METHOD FOR OPTIMALLY MANAGING THE CHARGING OF A FLEET OF ELECTRIC VEHICLES

Abstract

A method is performed for optimizing the management of the charge of a set of electric batteries, where each battery is recharged during a time interval during which it is connected to an electricity distribution network, according to a charge profile provided by a charge aggregator, by applying, under the control of a charge manager of the battery, a charging power level. The method involves determining, on the aggregator side, as a charge profile, a distribution curve of the charge per day over a given period of time, defined for all the batteries solely according to constraints specific to the network, transmitting the charge distribution curve to each charge manager and, on the charge manager side, adapting the received charge distribution curve to the time interval during which the battery is connected to the network and to the associated charging power level.

Claims

1. A charging method for optimally managing charging of at least one set of electric batteries for supplying power to electrical consumers, in which each of the batteries is recharged for at least one time interval during which it is connected to an electricity distribution grid, according to a charging profile supplied by a charging aggregator, by applying, under a control of a charging manager associated with each of the batteries, a charging power level associated with the time interval, the charging method comprising: on aggregator side, determining the charging profile to be a curve of distribution of charging per day over a given time period, defined for all the batteries solely depending on constraints specific to the electricity distribution grid comprising at least the power available on the electricity distribution grid; transmitting the charge distribution curve from the aggregator to each charging manager, and the charge distribution curve received on charging manager side is adapted to the at least one time interval during which the battery is connected to the grid and to the associated charging power level; and determining for all of the batteries a coefficient representative of a percentage to be reached per day with respect to a flexibility range permitted by the charging manager, which range is defined depending on the nominal capacity of the battery and on a minimum state of charge desired per day, and computing a daily target state of charge to be reached depending on the coefficient and on the minimum state of charge desired per day.

2. The charging method as claimed in claim 1, further comprising controlling charging and discharging moments of each of the batteries through the electricity distribution grid by monitoring the charge distribution curve.

3. The charging method as claimed in claim 1, wherein the determining of the charging profile depends on CO2 emissions associated with the generation of electricity on the electricity distribution grid over the given period.

4. The charging method as claimed in claim 3, further comprising receiving a prediction model for CO2 emissions from which model an average of the CO2 emissions over the given period and an average of the CO2 emissions per day are deduced and the level of the CO2 emissions at a time is compared to the average of the CO2 emissions per day, so that the charging profile is adapted to command charging only at moments when the level of CO2 emissions is below the average per day.

5. The charging method as claimed in claim 1, wherein the charge distribution curve is calculated in a percentage of charge with respect to time for a given sampling increment.

6. The charging method as claimed in claim 1, wherein the given time period corresponds to a time period of one week.

7. The charging method as claimed in claim 1, wherein the charging power level during each time interval is determined based on a comparison between the daily target state of charge to be reached and a current state of charge of the battery.

8. The charging method as claimed in claim 1, wherein the percentage to be achieved per day with respect to the flexibility range authorized by the charge manager varies inversely with the level of the CO2 emissions associated with the generation of electricity on the electricity distribution grid over the given period.

9. The charging method as claimed in claim 1, wherein the set of batteries consists of traction batteries of a fleet of electric or rechargeable-hybrid vehicles.

Description

BRIEF DESCRIPTION OF THE DRAWINGS

[0025] FIG. 1 illustrates a system for optimally managing charging of electric vehicles according to the present invention;

[0026] FIG. 2 shows a graph illustrating a charging law implemented vehicle-end, the law being obtained by employing the method for managing charging according to the present invention;

[0027] FIG. 3 shows a graph illustrating an example of variation over a week in a daily target state of charge to be reached, as a function of a percentage to be reached of a permitted flexibility range taking into account the variation in the CO2 emissions in the generation of electricity on the grid; and

[0028] FIG. 4 shows a graph illustrating an example of an ideal charging profile to be followed by all of the vehicles in order to reach the daily target state of charge.

DETAILED DESCRIPTION OF EMBODIMENTS

[0029] FIG. 1 schematically illustrates a set of electric vehicles VE each equipped with a traction battery and with an associated charging manager that is suitable for being connected to an electricity distribution grid 20. The charging of this set of electric vehicles VE is controlled by an aggregator 10, which is connected to the electricity distribution grid 20, to which the batteries are connected in a phase of charging thereof. The aggregator 10 is for example a dedicated server of a motor-vehicle manufacturer programmed to be able to communicate remotely with the vehicles of the manufacturer, for example via an over-the-air (OTA) communication. Thus, each of the vehicles VE possesses a communication module, for example a GSM module, allowing a communication, via a communication network, with the server 10 that forms the aggregator.

[0030] The aggregator is also designed to take into account the state of the electricity distribution grid 20, and in particular, as will be seen in more detail below, moments at which the power available on the grid is at least partially low-carbon, i.e. at which the generated power available on the grid is at least partially from low-carbon power sources, wind turbines or solar power plants for example.

[0031] The method of the invention is firstly based on determination of an ideal profile of charging over the day R.sub.ideal. This charging profile R.sub.ideal is computed by the aggregator 10 for all the vehicles VE controlled thereby and is defined as being the distribution of charging over each day, over a given time period, depending on constraints specific to the electricity distribution grid. This distribution of charging per day R.sub.ideal is computed in percentage of charge with respect to time for a given sampling increment, for example every quarter of an hour, for a given period of several days, one week for example.

[0032] The percentages of charge per given sampling increment may either be positive, meaning that the corresponding moment is propitious for the grid to deliver power to charge the battery, or negative, meaning that the corresponding moment is instead propitious to returning the energy stored in the battery to the electricity distribution grid, if the battery of the vehicle is connected to a reversible charging device.

[0033] FIG. 2 illustrates an example of a curve of distribution of charging per day thus computed for all of the vehicles, which corresponds to the optimal charging profile over the day in percentage per quarter of an hour R.sub.ideal, such as the profile is sent by the aggregator 10 to the vehicles VE.

[0034] The charging managers associated with the batteries of the various electric vehicles VE comprise software means for executing an algorithm for adapting this curve R.sub.ideal to the specific needs of the user of the vehicle and, in particular, to take into account moments at which the vehicle is connected to the network and the connection power available to it for charging. In other words, the charging orders intended for the batteries are not sent by the aggregator 10 but are generated locally by the charging manager associated with the battery, depending on the aggregated charge distribution curve as a function of time, which meets solely the needs of the electricity distribution grid, the curve being adapted, by each charging manager, depending on the moments of connection of the battery and on the power level of the charging. In yet other words, at the side of the aggregator 10, i.e. aggregator side, the needs of the users of the electric vehicles are ignored. In particular, the mobility requirements of users need not be known in advance, nor transmitted to the aggregator 10. This allows the exchanges of information between the aggregator 10 and the electric vehicles VE to be greatly decreased, with respect to the conventional aggregating methods used to control charging of a plurality of vehicles at the same time.

[0035] More precisely, the adapting algorithm implemented at the end of the charging manager associated with each battery, i.e. charging manager side, receives as input data the optimal profile of charging over the day R.sub.ideal in percentage per quarter of an hour, according to the example, and data representative of the mobility requirements of the user of the vehicle. These data representative of the mobility requirements of the user may take the form of a state of charge called the desired state of charge of the battery SOC.sub.desired, which may for example correspond to a minimum state of charge to be reached each day, typically at the end of the night, for example at six o'clock in the morning, knowing that this state of charge will be able to be exceeded. The charging manager also receives as input data the nominal capacity of the battery E.sub.max (in kWh). For example, a user with a battery the nominal capacity of which is equal to 60 kWh, will possibly believe that with a stored energy level of 30 kWh, this corresponding to a state of charge of 50%, he will be able to make most of the journeys he wants to. He will thus possibly set his desired state of charge SOC.sub.desired to 50%, thus leaving the charging manager associated with the battery a flexibility range of 30 kWh, for example, as will be described in more detail below, to perform the CO2 optimization, or in other words to take advantage of charging ranges that are optimal from a point of view of the availability of low-carbon power on the electricity distribution grid.

[0036] The charging manager of the battery will also receive as input data a coefficient representative of a percentage to be reached per day with respect to this flexibility range. In other words, according to the example given above, the coefficient K will indicate the percentage to be reached each day in the 30 kWh of flexibility permitted by the charging manager.

[0037] On the basis of these input data, the algorithm for adapting the curve R.sub.ideal is designed to compute a daily target state of charge SOC.sub.target, then the target energy E.sub.target to be reached depending on this daily target state of charge:

SOC.sub.target=SOC.sub.desired+K(1−SOC.sub.desired)  Math. 1

E.sub.target=SOC.sub.target*E.sub.max  Math. 2

[0038] At each time, the charging manager associated with the battery computes a difference ΔE between a current energy level of the battery E.sub.t and the target energy value of the day such as computed beforehand:

ΔE=E.sub.target−E.sub.t.  Math. 3

[0039] To adapt its own charging curve with respect to its times of connection and its charging power, the adapting algorithm uses proportional-integral regulation of the difference ΔE.

E.sub.delta=ΔE*(1+k*∫ΔE), with k=integral coefficient.  Math. 4

[0040] Thus, when the regulated difference E.sub.delta is positive, this means that the desired state of charge has not been reached and, thus, the adapting algorithm must command the battery to be charged as soon as the state of the network (resulting from the ideal charging profile R.sub.ideal sent to the vehicle) allows it. In contrast, when E.sub.delta is negative, this means that the battery has exceeded the desired state of charge and it is therefore necessary to command the battery to discharge to the grid at propitious moments, such as indicated by the ideal charging profile R.sub.ideal.

[0041] The adapting algorithm then deduces, from the difference E.sub.delta, the charging power level P.sub.charge to be applied to the battery, via the following strategy:

[0042] If E.sub.delta>0, then (case where the battery must be charged):

If R.sub.ideal>0, then P.sub.charge=E.sub.delta*R.sub.ideal

Else P.sub.charge=.sup.0

[0043] Else if E.sub.delta≤0, then (case where the battery must be discharged):

If R.sub.ideal<0, then P.sub.charge=−E.sub.delta*R.sub.ideal

Else P.sub.charge=0.  Math. 5

[0044] All that there remains to do, to determine the charging power P.sub.final to be applied in the end, is to take into account the moments at which the battery is connected to the electricity distribution grid and the power constraints in terms of the minimum and maximum charging powers P.sub.min and P.sub.max of the battery of the vehicle:

If |P.sub.charge|>P.sub.min, then: [0045] If the vehicle is not connected then P.sub.final=0 [0046] Else:

If P.sub.charge>0 then P.sub.final=min(P.sub.charge,P.sub.max)

Else P.sub.final=max(P.sub.charge,−P.sub.max)  Math. 6

[0047] Else P.sub.final=0.

[0048] Thus, a charging law that follows, as much as possible, the ideal charging profile R.sub.ideal received as input, without discontinuity in the amounts of charging power drawn from the grid, is obtained while needing to know nothing about how the vehicle will be used. As illustrated in FIG. 2 by curves representative of the consumption of the vehicle CONSO in kWH and of the charging power P.sub.final in kW resulting from the charging law set by the adapting algorithm, it may be seen, on the one hand, that the battery is indeed charged and discharged at the propitious moments indicated by the signal of the ideal charging profile R.sub.ideal and, on the other hand, that a higher consumption than usual, symbolized by the consumption peak in the signal CONSO, will lead to greater charging of the battery.

[0049] Management of charging is thus advantageously carried out in open loop: the charging profile computed by the aggregator allows an optimal control law to be defined, i.e. an ideal charging profile to be followed each day over a period of several days, taking into account solely constraints specific to the grid, and not feedback from end users (status of connection of the battery to the grid, current state of charge of the batteries, etc.), this ideal charging profile being transmitted to the charging managers associated with the batteries, which locally control the system for charging the battery. At regular intervals, for example once per week, a new control law is recomputed taking into account the state of the grid.

[0050] Advantageously, as indicated above in the description, it is possible to take into account an ideal charging profile computed on the basis of the CO2 emissions associated with the generation of the power available on the grid, to manage the phases of charging and discharging of the battery of vehicles connected to the grid and, in particular, to charge batteries more in periods in which the power available on the grid is the result of electrical power generation with low emissions of CO2.

[0051] Thus, in this context, which is the to be of optimization of CO2 emissions, it is recommendable for the ideal charging profile delivered by the aggregator 10 to also depend on the CO2 emissions associated with the generation of the electricity available on the electricity distribution grid 20. Thus, the aggregator 10 advantageously receives as input, with a view to establishing the ideal charging profile R.sub.ideal, a model predicting the CO2 emissions (or the price of the electricity, the latter being highly correlated to the CO2 emissions of the electricity generation) over a relatively long time period, typically over several days and preferably over one week. Specifically, the presence of two distinct periods in the variation in the CO2 emissions of electricity generation have been observed, namely a period of daily variation, which corresponds to the period of generation of solar power by photovoltaic solar plants connected to the grid, and a period of weekly variation, which corresponds to the period of generation of wind power by wind turbines connected to the grid. Thus, taking a model predicting CO2 emissions over several days and preferably over one week into account in the computation of the on-grid charging profile that is ideal for all the vehicles allows variations in the CO2 emissions of electricity generation and especially those associated with the electricity generated by wind turbines to be better integrated into the ideal charging profile R.sub.ideal. The model predicting CO2 admissions may for example be deduced by the aggregator from various sources of data, including metrological data (insolation, etc.), on which the generation of low-carbon electrical power in a given region depends.

[0052] The model predicting CO2 emissions associated with the generation of the power available on the grid for example delivers a CO2-emission value every 15 minutes during a time of one week and therefore delivers a CO2-emission prediction indicator structured in the form of a vector containing 672 components.

[0053] Initially, the daily target state of charge SOC.sub.target will be computed depending on the coefficient K, so as to allow the percentage of the flexibility range that the battery of the vehicle must reach, to charge more on days with low CO2 emissions, to be computed.

[0054] To do this, on the basis of the CO2 prediction provided as input, on the one hand the average of the CO2 emissions over the week CO2.sub.week-avg and on the other hand the average of the CO2 emissions per day CO2.sub.day-avg are computed for all the vehicles.

[0055] The integral I.sub.CO2 of the difference between these values is computed:

I.sub.CO2=∫(CO2.sub.week-avg−CO2.sub.day-avg)  Math. 7

[0056] and the maximum and minimum values I.sub.CO2max, I.sub.CO2min that the integral thus computed may take are determined:

I.sub.CO2max=max(I.sub.CO2)

I.sub.CO2=min=min(I.sub.CO2).  Math. 8

[0057] The coefficient K representative of the percentage of the permitted flexibility range that it is desired to reach is then deduced for each day of the week:

[00001]$\begin{array}{cc}K=\frac{I_{CO2}-I_{\mathrm{CO}2\min }}{I_{\mathrm{CO}2\max }-I_{\mathrm{CO}2\min }}& \mathrm{Math}.9\end{array}$

[0058] This coefficient K is identical for all the vehicles.

[0059] As explained above, this coefficient is then applied to the mobility requirements of each user, these requirements being defined in the form of a state of charge referred to as the desired state of charge of the battery SOC.sub.desired, in order to deduce therefrom the daily target state of charge SOC.sub.target to be reached:

SOC.sub.target=SOC.sub.desired+K(1−SOC.sub.desired).  Math. 10

[0060] Thus, on days on which the power available on the grid corresponds to high CO2 emissions, the daily target state of charge SOC.sub.target to be reached will be equal to the desired state of charge of the battery SOC.sub.desired, i.e. to the state of charge that the user desires to obtain at his departure time. In contrast, on days on which the power available on the grid corresponds to low CO2 emissions, the daily target state of charge SOC.sub.target to be reached will be higher. The battery is therefore charged more on days with low CO2 emissions.

[0061] FIG. 3 illustrates an example of variation over a week in the daily target state of charge SOC.sub.target to be reached, in the form of histograms, as a function of the percentage K of the flexibility range to be reached, which percentage is computed each day according to the principles exposed above, on the basis of the values of the average of the CO2 emissions over the week CO2.sub.week-avg and of the average of the CO2 emissions per day CO2.sub.day-avg. In FIG. 3, the portion of each daily histogram shaded in gray represents the desired state-of-charge level SOC.sub.desired, and the portion of each histogram that is not shaded in gray represents the permitted flexibility range that results therefrom with respect to the nominal capacity of the battery, the location of the daily target state of charge SOC.sub.target to be reached, which is symbolized by the horizontal line, in this flexibility range depending on the value of the coefficient K, which may take values between 0% and 100% depending on the level of any CO2 emissions.

[0062] In the same way, to compute the ideal distribution of charging over the day, the level of the CO2 emissions at a time t CO2.sub.current is compared to the average of the CO2 emissions per day CO2.sub.day-avg, so as to allow charging to be requested only at moments when the level of the CO2 emissions is below average. Thus, the following is computed:

Δ.sub.CO2_day=CO2.sub.day-avg−CO2.sub.current.  Math. 11

[0063] Subsequently, the distribution is normalized:

[00002]$\begin{array}{cc}{R}_{\mathrm{flee}t}=\frac{{\Delta }_{\mathrm{CO}2\_\mathrm{day}}}{\frac{\int \mathrm{abs}\left({\Delta }_{\mathrm{CO}2\_\mathrm{day}}\right)}{2}}.& \mathrm{Math}.12\end{array}$

[0064] Thus, a charging profile R.sub.fleet equivalent to the ideal charging profile R.sub.ideal but computed on the basis of CO2 emissions is obtained. This charging profile R.sub.fleet is then defined to be the charging distribution per day over a week, in percentage per quarter of an hour (or another sampling increment), computed on the basis of CO2 emissions, as explained above. The sum of all the (positive and negative) percentages over the day must be equal to 0. This ideal charging profile corresponds to the profile to be followed by all of the vehicles in order to achieve the daily target state of charge SOC.sub.target.

[0065] FIG. 4 is a graph illustrating an example of such a charging distribution over one day. When the level of the CO2 emissions at a time t CO2.sub.current is lower than the average over the day of the CO2 emissions CO2.sub.day-avg, the signal R.sub.ideal corresponds to a positive percentage of charge, meaning that the moment is propitious to charging if the battery is connected to the grid with a view to recharging thereof. In contrast, when the level of the CO2 emissions at a time t CO.sup.2.sub.current rises above the average over the day of the CO2 emissions CO2.sub.day-avg, the signal R.sub.fleet corresponds to a negative percentage of charge, meaning that the moment is instead propitious to returning energy stored in the battery to the electricity distribution grid, if the battery of the vehicle is connected to a reversible charging device. The percentage of charge per quarter of an hour in absolute value increases as the difference between the level of the CO2 emissions at a time t CO2.sub.current and the average over the day of the CO2 emissions CO2.sub.day-avg increases.

[0066] Control over several days and preferably over one week of the daily state of charge to be reached, i.e. of the daily charging target defined depending on the mobility requirements of the user, advantageously makes it possible to take maximum advantage of the electrical power made available to the grid by wind turbines and solar power plants. This allows the battery to be charged at moments that are optimal, both from the point of view of CO2 emissions but also from an economical point of view, these two factors being highly correlated.

[0067] Moreover, this method may be easily adapted to the needs of grid managers. Specifically, it is enough to adapt the scale of the model used to predict CO2 emissions to obtain, for example, control on a scale corresponding to a district, allowing congestion of the grid to be avoided, or to return to a model predicting CO2 emissions on a national scale to balance demand. It is also possible to use this method on the scale of an installation to optimize self-consumption.

[0068] The input model used to predict CO2 emissions with a view to determining the optimal charging profile allows the charging manager associated with each battery to compute charging requests for each vehicle of the set of electric vehicles, this allowing the moments most propitious to charging to be monitored without creating any discontinuity in the consumption from the electricity distribution grid and therefore without creating any additional disruption to the electricity grid, and all this to be achieved while taking into account CO2-emission differences related to the generation of electricity over several days.

[0069] The charging profile R.sub.fleet computed beforehand is advantageously the charging profile to be followed by all the vehicles, in particular when it is a question of a large set of vehicles. However, as a variant, for a smaller set of vehicles, the impact of which on the electricity distribution grid will be smaller, it may be more advantageous to seek to concentrate charging times at times determinable as being the most advantageous from the point of view of CO2 emissions and thus from an economical point of view. To do this, an ideal charging profile, referred to as the “concentrated” charging profile R.sub.fleet_concentrated, able to define these charging times, is defined by applying the following method:

[00003]$\begin{array}{cc}\mathrm{If}{R}_{\mathrm{fleet}}>0\mathrm{then}{R}_{\mathrm{fleet}\_\mathrm{concentrate}d}=\frac{R_{\mathrm{fleet}}}{\frac{1-C}{\max \left(R_{\mathrm{fleet}}\right)}*R_{\mathrm{flee}t}+C}\mathrm{Else}{R}_{\mathrm{fleet}\_\mathrm{concentrate}d}=\frac{R_{\mathrm{fleet}}}{\frac{1-C}{\min \left(R_{\mathrm{fleet}}\right)}*R_{\mathrm{flee}t}+C}& \mathrm{Math}.13\end{array}$

[0070] with c being a concentration coefficient comprised between 1 and Infinity (defined depending on the size of the set of vehicles). When c=1, no concentration is applied and R.sub.fleet_concentrated=R.sub.fleet.

[0071] Once again, the distribution is normalized:

[0072] If R.sub.fleet_concentrated>0 then:

R.sub.fleet_concentrated_pos=R.sub.fleet_concentrated

R.sub.fleet_concentrated_neg=0

[0073] If R.sub.fleet_concentrated<0 then:

R.sub.fleet_concentrated_pos=0

R.sub.fleet_concentrated_neg=R.sub.fleet_concentrated.  Math. 14

[0074] The final charging profile R.sub.fleet-final is then deduced for the set of vehicles:

[00004]$\begin{array}{cc}\mathrm{If}{R}_{\mathrm{fleet}\_\mathrm{concentrated}}>0\mathrm{then}:R_{\mathrm{fleet}\_\mathrm{final}}=\frac{R_{\mathrm{fleet}\_\mathrm{concentrated}\_\mathrm{pos}}}{\int R_{\mathrm{fleet}\_\mathrm{concentrated}\_\mathrm{pos}}}\mathrm{Else},R_{\mathrm{fleet}\_\mathrm{final}}=\frac{R_{\mathrm{fleet}\_\mathrm{concentrated}\_\mathrm{neg}}}{\int R_{\mathrm{fleet}\_\mathrm{concentrated}\_\mathrm{neg}}}.& \mathrm{Math}.15\end{array}$

[0075] Thus, for a set containing few vehicles, a high concentration coefficient c will be chosen as this will allow the vehicles to be charged at times that are the most advantageous from the point of view of CO2 emissions, a veritable grid discontinuity not being created because of the small impact that a small number of vehicles has with respect to overall power consumption. The larger the set of vehicles, the lower the concentration coefficient c, thus allowing grid constraints to be met as well as can be.