Contents
List of Figures II
List of Tables II
1 Introduction 1
1.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Research scope and Significance . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.3 Structure of the thesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
2 Literature Review 3
2.1 EV Charging Infrastructure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
2.2 Simulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.3 Discrete Event Simulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
3 Methodology 8
3.1 Model Conceptualization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
3.2 Model Assumptions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
3.3 Experiment Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
3.4 System and simulation specification . . . . . . . . . . . . . . . . . . . . . . . . 14
4 Model Evaluation 14
4.1 Validation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
4.2 Verification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
5 Simulation Results Analyze 14
5.1 Evaluation the key Performance . . . . . . . . . . . . . . . . . . . . . . . . . . 14
5.2 Sensitivity analyze . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
5.3 Optimal Scenario . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
6 Discussion 14
6.1 Limitation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
6.2 Future Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
7 Conclusion 14
Bibliography 17
A Appendix A 18
I
List of Figures
List of Tables
1 EV capacities and charging durations . . . . . . . . . . . . . . . . . . . . . . . 13
II
Abstract
Abstract
A brief summary of the procedure and the main results.
General Characteristics
• Objectivity: It should not contain any personal evaluation.
• Brevity: It should be as short as possible.
• Comprehensibility: It has a clear, comprehensible language and structure.
• Completeness: All essential facts should be included.
• Precision: It should accurately reflect the content and opinion of the original work.
III
1 Introduction
1 Introduction
1.1 Background
A shared global vision is to create a carbon-neutral future that ends climate change and
itsimpacts. Accordingly, nationsandcompaniesworldwidearestrivingtoachievenet-zero
carbonemissionsby2050(EsfahanianandMeech,2013). Greenhousegas(GHG)emissions
significantly influence the climate, triggering a range of adverse side effects (Stern, 2008).
InEurope,thisrealizationpromptedtheestablishmentoflong-termtargetstoreduceGHG
emissions, specifying an 80% reduction by 2050 compared to 1990 levels. In 2016, the
transport sector accounted for 24% of European greenhouse gas emissions (Kommission,
2019). In Germany, the transport sector ranks as the third largest contributor to green-
house gas (GHG) emissions, with passenger vehicles accounting for 59% of its emissions
(Bundesministerium für Umwelt, Naturschutz und nukleare Sicherheit, 2020). Thus, Ger-
many could benefit from cutting its greenhouse gas emissions by 78 million tons (48%) by
2030 relative to its 2019 levels (Kasten et al., 2023).
In pursuit of low-carbon goals, many nations now see the electrification of transport as
a central element of their strategic agenda and a major area for investment (Zhou et al.,
2013). To decrease these greenhouse gas emissions, electric vehicles could be viewed as
theoptimaltechnologicalsolution,sincetheyproducefeweremissionsthananyotherdrive
type even with today’s electricity mixes. Nevertheless, despite the clear environmental
advantagesofthisshift,theadoptionofelectricvehicles(EVs)hasbeenprogressingslowly.
The transition to this technology has been gaining traction, as evidenced by the fact that
in 2020 in Germany, electric passenger cars accounted for 8% of new registrations while
hybrids made up 20.4% (Zeng et al., 2021). This slow adoption maybe be linked to various
factors, including high purchase prices, technological constraints such as limited driving
range, and uncertainties or concerns about the capacity of the charging infrastructure (CI)
(Noel et al., 2020).
The type of housing an EV owner lives can even be a crucial factor in determining whether
home charging is feasible. Early adopters of electric vehicles are most likely to reside in
single-family homes, where the ability to charge at home is more common.(Nicholas and
Wappelhorst, 2020). Therefore, as time progresses and society advances, an increasing
number of consumers and families are opting for electric and hybrid vehicles. However,
not all types of residences can accommodate the charging needs of these vehicles, which
makes the strategic deployment of public EV charging stations particularly crucial.
Apart from this, beginning in 2009, the German government has actively promoted the
expansion of the charging network by investing in research, demonstration projects, and
funding schemes (Bundesministerium für Verkehr und digitale Infrastruktur [BMVI], 2020).
Charging infrastructure is the essential link between electric vehicles and the electricity
grid. Proper planning of charging infrastructure is therefore essential to promote the rapid
development of EVs while ensuring the normal operation of the electricity system (Xiang
et al., 2016).
1
1 Introduction
1.2 Research scope and Significance
This research focuses on optimising the deployment of electric vehicle (EV) charging in-
frastructure using a discrete event simulation (DES) model. The study addresses a critical
aspectofthetransitiontoelectricmobilitybyinvestigatingnotonlythequantityofcharg-
ing stations, but also the optimal mix of different charging technologies. The scope of this
researchisdefinedbyitsemphasisonsimulatingtheoperationaldynamicsofthecharging
infrastructureinacontrolledenvironment,wherekeyperformancemetricssuchascharging
unit utilisation, waiting times and queue management are systematically evaluated.
The significance of this work is twofold. First, by refining the configuration of EV charging
stations,thestudycontributestothereductionofgreenhousegasemissions-animportant
goalfornationsseekingacarbon-neutralfuture(EsfahanianandMeech,2013;Stern,2008).
Given that transport-related emissions account for a significant proportion of total green-
house gas emissions in Europe and Germany (Götz, Adisorn, and Tholen, 2021), optimising
charging infrastructure directly supports broader environmental and climate goals.
Second, the research provides critical insights for policy makers and industry stakehold-
ers. By identifying efficient configurations and deployment strategies, the study helps to
overcome one of the major barriers to EV adoption, the perceived inadequacy of charging
infrastructure(Noeletal.,2020). Thisisparticularlyrelevantinurbanareas,whereresiden-
tialconstraintslimitthefeasibilityofhomechargingandthereforearobustpubliccharging
network is required.
Overall, the results of this research are expected to facilitate the strategic planning of EV
charginginfrastructure,therebyacceleratingtheuptakeofelectricvehiclesandsupporting
sustainable urban mobility. The insights gained from the DES model not only improve our
understanding of the interplay between charging technology and user behaviour, but also
provide practical guidance for achieving long-term climate goals.
1.3 Structure of the thesis
This thesis focuses on the planning of electric vehicle charging infrastructure, employing
a discrete event simulation model to address a realistic problem. The primary objective is
to identify a more efficient configuration of EV charging stations—not merely in terms of
quantity but also by optimizing the mix of different charging station types.
The study begins with a comprehensive literature review that outlines related works and
introduces the selected methodologies. This section establishes the research context by
highlighting previous findings and identifying the gaps that this thesis aims to fill. Fol-
lowing the review, the thesis details the design of the experimental study. This includes a
clear presentation of the model assumptions, the overall framework of the discrete event
model, and the specific experimental setup. This rigorous design ensures that the simula-
tion reflects real-world conditions accurately. Subsequently, the fourth section is devoted
to the evaluation of the model, divided into two critical parts: validation and verification.
This dual assessment confirms the reliability and accuracy of the simulation, thereby re-
2
2 Literature Review
inforcing the credibility of the subsequent results. The fifth section presents an in-depth
analysisofthesimulationoutcomes. Here,theresultsarediscussedcomprehensively,with
a focus on uncovering the optimal configurations of EV charging stations and examining
the efficiency of different station combinations. In the following section, the limitations of
the simulation model are critically evaluated. This discussion not only highlights areas for
potentialimprovementbutalsosuggestsavenuesforfutureresearch,therebycontributing
to the ongoing development of EV infrastructure planning. Finally, the thesis concludes
with a concise summary that encapsulates the key findings and contributions of the study,
emphasizing its significance in advancing sustainable urban mobility and energy manage-
ment.
2 Literature Review
This section of the thesis provides an in-depth review of the existing literature and related
researchinfourkeyareas,withtheaimofprovidingarobusttheoreticalfoundationforthe
simulationmodellingthatfollows. Byclarifyingkeyprinciplesindetail,thisreviewsetsthe
context for the methodologies employed in the study.
Firstly,thediscussionfocusesontheEVcharginginfrastructure. Itoutlinesdifferentcharg-
ingmethodsandexaminescurrentpatternsofEVuse,providinginsightsintohowcharging
practices are evolving in response to technological advances and market demands. This
analysis helps to illustrate the challenges and opportunities that exist in the area of EV
charging infrastructure.
Secondly, the review highlights the widespread adoption of computational simulation as a
versatilemodellingtoolacrossmanyresearchareas. Itexplainshowsimulationhasbecome
anindispensabletechniquefortacklingcomplexreal-worldproblems,enablingresearchers
to replicate dynamic systems and predict outcomes with considerable accuracy.
The literature also emphasises that different real-world challenges often require tailored
simulation approaches. Different simulation methods are discussed, focusing on the fact
that each method is best suited to specific types of problems. In this context, discrete
eventsimulation(DES)modelsarehighlightedasbeingparticularlyeffectiveforanalysing
systems where discrete events occur at different times.
Finally, the section concludes with a comprehensive overview of the basic concepts un-
derlying DES models. This summary not only reinforces the theoretical underpinnings of
the simulation approach used in this thesis, but also highlights its relevance to the opti-
misation of EV charging infrastructure. Through this detailed examination, the reader will
gain a clear understanding of the principles and methods that will guide the subsequent
modelling and analysis.
3
2 Literature Review
2.1 EV Charging Infrastructure
With the rapid development of electric mobility, the current configuration of EV charging
points has become a critical research focus. The literature suggests that effective deploy-
mentofcharginginfrastructurerequiresamultifacetedapproachthatconsidersbothtech-
nologicalcapabilitiesandspatialdistribution. Unterluggaueretal.(2022)emphasisethata
well-organised network of charging stations is essential for large-scale EV adoption, while
Nicholas and Wappelhorst (2020) highlight Germany’s ambitious plan to expand its public
charging network to 1 million points by 2030 as a central element in reducing transport-
related emissions.
Existing studies categorise EV charging technologies into three main types, each designed
to meet different usage scenarios (Xi, Sioshansi, and Marano, 2013; Chen et al., 2020):
1. Level-one charging:
This method uses a standard household socket, typically 110V/15A. It is generally
intended for residential use where vehicles are parked for long periods - usually
overnight. Due to the relatively low power output, it can take between 12 and 18
hours to fully charge an EV battery (typically between 16 and 25 kWh). Level one
charging offers the advantage of convenience and low installation costs, but is lim-
ited by its slow charging rate.
2. Level-two Charging:
Leveltwochargingusesahighervoltagecircuit,typicallyaround220Vandratedbe-
tween15and30A.Thisapproachprovidesafasterchargingalternativetolevelone,
makingitwellsuitedtobothresidentialandpublicenvironmentssuchasworkplaces
or commercial car parks. The increased power output significantly reduces charging
time while maintaining a relatively moderate installation cost, providing a balance
between efficiency and accessibility.
3. DC Fast Charging:
Using high-voltage direct current - often in the range of 400-500V - DC fast charg-
ing is designed for rapid energy transfer. This technology can typically recharge a
standard EV battery in around 30 minutes, making it ideal for high-traffic areas such
as motorway service areas and city centres. However, due to the specialised equip-
ment and higher costs involved, DC fast charging is not generally installed in typical
residential environments.
Each of these charging technologies has an important role to play in developing a com-
prehensive EV infrastructure that meets the diverse needs of users and supports the wider
transition to electric mobility.
Furthermore,aheterogeneousmixofthesechargingtechnologiesisoftenrecommendedto
meetthediverseneedsofEVusers. Chenetal.(2020)arguethatwhileimprovementsinEV
technology are vital, the establishment of a reliable and efficient charging infrastructure
is equally important to encourage widespread EV adoption. The integration of different
chargingmethodscanimprovegridresiliencebybalancingrapidenergyprovisioninhigh-
4
2 Literature Review
traffic areas with more accessible, lower-power solutions in residential areas.
Policy initiatives and regulatory frameworks also shape the deployment of EV charging
points. Incentives such as subsidies and public-private partnerships have been instru-
mental in accelerating infrastructure deployment and optimising geographic distribution
(Nicholas and Wappelhorst, 2020). In addition, advances in digital technologies and smart
gridintegrationfacilitatesystemoptimisationthroughdynamicpricing,loadbalancingand
real-time monitoring, thereby improving both operational efficiency and user satisfaction
(Unterluggauer et al., 2022).
In summary, the literature highlights the importance of a strategically planned, heteroge-
neous charging network that uses a combination of Level 1, Level 2 and DC fast charging
technologies. Such a network not only supports the growing demand for electric vehicles,
but also plays a key role in achieving wider sustainability and urban air quality objectives.
2.2 Simulation
Simulationinvolvesdevelopingamodelofarealsystemandperformingexperimentswithit
tounderstandthesystem’sbehaviororassessvariousoperationalstrategieswithinspecific
criteria (Shannon, 1975). Simulating a probabilistic model involves generating its inherent
stochasticmechanismsandthenmonitoringhowthemodelevolvesovertime(Ross,2022).
Thisfundamentalmethodforproblem-solvingandoptimizationisemployedacrossawide
rangeofindustries,includingenergy,healthcare,publicservices,andmining(Moon,2017).
The application of computer simulations to address challenges in complex systems has
become widespread (Azab, AlGeddawy, et al., 2012).
Simulation methods have become indispensable tools for modeling complex systems. Re-
searchers have developed several approaches to capture dynamic interactions and uncer-
tainties. With the rapid growth of electric vehicles (EVs), charging stations have become
a vital part of urban infrastructure and energy management systems. Optimizing the lay-
out and operation of these stations is crucial for maximizing utilization, reducing waiting
times, and ensuring efficient energy distribution. Researchers have applied various simu-
lation techniques to address these challenges.
In the following, four key different approaches that can be used to simulate are described
andthosesimulationhighlightingtheirtheoreticalfoundations,practicalapplications,and
comparative advantages.
• Agent-based Simulation
Agent-based simulation represents a complex system as a collection of autonomous
agents interacting according to defined rules. This bottom-up approach is especially
useful for capturing the heterogeneity and decentralized decision-making processes
ofindividualEVusersandchargingstations. EarlyseminalworksbyBonabeau(2002)
and Macal and North (2005) demonstrated that ABS can effectively model emergent
phenomenainsystemsasdiverseasurbanplanning,marketdynamics,andsocialin-
teractions. The flexibility of ABS enables researchers to incorporate learning and
5
2 Literature Review
decision-making processes, providing a detailed understanding of system dynam-
ics In the context of EV charging infrastructure, ABS has been used to simulate the
micro-level interactions between EV owners, charging behaviors, and station perfor-
mance—revealing how individual decision-making can lead to macro-level network
effects(Bonabeau, 2002; Macal and North, 2005). Martí et al. (2022) introduced two
configurable generators designed to automate the creation of experiments in agent-
based simulations. These generators enhance the simulation of realistic movements
and the spatial distribution of vehicles, passengers, and other urban traffic system
users within a city. Empirical evaluations by Martí et al. (2022) have demonstrated
that these generators are highly effective for comparing different distributions of lo-
cations as well as varying agent movement behaviors based on real city data.
• User Charging Behavior
Incontextssuchaselectricvehicle(EV)infrastructure,modelinguserchargingbehav-
ior is critical for accurately simulating system performance. This approach integrates
empirical data on driver habits, such as charging frequency, duration, and decision-
making criteria, into simulation models. UBIS and the comfortable range concept
elucidate the typical battery charge level at which users choose to recharge. UBIS
is associated with users’ confidence in their internal understanding of range dynam-
ics, the effective utilization of their available range, and the incorporation of sur-
plusenergyfromrenewablesources. Theseinsightscarrysignificantimplicationsfor
enhancing the sustainability of electric mobility systems (Franke and Krems, 2013).
Electricvehicledrivers’chargingbehavioristypicallyinfluencedbypsychologicalfac-
torssuchaspersonalitytraitsandriskpreferences. Inthiscontext, Hu, Dong, andLin
(2019)presentsamodelingframeworkbasedoncumulativeprospecttheory(CPT)to
characterizethechargingdecisionsmadebythesedrivers,ultimatelyconcludingthat
riskpreferenceparameterssignificantlyimpactchargingbehavior(Hu,Dong,andLin,
2019). Incorporatingrealisticuserbehaviorsignificantlyimprovespredictionsregard-
ing charging station utilization and load balancing. Understanding these behavioral
patterns helps identify potential inefficiencies and informs infrastructure planning.
• Particle Swarm Simulation
Particle swarm simulation (PSS) is derived from particle swarm optimization (PSO),
a heuristic technique inspired by the collective behavior of flocks or swarms. Intro-
duced by Kennedy and Eberhart (1995), PSS is effective in exploring complex search
spaces and optimizing system parameters. This method has been applied to prob-
lems ranging from scheduling and routing to resource allocation. Its strength lies in
itsadaptiveconvergencetowardoptimalsolutionswhilesimulatingbothcooperative
and competitive interactions among particles (Kennedy and Eberhart, 1995). In this
approach,aclusterofcandidatesolutions(eachrepresentinga’particle’)exploresthe
multi-dimensionalsolutionspacetodeterminetheoptimalconfigurationofEVcharg-
ingstationlocationsandcapacityallocations. Aparticleswarmoptimisationmethod
basedondeepneuralnetworkmodifiedboundaries(DNNMBPSO)isusedtosolvethe
problemofnotbeingabletoquicklyandaccuratelycomputetheoptimalsolutionfor
charging station siting in the field of electric vehicle charging station siting research
6
2 Literature Review
(J. Liu, S. Liu, and Gong, 2021).
• Mento Carlo Simulation
Monte Carlo simulation is a stochastic method that employs random sampling to
understand the probabilistic behavior of complex systems. Since their inception by
Metropolis and Ulam (1949), Monte Carlo techniques have been widely adopted in
fields such as finance, risk management, and engineering. By simulating a vast num-
ber of scenarios, this approach provides insights into the variability and reliability of
model predictions, making it a robust tool for uncertainty quantification (Metropolis
and Ulam, 1949). The probabilistic modelling study, based on the Monte Carlo Simu-
lation(MCS)methodology,modelsEVchargingpointswithinthedistributionnetwork,
taking into account the influence of EV owner behaviour. This approach allows the
evaluation of the optimal location and sizing of EV charging stations (EVCS) to miti-
gatetheirnegativeimpactonthenetworkunderuncertainloadconditions(Shahbazi
et al., 2023). In Zhang et al. (2018), a Monte Carlo simulation-based approach was
proposed to model the daily charging load of an EV charging station while explic-
itlyaccountingforvariousstochasticfactors. Additionally,theauthorsincorporateda
threshold mechanism for failed charging attempts within the simulation framework,
thereby enhancing its capacity to capture the operational uncertainties inherent in
real-world charging environments.
2.3 Discrete Event Simulation
Depending on the simulation’s objective, there are specific metrics we aim to measure.
However,sincethemodel’sevolutionovertimetypicallyinvolvesacomplexinterplayamong
its components, tracking this progression to accurately determine these metrics is often
challenging. A comprehensive framework centered on the concept of "discrete events" has
been devised to facilitate tracking a model’s evolution over time and identifying the key
quantities of interest (Ross, 2022).
Discrete Event Simulation (DES) is based on the fundamental principle that the state vari-
ables of a system change only at specific, isolated points in time. It is at these specific
points that events occur, marking the only times when the state of the system changes.
Typically, DES models a complex system as an ordered sequence of events, although it
canaccommodatecomplexsequencesandhierarchicalstructures. Italsoexplicitlymodels
the uncertainties associated with these events and statistically analyses their cumulative
effects on the system (Moon, 2017).
Buildingonthesefundamentalprinciples,recentstudieshaveusedDEStoaddressthespe-
cificchallengesassociatedwithEVcharginginfrastructures. Forexample,DEShasbeenin-
strumentalinsimulatingthedynamicinteractionsbetweenvehicles,chargingstationsand
the electricity grid. Xiang et al. Xiang et al. (2016) demonstrated that DES can effectively
model how varying arrival rates and charging durations affect grid load management and
overall system performance. Their work highlights that strategic placement of charging
stations can mitigate peak load challenges and improve service reliability.
7
3 Methodology
FurtherresearchhasusedDEStoevaluatetheperformanceofheterogeneouschargingnet-
works. Nicholas and Wappelhorst Nicholas and Wappelhorst (2020) used DES to compare
different configurations of level one, level two and DC fast charging stations. Their results
suggest that an optimal mix of charging technologies not only reduces waiting times, but
also improves the overall utilisation of charging resources. Similarly, Chen et al. Chen et
al. (2020) used DES to assess the impact of charging infrastructure on urban mobility sys-
tems, providing evidence that a well-planned charging network is critical for accelerating
EV adoption and supporting sustainable transport.
In addition to infrastructure optimisation, DES has proven valuable in dealing with the un-
certainties inherent in the EV charging process. Moon (2017) demonstrated that DES is
particularly adept at capturingthe stochastic nature of vehicle arrivals and variablecharg-
ing durations. This capability is critical for accurately predicting key performance metrics
such as system dwell time and throughput, which in turn enables more resilient and re-
sponsive infrastructure planning Ross (2022) further emphasises that by focusing on dis-
crete events, DES provides a structured framework for monitoring system evolution and
identifying quantities of interest over time.
Overall, the literature highlights the versatility and robustness of DES as a methodology
for modeling complex systems such as EV charging infrastructures. Its ability to integrate
stochastic processes and capture the nuanced interplay of system components makes it an
invaluabletoolforbothresearchersandpolicymakers. Asthedemandforefficientandsus-
tainableEVchargingsolutionsgrows,theinsightsprovidedbyDESmodelswillbeessential
for guiding future infrastructure investments and policy decisions (Noel et al., 2020; Zeng
et al., 2021).
3 Methodology
Thischapterprovidesadetaileddescriptionoftheframeworkusedinthisstudy. Specifically,
a comprehensive DES model has been developed to replicate the operational dynamics of
electric vehicle charging infrastructure. This model is supported by a detailed case study
thatmeticulouslydescribestheunderlyingassumptionsthatareessentialtothesimulation
process. In addition, the application of DES to the evaluation of critical performance met-
ricsofthecharginginfrastructureandproceduresisdetailed. Keymetricsanalysedinclude
chargingunitutilisation,waitingtimesforservice,andqueueavailabilityandfrequency,all
of which are evaluated as a function of the number of charging stations. By systematically
varying the configuration of the charging network, the model provides valuable insights
into the operational efficiency and capacity of the system. Analysis of these performance
indicatorsnotonlyhelpstounderstandthecurrentstateofthecharginginfrastructure,but
also has important implications for optimising future electric vehicle charging networks.
Ultimately, this framework provides a robust basis for evaluating and improving the effi-
ciency of EV charging systems in real-world scenarios.
8
3 Methodology
3.1 Model Conceptualization
Themainaimofthisresearchistoanalyzeandoptimizethedistributionofcharginginfras-
tructureforElectricVehicles(EVs)inurbanareas. Awell-planneddistributionofelectricve-
hicle(EV)charginginfrastructurecansignificantlyenhancebothEVchargingefficiencyand
infrastructureutilization. Moreover, coordinatingeffectivelywiththenationalgridenables
more rational power allocation planning, thereby ensuring the expansion of the charging
network and the sustainable development of energy resources.
To accomplish this, a DES model was developed using Python, which accurately replicates
the operational environment, taking into account key attributes and variables reflective
of a real-world system. Since discrete event simulations based on real-world conditions
cannot fully replicate every scenario, defining the system boundaries in advance is crucial.
Consequently,thisstudyfocusesexclusivelyontheprocessesoccurringwithinthecharging
station from the moment a vehicle arrives until its charging is completed while excluding
factors such as vehicle travel and dynamic power limitations imposed by network loads.
Toaccuratelysimulatethisprocess,itisnecessarytoeliminateexogenousfactorsandfocus
only on the central elements. Accordingly, the model deliberately excludes several influ-
ences that are considered exogenous or outside the scope of the current analysis in order
to minimize statistical bias in the simulation. For example, the impact of price changes on
demand - such as variations in tariffs and different tariff strategies - is not considered as
these factors are beyond the focus of this study. In addition, network load constraints are
notconsidered,i.e.,bothfastchargersandtypeIIchargersareassumedtobecontinuously
available.
Furthermore, differences in vehicle arrival times are not modeled; instead, a constant dis-
tributionisused, andnospecialtreatmentisgiventopeakversusoff-peaktimes, allowing
the model to focus exclusively on arrival flows (i.e., arrival rates). Intrinsic differences in
vehicle state of charge (SOC) are also not explicitly modeled but are approximated by a
uniform time interval within the same framework.
By excluding these exogenous variables, the analysis is able to focus on the core issue:
assessing the impact of varying the number and performance of charging station configu-
rationsonsystemqueuingandwaitingtimes. Thedetailedassumptionswillbeelaborated
in the following section.
By abstracting and simplifying the process, we concentrate on the complete lifecycle of an
electric vehicle—from its arrival in an area equipped with charging stations to its eventual
departure. In practice, a full EV charging process involves multiple steps; however, in the
simulations conducted for this thesis, the EV is treated as a single unit, and the entire
sequence from arrival to departure is handled uniformly. This simplification reduces the
overallcomplexityofthemodel. Additionally,inpractice,EVownersoftendonotwaituntil
the charging process is completely finished—that is, they do not always allow the battery
to reach full capacity before ending the session. Specifically, two scenarios can occur:
9
3 Methodology
1. When the EV reaches full charge, the charging post automatically cuts off the power.
However, if the owner fails to move the vehicle promptly, the charging spot remains
occupied and cannot be utilized by others.
2. Alternatively, the owner may intentionally discontinue the charging process before
the EV is fully charged and then drive away.
Consequently, the charging process duration is determined by calculating the interval be-
tweenthecommencementofchargingandthevehicle’sdeparture,whiletheaveragewait-
ing time per hour is computed as the difference between the vehicle’s arrival time and the
start of its charging process.
Giventhelimitedavailabilityofreal-worlddataandthechallengesinaccuratelyestimating
parameters such as the arrival rate, we will employ hypothetical distributions to simulate
theprocessesofEVarrivalandcharging(orparking). Wewillthencomparekeyperformance
metrics—such as waiting time, system dwell time, and throughput—across three distinct
scenarios. Smallcarparksequippedwithcharginginfrastructurewillserveastheoperating
environment. Within this setting, the simulation will be configured to model interactions
between entities and resources—such as vehicle arrivals and the queuing behavior when
all charging points are occupied.
Within the system model, electric vehicles requiring charging services are treated as enti-
ties,whilevarioustypesofchargersserveasresourcescapableofaccommodatingmultiple
vehicles simultaneously in varying numbers. When a vehicle arrives and all charging sta-
tionsareoccupied,iteitherentersaqueueoroptstoleave,withthequeuingprocessfollow-
ingafirst-come,first-servedprotocol. Duetotheinsufficientavailabilityofreal-worlddata,
both the arrival process and charging durations must be estimated through assumptions.
Consequently, hypothetical distributions are employed to model the intervals between ve-
hicle arrivals. This thesis examines three distinct distributions for this purpose:
1. Uniform distribution
First,theuniformdistributionassumesaconstantprobabilityofarrivalovertheentire
time interval, meaning that it is equally likely that an EV will arrive at any moment
withintheinterval(Ross,2022). Forexample,ifanEVisequallylikelytoarriveatany
minute during a 10 minute period, the probability density function remains constant
throughout that period.
2. Triangular distribution
Second,thetriangulardistributionisparticularlyusefulwhenthereissomeempirical
or expert information about the minimum, maximum and most likely arrival times.
This distribution allows for a mode that reflects the peak arrival time and can be
skewed if the mode is not centred (Hillier and Lieberman, 2010). For example, if the
datasuggeststhatEVarrivalspeakaround8.30am,withfewerarrivalsbeforeandaf-
ter,atriangulardistributioncaneffectivelycapturethispatternbysettingappropriate
lower, upper and mode values.
3. Poisson distribution
Third, the Poisson distribution is commonly used to model the number of events oc-
10
3 Methodology
curring within a fixed interval, assuming that these events occur independently and
at a constant average rate (Banks et al., 2005). In the context of EV arrivals, if the av-
erage arrival rate is known (e.g. 5 vehicles per hour), the Poisson distribution allows
me to calculate the probability of observing a given number of arrivals in that time
period, thereby reflecting the inherent randomness of the process.
By including these three distributions in my simulation, I can assess the sensitivity of the
model to different assumptions about the arrival process. The specific assumptions of the
different distributions are explained in detail in the following sections. This approach not
only strengthens the robustness of the analysis, but also provides a more comprehensive
understandingofhowdifferentarrivalpatternsmightaffecttheperformanceofEVcharging
infrastructure.
3.2 Model Assumptions
Every simulation model starts with a defined set of assumptions that set the framework
within which the results are interpreted. In our study, we have explicitly stated these as-
sumptions to avoid misinterpretation and to ensure that our conclusions remain grounded
in the intended context of the model. By clearly articulating the rationale behind each as-
sumption, we provide the necessary transparency that not only clarifies the scope of our
analysis, but also enhances the validity of our findings.
Our deliberate formulation of these assumptions aims to isolate the core dynamics under
investigation. Specifically, the impact of different charging station configurations on sys-
temqueuingandwaitingtimes. Thisapproachallowsustofocusontheessentialelements
of the process, while minimizing extraneous influences that could obscure the interpreta-
tion of our simulation results. This ensures that our model serves its intended purpose: to
provide clear, actionable insights into the optimization of electric vehicle charging infras-
tructure.
MoststudiesonsimulationmodelsrelyonassumptionsaboutEVuserbehaviorratherthan
incorporating actual user data. In contrast, data-driven simulation models either extrapo-
late EV behavior from data collected on Internal Combustion Engine (ICE) vehicles or uti-
lize stated response data concerning EV infrastructure requirements (Philipsen, Schmidt,
and Ziefle, 2015). Very few effective simulation models are validated based on extensive
real-world data in the current research field. (Helmus et al., 2019).
In addition, a well-reasoned explanation of assumptions can contribute to the advance-
ment of sustainability science by guiding future research and promoting a more robust
understanding of complex systems. This approach not only enhances the scientific rigour
ofastudy,butalsosupportsmoreinformeddecisionmakinginenvironmentalandsustain-
ability initiatives (Moon, 2017).
Therefore, this thesis establishes key assumptions to ensure that the DES model functions
effectivelyandremainsreliable. Severalofthesecriticalassumptionsaredescribedbelow.
11
3 Methodology
1. Car parks equipped with EV charging infrastructure are assumed to operate continu-
ously24hoursaday,ensuringthatchargingservicesarealwaysavailableforelectric
vehicles in need. For the purposes of this study, we make the simplifying assump-
tion that these charging points operate under ideal conditions, without taking into
account special circumstances such as unplanned outages or technical failures. This
assumption allows the simulation to focus solely on the core dynamics of charging
demand and infrastructure performance, without the confounding influence of inter-
mittent service disruptions.
2. Given the wide variety of electric vehicles—such as Battery Electric Vehicles (BEVs)
andPlug-inHybridElectricVehicles(PHEVs)—thatrequirecharging,itisnecessaryto
simplifytheproblemforsubsequentscalabilitystudies. Tothisend, Iassumethatall
vehiclesarrivingatthechargingstationarehomogeneousintermsofbatterycapacity
and associated performance parameters. Although real-world BEVs may have battery
capacities ranging from 20 kWh to over 100 kWh, I adopt a standardized battery ca-
pacity value from Shao, Guo, and Qiu (2017). This assumption allows me to isolate
theessentialdynamicsofthechargingprocessandreducesthecomplexityassociated
with heterogeneous vehicle characteristics, thereby enabling a more robust analysis
of infrastructure performance.
3. It is generally accepted that electric vehicles should initiate recharging when their
battery charge falls below a critical threshold, typically set at 20%. In parallel, to
protect the batteries and extend their lifespan, I have established an upper charging
limit of 80% at the charging point. This dual threshold setting is based on industry
experienceandempiricalevidence,whichindicatethatlimitingthestateofchargeto
arangebetween20%and80%canhelpmitigatebatterydegradationwhileensuring
sufficient operational range. By standardizing these thresholds within the simula-
tion, I am able to focus on the core dynamics of charging behavior under controlled
conditions, thereby providing more precise insights into the performance of the EV
charging infrastructure.
4. Inthissimulation,Ifocusontwotypesofcharginginfrastructure: ACcharging(Level-
two) and DC charging. In general, DC charging points can provide charging rates of
up to 100 kW, while AC charging stacks provide a lower rate of 19.2 kW. By includ-
ing charging piles with these different rates in the simulation, the performance and
efficiency of high- and medium-power charging schemes can be compared within a
unified framework. This approach facilitates a nuanced analysis of the operational
trade-offs and infrastructure requirements associated with different charging tech-
nologies.
5. BasedontheresearchofErde!andKesen(2025),Imodeltheservicetimesforcharg-
ing using uniform distributions to capture the inherent variability in charging times.
Specifically, I assume that the service time for an AC charger follows a uniform(10,
120) distribution, while that for a DC charger follows a uniform(10, 30) distribution.
These assumptions are based on an evaluation of a 57 kWh battery being charged
from a critical threshold of 20% to full capacity. In this scenario, the slower charg-
ing capability of AC chargers requires a longer operating time - up to 120 minutes -
12
3 Methodology
while the higher performance of DC chargers enables a much faster process, taking
up to 30 minutes under optimal conditions. By using these uniform service duration
distributions, I not only capture the essential dynamics of the charging process, but
also establish a tractable simulation framework for evaluating the performance and
scalability of EV charging systems under variable operating conditions.
The capacity and duration-related parameters are provided in Table 1.
Definition Data
Battery capacities of all EVs 57 kWh
Charging rate of AC Charger 19.2 kWh
Charging rate of DC Charger 100 kWh
Service durations (min) of AC Charger UNIF (10, 120)
Service durations (min) of DC Charger UNIF (10, 30)
Critical Threshold(SOC) 20%
Maximum Threshold(SOC) 90%
Table 1: EV capacities and charging durations
3.3 Experiment Design
In this study, I use SimPy, a discrete event simulation framework in Python, to develop
and run the simulation experiments. In our simulation model, active components such
as electric vehicles and charging stations are represented as processes. These processes,
implementedasPythongenerators(oftenreferredtoasprocessfunctionsormethods),exist
within a unified simulation environment where they interact with each other and with the
environment through discrete events.
Throughout its lifecycle, each process generates events, which remain pending until they
are triggered. When a process generates an event, it is temporarily suspended or "hung"
until the event occurs. When the event is triggered, SimPy resumes the process, and if
multiple processes are waiting for the same event, they are resumed in the order in which
the events were generated. This structured sequence of suspensions and resumptions is
fundamental to capturing the dynamic interactions within the simulation.
Toensurethoroughtrackingandanalysis,IimplementedanEventLoggerclassthatrecords
allrelevanteventsthatoccurinthesimulation. Eachevent,alongwithitsassociatedvalues,
is logged via method parameters within this class. This detailed logging mechanism is
criticaltounderstandingthesequence,timingandimpactofeventsonsystemperformance,
ultimately allowing me to evaluate the effectiveness and efficiency of the simulated EV
charging infrastructure.
Inmysimulation,IstartbyinitialisingtheEVclassusingtheSimPyenvironment(self.env).
In parallel, I instantiate the ChargingStation class with the necessary resources and create
an instance of the EventLogger. These components work together to simulate the inter-
13
7 Conclusion
action between the vehicle stream and the charging process at the station - taking into
account variations in power levels and charging speeds, as well as the time elapsed be-
tweenthearrivalofanelectricvehicleandthecompletionofitscharging. Atcriticalpoints
in the charging process, I use the EventLogger to record relevant events.
I also define key properties for each vehicle, such as battery capacity (self.car_battery_size)
and current state of charge (self.car_battery_level). In this simulation, I assume that all
vehicles have the same battery capacity, but arrive with different initial charge levels. The
entire entity flow through the system is orchestrated within the specified run method.
3.4 System and simulation specification
4 Model Evaluation
Before any simulation model is used in research or further analysis, it must be thoroughly
verifiedandvalidated. Istartbycheckingthatthemodeliscorrectlyconstructed,confirming
thateachcomponentandinteractionisconsistentwithitstheoreticaldesign. Ithenvalidate
themodeltoensurethatitaccuratelyrepresentsthereal-worldsystemorphenomenonitis
intendedtosimulate. Theseprocessesareessentialtoestablishthereliabilityandaccuracy
ofthemodel,therebyprovidingarobustbasisforsubsequentscientificinvestigation(Moon,
2017).
4.1 Validation
4.2 Verification
5 Simulation Results Analyze
5.1 Evaluation the key Performance
5.2 Sensitivity analyze
5.3 Optimal Scenario
6 Discussion
6.1 Limitation
6.2 Future Work
7 Conclusion
14
References
Azab, A, T AlGeddawy, et al. (2012). “Simulation methods for changeable manufacturing”.
In: Procedia CIRP 3, pp. 179–184.
Banks,Jerry,JohnS.Carson,BarryL.Nelson,andDavidM.Nicol(2005).Discrete-EventSystem
Simulation. 4th. Prentice Hall.
Bonabeau,Eric(2002).“Agent-basedmodeling:Methodsandtechniquesforsimulatinghu-
man systems”. In: Proceedings of the national academy of sciences 99.suppl_3, pp. 7280–
7287.
Bundesministerium für Umwelt, Naturschutz und nukleare Sicherheit (2020). Klimaschutz-
zahlen 2020. [Online]. URL: https://www.bmu.de/fileadmin/Daten_BMU/
Pools/Broschueren/klimaschutz_zahlen_2020_broschuere_bf.pdf.
Bundesministerium für Verkehr und digitale Infrastruktur [BMVI] (2020). Title of the Web
Page.2024-07-25.URL:%5Chttps://www.bmvi.de/SharedDocs/DE/Artikel/
G/%20foerderrichtlinie-ladeinfrastruktur-elektrofahrzeuge.html.
Chen, Tianjin, Xiao-Ping Zhang, Jianji Wang, Jianing Li, Cong Wu, Mingzhu Hu, and Huiping
Bian (2020). “A review on electric vehicle charging infrastructure development in the
UK”. In: Journal of Modern Power Systems and Clean Energy 8.2, pp. 193–205.
Erde!,HakanandSaadettinErhanKesen(2025).“Searchingforadoptionofelectricvehicles
toourmodernlife:adiscreteeventsimulationanalysis”.In:SIMULATION101.2,pp.177–
194.
Esfahanian, E. and J. A. Meech (2013). “Hybrid electric haulage trucks for open pit mining”.
In: IFAC Proceedings Volumes (IFAC-PapersOnline). IFAC Secretariat, pp. 104–109. DOI:
10.3182/20130825-4-US-2038.00042.
Franke, Thomas and Josef F Krems (2013). “Understanding charging behaviour of electric
vehicle users”. In: Transportation Research Part F: Traffic Psychology and Behaviour 21,
pp. 75–89.
Götz,Thomas,ThomasAdisorn,andLenaTholen(2021).DerdigitaleProduktpassalsPolitik-
Konzept: Kurzstudie im Rahmen der umweltpolitischen Digitalagenda des Bundesministeri-
ums für Umwelt, Naturschutz und nukleare Sicherheit (BMU). Tech. rep. Wuppertal Report.
Helmus, Jurjen R, Seyla Wachlin, Igna Vermeulen, and Mike H Lees (2019). “SEVA: A data
driven model of electric vehicle charging behavior”. In: arXiv preprint arXiv:1904.08748.
Hillier,FrederickS.andGeraldJ.Lieberman(2010).IntroductiontoOperationsResearch.10th.
McGraw-Hill Education.
Hu, Liang, Jing Dong, and Zhenhong Lin (2019). “Modeling charging behavior of battery
electricvehicledrivers:Acumulativeprospecttheorybasedapproach”.In:Transportation
Research Part C: Emerging Technologies 102, pp. 474–489.
Kasten,Peter,LukasMinnich,JoelleRandrianarisoa,DavidRitter,MoritzVogel,KayaDünzen,
Lisa Just, Amir Ashour Novirdoust, Hendrik Diers, Nicole Niesler, et al. (2023). Szenarien
und regulatorische Herausforderungen für den Aufbau der Ladeinfrastruktur für elektrische
Pkw und Lkw.
Kennedy,JamesandRussellEberhart(1995).“Particleswarmoptimization”.In:Proceedings
of ICNN’95-international conference on neural networks. Vol. 4. ieee, pp. 1942–1948.
15
Kommission, Europäische (2019). Langfristige Strategie—Zeithorizont 2050. URL: https:
//ec.europa.eu/clima/policies/strategies/2050_de (visited on
02/26/2019).
Liu, JY, SF Liu, and DQ Gong (2021). “Electric vehicle charging station layout based on par-
ticle swarm simulation”. In: International Journal of Simulation Modelling 20.4, pp. 754–
765.
Macal, Charles M and Michael J North (2005). “Tutorial on agent-based modeling and sim-
ulation”. In: Proceedings of the Winter Simulation Conference, 2005. IEEE, 14–pp.
Martí, Pasqual, Jaume Jordán, Javier Palanca, and Vicente Julian (2022). “Charging stations
and mobility data generators for agent-based simulations”. In: Neurocomputing 484,
pp. 196–210.
Metropolis, Nicholas and Stanislaw Ulam (1949). “The monte carlo method”. In: Journal of
the American statistical association 44.247, pp. 335–341.
Moon, Young B (2017). “Simulation modelling for sustainability: a review of the literature”.
In: International Journal of Sustainable Engineering 10.1, pp. 2–19.
Nicholas, Michael and Sandra Wappelhorst (2020). “Regional charging infrastructure re-
quirementsinGermanythrough2030”.In:Beijing,Berlin,SanFrancisco,SaoPaulo,Wash-
ington: icct.
Noel,Lance,GerardoZarazuadeRubens,JohannesKester,andBenjaminKSovacool(2020).
“Understandingthesocio-technicalnexusofNordicelectricvehicle(EV)barriers:Aqual-
itativediscussionofrange,price,chargingandknowledge”.In:EnergyPolicy138,p.111292.
Philipsen, Ralf, Teresa Schmidt, and Martina Ziefle (2015). “A charging place to be-users’
evaluation criteria for the positioning of fast-charging infrastructure for electro mobil-
ity”. In: Procedia Manufacturing 3, pp. 2792–2799.
Ross, Sheldon M (2022). Simulation. academic press.
Shahbazi, A, H Moradi CheshmehBeigi, H Abdi, and M Shahbazitabar (2023). “Probabilistic
optimal allocation of electric vehicle charging stations considering the uncertain loads
byusingtheMonteCarlosimulationmethod”.In:JournalofOperationandAutomationin
Power Engineering 11.4, pp. 277–284.
Shannon, Robert E (1975). Systems simulation-the art and science.
Shao, Sujie, Shaoyong Guo, and Xuesong Qiu (2017). “A mobile battery swapping service
for electric vehicles based on a battery swapping van”. In: Energies 10.10, p. 1667.
Stern, Nicholas (2008). “The economics of climate change”. In: American Economic Review
98.2, pp. 1–37.
Unterluggauer, Tim, Jeppe Rich, Peter Bach Andersen, and Seyedmostafa Hashemi (2022).
“Electricvehiclecharginginfrastructureplanningforintegratedtransportationandpower
distribution networks: A review”. In: ETransportation 12, p. 100163.
Xi, Xiaomin, Ramteen Sioshansi, and Vincenzo Marano (2013). “Simulation–optimization
modelforlocationofapublicelectricvehiclecharginginfrastructure”.In:Transportation
Research Part D: Transport and Environment 22, pp. 60–69.
Xiang, Yue, Junyong Liu, Ran Li, Furong Li, Chenghong Gu, and Shuoya Tang (2016). “Eco-
nomic planning of electric vehicle charging stations considering traffic constraints and
load profile templates”. In: Applied Energy 178, pp. 647–659.
16
Zeng, Dan, Yan Dong, Huajun Cao, Yuke Li, Jia Wang, Zhenbiao Li, and Michael Zwicky
Hauschild (2021). “Are the electric vehicles more sustainable than the conventional
ones? Influences of the assumptions and modeling approaches in the case of typical
cars in China”. In: Resources, conservation and recycling 167, p. 105210.
Zhang,Tianyang,XiChen,ZheYu,XiaoyanZhu,andDiShi(2018).“AMonteCarlosimulation
approach to evaluate service capacities of EV charging and battery swapping stations”.
In: IEEE Transactions on Industrial Informatics 14.9, pp. 3914–3923.
Zhou, Lin, Furong Li, Chenghong Gu, Zechun Hu, and Simon Le Blond (2013). “Cost/benefit
assessmentofasmartdistributionsystemwithintelligentelectricvehiclecharging”.In:
IEEE Transactions on Smart Grid 5.2, pp. 839–847.
17
A Appendix A
18
