METHOD FOR DETECTING AN ANOMALIA IN THE CONSUMPTION OF A LIQUID OF INTEREST BY A CONSUMPTION SYSTEM

Description

TECHNICAL FIELD

[0001] The field of the invention is that of the consumption of a fluid of interest by a consuming system, for example the consumption of liquid water by a house, a building, a school, a hospital, etc... The invention relates more specifically to the monitoring and detection of an anomaly in the consumption of the fluid of interest by the consuming system. PREVIOUS STATE OF THE ART

[0002] A system consuming a fluid of interest can be a house, an apartment building, a school, a hospital, a factory, a corporate hotel, etc. It comprises several devices that consume the fluid of interest, connected by a fluid circuit to a source. In the case of a factory, the fluid of interest can be a liquid such as water, hydrogen, oxygen, nitrogen, etc. In the case of a house, the fluid of interest can be liquid water, and the consuming devices can include a dishwasher, a washing machine, a shower, a toilet with a flush system, and the various taps (bathroom, kitchen, etc.).

[0003] US documents 2018 / 143056 A1, WO 2007 / 059592 A1, and CN 105 185 051 A describe examples of methods for detecting an anomaly in fluid consumption. However, there is a need for a method to effectively detect the presence of an anomaly in the consumption of the fluid of interest, such as an intermittent or continuous leak, or a valve that is not properly closed. DESCRIPTION OF THE INVENTION

[0004] The invention aims to provide a method that allows for the efficient and simple detection of an anomaly in the consumption of a fluid of interest by a consuming system, while taking into account any possible evolution of the consumption of the fluid of interest by the consuming system, from one monitoring period to another.

[0005] To this end, the object of the invention is a method for detecting an anomaly in the consumption of a fluid of interest by a consuming system. The consuming system comprises: at least one device consuming the fluid of interest, connected to a source by a fluid circuit; a measuring device, connected to the fluid circuit between the source and the consuming device, adapted to measure and transmit to a processing unit a measurement signal Sm representative of a variation, during a predefined monitoring period Δt s, of an average flow rate D of the fluid of interest and of a flow duration Δt e|D associated with the average flow rate D; the processing unit, adapted to determine from the measurement signal Sm, a consumption profile corresponding to a variation as a function of the average flow rate D of a flow indicator representative of a proportion %Δt s of a cumulative flow duration Δt ec|D of the fluid of interest during the monitoring period.

[0006] The detection process involves the following steps: a) provide a database containing a plurality of consumption profiles previously identified as being without consumption anomalies; b) determine an anomaly threshold Sa (ic) , for at least one so-called reference value of average flow rate, from the consumption profiles in the database; c) acquire, by the measuring device, a measurement signal Sm (ic) called current, during a monitoring duration Δt s(ic); d) determine, by the processing unit, a consumption profile called current from the measurement signal Sm (ic) current;e) Compare the current consumption profile to the anomaly threshold Sa(ic): if the current consumption profile shows, for at least one reference value of the average flow rate, a value below the anomaly threshold Sa(ic): identify the absence of a consumption anomaly and add the current consumption profile to the database, then repeat steps b) to e) for a subsequent monitoring period Δt s(ic+1); if the current consumption profile shows, for at least one reference value of the average flow rate, a value at least above the anomaly threshold Sa(ic): identify the presence of a consumption anomaly and do not add the current consumption profile to the database, then repeat steps c) to e) for a subsequent monitoring period Δt s(ic+1).

[0007] Some preferred but not exhaustive aspects of this detection method are as follows.

[0008] Step a) of providing the database may include the following steps: a1) acquire, by the measuring device, a measurement signal Sm (n) during a monitoring time Δt s(n); a2) determine, by the processing unit, a consumption profile from the measurement signal Sm (n), identified as being without consumption anomaly, and add the determined consumption profile to the database; repeat steps a1) and a2) until the database contains a number N of consumption profiles at least equal to a predefined minimum number N min.

[0009] The step of adding the current consumption profile to the database may involve a comparison of the number N of consumption profiles in the database, and if this number N is greater than a predefined maximum number N max, a deletion of the consumption profile associated with the oldest monitoring duration.

[0010] The step of determining the consumption profile may include determining a variation, as a function of the average flow rate, of the cumulative flow time by adding the values ​​of the flow times associated with the same average flow rate value.

[0011] Each consumption profile can be determined by means of the following steps: determine a distribution function FD(d) from the variation of the cumulative flow time as a function of the mean flow rate, the distribution function indicating a probability P(D≤d) of obtaining a value of the mean flow rate D less than or equal to a value d; define the consumption profile as being equal to 1-FD(d) or to FD(d).

[0012] Alternatively, each consumption profile can be determined by calculating a ratio of the cumulative flow time to the monitoring time, based on the average flow rate.

[0013] The step of determining the anomaly threshold may involve calculating an average or median of the values ​​of the consumption profiles in the database, for at least one predefined reference value of the average flow rate.

[0014] The step of determining the anomaly threshold may also include a calculation of a standard deviation associated with the values ​​of the consumption profiles in the database, for at least one predefined reference value of the average flow rate.

[0015] The measuring device may include a flow meter or a pulse counter.

[0016] The fluid of interest can be a liquid or a gas, chosen from water, hydrogen, oxygen, nitrogen and helium.

[0017] The consumer system can be a house, a building with several apartments, a factory, a hospital, a school, a campsite.

[0018] The consumer system may include a plurality of consumer devices, the measuring device being connected to the fluidic circuit between the source and the consumer devices. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Other aspects, objectives, advantages, and features of the invention will become clearer upon reading the following detailed description of preferred embodiments thereof, given by way of non-limiting example, and made with reference to the accompanying drawings in which: there figure 1 is a schematic and partial view of a consuming system, comprising devices consuming the fluid of interest, a measuring device, and a processing unit enabling the detection of the presence or absence of a consumption anomaly; the figure 2 is a flowchart of a method for detecting an anomaly in the consumption of the fluid of interest by the consuming system, according to a first embodiment; the figure 3A illustrates an example of consumption profiles %Δt s|D(ic) (d) derived from measurement signals acquired by a flowmeter water meter; the figure 3B illustrates three types of consumption profiles %Δt s|D(ic) (d) from measurement signals acquired by a pulse water meter; the figure 3C illustrates another example of consumption profiles %Δt s|D(ic) (d) derived from measurement signals acquired by a pulse water meter; the figure 4 is a flowchart of a method for detecting an anomaly in the consumption of the fluid of interest by the consuming system, according to a second embodiment; the figure 5 illustrates an example of changes in consumption profiles %Δt s|D(ic) (D) from measurement signals acquired by the measurement device. DETAILED DESCRIPTION OF SPECIFIC METHODS OF IMPLEMENTATION

[0020] In the figures and throughout the description, the same reference numerals represent identical or similar elements. Furthermore, the various elements are not drawn to scale to ensure clarity. Unless otherwise indicated, the terms "approximately," "about," and "in the order of" mean within 10%, and preferably within 5%. Moreover, the terms "between ... and ..." and equivalent mean that the limits are inclusive, unless otherwise stated.

[0021] There figure 1 is a schematic and partial view of a consumer system 1 according to one embodiment.

[0022] The consumer system 1 is a structure that consumes a fluid of interest through at least one consumer device 2 (and preferably several consumer devices 2), and whose consumption must be monitored to detect any anomalies. Consumption is understood to mean that the consumer system 1 receives the fluid of interest from a source 3 and uses it (the "consumes" it) for various purposes, which may be personal and / or professional. Such a consumer system 1 could be, for example, a house, an apartment, a multi-unit building, a school, a factory, a hospital, a campsite, or other similar structures. In general, it is any type of structure or assembly for personal and / or professional use, comprising at least one device 2 that consumes the fluid of interest, and preferably a plurality of consumer devices 2.

[0023] The fluid of interest can be a liquid or a gas, such as water, hydrogen, oxygen, nitrogen, helium, etc. Thus, consumer system 1 could be a factory that consumes, for example, hydrogen, oxygen, or nitrogen, in liquid or gaseous form. It could also be an educational building (school) that uses liquid water. In the following description, consumer system 1 is a dwelling, and the fluid of interest is liquid water.

[0024] The consumer system 1 includes at least one consumer device 2 that ensures the actual consumption of the fluid of interest, and preferably a plurality of consumer devices 2. In our example of a dwelling, the consumer devices 2 could be toilet, kitchen, and bathroom taps, as well as household appliances such as a dishwasher and a washing machine. They could also include flush toilets. The consumer system 1 may also include consumer devices whose consumption is not monitored by the measuring device 4 and the treatment unit 5.

[0025] The consumer system 1 also includes a fluid circuit that distributes the fluid of interest from the source 3 to the consumer device(s) 2. This consists of distribution pipes, possibly equipped with valves. The source 3 of the fluid of interest can be a supply network, for example, a city supply, a reservoir, or something else.

[0026] The consumer system 1 includes a measuring device 4, connected to the fluid circuit and located between the source 3 and the consumer device(s) 2. This could be, for example, a flow meter, a pulse water meter, or another type. It is adapted to measure and provide a processing unit 5 with a measurement signal (Sm(t (m))) 1≤m≤M representing an average flow rate D(t (m)) and a flow duration Δt e|D (t (m) ) associated with the average flow rate D, over a predefined duration Δt s called the monitoring duration. It can transmit this measurement signal Sm in real time, at regular intervals, or at the end of the monitoring duration Δt s. Here, t (m) denotes the measurement instant to which the average flow rate D(t (m) ) and the flow duration Δt e|D (t (m) ) are associated. This can be a time instant during the monitoring duration Δt s , or a simple increment corresponding to a measurement event.Thus, the measurement signal (Sm(t (m) )) 1≤m≤M has M measurement instants, where the increment m goes from 1 to M. The measured flow rate D(t (m) ) is said to be average in the sense that it is the average of the effective flow rate over the associated flow time Δt e|D (t (m) ), at the measurement instant t (m) .

[0027] The monitoring duration Δt s is characteristic of the time scale associated with a consumption profile of the consumer system 1. It can be one to several hours, one or more days, or even weeks or months. In our example of a house, the monitoring duration Δt s corresponds to one day.

[0028] For example, if the measuring device 4 is a turbine-driven water meter, the measurement signal (Sm(t (m) )) 1≤m≤M is a matrix formed, for example, by the measurement time t (m), the average flow rate D(t (m) ) measured or, equivalently, the volume of fluid of interest that has flowed, and the corresponding flow duration Δt(t (m) ). The turbine can be the flow meter itself, or it can be a power supply for transmitting the measurement signal.

[0029] Alternatively, the measuring device 4 can be a pulse water meter. In this example, the measurement signal (Sm(t (m) )) 1≤m≤M can be a matrix containing the instant or event at which a pulse is emitted. This pulse is emitted when a predefined volume of the fluid of interest has flowed (e.g., 1 pulse per liter). The processing unit 5 can then deduce the values ​​of the average flow rate D(t (m) ), as well as the flow duration Δt(t (m) ) associated with each value of the average flow rate.

[0030] The measuring device 4 can transmit the measurement signal Sm at the end of the monitoring time Δt s, or even can transmit the measurement signal Sm during the monitoring time Δt s, the measurement signal then being acquired.

[0031] The consumer system 1 also includes a processing unit 5, connected to the measuring device 4. It comprises at least one computer and at least one memory. It enables the implementation of the processing operations for the anomaly detection method described later. The computer includes a programmable processor capable of executing instructions stored on a data storage medium. The memory contains instructions for implementing the detection method. It is also adapted to store information received by the measuring device 4, and here includes a database (DB) of consumption profiles of the consumer system 1 identified as having no consumption anomalies. It may be fully or partially integrated into the measuring device 4 or be located remotely from it, and the transmission of information from the measuring device to the processing unit 5 may be wired or wireless.

[0032] The processing unit 5 is particularly suited to determining a consumption profile from the measurement signal Sm received by the measuring device 4. The consumption profile corresponds to a variation, as a function of the average flow rate D, of a flow indicator representing a proportion %Δt s|D of a cumulative flow time Δt ec|D of the fluid of interest during the monitoring period Δt s. In other words, for each value of the average flow rate D measured by the measuring device 4, the corresponding cumulative flow time Δt ec|D is determined, which is the sum of the flow times Δt e|D measured and associated with the same value of the average flow rate D (or the same class of values ​​in the case of a histogram). The flow indicator is thus representative of the proportion of the cumulative flow time Δt ec|D during the monitoring period Δt s, for each value (or class) of the average flow rate.Subsequently, we refer to a value of the average flow rate, but this also covers a class of average flow rate values.

[0033] As detailed later, the detection process comprises a preparation phase, followed by a monitoring and detection phase. During the preparation phase, a database is created, consisting of consumption profiles of consumer system 1 identified as having no consumption anomalies. This database is used to subsequently determine an anomaly threshold Sa and is intended to be updated regularly so that the anomaly threshold adapts to changes in the consumption of consumer system 1 over the monitoring periods. Then, during the monitoring and analysis phase, a new consumption profile, referred to as the current profile, is determined from a measurement signal and then compared to the anomaly threshold Sa previously determined from the database DB to determine whether or not the consumption profile exhibits a consumption anomaly.In the event that there are no anomalies, the current consumption profile is added to the database and the anomaly threshold Sa is then updated.

[0034] There figure 2 This is a flowchart of an anomaly detection process according to a first embodiment. In the following description, the consumer system 1 is a family home and the fluid of interest is liquid water.

[0035] As previously stated, a consumption profile corresponds to a variation, as a function of the average flow rate D measured by the measuring device 4, of a flow indicator representative of a proportion %Δt s|D of a cumulative water flow duration Δt ec|D relative to the monitoring duration Δt s. It is determined by the processing unit 5 from the measurement signal Sm acquired by the measuring device 4 during the monitoring duration Δt s, here during one day.

[0036] In this first embodiment, the consumption profile is determined from the distribution function FD(d) (also called the cumulative distribution function, or Cumulative Distribution Function, (in English). This can therefore be either the cumulative distribution function FD(d) itself, or the complementary function %Δt s|D(d) = 1-FD(d). Subsequently, the consumption profile is the function %Δt s|D(d).

[0037] Note here that the cumulative distribution function FD(d) corresponds to the probability of obtaining a mean flow rate value less than or equal to d: FD(d) = P(D≤d). This probability P(D≤d) is equal to the proportion %Δt s|D of the cumulative flow duration Δt ec|D during the monitoring period Δt s (here, the day) for which the mean flow rate D is zero or at most equal to d. The cumulative distribution function FD(d) is between 0% and 100%, and the mean flow rate D varies between zero and the maximum flow rate measured by the measurement signal Sm.

[0038] Thus, for example, a value of FD(Ds) of 80% for a threshold average flow rate Ds of 0.1 L / min corresponds to 80% of the monitoring time Δts for which the average flow rate is zero (no flow) or at most equal to 0.1 L / min. It follows that the function %Δts|D(Ds) has a value of 20%, which corresponds to the proportion of the monitoring time Δts for which the average flow rate is strictly greater than 0.1 L / min (flow). Preparation phase 110: Provide the database database

[0039] The preparation phase 110 therefore consists of creating a database (DB) of consumption profiles identified as having no consumption anomalies. It includes step 111 of acquiring a measurement signal, step 112 of determining the corresponding consumption profile and creating the database (BD), and then step 113 of verifying that the database (BD) contains the minimum number (N min) of consumption profiles.

[0040] The database consists of a plurality N of consumption profiles {%Δt s|D(n) (d)} 1≤n≤N of the consuming system 1, where %Δt s|D(n) (d) = 1-FD (d), and where N is at least equal to a minimum value N min >1. N min can be equal to at least 5, 10, or 20. As an example, N min can be equal to 7, so that the database contains consumption profiles %Δt s|D(n) (d) associated with successive monitoring durations Δt s(n), thus covering at least one week. The consumption profiles in the database are identified as having no consumption anomalies (and with consumption of the fluid of interest), and will therefore allow for the subsequent determination of an anomaly threshold Sa.

[0041] The consumption profile database can be created in various ways. The database can be obtained through numerical simulation based on a physical model of the consumer system 1. Alternatively, as described in detail below, it can be obtained during monitoring of the consumer system 1 by the measuring device 4 and the processing unit 5.

[0042] During a step 111, the measuring device 4 acquires the measurement signal Sm (1) during a first monitoring duration Δt s(1) (ic=1). The measurement signal Sm (1) = (Sm (1) (t (m) )) 1≤m≤M is therefore formed of M successive measurement instants t (1,m), where the increment m goes from 1 to M, of M mean flow rate values ​​D (1) (t (m) ), and of M flow duration values ​​Δt e|D(1) (t (m) ) associated with the mean flow rate D (1) (t (m) ) measured.

[0043] In step 112, the processing unit 5 then determines the consumption profile %Δt s|D(1) (d) from the measurement signal Sm (1). To do this, it determines the cumulative flow duration Δt ec|D(1) associated with the measured mean flow rate values ​​D (1), by summing the flow durations Δt e|D(1) having the same measured mean flow rate value D (1). Then, it determines the associated distribution function FD(1) (d), and deduces the consumption profile %Δt s|D(1) (d) = 1-FD(1) (d). The consumption profile %Δt s|D(1) (d) thus indicates the proportion of time within the monitoring duration Δt s(1) for which the measured mean flow rate D (1) is greater than the value d.

[0044] The consumption profile %Δt s|D(1) (d) is identified as having no consumption anomalies and is therefore integrated into the database. More precisely, the profile can be identified by the user as having no anomalies. Alternatively, as described later with reference to step 141, the processing unit 5 can compare the value of the consumption profile %Δt s|D(1) (D s ) for a reference average flow rate D s, for example equal to 0.1 L / min, to a predefined anomaly threshold, for example equal to 20%.

[0045] Steps 111 and 112 are repeated for successive monitoring durations Δt s(n) until, as shown in step 113, the number N of consumption profiles (%Δt s|D(n) (d)) 1≤n≤N is at least equal to the predefined minimum value N min. This yields a database BD containing at least N min consumption profiles %Δt s|D(n) (d) identified as being without consumption anomalies, and therefore from which an anomaly threshold Sa can then be defined. Monitoring and detection phase 120

[0046] The monitoring and detection phase 120 comprises a step 121 for determining the anomaly threshold Sa, steps 131 and 132 for measuring and determining a current consumption profile %Δt s|D(ic) (d), and then steps 141, 142, and 143 for analyzing the current profile and detecting the presence or absence of a detection anomaly. The monitoring phase 120 is performed for a current monitoring duration Δt s(ic) and is repeated when moving to the next monitoring duration Δt s(ic+1), with the indicator ic being incremented by one.

[0047] In step 121, the processing unit 5 determines the anomaly threshold Sa(ic)(Ds), associated with at least one reference value Ds of the average flow rate D, from the consumption profiles (%Δts|D(n)(d))1≤n≤N of the database. In this first embodiment, the anomaly threshold is a scalar associated with a single predefined reference value Ds of the average flow rate, here for example equal to 0.1 L / min.

[0048] The reference value Ds of the average flow rate D depends on the water consumption patterns of the consuming system 1, and therefore on the type of consuming devices 2 present. The value of 0.1 L / min is a sufficiently low non-zero value to correspond to the flow rate of a poorly closed tap or a leak in a shower or toilet. In a residential setting, the proportion of daily flow time for which the average flow rate exceeds 0.1 L / min is usually around 20% or even less. Therefore, it is understandable that if a typical consumption profile %Δts|D(ic) (d) shows a proportion of flow time associated with a flow rate Ds of 0.1 L / min that exceeds the anomaly threshold set, for example, at 40%, it can be deduced that water consumption has increased, and thus that there is a consumption anomaly, such as an intermittent or continuous leak.

[0049] The anomaly threshold Sa (ic) (D s ) is defined from the values ​​(%Δt s|D(n) (D s )) 1≤n≤N for the reference value D s of the mean flow rate D. Here, it is equal to an average <%Δt s|D(n) (D s )> 1≤n≤N of the values ​​(%Δt s|D(n) (D s )) 1≤n≤N of the consumption profiles at the reference value D s, to which a predefined coefficient has been added. The coefficient can be equal to at least one standard deviation σ %Δts(n) (D s ) associated with the values ​​(%Δt s|D(n) (D s )) 1≤n≤N , and here is equal to three times the standard deviation. It may also include a so-called tolerance term tt, for example equal to 5%.

[0050] In other words, the anomaly threshold Sa(D s ) is defined in this example by the following relationship: Sa ic D s = % Δt s D n D s 1 ≤ n ≤ N + 3 × σ % Δts n D s + tt

[0051] Other definitions of the anomaly threshold are obviously possible, for example, using a median instead of a mean. Generally speaking, the anomaly threshold can be calculated from a measure of central tendency (mean, median, etc.). - central tendency, (in English) to which a coefficient is added or multiplied. The mean can be an arithmetic mean, a weighted mean, or equivalent. The standard deviation term and / or the tolerance term may or may not be present. One can also use, among other things, a term corresponding to a predefined percentile associated with the consumption profiles (%Δt s|D(n) (D s )) 1≤n≤N, for example, instead of the standard deviation term.

[0052] Next, during steps 131 and 132, we acquire the current measurement signal Sm (ic) during the current monitoring time Δt s(ic) , then we determine the current consumption profile %Δt s|D(ic) (d).

[0053] During step 131, during the current monitoring time Δt s(ic) , the measuring device 4 acquires the current measurement signal Sm (ic) , which is therefore formed of M values ​​of successive measurement instants t (ic,m) , of the average measured flow D (ic) (t (m) ), and of the flow duration Δt e|D(ic) (t (m) ) associated with the average measured flow.

[0054] In step 132, the processing unit 5 then determines the current consumption profile %Δt s|D(ic) (d) from the current measurement signal Sm (ic). As previously mentioned with reference to step 112, the processing unit 5 determines the cumulative flow time Δt ec|D(ic) associated with the measured mean flow rate values ​​D (ic) by summing the flow times Δt e|D(ic) having the same measured mean flow rate value D (ic). It then determines the associated distribution function FD(ic) (d) and deduces the current consumption profile %Δt s|D(ic) (d) = 1-FD(ic) (d).

[0055] Then, during steps 141, 142 and 143, we compare the current consumption profile %Δt s|D(ic) (d) to the anomaly threshold Sa (ic) (D s ), and we deduce whether an anomaly is present or not, in order to then possibly update the database BD.

[0056] In step 141, the value of the consumption profile %Δt s|D(ic) (D s ) is compared to the reference value D s of the average flow rate with respect to the anomaly threshold Sa (ic) (D s ). The reference value D s is here equal to 0.1L / min.

[0057] If the value %Δt s|D(ic) (D s ) is at least above the anomaly threshold Sa (ic) (D s ), that is, here greater than or equal to the anomaly threshold, then the current consumption profile shows abnormal water consumption, and a consumption anomaly is detected. The process continues to step 142. Note that in the case where the current consumption profile is defined as %Δt s|D(ic) (d) = FD(ic) (d ), the value %Δt s|D(ic) (D s ) is said to be at least above the anomaly threshold Sa (ic) (D s ) when it is less than or equal to the anomaly threshold.

[0058] If the value %Δt s|D(ic) (D s ) is below the anomaly threshold Sa (ic) (D s ), that is, in this case, less than the anomaly threshold, then the current consumption profile does not show excessive water consumption, and the profile is identified as having no consumption anomaly. The process continues to step 143. Furthermore, in the case where the current consumption profile is defined as %Δt s|D(ic) (d) = FD(ic) (d ), the value %Δt s|D(ic) (D s ) is said to be below the anomaly threshold Sa (ic) (D s ) when it is greater than the anomaly threshold.

[0059] During step 142 (presence of a consumption anomaly), the user is alerted by processing unit 5 to the presence of a consumption anomaly during the current monitoring period Δt s(ic). The user can then mark the current consumption profile as having an anomaly, so that it is not included in the database. Monitoring then continues until the end of the current monitoring period Δt s(ic), and then monitoring phase 120 resumes at step 131, acquisition of the measurement signal Sm (ic+1), for a new monitoring period Δt s(ic+1).

[0060] An "intensity" indicator for the anomaly can be calculated and displayed to the user. This can be defined as the deviation of the value %Δt s|D(ic) (D s ) from the anomaly threshold Sa(D s ). Thus, the larger this deviation in absolute value, the longer the flow duration associated with the consumption anomaly. This deviation is maximal when the value %Δt s|D(ic) (D s ) equals 100%, indicating that the consumption anomaly is present throughout the entire current monitoring period Δt s(ic) , and may therefore represent a continuous leak.

[0061] During step 143 (absence of an anomaly), the current consumption profile %Δt s|D(ic) (d) is added to the database BD. Then the monitoring phase 120 resumes at step 121 of determination of the anomaly threshold Sa (ic+1) (D s ), for a new monitoring duration Δt s(ic+1) .

[0062] Thus, the detection process allows for the simple and efficient identification of the presence or absence of an anomaly in the current consumption profile, while also being able to adapt to changes in the consumption of consumer system 1 over monitoring periods. This is achieved by regularly updating the anomaly threshold value through the integration of current consumption profiles into the database when these profiles are identified as being free of anomalies. In this way, the anomaly threshold adapts according to the actual consumption of consumer system 1, providing more relevant and effective anomaly detection. The anomaly threshold is therefore not defined in such a way as to a priori,i.e. without taking into account the type of consumer system 1 and the evolution of its consumption. On the contrary, by using a regularly updated consumption profile database through the integration of current consumption profiles without anomalies on the one hand (these profiles being specific to the monitored consumer system 1), and by calculating the anomaly threshold from the database, anomaly detection is of better quality.

[0063] There figure 3AThis illustrates an example of consumption profiles %Δt s|D(ic) (d) derived from measurement signals acquired by a water flow meter, where the monitoring period Δt s(ic) corresponds to one day. We observe that most of the curves (referenced by "A") show a value of %Δt s|D (d=0) of at most approximately 5% for a zero average flow rate. This reflects the fact that for 5% of a day, the average flow rate is non-zero (flow is present). In other words, consumer system 1 consumes water only during 5% of the day. Conversely, it appears that for one of the curves (referenced by "B"), the value of %Δt s|D (d=0) is approximately 35%, therefore much higher than 5%. This reflects the fact that, for 35% of the day, the average flow rate is non-zero, and in this case, greater than approximately 4 L / min. In other words, for 35% of the day, consumer system 1 consumes water with an average flow rate greater than 4L / min.Such a curve can therefore be representative of an intermittent leak.

[0064] There figure 3B This illustrates three types of consumption profiles %Δt s|D(ic) (d) derived from measurement signals acquired by a pulse water meter, where the monitoring period Δt s(ic) also corresponds to one day. First, note that all the curves begin at 100%, in other words, all values ​​of %Δt s|D (d=0) are equal to 100%. This stems from the fact that a pulse water meter cannot determine the actual flow between two successive pulses. Thus, although there may be a period of no flow between two pulses, the average flow rate will always be non-zero between two pulses. The consumption profiles therefore have a 100% probability of having a non-zero average flow rate.

[0065] A first set (referenced "A") of consumption profiles shows a %Δt s|D (d) value that decreases sharply from approximately 100% to 5% between 0 and 0.1 L / min, then remains essentially constant above 0.1 L / min or decreases slowly towards zero. This type of consumption profile is representative of an absence of consumption anomalies. The anomaly threshold can therefore be set at 25-30% for a reference value D s equal to 0.1 L / min.

[0066] A second set (referenced "B") of consumption profiles shows a %Δt s|D (d) value that decreases sharply from approximately 100% to 70% between approximately 0 and 0.1 L / min, then gradually decreases from approximately 70% to 5% between approximately 0.1 L / min and 0.3 L / min, and then remains essentially constant from 0.3 L / min onwards or decreases slowly towards zero. This type of consumption profile is representative of the presence of a consumption anomaly such as an intermittent leak.

[0067] Finally, a third set (referenced "C") of consumption profiles shows a %Δt s|D(d) value that remains approximately 100% up to about 0.4 L / min, then drops sharply from 100% to about 10% between approximately 0.4 L / min and 0.6 L / min, and then remains essentially constant from 0.6 L / min onwards. The fact that the %Δt s|D(d) value remains at 100% up to about 0.4 L / min indicates the presence of a consumption anomaly, such as a continuous leak of at least 0.4 L / min. Indeed, this value indicates that there is a 100% probability, during the monitoring period, of having an average flow rate greater than 0.4 L / min.

[0068] THE fig. 3A and 3B thus show that it is possible to determine a reference value D s of the average flow rate D allowing to discriminate the consumption profiles with and without consumption anomaly, whether the measuring device 4 includes a water meter with flow meter or a pulse water meter.

[0069] There figure 3C This illustrates another example of consumption profiles %Δt s|D(ic) (d) derived from measurement signals acquired by a pulse water meter, where the monitoring duration Δt s(ic) corresponds to one day. Consumption profiles %Δt s|D(n) (d) from the database are shown here, along with three examples of common consumption profiles exhibiting consumption anomalies.

[0070] The consumption profiles (%Δt s|D(n) (d)) 1≤n≤N from the database DB form reference profiles from which the anomaly threshold can be determined. In the example given previously, the anomaly threshold Sa(D s ) corresponds to the average of the values ​​(%Δt s|D(n) (D s )) 1≤n≤N where the reference value D s of the average flow rate is equal to 0.1 L / min, from which a coefficient corresponding to three times the standard deviation associated with these values ​​and a 5% tolerance has been subtracted. Here, it is approximately 20-25%.

[0071] Three typical consumption profiles are illustrated. The first profile shows a %Δt s|D (D s ) value of approximately 30%, so the proportion of flow %Δt s during the monitoring period Δt s with an average flow rate greater than 0.1 L / min is 30% and here is roughly equal to the anomaly threshold Sa(D s ). The second profile shows a %Δt s|D (D s ) value of approximately 60%, so the proportion of flow %Δt s with an average flow rate greater than 0.1 L / min is 60% and here again exceeds the anomaly threshold Sa(D s ). These two examples are representative of an intermittent leak. Finally, the third profile has a value of %Δt s|D (D s ) of 100%, so that the flow %Δt s during the monitoring time Δt s with an average flow rate greater than 0.1L / min is continuous (continuous leakage especially in the case of a flow meter).

[0072] The detection method can have several advantageous variations. For example, the database can contain a maximum number of consumption profiles, Nmax, for instance, 30 to cover a month of consumption. It is advantageous for the maximum number of consumption profiles, Nmax, to be a sliding number. Thus, when the database contains Nmax profiles, adding the current profile results in the deletion of the oldest profile. Therefore, the anomaly threshold, which is updated each time a current profile without an anomaly is identified, will be relatively insensitive to sudden changes in a current profile without anomaly, but will still be sensitive to gradual changes in consumption profiles.

[0073] Furthermore, the anomaly threshold Sa (ic) (D s ) can vary within a predefined range, delimited by a minimum and a maximum value. The minimum value can be zero, but it does not have to be. Conversely, the maximum value prevents an excessively large standard deviation from generating an anomaly threshold that is too high, thus making it impossible to detect actual consumption anomalies. This maximum value depends in particular on the type of consumer system 1. It can be set at 30% for a residential building.

[0074] As mentioned previously, during preparation phase 10, consumption profiles can be compared to an initial anomaly threshold predefined by the user, while waiting for the number N of profiles in the database to reach the value N min. The initial value of the anomaly threshold can, for example, be set at 20%.

[0075] Furthermore, step 141 can be carried out regularly during the same monitoring period Δt s(ic) , as the values ​​of the measurement signal Sm (ic) are transmitted to the processing unit 5. This allows a consumption anomaly to be detected as early as possible, without waiting for the monitoring period Δt s(ic) to be completed.

[0076] Finally, it should be noted that the detection method according to this first embodiment can advantageously take advantage of the high sensitivity of certain types of measuring devices to low average flow rates, such as pulse counters.

[0077] There figure 4 is a flowchart of an anomaly detection process according to a second embodiment. The consumer system 1 is also a family home and the fluid of interest is liquid water.

[0078] In this second embodiment, the consumption profile is the variation %Δt s (D), as a function of the average flow rate, of the proportion %Δt s of cumulative flow duration Δt ec|D over the monitoring duration Δt s. This variation is denoted here as %Δt s (D) because it is a function of the average flow rate D, and differs from the function %Δt s (d) of the first embodiment, which was a probability associated with an average flow rate D greater than d.

[0079] The consumption profile %Δt s (D) may show peaks depending on the average flow rate, these peaks being characteristic of the usual average flow rates corresponding to the different consumer devices 2. Indeed, a shower has an average flow rate which is not that of a washing machine or that of a toilet flush.

[0080] The consumption profile %Δt s (D) is therefore a profile of the cumulative flow duration, or here, its proportion over the monitoring period, as a function of the average flow rate. It represents the flow rate distribution of the flow duration (or the time proportion of the flow over the monitoring period). Note that this distribution is the derivative of the cumulative distribution function FD (d).

[0081] The detection process also includes an initial preparation phase 10, followed by a monitoring and detection phase 20. Several steps are identical or similar to those of the process according to the first embodiment, and will therefore not be detailed again. Preparation phase 210: Provide the database database

[0082] The preparation phase 210 consists of forming a database DB of consumption profiles identified as being without consumption anomalies.

[0083] During a step 211, the measuring device 4 acquires a first measurement signal Sm (1) during the monitoring time Δt s(1) , where the indicator ic is equal to 1. The measurement signal Sm (1) is identical to that of step 111.

[0084] In step 212, the processing unit 5 then determines the consumption profile %Δt s|D(1) (D) from the measurement signal Sm (1). To do this, it determines the cumulative flow duration Δt ec|D(1) associated with the measured mean flow rate values ​​D (1), by summing the flow durations Δt e|D(1) having the same measured mean flow rate value D (1). Then, it determines the proportion %Δt s|D(1) (D) = Δt ec|D(1) (D) / Δt s, which is the ratio of the cumulative flow duration Δt ec|D(1) to the monitoring duration Δt s, for each mean flow rate value. The variation %Δt s|D(1) (D) of the proportion of the cumulative flow duration as a function of the mean flow rate D thus forms the consumption profile.

[0085] The consumption profile is identified as being without anomalies, in the same way as described in the process according to the first embodiment. It is integrated into the database. Steps 211 and 212 are repeated for successive monitoring durations Δt s(n) until, as shown in step 213, the number N of consumption profiles (%Δt s|D(n) (D)) 1≤n≤N is at least equal to the predefined minimum value N min. This yields a database BD containing at least N min consumption profiles %Δt s|D(n) (D) identified as being without consumption anomalies (and with consumption of the fluid of interest), and therefore from which an anomaly threshold Sa can then be defined. Monitoring and detection phase 220

[0086] As before, the monitoring phase 220 is carried out for a current monitoring duration Δt s(ic), and is repeated when moving to the next monitoring duration Δt s(ic+1), with the indicator ic then being incremented by one unit.

[0087] In step 221, the processing unit 5 determines the anomaly threshold Sa(ic)(D), associated with each value of the mean flow rate D, from the consumption profiles (%Δt s|D(n)(D)) 1≤n≤N in the database. In this second embodiment, the anomaly threshold is not a scalar (as in the first embodiment), but a vector that has a value for each value of the mean flow rate D. Thus, each value of the mean flow rate D is a 'reference value' against which the current consumption profile will be compared to the anomaly threshold.

[0088] The anomaly threshold Sa (ic) (D) is defined from the values ​​(%Δt s|D(n) (D)) 1≤n≤N. Here, it is equal to a mean <%Δt s|D(n) (D)> 1≤n≤N as a function of the mean flow rate D, to which a predefined coefficient has been added. The coefficient can be equal to at least one standard deviation σ %Δts(n) (D) associated with the values ​​(%Δt s|D(n) (D)) 1≤n≤N, and here is equal to three times the standard deviation. It can also include a so-called tolerance term tt, for example, equal to 0.5% of the monitoring time (here, 7 min and 12 s). Other definitions of the anomaly threshold are obviously possible, as indicated previously with reference to step 121.

[0089] In other words, the anomaly threshold Sa(D) is defined in this example by the following relationship: ∀ D , Sa ic D = % Δt s D n D 1 ≤ n ≤ N + 3 × σ % Δts n D + tt

[0090] Next, during steps 231 and 232, we acquire the current measurement signal Sm (ic) during the current monitoring time Δt s(ic) , then we determine the current consumption profile %Δt s|D(ic) (D).

[0091] During step 231, during the current monitoring time Δt s(ic) , the measuring device 4 acquires the current measurement signal Sm (ic) .

[0092] In step 232, the processing unit 5 then determines the current consumption profile %Δt s|D(ic) (D) from the current measurement signal Sm (ic). Thus, the processing unit 5 determines the cumulative flow time Δt ec|D(ic) (D) for each value of the measured average flow rate D (ic), by summing the flow times Δt e|D(ic) having the same measured average flow rate value D (ic). Then, it determines the proportion of flow time %Δt s|D(ic) (D) as a function of the average flow rate D.

[0093] Then, during steps 241, 242 and 243, we compare the current consumption profile %Δt s|D(ic) (D) to the anomaly threshold Sa(D), and we deduce whether an anomaly is present or not, in order to then possibly update the database BD.

[0094] In step 241, the values ​​of the consumption profile %Δt s|D(ic) (D s ) are compared to those of the anomaly threshold Sa(D).

[0095] If there exists a value Ds of the average flow rate for which the value %Δts|D(ic) (Ds) is above (here greater than or equal to) the corresponding value of the anomaly threshold Sa(ic) (Ds), in other words: ∃Ds / %Δts|D(ic) (Ds) ≥ Sa(ic) (Ds), then the current consumption profile shows water consumption exceeding the threshold, and a consumption anomaly is detected. The process continues to step 242.

[0096] If, on the other hand, all the values ​​%Δt s|D(ic) (D) are below (here, lower than) the corresponding values ​​of the anomaly threshold Sa (ic) (D), in other words: ∀D, %Δt s|D(ic) (D) < Sa (ic) (D), then the current consumption profile does not show excessive water consumption, and the profile is identified as having no consumption anomalies. The process continues to step 243.

[0097] Steps 242 and 243 are identical to steps 142 and 143 and are not detailed again.

[0098] There figure 5 This illustrates an example of changes in consumption profiles %Δt s|D(ic) (D) from measurement signals acquired by the measuring device 4 during a day. The dashed line corresponds to the sum of the components of the anomaly threshold Sa (ic) (D) of the mean and the standard deviation (therefore without the tolerance term), and the solid line corresponds to the anomaly threshold Sa (ic) (D).

[0099] The consumption profiles %Δt s|D (D) therefore correspond to the variation, as a function of the average flow rate D, of the proportion of flow duration associated with each value of the average flow rate during the monitoring period (the day). The profiles highlight several peaks in flow duration for certain average flow rate values, which may correspond, for example, to the use of a washing machine, a dishwasher, a shower, and a toilet flush. By defining an anomaly threshold that varies according to the average flow rate, it is then possible to detect a consumption anomaly on the one hand, and also to identify the energy-consuming device that is causing it on the other.

[0100] The figure also shows the variation, as a function of average flow rate, of the sum of the mean calculated on the corresponding values ​​of the consumption profiles and three times the standard deviation, as well as the variation of the anomaly threshold. The difference between these two curves corresponds to the tolerance term.

[0101] It is observed that the cumulative flow duration associated with each average flow rate does not exceed approximately 1 to 2% of the monitoring period, so the anomaly threshold here varies between approximately 0.5 and 1.5%. The anomaly threshold value is therefore significantly lower than the 20 to 30% value it exhibits in the first embodiment. Consequently, the detection of consumption anomalies is more sensitive in this second embodiment.

[0102] Thus, the detection process according to this second embodiment makes it possible to detect consumption anomalies of the consumer system 1, with an anomaly threshold which is regularly updated and therefore takes into account the evolution of consumption of the consumer system 1. It also presents a higher sensitivity, and makes it possible to identify which consumer device would be at the origin of the anomaly.

[0103] Note that each consumption profile can be represented as a histogram where the average flow rate values ​​are divided into classes, for example, with a width of 0.2 L / min. Thus, the average flow rate varies from one class to another: ]0; 0.2], ]0.2; 0.4], ]0.4; 0.6], etc. The flow durations Δt e|D for each average flow rate class are then summed to obtain the cumulative flow duration Δt ec|D for each average flow rate class. It is advantageous to perform a sliding summation, so that the classes partially overlap in pairs, for example: ]0; 0.4], ]0.2; 0.6], ]0.4; 0.8], etc. This approach preserves flow duration peaks in cases where a peak is present between two successive classes. The anomaly threshold can be calculated based on consumption profiles constructed from the sliding sum.

Claims

1. Method for detecting an anomaly in consumption of a fluid of interest by a consuming system (1), ∘ the consuming system (1) comprising: at least one consuming device (2) consuming the fluid of interest and connected to a source (3) by a fluid circuit; and • a measuring device (4), connected to the fluid circuit between the source (3) and the consuming device (2), capable of measuring and transmitting to a processing unit (5) a measurement signal Sm representing a variation, over the course of a predefined monitoring time Δts, of an average flow rate D of the fluid of interest and of a flow time Δte|D associated with the average flow rate D; • the processing unit (5), which is adapted to determine, from the measurement signal Sm, a consumption profile corresponding to a variation, as a function of the average flow rate D, of a flow indicator representative of a proportion %Δts of a cumulative flow time Δtec|D of the fluid of interest during the monitoring time; ∘ the method comprising the following steps: ∘ a) providing (110; 210) a database comprising a plurality of consumption profiles previously identified as having no consumption anomaly; ∘ b) determining (121; 221) an anomaly threshold Sa(ic), for at least one so-called average flow rate reference value, from the consumption profiles in the database; ∘ c) acquiring (131; 231), using the measuring device (4), a so-called current measurement signal Sm(ic), over the duration of a monitoring time ΔtS(ic); ∘ d) determining (132; 232), using the processing unit (5), a so-called current consumption profile from the current measurement signal Sm(ic); ∘ e) comparing (141; 241) the current consumption profile to the anomaly threshold Sa(ic): • if the current consumption profile has, for at least one average flow rate reference value, a value below the anomaly threshold Sa(ic): identifying that no consumption anomaly is present and adding the current consumption profile to the database, then repeating steps b) to e) for a subsequent monitoring time ΔtS(ic+1); • if the current consumption profile has, for at least one average flow rate reference value, a value at least above the anomaly threshold Sa(ic): identifying that a consumption anomaly is present and not adding the current consumption profile to the database, then repeating steps c) to e) for a subsequent monitoring time ΔtS(ic+1).

2. Detection method according to claim 1, wherein the step of providing the database comprises the following steps: ∘ a1) acquiring (111; 211), using the measuring device (4), a measurement signal Sm(n) for the duration of a monitoring time ΔtS(n); ∘ a2) determining (112; 212), using the processing unit (5), a consumption profile from the measurement signal Sm(n), identified as having no consumption anomaly, and adding the determined consumption profile to the database; ∘ repeating (113; 213) steps a1) and a2) until the database comprises a number N of consumption profiles at least equal to a predefined minimum number Nmin.

3. Detection method according to claim 1 or 2, wherein the step (143; 243) of adding the current consumption profile to the database comprises comparing the number N of consumption profiles in the database, and if this number N is greater than a predefined maximum number Nmax, deleting the consumption profile associated with the least recent monitoring time.

4. Detection method according to any one of claims 1 to 3, wherein the step of determining the consumption profile comprises determining a variation, as a function of the average flow rate, of the cumulative flow time by adding the values of the flow times associated with a same average flow rate value.

5. Detection method according to claim 4, wherein each consumption profile is determined by carrying out the following steps: ∘ determining a distribution function FD(d) from the variation of the cumulative flow time as a function of the average flow rate, the distribution function indicating a probability P(D≤d) of obtaining a value of the average flow rate D that is less than or equal to a value d; ∘ defining the consumption profile as being equal to 1-FD(d) or to FD(d).

6. Detection method according to claim 4, wherein each consumption profile is determined by calculating a ratio of the cumulative flow time over the monitoring time, as a function of the average flow rate.

7. Detection method according to any one of claims 1 to 6, wherein the step of determining the anomaly threshold comprises calculating an average or a median of the values of the consumption profiles of the database, for at least one predefined reference value of the average flow rate.

8. Detection method according to claim 7, wherein the step of determining the anomaly threshold further comprises calculating a standard deviation associated with the values of the consumption profiles of the database, for at least one predefined reference value of the average flow rate.

9. Detection method according to any one of claims 1 to 8, wherein the measuring device comprises a flow meter or a pulse counter.

10. Detection method according to any one of claims 1 to 9, wherein the fluid of interest is a liquid or a gas, chosen from water, hydrogen, oxygen, nitrogen, and helium.

11. Detection method according to any one of claims 1 to 10, wherein the consuming system (1) is a house, a building comprising several apartments, a factory, a hospital, a school, or a campsite.

12. Detection method according to any one of claims 1 to 11, wherein the consuming system (1) comprises a plurality of consuming devices (2), the measuring device (4) being connected to the fluid circuit between the source (3) and the consuming devices (2).