Understanding Car Usage Patterns for V2G Integration: Insights

from Dutch Travel Diaries

Simon Leu

, Gonçalo Homem de Almeida Correia

, Hans van Lint

and Axel Leonhardt

Institute of Transport and Spatial Planning, University of the Bundeswehr Munich,

Werner-Heisenberg-Weg 39, 85577 Neubiberg, Germany

Department of Transport and Planning, TU Delft, Stevinweg 1, 2628 CN Delft, The Netherlands

Keywords: Vehicle-to-Grid (V2G), Mobility Data Analysis, Car Usage Behavior, Clustering Analysis.

Abstract: Integrating renewable energy sources, such as solar and wind, challenges grid stability due to their intermittent

nature. Vehicle-to-grid (V2G) technology provides a promising solution by utilizing electric vehicles (EVs)

as decentralized energy storage systems, enabling the storage of surplus energy during low demand and its

release during peak demand. The effectiveness of V2G depends critically on car usage patterns. Data from

the Netherlands Mobility Panel (MPN) of 2022, comprising travel diaries from 2,505 households, was

analyzed to explore this. A methodology was developed to create car usage profiles based on parking durations

and locations, distinguishing weekday and weekend patterns. The analysis shows that vehicles are

predominantly parked at home, with weekday profiles reflecting work-related parking and weekend profiles

highlighting increased leisure activity. Households with shared cars showed higher driving activity and shorter

parking durations than households with a 1:1 car-to-license ratio or surplus vehicles. Six distinct car usage

clusters were identified for weekdays and four for weekends.

1 INTRODUCTION

The transition to renewable energy sources such as

solar, wind, and hydro is accelerating worldwide. In

2023, renewable energy accounted for 48% of

electricity generation in the Netherlands, equivalent

to more than 55 billion kWh (Statistics Netherlands

(CBS), 2024). While this progress brings significant

environmental benefits, it also poses new challenges

for the energy grid. The intermittent nature of solar

and wind energy production means that supply often

peaks when demand in residential areas is low, such

as in the afternoon when the sun unfolds its full power

or during periods of high wind, which are also

typically in the afternoon, while the highest energy

demand in these areas commonly occurs in the

mornings and evenings when people are at home to

take a shower or charge their cars, for example.

This mismatch between supply and demand is

increasingly pushing the energy grid to its limits, as

https://orcid.org/0009-0000-3173-0995

https://orcid.org/0000-0002-9785-3135

https://orcid.org/0000-0003-1493-6750

https://orcid.org/0009-0000-1382-3231

shown by the grid congestion in Figure 1

(Capaciteitskaart 2024), which occurs when

electricity cannot be transported through the grid at

that time. Without adequate energy storage solutions,

excess renewable energy is wasted during periods of

low demand, while fossil fuel-based generation may

still be required to meet peak demand in the evenings.

An ideal solution is to store excess renewable

energy at times of low demand and release it at peak

times. Vehicle-to-grid (V2G) technology offers a

promising approach to achieving this goal by

leveraging electric vehicle (EV) batteries as

distributed energy storage systems.

Although V2G is technically mature and ready for

deployment, its adoption remains limited. Key

barriers include a lack of standardization, limited

availability of V2G-compatible vehicles,

infrastructure challenges, battery degradation

concerns, and insufficient regulatory and policy

support. Besides infrastructural, technical, and

Leu, S., Correia, G. H. A., van Lint, H. and Leonhardt, A.

Understanding Car Usage Patterns for V2G Integration: Insights from Dutch Travel Diaries.

DOI: 10.5220/0013412500003941

Paper published under CC license (CC BY-NC-ND 4.0)

In Proceedings of the 11th International Conference on Vehicle Technology and Intelligent Transport Systems (VEHITS 2025), pages 189-198

ISBN: 978-989-758-745-0; ISSN: 2184-495X

189

regulatory boundary conditions, the actual

availability of EVs to serve as storage devices is key

to the successful deployment of V2G. Therefore,

gaining insights into car usage patterns is critical for

assessing the practical potential of V2G.

Figure 1: Grid congestion for consumption (left) and feed-

in (right) in the Netherlands on December 9, 2024

(Capaciteitskaart 2024).

This paper examines car usage patterns using real-

world mobility data from the Netherlands Mobility

Panel (Mobiliteitspanel Nederland, MPN). The MPN

includes travel diaries from individuals residing in the

Netherlands and offers comprehensive household-

level information such as details about household

members, vehicle characteristics (e.g., engine type),

and trips made.

The structure of the paper is as follows: Chapter 2

reviews relevant literature, highlighting key studies

and gaps in the context of V2G applications.

Chapter 3 presents an overview of the dataset and

details the data preparation process, including

filtering criteria and the estimation of trip departure

times. In Chapter 4, car usage profiles are developed

based on individual mobility patterns, and clustering

analysis is conducted to identify relevant usage

patterns for weekdays and weekends. The clustering

analysis results are shown and discussed in

Chapter 5, followed by a conclusion outlining the

findings' implications and directions for future

research.

2 STATE OF THE ART AND

LITERATURE REVIEW

V2G technology has been extensively studied, with

early research primarily focusing on its technical

feasibility and potential benefits, such as peak-load

shaving and reductions in total generation costs

(Zheng et al., 2019). Kempton and Tomić (2005)

demonstrated the viability of using EV batteries for

grid stabilization and renewable energy integration.

Their research highlights that automobiles are

typically used only about 4% of the time, suggesting

that V2G systems could utilize the parking time of

electric vehicles to store and supply energy. This

foundational work laid the groundwork for further

studies exploring V2G's applications, including

frequency regulation, peak shaving, and renewable

energy integration.

Building on these technical foundations,

subsequent research addresses gaps in understanding

user behavior and mobility patterns for V2G

implementation. Noel et al. (2019) highlighted a lack

of research into user behavior in the context of V2G

and stressed the importance of incorporating mobility

patterns into V2G planning. Their findings revealed

that a typical vehicle is used for driving only 4-5% of

the day.

Several studies have delved into the relationship

between mobility patterns and parking durations,

showing the temporal availability of EVs for grid

services. For instance, Fu et al. (2021) used travel

surveys from the German Mobility Panel (MOP) to

identify the V2G potential of passenger cars. They

applied a two-level clustering method to analyze

driving and parking patterns, focusing on parking

locations and durations. The study identified ten

different weekday driving patterns, highlighting a

significant potential for V2G participation. Similarly,

Demirci et al. (2023) noted that many studies fail to

consider how driving and charging behavior patterns

influence V2G integration. Their research proposed a

framework for processing EV driving and charging

behaviors to improve charging management

operations, incorporating recent advancements and

real driving data. By evaluating attributes such as

charging location, charging duration, charging levels,

and charging times, the study aims to create a realistic

and consistent dataset reflecting new electro-mobility

trends.

Crozier et al. (2018) clustered data from the UK

National Travel Survey to identify five typical

conventional vehicle usage profiles. They found that

70% of vehicles fall into the lowest usage group,

while 30% account for 65% of total fleet mileage.

These findings emphasize the importance of

identifying underutilized vehicles as potential

candidates for V2G integration. Sovacool et al.

(2017) highlighted a significant gap in research on

customer acceptance and driving behavior in the

context of vehicle-grid integration.

VEHITS 2025 - 11th International Conference on Vehicle Technology and Intelligent Transport Systems

190

Using real-world mobility data, such as the

German travel survey analyzed by Fu et al. (2021),

has significantly advanced our understanding of

users' driving behavior. Building upon these methods,

this study leverages data from the MPN, which

provides comprehensive information on household

travel patterns. To the best of our knowledge, this is

the first study to process and analyze real-world

mobility data from the MPN to gain comprehensive

insights into the general car usage profiles of the

Dutch population. By focusing specifically on home

and workplace parking durations, this research offers

a nuanced understanding of the temporal availability

of vehicles for possible V2G applications. These car

usage profiles form the foundation for assessing V2G

potential in the Netherlands.

3 DATA: OVERVIEW AND

PREPARATION

This chapter details the most recent MPN dataset of

2022 and its preparation for analysis. A

comprehensive filtering process was applied to

ensure data reliability. Trip departure times were

estimated to improve the temporal accuracy of car

usage profiles, as the raw data provided only

aggregated time intervals.

3.1 Netherlands Mobility Panel

The MPN (Hoogendoorn-Lanser et al., 2015) gives

insights into the travel behavior of fixed groups of

individuals and households since 2013 of the Dutch

population. Participants in the panel maintain a travel

diary over three consecutive days (including

weekdays and sometimes weekends) to record their

mobility patterns. However, the MPN dataset only

allows a general distinction between day types

(weekday or weekend), as it does not specify the

exact day of the week. These anonymous travel

diaries capture detailed information such as travel

times, modes of transport, the type of start and end

location, and the purpose of each trip. Additionally,

the diaries are supplemented with personal data (e.g.,

job or driver's license) and household details (e.g.,

number of people or cars).

Based on the travel diaries, parking times of cars

at home, work, and other locations can be identified

through a series of processing steps. The analysis

begins with data preprocessing, which includes

filtering out implausible entries and irrelevant

households. Although the original MPN dataset was

designed to be nationally representative, the sample's

representativeness after filtering was not explicitly

assessed. However, as filtering was based on diverse

criteria, significant biases in spatial distribution are

not expected.

3.2 Data Filtering

To ensure the reliability and consistency of the

dataset, a filtering process was applied. The filtered

dataset includes only complete households with at

least one licensed driver and one car. A complete

household is one, as Hoogendoorn-Lanser et al.

(2014) defined it as when all members aged 12 and

older fully complete the three-day online travel diary.

Incomplete households were not included in the

analysis, as missing data could compromise the

accurate reconstruction of vehicle usage behavior.

This step is crucial because individuals with car usage

without a completed travel diary would result in

incomplete or misleading data for household car

usage. Table 1 outlines the sequential filtering steps

and the corresponding number of households retained

at each stage. The initial dataset consisted of 3,108

households. First, households without at least one

member completing the full three-day online travel

diary were excluded, reducing the sample to 2,505.

Next, households without at least one licensed driver

were removed, leaving 2,227 households. A further

refinement excluded those without at least one car,

leaving 2,059 households. Finally, only complete

households that met all three previous criteria were

included in the final dataset of 1,661 households,

which formed the basis for detailed analysis.

Table 1: Filtering process to identify relevant households.

Travel diary

≥ 1

Driver’ s license

≥ 1

Car

≥ 1

Complete

households

Number of

households

3,108

x 2,505

xx 2,227

xxx 2,059

xxxx 1,661

Table 2 provides an overview of the relevant

information extracted about the individuals and cars

in these households.

Understanding Car Usage Patterns for V2G Integration: Insights from Dutch Travel Diaries

191

Table 2: Details of relevant households.

umber of persons 3,246

umber of persons with driver’s licenses 2,628

umber of cars 2,118

umber of households with fewe

cars

than persons with driver’s licenses

546

In 546 households, there are fewer cars than

individuals with driver’s licenses, indicating that car

sharing is necessary among household members. This

aspect is particularly relevant for the methodology

used to generate car usage profiles, as it directly

impacts vehicle availability and utilization patterns.

A "record," as referred to in

Table 3, can be either

a trip segment or a survey day without any recorded

trips by the members of the household. A trip must

consist of at least one trip segment. If the mode of

transport is changed during a trip or there is a brief

interruption, but the destination remains the same, a

new trip segment is created. The dataset contains

22,487 trip segments and 1,696 records representing

days without trips.

Table 3: Overview of records in the dataset.

umber of records in travel data 24,183

umber of records in travel data with

transport mode ca

8,828

The trip segments in the dataset lack detailed

information such as duration, segment destinations,

or mileage. Instead, the dataset only provided

aggregate information for the entire trip (e.g., total

duration, final destination, and total mileage). This

limitation did not pose issues for trips where all

segments shared the same transportation mode.

However, it became problematic for trips involving

multiple modes, particularly those where the car was

used as the driver for at least one segment. Without

detailed segment-level data, it is impossible to

reconstruct car usage or meaningfully analyze such

trips comprehensively. To address this issue, all

households where members recorded trips with

multiple transportation modes and at least one trip

segment involving a car as the driver were excluded

from the relevant dataset. This adjustment impacted

72 households, resulting in a refined dataset of 1,589

relevant households with 2,027 cars. Among these

households, 985 have an equal number of cars and

individuals with driver’s licenses, indicating a one-to-

one correspondence between vehicle ownership and

potential users. In contrast, 522 households have

fewer cars than individuals with driver’s licenses,

which suggests that car sharing is necessary among

household members. Additionally, 82 households

have more cars than individuals with driver’s

licenses, indicating surplus vehicle availability.

Some inconsistencies were identified in the data

set. In the travel diary, specific entries show

individuals with two consecutive trips, both labeled

with the objective "going home," where the first trip

was not part of a round trip. Such a trip chain is

logically impossible, as an individual already at home

cannot take another trip with the objective of "going

home" without leaving first. This inconsistency was

observed in 18 instances, which were carefully

analyzed. It was determined that, in most cases,

respondents appeared to forget to mark a subsequent

trip as part of a round trip. For example, a respondent

might return "to home" by car and then take a walk,

also marked as "to home." In such cases, the walking

trip was corrected and classified as a round trip.

All respondents completed the travel diary in at

least three days, though three recorded trips occurred

before the first official day of the survey. To maintain

consistency and comparability across the dataset,

only the data from the officially recorded days for

each respondent were included in the analysis.

3.3 Estimating Trip Departure Times

One of the most critical steps in the preprocessing of

the data was the estimation of the departure times of

the trips in the travel diaries. In the original dataset,

departure times are summarized in predefined time

classes, as shown in Table 4.

Table 4: Allocation of departure time classes to time ranges.

Departure Time Class Time Ranges

1 00:00 to 04:00

2 04:00 to 07:00

3 07:00 to 08:00

4 08:00 to 09:00

5 09:00 to 12:00

6 12:00 to 13:00

7 13:00 to 14:00

8 14:00 to 16:00

9 16:00 to 17:00

10 17:00 to 18:00

11 18:00 to 19:00

12 19:00 to 20:00

13 20:00 to 23:59

However, departure times with a higher resolution

are essential for the creation of vehicle usage profiles

for later investigation of V2G potential. In this study,

a time resolution of five minutes was chosen to

improve accuracy. The process of estimating trip

VEHITS 2025 - 11th International Conference on Vehicle Technology and Intelligent Transport Systems

192

departure times is shown in Figure 2. At the

beginning of the process, empirical normalized traffic

count data at a one-hour resolution is interpolated to

a five-minute resolution, incorporating departure time

classes. Next, trip times for MPN travel diaries are

estimated based on the empirical traffic count

distribution. Finally, the refined normalized departure

times are compared with the empirical traffic counts

to ensure consistency and accuracy.

Figure 2: Process of estimating trip departure times.

Time series of traffic counts for the Netherlands

were analyzed to estimate trip departure times with a

five-minute resolution, referring to the empirical

work of Andriesse et al. (2021)

The empirical traffic count time series used

(orange curve in Figure 3) is the normalized one of

the Waterleidingstraat, Rosmalen, as reported by

Andriesse et al. (2021). This time series of traffic

counts was taken from a cross-section of the street,

covering both directions of traffic, in order to reduce

peaks in a single direction. Although other traffic

count time series could also be applied, the one

presented here is representative. The time series

chosen was the one that was most similar to the MPN

departure time classes in the normalized form at one-

hour resolution (blue curve in Figure 3). This curve

was constructed by summing all values within each

time class, normalizing them, and converting them to

hourly values based on the intervals defined in Table

4. In particular, the MPN time series of departure

times shows a less pronounced drop after the morning

peak hour and a more subdued afternoon peak than

the empirical time series.

To generate departure times in five-minute

resolution, the trips were assigned within their

corresponding time ranges while ensuring adherence

to observed traffic patterns. The goal was to align the

values within each time class with the empirical time

series of traffic counts as closely as possible. To

achieve this, the empirical curve from Andriesse et al.

(2021) was divided into the same time classes as the

MPN data for direct comparison. Every class was

normalized, and all values from each class were

interpolated into five-minute intervals for higher

resolution. Then, trip departure times were assigned

based on the empirical normalized times series of

traffic counts for each time class.

The departure times directly affect the calculation

of arrival times, which are calculated by adding the

travel time, rounded to the nearest five-minute

interval, to the departure time. For travel times

between one and four minutes, the values are always

rounded to five minutes to ensure they remain

recognizable as trips.

Additionally, constraints were implemented to

ensure data integrity, which may have caused slight

deviations from the empirical curve.

 If a person makes a trip within a specific time

range on a given day, any subsequent trip (with a

higher trip ID) can only start after the preceding

trip has ended.

 All records were adjusted to conclude at the end

of the survey day for consistency. For instance, if

a trip extended into the following day (e.g.,

ending at 01:25), it was truncated at 23:59 of the

survey day. In the 77 cases where this occurred,

the record reflects that the individual was still

traveling at the cutoff time.

 While the departure times were aligned

with the empirical time series of traffic counts,

the MPN time series of traffic counts (red curve

in Figure 3) is additionally influenced by the

recorded travel times. This curve represents the

normalized time series of traffic counts at a one-

hour resolution, calculated based on the time

trips occur—encompassing everything between

departure and arrival times. It demonstrates that

the calculated times align well with the empirical

traffic counts, except for a slight deviation at the

midday peak.

4 METHODOLOGY:

IDENTIFICATION OF CAR

USAGE PATTERNS

The principal aim of this study is to derive typical car

usage profiles from the MPN data. Thus far, the

analysis has considered complete travel diaries

Understanding Car Usage Patterns for V2G Integration: Insights from Dutch Travel Diaries

193

encompassing trips made via various transport

modes. This section outlines the methodology for

creating car mobility profiles, with a specific focus on

trips where the transportation mode is recorded as

"car as a driver."

The process comprises three principal stages.

Initially, car usage profiles are generated for each

individual based on the recorded trips. Subsequently,

these profiles are aggregated to create actual car

usage profiles for cars. Finally, clustering algorithms

are employed to identify distinct usage patterns. The

car usage profiles are generated based on recorded

travel diaries, thereby providing insights into overall

usage behavior without distinguishing between EVs

and internal combustion engine vehicles. This

approach ensures a comprehensive analysis of trip

chains across the entire vehicle fleet.

4.1 Creation of Car Usage Profiles per

Individual

A five-minute mobility profile (288 entries) was

created for each person and survey day to capture

individuals' daily car mobility patterns. Each profile

tracks the individual's location and activity status

throughout the day with the car, represented by the

following states: H (Home), W (Work), and O (Other

locations).

If no trips with a car were recorded for an

individual on a specific day, the profile was filled

entirely with H (Home), which means the individual

had no car activity that day.

For individuals with recorded car trips, the

departure and arrival times were used to mark periods

of driving (D) and location changes based on the

trip’s purpose (H, W, and O).

4.2 Creation of Car Usage Profiles

The car mobility profiles of individuals are

aggregated into car usage profiles per car and day

type. Therefore, the households were classified into

three groups based on the ratio of vehicles to licensed

drivers:

 Households where the number of cars matches

the number of individuals with driver’s licenses

 Households with more vehicles than licensed

drivers

 Households with fewer than licensed drivers

For households with the same number of cars and

licensed drivers, each vehicle was directly assigned to

a single individual’s car usage profile.

For households with more cars than licensed

drivers, unused vehicles were assumed to remain at

home (H) throughout the day.

Conversely, for households with fewer vehicles

than licensed drivers, individual mobility profiles are

aggregated to simulate shared car usage. This

approach aggregates vehicle usage profiles within

households based on data on car usage profiles per

individual. These data are assigned to vehicles, with

vehicle locations updated accordingly to reflect their

use. The final output is a structured dataset that

captures the usage patterns of each car throughout the

day.

The final dataset offered vehicle usage profiles,

detailing each car's location and activity status

throughout the day and distinguishing between

weekday and weekend patterns, as shown in Table 5.

Table 5: Excerpt of car usage profiles per car.

Car ID

Usage

Profile

Day Type

3026

…WWWWWWWWWWW

WWWWDDDDDDDDDD

DDDDDDDDDDHHHH…

Weekday

3005

…HHHHHHHHHHHHHH

HHHDDDDDDDDDDDDD

DDDDDDDOOOOOOO…

Weekend

4.3 Clustering of Car Usage Profiles

The k-medoids clustering algorithm, using Hamming

distance, was applied to the aggregated car usage

profiles, to identify patterns in car usage on weekdays

and weekends. A critical parameter for the k-medoids

algorithm is the number of clusters, which

significantly influences the interpretability and

accuracy of the clustering results. Both the elbow

method (Thorndike, 1953) and silhouette analysis

(Rousseeuw, 1987) were employed to determine the

optimal number of clusters.

The elbow method evaluates the total within-

cluster variance (inertia) for different cluster counts.

The "elbow" point, where the variance rate decreases

significantly, indicates the optimal number of

clusters. The silhouette analysis measures the

cohesion and separation of clusters, with higher

silhouette scores suggesting better-defined clusters.

According to Januzaj et al. (2023), the highest

silhouette score generally corresponds to the optimal

number of clusters.

VEHITS 2025 - 11th International Conference on Vehicle Technology and Intelligent Transport Systems

194

For weekdays, the silhouette analysis indicated

the highest score with two clusters, as shown in

Figure 3. However, this result was deemed

insufficient to capture the diversity of car usage

patterns, as it oversimplifies the behavior observed in

the dataset. While not yielding a definitive "elbow,"

the elbow method suggested a potential range

between four and six clusters. Upon analyzing results

with more than six clusters, it became evident that the

additional clusters offered little meaningful

distinction. For example, with seven clusters, two

clusters represented shopping and leisure activities

with only minimal differences in timing, making

them difficult to interpret or justify as separate

groups. Therefore, six clusters were selected for

weekday data to balance interpretability and detail.

Although the silhouette score for six clusters was

slightly lower than for five, the additional cluster

provided more nuanced and precise insights into car

usage profiles.

Figure 3: Determining the number of clusters for weekdays.

For weekends, the elbow method suggested a

similar range, with a noticeable bend around four or

five clusters (see Figure 4). However, the silhouette

score analysis showed a significant decline in

cohesion with five clusters compared to four. Based

on these findings, four clusters were chosen for

weekend data.

In summary, six clusters were selected for

weekdays and four for weekends, showing

characteristics of Dutch vehicle usage during these

periods.

Figure 4: Determining the number of clusters for weekends.

5 RESULTS

The six distinct car usage patterns for weekdays are

illustrated in Figure 5. Each cluster represents a

typical car usage profile, showing obvious differences

in the temporal distribution of vehicle activity states

(home, work, other locations, driving).

The largest cluster (#1), representing 71.7% of the

dataset, is characterized by cars that remain

predominantly at home throughout the day with

minimal driving activity. This cluster indicates a

significant portion of vehicles that are primarily

parked, suggesting a high potential for V2G

applications, as these vehicles are readily available

for energy storage and grid interaction. The

dominance of this cluster reflects the overall low

utilization of vehicles during weekdays, consistent

with findings in mobility studies in Chapter 2, where

most private vehicles remain unused for most of the

day.

The second-largest cluster (#2), comprising

15.0% of the dataset, corresponds to the classic

commuter profile. Vehicles in this cluster are

primarily driven in the morning and evening, with

extended parking durations at work during the day.

This pattern emphasizes the potential for workplace-

based V2G systems.

Cluster #4, comprising 3.6% of the dataset, also

presents a commuter profile, indicating the driving

activities in the morning and evening and parking at

work during the day. However, the only difference

with cluster #2 is that the car is parked overnight at a

different location from home.

Cluster #3, accounting for 4.0% of the dataset,

represents vehicles primarily used for shopping.

These cars exhibit sporadic driving activity

Understanding Car Usage Patterns for V2G Integration: Insights from Dutch Travel Diaries

195

throughout the day and are parked at various non-

work locations for extended periods.

In clusters #5 and #6, representing 3.5% and 2.2%

of the dataset, respectively, the car is parked at

another place at the beginning and at home at the end

of the day. The other place could be a hotel, a second

home, or a partner's place. During the day, cars show

sporadic driving activity. They are parked for

extended periods at different non-work locations,

which could indicate less typical weekday use, such

as non-routine travel for work or leisure. These

patterns highlight the diversity of weekday travel

scenarios, including overnight stays or irregular

driving patterns.

Figure 5: Car usage profiles for weekdays.

Figure 6 presents the clustering results for

weekend car usage, showing four usage patterns.

The largest cluster (#1), representing 81.7% of the

dataset, captures vehicles that remain predominantly

at home throughout the day. This cluster highlights a

significant portion of cars with no activity on

weekends, reflecting limited driving needs during

these days.

The second-largest cluster (#2), comprising 4.9%

of the dataset, corresponds to cars used primarily for

shopping and leisure activities. Vehicles in this

cluster are typically driven mid-morning, parked at

non-home locations throughout the day, and return to

their home by evening.

Cluster #3 and #4, accounting for 4.6% and 3.1%

of the dataset, reflect cars used for extended weekend

trips (e.g., visiting friends, family, or travelling).

Vehicles in cluster #3 start their day at home, are

driven throughout the day, and end at another

location. Cluster #4 describes the opposite, starting at

another place and ending at home. These profiles

align with weekend getaways or extended leisure

trips, where vehicle availability for V2G is limited

during daytime hours. The unequal ratio of vehicles

leaving home and returning at the end of the day may

be attributed to some individuals being driven away

already during the week or to the dataset’s uneven

representation of Saturdays and Sundays. However,

precise information on this distribution is unavailable

because the MPN dataset, as mentioned above, is only

distinct between weekdays and weekends.

Figure 6: Car usage profiles for weekends.

The clustering results reveal distinct differences

in car usage between weekdays and weekends.

Cars are predominantly parked at home during

both periods, with slightly higher home parking times

on weekends. The stationary behavior of vehicles at

home, evident in the largest cluster for both periods,

underscores the potential for V2G applications,

particularly in residential settings.

VEHITS 2025 - 11th International Conference on Vehicle Technology and Intelligent Transport Systems

196

While weekday profiles are dominated by work-

related activity and commuting, weekend profiles

emphasize leisure-related mobility (e.g., for leisure or

shopping) and less work-related parking. The

shopping and leisure cluster highlights midday

availability, while the travel-related clusters capture

more dynamic and less predictable usage patterns.

Driving times remain minimal on both weekdays and

weekends. Table 6 provides a detailed summary of

mean parking times, standard deviations, and

differences based on the ratio of vehicles to driver’s

license holders.

Table 6: Mean times of activity state and standard deviation

for weekdays and weekends for different ratios of cars to

license holders.

Ratio of Cars to

License Holders

State

Mean Weekday

(hours)

Standard Deviation

Weekday (hours)

Mean Weekend

(hours)

Standard Deviation

Weekend (hours)

Less

Parking

Home

19.73 6.12 19.93 6.05

Parking

Wor

1.46 3.19 1.12 2.87

Parking

at other

2.24 4.64 2.41 4.84

Drivin

0.57 1.09 0.54 1.14

Equal

Parking

Home

19.68 6.10 20.06 6.02

Parking

Wor

1.91 3.59 1.41 3.19

Parking

at other

1.89 4.19 2.03 4.43

Drivin

0.52 0.92 0.50 1.14

Parking

Home

21.40 5.14 21.73 4.93

Parking

Wor

1.22 2.97 0.90 2.61

Parking

at other

1.06 3.02 1.09 3.30

Drivin

0.31 0.64 0.27 0.59

The analysis of vehicle usage patterns based on

the ratio of vehicles to individuals with driver’s

licenses shows that households with fewer cars than

driving license holders exhibit higher driving activity

and lower parking durations compared to households

with one or more cars per driving license holder.

Specifically, in households where cars are shared,

vehicles spend 19.73 hours on weekdays, and 19.93

hours on weekends parked at home, and 0.57 resp.

0.54 hours being driven. In households with a 1:1

ratio, the corresponding figures are 19.68 hours on

weekdays and 20.06 hours on weekends parked at

home, and 0.52 resp. 0.50 hours driven, while in

households with more cars than licensed individuals,

vehicles spend 21.40 hours on weekdays and 21.73

hours on weekends at home and are driven for only

0.31 resp. 0.27 hours.

6 CONCLUSIONS AND

OUTLOOK

This study presents a methodology for creating and

analyzing car usage profiles from the Netherlands

Mobility Panel (MPN) data, laying the groundwork

for assessing Vehicle-to-Grid (V2G) potential. In the

data, six car usage patterns are identified for

weekdays and four for weekends. For weekdays and

weekends, the most significant cluster is parking at

home over the whole day. On weekdays, there is also

a substantial share of parking at work, whereas at the

weekends, cars are often parked at other locations

than home or work. Households with shared vehicles

exhibit higher driving activity and lower parking

durations, whereas households with more vehicles

than driver license holders demonstrate longer

stationary periods. The clustering results further

illustrate the diversity in vehicle usage, capturing

patterns ranging from daily commuting to irregular

travel scenarios, such as overnight trips or extended

errands. The breakdown of these usage patterns

highlights that vehicles are predominantly parked at

home and additionally at work on weekdays.

The vehicle usage profiles developed in this study

are critical for evaluating V2G potential, as they

provide insights into when and where vehicles are

stationary and available for grid interaction.

Understanding these patterns enables the

development of tailored strategies for energy storage

and grid stabilization, optimizing V2G integration

into residential, workplace, and public settings.

Future research on creating and analyzing car

usage profiles should prioritize using more recent and

Understanding Car Usage Patterns for V2G Integration: Insights from Dutch Travel Diaries

197

detailed datasets to capture better current trends in

vehicle usage and adoption of electric vehicles (EVs).

Incorporating a classification by specific days of the

week, rather than the general distinction between

weekdays and weekends, would provide a more

accurate representation of mobility patterns, as travel

behavior is likely to vary across different days.

Additionally, improving spatial granularity—such as

distinguishing between urban, suburban, and rural

areas—would allow for a more nuanced analysis of

car usage. This enhanced approach would help

identify regional variations and offer deeper insights

into V2G potential across diverse geographic and

socio-economic contexts.

ACKNOWLEDGMENTS

This research is part of the project MORE – Munich

Mobility Research Campus. The project is funded by

dtec.bw – Digitalization and Technology Research

Center of the Bundeswehr. dtec.bw is funded by the

European Union – NextGenerationEU.

This publication makes use of data from the

Netherlands Mobility Panel, which is administered by

KiM Netherlands Institute for Transport Policy

Analysis.

The first author conducted this research as a guest

researcher at the Department of Transport & Planning

(TU Delft), as part of the V2G-QUESTS project (F-

DUT-2022-0241). This project was funded by the

Dutch Research Council (NWO) under the Driving

Urban Transitions Partnership which is co-funded by

the European Commission.

REFERENCES

Andriesse, Rico, Nico Dogterom, and Joran van Kessel.

2021. “Fietsstraten Buiten de Bebouwde Kom

Aanbevelingen Voor Toepassing En Vormgeving.”

https://www.fietsberaad.nl/getmedia/1138dea5-b287-

4aa0-8dd4-87a7b0a70574/Rapportage-Fietsstraten-

buiten-de-bebouwde-kom-(007309-20201204-R1-

02).pdf.aspx?ext=.pdf (November 27, 2024).

Capaciteitskaart. 2024. “Capaciteitskaart invoeding

elektriciteitsnet.” https://capaciteitskaart.

netbeheernederland.nl (December 9, 2024).

Crozier, Constance, Dimitra Apostolopoulou, and Malcolm

McCulloch. 2018. “Clustering of Usage Profiles for

Electric Vehicle Behaviour Analysis.” In 2018 IEEE

PES Innovative Smart Grid Technologies Conference

Europe (ISGT-Europe), Sarajevo, Bosnia and

Herzegovina: IEEE, 1–6. doi:10.1109/

ISGTEurope.2018.8571707.

Demirci, Alpaslan, Said Mirza Tercan, Umit Cali, and

Ismail Nakir. 2023. “A Comprehensive Data Analysis

of Electric Vehicle User Behaviors Toward Unlocking

Vehicle-to-Grid Potential.” IEEE Access 11: 9149–65.

doi:10.1109/ACCESS.2023.3240102.

Fu, Yuzhuo, Kathrin Walz, and Krzysztof Rudion. 2021.

“Analysis of Driving Patterns in Car Traffic and Their

Potential for Vehicle-to-Grid Applications.” In 2021

IEEE Madrid PowerTech, , 1–6. doi:10.1109/

PowerTech46648.2021.9494995.

Hoogendoorn-Lanser, Sascha, Nina T. W. Schaap, and

Marie-José Olde Kalter. 2015. “The Netherlands

Mobility Panel: An Innovative Design Approach for

Web-Based Longitudinal Travel Data Collection.”

Transportation Research Procedia 11: 311–29.

doi:10.1016/j.trpro.2015.12.027.

Januzaj, Ylber, Edmond Beqiri, and Artan Luma. 2023.

“Determining the Optimal Number of Clusters Using

Silhouette Score as a Data Mining Technique.”

International Journal of Online and Biomedical

Engineering (iJOE) 19(04): 174–82. doi:10.3991/

ijoe.v19i04.37059.

Kempton, Willett, and Jasna Tomić. 2005. “Vehicle-to-

Grid Power Implementation: From Stabilizing the Grid

to Supporting Large-Scale Renewable Energy.”

Journal of Power Sources 144(1): 280–94.

doi:10.1016/j.jpowsour.2004.12.022.

Noel, Lance, Gerardo Zarazua De Rubens, Johannes

Kester, and Benjamin K. Sovacool. 2019. Vehicle-to-

Grid: A Sociotechnical Transition Beyond Electric

Mobility. Cham: Springer International Publishing.

doi:10.1007/978-3-030-04864-8.

Rousseeuw, Peter J. 1987. “Silhouettes: A Graphical Aid to

the Interpretation and Validation of Cluster Analysis.”

Journal of Computational and Applied Mathematics

20: 53–65. doi:10.1016/0377-0427(87)90125-7.

Sovacool, Benjamin K., Jonn Axsen, and Willett Kempton.

2017. “The Future Promise of Vehicle-to-Grid (V2G)

Integration: A Sociotechnical Review and Research

Agenda.” Annual Review of Environment and

Resources 42(1): 377–406. doi:10.1146/annurev-

environ-030117-020220.

Statistics Netherlands (CBS). 2024. “Nearly Half the

Electricity Produced in the Netherlands Is Now

Renewable.” Statistics Netherlands. doi:10/nearly-half-

the-electricity-produced-in-the-netherlands-is-now-

renewable.

Thorndike, Robert L. 1953. “Who Belongs in the Family?”

Psychometrika 18(4): 267–76. doi:10.1007/

BF02289263.

Zheng, Yanchong, Songyan Niu, Yitong Shang, Ziyun

Shao, and Linni Jian. 2019. “Integrating Plug-in

Electric Vehicles into Power Grids: A Comprehensive

Review on Power Interaction Mode, Scheduling

Methodology and Mathematical Foundation.”

Renewable and Sustainable Energy Reviews 112: 424–

39. doi:10.1016/j.rser.2019.05.059.

VEHITS 2025 - 11th International Conference on Vehicle Technology and Intelligent Transport Systems

198