Energy-Management-as-a-Service: Mobility Aware Energy

Management for a Shared Electric Vehicle Fleet

Julien Ostermann and Falko Koetter

Fraunhofer Institute for Industrial Engineering IAO, Nobelstr. 12, 70569 Stuttgart, Germany

Keywords: Energy Management System, Electric Vehicle, Distributed Energy Resources, Cloud-based, Information

and Communication Technology, Optimal Charging, Direct Load Control.

Abstract: The combination of sustainable energy generation and transportation is one of the biggest challenges of the

21st century. In this work, an energy management system is presented which provides energy-management-

as-a-service for electric vehicle fleet operators. Energy production and price forecasts are integrated with

near real-time telematics data from a shared electric vehicle fleet, to optimize charging profiles for multiple

charging sites of fleet operators. For this purpose, system architecture and a proper optimization method

enabling different charging strategies are introduced. The presented system is finally evaluated by real

world model trails and an optimization benchmark.

1 INTRODUCTION

One of the biggest challenges of the 21st century is

the transition to sustainable transportation and

energy generation (Chu and Majumdar, 2012). The

need for ensuring sustainability is not only driven by

rising fuel cost or the increase of CO

emissions, but

also due to economic and political issues. In the field

of transportation, battery electric vehicles (BEV) and

plug-in hybrids (PHEV) help to reduce CO

emissions as well as fossil fuel dependency.

Unfortunately, using these vehicle types incurs high

investment costs, due to the cost of vehicles

themselves and the cost of charging infrastructure.

Nevertheless, these cost issues can be mitigated,

using electric vehicles (EV) in corporate car pools.

The study presented in (Plötz et al. 2014) calculates

that the total cost of ownership (TCO) of EVs is

lower compared to vehicles with internal

combustion engines (ICE) when used in corporate

vehicle fleets due to uniform driving profiles. Thus,

corporate car fleets are a field of high potential for

electric vehicle usage, especially since the study

shows that 60 percent of today’s newly registered

cars are equipped as commercial cars in Germany.

Furthermore, the German Federal Government

(German Federal Government 2009) projected a

market ramp-up of 1 million EVs by the year 2020.

The project Shared E-Fleet (Ostermann et al. 2014)

aims to leverage this potential by providing a

software solution to enable small and medium sized

enterprises which are not able to operate a sufficient

amount of EVs economically on their own. The

solution helps sharing vehicles between companies

and users, realizing cross-company electric car

pools.

Compared to conventional cars, the usage of EVs

in a corporate shared vehicle fleet imposes unique

challenges. Besides the limited range of EVs and

indispensable charging times, concurrent charging of

EVs might result in power peaks at a common

charging site that might violate local grid constrains.

Especially the operation of large fleets will not only

affect fleet operators but also energy grid operators.

These must provide the grid with a suitable amount

of energy during peak times (Clement-Nyns 2010,

Lopes et al. 2010, Deilami et al. 2011). The grid

operators fear that uncontrolled integration of a large

amount of EVs into the distribution grid might have

a huge impact on the grid stability (Lopes et al.

2010). One solution is the integration of EV fleets as

part of the smart grid. In this way, grid operators can

prevent cost intensive grid expansion given the fact,

that information and communication technologies

(ICT) are harnessed to realize coordinated charging

strategies. On the other hand, fleet operators can

utilize coordinated charging to prevent demand

peaks caused by concurrent charging of their vehicle

fleet and thus enable scheduled charging in order to

340

Ostermann, J. and Koetter, F.

Energy-Management-as-a-Service: Mobility aware energy management for a shared electric vehicle ﬂeet.

In Proceedings of the 5th International Conference on Smart Cities and Green ICT Systems (SMARTGREENS 2016), pages 340-350

ISBN: 978-989-758-184-7

preferably consume locally generated renewable

energy as it is produced.

In order to leverage these capabilities of EV

fleets, an energy management system (EMS) is

required which is able to easily integrate with

various system services and business processes to

provide dedicated control of the operation.

Furthermore, it enables coordinated grid to vehicle

(G2V) charging of EV fleets at multiple locations

under consideration of local and decentral energy

production. In the process, forecast services shall be

used to take into account weather-dependent

renewable energy production.

In this work, we present the results of our

research in a mobility aware intelligent energy

management aggregator, serving as an EV virtual

power plant (VPP) as part of the Shared E-Fleet

architecture. It enables a direct load control of

intelligent charging stations. Thus, a centralized

control architecture is introduced, interacting with

multiple ICT components to apply optimized

charging schedules, based on real time needs of a

shared EV fleet.

This work is structured as follows: Section 2

gives an overview of relevant related work. Section

3 introduces the use case of the project Shared E-

Fleet from an energy related viewpoint. Section 4

describes the architecture of the aggregator. In

Section 5 the applied energy optimization algorithm

is outlined. The prototype and evaluation are given

in Section 5. Finally, Section 6 concludes this work

and gives an outlook on future work.

2 RELATED WORK

Considerable research has been contributed already,

investigating the integration of all kinds of EVs into

a future smart grid. Since research in this area

started already when adaption of PHEV began, not

all work does solely concentrate on BEVs.

Nevertheless, both types of vehicles do behave the

same regarding the goal to develop coordination of

charging and smart grid integration. In (Jansen et al.

2010), the authors present a modular VPP

centralized architecture and necessary

communication protocols to realize coordinated

charging for a fleet of EVs as part of the EDISON

project. Similar approaches concerning VPP for EV

fleets with integration of distributed energy

resources (DER) were already explicitly investigated

by (Raab et al. 2011, Vandoorn et al. 2011). In

literature, two different approaches of integration of

DER are discussed. Besides VPP which aggregate

DER units to provide controllability and enable

market participation, micro grids (MG) aggregate

local DER to provide a controllable entity that can

operate in grid-connected and islanded mode

(Vandoorn et al. 2011). In both concepts, an EV

fleet acts as controllable battery storage system.

Beside energy storage systems a MG or VPP can

also include charging stations or a PV plant.

Although the MG and the VPP concepts are similar,

they can be differentiated by seeing a VPP

aggregator as virtual, software-based aggregation

and the MG as physical aggregation of DER units.

In this work, we introduce an approach, not only

focusing on a central VPP acting as aggregator but a

multi-station aggregator, being able to provide

second level control and operate multiple MGs

independently. Thus, it enables the aggregator to

consider preferences and local load management

constraints defined by the respective MG operator.

The later proposed aggregator is able to perform

smart charging. With smart charging, charging

stations are basically provided with predefined

charging profiles. These profiles have to be followed

by the battery management systems (BMS) of the

vehicle. Smart charging has the advantage to enable

charging stations and subsequently each connected

EV to defer charging processes to a later point in

time or directly control the drawn current according

to a given profile. Smart charging for EV fleets is

already addressed in previous research. For example,

in the work of Hu et al. (Hu et al. 2014),

optimization and control methods are summarized to

present an overview of this field regarding smart

charging as part of EV aggregators. The authors in

(Valogianni et al. 2014) are presenting a

management system leveraging the battery storage

capabilities provided by EVs. An extensive review

of smart charging approaches and architectures is

presented in (García-Villalobos et al. 2014).

Especially the sprawl of distributed energy

generation, energy storage systems, privately owned

photovoltaic (PV) power plants, wind power

generation, or combined heat and power-plants

(CHP) presents a challenge for future charging

aggregator systems, making support of smart

charging necessary. The flexibility of EVs as

controllable loads to mitigate uncoordinated

charging impacts was investigated in (Han et al.

2010, Saker et al. 2011, Sundström and Binding

2012, Alonso et al. 2014, Valogianni et al. 2014).

Extensive reviews regarding charge scheduling for

EVs is given by (García-Villalobos et al. 2014,

Mukherjee and Gupta, 2014).

Nonetheless, previous work considered only

Energy-Management-as-a-Service: Mobility aware energy management for a shared electric vehicle ﬂeet

341

optimization methods isolated from productive

systems and not integrated into working prototypes.

VPP aggregators realizing smart charging as part of

a smart grid have been presented (Chynoweth et al.

2014, Lutzenberger et al. 2014, Zuccaro et al. 2014).

Additionally, various researchers developed multi

agent systems to accomplish a VPP aggregator

(Lutzenberger et al. 2014). Although the approach in

(Mültin et al. 2012) is similar to our approach, we

provide a hierarchical controlled VPP, monitoring

and controlling multiple sites as a cloud-based

solution. As proposed in (Hu et al. 2013, Mukherjee

and Gupta 2014), we orchestrate a distributed

service oriented architecture (SOA) as a cloud-based

solution, enabling the aggregator to react in near

real-time to the fleet operation. Hence, we develop a

solution for fleet operators, with the goal to enable

their fleet to participate in the smart grid, without

being dependent on solutions provided by a utility

company, as well as being flexible to scale.

3 SHARED E-FLEET SCENARIO

In the research project Shared E-Fleet (SEF) a

cloud-based solution was investigated which enables

a car fleet operator to manage and provide a fleet of

BEV across several companies at one or multiple

sites. Different works (see (Barth et al. 2000,

Delucchi and Lipman, 2001, Lee et al. 2005))

suggested that increasing the utilization of EV fleets

for example by increasing trips per day, decreases

the TCOs and make them more economic than

combustion engine vehicles regarding short range

trips. Hence, the SEF IT solution provides a system

including mandatory functionalities for car fleet

management like booking, billing and operation, to

be able to realize a corporate car sharing platform.

Compared to other already available commercial

solutions, the SEF solution was intended as an

extensible, service-based cloud-platform to integrate

various fleet management services in a highly

configurable matter, making the operator

independent of vendor specific solutions (Ostermann

et al. 2014).

Unlike state of the art solutions, in the SEF use

case, instead of booking a car, the user books a

mobility demand by defining the start and end point

and time of his business trip. Since SEF uses a

station based car sharing approach, the start and

endpoint of each mobility demand is always at a

dedicated station of the SEF system. Certainly, the

start and end station do not have to be the same.

Until one hour before the respective beginning of a

trip, a booking is not explicitly bound to one vehicle.

Only after reaching this time, it is fixed to a

dedicated vehicle. A one hour time frame was

chosen to assure the user a safe operation of the

system by this feedback and leave a margin in case

of failure. In this way, the disposition optimization

management service reschedules journeys in real

time and reacts to unforeseen issues like delayed

vehicle returns or unexpected state of charge (SOC)

at the time of return (Koetter 2015). The latter one is

especially crucial in EV fleets, since a properly

predicted SOC influences all succeeding, already

booked journeys on the same vehicle. Thus, already

booked trips may be canceled in case of unforeseen

issues. During the booking process, the disposition

optimization assesses the current schedule whether

the user request can be fulfilled even considering a

suitable buffer in time and SOC, but large deviations

of the expected SOC or return time can only be

intercepted by standby vehicles.

Thus, the platform is constantly aware of the vehicle

state. During each trip, the vehicles are monitored,

predicting their estimated return time and their SOC

based on real-time data from an on-board unit

(OBU). Future states of the vehicles at the end of a

journey can be estimated and used for future

optimization procedures. On return of a vehicle, its

user has to reconnect the vehicle to the charging

station, enabling them to recharge for the upcoming

trip.

As part of the SEF ecosystem, a service is

required for managing the charging processes and

energy demand of the vehicles according to grid

constrains and operator specification as part of a

smart grid. Considering that, an energy management

system controls the charging of the EVs at the SEF

charging sites. To perform these tasks, the following

requirements have to be fulfilled:

R1. All EVs of the fleet must always be able to

satisfy the needs of the mobility demand of the user.

Hence, the EMS is responsible for sufficiently

charged vehicles according to the journey schedules

of the disposition management. The schedules for

the vehicles must contain a predefined start and end

time as well as a consumption forecast for each

journey.

R2. Vehicle disposition schedules may change

over time. Due to unforeseen influences, vehicles

may return with different SOC or return late at the

station. The EMS has to update the charging

schedules continuously according to the current state

of the system.

R3. In order to sufficiently fulfill R1 and R2, an

algorithm is required which is capable to calculate

SMARTGREENS 2016 - 5th International Conference on Smart Cities and Green ICT Systems

342

an optimal charging schedule regarding the

constrains of the system. Thus, optimal charging

schedules have to be computed based on different

charging strategies which are ought to be selected by

the operator of the respective fleet ensuring R1 and

R2.

R4. In order to perform tasks as part of a smart

grid, the EMS is supposed to be able to monitor and

control distributed and local energy resources. A

model is required, describing the involved

components and their properties to provide

optimization algorithms with system constraint

boundaries.

R5. Especially renewable energy resources are

dependent on weather conditions. Thus, weather

conditions have to be considered during charge

scheduling. A forecasting system has to be

integrated which provides information about the

prospected energy production of individual

components.

R6. As stated in R1, the state of the complete

EMS is time dependent. The EMS has to be

provided with information about the state of all

relevant system components. The EMS has to store

these states in order to enable the user to keep track

of them at a later point in time.

R7. The EMS must integrate intelligent charging

stations which are capable of directly controlling the

output current. This way, pre-calculated charging

profiles can be applied to the EV.

R8. Furthermore, various fleet operators with

multiple fleet charging sites must be able to use and

integrate their master data management with the

EMS.

R9. The EMS must keep track of the current

energy production and EV charging at the respective

charging sites and always ensure safe operation of

the energy system by staying in system boundaries.

R10. The system must also be able to fulfil non-

functional requirements. Thus, it must be able to

scale in order to provide services to multiple users.

Robustness is required to provide services even in

case of failure. Additionally, the EMS must perform

well, even with a large amount of managed

components.

4 AGGREGATOR SYSTEM

ARCHITECTURE

In the following, we present the architecture of the

EMS, showing how it is able to provide energy-

management-as-a-service. The system architecture

of the complete SEF ecosystem for a shared EV fleet

is described in (Ostermann et al. 2014). The EMS is

part of this ecosystem. Its system architecture is

depicted in Figure 1. The EMS is acting as central

node, integrating all mandatory components to fulfil

all previously stated requirements. All components

are integrated by connecting to their web services.

Through the integration of master data management

(MDM) services, the EMS can obtain properties of

all physical system components. Thus, by defining a

mutual data exchange interface, various MDM

systems can provide information about the deployed

EVs, the energy production resources, the charging

stations and about charging sites of the individual

fleet operator.

Figure 1: EMS system architecture.

An interface to a charging station operator (CSO)

provides the EMS with means to monitor and

control charging stations. CSO are business entities,

owning, managing, maintaining, and operating an

aggregation of charging stations. Note that EV fleet

operators can also be CSOs themselves. In order to

provide the EMS with real-time state data of the

vehicles, a external database is interconnected to the

EMS which stores telematics from vehicles.

Telematic real-time data of the vehicles might be

provided by OEMs in near future. However, today

this data is provided by third-party developers using

proprietary hardware which is amalgamated in

separate provider specific databases.

A forecast provider service supplies the EMS

with up-to-date and day-ahead energy production

forecasts. Similarly, the energy grid operator, e.g.

the transmission system operator (TSO) or

distribution system operator (DSO), is connected

with the EMS. Thus, the operator can either request

the EMS to perform ancillary services on the grid or

send price signals or dynamic day-ahead price

curves. By integrating a building energy

management system, information about locally

generated energy (e.g. by a photovoltaic power

plant) and the energy consumption of the building

can be obtained.

The software architecture is depicted in Figure 2.

The EMS is constructed as a multi-tenant platform.

Energy-Management-as-a-Service: Mobility aware energy management for a shared electric vehicle ﬂeet

343

In this way, collaboration between different

customers (EV fleet operators) can be implemented

instantaneously without the need to

programmatically extend the platform or integrate

multiple instances. Additionally, updates and

maintenance of the platform is more flexible while

ensuring contracted service-level agreements and

quality of service without the need for extensive IT

infrastructure at the customer site (Buyya et al.

2009).

Consequently, managing the tenants of the

platform is a crucial part. A tenant manager is

responsible for managing all data belonging to its

respective customer avoiding and handling cross-

access to other customer’s data. All collected data of

all tenants is stored in one shared database.

Therefore, each fleet operator can configure the

setting of its own controllable charging sites.

Figure 2: Software architecture of the EMS.

A frontend visualizes the system state for each

customer and provides the means to control all

integrated components. A screenshot of the actual

frontend of the EMS prototype is depicted in figure

3 showing the real-time load at a charging site, the

current energy production and an overview of the

vehicle states.

Figure 3: Frontend visualization showing the load and

production parameters and the state of the cars for the

customer.

As depicted in Figure 2, each tenant possesses a

set of the master data of all components in their

system. Furthermore, each tenant possesses a

controller which has timed tasks that run multiple

times a minute to read out the value of the pre-

calculated charging profiles from the charging

schedule and set the power-output set-point to the

charging station while minimizing the error between

production and consumption.

Here, the association between charging station

and physically connected vehicle is important. The

identification of the car which is connected to the

charging station can be done in several ways. Each

user is assigned a user-specific token. A token is an

identification number which uniquely identifies each

user in the system. In this way, the EV can be

identified by looking up the vehicle which was used

by the user who checked in to the charging station.

During this procedure, the token can either be

transmitted via a smartphone application or be read

from a RFID-card in the charging station.

Unfortunately, current EVs do not transmit any

unique identification to the charging station. Hence,

in the SEF project, we used a user identification

which is transmitted to the charging station for

authentication with the system via a smartphone

application.

During charging, the controller continuously

measures the power drawn from the connected EV

and compares it to the pre-calculated profile. Any

occurring deviation is reduced according to the

individually selected strategy. This ensures that grid

limitations and optimization goals are met.

During each charging session of the EV the

smart meter values of the charging stations are

sampled in a 15 minute interval. Using this data, the

consumption module computes time dependent

energy costs of each session. This enables the fleet

operators to use dynamical energy plans of their

DSO.

A charging schedule holds the charging profile

for each EV in the system. It is calculated based on

the EV disposition schedule. As soon as the

disposition schedule is updated by any entity, the

charging schedule is updated subsequently. For

planning the charging schedule, it is assumed that a

vehicle is connected to the charging station as soon

as it returns from its trip. Connecting the vehicle is

mandatory for each user. In case of a missing, but

expected connection, the user is notified. As a

consequence, the time between the end of a

proceeding and the start of an upcoming trip is

considered available for charging. Only if a user

explicitly wishes to charge in the meantime, during

SMARTGREENS 2016 - 5th International Conference on Smart Cities and Green ICT Systems

344

an ongoing booking, this request is taken into

account in the charging optimization. The charging

optimization is performed by the optimizer. Based

on the selected charging strategy, the optimizer

schedules and performs an optimization calculation

which is described in detail in the next section.

Through this architecture a multi-tenant, flexible and

scalable EMS aggregator is realized connecting

multiple sites respectively integrating multiple self-

sufficient micro grid into one control unit.

5 CHARGING PROFILE

OPTIMIZATION

The central component of the EMS is the

optimization module. Controlling the charging

processes of a fleet of EVs requires the system to

continuously update the charging schedule based on

the current system state. Two distinctive events

trigger the optimization of the charging plan. Since

the EMS does not plan the disposition of the fleet, it

has to react to changes of the disposition schedule.

Hence, an event is sent by the disposition optimizer,

notifying the EMS about the updated or changed

disposition schedule. Even without disposition

changes, the charging schedule is periodically

updated in constant intervals by the EMS. Thus, it is

always able to react to the actual state of the system.

In this way, deviations of the expected SOC can be

taken into account. A task manager triggers an

optimization every 15 minutes if the schedule has

not changed in the preceding 5 minutes. On this

way, the optimization process has enough time to

perform a complete iteration before the next start is

triggered.

Especially for larger fleets, a fast optimization

process is mandatory, since the amount of vehicles

in the fleet have an impact on the computation time.

This can be reached for example by reducing the

complexity of the optimization problem to speed up

computation time. Different authors already found

out that linearization of the charging processes

respectively the battery behavior of EVs is sufficient

for coordinated smart charging. Nonlinear

approximation of the charging behavior does not

justify the increase in computation time (Sundstr.

2010; Hu et al., 2013). Because of that, we followed

the recommendation and chose linear programs (LP)

(Rardin, 1997) to perform optimized charging

profiles for EV charging. Thus, we picked a LP to

describe the constrains of the system at hand and

solve the objective function considering a

specifically selected charging strategy. For the

algorithm we discretized the time into timeslot

intervals of 15 minutes. Hence, a day of 24 hours

has 96 individual timeslots. We designed the

objective function of our LP in order to minimize

cost.

Based on the selected strategy the cost vector can

either be a time dependent energy price or the

amount of CO

emission produced by a certain

energy source when used for charging. Thus, an

economic or ecologic charging strategy can be

realized by adjusting the cost vector appropriately.

This way, the optimizer is capable of calculating the

optimal amount of power used at a designated

timeslot under consideration of the time dependent

cost of a specific energy source. Therefore, the

objective function of our LP looks as follows:

min  

,,





;



;



; ;

(1)

Here,  describes the index for the vehicle and 



the maximum number of vehicles available to

optimize. Furthermore,  describes the index of the

energy source and 



the maximal number of energy

sources available to use. Each timeslot is denoted by

the index , whith 



being the number of total

timeslots and by that the total timespan to be

optimized. Following, the parameter 

,,

is defined



,,



,

∗

,,

∗0.25,

(2)

where parameter 

,,

 from equation 2 being

the cost of the energy charged into a EV v, using

energy source g at timeslot t. The parameter to be

optimized by the solver will be the output power



,,

. With the objective function given, the

following constrains are limiting the solution space.

First, the sum of the allocated power for charging all

vehicles using an energy source at timeslot t being

denoted as 

,

must be lower or equal to the

forecasted amount of power for this respective

timeslot 



,

 :



,



,,

,…,



(3)



,





,

(4)

Vehicles can only be charged while connected to a

Energy-Management-as-a-Service: Mobility aware energy management for a shared electric vehicle ﬂeet

345

charging station. Thus, by using a vector modelling

the connection of the EV to the charging stations,

the charged energy up to the trip k subsequent to a

charging session, can be defined to:



,



,

  

,

∗

,,

…





…



(5)



,





(6)



,



,

(7)

where 

,

is the total amount of energy charged for

EV v, 

,

the maximum amount energy which

can be charged into the battery of EV v.

Further,

,,

describes the parameter, denoting each

timeslot t in which the EV v is connected to a

charging station up to the beginning of trip k. The

parameter 

,

describes the state of the connection

with 

,

1 as connected and 

,

0 as not

connected to the charging station. Equation 5

ensures that the sum of the energy charged up to the

start of trip k plus the initial SOC of the EV 

,

has

to be at least the size of 



, the amount of energy

required for trip k. Equation 7 ensures on the other

side, that the charged amount of energy does not

exceed the battery capacity of the respective EV. In

addition, assume 

,

being the sum of the energy

charged at timeslot t:



,



,,

,…,



(8)

At any time, the amount of energy 

,

drawn from

the charging station should neither exceed its

maximal power output capabilities nor the maximal

input power of the respective EV v, letting the

maximal power output be set to 

,,



,



,,

(9)

Additionally, charging has always to be limited to

the times the EV is physically connected to the

charging station. From this follows:



,



,

∗

,

(10)

If the vehicle is not connected the parameter 

,

this timeslot has to be zero.

If these constrains can be satisfied, the optimizer

calculates the best fitting charging profile for the

vehicles. Note, that the problem space of the

described LP scales linearly with the amount of

regarded EV, energy sources and days. Computation

time scales linearly with the size of the problem

space. Both can be reduced by either simplifying the

constrains to only focus on one single energy source

making the sum of all forecasts the maximum

available power or by reducing the amount of

regarded timeslots by decreasing computation range

or increasing slot duration.

In the Shared E-Fleet project, we focused on two

strategies for charging plan generation. We

differentiated between economic and ecologic

charging. Based on price forecasts only time

intervals will be considered for charging which are

ultimately required for charging the car to a specific

SOC necessary to perform a journey or which offer

considerably lower prices in comparison to another

time of day. With an ecologic strategy the algorithm

is supposed to charge the EVs mainly at times of the

day when renewable energy is generated.

Subsequently of the computation, the set points for

the maximum allowed output power is transformed

in a charging profile and stored as a charging

schedule. Based on the chosen strategy, different

charge profile schedules may emerge from the

calculations. After having stored the charging profile

for each vehicle, it is possible to calculate an

estimation for the SOC of each EV at a specific

point in time based on current system state. Thus, it

can be validated if the computed charging profiles

are able to perform all booked trips sufficiently.

6 PROTOTYPE AND

EVALUATION

Based on the proposed system architecture, we

developed a multi-tenant web application using Java

as a framework system. Hence, it was possible to

evaluate the previously introduced architecture and

algorithms as part of two model trials at two

charging site locations. As parts of this system we

integrated intelligent AC Level 2 charging stations at

each location with a capable maximum power output

of 22kW with our EMS. However, the here applied

EVs, four BMW i3, which were selected due to

overall project requirements, were only capable to

draw a maximum power of 4.6 kW. The vehicles

were equipped with specifically developed

telematics units, sending data to a real-time data

pool. This way, data is read out from the vehicles’

CAN-Bus. The collected data included among others

SMARTGREENS 2016 - 5th International Conference on Smart Cities and Green ICT Systems

346

the SOC, position, and estimated range. Data was

continuously sent to the data pool in five minute

intervals via a GSM module. The charging stations

were connected through a proprietary manufacturer

specific interface. This enabled the EMS to set the

maximum power output, as the acquired charging

stations were only supporting Open Charging Point

Protocol (OCPP) version 1.5 (Open Charge Alliance

2015). At the time of purchase, OCPP version 2.0

which supports power output set-point specification

had neither been finalized nor been implemented in

the hardware by any manufacturer.

Additionally, we monitored a PV power plant at

one of the locations continuously as a reference for

locally generated energy. Although it was not

physically connected to the charging stations due to

regulatory issues, the obtained data served as real

time data to model the behavior of a real PV plant.

In addition to the real time data of the PV plant, we

connected a production forecast system called

PVCast (Klein 2013). PVCast is a commercial

service, which predicts the generated power of a

given PV power plant, based on previously

measured data and increases its accuracy by new

measurements. Master data of charging stations,

vehicles and charging sites is provided by services

of the SEF system. As part of the SEF system, a

dynamic disposition management reschedules the

vehicles according to the vehicle SOC states. In the

prototype, the disposition management is connected

with the EMS and triggers it as soon as the schedule

is changed. With every trigger, it transmits the

complete disposition of the vehicles including the

start and end time of each trip, the distance and the

expected consumption.

With all services connected to the EMS, new

charging schedules are created as soon as the

disposition schedule changes. Unforeseen changes in

the states of the vehicles are mitigated by

periodically conducted optimizations taking into

account the holistic system state. Through that,

requirements R1 and R2 are fulfilled. Applying the

previously presented LP different charging strategies

are possible, fulfilling the requirement R3.

Furthermore, the integration of a building energy

management provides the EMS with data about the

connected electric power system components.

Therefore, R4 is fulfilled. Requirement R5 is

fulfilled by introducing a forecast system providing

the EMS with day-ahead energy production

information. Through the integration of a database,

storing the states of all integrated components and a

frontend, history data is always accessible fulfilling

R6. In the prototype, intelligent charging stations

were integrated and direct load control was possible.

By this, requirement R7 is fulfilled. The requirement

R8 is fulfilled by using a tenant manager in the

software architecture to support different MDM

clients. Furthermore, the tenant manager enables the

EMS to be dynamically scalable fulfilling R10.

However, the model trails could not be used to fully

evaluate all requirements. Due to real time operation

and its prototypic state, the system was not running

without interruption. Additional to that, not all EVs

were available at the same time because of

maintenance reasons. Because of that, a synthetic

scenario was created, using a fleet of 30 EVs This

scenario is similar to the situation at our institute

campus, if all vehicles with ICE were replaced with

BEV. Furthermore each of the EVs has its own

charging station. In this way, the vehicle will always

have the ability to charge . It is assumed that the EV

as well the charging stations are homogenous. For

the sake of the project context, the charging

characteristic of a BMW i3 was used to model the

charging behavior. That means that each vehicle has

a battery capacity of 18.8 kWh which is useable for

storing energy. The average power consumption of

the vehicles is assumed to be 12.3 kW per 100 km

distance. Making each car able to reach a total range

of approx. 147 km. Each vehicle can charge a

maximum power of up to 4.6 kW. That would

require the transformer to handle a maximum

concurrent peak power demand of 138 kW. To

simulate coordinated charging, the maximum usable

power is limited to 100 kW. Any additionally

required power can be provided by a physically

connected PV power plant at the charging site where

the charging stations are setup. For the evaluation,

the grid as well as the PV power plant serve as

energy source, supplying the charging station and by

this the connected EVs, to always provide energy

when needed. To present the results of the

optimization algorithm appropriately, three different

disposition schedules are applied to the fleet. Each

profile is applied to a third of the fleet. If EVs were

charged concurrently in this condition, they would

exceed the given local maximum power. The

consumption of each trip is assumed to be linear

according to the EV’s average consumption per

distance. The three profiles are listed in Table 1.

In this evaluation, optimizations regarding an

ecologic and an economic strategy are conducted.

Each result is compared with dumb charging. Dumb

charging means that the vehicles are charged directly

after plugging in the cable of the vehicles into the

charging station. The charging session is then

performed as long as the EV is connected and not

Energy-Management-as-a-Service: Mobility aware energy management for a shared electric vehicle ﬂeet

347

Table 1: Applied disposition schedule profile types.

Profile

Number

Start

Time End Time

Distance

[km]

Consumption

[kW]

05:30:00 14:10:00 65 8,00

17:30:00 18:30:00 85 10,46

21:00:00 23:30:00 45 5,54

07:30:00 11:30:00 75 9,23

14:45:00 16:00:00 30 3,69

18:00:00 22:30:00 100 12,30

05:30:00 09:30:00 55 6,77

15:00:00 18:30:00 30 3,69

fully charged. If not directly controlled by the

charging station itself, concurrent charging of the

vehicles would lead to stress on the transformer at

the site which might even lead to an overload or

fatal hardware failure. An production forecast is

created using a forecast given by PVCast with an

adjusted peak power of 40 kW. Additionally, an off-

peak tariff is assumed. In the peak time, between 6

am and 8 pm, the energy costs 11 cent/kWh,

whereas in the off-peak time at the rest of the day,

energy costs 7 cent/kWh. The cost of using solar

energy is set to be 9 cent/kWh for the whole

regarded time period and hence being cheaper than

the energy from the grid in the peak time period.

Each optimization is performed considering a time

span of 24 hours, resulting in a total time span of 96

timeslots. Each vehicle is supposed to be completely

empty at the start of the day. As a solver, the solver

of the free Apache Commons Library (The Apache

Software Foundation 2016) was used. The result of

the economic strategy is depicted in Figure 4.

Figure 4: Comparison of optimization results of economic

charging strategy and dumb charging.

Figure 4 compares dumb charging with economic

charging. In the applied scenario, the concurrent

charging is limited by the local load constrain to the

maximum power limitation of 100 kW. Only as

much energy is charged as is required to conduct the

trips. All charging operation is only performed

during the morning hours of the day, due to the low

energy costs. With dumb charging, the fleet would

be charged two additional times during the day.

With a smart charging strategy, this is not necessary.

Since in case of the economic and the ecologic

strategy, only the trip demand is charged, 64% of the

energy can be saved compared to dumb charging

and thus being more energy efficient. Regarding the

energy cost, in the economic scenario, 30% of the

energy cost could be saved. In Figure 5, the results

of optimization with the ecologic optimization

strategy are depicted.

Figure 5: Comparison of optimization results of ecologic

charging strategy and dumb charging.

In Figure 5, it can be observed, that charging in the

morning is deferred and continued in the afternoon,

when solar energy is available. Thus, CO

emission

is mitigated by using renewable energy instead of

energy from the grid. With the ecologic strategy, it’s

possible to save even 76% of CO

emission by not

immediately start charging the EVs and by deferring

the charging sessions into times of renewable energy

production. This synthetic evaluation shows, that

local grid constrains are not exceeded, and thus

requirement R9 is always fulfilled. Furthermore,

requirement R5 is satisfied by taking into account

the production forecast of the PV power plant with

the ecologic evaluation.

By fulfilling all requirements, a fully functional

framework is created, serving as EMS and being

able to perform smart charging as part of an

intelligent energy management service.

7 CONCLUSION AND OUTLOOK

In this work, we introduced the architecture of a

cloud-based EMS, serving as a scalable and flexible

SMARTGREENS 2016 - 5th International Conference on Smart Cities and Green ICT Systems

348

system to monitor and control the charging of a fleet

of EVs. It enables fleet operators to integrate their

energy components and a fleet of EV into an EMS,

providing them with the means to systematically

improve the usage of these components. For this, we

included an optimizer to calculate charging profiles

for an EV fleet, in order to exploit energy production

forecasts and dynamic price tariffs. The evaluation

showed that optimization of the charging profile

delivered charging profiles which sufficiently served

the mobility demands of the user, kept the

boundaries of the energy system and minimize costs.

By this, the here presented aggregator can

accomplish primary objectives of energy

management systems like increasing energy

efficiency, reduction of the energy used for charging

and maximization of profits by minimization of

costs. (Barney et al. 2008)

In the future, we plan to extent this approach to

completely control different micro grids in islanded

mode operation resulting in a smart micro grid. By

providing this framework for a cloud-based and

flexible EMS, we plan to integrate more energy

components, further investigating cloud-based

hierarchical control. Thus, enabling the combination

of free floating EV fleets with multiple micro grid

controls to provide ancillary services and to

maximize profits and simultaneously stabilizing the

grid. As a consequence, we will conduct further

research to decrease runtime of optimization

processes in order to tackle a growing amount of

optimization parameters. Especially interesting is the

coordination of EMS optimization and disposition

planning considering different charging sites and

energy-aware routing of vehicles.

ACKNOWLEDGEMENTS

This research has been supported by the IKT II

program in the Shared E-Fleet project. They are

funded by the German Federal Ministry of

Economics and Technology under the grant number

01ME12105. The responsibility for this publication

lies with the authors.

REFERENCES

Alonso, M., Amaris, H., Germain, J. G., and Galan, J. M.,

2014. Optimal charging scheduling of electric vehicles

in smart grids by heuristic algorithms. Energies, 7 (4),

2449–2475.

Barney, L. C., Wayne, C. T., and William, J. K., 2008.

Guide to Energy Management. Lilburn, GA 30047,

US: 6th Edition, 1.

Barth, M., Todd, M., and Murakami, H., 2000. Intelligent

Transportation System Technology in a Shared

Electric Vehicle Program. Transportation Research

Record, (1731), 88–95.

Buyya, R., Yeo, C. S., Venugopal, S., Broberg, J., and

Brandic, I., 2009. Cloud computing and emerging IT

platforms: Vision, hype, and reality for delivering

computing as the 5th utility. Future Generation

Computer Systems [online], 25 (6), 599–616.

Available from: http://portal.acm.org/citation.cfm?

id=1528937.1529211.

Chu, S. and Majumdar, A., 2012. Opportunities and

challenges for a sustainable energy future. Nature

[online], 488 (7411), 294–303. Available from:

http://www.nature.com/doifinder/10.1038/nature11475.

Chynoweth, J., Chung, C. Y., Qiu, C., Chu, P., and Gadh,

R., 2014. Smart electric vehicle charging infrastructure

overview. 2014 IEEE PES Innovative Smart Grid

Technologies Conference, ISGT 2014.

Clement-Nyns, K., 2010. The impact of charging plug-in

hybrid electric vehicles on a residential distribution

grid. Power Systems, IEEE… [online], 25 (1), 371–

380. Available from: http://ieeexplore.ieee.org/xpls/

abs_all.jsp?arnumber=5356176.

Deilami, S., Masoum, A. S., Moses, P. S., and Masoum,

M. a S., 2011. Real-time coordination of plug-in

electric vehicle charging in smart grids to minimize

power losses and improve voltage profile. IEEE

Transactions on Smart Grid, 2 (3), 456–467.

Delucchi, M. A. and Lipman, T. E., 2001. An analysis of

the retail and lifecycle cost of battery-powered electric

vehicles. Transportation Research Part D: Transport

and Environment.

García-Villalobos, J., Zamora, I., San Martín, J. I.,

Asensio, F. J., and Aperribay, V., 2014. Plug-in

electric vehicles in electric distribution networks: A

review of smart charging approaches. Renewable and

Sustainable Energy Reviews [online], 38, 717–731.

Available from: http://dx.doi.org/10.1016/j.rser.

2014.07.040.

German Federal Government, 2009. National

Electromobility Development Plan [online]. Available

from: http://www.bmwi.de/Dateien/BMWi/PDF/

nationaler-entwicklungsplan-elektromobilitaet-der-

bundesregierung,property=pdf,bereich=bmwi,sprache

=de,rwb=true.pdf [Accessed 26 Oct 2015].

Han, S., Han, S., and Sezaki, K., 2010. Development of an

optimal vehicle-to-grid aggregator for frequency

regulation. IEEE Transactions on Smart Grid, 1 (1),

65–72.

Hu, J., You, S., Si, C., Lind, M., and Østergaard, J., 2013.

Optimization and control method for smart charging of

EVs facilitated by Fleet operator. Review and

classification, 10 (1), 383–397.

Hu, J., You, S., Si, C., Lind, M., and Østergaard, J., 2014.

Optimization and Control Methods for Smart

Charging of Electric Vehicles Facilitated by Fleet

Operator: Review and Classification. International

Energy-Management-as-a-Service: Mobility aware energy management for a shared electric vehicle ﬂeet

349

Journal of Distributed Energy Resources and Smart

Grids, 10 (1), 383–397.

Jansen, B., Binding, C., Sundström, O., and Gantenbein,

D., 2010. Architecture and Communication of an

Electric Vehicle Virtual Power Plant. Smart Grid

Communications (SmartGridComm), 2010 First IEEE

International Conference on, 149–154.

Klein, M., 2013. PVCast- Simple and flexible energy yield

forecasts for photovoltaic plants. [online]. Available

from: http://www.pvcast.de.

Koetter, F., 2015. Dynamische Einsatzoptimierung von

gemeinsam genutzten Elektrofahrzeugfotten [online].

Available from: http://www.shared-e-fleet.de

images/Dynamische_Einsatzoptimierung_von_gemein

sam_genutzten_Elektrofahrzeugflotten.pdf [Accessed

2 Jun 2015].

Lee, D. A., Cook, G., Ford, N. P., Freeland, R. L., Gilliam,

F. M., Hough, J. A., Bridge, G. G., and Scott, B. A.,

2005. Car-Sharing: Where and How It Succeeds.

Transit Cooperative Research Program Report 108.

Lopes, J. a P., Soares, F. J., and Almeida, P. M. R., 2010.

Integration of Electric Vehicles in the Electric Power

System. Proceedings of the IEEE, 99 (1).

Lutzenberger, M., Kuster, T., and Albayrak, S., 2014.

Integrating electric vehicles into smart grid

infrastructures a simulation-based approach that

became reality. In: Proceedings of the Winter

Simulation Conference 2014 [online]. IEEE, 1061–

1072. Available from: http://ieeexplore.ieee.org/

lpdocs/epic03/wrapper.htm?arnumber=7019965.

Mukherjee, J. C. and Gupta, A., 2014. A Review of

Charge Scheduling of Electric Vehicles in Smart Grid.

IEEE Systems Journal [online], 1–13. Available from:

http://ieeexplore.ieee.org/lpdocs/epic03/wrapper.htm?

arnumber=6919255.

Mültin, M., Allerding, F., and Schmeck, H., 2012.

Integration of electric vehicles in smart homes - An

ICT-based solution for V2G scenarios. 2012 IEEE

PES Innovative Smart Grid Technologies, ISGT 2012.

Open Charge Alliance, 2015. OCPP – Open Charge Point

Protocol [online]. Available from: http://www.ocpp.nl

[Accessed 9 Nov 2015].

Ostermann, J., Koetter, F., Renner, T., and Hudert, S.,

2014. Leveraging Electric Cross-Company Car Fleets

through Cloud Service Chains: The Shared E-Fleet

Architecture. In: 2014 Annual SRII Global Conference

[online]. San Jose, CA: IEEE, 290–297. Available

from: http://ieeexplore.ieee.org/lpdocs/epic03/wrapper

.htm?arnumber=6879697.

Plötz, P., Gnann, T., Kühn, A., and Wietschel, M., 2014.

Markthochlaufszenarien für Elektrofahrzeuge.

Fraunhofer ISI, 2013 (September 2013).

Raab, A. F., Ferdowsi, M., Karfopoulos, E., Unda, I. G.,

Skarvelis-Kazakos, S., Papadopoulos, P., Abbasi, E.,

Cipcigan, L. M., Jenkins, N., Hatziargyriou, N., and

Strunz, K., 2011. Virtual Power Plant Control

concepts with Electric Vehicles. 16th International

Conference on Intelligent System Applications to

Power Systems [online], 1–6. Available from:

http://ieeexplore.ieee.org/lpdocs/epic03/wrapper.htm ?

arnumber=6082214.

Rardin, R. L., 1997. Optimization in Operations Research.

New Jersey, US: Prentice Hall (Higher Education

Division, Pearson Education), 131–174.

Saker, N., Petit, M., and Vannier, J. C., 2011. Electric

vehicles charging scenarios associated to Direct Load

Control programs (DLC). NAPS 2011 - 43rd North

American Power Symposium, (Dlc).

Sundstr, O., 2010. Optimization Methods to Plan the

Charging of Electric Vehicle Fleets. Proceedings of

the International Conference on Control

Communication and Power Engineering [online], 28–

29. Available from: http://www.zurich.ibm.com/

pdf/csc/EDISON_ccpe_main.pdf.

Sundström, O. and Binding, C., 2012. Flexible charging

optimization for electric vehicles considering

distribution grid constraints. IEEE Transactions on

Smart Grid, 3 (1), 26–37.

The Apache Software Foundation, 2016. Commons Math:

The Apache Commons Mathematics Library [online].

Available from: http://commons.apache.org/proper/

commons-math/ [Accessed 11 Feb 2016].

Valogianni, K., Ketter, W., Collins, J., and Zhdanov, D.,

2014. Effective Management of Electric Vehicle

Storage using Smart Charging.

Vandoorn, T. L., Zwaenepoel, B., Kooning, J. D. M. De,

Meersman, B., and Vandevelde, L., 2011. Smart

microgrids and virtual power plants in a hierarchical

control structure. 2nd IEEE PES International

Conference and Exhibition on Innovative Smart Grid

Technologies [online], 1–7. Available from:

http://ieeexplore.ieee.org/lpdocs/epic03/wrapper.htm?

arnumber=6162830.

Zuccaro, L., Di Giorgio, A., Liberati, F., Canale, S.,

Lanna, A., Pallares, V. F., Blanco, A. M., Escobar, R.

U., Ratej, J., Mehle, B., and Krisper, U., 2014. Smart

vehicle to grid interface project: Electromobility

management system architecture and field test results.

In: 2014 IEEE International Electric Vehicle

Conference (IEVC) [online]. 1–7. Available from:

http://arxiv.org/ftp/arxiv/papers/1503/1503.01266.pdf.

SMARTGREENS 2016 - 5th International Conference on Smart Cities and Green ICT Systems

350