PROCEDURE FOR CONTROLLING PUBLIC LIGHTING

Description

[0001] The invention relates to the field of public lighting. More specifically, the invention relates to a method for controlling public lighting.

[0002] Conventionally, public lighting, particularly urban lighting, in a given geographical area is switched on automatically between a start time of service and an end time of service, which are generally determined according to local sunset and sunrise times. This type of automatic control is not satisfactory from an ecological and economic point of view, since the public lighting then remains on during periods of the night when there is no need for lighting.

[0003] In order to meet public lighting policies imposing ecological or economic constraints, it has been considered to turn off public lighting during certain predefined periods of the night. However, this solution is also not satisfactory because it does not take into account the needs of people living or traveling in the geographical area. Thus, in the event of a change in residents' habits, particularly with regard to the seasons, or in the event of particular events involving a change in population flow through the area, turning off public lighting may become unsuitable for users' needs. Similarly, the need for lighting may vary depending on the number of people moving around the area.For example, it may be necessary to significantly illuminate a street when only a few people are moving along it at night, in order to generate a feeling of security, while dim lighting may be sufficient when a significant number of people are crossing it, the sensation of crowding reducing the feeling of insecurity.

[0004] In order to be able to control public lighting in a more appropriate way, so-called smart public lighting has also been imagined, in which the luminaires, candelabras and other light sources of public lighting are equipped with presence sensors allowing the automatic emission of light when a person is detected. While this type of lighting effectively allows the lighting to be modulated according to the presence of people, it has several drawbacks. On the one hand, in this type of solution, the lighting is activated as soon as the sensor detects the presence of a person, which does not meet the need to adapt the lighting to the number of people passing through the geographical area. On the other hand, this type of lighting requires specific and expensive equipment for each light source.Finally, this type of lighting does not allow for a restrictive policy in terms of economy or ecology to be met, since it is not possible to modulate, at the level of a given area, the overall electricity consumption of public lighting in that area. FR 3 081 646 A1 describes a process for deactivating public lighting during predetermined time slots when it is unlikely that it will be useful.

[0005] WO 2018 / 091315 A1 describes a system for improving the experience of customers moving around a commercial area, in which telecommunications data makes it possible to determine, by means of machine learning algorithms, movement patterns.

[0006] In this context, there is thus a need for a method for controlling public lighting in a given geographical area, making it possible to adapt the lighting in particular to the needs of users, and in particular to dynamic changes in the travel habits of users, which is universal for all types of light sources in this public lighting and does not require specific and expensive equipment. The present invention is placed in this context and aims to meet this need.

[0007] For these purposes, the invention relates to a method for controlling public lighting in a predetermined geographical area, the method comprising the following steps: a. Provision of a data set relating to the presence of people in a geographical estimation area, obtained from mobile terminals of these people; b. Determination of at least one population flow model in said predetermined geographical area from said data set; c. Definition of a control law for public lighting in said geographical area based on said population flow model and application of said control law to this public lighting.

[0008] According to the invention, the geographical estimation area may correspond to the predetermined geographical area or, alternatively, cover all or part of the predetermined geographical area. As a further alternative, the estimation area may be an area different from the predetermined geographical area, and of the same category as the predetermined geographical area. Geographical areas of the same category are understood to mean geographical areas having at least one identical intrinsic characteristic, for example the presence of a shopping centre, a stadium, a residential area. The population flow model may thus be determined directly for the geographical area concerned or may be determined for a geographical area of ​​the same category and transposed to the predetermined geographical area.

[0009] According to the invention, the data set provided may comprise a plurality of indicators relating to the presence and / or mobility of people in the geographical estimation area over time in a given overall time interval. Where appropriate, the population flow model may comprise one or more values ​​representative of the correlation, linear or not, of these indicators with each other. The population flow model is thus a representation of the movement habits of the population within the area.

[0010] It is thus understood that thanks to the invention, the population flow in the predetermined geographical area can be modeled from a data set reflecting the position of users over time in an estimation zone. This model is then used to define a public lighting control law which is particularly adapted to the movement habits of the population in this predetermined geographical area while respecting constraints imposed by an economic and / or ecological policy of the predetermined geographical area. It is further understood that public lighting can be controlled in a simple manner, without requiring each light source to be equipped with a presence sensor.

[0011] In one embodiment of the invention, the data set provided comprises an estimate of the number of people entering, present and leaving the estimation zone during a plurality of elementary time intervals of a given overall time interval. Other indicators relating to the presence of people in the estimation zone may be provided, cumulatively or alternatively, and in particular an estimate of the change in population density within the estimation zone over time, an estimate of the change in the number of residents of the estimation zone present in the estimation zone over time, an estimate of the change in the time of presence of people present in the estimation zone over time, or any other indicator representative of the mobility of people in the estimation zone over time.

[0012] For example, said estimation of the number of people entering, present in and leaving the estimation zone can be obtained from information relating to the position over time of a plurality of mobile terminals, this information being determined from data generated by relay antennas of a mobile telephone network located in the estimation zone, to which or to which these mobile terminals have connected at a given time. Indeed, for each elementary time interval, it is possible to determine the position of each of the mobile terminals of said plurality of mobile terminals, for example by triangulation.Each mobile terminal connecting regularly, during the same elementary time interval, to at least two or three relay antennas which are close to it, it is possible to determine the distance of this terminal, for example by measuring the flight time, to each of these relay antennas and thus to estimate, by triangulation, the position of this mobile terminal during this elementary time interval. According to the invention, it is thus possible to determine, from these distances, whether a mobile terminal has entered or left the estimation zone.

[0013] Alternatively or cumulatively, the data set provided may include information relating to the mobility of people in the estimation area over time, obtained from geolocation data provided directly by the mobile terminals of these people, for example collected and aggregated by means of one or more software applications installed on these mobile terminals.

[0014] According to the invention, the step of determining the model comprises a sub-step of generating a plurality of population flow models in said predetermined geographical area, called test models, each test model being generated by a separate machine learning algorithm from said data set, each test model being associated with an evaluation metric, and a sub-step of selecting a population flow model from among the plurality of test models as a function of the evaluation metrics of said plurality of test models. Where appropriate, the model determined at the end of the determination step is the selected test model.

[0015] Advantageously, the data set comprising a plurality of indicators relating to the presence and / or mobility of people in the geographical area of ​​estimation over time in a given overall time interval, each machine learning algorithm determines, in the generation sub-step, regression or correlation coefficients between two or more of these indicators, according to a given mathematical method and one or more given hyperparameters, the set of coefficients forming said test model. For example, the distinct machine learning algorithms may be algorithms implementing distinct mathematical methods and / or algorithms implementing the same mathematical method with one or more distinct hyperparameters.For example, we can use regressor-type algorithms, supervised or not, and in particular algorithms such as: decision trees, random forest, simple linear regression, multiple linear regression, logistic regression, k-nearest neighbors method (also called KNN), data partitioning, in particular K-means, neural network.

[0016] Advantageously, the generation sub-step comprises a step of separating the provided data set into a so-called training set and a so-called evaluation set, each test model being generated from said training set and a step of evaluating each test model, in which the evaluation metric is determined by comparing data predicted by the test model and the evaluation set. A simple separation or a cross-validation type separation may be provided. For example, an evaluation metric of one of the following types may be determined for each test model: confusion matrix, area under the ROC curve (Receiver Operating Characteristic), root mean square error (also called RMSE), relative square error, coefficient of determination.Where appropriate, the step of selecting a population flow model from among the plurality of test models may, for example, consist of selecting the test model with the largest area under the ROC curve.

[0017] According to one embodiment of the invention, the method may comprise a step of providing at least one additional set of contextual data relating to said predetermined geographical area. Where appropriate, the population flow model in said predetermined geographical area is determined from said data set and the additional data set. It may, for example, be data relating to an event in the predetermined geographical area, meteorological data in the predetermined geographical area and / or so-called counting data in the predetermined geographical area, for example corresponding to a count of the number of people crossing one or more streets, or a sub-area of ​​the predetermined geographical area over time, in particular obtained by means of one or more magnetic induction loops installed in this or these streets or this sub-area.The provided dataset is thus enriched with data from other sources in order to obtain a population flow model allowing a reliable prediction in the predetermined geographical area.

[0018] According to one embodiment of the invention, the method comprises a step of classifying the data of the data set among data categories each associated with a geographical sub-area. Where appropriate, the step of determining the model comprises, for each geographical sub-area, a sub-step of determining a sub-model of population flow in this geographical sub-area from the data classified in the data category associated with this geographical sub-area. Advantageously, the method may comprise a step of providing one or more additional sets of data relating to the presence of people in another or more other geographical estimation areas, obtained from mobile terminals of these people. Advantageously, the classification step may be carried out using one or more automatic classification algorithms based on supervised or unsupervised machine learning.It is thus possible to categorize the data using sub-zones, predetermined or not, for example to determine categories of zones, in each of which the method according to the invention will thus define a sub-model of population flow, which can then be used to define a control law for public lighting in another zone of the same category.

[0019] According to one embodiment of the invention, the method comprises a step of providing a prior data set relating to the presence of people in the geographical estimation area at a given time, a step of predicting a population flow model at a future time relative to the given time from the population flow model determined during the determination step and the prior data set, the control law being defined from the predicted population flow model. For example, the population flow model at a future time may be predicted by means of a machine learning algorithm, from said prior data set and the model determined at the end of the determination step.For example, similarly to the step of determining the population flow model, the step of predicting the population flow model may consist of selecting a model from a plurality of test models generated using a plurality of machine learning algorithms from the model determined during the determining step and said prior data set, the model being selected based on an evaluation metric.

[0020] If desired, the method may comprise a step of providing at least one piece of information relating to a future state of said predetermined geographical area at said future time, the step of predicting a population flow model at a future time being carried out from the population flow model determined during the determination step, the previous data set and said information relating to the future state of said predetermined geographical area. The information relating to a future state may be information relating to an upcoming event in the predetermined geographical area, a meteorological prediction in the predetermined geographical area.

[0021] Advantageously, the method may comprise a step of providing a posterior data set relating to the presence of people in the geographical area of ​​estimation at said future time, this step being carried out at the future time or later than this future time, and a step of correcting said predicted population flow model using said posterior data set. For example, similarly to the step of determining the population flow model, the step of predicting the population flow model may consist of selecting a corrected model from a plurality of test models generated using a plurality of machine learning algorithms from the model determined during the determination step and said prior data set, the corrected model being the one that minimizes a prediction error, measured from a comparison between data predicted by the predicted model and the posterior data set.This allows for relearning to adapt the model to changes in the population's travel habits.

[0022] Advantageously, the steps of providing a prior data set, predicting a population flow model, defining and applying a control law to public lighting, providing a posterior data set and correcting the predicted population flow model are repeated recurrently. Where appropriate, the corrected population flow model and the posterior data flow provided during an iteration of these steps being respectively the population flow model provided for the prediction step and the prior data set provided during the following iteration of these steps.

[0023] In one embodiment of the invention, the step of defining a control law for public lighting in said geographical area comprises defining at least one period of switching off and / or decreasing and / or switching on and / or increasing the public lighting as a function of said population flow model.

[0024] For example, the method may comprise a step of determining, from the determined or predicted population flow model, at least a first period in which the population flow is lower than a given first threshold and / or at least a second period in which the population flow is higher than a given second threshold, in which the public lighting control law defines a setpoint for the quantity of luminous flux to be emitted for each light source of the public lighting in the predetermined geographical area, the control law defining a first luminous flux setpoint for the first and / or second period. It is thus understood that it is possible to adapt the public lighting to the number of people moving in the geographical area, for example by reducing the public lighting when there are no or few people and / or on the contrary when there are many people, the need for lighting is then low or even non-existent.

[0025] If desired, the first luminous flux setpoint may be identical for all the light sources of the public lighting in the predetermined geographical area or, alternatively, may be specific to each light source of the public lighting in the predetermined geographical area. For example, in the case of an incandescent or discharge lamp type light source, the control law may consist of switching off this light source, the first luminous flux setpoint then being zero. In the case of a light source of the light-emitting diode type, the first luminous flux setpoint may be zero or, alternatively, may be a non-zero value and substantially lower than the nominal luminous flux that this light source is capable of emitting.

[0026] Advantageously, the method comprises a step of determining, from the determined or predicted population flow model, at least a third period in which the population flow is between the first given threshold and the second given threshold, the control law defining a second luminous flux setpoint for the third period. It is thus understood that it is possible to adapt the public lighting to the number of people moving in the geographical area, by switching on the public lighting when the number of people justifies it, so as to create in particular a feeling of security. For example, in the case of an incandescent or discharge lamp type light source, the control law may consist of switching on this light source, the second luminous flux setpoint then being identical to the nominal luminous flux that this light source is capable of emitting.In the case of a light source of the light-emitting diode type, the second luminous flux setpoint may be substantially equal to the nominal luminous flux that this light source is capable of emitting or, alternatively, be greater than the first luminous flux setpoint.

[0027] Advantageously, the method comprises a step of providing an overall energy consumption setpoint, and the periods of switching off, decreasing, switching on and / or increasing the public lighting are defined according to the population flow model and so that the energy consumption of the public lighting during all of these periods complies with said setpoint.

[0028] If desired, the method may include a step of providing information relating to the ambient brightness of the predetermined geographical area, the periods of switching off, decreasing, switching on and / or increasing the public lighting being defined as a function of the population flow model and this ambient brightness information. The information relating to the ambient brightness may, for example, be meteorological information, information relating to the duration of dawn and / or dusk, information on the lunar phase, information relating to the presence of commercial lighting installations. It is thus also possible to adapt the public lighting according to the local ambient brightness of the predetermined geographical area.

[0029] The present invention is now described with the aid of examples which are solely illustrative and in no way limitative of the scope of the invention, and from the attached illustrations, in which: [ Fig. 1 ] represents, schematically and partially, a view of a method for controlling public lighting according to an embodiment of the invention; [ Fig. 2 ] represents, schematically and partially, a top view of a geographical area with public lighting controlled according to the process of [ Fig. 1 ]; [ Fig. 3 ] represents, schematically and partially, a plurality of indicators relating to the presence of people in the area of ​​the [ Fig. 2 ] ; [ Fig. 4 ] represents, schematically and partially, a step of prediction of a population flow model in the area of ​​the [ Fig. 2 ], according to the process of [ Fig. 1 ]; [ Fig. 5 ] represents, schematically and partially, a stage of definition of a law of control of the public lighting of the zone of the [ Fig. 2 ], according to the process of [ Fig. 1 ]; And [ Fig. 6 ] represents, schematically and partially, a stage of definition of another law of control of the public lighting of the zone of the [ Fig. 1 ].

[0030] In the following description, elements which are identical, by structure or by function, appearing in different figures retain, unless otherwise specified, the same references.

[0031] We have represented in [ Fig. 1 ] a method of controlling 1 public lighting in a geographical area ZG, which has been represented in [ Fig. 2 ].

[0032] The geographical area ZG (represented by long dotted lines) is an urban residential area, gridded by a set of streets equipped with light sources, for example street lamps, candelabras, lampposts, forming public lighting for this area ZG.

[0033] A plurality of relay antennas 2 of a mobile telephone network are distributed in and around the geographical area ZG by defining an estimation zone ZE (represented by short dotted lines). The estimation zone ZE can for example be defined by the perimeter of a Voronoi diagram obtained from the positions of these antennas 2.

[0034] When a user's mobile terminal is in the estimation zone ZE, it connects, within a sufficiently short time interval, to at least three relay antennas 2 that are close to it. It is thus possible to estimate, by triangulation, its position in the zone ZE. These connections to the relay antennas 2 being made regularly, it is thus possible to follow the position of the mobile terminal over time. The relay antennas 2 can thus generate data reflecting the movements of the users in the estimation zone ZE. For example, it is possible to obtain a series of data I1 indicating the number of users entering the zone ZE during an elementary time interval, for example every 30 minutes, during a global time interval, for example a day, by identifying the new mobile terminals entering the zone.It is also possible to obtain a data series 12 indicating the number of users leaving the ZE zone in each elementary time interval of this global time interval, and a data series 13 indicating the number of users present in the ZE zone in each elementary time interval of this global time interval. These data series I1, I2 and 13 have been represented in [. Fig. 3 ]

[0035] These data series I1 to 13 are thus aggregated into a data set D, which is provided, in a first step E11, as input to method 1 according to the invention.

[0036] It should be noted that the number and types of indicators in the example described are not limiting, and that the data set provided in step E11 may comprise M series of data I 1 to IM of different types, acquired or not by means of the relay antennas 2, each series of data I 1 to IM being an indicator relating to the presence and / or mobility of people in the zone ZE over time.

[0037] In a step E2, the method determines, from the data set D, a model M of population flow in the geographical area ZG.

[0038] For these purposes, the dataset D is enriched using an additional dataset DG, provided in a step E12. The set DG is a contextual dataset relating to the geographical area ZG, comprising information relating to events having taken place during the overall time interval in the area ZG, meteorological data in the area ZG during the overall time interval, as well as data on counting people moving in the area ZG during the overall time interval, recorded in situ.

[0039] The enriched dataset D, DG is separated into a training set and a validation set. In a step E21, a plurality of test population flow models M 1 to MN is generated. More specifically, step E21 comprises a plurality of steps E31 to E3N. In each step E3i, a test model M i is generated by a separate machine learning algorithm from the training set. The machine learning algorithms employed in these steps E31 to E3N may implement separate mathematical methods and / or the same mathematical method employing one or more separate hyperparameters. In the example described, each of the models M 1 to MN is a model of the number of people moving in the zone ZG over time, generated by an algorithm implementing a multiple polynomial regression type method with a separate set of hyperparameters for each step E31 to E3N, from the training set.Each population flow model is thus defined by a set of regression coefficients allowing the mathematical prediction of the number of people moving in the ZG zone over time based on new data.

[0040] Furthermore, in each step E3i, each test model M 1 to MN is evaluated by comparing data predicted by this test model and the data of the validation set, and thus determining an evaluation metric, for example a root mean square error.

[0041] In a step E22, the population flow model M having the best metric among the test models M1 to MN is selected.

[0042] There [ Fig. 4 ] represents, at the top, the population flow model M, for example representing an estimate of the number of people moving in the geographical area ZG over time. In a step E4, a population flow model at a future time M' is predicted, from the model M. For this purpose, in a step E41, a so-called prior data set DT, relating to the presence of people in the area ZE during a time interval preceding the future time, is provided. For example, this prior data set DT can be acquired in a similar way to the data set D. We have represented, in the middle graph of the [ Fig. 4 ] an example of a previous DT dataset.

[0043] A model M' is then predicted, at step E4, using the model M from step E2 and the previous dataset DT . The bottom graph of the [ Fig. 4 ] represents an example of a predicted model M', with model M represented by a dotted line. Note that the prediction step E4 can be similar to step E2, by selecting the best metric model from a plurality of models generated by distinct machine learning algorithms, from model M and the prior dataset DT .

[0044] In a step E5, a control law LC for the public lighting of the geographical zone ZG is defined, using the model M' predicted in step E4.

[0045] For this purpose, the method comprises a step E51 of providing an overall energy consumption instruction C and a step of receiving information D' G relating to the ambient brightness of the geographical area ZG. In the example described, this involves meteorological information and the times of sunrise and sunset, as well as the durations of dawn and dusk.

[0046] In the example described, the step of defining the control law LC comprises a step of defining periods T1 in which the model M' is lower than a given threshold A1 defined as a function of the consumption setpoint C, for example 2 people, periods T2 in which the model M' is higher than a given threshold A2 higher than the threshold A1 and defined as a function of the consumption setpoint C, for example 30 people, and periods T3 in which the model M' is between the threshold A1 and the threshold A2.

[0047] We have represented in [ Fig. 5 ] on the top graph the model predicts M', these different periods T1, T2, T3 as well as the information D' G .

[0048] The bottom graph of the [ Fig. 5 ] represents a control law LC defined, in a step E32, as a function of these periods T1, T2 and T3 and of this information D' G . This control law LC defines a luminous flux instruction to be emitted by each of the light sources of the public lighting of the geographical zone ZG.

[0049] In the first period T2, due to sunlight, there is no need to illuminate the geographical area ZG, so the luminous flux setpoint of the public lighting of the area ZG is zero. In the second period T2, the brightness decreases due to sunset. However, this brightness is still sufficient, and the number of people moving in the geographical area ZG according to the model M' does not require lighting.

[0050] In the third period T2, we observe the appearance of clouds reducing the ambient brightness. Although the number of people moving in the geographical area ZG according to the model M' is significant, it is still necessary to turn on the public lighting in order to meet the needs of these people. The meteorological prediction being provided to the process, the control law LC thus defines a non-zero luminous flux setpoint, corresponding for example to 100% of the nominal luminous flux that each light source of the public lighting in the geographical area ZG is likely to emit.

[0051] In the first period T3 following the third period T2, clouds still impact the ambient brightness, while the number of people moving in the geographical area ZG according to the M' model decreases. The LC control thus maintains the luminous flux setpoint at 100% of the nominal flux.

[0052] In the fourth period T2 following the first period T3, the clouds have disappeared and the ambient brightness increases. In addition, the number of people moving in the geographical area ZG according to the M' model increases. Under these conditions, it is considered that it is not necessary to illuminate the geographical area ZG, the number of people being sufficient in itself to generate a feeling of security. The LC control law thus defines a zero luminous flux setpoint.

[0053] In the second period T3 following the fourth period T2, the situation is equivalent to the first period T3, so that the LC control law defines a luminous flux setpoint at 100% of the nominal flux.

[0054] In the first period T1 following the second period T1, the number of people moving in the geographical area ZG according to the model M' is particularly low. There is therefore no need for public lighting. The control law LC thus defines a zero luminous flux setpoint.

[0055] Finally, in the third period T3 following the first period T1, the situation is equivalent to the first period T3, so that the LC control law defines a luminous flux setpoint at 100% of the nominal flux.

[0056] The different periods for switching on and off public lighting are thus defined in step E5 so as to comply with consumption instruction C.

[0057] In a step E6, the control law LC is then transmitted to the light sources of the public lighting of the geographical zone ZG, which apply it to respect the setpoint values ​​that this control law LC defines.

[0058] It should be noted that step E4 of predicting the model M' is repeated recurrently. In order to be able to adapt the model M to changes in the population's travel habits, the method comprises a step E42 of providing a posterior data set D T+1 , relating to the presence of people in the zone ZE during a time interval at the future time, is provided. For example, this posterior data set D T+1 may be acquired in a similar manner to the data set D, at the future time, and provided during step E42, during the future time or later.

[0059] The method comprises a step E43 of correcting the model M' into a corrected model M, using this posterior data set D T+1 . It will be noted that the correction step E43 can be similar to step E2, by selecting the model M from a plurality of models generated by distinct machine learning algorithms, from the model M', which minimizes the prediction error between the data it predicts and the posterior data set D T+1 .

[0060] In the next iteration of the prediction step E4, the corrected model M from the correction step E43 of the previous iteration as well as the posterior dataset D T+1 are then used to predict the next model M'.

[0061] In the example of the [ Fig.5 ], the LC control law defines two binary setpoint values, since the public lighting in the geographical area considered is discharge lamp type lighting, the light intensity of which can be difficult to control.

[0062] We have represented in [ Fig. 6 ] another example of LC2 control law, suitable for public lighting of the light-emitting diode type, the light intensity of which can be easily controlled.

[0063] We can thus see that during the second and third periods T2, due to the progressive decrease in the number of people moving in the geographical zone ZG according to the model M', and the progressive decrease in ambient brightness, the luminous flux setpoint gradually increases from a zero value to a value of 100% of the nominal flux.

[0064] During the second period T3, this instruction decreases then increases, depending on the number of people moving in the geographical zone ZG according to the model M.

[0065] Finally, during the third period T3, this setpoint decreases from a value of 100% of the nominal flow to a zero value.

[0066] The preceding description clearly explains how the invention makes it possible to achieve the objectives it has set for itself, and in particular by proposing a method modeling a population flow in a geographical area using data obtained from mobile terminals of people present in an estimation area and defining a control law for public lighting from this population flow model. The control law is thus adapted to the movement habits of the population in this geographical area.

[0067] In any event, the invention cannot be limited to the embodiments specifically described in this document, and extends in particular to any equivalent means and to any technically effective combination of these means. In particular, provision may be made to provide other indicators in order to generate the population flow model than those which have been described, and in particular an indicator of the number of people residing in the estimation area or an indicator of the time spent in the estimation area. Provision may also be made for other machine learning algorithms than those described.

Claims

1. Method (1) for controlling street lighting in a predetermined geographical zone (ZG), the method comprising the following steps: • (E11) Providing a set of data (D) relating to the presence of people in an estimation geographical zone (ZE), obtained from mobile devices of these people; • (E2) Determining at least one model (M) of population flow in said predetermined geographical zone from said set of data; • (E5) Defining a control law (LC) for the street lighting of said geographical zone on the basis of said population flow model and (E6) applying said control law to this street lighting; characterized in that the step (E2) of determining the model (M) comprises a sub-step (E21) of generating a plurality of models (M1, M2, MN) of population flow in said predetermined geographical zone (ZG), so-called test models, each test model being generated by a distinct machine learning algorithm from said set of data (D), each test model being associated with an evaluation metric, and a sub-step (E22) of selecting a population flow model (M) from the plurality of test models on the basis of the evaluation metrics of said plurality of test models.

2. Method (1) according to the preceding claim, wherein the provided set of data (D) comprises an estimate (I1, I2, I3) of the number of people entering, present in and leaving the estimation zone (ZE), during a plurality of elementary time intervals of a given overall time interval.

3. Method (1) according to the preceding claim, wherein said estimate (I1, I2, I3) of the number of people entering, present in and leaving the estimation zone (ZE) is obtained from information relating to the position over time of a plurality of mobile devices, this information being determined from data generated by relay antennas (2) of a mobile telephony network located in the estimation zone, to which these mobile devices have connected at a given instant.

4. Method (1) according to any of the preceding claims, characterized in that it comprises a step (E12) of providing at least one additional set of contextual data (DG) relating to said predetermined geographical zone (ZG), and wherein the model (M) of population flow in said predetermined geographical zone is determined from said set of data (D) and the additional set of data.

5. Method (1) according to any of the preceding claims, characterized in that it comprises a step of classifying the data of the set of data (D) among data categories each associated with a geographical sub-zone, and wherein the step (E2) of determining the model (M) comprises, for each geographical sub-zone, a sub-step of determining a sub-model of population flow in this geographical sub-zone from the data classified in the data category associated with this geographical sub-zone.

6. Method (1) according to any of the preceding claims, characterized in that it comprises a step (E41) of providing a prior set of data (DT) relating to the presence of people in the estimation geographical zone (ZE) at a given instant, a step (E4) of predicting a model (M') of the population flow at a future instant relative to the given instant from the population flow model (M) determined during the determination step (E2) and from the prior set of data, the control law (LC) being defined on the basis of the predicted population flow model.

7. Method (1) according to the preceding claim, the method comprising a step of providing at least one piece of information relating to a future state of said predetermined geographical zone (ZG) at said future instant, the step (E4) of predicting a model (M') of the population flow at a future instant being carried out on the basis of the model (M) of the population flow determined during the determination step (E2), the prior set of data (DT) and said piece of information relating to the future state of said predetermined geographical zone.

8. Method (1) according to either of claims 6 or 7, the method comprising a step (E42) of providing a later set of data (DT+1) relating to the presence of people in the estimation geographical zone (ZE) at said future instant, this step being carried out at the future instant or subsequently to this future instant, characterized in that it comprises a step (E43) of correcting said predicted population flow model (M') using said later set of data.

9. Method (1) according to any of the preceding claims, wherein the step of defining a control law (LC) for the street lighting of said geographical zone (ZG) comprises defining at least one period of switching off and / or reducing and / or switching on and / or increasing (T1, T2, T3) the street lighting on the basis of said population flow model (M).

10. Method (1) according to the preceding claim, the method comprising a step (E51) of providing an overall energy consumption setpoint (C), wherein the periods for switching off, reducing, switching on and / or increasing the street lighting (T1, T2, T3) are defined on the basis of the population flow model (M) and such that the energy consumption of the street lighting during all of these periods conforms to said setpoint.

11. Method (1) according to either of claims 9 to 10, wherein the method comprises a step (E52) of providing information relating to the ambient brightness (IL) of the predetermined geographical zone (ZG), wherein the periods for switching off, reducing, switching on and / or increasing the street lighting (T1, T2, T3) are defined on the basis of the population flow model (M) and this ambient brightness information.