An App-based Algorithmic Approach for Harvesting Local and
Renewable Energy using Electric Vehicles
Antoine Dubois, Antoine Wehenkel, Raphael Fonteneau, Fr
eric Olivier and Damien Ernst
Department of Electrical Engineering and Computer Science, University of Li
ee de la D
ecouverte, 10, 4000 Li
ege, Belgium
Multi-agent System, Electric Vehicles, Renewable Energy.
The emergence of electric vehicles (EVs), combined with the rise of renewable energy production capacities,
will strongly impact the way electricity is produced, distributed and consumed in the very near future. This
position paper focuses on the problem of optimizing charging strategies for a fleet of EVs in the context where
a significant amount of electricity is generated by (distributed) renewable energy. It exposes how a mobile
application may offer an efficient solution for addressing this problem. This app can play two main roles.
Firstly, it would incite and help people to play a more active role in the energy sector by allowing photovoltaic
(PV) panel owners to sell their electrical production directly to consumers, here the EVs’ agents. Secondly, it
would help distribution system operators (DSOs) or transmission system operators (TSOs) to modulate more
efficiently the load by allowing them to influence EV charging behaviour in real time. Finally, the present
paper advocates for the introduction of a two-sided market-type model between EV drivers and electricity
The past decade has seen a steep rise in the devel-
opment of renewable energy production capacities,
mainly driven by the willingness to (i) reduce pollu-
tion and greenhouse gas emissions and (ii) limit the
dependency on fossil fuels. In addition, electric ve-
hicles (EVs) are now emerging rapidly. The conjunc-
tion of these factors may be an opportunity to set the
basis of a joint optimization approach for charging
EVs with regard to the fluctuation of renewable en-
ergy production. This has already been the topic of
academic research in the past decade (see (Palensky
et al., 2013) for a comprehensive review). In partic-
ular, it has been shown that EVs may be efficiently
used in order to balance the load within distribution
networks (Caramanis and Foster, 2009). Also, vehi-
cle charging can be influenced by dynamically adapt-
ing tariffs as shown in (O’Connell et al., 2011). Even
though these results are of significant importance,
none of them propose a convenient way to apply them
to real situations.
In this paper, we focus on the problem of optimiz-
ing the charge of a fleet of EVs in a context where it is
important to match, at best, local energy consumption
and renewable electricity production. We argue that a
mobile application would be an efficient and elegant
support to deploy algorithms to dispatch EVs to elec-
tricity sources. In Section 2, we first explain how this
app should be designed. In Section 3, we describe a
few problems that may be efficiently addressed by ad-
hoc algorithms integrated into the app. Section 4 lists
a few generic algorithmic techniques that could be
used for addressing decision-making problems within
the app. Finally, Section 5 concludes.
The combination of (mobile) EVs and (static) elec-
tricity sources is seen as a multi-agent system, where
mobile agents (the EVs) should gather, at best, re-
newable energy under travel distances and electricity
production fluctuations constraints. On the one hand,
electricity sources are static, but electricity production
is fluctuating, depending on solar irradiation or wind
speed. On the other hand, EVs are mobile, but are
subject to travel constraints.
One main objective is to allocate electricity
sources to EVs based on geographical parameters
(starting point, destination, etc.). A second aspect
Dubois A., Wehenkel A., Fonteneau R., Olivier F. and Ernst D.
An App-based Algorithmic Approach for Harvesting Local and Renewable Energy using Electric Vehicles.
DOI: 10.5220/0006250803220327
In Proceedings of the 9th International Conference on Agents and Artificial Intelligence (ICAART 2017), pages 322-327
ISBN: 978-989-758-219-6
2017 by SCITEPRESS Science and Technology Publications, Lda. All rights reserved
is to take into account the fact that several EVs may
compete for the same electricity source. So, it ap-
pears that the fleet of should be coordinated in order
to attain a consensus so that the amount of electricity
harvested by the whole fleet of EVs is maximized. In
addition to this, one should take into account the fact
that a specific subclass of sources of electricity may
be used to rapidly charge EVs, thus offering the op-
portunity to make a stop when travelling, or even to
take a small detour, in order to get to a fast-charging
The app should be an efficient tool for reach-
ing such a consensus among the static and dynamic
agents. To this end, we suggest building an app in the
form of a two-sided market platform.
2.1 Two-sided Market
One main role of the app is to create a convenient way
for PV panel owners to share (and sell) their electric-
ity to EV drivers. This is particularly important, since
the electrical network has not been designed to absorb
electricity production from private dwellings. For ex-
ample, during solar production peaks, PV panels may
have to be disconnected because of overvoltages on
the network (see e.g., (Olivier et al., 2016)). Neither
has the electric network been designed to supply a
sufficient amount of power sufficient to fully recharge
a large fleet of EVs in a few hours within a neighbour-
Consequently, on one side we have PV panels
owners that may want to find a way to profitably ex-
ploit their installation, and on the other side we have
EV drivers who may be constrained to charge their
vehicles while they are occupied by other tasks. The
app is aimed at satisfying both requirements and to
make it easy for each sides’ to benefit from the other
2.2 Driver Services
To be successful, the app must be attractive to EV
drivers. For this purpose, a booking service will be
provided to guarantee a charge level for the drivers.
Moreover, the app intelligence should be able to pro-
pose a flexible and suitable choice of charging points
depending on the user needs.
One way for the app to output convenient charging
point suggestions would be to specify journeys using
a set of geographical points. The simplest case would
be to specify the starting point (A) and the destina-
tion (B). In this configuration, the app would compute
the optimal stations at which the driver should stop as
they travel from A to B, knowing the level of charge,
consumption and battery autonomy of the vehicle. A
more-advanced case would be to specify three geo-
graphical points: the starting point (A), a stop (B)
and a final destination (C), as well as the time dura-
tion spent at location B, where it may be possible to
recharge the EV. Taking all these parameters into ac-
count, plus eventually additional constraints that the
drivers may have, it would suggest the best stations
for drivers where they should stop during the jour-
ney. Obviously, suggestions should also be optimized
in order to minimize energy costs and/or curtailment.
As the app interacts with a fleet of EVs and not just
one single EV, it may also be interesting to optimize
charging station suggestions globally rather than on a
single EV basis. This would lead to a better solution
at the level of an EV fleet but may penalize some EV
owners due to a possible lack of fairness of the global
solution. This fairness issue could be addressed by
implementing a compensation mechanism. It could
consist in monetary terms, but also in other advan-
tages like free charge or booking priority.
2.3 Producer Services
Producers can be divided into two categories: compa-
nies and individuals. Companies can either be large
renewable energy source owners, DSOs or substantial
charging-point owners, whereas the individuals con-
sist mainly of private owners of PV units, with a nom-
inal power less than 10 kWp, and who also want to
sell the electricity they produce.
We may reasonably assume that, in many cases,
EV users may want to book a charging station,
through the app, that has a significant charging power
(e.g., higher than 20 kW). This may penalize individ-
ual charging stations, which, if they want to sell their
green electricity, are limited to the power produced
by their PV installation. However, we can reason-
ably assume that in the coming years, with the rise
of batteries, these individual producers should be able
to store their excess of green electricity. In such a
context, during certain periods of the day, they could
offer a charging power close to the power of the PV
installation plus the power that their batteries can de-
liver. Indeed, with a 10 kWp PV installation, an indi-
vidual producer will not be able to sell many green
charges per week due to the rather limited amount
of energy that such an installation can produce. In-
deed, if we assume a load factor of 8.9% for the PV
installation, the average European load factor for PV
(Energy - Yearly statistics 2008 (Eurostat)), the 10
kWp PV installation would produce on a daily ba-
sis 10 × 0.089 × 24 = 21.36 kWh. So, at most, only
once every three days, the individual producer will be
An App-based Algorithmic Approach for Harvesting Local and Renewable Energy using Electric Vehicles
able to sell a full fast charge to an EV with a 60 kWh
battery. This illustrates why individual charging may
have a low level of availability.
The app could offer many services to the owners
of the charging stations. It would include a booking
system, but also other services based, for example, on
data to manage prices and attractiveness of the charg-
ing points. More specifically, the app could allow
for the efficient retrieval of data about those stations
(such as price, power, etc.). From these data, it would
be able to carry out an analysis in order to establish
how pricing can influence the attractiveness of charg-
ing points. Based on such an analysis, the app could
also help the producers to better manage their charg-
ing stations, by having, for example, dynamic pricing
to adapt demand to production. It could also help in-
dividual users to manage their batteries and the flex-
ible loads that they may have at home (e.g., washing
machine, heat pump, etc.).
In addition to creating a two-sided market, a mo-
bile application could also provide a set of services
for addressing underlying problems associated with
EV charging and network management. We believe
that the opportunity to solve these problems via the
app may be a game changer in the electricity sector,
mainly due to the app’s ability to enable producers
and consumers react quickly and in a very flexible
3.1 Booking Service
In the short-term, congestion problems are going to
occur close to well-located charging points as a full
recharge typically takes at least 30 minutes with su-
perchargers, and more generally many hours with
classic chargers. As a result, we can expect that know-
ing in advance whether a charging station will be
available or not may probably become one of the main
problems for EV drivers. For now, no suitable solu-
tion to address this problem has emerged, although
most EV drivers have already experienced this incon-
The app’s booking service would be a solution to
this problem, and it should work as follow: Firstly,
stations will only be bookable a short period in ad-
vance (say, for example, 24 hours), and this for sev-
eral reasons. For example, if someone could book a
station for an entire year, this would lead to a lack
of attractiveness of the app for new users. As an-
other example: the app intelligence will be respon-
sible for organizing a fair distribution of charging sta-
tions among users to avoid the same type of problem.
Indeed, we want to avoid situations where far-sighted
users would always book the most attractive stations,
at the expense of other users. Actually, two main rea-
sons motivate this proposal: on the one hand, the app
must remain attractive by providing each type of user
with a good suggestion, while on the other hand, one
of the main goals of the app is to retain and maintain
as much flexibility as possible in the demand for elec-
We propose two solutions to maintain flexibility of
the demand while allowing charging station booking:
The first solution is to adapt electricity pricing.
By booking a charging station suggested by the
app just prior to stopping to charge their EV, the
driver is more likely to offer a valuable flexibil-
ity service to the network (provided that the app
is smart enough) that should be rewarded, for in-
stance, with a reduction in the cost of the charge.
On the other side of the coin, the early booking
driver puts advance constraints in advance on the
network that may not match renewables produc-
tion (corresponding to the situation where the bid
is too early compared, for example, to the the time
horizon at which it is possible to get accurate pre-
dictions of renewable production). Thus such a
user would have to pay an additional fee for an
early booking. It is important to note that the fron-
tier between early and late booking is not clearly
defined a priori. However, one could imagine de-
signing learning algorithms that are able to clas-
sify users’ behaviour according to the service they
are actually offering to the network.
In the second solution, the booking transactions
could be stored in blocks, which would only
be accepted after a certain time, thus allowing
more flexibility in terms of charging station al-
location. This solution could be based on dis-
tributed ledgers (such as blockchain technology
(Nakamoto, 2008)) associated with smart con-
which could be used to determine the con-
Smart contracts (also called self-executing contracts,
blockchain contracts, or digital contracts) are simply com-
puter programs that act as agreements where the terms of
the agreement can be coded in advance with the ability to
self-execute and self-enforce themselves. This code defines
the rules and consequences in the same way that a tradi-
tional legal document would, stating the obligations, bene-
fits and penalties which may be due to either party in various
different circumstances. This code can then be automati-
cally executed by a distributed ledger system.
ICAART 2017 - 9th International Conference on Agents and Artificial Intelligence
ditions of the transactions without passing by a
third party, so adding more dynamism and auton-
omy to the app.
Finally, a problem which rapidly comes to mind
when proposing the idea of a global booking service
is the difficulty to control whether bookings are re-
spected. Indeed, unless you own every single station
that you propose to drivers, you cannot guarantee that
a station booked by one driver will not be occupied by
another driver that is not using the app. The solution
we propose amounts to creating a global booking ser-
vice, which would not be part of the one single app,
but should also be available for other app creators. It
would retain all the booking data about each station.
Station owners would just have to rely on this tool to
manage the booking around their stations. The tool
could probably have some specific features and pro-
pose some additional services but this is not the sub-
ject we want to explore here.
3.2 Demand Management and Dynamic
Managing the demand is probably one of the best as-
sets of the app. By maintaining a certain degree of
flexibility in booking, the app allows to better cor-
relate local consumption with production by making
EVs charge close to energy sources. As shown in
(O’Connell et al., 2011), a dynamic pricing system
could allow the app to boost the demand in order to
follow, more closely, the production. Moreover, with
smart chargers (i.e. chargers that may automatically
change the amount of power they output) and accu-
rate weather forecasting, we can hope to fit, exactly,
the (over)production to the EV charging by allocating
each vehicle to an appropriate station for a precise du-
ration of time.
To allow for such a load management, a conve-
nient dynamic pricing system must be implemented.
It therefore has to allow producers to keep a hand
on the demand simply by raising or lowering charg-
ing prices. For instance, they could lower prices at
overproduction times to increase demand and do the
opposite when domestic consumption is peaking, or
when the level of renewable energy production is low.
This will all be done through the app that should help
charging point owners to understand the influence of
these pricing choices. For instance, the app could pro-
vide estimations of the demand as a function of the
price chosen for one charging point.
3.3 Data Mining
Another strength of the app is the flow of data passing
through it and that can be stored. These data con-
sist mainly of information about drivers trips (travel
time, distance, ...) or the performance of EVs (mean
consumption, charge level variations, ...). More gen-
erally, we can summarize this information as when,
where, and at which frequency EV drivers charge
their cars and how it influences the vehicles’ perfor-
mances, the vehicles’ fleet dynamics, and the network
load. This information may be of interest to a num-
ber of actors in the automotive and energy industries.
Here is a short list of such interested actors, as well as
of their potential use of the app’s data:
Drivers: The drivers themselves would be the
first to benefit from these data through a self-
learning improvement of the app’s dispatching
strategy based on previous results.
Car manufacturers: More practically, EV
real-time performances could help manufacturers
identify key points for future improvements.
Producers: Forecast of the density of traffic at
different locations, at different times could help
producers apply an appropriate dynamic pricing
Electricity Networks: These same forecasts,
combined to renewable energy production pre-
dictions, can be used by TSOs/DSOs to find
methods to balance the load and prevent over-
loads/overvoltages. Dynamic pricing can be used
in two different schemes: (i) it can influence the
time at which EV users decide to charge their cars,
or (ii) it can be used to orientate the EV users to
the right charging station where they would not
cause any problems, or even be beneficial to the
3.4 Shared Economy Model
EV drivers can already use a panel of apps help-
ing them to find information about charging sta-
tions that they could possibly use (PlugShare
, NextCharge
). However, all of
them work with a one-view driver-oriented system.
An advantage of our app would be its multi-view or-
ganization aiming to be profitable for every actors in-
volved. This approach, supported by a shared econ-
omy model, would benefit drivers by optimizing their
travel and charging time, and also the producers by al-
lowing them to implement dynamic pricing methods
to modulate the attractiveness of their products.
An App-based Algorithmic Approach for Harvesting Local and Renewable Energy using Electric Vehicles
In addition, shared economy models are becoming
more and more popular and are reaching every branch
of capitalism, as stated in (Rifkin, 2014). The success
of Uber
or Airbnb
prove that people are eager
for a change of economic model. Adequacy of supply
and demand, efficiency, affordability and scalability
are all different reasons why these models are so suc-
cessful and are all properties that would be beneficial
for an EV driver app. As a matter of fact, it would
seem more convenient nowadays to directly book a
station to its owner than having to solicit the interven-
tion of an external organization.
3.5 Reactivity
To work properly, all the solutions proposed have to
be extremely reactive. Nowadays, as the number of
EVs and charging stations increases, there is a grow-
ing need for a real-time information platform. Mo-
bile applications currently offer one of the most reac-
tive platforms and a true answer to this kind of prob-
lems. In particular, the booking can be easily and dy-
namically managed by connecting together a series of
drivers’ and producers’s phones. More than connect-
ing people, it would also offer a direct connection to
smart electric cars, instantly collecting immediately
all the data needed to predict when and where the next
charge should be done. Moreover, the app would open
the door to more connected payment methods like vir-
tual money or phone payment which would add some
more fluidity to the system.
Finally, a mobile application adds a layer of re-
activity over all these features due to its proximity
to people. More than 2 billion people have a smart-
. Therefore, apps are always with us allow-
ing them to inform us of any important update like
price modifications, overproduction risks or reserva-
tion problems directly when they happen so that we
can react in a minimum amount of time.
As showed in previous sections, a mobile application
could work as a good support to implement efficient
EV charging, while favouring the integration of re-
newables into electricity networks. Nevertheless, as
pointed out previously, a series of algorithms is neces-
sary to achieve this implementation. In the following,
we describe a few possibilities for designing them.
According to the website
4.1 Discrete Optimization and Machine
Dispatching EVs to a set of charging points while
minimizing certain constraints for the user (e.g. costs,
detour time) and/or for the network (e.g. amount of
energy curtailed) can be modelled as a discrete op-
timization problem. A possible solution is therefore
to use graph theory and algorithms, such as the mini-
mum spanning tree-type algorithms, where the nodes
represent either a charging point or a driver destina-
tion, while the weights of the edges are equivalent to
the constraints of the problem. Meta-heuristics, such
as genetic algorithms, have also proven to be useful
in the resolution of this kind of problems.
Pricing decisions is another important aspect of
the problem these should take into account not only
the variability of renewable energy but also the be-
haviour of the different actors. This could typically
be solved with some machine learning algorithms.
4.2 Multi-agent Systems
A last generic domain that is also closely related to
our problematic is, of course, the one of multi-agents
systems. Multi-agent technology has already been
used in power networks, leading, for example, to
ways to reduce imbalances in distribution networks
(Kok et al., 2008) or to manage congestions caused
by electric vehicles in a distribution grid (Hu et al.,
2015). Matching this technology with our mobile ap-
plication could probably enhance its performances.
This article shows how a mobile application (app)
may be an appropriate support for tackling several
upcoming obstacles associated with the rise of EVs
and renewable energy production capacities, com-
bined with the management of the electrical network
load. This app may be seen as an interface between
drivers and producers, providing services to both of
them. This paper also lists a series of algorithms that
could be used in this app to solve several decision-
making issues related to this app.
The main purpose of this paper is to make the
public aware that the upcoming issues regarding EV
charging and network management could be solved
by using an app-based strategy. Several research di-
rections should be taken for accelerating the develop-
ment of such an app. They should encompass algo-
rithmic research for tackling all the decision-making
ICAART 2017 - 9th International Conference on Agents and Artificial Intelligence
issues related to this app. They should also relate to
all the more “practical aspects” to be put in place for
such an app to be successful (e.g., design of the right
user interface, data management issues, etc.).
This research was carried out thanks to the support
of ENGIE - Electrabel. We also thank Powerdale for
many valuable discussions.
Caramanis, M. and Foster, J. M. (2009). Management of
electric vehicle charging to mitigate renewable gen-
eration intermittency and distribution network con-
gestion. In Decision and Control, 2009 held jointly
with the 2009 28th Chinese Control Conference.
CDC/CCC 2009. Proceedings of the 48th IEEE Con-
ference on, pages 4717–4722. IEEE.
Hu, J., Saleem, A., You, S., Nordstr
om, L., Lind, M., and
Østergaard, J. (2015). A multi-agent system for distri-
bution grid congestion management with electric ve-
hicles. Engineering Applications of Artificial Intelli-
gence, 38:45–58.
Kok, K., Derzsi, Z., Gordijn, J., Hommelberg, M., Warmer,
C., Kamphuis, R., and Akkermans, H. (2008). Agent-
based electricity balancing with distributed energy re-
sources, a multiperspective case study. In Hawaii in-
ternational conference on system sciences, proceed-
ings of the 41st annual, pages 173–173. IEEE.
Nakamoto, S. (2008). Bitcoin: A peer-to-peer electronic
cash system.
O’Connell, N., Wu, Q., Østergaard, J., Nielsen, A. H., Cha,
S. T., and Ding, Y. (2011). Electric vehicle (ev) charg-
ing management with dynamic distribution system tar-
iff. In Innovative Smart Grid Technologies (ISGT Eu-
rope), 2011 2nd IEEE PES International Conference
and Exhibition on, pages 1–7. IEEE.
Olivier, F., Aristidou, P., Ernst, D., and Van Cutsem, T.
(2016). Active management of low-voltage networks
for mitigating overvoltages due to photovoltaic units.
IEEE Transactions on Smart Grid, 7(2):926–936.
Palensky, P., Widl, E., Stifter, M., and Elsheikh, A. (2013).
Modeling intelligent energy systems: Co-simulation
platform for validating flexible-demand ev charging
management. IEEE Transactions on Smart Grid,
Rifkin, J. (2014). Capitalism is making way for the age of
free. The Guardian, 31.
An App-based Algorithmic Approach for Harvesting Local and Renewable Energy using Electric Vehicles