EP3005517A1

EP3005517A1 - Estimation of a deleted fluid consumption

Info

Publication number: EP3005517A1
Application number: EP14731310.0A
Authority: EP
Inventors: Benoît GROSSIN; Philippe CHARPENTIER; Leslie Hatton; Eric MATZNER-LOBER
Original assignee: Electricite de France SA
Current assignee: Electricite de France SA
Priority date: 2013-05-24
Filing date: 2014-05-23
Publication date: 2016-04-13
Also published as: JP6151440B2; FR3006075B1; FR3006075A1; US10255553B2; JP2016526369A; US20160132782A1; CA2913250A1; WO2014188140A1

Abstract

The present invention concerns a device (100) for estimating a deleted fluid consumption (E) deleted during a deletion phase (Te), said device (100) comprising: - a collection module (10) configured to collect a) first consumption data (D_CF₁,1, D_CF1,2, D_CF1,n) comprising information related to the fluid consumption of n fluid counters (CF1,1, CF1,2, CF1,n) from a first group (G1), and b) second consumption data (D_CF2,1, D_CF2,2, D_CF2,m) comprising information related to the fluid consumption of m fluid counters (CF2,1, CF2,2, CF2,m) from a second group (G2), and - an analysis computer module (20) configured to calculate, on the basis of said first and second consumption data, weighting coefficients βi that minimise the distance between the fluid consumptions of the first (G1) and second (G2) groups, i being a positive integer between 1 and m.

Description

ESTIMATING DEPRESSED FLUID CONSUMPTION

Technical area

The subject of the present invention relates to the field of the management of fluid consumption; the subject of the present invention relates more particularly to the reduction of fluid consumption.

One of the objectives of the present invention is to accurately estimate the amount of fluid that would have consumed an installation (domestic or industrial) if it had not been erased during a so-called erasure phase.

The present invention thus finds many advantageous applications for example for energy operators by allowing them to manage in real time and in an optimized manner their production of fluid in particular to ensure a balance between supply and demand of fluid.

The present invention also finds other advantageous applications in particular for adjustment operators by allowing them to accurately quantify the fluid consumption erased during an erasure period.

By fluid within the meaning of the present invention, it is to be understood here throughout the present description that any energy source, such as for example electricity, water, or gas or fuel oil, likely to be consumed by equipment of an installation (domestic or industrial) in particular for its operation.

State of the art

Controlling the consumption of fluids has become a daily and growing challenge for both individuals and industry: the reasons for controlling this consumption are both economic (high financial costs) and ecological (pollution , greenhouse gas emissions, natural resources management).

To control this consumption, energy operators have been implementing energy efficiency policies for several years aimed at reducing fluid consumption, especially during periods of peak consumption.

For individuals, this peak in consumption occurs most often in winter between

18 and 20 hours; this is explained in particular by the weather conditions at this time of the year and the traditional domestic uses. Generally, this peak of energy fluid consumption is satisfied by "fast" production means; however, these means are often polluting.

Thus, for example, for the production of electrical energy, combustion turbines are used.

In the field of electrical energy, to limit this increased consumption during such a peak, we have known for several decades the tariff incentive.

Most electricity providers have indeed set specific rates for so-called "hollow" hours and so-called "full" hours: the price of electricity is thus increased over a determined time period in order to reduce or postpone the request.

Other solutions are now put in place to better control this consumption; energy operators have in fact implemented an increased policy on Active Demand Management (also known as "GAD"); this management aims to control and reduce energy fluid consumption in both the residential and industrial markets.

Among these solutions, one of them is to directly control the electrical load of certain equipment.

Thus, certain electrical uses such as heating and / or domestic hot water can be interrupted at times of high demand, for example for a period of two hours (preferably between 18 and 20 hours).

During these hours of power consumption reduction, it is said that the customer "fades", the customer having of course voluntarily subscribed to such an erasure service (often with preferential rates in return).

These means of control of the load are generally operated only a few days per year (15 to 20 days) during the winter and can significantly reduce the fluid consumption and the final bill of the consumer.

It is therefore decisive to estimate precisely this quantity of erased fluid, also called "erasure". This amount of erased fluid therefore corresponds to the difference between the amount of fluid actually consumed and the amount of fluid that would have been consumed if the customer had not cleared; this theoretical amount is also called "baseline".

This estimation of the quantity of fluid erased, or erasure, is all the more strategic since it is now possible to value this erasure: there are indeed adjustment operators who undertake contractually to sell an erasure for example every half hour over a predetermined period of time to allow regulation of fluid consumption, and possibly sell energy to other operators during these periods. peaks to cope with the demand.

Faced with this new problem, many methods have been proposed to calculate this erasure by estimating the "baseline" (also called counterfactual or non-consumption): the accuracy of these different methods of estimating the erasure therefore depends on the accuracy with which the baseline is estimated.

A first approach is based on the observation of the electric power before and after the erasure phase, and proposes a linear interpolation of the actual consumption between the beginning and the end of the erasure.

Such an approach is described in particular in document WO 2008/017754.

This first approach has two main disadvantages.

On the one hand, this approach implicitly assumes that the behavior of the consumer between the beginning and the end of the erasure is linear; such a working hypothesis is not correct and the studies conducted by the Applicant have clearly shown that the implementation of methods based on this approach inevitably leads to an overestimation or underestimation of the "baseline", and therefore of the quantity erased fluid.

On the other hand, it is systematically observed on the consumption curves an anticipation effect before the beginning of the erasure (artificial increase of the consumption) and a rebound effect after the end of the erasure (technical increase from one to one). catch-up effect, especially heating).

Thus, by estimating the "baseline" according to the first approach, the anticipation and rebound effects invariably accentuate the estimation error.

Such approximations to estimate the "baseline" are not acceptable, especially if the estimated erasure in the end gives rise to remuneration.

A second approach is to estimate the "baseline" by the average load curve of a group of customers, also called control group or control group.

According to this approach, control group clients can not fade away.

However, such an approach is not free of defects, because there is inevitably a bias between the control group and the group of so-called erased customers having subscribed to the erasure service. Indeed, the subscription to an erasure service inevitably implies that the customer is concerned by the environment and / or is economical.

This approach does not seek to build a control group comparable to the group of customers who subscribed to the erasure service and therefore having similar behavior in terms of consumption, especially during erasure phases.

The control group is randomly selected from the panel of clients who did not subscribe to the erase service.

On the theoretical side, it is possible to envisage the construction of a control group comparable to the group of customers having subscribed to the erasure service by randomly selecting from this group of erased clients clients for the control group.

Nevertheless, such a solution can not be validly implemented in practice, particularly for reasons of costs and means. Indeed, the customers having subscribed to the erasure service are all solicited and adhere to the erasing process; it is therefore impossible to set aside some of these clients to form a control group.

It is known in the state of the art the document entitled "Navigant Consulting:" Evalulation Report: Home Energy Reports - Plan Year 4 "which describes in detail an approach similar to the above.

Indeed, in this document, recruited customers are randomly assigned to the control group or group of deleted clients.

This approach is coherent from a theoretical point of view; however, in operational mode, this approach is not exploitable.

On the one hand, as mentioned above, it is not possible to randomly allocate a customer who has subscribed to the cancellation offer: we can not prevent a customer who has subscribed to the cancellation offer from not fade for the construction of the control group. Similarly, it is not possible to force a client to fade if he does not want to subscribe to the offer. The results obtained with this approach would necessarily be biased.

On the other hand, such an approach requires too much processing time: at least two seasons are needed. Indeed, to provide relatively stable results with similar numbers, the approach requires measurements on at least two consecutive seasons: a first season to check if the samples measured for each group are similar and a second season to measure the erasures.

Thus, the construction of the control group proposed in this document can only be used as part of an experiment; this approach is unusable in operational mode when it is necessary to dynamically build a control group and determine in real time the erasure, in particular for the readjustment market.

There is currently no state-of-the-art method for building a relevant control group.

Summary and object of the present invention

The object of the present invention is to improve the situation described above.

Thus, one of the objectives of the present invention is to allow the construction of a control group having a behavior comparable to a group of customers having subscribed to an erasure service to estimate accurately and in real time a quantity of consumption of fluid erased during an erasure phase.

Another objective of the present invention is to allow the construction of a flexible and adaptable control group that can be applied to other groups to calculate other erasures.

For this purpose, the object of the present invention relates to a method for estimating a consumption of fluid erased during an erasure phase. Preferably, this estimation is done in real time or a posteriori (that is to say after erasure).

According to the present invention, the estimation method is implemented by computer means, and comprises during a learning phase:

a first collecting step which consists in collecting first consumption data comprising information relating to the fluid consumption of n fluid counters from a first group, n being a positive integer, and

a second collection step of collecting second consumption data comprising information relating to the fluid consumption of m fluid meters from a second group, m being a positive integer.

This data collection is carried out continuously or periodically.

Preferably, m is strictly greater than n. This allows for more accurate results. Advantageously, the n fluid counters of the first group have subscribed to an erase service, and the m fluid meters of the second group have not subscribed to an erase service.

According to the invention, the estimation method then comprises a step of analyzing the first and second consumption data collected.

During this step, weighting coefficients β; (i being a positive integer between 1 and m) are calculated according to these consumption data so as to minimize the distance between the fluid consumptions of the first and second groups.

During the learning phase, the method according to the present invention also comprises a determination step in which a control group is determined from the second consumption data from the m second group of fluid meters, and function of the β weighting coefficients; calculated during the analysis step.

Advantageously, the method according to the present invention comprises, during a determined erasure phase, an estimation step which consists in estimating in real time the consumption of erased fluid by calculating the difference between the fluid consumption of the control group and the average fluid consumption of the n fluid counters of the first group.

Thanks to this succession of technical steps, characteristic of the present invention, it is possible to collect, from a distance and in real time, consumption data coming from fluid meters connected to installations (domestic and / or industrial) and to calculate ( in real time or a posteriori) according to these data of the weights.

The calculation of the weighting coefficients is characteristic of the present invention. Indeed, these coefficients will allow to build, quickly and accurately, a control group comparable to a group of fluid meters having subscribed to an erase service. Thus, according to the invention, the control group is not built randomly, contrary to the state of the art mentioned in the preamble.

According to the invention, the control group is presented as a linear combination of the individual fluid consumption of the fluid meters of the second group approaching the average fluid consumption of the fluid meters of the first group during the learning phase.

It is thus possible thanks to this control group to estimate in real time and with precision the consumption of fluid erased during an erasure phase, this by observing the fluid consumption of the control group.

The present invention thus applies to the estimation of the quantity of fluid erased, during any type of erasure (by direct control of the charge or any other type of erasure such as for example a voluntary and punctual erasure of a customer for example to reduce his bill).

Preferably, the learning phase is prior to the erasure phase.

In an advantageous variant, during the analysis step, the weighting coefficients are calculated so as to solve the following equation:

arg min (distfat) -Σ " _t ()) in which:

P _{2 j} (t) is the individual fluid consumption of the fluid counter j belonging to the second group during the learning phase, and

- is the average fluid consumption of the n fluid counters of the first group during the learning phase.

Preferably, the fluid consumption is an electrical power representative of the electrical power consumed; this fluid consumption can also be representative of a consumed energy.

The present invention provides different embodiments for calculating the β weighting coefficients; performed during the analysis step.

In a first embodiment, the calculation of weighting coefficients β; comprises a sequential selection of the second group fluid meters minimizing the distance between the fluid consumptions of the first and second groups.

Preferably, a value "1" is assigned to the weighting coefficient if the fluid consumption of the fluid counter k of the second group is selected, and a value "0" is assigned otherwise, where k is a positive integer between 1 and m.

This so-called sequential selection algorithm is simple to implement. In addition, this algorithm assigns a weight of "1" to the fluid counter of the second group if it is selected, which makes sense from the operational point of view, especially when m is of size m "n.

In this mode, the step of determining the control group comprises calculating the average fluid consumption of the second group as a function of the weighting coefficients, such a calculation being preferably performed according to the following mathematical formula:

wherein P _{2 j} (t) is the individual fluid consumption of the fluid counter j belonging to the second group during the learning phase.

Preferably, the fluid consumption is an electrical power representative of the electrical power consumed; we also talk about a customer's load curve; this fluid consumption can also be representative of a consumed energy.

In a second embodiment, the calculation of weighting coefficients β; comprises a linear regression, preferably under stress, of the average fluid consumption of the n fluid counters of the first group on the individual fluid consumptions of each of the m fluid meters of the second group.

By linear regression under stress, we mean here a linear regression introducing a constraint on the norm of the coefficients of the model of the type:

|| /? || R ^m

with = (A, ..., A.)

l a standard type Ll, L2 or others /? the vector of the parameters of the m + 1 coefficients (a coefficient for the constant + a coefficient per variable (customers present in the control group)).

The variable here corresponds to the load curve of a customer.

Such a linear regression under stress can be carried out according to several approaches.

According to a first approach, the linear regression under stress is of the Ridge type. According to a second approach, the linear regression under stress is of the type

Lasso.

In an advantageous sub-variant, it is also possible to use a linear regression under stress of the Lasso positive type.

Alternatively, it may be linear regression type PLSPour "Partial

Least Squares "or PCR for" Principal Components Regression ".

The different linear regressions proposed below have the following advantages:

- to have permanently the control group closest to the group of customers erased even if it changes over time (new customers, customer losses) because the estimate of the model is based on a short history.

- To allow a network operator to monitor the erasure in real time and to be able to act in case of imbalance of the electrical system by an increase / decrease in the number of customers erased or by stopping / starting means of production.

- to allow an adjustment operator to respect his erasure contract and to be able to intervene by increasing or decreasing the number of erased customers in the event of non-compliance with the erasure contract.

- to control the realization of the erasures at the pace of 10 minutes or 30 minutes on the valuation markets of erasure.

When the weighting coefficients are calculated according to a linear regression, the step of determining the control group comprises calculating the average fluid consumption of the second group as a function of the weighting coefficients, such a calculation being preferably performed according to the formula following mathematical:

Correlatively, the object of the present invention relates to a computer program which includes instructions adapted for the execution of the steps of the estimation method as described above, this in particular when said computer program is executed by at least one processor.

Such a computer program can use any programming language, and be in the form of a source code, an object code, or a code intermediate between a source code and an object code, such as in a partially compiled form, or in any other desirable form.

Similarly, the subject of the present invention relates to a computer-readable recording medium on which is recorded a computer program comprising instructions for carrying out the steps of the method as described above.

On the one hand, the recording medium can be any entity or device capable of storing the program. For example, the medium may comprise storage means, such as a ROM, for example a CD-ROM or a microelectronic circuit-type ROM, or a magnetic recording means, for example a diskette of the type " floppy says "or a hard drive.

On the other hand, this recording medium can also be a transmissible medium such as an electrical or optical signal, such a signal can be conveyed via an electric or optical cable, by conventional radio or radio or by self-directed laser beam or by other ways. The computer program according to the invention can in particular be downloaded to an Internet type network.

Alternatively, the recording medium may be an integrated circuit in which the computer program is incorporated, the integrated circuit being adapted to execute or to be used in the execution of the method in question.

The subject of the present invention also relates to a device for estimating a consumption of fluid erased during an erasure phase.

For this purpose, the estimation device according to the present invention comprises computer means configured for the implementation of the steps of the estimation method as described above.

More specifically, the estimation device includes a collection module configured to collect:

a) first consumption data that includes fluid consumption information of n fluid counters from a first group, n being a positive integer, said n fluid counters of the first group having subscribed to a service of erasure, and

b) second consumption data which comprise information relating to the fluid consumption of m fluid meters from a second group, m being a positive integer, said m second group of fluid meters having not subscribed to a erasure service. Advantageously, the estimation device further comprises:

an analysis computer module, or calculator, which is configured to calculate, according to the first and second consumption data, weighting coefficients β; minimizing the distance between the fluid consumptions of the first and second groups, i being a positive integer between 1 and m,

a processor configured to determine a control group from the second consumption data collected from the m fluid meters of the second group, and as a function of said weighting coefficients β; calculated by the analysis computer module, and

an estimation computer module configured to estimate the consumption of fluid erased during an erasure phase by calculating the difference between the fluid consumption of the control group and the average fluid consumption of the n first group of fluid meters .

Thus, by its various functional and structural aspects described above, the present invention makes it possible to estimate in real time (or a posteriori) and accurately the amount of fluid consumption erased during an erasure phase, the present invention applying for any type of erasure (erasure by direct control of the load or voluntary deletion of the customer).

Brief description of the attached figures

Other characteristics and advantages of the present invention will emerge from the description below, with reference to FIGS. 1 to 3, which illustrate an exemplary embodiment without any limiting character and on which:

FIG. 1 schematically represents an estimation device according to an exemplary embodiment of the present invention;

FIG. 2 represents a flowchart comprising the different steps implemented according to an exemplary embodiment of the method of the invention;

FIG. 3a is a graph showing the variation over one day of the average fluid consumption of the first group of fluid meters and that of the control group, when the erasure is effective between 18h and 20h; and

- Figure 3b is a graph representing an estimate of the erasure between 18h and 20h when the erase service is effective.

Detailed description of an embodiment of the invention A method for estimating an erased fluid consumption E and associated device 100 will now be described in the following with reference to Figures 1 to 3.

As a reminder, during an erase phase Te, it is not possible to directly measure the erased fluid consumption E (illustrated in FIGS. 3a and 3b) of a first group Gl of n fluid counters CF _{1; 1} , CF _{1> 2} , CF _{1> n} having subscribed to an erase service.

Indeed, such an erasure value E is not measurable since it is purely fictitious and theoretical.

As explained above, estimating this consumption of erased fluid E represents a strategic issue for the actors of the energy activity.

The approaches proposed so far are not satisfactory: they remain too approximate and present estimation errors that are not acceptable, particularly given the economic and financial stakes related to erasure (which now represents a market in its own right). ).

Improving the estimation of the effected fluid consumption E is therefore one of the objectives of the present invention.

Thus, in the example described here, and as illustrated in FIG. 1, the estimation device 100 according to the present invention comprises a collection module 10, for example a "Smartmeter" automatic remote reading equipment, which during a determined learning phase Ta simultaneously collects consumption data during a first SI and a second S2 collection stages.

More specifically, in the example described here, and as illustrated in FIG. 1, the collection module 10 collects, continuously or periodically, first data D_CF _{1; 1} , D_CF _{1 2} , D_CF _ln and second data D_CF _{2 1} , D_CF _{2 2} , D_CF _{2 m} of consumption. A collection periodically performed in a step of 15 or 30 minutes during the learning phase Ta gives quite satisfactory results.

In the example described here, the first data O_ _{\, \} , D_CF _{1 2} , D_CF _ln comprise information relating to the individual fluid consumption P _1: (ί), P _{1 2} (t), P _ln (t) n fluid counters CFi _j , CF _{1> 2} , CF _{1> n} from the first group G1; and the second consumption data D_CF _{2 1} , D_CF _{2 2} , D_CF _{2 m} comprise information relating to the individual fluid consumption P _{2 1} (t), P _{2 2} (t), P _{2 m} (t) of m fluid meters CF _{2> 1} , CF _2j2 , CF _2jm from the second group G2, the m fluid meters CF _{2> 1} , CF _2j2 CF _2jm not having subscribed to the erasure service.

As illustrated in FIG. 1, each of the above fluid meters CF _{1; 1} , CF _{1 2} , CF _1n and CF _{2> 1} , CF _2j ₂ , CF _{2> m} are respectively connected to installations I _{1; 1} , 1 _{1 2} , 1 _{1> n} and I _{2> 1} , I _2j2 , h, m, and are configured to count a quantity of fluid consumed by all the equipment respectively of the installation I _{1; 1} , I _{1> 2} , Ii, _n and I _{2> 1} , 1 _{2> 2} , I _{2> m} .

In the example described here, m is strictly greater than n (here, m "n); this allows for finer results.

The concept underlying the present invention is therefore here to construct a GC control group from this second group G2 of fluid meters CF _{2> 1} , CF _2j ₂ , CF _{2> m} , and to ensure that this GC control group is closest to the behavior of the first group Gl in terms of consumption.

Thus, to coherently construct such a group GC and accurately estimate the consumption of erased fluid E, the present invention aims to select in the second group G2 the fluid meters so that the distance between their average consumption P ₂ (t ) and the average consumption i> (t) of the first group G1 is minimal.

The search for this minimum distance requires the calculation of weighting coefficients β _ί , where i is a positive integer between 1 and m.

For this purpose, in the example described, and as illustrated in FIG. 1, the estimation device 100 comprises an analysis computer module 20 which, during an analysis step S3, analyzes the first and second data of the analysis. consumption collected, and calculates on the basis of these data the weighting coefficients β; in order to minimize the distance between the fluid consumptions of the first G1 and second G2 groups.

In the example described here, the analysis computer module 20 is configured so that the weighting coefficients β; be calculated to solve the following equation (1):

arg min (eUstfa (-Σ ^™ _l ()) in which:

P _{2 j} (t) is the individual fluid consumption of the fluid counter j belonging to the second group G2 during the learning phase Ta, and is the average fluid consumption of the n fluid counters CF _{1; 1} ,

CF _{1> 2} , CF _ln of the first group G1 during the learning phase Ta.

In the example described here, the average fluid consumption i> (t) is calculated as follows:

where P _{1 j} (t) is the individual fluid consumption of the fluid counter j belonging to the first group G1 during the learning phase Ta.

In the example described here, several approaches described above are provided for programming the analysis module 20 in order to solve the equation (1) and calculate the weighting coefficients β _; .

In the example described here, the device 100 further comprises a determination module (or processor) 30 configured to use, during a determination step 40, the weighting coefficients β; calculated by the module 20 and determine the GC control group from the second group G2, this GC control group being a linear combination of the individual fluid consumptions P _{2 j} (t) fluid meters CF _{2 1} , CF _{2 2} CF _{2 m} of the second group G2 best approximating the average fluid consumption i> (t) of the n fluid counters CF, i, CF _{1 2} , CF _ln of the first group G 1 during the learning phase Ta:

Calculate the β weighting coefficients; is characteristic of the present invention.

To calculate these coefficients, a first approach already mentioned above provides for sequential selection of the fluid counters CF _{2 1} , CF _{2 2} , CF _{2 m} of the second group G 2.

The number of GC control groups that can be constituted with m fluid meters of weight "0" or "1" is of complexity 2 ^m-1 . In order to reduce the number of possible subgroups, the analysis module 20 seeks to reduce the distance - ^m β _] Ρ ₂ , (t)) by sequentially selecting, Vi = 1, ..., /, the individual consumption P _{2 k} of the fluid meters CF _{2> 1} , CF _{2> 2} , CF _{2> m} of the second group G2. In this variant, the algorithm implemented in the analysis module 20 takes the following form:

• Initialization at k = 1:

S _u = dist (ï> (t) - P _2ii (t))

i, = arg min { _u }

i = l, ..., m

(= ^ (S, = dist ^ (t) - P _{2 1)} (t))

• Loop on k = 2, ..., m:

We thus obtain the set of individuals {¾} cz ... <z {¾, ..., i _m } and the respective distances S _l , ..., S _m , and we select the set of individuals { ¾, ..., ¾} minimizing the distance S _k .

According to this variant, the control group GC is obtained by the processor 30 which determines this group GC during the step S4 according to the following mathematical formula:

Other more efficient approaches are also part of the present invention for calculating the weighting coefficients.

As stated above, it is desirable that the number of observations be limited.

The explanatory variables here are the customer load curves available to constitute the control group: the number of explanatory variables m can therefore be counted in the hundreds of thousands. Therefore, we have m "n.

In addition, customers can be correlated with each other; therefore, even if m <n, then the family of explanatory variables may be linked, which is not possible with classical linear regressions as proposed in the document "Navigant Consulting:" Evalulation Report: Home Energy Reports - Year 4 Plan ".

To solve this problem, the present invention provides an approach in which the matrix of explanatory variables is of full rank.

Two large families of methods allow it:

- Linear regressions under Lasso, Ridge, Elastic net constraints and their derivatives.

It is a linear regression method introducing a constraint on the standard of the coefficients of the type model:

|| /? || R ^m

with = (A, ..., A.)

l a standard type Ll, L2 or others

β the vector of the parameters of the m + 1 coefficients (a coefficient for the constant + a coefficient per variable).

- The regressions on the principal components and the regressions of the PLS type whose objective is to orthogonalize and reduce the space of the variables. The so-called Ridge and Lasso approaches will be explained in more detail in the following.

According to the so-called Ridge approach, solving equation (1) above consists in performing a linear regression of the average fluid consumption of the first group G1 on the individual consumptions of the second group G2 using the norm L2.

In the example described here, this resolution of equation (1) therefore amounts to solving the minimization below:

arg min P, (0 -Σ ^m (0 with a constraint on the norm L2 of the vector β: λ, the parameter λ being selected by cross-validation.

Alternatively, a regression algorithm according to the Lasso approach is implemented in the analysis computer module 20 to solve equation (1).

According to this approach, solving equation (1) amounts to solving the minimization below:

arg min with a constraint on the norm L1 of the vector β '· , the parameter τ being selected by cross-validation.

The constraint on the standard L1 has the advantage of selecting in a more parsimonious manner the fluid counters (and therefore the customers) and thus reduce the bias between the GC control group and the erased group G1.

In other words, with this constraint on the norm L1 of the vector β, more clients of the control group GC will have a weighting coefficient (or weight) equal to "0".

It is also possible to add an additional constraint on the weights by forcing the selection of individuals whose coefficients are positive. We speak of linear regression under Lasso positive stress.

According to these different approaches using a calculation of the weighting coefficients by a linear regression, the control group GC is calculated by the processor 30, during the determination step 40, according to the following formula Vt = 1, ..., / : Â (= Σ ^ A, (

In the example described here, once this GC group is determined, the estimating computer module 40 or estimator can, during an estimation step S5, estimate in real time during an erase phase Te the erased fluid consumption E by calculating the difference between the fluid consumption of the GC control group and the average fluid consumption of the n fluid counters CF _{1; 1} , CF _{1> 2} , CF _{1> n} of the first group G 1.

The Applicant during the various tests carried out has demonstrated that the Lasso approach allows to obtain particularly efficient results, with an error rate of less than 8% against an error rate of more than 10% with the proposed approaches. in the state of the art.

Such results are made possible by constructing the control group from the intraday load curves available prior to the event day.

This principle has the advantage of capturing the effects of constant observable variables over time (location, type of housing, ...), the effects of observable variables fluctuating over time (weather) as well as unobservable variables (political opinion, ecological sensitivity, management profile, etc.). Moreover, in addition to a finer estimation of the erasure, the present invention makes it possible to avoid the effects of overestimation induced by the anticipation and rebound effects caused in the consumer before and after the erasure.

Finally, the present invention differs from other approaches of the state of the art in allowing real-time estimation that adapts over time depending on the behavior of customers, fluid consumers, and the arrival of new ones. customers who are members of the erasure service.

The built control group can also be applied to other groups to calculate other erasures.

The present invention finds a particularly advantageous application in the estimation of the consumption of erased fluid with erasures by direct control of the load, for example by the operator. Obviously, the present invention applies to all types of erasure, such as for example a voluntary and punctual deletion of a customer for example during a peak rate.

In the example described here, the set of technical functionalities described above for each of the physical entities of the device 100 is here controlled by a computer program PG which is recorded on a recording medium CI.

It should be observed that this detailed description relates to a particular embodiment of the present invention, but in no case this description is of any nature limiting to the subject of the invention; on the contrary, its purpose is to remove any imprecision or misinterpretation of the claims that follow.

Claims

1. Method for estimating an erased fluid consumption (E) during an erasure phase (Te), said method implemented by computer means comprising: during a learning phase (Ta):

a first collection step (SI) of collecting first consumption data (D_CF _{1; 1} , D_CF _{1 2} , D_CF _ln ) comprising information relating to the fluid consumption of n fluid counters (CF _{1; 1} , CF _{1 2} , CF _ln ) derived from a first group (Gl), n being a positive integer, said n fluid counters (CF _{1; 1} , CF _{1> 2} , CF _{1> n} ) of the first group (G 1) having subscribed to an erasure service,

a second collection step (S2) of collecting second consumption data (D_CF _{2 1} , D_CF ₂₂ , D_CF _{2 m} ) comprising information relating to the fluid consumption of m fluid meters (CF _{2> 1} , CF _{2> 2} , CF _{2> m} ) from a second group (G2), m being a positive integer, said m fluid meters (CF _{2> 1} , CF _{2> 2} , CF _{2> m} ) of the second group ( G2) who did not subscribe to an erase service,

an analysis step (S3) of the first and second collected consumption data (D_CF _U , D_CF _{1> 2} , D_CF _{1> n} , D_CF _{2> 1} , D_CF _{2> 2} , D_CF _{2> m} ) during which β weighting coefficients; are calculated according to these first and second consumption data so as to minimize the distance between the fluid consumptions of the first (Gl) and second (G2) groups, i being a positive integer between 1 and m, and

a determination step (S4) during which a control group (Gc) is determined from the second collected consumption data (D_CF _{2> 1} , D_CF _{2 2} , D_CF _{2 m} ) resulting from the m fluid meters ( CF _{2 1} , CF _{2 2} , CF _{2 m} ) of the second group (G2), and as a function of the weighting coefficients β; calculated during the analysis step (S3), and

during the erasure phase (Te):

an estimation step (S5) of estimating the effected fluid consumption (E) by calculating the difference between the fluid consumption of said control group (GC) and the average fluid consumption of the n fluid counters (CFi, CF _{1> 2} , CF _{1> n} ) of the first group (Gl).

2. Method according to claim 1, characterized in that, during the analysis step (S3), the weighting coefficients are calculated so as to solve the following equation:

arg min (eUstfa (-Σ ^™ _l ()) in which:

P _{2 j} (t) is the individual fluid consumption of the fluid counter j belonging to the second group (G2) during the learning phase (Ta), and

- is the average fluid consumption of the n fluid counters (CF _{1; 1} , CF _{1> 2} , CF _{1> n} ) of the first group (Gl) during the learning phase (Ta).

3. Method according to claim 1 or 2, characterized in that m is strictly greater than n.

4. Method according to any one of claims 1 to 3, characterized in that, during the analysis step (S3), the calculation of weighting coefficients β; includes a sequential selection of the second group fluid counters (G2) minimizing the distance between the fluid consumptions of the first (G1) and second (G2) groups.

5. Method according to claim 4, characterized in that, during the analysis step (S3), a value "1" is assigned to the weighting coefficient if the individual fluid consumption of the fluid counter k of the second group (G2) is selected, and a value "0" is assigned otherwise, where k is a positive integer between 1 and m.

6. Method according to any one of claims 4 or 5, characterized in that the determination step (S4) of the control group (GC) comprises the calculation of the average fluid consumption ₂ (i) of the second group ( G2) according to the weighting coefficients, such a calculation being made according to the following mathematical formula:

wherein P _{2 j} (t) is the individual fluid consumption of the fluid counter j belonging to the second group (G2) during the learning phase (Ta).

7. Method according to any one of claims 1 to 3, characterized in that, during the analysis step (S3), the calculation of weighting coefficients β; comprises a linear regression of the average fluid consumption of the n fluid counters (CF _{1; 1} , CF _{1 2} , CF _ln ) of the first group (Gl) on the individual fluid consumptions of each of the m fluid meters (CF _{2 > 1} , CF _{2> 2} , CF _{2> m} ) of the second group (G2).

8. The method of claim 7, characterized in that the linear regression is a linear regression under stress.

9. The method of claim 7 or 8, characterized in that the linear regression is a linear regression of the Ridge type.

10. The method of claim 7 or 8, characterized in that the linear regression is a linear regression under constraint of the Lasso type.

11. The method of claim 7 or 8, characterized in that the linear regression is a stress linear regression of the net Elastic type.

12. Method according to claim 7, characterized in that the linear regression is of the PLS or PCR type.

13. Method according to any one of claims 7 to 12, characterized in that the determination step (S4) of the control group (Gc) comprises the calculation of the average fluid consumption ₂ (i) of the second group ( G2) according to the weighting coefficients, such a calculation being made according to the following mathematical formula:

14. Method according to any one of the preceding claims, characterized in that the fluid consumption is an electrical power representative of the electrical power consumed.

15. Computer program (PG) comprising instructions adapted for performing the steps of the estimation method according to any one of claims 1 to 14 when said computer program (PG) is executed by at least one processor .

16. Computer-readable recording medium (CI) on which a computer program (PG) is recorded including instructions for performing the steps of the estimation method according to any one of claims 1 to 14.

17. Apparatus for estimating (100) an erased fluid consumption (E) during an erasure phase (Te), said device (100) comprising:

a collection module (10) configured to collect:

a) first consumption data (D_CF _{1; 1} , D_CF _{1 2} , D_CF _ln ) comprising information relating to the fluid consumption of n fluid counters (CF _{1; 1} , CF _{1 2} , CF _ln ) derived from a first group (Gl), n being a positive integer, said n fluid counters (CF _{1; 1} , CF _{1 2} , CF _ln ) of the first group (Gl) having subscribed to an erase service, and

b) second consumption data (D_CF _{2> 1} , D_CF _{2> 2} , D_CF _{2 m} ) comprising information relating to the fluid consumption of m fluid meters (CF _{2 1} , CF _{2 2} , CF _{2 m} ) derived from a second group (G2), m being a positive integer, said m fluid meters (CF _{2 1} , CF _{2 2} , CF _{2 m} ) of the second group (G 2) not having subscribed to an erasing service , and

an analysis computer module (20) configured to calculate as a function of the first and second consumption data (D_CFi _1; D_CF _{1 2} , D_CF _ln , D_CF _{2 1} , D_CF _{2 2} , D_CF _{2 m} ) of the weighting coefficients β; ; minimizing the distance between the fluid consumptions of the first (Gl) and second (G2) groups, i being a positive integer between 1 and m,

a processor (30) configured to determine a control group (Gc) from the second collected consumption data (D_CF _{2 1} , D_CF _{2 2} , D_CF _{2 m} ) coming from the m fluid meters (CF _{2> 1} , CF _{2> 2} , CF _{2> m} ) of the second group (G2), and according to said weighting coefficients β; calculated by the analysis computer module (20), and

an estimation computer module (40) configured to estimate the effected fluid consumption (E) during an erasure phase (Te) by calculating the difference between the fluid consumption of said control group (GC) and the average fluid consumption of the n fluid counters (CF _{1; 1} , CF _{1> 2} , CF _{1> n} ) of the first group (Gl).

18. Device (100) according to claim 17, characterized in that it further comprises computer means configured for the implementation of the steps of the estimation method according to any one of claims 1 to 14.