enhanced control achieves p. 1–5. http://dx.doi.org/10.1109/ISGT45199.2020.9087792.
[11] Luthander R, Widén J, Nilsson D, Palm J. Photovoltaic self-consumption in
a significantly higher self-sufficiency of the hybrid system at 87.3%,
buildings: A review. Appl Energy 2015;142:80–94. http://dx.doi.org/10.1016/
while the base control yields 73.8%. In winter with significantly lower j.apenergy.2014.12.028.
PV production, however, the self-sufficiency drops to 9.3% (enhanced) [12] Sbordone D, Bertini I, Di Pietra B, Falvo MC, Genovese A, Martirano L. EV
and 9.1% (base). In November, the base control achieves a higher fast charging stations and energy storage technologies: A real implementation
success rate than the enhanced control, although both strategies still in the smart micro grid paradigm. Electr Power Syst Res 2015;120:96–108.
http://dx.doi.org/10.1016/j.epsr.2014.07.033.
complete 90.2% (enhanced) and 93% (base) of the charging processes,
[13] Tran VT, Islam MR, Muttaqi KM, Sutanto D. An efficient energy management
considering that they prioritise low local PV production over grid
approach for a solar-powered EV battery charging facility to support distribution
import. grids. IEEE Trans Ind Appl 2019;55(6):6517–26. http://dx.doi.org/10.1109/TIA.
All in all, the enhanced control reduces the number of battery 2019.2940923.
cycles, at the cost of higher average charge levels, compared with the [14] Khalid MR, Alam MS, Sarwar A, Jamil Asghar MS. A comprehensive review
base control. The ongoing demonstration activities within the InsulaE on electric vehicles charging infrastructures and their impacts on power-quality
of the utility grid. ETransportation 2019;1:100006. http://dx.doi.org/10.1016/j.
project invite for model improvements based on real EV charging data.
etran.2019.100006.
[15] Sharma G, Sood VK, Alam MS, Shariff SM. Comparison of common DC and
CRediT authorship contribution statement AC bus architectures for EV fast charging stations and impact on power quality.
ETransportation 2020;5:100066. http://dx.doi.org/10.1016/j.etran.2020.100066.
Jan Engelhardt: Conceptualization, Methodology, Software, For- [16] Vermaak HJ, Kusakana K. Design of a photovoltaic–wind charging station for
mal analysis, Investigation, Writing – original draft, Writing – re- small electric Tuk–tuk in D.R.Congo. Renew Energy 2014;67:40–5. http://dx.
view & editing, Visualization. Jan Martin Zepter: Conceptualization, doi.org/10.1016/j.renene.2013.11.019.
[17] Cunha A, Brito FP, Martins J, Rodrigues N, Monteiro V, Afonso JL, et al.
Methodology, Software, Formal analysis, Investigation, Writing – orig-
Assessment of the use of vanadium redox flow batteries for energy storage
inal draft, Writing – review & editing, Visualization. Tatiana Gabder- and fast charging of electric vehicles in gas stations. Energy 2016;115:1478–94.
akhmanova: Conceptualization, Writing – review & editing, Supervi- http://dx.doi.org/10.1016/j.energy.2016.02.118.
sion. Mattia Marinelli: Conceptualization, Writing – review & editing, [18] Ding H, Hu Z, Song Y. Value of the energy storage system in an electric bus
Supervision, Funding acquisition. fast charging station. Appl Energy 2015;157:630–9. http://dx.doi.org/10.1016/
j.apenergy.2015.01.058.
Declaration of competing interest [19] Hafez O, Bhattacharya K. Optimal design of electric vehicle charging stations
considering various energy resources. Renew Energy 2017;107:576–89. http:
//dx.doi.org/10.1016/j.renene.2017.01.066.
The authors declare that they have no known competing finan- [20] Calearo L, Marinelli M, Ziras C. A review of data sources for electric vehicle
cial interests or personal relationships that could have appeared to integration studies. Renew Sustain Energy Rev 2021;151:111518. http://dx.doi.
influence the work reported in this paper. org/10.1016/j.rser.2021.111518.

9
J. Engelhardt et al. eTransportation 14 (2022) 100198

[21] Domínguez-Navarro JA, Dufo-López R, Yusta-Loyo JM, Artal-Sevil JS, Bernal- [34] Lage JG. Design of energy management strategies for a reconfigurable battery
Agustín JL. Design of an electric vehicle fast-charging station with integration system coupled with an electric vehicle fast charger and photovoltaics. (Master’s
of renewable energy and storage systems. Int J Electr Power Energy Syst thesis), Kgs. Lyngby, Denmark: Technical University of Denmark; 2021.
2019;105:46–58. http://dx.doi.org/10.1016/j.ijepes.2018.08.001. [35] Wolbertus R, van den Hoed R. Electric vehicle fast charging needs in cities and
[22] Li D, Zouma A, Liao J-T, Yang H-T. An energy management strategy with along corridors. World Electr Veh J 2019;10(2):45. http://dx.doi.org/10.3390/
renewable energy and energy storage system for a large electric vehicle charging wevj10020045.
station. ETransportation 2020;6:100076. http://dx.doi.org/10.1016/j.etran.2020. [36] Wolbertus R, van den Hoed R. Fast charging systems for passenger elec-
100076.
tric vehicles. World Electr Veh J 2020;11(4):73. http://dx.doi.org/10.3390/
[23] Tushar MHK, Assi C, Maier M, Uddin MF. Smart microgrids: Optimal joint
wevj11040073.
scheduling for electric vehicles and home appliances. IEEE Trans Smart Grid
[37] Zepter JM, Engelhardt J, Gabderakhmanova T, Marinelli M. Re-thinking the
2014;5(1):239–50. http://dx.doi.org/10.1109/TSG.2013.2290894.
definition of self-sufficiency in systems with energy storage. In: 5th International
[24] Chaudhari K, Ukil A. TOU pricing based energy management of public EV
charging stations using energy storage system. In: 2016 IEEE International Conference on Smart Energy Systems and Technologies (SEST 2022). IEEE; 2022.
Conference on Industrial Technology. IEEE; 2016, p. 460–5. http://dx.doi.org/ [38] Barré A, Deguilhem B, Grolleau S, Gérard M, Suard F, Riu D. A review on lithium-
10.1109/ICIT.2016.7474795. ion battery ageing mechanisms and estimations for automotive applications. J
[25] García-Triviño P, Torreglosa JP, Fernández-Ramírez LM, Jurado F. Control and Power Sources 2013;241:680–9. http://dx.doi.org/10.1016/j.jpowsour.2013.05.
operation of power sources in a medium-voltage direct-current microgrid for 040.
an electric vehicle fast charging station with a photovoltaic and a battery [39] Keil P, Schuster SF, Wilhelm J, Travi J, Hauser A, Karl RC, et al. Calendar
energy storage system. Energy 2016;115:38–48. http://dx.doi.org/10.1016/j. aging of Lithium-ion batteries. J Electroch Soc 2016;163(9):A1872–80. http:
energy.2016.08.099. //dx.doi.org/10.1149/2.0411609jes.
[26] Deng J, Shi J, Liu Y, Tang Y. Application of a hybrid energy storage system [40] Hong S, Yue T, Liu H. Vehicle energy system active defense: A health assessment
in the fast charging station of electric vehicles. IET Gener Transm Distribution of lithium-ion batteries. Int J Intell Syst 2020. http://dx.doi.org/10.1002/int.
2016;10(4):1092–7. http://dx.doi.org/10.1049/iet-gtd.2015.0110. 22309.
[27] Mumtaz S, Ali S, Ahmad S, Khan L, Hassan S, Kamal T. Energy management and [41] Preger Y, Barkholtz HM, Fresquez A, Campbell DL, Juba BW, Romàn-Kustas J, et
control of plug-in hybrid electric vehicle charging stations in a grid-connected al. Degradation of commercial Lithium-ion cells as a function of chemistry and
hybrid power system. Energies 2017;10(11):1923. http://dx.doi.org/10.3390/
cycling conditions. J Electroch Soc 2020;167(12):120532. http://dx.doi.org/10.
en10111923.
1149/1945-7111/abae37.
[28] Gamboa G, Hamilton C, Kerley R, Elmes S, Arias A, Shen J, et al. Control strategy
[42] CATL. CATL BESS product brochure. 2019, URL https://www.catl.com/en/ess/.
of a multi-port, grid connected, direct-DC PV charging station for plug-in electric
[43] Simolin T, Rauma K, Viri R, Mäkinen J, Rautiainen A, Järventausta P. Charging
vehicles. In: 2010 IEEE Energy Conversion Congress and Exposition. IEEE; 2010,
powers of the electric vehicle fleet: Evolution and implications at commer-
p. 1173–7. http://dx.doi.org/10.1109/ECCE.2010.5617838.
[29] Acha S, Green TC, Shah N. Effects of optimised plug-in hybrid vehicle charging cial charging sites. Appl Energy 2021;303:117651. http://dx.doi.org/10.1016/
strategies on electric distribution network losses. In: IEEE PES T&D 2010. IEEE; j.apenergy.2021.117651.
2010, p. 1–6. http://dx.doi.org/10.1109/TDC.2010.5484397. [44] Dominguez-Jimenez JA, Campillo JE, Montoya OD, Delahoz E, Hernández JC.
[30] Engelhardt J, Gabderakhmanova T, Rohde G, Marinelli M. Reconfigurable Seasonality effect analysis and recognition of charging behaviors of electric
stationary battery with adaptive cell switching for electric vehicle fast-charging. vehicles: A data science approach. Sustainability 2020;12(18):7769. http://dx.
In: 55th International Universities Power Engineering Conference. IEEE; 2020, doi.org/10.3390/su12187769.
p. 1–6. http://dx.doi.org/10.1109/UPEC49904.2020.9209774. [45] Cheng Y, Wang W, Ding Z, He Z. Electric bus fast charging station resource
[31] Engelhardt J, Zepter JM, Gabderakhmanova T, Rohde G, Marinelli M. Double- planning considering load aggregation and renewable integration. IET Renew
string battery system with reconfigurable cell topology operated as a fast Power Gener 2019;13(7):1132–41. http://dx.doi.org/10.1049/iet-rpg.2018.5863.
charging station for electric vehicles. Energies 2021;14(9):2414. http://dx.doi. [46] Liu T, Yang X-G, Ge S, Leng Y, Wang C-Y. Ultrafast charging of energy-dense
org/10.3390/en14092414. lithium-ion batteries for urban air mobility. ETransportation 2021;7:100103.
[32] Gabderakhmanova T, Engelhardt J, Zepter JM, Sørensen TM, Boesgaard K, http://dx.doi.org/10.1016/j.etran.2021.100103.
Ipsen HH, et al. Demonstrations of DC microgrid and virtual power plant
[47] Karimi S, Zadeh M, Suul JA. Evaluation of energy transfer efficiency for shore-
technologies on the danish island of Bornholm. In: 55th International Universities
to-ship fast charging systems. In: 2020 IEEE 29th International Symposium
Power Engineering Conference. IEEE; 2020, p. 1–6. http://dx.doi.org/10.1109/
on Industrial Electronics. IEEE; 2020, p. 1271–7. http://dx.doi.org/10.1109/
UPEC49904.2020.9209853.
ISIE45063.2020.9152219.
[33] Ledro M, Pinter ZM, Gabderakhmanova T, Marinelli M. Testing procedures of
active power functions for grid code compliance of a battery system. In: 1st
International Conference on Renewable Energies and Smart Technologies. IEEE;
2022.

10