METHOD FOR DESIGNING A NETWORK OF ELECTRIC VEHICLE CHARGING STATIONS

Abstract

The invention relates to a method for sizing a network of charging stations for electric vehicles on a motorway network comprising a plurality of candidate configurations for the distribution of charging stations distributed across the motorway network, comprising: a- acquiring actual vehicle journeys made on the motorway network, b- defining a target fleet of vehicles assigned target journeys based on the actual journeys, c- a charging schedule, with a planning agent assigned to each vehicle, and a station agent for each station, said agents exchanging information iteratively in order to construct a charging schedule projection for all or part of the plurality of candidate configurations, d- calculating an overall cost function, representing vehicle waiting times in the charging schedule projection and infrastructure cost.e- a step of identifying a minimum of the overall cost function. Figure 1,

Description

Title of the invention: METHOD FOR SIZING A NETWORK OF ELECTRIC VEHICLE CHARGING STATIONS

[0001] The invention relates to a method for dimensioning a network of electric vehicle charging stations, the charging stations being distributed over a motorway network.

[0002] The logic of deploying charging stations is currently carried out using a statistical approach section by section for motorway networks.

[0003] We are currently seeing an increase in the number of electric vehicles in circulation, as well as the possibility of using rapid charging with rates of 50kW, 75kW, 150kW, 300kW, 350kW. Rapid charging is a significant need for long journeys where the distance traveled is greater than the range permitted by the battery capacity.

[0004] It is desirable to allow electric vehicle drivers to recharge their vehicle batteries without waiting or too long at charging stations. A fast charge typically takes 30 to 45 minutes.

[0005] The need to recharge the batteries of electric vehicles is particularly critical during long journeys on the motorway, where electric vehicles travel distances which are often greater than their range provided by 100% of the battery capacity.

[0006] To overcome peak travel times during holiday departure periods, current logic would require significantly increasing the number of terminals and the number of charging stations, in order to limit waiting times to access a charging station.

[0007] The inventors sought to propose an optimized solution to this need.

[0008] To this end, the present invention proposes a method for dimensioning, and where appropriate installing and operating, a network of charging stations for electric vehicles, on a motorway network, the motorway network comprising a plurality of possible candidate configurations for distributing charging stations distributed over the motorway network, each charging station comprising a candidate number of charging terminals depending on the candidate configuration concerned, the method comprising: a- a step of acquiring actual journeys of electric vehicles made on the motorway network, b- a step of defining at least a first target fleet of vehicles assigned target journeys based on the actual journeys, representative of a level of traffic on the network of charging stations, c- a step of planning the recharges, the method using a vehicle planning agent assigned to each vehicle, and a station agent for each recharge station, the vehicle planning agents and the station agents exchanging information iteratively in order to construct by recurrence a recharge schedule projection for the first target fleet of vehicles, and this for each of all or part of the plurality of candidate configurations, d- a step of calculating a global cost function, representative of a sum of vehicle waiting time in the recharge schedule projection and infrastructure cost, for the candidate configurations considered in the planning step, e- a step of identifying at least one minimum of the global cost function, with a view to determining a preferred candidate configuration that is the most relevant.

[0009] Thanks to the provisions promoted above, the network of charging stations can be sized appropriately in relation to particularly high numbers of people, for example when leaving for or returning from vacation.

[0010] The overall cost function FCG is minimized to optimize the ratio of service provided to an overall cost. From another point of view, the investment to be made is minimized while simultaneously guaranteeing a maximum waiting time tolerated at the stations for recharging.

[0011] In a candidate configuration, each charging station CSj of index j comprises Nj terminals, the method leads to determining the most relevant Nj, by increasing Nj for the existing MSC stations, and including by proposing new installations on areas not yet equipped.

[0012] In the context of the document, the station agent acts as a local manager whose function is to manage charging sequences on its terminals and the future planning of charging sequences.

[0013] Under the term "for each of all or part of the plurality of candidate configurations", it should be understood that steps c- and d- are carried out either for all of the possible candidate configurations if this set remains within reasonable proportions in terms of calculation time, or for a subset (therefore only a part by sampling) of the possible candidate configurations in the case where the systematic exploration of all of the possible candidate configurations represents prohibitive calculation time, for example several weeks.

[0014] We intuitively feel that if we increase the total number of terminals / yMSC ^j=ti then the waiting times will be low or reduced, but in return the financial investment is very substantial and the terminals will be unoccupied for most of the time, while their footprint will be very significant. Conversely, if we reduce the total number of terminals (ijrM) then the financial investment can remain reasonable as well as the space occupied by the terminals, but the waiting times risk being sufficiently long to displease many drivers and dissuade other potential drivers from converting to the electric car. The FCG formula based on a behavioral simulation embodies the mathematical formulation of this intuitive approach and in practice allows us to identify the best solution based on a series of input parameters which will be detailed later. As we will see later, the combinatorics to be studied makes it totally unrealistic to carry out the necessary calculations by hand.

[0015] According to one embodiment, an evolutionary metaheuristic optimization algorithm is used to search for the minimum of the overall cost function, and steps c- to e- are iterated a number NN of times with another subset of candidate configurations at each iteration.

[0016] As a result, we obtain a convergence of the algorithm in a number NN of iterations which remains modest with a reasonable number of simulated candidate configurations. This is to be compared to a hypothesis where we run the simulation for all possible candidate configurations, which represents a very important complete combinatorics.

[0017] According to one embodiment, the optimization algorithm is a gray wolf pack type algorithm.

[0018] The inventors found that this algorithm converged fairly quickly and that it presented a good fit with respect to the problem posed by candidate configurations of charging stations.

[0019] According to one embodiment, the method further comprises a step of deploying the candidate configuration on the motorway network. The result of the simulation and the result of the calculation of the optimization function lead to a practical application, in this case the installation of new charging stations on the motorway network. The high financial investment relating to these new terminal installations must be made with a supported technical motivation, and this is the aim of the dimensioning method that the inventors propose.

[0020] According to one embodiment, steps b- to e- are repeated with at least a second target fleet of vehicles, a third target fleet of vehicles, a fourth target fleet of vehicles, the different target fleets of vehicles being representative of different high traffic volumes.

[0021] Thus, the global cost function FCG is averaged with respective weights (identical or differentiated) to obtain an optimization valid for all types of large crowds.

[0022] The solution chosen can thus be an optimum which takes into account peak times such as departures on holiday, return from holiday and certain weekends accompanying public holidays or bank holidays.

[0023] According to one embodiment, the overall cost function FCG is expressed as follows for the first fleet: FCG = C ra (cc) +L^penalty kv (œ) P our une llollc d ' indicek - CFC representing the cost of installing, operating and maintaining charging stations, and penalty being a penalty representing a waiting time greater than a predetermined waiting time.

[0024] WTS is considered an allowable waiting time.

[0025] According to one embodiment, the overall cost function FCG is expressed as follows for a number NF of fleets: FCG = C Fcl (cc) + P™“ ensemblede NF HAS"" 8 - CFC representing the cost of installing, operating and maintaining charging stations, and penalty embodying a penalty representative of a waiting time WT greater than a predetermined waiting time WTS.

[0026] According to one embodiment, the penalty value is written as follows: penalty = K x max(WT - WTS,0), max being the conventional maximum function, K being a predetermined multiplier coefficient and WT being the longest waiting time on a complete journey of a vehicle.

[0027] According to one embodiment, CFCi is the addition of three terms C1, C2, C3, i.e. CFCi = C1 + C2 + C3, Cl representing a daily charge relating to a monthly rent to be paid as well as subscriptions to an electricity supplier and maintenance costs, C2 representing a cost of installing a new terminal in an already existing station, C3 representing a cost of installing a new charging station in a service area not yet equipped with a charging station.

[0028] The invention further relates to an installation of a network of charging stations for electric vehicles on a motorway network dimensioned according to the method as described previously.

[0029] The invention will be further detailed by the description of non-limiting embodiments, and on the basis of the appended figures illustrating variants of the invention, in which: [Fig.l] is a schematic representation of an example of a motorway network, [Fig.2] is a schematic representation of a fleet of electric vehicles, [Fig.3] is a schematic representation of a vehicle charging station electric, [Fig.4] represents an example of a flowchart showing the exchanges between vehicle planning agents and station agents, [Fig.5] represents a flowchart illustrating an example of construction of the global cost function, [Fig.6] represents an example of a flowchart illustrating the steps of the process, [Fig.7] represents an example of construction of a charging plan for a charging station.

[0030] In the various figures, the same references designate identical or similar elements. For reasons of clarity of the description, certain elements are not necessarily represented to scale.

[0031] Motorway network

[0032] A general diagram of a motorway network 100 according to a non-limiting example is now described with reference to [Fig. 1].

[0033] The topology of the motorway network of interest can be more or less complex. In a very simple case, the motorway network can be monolithic without branches or diversions or connections to another motorway.

[0034] In the illustrated example, the motorway network 100 comprises several branches, the end entry points have been represented, the intermediate entry and exit interchanges have not been represented.

[0035] On the left of the diagram, the motorway network has three end points A, B, C and on the right of the diagram the motorway network has three other end points F, G, H.

[0036] Three-armed intersections are found in the illustrated cases, although a four-armed cross intersection is also possible. In fact, the present invention applies to any motorway network topology, with or without tolls. As an illustration, the French motorway networks ASF, APRR, SANEF can be taken as examples, without these examples being limiting.

[0037] For the optimization study promoted here, one can consider either parts of the network only or a superset of networks, i.e. a collection of motorway networks.

[0038] Each branch of the network comprises traffic lanes in a first direction and traffic lanes in a second direction opposite to the first. The network extends over at least 100 km, in practice the network extends over several hundred kilometers.

[0039] Along the branches of the motorway network, service areas are provided, also sometimes called 'rest areas'. Service areas equipped with a charging station for electric vehicles are generically identified by 1. Most of these areas are doubled, that is to say that there is an area 1a accessible by the traffic lanes in the first direction of circulation and another separate area 1b, accessible by the traffic lanes in the second direction of circulation. Each charging station 1 comprises a plurality of charging terminals as will be seen in detail below. It should be noted that we speak of 'charging terminals', but we can just as well speak of 'charging terminals', there is no difference in the sense of the present invention between "charge" and "recharge".

[0040] Service areas equipped with charging stations are represented by a small solid black dot in [Fig. 1].

[0041] Furthermore, certain service areas equipped with charging stations are accessible from both sides of the motorway. These particular service areas with dual access are marked 12 in [Fig.l], and represented by a small rectangle straddling the line symbolizing the motorway lane.

[0042] Generally, service areas that are equipped with charging stations for electric vehicles are also equipped with gas pumps.

[0043] In addition, the motorway network 100 includes other service areas marked 10 which are not equipped with charging stations, but which could be equipped with them in the future. These areas are represented by a small white square in [Fig.l].

[0044] To give an idea, depending on each direction of traffic, the service areas 1 equipped with charging stations are spaced a few dozen kilometers apart.

[0045] Motorway traffic varies daily, weekly, seasonal, and annual. There are a few days a year when traffic is at its annual peak. This is the case, for example, on days when people go on or return from vacation. There are even so-called crossover days when traffic is heavy in all directions.

[0046] Busy days are critical from the point of view of charging electric vehicles. If the supply of charging services is insufficient, this will cause significant waiting times at charging stations to be able to access a free charging station.

[0047] A sizing of the charging network which would ensure that waiting times are avoided during the busiest days would represent a prohibitive cost and a very significant footprint.

[0048] Therefore, the aim pursued by the proposal made in this document is to dimension the infrastructure to the exact amount needed in a realistic way. Starting from an existing network, such as the one illustrated in [Fig.l], the question is how it should be developed to correctly accept high traffic in the future. Starting from a blank sheet of paper, the ideal optimized service offer can be defined, for example for a motorway network defined by a computer model or ma- thematic.

[0049] If we consider that the number of charging stations can be quite high, for example several dozen, the combination of possibilities for the number of terminals in each of the charging stations turns out to be gigantic.

[0050] If we add to this the fact that certain service areas can come into play, that is to say be the subject of a decision to install charging terminals to create a new charging station, then this increases the number of possibilities even further.

[0051] We identify by the variable cc a candidate configuration among the set of possible candidate configurations.

[0052] In a candidate configuration cc, with a number MSC of charging stations, each charging station SCj of index j comprises Nj terminals. The method leads to determining the most relevant number Nj, ​​for example by increasing Nj for some of the existing MSC stations, and including by proposing new installations on areas not yet equipped.

[0053] For the optimization approach, the set of possible candidate configurations cc is located in a multi-dimensional vector space, for example as many dimensions as the number of charging stations equipped with terminals, namely MSC.

[0054] For the optimization approach, we can give ourselves a constraint concerning the total number of charging stations TNj = j- ct c cst 'a geo distribution graph that will be the result. There may also be a maximum number Njmax of possible terminals, specific to each CSj charging station, due to the limitation of the available space or the limitation of the available electrical power.

[0055] Fleets

[0056] In order to determine the most relevant candidate cc configuration(s), fleets of electric vehicles are defined, such as the example illustrated in [Fig.2]. These are logical fleets and not physical fleets. The number of vehicles in the fleet is typically between 250 and 1000. Of course, the size of the fleet can be greater than 1000.

[0057] According to one option, the number of vehicles in the fleet is representative of a number of electric vehicles using the motorway network over a day and traveling more than 300 km on this network.

[0058] The number of vehicles included in the fleet can be made dependent on a projection on the growth of the electric vehicle fleet and the time horizon that one wants to cover, that is to say for example a 3-year projection, a 5-year projection, a 10-year projection. The growth of the electric vehicle fleet can be estimated by the sales of new electric vehicles at the time or by assumptions linked for example to tax aid for purchase present or announced for the future.

[0059] It is noted that certain plug-in hybrid vehicles may, where appropriate, be included in the fleets of electric vehicles considered in this document.

[0060] When handling several fleets, each is identified by an index k, the fleet concerned being noted FLk. A given fleet FLk contains a number of vehicles noted PFk. We note NF the number of fleets considered in the optimization study.

[0061] Vehicles are designated by an index v. For a fleet FLk, the index v ranges from 1 to PFk.

[0062] Each VEv vehicle has an identifier and intrinsic characteristics, such as the battery capacity, the average consumption per 100 km, the maximum acceptable power when recharging (respectively the columns named BatCap, Conso and Max Pow Recharge in the table in [Fig.2]).

[0063] Furthermore, in the logical fleet, auxiliary information is linked to the vehicle, in particular relating to the journey. For each vehicle in the fleet (one line of the table) an entry point on the motorway network, an entry timestamp and an exit point from the network are defined. Among the auxiliary information, we also find the battery charge level at the time of vehicle entry (named %departure charge in the table [Fig.2]).

[0064] Each vehicle is assigned a vehicle planning agent APV which is a behavioral model of the vehicle and its driver during a journey on the motorway network, in particular for the decision(s) on battery charging stops.

[0065] A threshold waiting time WTS is also defined, beyond which the driver is dissatisfied. WTS can be chosen between 5 and 20 minutes, for example around 10 minutes or 15 minutes.

[0066] For the journeys considered in a fleet, we base ourselves on real journeys of vehicles (electric or not) actually made on the motorway network. For example, we can collect the journeys and trips made the previous summer during the busiest day on the motorway network, all vehicles combined, e.g. electric, hybrid and thermal. However, we can exclude trucks and utility vehicles from the journeys collected.

[0067] According to another option, the journeys and trips made during the 3 or 5 known busiest days on the motorway network are collected.

[0068] The process of collecting actual journeys forms step a- of the method promoted in this document.

[0069] The process of defining fleets forms step b- of the method.

[0070] Charging stations

[0071] As illustrated in [Fig.3], each charging station 1,12 comprises terminals Bi charging stations, some of them are fast charging stations, their caliber can be 50kW, 75kW, 150kW, 300kW or 350kW. Other calibers are also possible.

[0072] We are talking here about rapid charging, that is to say using a direct current vector between the charging station and the vehicle.

[0073] For each station CSj, there is a number Nj of terminals satisfying the power criteria defined above. The lower power terminals are neglected.

[0074] We use the index i for the limits from 1 to Nj.

[0075] We use the index j for the CSj stations, j from 1 to MSC stations.

[0076] The case Nj = 0 corresponds to the case of a candidate service area which is not yet equipped. If we start from an existing motorway network, we can assume that the candidate configuration for a charging station already including Njl terminals only explores cases where the number of terminals is greater than the current number Njl.

[0077] Each charging station is characterized by its identifier idSC, its geographical position 'Geoloc', its road link, i.e. whether it is accessible from only one side of the highway, or from both sides of the highway. Furthermore, each charging station has a number Nj of fast charging terminals, i.e. a power greater than or equal to 50 kilowatts. It should be noted that this number of terminals Nj can change over time and is considered a variable within the meaning of the present invention. Instead of a threshold of 50 kilowatts to consider a station as fast charging, a threshold of 100 kilowatts can also be taken. Each charging station also indicates its current status in its communications to the requesting entities.

[0078] Each terminal Bi is characterized by its maximum deliverable power Pi. Each terminal Bi also indicates its operating status.

[0079] The sum of the powers of the terminals is The total power installed for a station, noted PM(j) is limited. This can be capped, the cap can be dependent on the current time of day. In practice, the sum of the deliverable powers of the terminals can be greater than the total power installed for the station and in this case the power of the terminals can be capped at a percentage of the maximum deliverable power.

[0080] In the other case, if ^=Njpj < PM(j), then there is no need for clipping.

[0081] We also note that the price charged per kWh of charging can be indicated by the terminal or by the charging station, this information can be made available remotely to third parties, in particular future users of the charging service.

[0082] It is noted that the APV vehicle planning agent can act according to the price at kWh. The price criterion can be taken into account at first order or second order behind travel time.

[0083] Each CSj charging station is assigned a station agent APS.

[0084] The APS station agent is a local manager in charge of managing sequences of recharge and the upcoming schedule of recharge sequences.

[0085] The APS station agent receives requests from the vehicle planning agents.

[0086] The APS station agent can return an updated waiting time table.

[0087] Note that the station agent may practice a first-come, first-served policy. According to a variant, the station agent may accept a short-term reservation logic, for example up to 15 or 20 minutes after the current time.

[0088] Simulation and Optimization

[0089] Step c- of the method promoted here comprises a recharging planning step. The APV vehicle planning agents and the APS station agents exchange information iteratively in order to recursively construct a recharging planning projection.

[0090] This process is described with reference to Figures 4 and 7.

[0091] For a fleet of vehicles, for example the first target fleet of vehicles FL1, we run a discrete event simulation based on time and the progress of states and transitions as a function of time.

[0092] This simulation is carried out for each of all or part of the plurality of candidate configurations cc, that is to say on a subset which can be in the maximum hypothesis the complete set of possible candidate configurations.

[0093] According to a systematic calculation hypothesis called 'brute force', all the possibilities of candidate configurations are scanned in a single pass. The calculation time is reasonable only for a small motorway network with a small number of charging stations and a limited total number of terminals.

[0094] According to another possibility, the most used in practice in view of the large combinatorics to be studied, the simulations are only carried out on a subset of the plurality of candidate configurations. More precisely, a subset is selected by a choice of samples. This selection can be made according to the recommendations of so-called metaheuristic algorithms.

[0095] After running the simulation engine on the first subset, a second subset of candidate configurations is defined and then the simulation engine is run again, and so on.

[0096] An evolutionary metaheuristic optimization algorithm is used to search for the minimum FCG, and steps c- to e- are iterated NN times with another subset of candidate configurations at each iteration.

[0097] Figures 4 and 7 illustrate the iterative planning process.

[0098] A generic iteration is described below. Each vehicle in the VEv fleet sends, via its APV planning agent, requests to the different charging stations along its route to find out their occupancy or their expected waiting time based on the vehicles already charging there. This step is called Stepl in [Fig.4].

[0099] Generally for Figure 4, the legend logic is as follows: WTj is the estimated waiting time at the charging station CSj, AT is an estimate of the arrival time stamp at the charging station CSj, and CT is an estimate of the charging time at the charging station.

[0100] The flowchart illustrated in [Fig.4] is produced for each vehicle in the fleet considered, one after the other, in any order or in a particular order linked to the entry timestamp or even linked to a priority criterion of one vehicle compared to another.

[0101] In response to this request, the station agent APS of each of the requested charging stations sends to the vehicle that requested it a table of current and short-term WTi waiting times. This step is called Step2 in [Fig.4].

[0102] Depending on the responses obtained, the vehicle planning agent APV calculates all possible schemes or hypotheses of the COV charging plan on its route for the vehicle in question. Among these possible schemes, it selects the one which gives the shortest total journey time noted This step is called Step3 in [Fig.4],

[0103] Then, the APV planning agent sends, to the charging stations which are involved in the Wy charging plan giving the shortest route, the planned charging data of the VEv vehicle.

[0104] To complete the process of a basic cycle, each of the charging stations having received the expected charging data of the VEv vehicle updates its waiting time table WT;. This step is called Step5 in [Fig.4].

[0105] This process is repeated individually for each of the VEv vehicles in the fleet from 1 to Nj.

[0106] Then, the simulation considers that time advances by one unit of time. We go from t-1 to t in [Fig.4].

[0107] The basic cycle described above is repeated again for each of the vehicles in the fleet. The iteration continues until the end of the day, or optionally until the end of the journey of the vehicle with the longest journey.

[0108] In [Fig.7], we are at time tz, after having carried out a certain number of simulation cycles, terminal Bi has seen the recharging of vehicle VEi then the recharging of vehicle VE5, which ends. Terminal B2 has seen the recharging of vehicle VE3 then the charging of vehicle VE4. Terminal B3 saw the charging of vehicle VE8 then the charging of vehicle VE7>then the charging of vehicle VE3. Terminal B4 saw the charging of vehicle VE2 then the charging of vehicle VE6. The dotted charging lines correspond to expected charging data for vehicles that are expected to arrive at the terminals in the near future compared to tz.

[0109] The current waiting time for the first terminal Bi is almost 0. The current waiting time for the first terminal B2 is almost WT2. The current waiting time for the first terminal B3 is almost WT3. The current waiting time for the first terminal B4 is almost WT4.

[0110] Note that this simulation was carried out for a given candidate configuration cc.

[0111] The simulation is different for another candidate configuration. More precisely, the cov charging plans calculated by the vehicle planning agents are different because the number of terminals Nj in the charging stations are different, and therefore the vehicle assignments to the terminals differ from the beginning.

[0112] According to the method presented, a step d- is provided for calculating a global cost function noted FCG, representative of a sum of vehicle waiting times in the projection of recharging planning and infrastructure cost, for each candidate configuration having been the subject of the simulation using the vehicle and station agents.

[0113] The cost function is expressed as follows:

[0114] FCO = Cra(cc) + ^penaltyk^ For a fleet of index k.

[0115] The cost function is expressed as follows for several fleets:

[0116] FCG = Cref(œ) + ^k^iPenaltykv(cc) for a set of NF fleets. In this example, it is a simple average. Alternatively, a weighted average with specific weights can be provided.

[0117] CFC represents the cost of installing, operating and maintaining charging stations.

[0118] More precisely, CFCi is the sum of several components, i.e. CFCi = C1 + C2 + C3. Cl represents a daily charge relating to a monthly rent to be paid as well as subscriptions to an electricity supplier and maintenance costs. C2 represents a cost of installing a new terminal in an existing station. C3 represents a cost of installing a new charging station in a service area not yet equipped with a charging station.

[0119] The term 'penalty' embodies a penalty representative of a waiting time WT greater than a predetermined waiting time WTS.

[0120] The penalty can be directly proportional to the waiting time, penalty = K x WT or proportional to the waiting time beyond the WTS threshold, therefore penalty = K x max(WT-WTS,0).

[0121] [Fig.5] illustrates the functions leading to the calculation of the overall cost function.

[0122] The simulation engine takes as input the aforementioned vehicle fleets FLk (here three fleets FL1, FL2, FL3) and a vector space of candidate configurations cc with the variables of number of terminals as already explained above. In the example illustrated in [Fig.5], we are interested in fast charging terminals from 100 kilowatts, but of course we can consider fast charging terminals from 50 kilowatts or another threshold to be defined.

[0123] The simulation shows the waiting time for each electric vehicle during its journey.

[0124] For example, if an electric vehicle has waited at least once more than WTS, then the penalty score is applied, otherwise no penalty is applied.

[0125] In [Fig.6] the sequence of steps of the proposed method is summarized. We note in particular the looping arrow after step e- which returns to step c-, this applies to recursive optimization algorithms with subsets of candidates, as explained above.

[0126] The method may further comprise a step of deploying or installing the candidate configuration on the motorway network, this step is called f-,

[0127] To give a simple numerical illustrative example, in a motorway network comprising ten charging stations, each having ten charging terminals, then the combinatorics studied and of the order of 1010, which is already considerable.

[0128] Evolutionary algorithms are algorithms whose principle is inspired by living nature to solve a problem of a multidimensional function to be optimized. These are therefore bio-inspired calculation methods where the configuration subset considered is evolved with regard to biomimetic behavior, for example by drawing inspiration from ants, bees, and wolf packs.

[0129] Concerning the gray wolf pack type algorithm, the reader can consult the following article: S. Mirjalili, SM Mirjalili, A. Lewis, “Grey Wolf Optimizer, Advances in Engineering Software”, vol. 69, pp. 46-55, 2014, DOI: http: / / dx.doi.org / 10.1016 / j.advengsoft.2013.12.007.

Claims

Claims

1. Method for dimensioning a network of charging stations for electric vehicles (EV), on a motorway network (100), the motorway network comprising a plurality of possible candidate configurations for distributing charging stations (1, 12) distributed over the motorway network, each charging station comprising a candidate number (Nj) of charging terminals (Bi) depending on the candidate configuration concerned, the method comprising: a- a step of acquiring actual journeys of electric vehicles made on the motorway network, b- a step of defining at least a first target fleet of vehicles assigned target journeys based on the actual journeys, representative of a level of traffic on the network of charging stations, c- a step of planning the recharges, the method using a vehicle planning agent (VPA) assigned to each vehicle, and a station agent (SPA) for each charging station,the vehicle planning agents (VPAs) and the station agents exchanging information iteratively in order to construct by recurrence a projection of the recharging schedule for the first target fleet of vehicles, and this for each of all or part of the plurality of candidate configurations (cc), d- a step of calculating a global cost function, representative of a sum of vehicle waiting time in the projection of the recharging schedule and of infrastructure cost for the candidate configurations considered in the planning step (c-), e- a step of identifying at least one minimum of the global cost function, with a view to determining a most relevant preferred candidate configuration.,

2. Method according to claim 1, characterized in that an evolutionary metaheuristic optimization algorithm is used to search for the minimum of the global cost function (FCG), and steps c- to e-, i.e. from the step of planning recharges to the step of identifying at least one minimum of the global cost function, are iterated a number NN of times with another subset of candidate configurations at each iteration.

3. Method according to claim 2, characterized in that the algorithm optimization is a gray wolf pack type algorithm.

4. Method according to any one of claims 1 to 3, further comprising: f- a step of deploying the candidate configuration on the motorway network.

5. Method according to any one of claims 1 to 4, characterized in that steps b- to e-, i.e. from the step of defining at least a first target fleet of vehicles to the step of identifying at least one minimum of the overall cost function, are repeated with at least a second target fleet of vehicles, a third target fleet of vehicles, a fourth target fleet of vehicles, the different target fleets of vehicles being representative of different high traffic volumes.

6. Method according to any one of claims 1 to 4, in which the global cost function FCG is expressed as follows for the first fleet: cFCAcc) penalty k^cc) Cparepresenting the cost of installation, operation and maintenance of the charging stations, penalty being a penalty representative of a waiting time greater than a predetermined waiting time.

7. Method according to any one of claims 1 to 5, in which the overall cost function FCG is expressed as follows for a number NF of fleets Cpcirepresenting the cost of installation, operation and maintenance of the charging stations, penalty being a penalty representative of a waiting time greater than a predetermined waiting time.

8. Method according to claim 6 or claim 7, characterized in that the penalty value is written as follows: penalty = K x max((WT -WTS),0), max being the conventional maximum function, K being a predetermined multiplier coefficient and WT being the longest waiting time on a complete journey of a vehicle.

9. Method according to claim 6 or claim 7, characterized in that the term CFCi is the addition of three terms C1, C2, C3, i.e. CpCi = C1 + C2 + C3, Cl representing a daily charge relating to a rent monthly to be paid as well as subscriptions to an electricity supplier, C2 representing a cost of installing a new terminal in an existing station, C3 representing a cost of installing a new charging station in a service area not yet equipped with a charging station.

10. Installation of a network of charging stations for electric vehicles (EV) on a motorway network (100) characterized in that the installation is dimensioned according to the method according to one of claims 1 to 9.

Images