Evaluating the Fulﬁlment Rate of Charging Demand for Electric Vehicles

Using Open-Source Data

Hana Elattar, Ferdinand von T

ullenburg

, Sebastian W

ollmann and Javier Valdes

Institute for Applied Computer Science, Deggendorf Institute of Technology, Technologie Campus Freyung,

Freyung, Germany

Keywords:

Electric Vehicles (EVs), Energy Demand, Charging Infrastructure.

Abstract:

With the shift towards electric vehicles accelerating; we are working with open-source data to estimate to

which degree existing charging infrastructure is fulﬁlling the demand created by electric vehicles. This paper

is explaining how to create such a calculation by extracting data from large public areas in the city of Lindau

(Bodensee), Germany as a showcase. With this data we aim to evaluate whether charging stations located

in the premises of public and commercial buildings cover the demand of electric vehicles reaching the said

buildings. This research is conducted as a ﬁrst step of methodologies development that aims on the long term

to create a tool that supports in the optimal placement of new charging stations. The methodology chosen

is inspired by two main concepts: the ﬁrst is the attractiveness factor concept used for the creation of travel

models, while the second is the classiﬁcation of charging stations based on location to determine their rate of

occupancy. They are both used to cluster buildings and charging stations respectively to be able to determine

the number of users in the area of study (AOS) compared to the overall number of electric vehicles reaching

the destination in a given day. This paper takes the island in the centre of the city of Lindau (Bodensee) as

its area under investigation and uses open-source data along with the appropriate assumptions as a base for its

calculations.

1 INTRODUCTION

In recent years, the distribution of Electric Vehicles

(EVs) has signiﬁcantly increased. More and more

people switch from Combustion Vehicles (CVs) re-

place them especially with Battery-Driven Electric

Vehicles (BEVs). Together with the increasing avail-

ability of green energy, this development leads to a

positive effect for the greenhouse gas balance of our

economies. On the other side of the coin, Charg-

ing Infrastructure (CI) is being increasingly requested

and utilised. Following the initial spread of publicly

available Charging Stations (CSs) across the countries

in order to decrease the hurdles for the shift to more

sustainable transportation, now the question arises, at

which locations the CI should be strengthened in or-

der to satisfy the needs of the EV users (Klinkhardt

et al., 2021).

In recent years, there has been several research ap-

proaches and studies to evaluate the energy demand

and market development of electric vehicles including

extrapolations into the nearer future either for partic-

https://orcid.org/0000-0002-2326-1815

ular cities (Schlote et al., 2021) or for whole countries

(Zhou et al., 2015). Other studies answered the ques-

tion about the current utilisation of CI (Hecht et al.,

2020). In order to estimate arrivals at certain loca-

tions it has been proposed to analyse the popularity

index (as known, e. g., from Google maps), show-

ing the occupation of certain locations by time. From

this studies, we know in rather general terms how CI

needs to be developed in the upcoming years.

However, an open challenge is to precisely esti-

mate the demand for CI around a particular location

or set of locations. With this information, more tar-

geted investments into CI would be possible taking

the actual demands of EV drivers into its focus. On

the one hand it needs to be answered, if visitors of a

certain place ﬁnd sufﬁcient charging possibilities for

their EVs within walking distance. On the other hand,

it needs to be considered, if it makes sense for EV

users to charge their vehicle at a given location, which

basically is related to the time, that people typically

stay at a given location, State of Charge (SOC) of the

EV and the user preferences. The approaches, which

analyse trafﬁc situations at particular locations in or-

Elattar, H., von Tüllenburg, F., Wöllmann, S. and Valdes, J.

Evaluating the Fulﬁlment Rate of Charging Demand for Electric Vehicles Using Open-Source Data.

DOI: 10.5220/0011849400003473

In Proceedings of the 9th International Conference on Geographical Information Systems Theory, Applications and Management (GISTAM 2023), pages 159-166

ISBN: 978-989-758-649-1; ISSN: 2184-500X

 2023 by SCITEPRESS – Science and Technology Publications, Lda. Under CC license (CC BY-NC-ND 4.0)

159

der to estimate arrivals do not take into consideration

typical stay times.

Our research objective has been to develop an ap-

proach to estimate the demand for charging in walk-

ing distance to particular locations or buildings using

publicly available sources of data. In order to estimate

how many CSs should be available in walking dis-

tance of a particular building, we propose a method-

ology to determine the purpose of the building, which

gives an indication for typical stay times and the open-

ing hours of a building. Additionally, we consider the

characteristics of the building such as the size of the

building, which gives, together with the building / lo-

cation purpose, an estimation on the number of people

that arrive to the building based on statistical data and

building regulations. For the analysis, we limit our-

selves to publicly available OpenStreetMap (OSM)

data to ensure that the approach is transferable to ar-

bitrary regions and we also consider if buildings have

multiple purposes, for instance, if an apartment build-

ings has a shop ﬂoor. The methodology we present

is based on a re-classiﬁcation and standardization of

OSM data. We feel that our approach can be well used

as an extension to already existing approaches, which

take trafﬁc data and car arrivals at particular locations

into their consideration.

From a scientiﬁc point of view, our work con-

tributes to the development of methodologies aimed

at estimating the demand and capacity of CIs. In ex-

tension to the existing work, we particularly empha-

size the use of open data, and how types of real world

data (building data, location, charging station, loca-

tion attractiveness data, and statistical information) is

merged into a single automated workﬂow.

2 RELATED WORK

Among the various approaches for estimating the

charging demand for EVs, some focus on the charging

stations while some approaches focus on the tracks

the cars travel in. For both approaches two general

directions can be found: empirical investigations and

simulations.

Empirical investigations: van den Akker (J.M.

van den Akker, 2020) used a preprocessed data set

of 4.9M charging processes provided externally, con-

taining: arrival time, connection time, distance (be-

tween adjacent chargers in a straight line), time be-

tween sessions, category of a charging session (13

different types differentiating charging time, same or

different charging location, duration of the session -

same, medium, long-, time between charging events,

and type of a distribution – narrow, wide, broad-). Ad-

ditionally, drivers are modelled by their behaviour de-

scribed through the probability over the 13 types of

charging session the users perform. For each individ-

ual EV the distances are computed through the energy

consumption between charging processes.

An important empirical approach has been pro-

vided by Hecht et al. (Hecht et al., 2020), who anal-

ysed the occupancy rate of charging stations all over

Germany. For this, the authors classiﬁed charging sta-

tions by their power and their location (urban, subur-

ban, industrial, uninhabited). The work provides aver-

age statistical data on when charging stations are used

and how utilised they are being.

Another empirical approach is given by Draz and

Albayrak (Draz and Albayrak, 2019), who estimate

the energy demand of a vehicle through stochastically

deﬁning the SOC when the vehicle arrives and the

battery size. Based on this it is estimated whether a

car uses a high power CS (with low SOC) or a nor-

mal charging power. Based on the arrival times of

EVs and the number of charging points at a location,

the number of charging services is deﬁned. Given a

charging proﬁle for a EV, it is now possible to calcu-

late the energy demand.

simulative investigation: A simulative approach

has been described by Schlote et al. (Schlote et al.,

2021), where the energy consumption of EV is mod-

eled on the basis of the Markovian net, taking some

parameters such as slope of the road segment, aver-

age speed, acceleration, and potential of regenerative

breaking into account. Technically, the SUMO traf-

ﬁc simulator is utilised for evaluation of the model.

The motion trajectories of cars are based again on the

Markov model where the probabilities that a car is

driving along certain road segments are derived; the

segments being deﬁned by the road intersections. The

destinations of cars (which could give an indication

where cars recharge), are determined by a random

process estimating how many cars are travelling from

which origin to which destination. Different general

strategies for route selection are taken such as a ”min-

imum popularity routing”, or ”minimum energy rout-

ing”. As a vehicle model, a single standard vehicle

has been used by the authors to be modeled via a small

set of parameters.

Another simulative approach uses the SUMO traf-

ﬁc simulator and extends it with a physics-based

model for the energy consumption of different classes

of vehicles (Koch et al., 2021). Different power-trains

can be simulated with this extension. In a large-scale

study, the authors investigated the energy consump-

tion of the German city of Paderborn in various sce-

narios describing the total share of EV of correspond-

ing car segments. For this purpose, the authors used

GISTAM 2023 - 9th International Conference on Geographical Information Systems Theory, Applications and Management

160

three potential scenarios for the development of the

local trafﬁc including the share and type of EVs in

Paderborn based on recent developments with an out-

look towards 2030. For the ﬁrst scenario, the current

set of car segments is used, for the second scenario a

dominance of small vehicles is assumed. In scenario

3, the increasing trend of SUVs of 13% per year with

1% decrease for the next 10 years is assumed, making

SUVs the most dominant car by 2030. For technical

information about cars, statistical data for the devel-

opment of the EV market has been used.

A third simulative approach is the demand estima-

tion of the EVs using Multi-Agent-System (MAS)-

based open source trafﬁc simulation software called

MATSim (Jahn et al., 2020). Four car classes with

different average consumption per km are used for the

car simulation. In the MAS, the agents follow their

daily schedule composed by the activities of work-

ing, shopping, and other leisure activities, which have

been derived from census data. Each agent is applied

with transportation modes such as car, bike, public

transport, etc. Distances, transport modes, etc. are

backed with open source and census data.

Also Hern

andez-Moreno et al. provided a simu-

lative approach for estimating the energy consump-

tion of a single EV using the MATLAB/SIMULINK

package ADVISOR, which is capable of estimat-

ing the energy consumption of arbitrary power-trains

(Hern

andez-Moreno et al., 2022). The cars are mod-

elled through parameters describing the chassis, the

e-motor, the battery set, and a storage block for regen-

erative breaking. The authors used the Tesla Model 3

as their base model due to the public availability of

the required data. Dynamic trafﬁc parameters have

been modelled with a Markov M/D/n queuing model,

the parameters of which have been captured from a

real-world trafﬁc observation based on video track-

ing. From the computation of the queuing model the

parameters for a SUMO MAS simulation have been

derived. From that simulation, driving parameters for

individual cars have been gathered, which, in turn,

have been used for analysing the energy consumption

of these cars.

3 DATA

Continuing on the carried-on research, our aim is to

come up and apply a methodology to create a model

that depends solely on open-data, and hence could

further on be applied to different contextual locations

to calculate more precisely the required expansions in

the local CS infrastructures for cities and the optimal

locations of new stations. The methodology to be ex-

Figure 1: Methodology Logic.

plained in this section and demonstrated in Fig 1 is

inspired by the overlapping of the work of Klinker-

hardt et.al. (Klinkhardt et al., 2021) and that of Hecht

et.al. (Hecht et al., 2020). Using classiﬁed OSM in-

formation on buildings and Points of Interests (POIs)

data organized by tags representing particular features

of locations, we can can provide an estimation of the

maximum number of people visiting a certain build-

ing during a chosen time interval. This data is then

overlapped with open-data on CS which can be cate-

gorized based on the study from Hecht et al. (Hecht

et al., 2020) and therefore have an average percentage

of occupancy rate during these chosen time intervals.

3.1 Working with Geographical Data

POIs and Building uses construct the basis of our Trip

Purpose tables. We use the following “Trip Purpose”

deﬁnition extracted from different research on travel

model: “the categorized list of destinations based on

the function” (Klinkhardt et al., 2021). To ensure the

adaptability of the methodology to different scenar-

ios, this research paper uses OSM as a main source

for geographical data. This decision ensures avail-

ability of data for most locations, the ability to en-

hance the data thanks to the open-source character-

istic of the platform, and the ability to calculate the

accuracy of data in comparison to other areas. The

use of OSM as a source has also allowed for the ac-

quisition of what is considered “raw data”; meaning

that no pre-processing has been carried out on it and

therefore the data can be ﬁltered and clustered to ﬁt

into our travel model.

Evaluating the Fulﬁlment Rate of Charging Demand for Electric Vehicles Using Open-Source Data

161

Figure 2: Classiﬁed buildings.

To ensure our travel model includes all types of

trips; we extracted all buildings from OSM and then

proceeded to assign different uses to buildings based

on the tags assigned to them. This way we ensure

all tags are considered. We therefore extracted the

list of attributes related to our downloaded buildings

and created a prioritization list based on the data ﬁlled

in. We were then, able through comparison of dif-

ferent attributes and their rate of usability, and using

PyQGIS script for the automation of the process; to

create a new “fclass” local feature attribute to contain

the identiﬁed use of the building. During the prioriti-

zation, some keys’ values were eliminated; those were

values that didn’t provide an indication of a place that

could host people but rather indicated either a quick

stop or an attachment to a place (as with the key value

”attached” often found). The ﬁrst “quick stop” elimi-

nation is to adapt the data to our need, which is the

allocation of new Charging Stations. Quick stops

are not in themselves destinations and don’t provide

a time long enough for charging an E-vehicle. The

second elimination of “attachments” was to avoid du-

plication of data by eliminating ”attached buildings”

whereas the main building was found to be sufﬁcient

for the calculations. A third type of elimination also

took place. This was to eliminate polygons features

that were mistakenly mapped as buildings when they

maintain the function of a plot of land.

For the POIs, a different type of features -

“Points”- was studied. What created the challenge,

with handling the POIs, is their inaccurate allocation

where they sometimes overlapped with buildings with

different uses. Nevertheless, POIs add a layer of ac-

curacy as they include uses and places of attractions

that may not be added to the buildings’ attributes. To

calculate the factor of attractiveness in this research,

however, area calculations are necessary. For Lin-

dau, we could then merge POIs with their nearest

buildings- which is usually the case in reality- and

proceed to treat the resulting building as a multipur-

pose. This sequence of processes resulted in the map

shown in Fig 2, where an overall use of each building

in the city of Lindau (Bodensee) can be seen.

3.2 Building Areas / Points of Interest

Areas

Having performed a ﬁrst clustering of buildings based

on usage, a second layer of clustering is carried out

based on trip purposes and the type of visitors. Ac-

cording to Klinkerhart et al., a speciﬁed list of pur-

poses can be created to identify the reason a person

would want to reach a destination as well as the type

of person (i.e., student – customer – worker – etc.).

This table can be matched with open databases on at-

tractiveness factor. An attractiveness factor describes

the rate by which this destination is likely to be chosen

for a trip. Our next step to identify an exact number of

visitors is therefore to multiply this factor by the num-

ber of visitors in the area of the building; assuming the

capacity of 1 person per square metre (pers/sqm).

3.3 Open Data on Charging Stations

In this research we focus on two types of informa-

tion concerning the charging stations in Lindau: geo-

graphical and statistical. Being that geographical in-

formation refers to the location of charging stations;

the statistical information refers to the statistical data

on the usage rate of charging stations. The later is

based on the study by Hecht et al.(Hecht et al., 2020),

classifying the CS based on location and assigning us-

age rates to each class.

From OSM and other public sources we acquire

locations and some of the other parameters of the

charging stations currently available in the area of

study such as charging power, accessibility, opening

hours. However, this data does not contain informa-

tion of their typical utilisation which can be described

mainly by the typical number of charging events per

day. As a starting point we use the classiﬁcation as

proposed by Hecht et al., who classify charging sta-

tions according to the location and the nominal charg-

ing power. Location classes are “industrial”, “ur-

ban”, “suburban”, “uninhabited”, while classes for

the charging power are “P <= 4kW ”, “4kW < P <=

12kW ”, “12kW < P <= 25kW”, “25kW < P <=

GISTAM 2023 - 9th International Conference on Geographical Information Systems Theory, Applications and Management

162

100kW ” “P < 100kW ”). The authors provide utili-

sation proﬁles for each of these classes by weekday.

For our analysis and considering our area of study, we

will focus on the data for urban-class CSs with 22 kW

(class: “12kW < P <= 25kW ”) and use a correcting

factor according to the building classiﬁcation and the

typical stay time.

3.4 Data on Electric Vehicles Stock

For the estimation of the number of e-vehicles (BEV

and PHEV) in Lindau we take general statistical val-

ues for Germany into consideration. Ofﬁcial statistics

indicate a market share of 2.6% of BEV (1.3%) and

PHEV (1.3%) of the total number of vehicles in 2022

(Kraftfahrt-Bundesamt, 2022). For the development

of the market share until 2030 several studies have

been conducted, most of them aiming at a total market

share of BEV and PHEV between one-fourth (Detlef

Borscheid and Kraftfahrt-Bundesamt, 2020) and one-

third (Center of automotive management, 2022). For

2025, Borscheid (Detlef Borscheid and Kraftfahrt-

Bundesamt, 2020) assumes a stock share of about

11%.

Thus, for our study, we assume a stock share of

BEV and PHEV vehicles of 11% for 2025. For 2030,

we assume 30% stock share.

4 METHODOLOGY

The aim of the applied methodology is to prepare

a model that could further on be used for differ-

ent contextual locations and for more precise calcu-

lations on the required CS infrastructure for cities

and their optimal locations. The methodology is in-

spired by the overlapping of the work of Klinkerhardt

et.al. (Klinkhardt et al., 2021) and that of Hecht et.al.

(Hecht et al., 2020). Using classiﬁed OSM buildings

and POIs data, we can provide an estimation of the

maximum number of people visiting a certain build-

ing during a chosen time interval. This data is then

overlapped with open-data on CS which can be cate-

gorized based on the study from Hecht et al. (Hecht

et al., 2020) and therefore have an average percentage

of occupancy rate during these same time intervals.

4.1 Methodology Logic

Having gathered and aggregated the building data

from OSM, we used the calculation of the building

capacity based on the usage regulations to identify

Figure 3: Buffer of OSM available CS.

the buildings with the highest capacity. The assump-

tion taken at this point is that during rush hours these

buildings will be at full capacity. In the work of

Klinkhardt et al, attractiveness factor databases were

used as a solution against this assumption and to cal-

culate a concrete number of visitors. This attractive-

ness factor can differ based on the use of the space

but also based on more case-speciﬁc factors such as

the brand name for commercial buildings for exam-

ple. Sources for attractiveness factors were found to

be proprietary, as opposed to common architectural

recommendations for sqm/pers. For the purpose of

this research, which is to investigate the reliability of

open-data in conducting such a study; the attractive-

ness factor has been substituted by the building ca-

pacity. These recommendation tables and tools such

as found in the Neufert’s book (Neufert et al., 2012)

can help determine the standard sqm/pers based on

which a maximum ﬂoor building capacity could be

calculated. The assumption then made is that the ﬂoor

area is at full capacity during the simulation inter-

val. The next layers to be overlapped are the num-

ber of EVs arriving to each destination and the CS

data layer. Assuming all visitors arrive with personal

vehicles to the destination, and taking the EVs stock

shares mentioned in subsection 3.6 into considera-

tion, we can use the following equation to estimate

the maximum number of EVs (n

). Number of ve-

hicles being n

, number of EVs is n

, and building

capacity is b, if we assume the building is at full ca-

pacity we would have (n

) = (b) and since in 2022 the

share of EV in Germany was 2.6 n

= 2.6/100 ∗ (n

)

then n

= 2.6/100 ∗ b.

Evaluating the Fulﬁlment Rate of Charging Demand for Electric Vehicles Using Open-Source Data

163

Figure 4: Automation Framework.

Some more sophisticated models such that created

by the ”Ver-Bau” tool work on the same basis but in-

clude however more case speciﬁc factors such as the

time of day, etc. The Ver-Bau tool was used, for ex-

ample, to determine the needed number of parking

slots attributed to the commercial building Lindau-

park (Engstler et al., 2021).To expand the model by

specifying when building are visited based on open

data, in the case of Germany or Austria, standard load

proﬁles of buildings can be used. These provide infor-

mation on the distribution of electricity demand gen-

erated by the economic activity. A classiﬁcation of

POIs with regard of this demand proﬁle, can help to

identify the moments in which the POI is going to be

more visited.

Allocating the buildings with the highest number

of EVs reaching them, we can investigate the sur-

rounding CSs which are, in our case study, catego-

rized as urban CS.From the literature review and more

speciﬁcally based on the work of Hecht et al. (Hecht

et al., 2020), we can conclude that on average an ur-

ban CS is occupied and used 20% of the working

time. Time is therefore our point of intersection be-

tween the number of EVs in a destination and the CS

demand. For a building (A) in the time interval of 5

hours between 12 to 17h, the number n

is stationed

in front of the building each hour. We know from

charging behaviour studies, that an EV stationed in

front of a building for a certain duration, will occupy

the CS for the same duration; regardless of the time

needed to complete a charging proﬁle. Looking at the

time of the study, it is possible then to assume that a

CS is occupied by the same vehicle. In the 400m rec-

ommended buffer, an Urban Charging Station (UCS)

is stationed, which is occupied 20% of the time; ap-

proximately 1h of the 5-hour study time. Since the

number n

is taken as a constant for each hour, UCS

is then fulﬁlling 20% of the demand in the span of

these 5 hours.

4.2 Methodology Application

Taking the characteristics of the city of Lindau (Bo-

densee) as inputs, we started the process of applying

our logic and visualizing the results on the map us-

ing QGIS. Because the whole process of downloading

and analyzing the OSM data manually is time con-

suming and difﬁcult to replicate, we started to auto-

mate the process via PyQGIS. PyQGIS is the python

interface for QGIS which allows us to build a modular

standalone data processing pipeline, with the work-

ﬂow as explained in Fig.4. The only dependency is a

valid, platform independent, QGIS (in our case QGIS

LTR 3.22.x). The CLI interface expects two input pa-

rameters: the name of the are to be processed, and a

directory path to save the output data. To begin, the

“Downloader”-Module tries to ﬁnd the corresponding

OSM-ID of the given area name. It sends a POST-

Request to the Overpass Interpreter Endpoint with the

following query:

[out:json][timeout:900];

relation["boundary"="administrative"]

["name"="{0}"]["type"="boundary"];

(._;>;);

out body;

This query searches for all administrative bound-

aries with the given name. The result of that

query contains information about the administra-

tive level of the boundary area, which can then be

split up by searching for boundaries with a higher

OSM administrative level (smaller real-world area)

to avoid maxing out the run-time of the API. Be-

cause each OSM building is referenced to be within

a boundary, we can query for all buildings in the

resulting boundaries. After downloading all build-

ings in .osm formats, we continue to merge them

back together into one ﬁle. At ﬁrst, we convert

each downloaded “.osm” ﬁle into two shapeﬁles: a

polygon and a point, by using the “ogr2ogr” CLI

tool. Using the “native:mergevectorlayers” algo-

rithm, the “LayerMerger”-module merges all poly-

gon ﬁles and all point ﬁles into corresponding joint

ﬁles. Moving forward, the “LayerCleaner”-Module

ﬁrst deletes duplicated geometries by using a cus-

tom algorithm, that iterates over each feature, saves

its ID into a list; if an ID is already included in the

list, the feature gets discarded. Otherwise, the fea-

ture gets copied into a new in-memory layer. Si-

multaneously every attribute column is checked to

ﬁnd completely empty attributes across the whole

data set. These attributes get then deleted by

the ”layer.dataProvider().deleteAttributes()” function.

The “LayerCleaner”-Module ﬁnishes by saving the

newly created in memory layer as a shapeﬁle on disk.

GISTAM 2023 - 9th International Conference on Geographical Information Systems Theory, Applications and Management

164

Table 1: Percentage of fulﬁlment of CS of the overall charging demand in the AOS.

Current rate of contribution Rate by 2025 Rate by 2030

Existing CS 20 % 4.7 % 1.7%

The last and biggest processing step is handled by

the “FieldCalculator”-module. Its purpose is to apply

the above-described in the methodology logic and in

Section 3 to calculate the measure and travel purpose

for each building. Based on our current dataset we

could attribute attractiveness factors to certain types

of buildings. For that, we begin with creating new

attribute ﬁelds (“fclass t”, “measure”, “trip purp”,

“attr rate”, “buil cap” and “area”) on the polygon

layer by using the “layer provider.addAttributes()”

function. After that the module iterates over

each polygon feature. We begin be determining

the “fclass” by comparing the “leisure”, “ofﬁce”,

“tourism”, “shop” and “building” ﬁelds. Going for-

ward the polygon area in square meters is calcu-

lated and saved in the “area” ﬁeld by using the

“QGSDistanceArea().measureArea()” function. Af-

ter that the “measure”, “trip purpose” and “attr rate”

are determined by looking up the corresponding

value in relation to the “fclass” in a predeﬁned

“fclass

trafﬁc matrix. An extract of that matrix looks

like this:

FCLASS_TRAFFIC_MATRIX = {

’office’: {

’measure’: ’Workplaces’,

’trip_purpose’: ’Business’,

’attraction_rate’: 0.1

}, ... }

We deﬁned for each possible fclass a measure,

trip purpose and attraction rate. The module ﬁn-

ishes by selecting all features in the layer and saving

them to the user-deﬁned output directory as a copy

with the “native:saveselectedfeatures” algorithm. The

“TrafﬁcChargingStationAnalyzer”-module proceeds

then to iterate over OSM mapped charging stations

(downloaded again by the “Downloader”-Module, but

querying this time for “charging stations”), calcu-

lates a 150-meter buffer (native “geometry().buffer()”

function) around it as shown in Fih 3 and sums up the

previously determined building capacity of all build-

ings inside the buffer. Based on the stock shares con-

cluded for EVs (2,6% for 2022, 11% for 2025 and

30% for 2030), the total number of EVs and the charg-

ing station usage is calculated for each year and re-

sults in the percentages demonstrated in Table 1.

As a result, there are two layers, one shapeﬁle with

every building and one shapeﬁle with all charging sta-

tions and their corresponding calculated values. The

pipeline ﬁnishes by validating if the output ﬁle exists

and then deleting all temporal ﬁles. In case something

goes wrong, the pipeline can pick up at a user con-

trollable step so not everything needs to be run again.

Having automated these calculations can demonstrate

the variation in which existing charging stations will

answer to the increased demand.

5 CONCLUSIONS AND FUTURE

WORK

The ﬁnal goal of our research is the development of

a method for gaining insight into the sufﬁciency of

CI for EVs, particularly BEV. In this article, we re-

ported about the ﬁrst results in our pursue for the over-

all goal: a method for determining to which degree

available CI is fulﬁlling the demands created by EV.

The motivation behind this work stems from the

support of both, the movement towards the trafﬁc

electriﬁcation, which subsequently results into a rise

in the use of EVs, the movement towards enhanc-

ing and providing more open-source data that can be

used for social and civil improvements. The presented

methodology therefore provides groundwork for how

planning CS infrastructure could be carried on us-

ing open-source data but also provides an overview

of the challenges that researchers would face in that

attempt. As open-source data would fall short in ac-

curacy, the results acquired through this work don’t

represent, therefore, a real-life estimation but rather

showcases the falling-short of CS infrastructure. In

the current step of our research we are also able to

analyse different scenarios with larger shares of EVs

and/or more CSs in the region.

In our planned future work, this methodology is

enhanced through the acquisition and integration of

more case-speciﬁc data such as CS proﬁles, demo-

graphic data for drivers behaviors and used cars, more

speciﬁc data on building uses and, vehicles’ energy

usage and charging behaviour. It is, hence, important

to keep track of the rate of accuracy of the results and

to provide more validation methodologies in the fu-

ture.

More speciﬁcally, our planned next steps will par-

ticularly include detailed information about trafﬁc

and vehicles. To go about that, ﬁrst, additional data

sources needs to be processed and included. Most im-

portantly, Floating Car Data (FCD) and trafﬁc census

data will be used to get insights into the local trafﬁc

Evaluating the Fulﬁlment Rate of Charging Demand for Electric Vehicles Using Open-Source Data

165

situation by capturing the trafﬁc in the region under

investigation. This particularly includes the number

of cars in the region and the origin-destination rela-

tions they typically exhibit. With this, we are able

to estimate how probable or important a recharging

of a car is, when it arrives (short distance drivers are

less probable to recharge then long-distance drivers).

We expect a more detailed estimation of the CS oc-

cupation. Furthermore, FCD would allow us also to

capture hourly distributions of EV arrivals in order to

estimate the CS occupation by our of day, this also in-

cludes the extension of our building usage model by

opening or typical usage times.

In a next step, also the car types, which are typi-

cal in the region will be included. This information is

used together with the typical travel distances, typical

stay times, energy consumption data for charging pro-

cesses, and a vehicle energy consumption simulation

to ﬁnally calculate the actual energy demand for EVs

in the investigated region. This is particularly impor-

tant for estimating the potentials for reinforcement of

CI. Particularly, when information about the power

distribution grid are taken into account.

Another envisioned step is the improvement of the

attractiveness factor by adding a metric describing the

reach of a particular building. Based on the building

use, the reach metric describes from which distance

cars typically arrive as this could essentially differ,

for instance, at a supermarket from a public author-

ity building with supra-regional signiﬁcance.

ACKNOWLEDGEMENTS

This work was created as part of the ERA-Net Smart

Energy Systems project CrossChargePoint (CCP),

funded from the European Union’s Horizon 2020 re-

search and innovation programme under grant agree-

ment no. 775970 and the German Ministry for Eco-

nomic Affairs and Climate Action.

REFERENCES

Center of automotive management (2022). Electromoility

report 2022.

Detlef Borscheid and Kraftfahrt-Bundesamt (2020). Prog-

nose: Mehr als elf Millionen Elektroautos und Plug-

Ins bis 2030. Autohaus.

Draz, M. and Albayrak, S. (2019). A Power Demand Esti-

mator for Electric Vehicle Charging Infrastructure. In

2019 IEEE Milan PowerTech, pages 1–6. IEEE.

Engstler, G., Weiß, A., and Doller, S. (2021). Er-

weiterung Lindaupark. Technical report, Feneberg

Grundst

ucksgesellschaft GbR.

Hecht, C., Das, S., Bussar, C., and Sauer, D. U. (2020).

Representative, Empirical, Real-World Charging Sta-

tion Usage Characteristics and Data in Germany, vol-

ume 6. eTransportation.

Hern

andez-Moreno, A., Diego-Ayala, U., Jazcilevich, A.,

Hern

andez-Paniagua, I. Y., Zavala-Reyes, J. C.,

arez, A., and Rosas-P

erez, I. (2022). Transient traf-

ﬁc energy-use analysis employing video-tracking and

microscopic modeling techniques: A case study us-

ing electric and combustion engine vehicles. Energy

Science & Engineering, 10(7):2022–2034.

Jahn, R. M., Syr

e, A., Grahle, A., Schlenther, T., and

ohlich, D. (2020). Methodology for Determining

Charging Strategies for Urban Private Vehicles based

on Trafﬁc Simulation Results. Procedia Computer

Science, 170:751–756.

J.M. van den Akker (2020). E-Vehicles: Interaction of

Smart Charging and DSO Strategies. Utrecht Univer-

sity.

Klinkhardt, C., Woerle, T., Briem, L., Heilig, M., Kager-

bauer, M., and Vortisch, P. (2021). Using Open-

StreetMap as a Data Source for Attractiveness in

Travel Demand Models. Transportation Research

Record, 2675(8):294–303.

Koch, L., Buse, D. S., Wegener, M., Schoenberg, S.,

Badalian, K., Dressler, F., and Andert, J. (2021). Ac-

curate physics-based modeling of electric vehicle en-

ergy consumption in the SUMO trafﬁc microsimula-

tor. In 2021 IEEE Int. Intell. Transp. Syst. Conf. ITSC,

pages 1650–1657. IEEE.

Kraftfahrt-Bundesamt (2022). Der fahrzeugbestand am 1.

januar 2022.

Neufert, E., Neufert, P., and Kister, J. (2012). Architects’

data. Wiley-Blackwell, Chichester, West Sussex, UK

; Ames, Iowa, 4th ed edition. OCLC: ocn775329524.

Schlote, A., Crisostomi, E., Kirkland, S., and Shorten, R.

(2021). Trafﬁc modelling framework for electric ve-

hicles. International Journal of Control, 85(7):880–

897.

Zhou, Y., Wang, M., Hao, H., Johnson, L., Wang, H., and

Hao, H. (2015). Plug-in electric vehicle market pen-

etration and incentives: A global review. 20(5):777–

795.

GISTAM 2023 - 9th International Conference on Geographical Information Systems Theory, Applications and Management

166