FR3075969A1 - DEVICE FOR DETECTING FAILURE IN MONITORING AN ELECTRICITY NETWORK - Google Patents

Abstract

Device for detecting failure in the monitoring of an electrical network, characterized in that it comprises, at least one measuring device, at least one transmission line for conveying at least one measurement from the measuring device to a support designed to maintain the value and nature of the measurement, at least one clock collaborating with the medium to determine the date of transmission or reception of the initiated measurement set, a consistency check tool designed to handle the set of measurements initial date received by the medium within a predetermined time interval, generating an alert if one of the received measurements is inconsistent with the others, and identifying the faulty measuring apparatus corresponding to the incoherent measurement.

Description

Field of the invention The invention relates to a device for managing an electrical network. The invention also relates to a method for managing an electrical network implemented by such a device.

Presentation of the Prior Art

The management of an electrical network is based on different physical quantities (frequency, voltage by zone, temperature, production, consumption) detected by sensors that are possibly faulty or collected declaratively with possible errors.

The management of wear and tear on the mechanical parts of a power plant depends on sensors which themselves may be faulty. A sensor failure can lead to the decision not to replace a mechanical element that can cause a major failure for the plant as well as for the electrical network.

A decision based on an incorrect measurement can lead to high voltage on the electrical network or malfunction within a power plant, which can endanger the entire electrical network.

A failure in the monitoring of parts or control devices of the electrical network which is not identified therefore constitutes a major security problem.

Presentation of the invention

A general aim of the invention is to overcome the drawbacks of prior art electrical network management systems.

This object is notably achieved by means of a failure detection device in the monitoring of an electrical network, characterized in that it comprises: - at least one measuring device; - at least one transmission line to route at least one measurement from the measuring device to a support designed to preserve the value and nature of the measurement; - at least one clock collaborating with the support in order to determine the date of emission or reception of the initial set of measurements; - a consistency verification tool designed to process the initial dated set of measurements received by the support within a predetermined time interval, generate an alert if one of the measurements received is inconsistent with the others, and identify the faulty measurement device corresponding to the inconsistent measure.

Measurements made at different locations in the network to ensure monitoring are transmitted to a management center.

It is possible to exploit the fact that these measures, although they come from different locations, can be deeply linked to each other. For example, the measurements taken on a mechanical element of a power plant can be deeply linked to the measurements taken by another sensor or even to more global measurements related to the production of the power plant, at outdoor temperature.

An advantage of such a measurement consistency verification tool which exploits this link is to reveal an anomaly in a set of measurements received in full, on time, in the expected form and without any apparent defect, and to clearly identify which measurement is not reliable so as not to take it into account.

The device is advantageously supplemented by the fact that the consistency verification tool is designed to propose a coherent replacement value for the measurement identified as inconsistent.

Such a measurement consistency verification tool enables decisions to be taken quickly, in particular at times when consumption levels increase or decrease significantly within the electricity network. At such a time, the reception of a set of measures identified as carrying an anomaly, with a coherent replacement value proposition, can make it possible to make a decision, and not to jeopardize the proper functioning of the network.

According to a particular embodiment of the invention, the consistency verification tool is designed to process heterogeneous measurements.

The measurements made at different locations on the network are different in nature, that is to say that they have different dimensions: voltage in an area of the electrical network, frequency of the voltage in an area of the electrical network, outside temperature, electrical power produced, electrical consumption in the network, wear of a physical part of the electrical network, etc.

The link between the measures exists despite their heterogeneous nature. The more different sensors there are, the more different links can be exploited between different measurements. The accuracy of fault identification and network security are improved when the verification tool is able to process heterogeneous measurements.

On the other hand, the heterogeneous nature of the measurements received makes checking for consistency more difficult to process. The greater the number of different dimensions in the measurement set received, the more complex the verification of the link.

To solve this problem, it is possible to normalize each measurement received into a dimensionless value. By having historical reference measurements on the physical quantity measured, different techniques are possible, among them: - an operation which associates with the measurement received a value between "0" and "1", so that the value "0" corresponds the minimum measurement of historical reference measurements, and that the value "1" corresponds to the maximum measurement of historical reference measurements; this can in particular be obtained by a first step of subtraction from the measurement received from the minimum value of the historical reference measurements, then a second step of dividing by the difference between the maximum value and the minimum value of the historical reference measurements.

According to another embodiment of the invention, the network measurement devices are divided into several groups and each group is associated with a tool for checking the consistency of the measurements received.

The management of an electrical network requires taking decisions at different scales of the network, on the one hand because the number of measures to be taken into account is too large to centralize and process them in a single management center, on the other hand part because there may be a strong link between certain measures. For example, there is a geographic link between measurements of sunshine time, outside temperature and electricity consumption. For these two reasons, network management by zones is more effective than centralized management.

The definition of groups of measuring devices each associated with a consistency verification tool makes it possible to adapt to management by zones and to exploit the strong link between certain measurements placed in the same group.

According to another embodiment of the invention, the consistency verification tool which checks the consistency of a set of measurements associated with a date is designed to take more account of the reference measurement sets closest to this dated.

According to a particular embodiment of the invention, the consistency verification tool can implement or collaborate with dictionaries constructed from historical reference measurements. The consistency of a set of measures associated with a date is evaluated with respect to the historical reference sets which represent usual states of the network. The oldest historical reference games may, however, correspond to periods of time in which the network included fewer or less performing measuring devices. To assess the consistency of a recent set of measures, it seems appropriate to increase the weight of the historical reference sets which are contemporary with it.

This embodiment of the invention makes it possible to preferentially take into account the sets of historical reference measurements of the same era as that for which it is desired to verify the consistency.

Presentation of the figures Other characteristics and advantages of the invention will become more apparent from the description which follows, which is purely illustrative and not limiting, and must be read with reference to the appended drawing of Figure 1. This diagrammatically represents a device detection of failure in the monitoring of an electrical network.

detailed description

Figure 1 shows schematically a failure detection device in the monitoring of an electrical network. The device includes m measuring devices A1, A2 ... Am connected by the transmission line 1 to the support 2. The clock 3 and the consistency checking tool 4 are connected to the support 2.

When the various AltA2 ... Am measuring devices take a measurement reading of the state of the electrical network, the measurements are transmitted to the medium 2 by the transmission line 1. The measurements are time-stamped by the clock 3 and the medium keeps for each measure its origin, its value, its nature and its date.

The dated measurement set is of the form yt = (yt, i, yt, 2 -yt, m) with Ιθ date t and the m measurements yt> i, yt> 2 -yt.m received

The dates t can be presented as whole numbers or indexed by whole numbers whose ascending order corresponds to the chronological order of the measurements.

This set is sent to the consistency check tool designed to process the initial dated set of measurements received by the medium in a predetermined time interval.

The consistency of the dated measurement set yt is checked in m steps, each corresponding to one of the measuring devices: the consistency of each of the measurements yt i is checked with each index i corresponding to one of the measuring devices, with respect to the other measurements. ytp with p ψ i. The operations described below are performed for each index i e

We create xj = (yt, i, yt, 2 -yt, i-i> yt, i + i a set of m-1 measures.

The consistency of the dated measurement set is checked from n historical reference measurement sets noted Hky = (ïyi’ïyz -kym) with k e [1; n] associated with a measurement date tk and with the m measurements received on this date. We create in the same way the sets of m - 1 measures in the form Hkxl = (^ yi4y2 - ^ - 1, ^ + 1 -Hkym) · We note Ηΰι the set of ^ x ‘with k e [1; nj.

We thus associate with any Hky element an element ^ xL

We build a dictionary D1 made up of sets of m - 1 measures d) = (dj ^, dj2 ... dj (i + 1) ... djm) with j e [1; Kjl · D1 is presented in the form of a matrix D1 = [d [, dl2 ... of dimension (m - 1) x / fj.

Each dictionary D 'corresponds to the measuring device corresponding to the index i.

According to one embodiment, the dictionary D1 is created by the following operation:

With:

Argmin (/ (x)) the function that returns the set of points x at which the expression / (x) reaches its minimum value. || z | l2 for an element z of is the sum of the squares of the m-1 coordinates or measures constituting z. || "i || x the sum of the absolute values of the Ki coordinates or measures constituting at. λ an input parameter whose value is predetermined by ex posts tests (or in English "backtests").

Once the dictionary Ώ1 has been determined, we build a set of (m - 1) measurements image of the initial dated measurement set yt from Ώ1 by identifying the element d [which is the element d) of the nearest neighboring dictionary of xlt: d [= Argmin || x £ - dj || 2

We determine V closest neighbors of d [, in the set HDl of fcXÉ

According to one embodiment, these V nearest neighbors are denoted dllp with pe [l; l / j] and determined according to the following successive operations:

Vp e [1; d [p = Argmin || dz “Ml2 kxiÊHDi \ idi, i'di.2-dlp-i}

Once the V nearest neighbors dll p have been identified, a coherent dated measurement set ÿl is generated according to the following operations:

We associate with each dllp e HD \ set of sets of m - 1 measures Hkx \ the set of m measures Hky corresponding that we denote dllp.

We generate the matrix Vt = [djp dll2, ..., άιιν] of dimension (m - 1) x V.

We generate the matrix 7, = [d1 ^, dll2, ..., άιι ν] of dimension m x V.

We generate the operator = 7, (½7½ + /////) - 1½7.

With: K7 the transposed matrix of Vt. μ a predetermined parameter.

We determine y (= Argmin || y - Pkxt ^ 2 yeKm

The initial dated measurement set is compared to the coherent dated measurement set by calculating | yt, - ÿ ^ l There is generation of an alert if | yt, - ÿ ^ l is greater than a predetermined threshold and identification of the faulty measuring device corresponding to the indexed measurement i if | yt, - ÿ ^ l is greater than a predetermined threshold.

In this case, the value y- can be proposed as a coherent replacement value to replace the measure yt identified as inconsistent.

According to another embodiment, the identification of the V nearest neighbors dllp with p e [1; KD, of d [, in the set HDl of Hkxl is carried out according to the following successive operations:

Vp e [1; Π, dll p = Argmin d (Hkxl, d [)

Hkxi HDi ^ \ {di1, DIL2 dÎlp_1 ...}

The function ά (Ηκχι, d [) is expressed in the form of a product: d (Hkxl, d [) = ee ^ k ~ hGi (Hkxl, d [)

With: Θ is a strictly positive predetermined parameter.

Gi is a positive or zero function. The Gj function can for example be written in the form:

where y is a strictly positive predetermined parameter, p (i, k) is an estimate of the correlation coefficient between the measured quantities which appear at the i-th and the k-th position in the dated measurement sets and the functions fq have the purpose of which is to modify the value of the measurements in order to take into account the heterogeneous nature of the measurements.

By having historical reference measurements on the physical quantity measured, various techniques are possible for proposing the functions fq, among them: - an operation which associates with the measurement received a value between “0” and “1”, so that the value "0" corresponds to the minimum measurement of historical reference measurements, and the value "1" corresponds to the maximum measurement of historical reference measurements; this can in particular be obtained by a first step of subtraction from the measurement received from the minimum value of the historical reference measurements, then a second step of dividing by the difference between the maximum value and the minimum value of the historical reference measurements.

As in the previous mode, once the V nearest neighbors dll p have been identified, a coherent dated measurement set y ^ is generated, compared to the initial dated measurement set, the faulty measurement device corresponding to the possible inconsistent measurement is identified and, where appropriate, a replacement value for the measure identified as inconsistent may be proposed.

According to another embodiment, all the historical and received measurements are normalized to a dimensionless value, according to one of the

techniques already mentioned. We work with Hky and xlt, which are the normalizations of Hky and xlt games.

We simultaneously determine the dictionary D1 = [dj, dl2 ... dj-.] And n sets of Kt values (a (, al2 ... aQ according to the following operation: in (4-4 - ^ - 4-4 -4) = Argmin - Y (|| Hkxl - Mx ak \\ 22 +

Mx = [di ..... dK.] ERlm- ^ Ki Δη [αΊ, a2 ... an] eRK ixn || d; || 2 <i D1 consists of Kt sets of m-1 measures dj = (dj, i 'dj, 2 - dlj' (i-iy djçi + 1y ... dljm) with; e [1; KJ.

Once the dictionary D1 and the n sets of Kt values (aj, a2 ... aQ determined, we construct Aj, a second dictionary consisting of Ki sets of m measures according to the following operation:

Once the second dictionary Aj has been determined, the set of Ki values â is identified from xlt and from the dictionary D \ according to: 1 II 'l | 2 â = Argmin - || xj - D'cql + A || < z | li · a £ RKi n

We thus determine the set D'd comprising (m - 1) measures, which is an image of the initial dated set of measures yt constructed from DE

A coherent and standardized dated measurement set is generated by the operation: = Argmin || y -c yERm 2 o We then perform an inverse operation to the initial normalization operation to determine the coherent dated measurement set yj. o As for the previous modes, once the coherent dated measurement set ÿj has been generated, it is compared to the initial dated measurement set, the faulty measuring device corresponding to the inconsistent measurement

possible is identified and if necessary a replacement value for the measure identified as inconsistent can be proposed.

Claims (24)

1. Failure detection device in the monitoring of an electrical network, characterized in that it comprises: - at least one measuring device; - at least one transmission line to route at least one measurement from the measuring device to a support designed to preserve the value and nature of the measurement; - at least one clock collaborating with the support in order to determine the date of emission or reception of the initial set of measurements; - a consistency verification tool designed to process the initial dated set of measurements received by the support within a predetermined time interval, generate an alert if one of the measurements received is inconsistent with the others, and identify the faulty measurement device corresponding to the inconsistent measure. 2. Failure detection device in the monitoring of an electrical network according to claim 1 characterized in that the consistency verification tool is further designed to provide a replacement value consistent with the identified inconsistent measurement. 3. Device for detecting failure in monitoring an electrical network according to claim 1 or 2 characterized in that it comprises several measuring devices. 4. Failure detection device in the monitoring of an electrical network according to claim 3 characterized in that at least one measuring device provides a measurement heterogeneous to the measurements of other measuring devices. 5. Device for detecting failure in monitoring an electrical network according to claim 3 or 4 characterized in that the electrical network measuring devices are divided into several groups, and that for each group the device comprises - a support designed to preserve the value and the nature of the measurements carried out in the group; - one or more transmission lines designed to route the measurements from the measuring devices to the support; - at least one clock collaborating with the support in order to determine the date of emission or reception of the initial set of measurements; - a consistency verification tool designed to process the initial dated measurement set received by the group's support within a predetermined time interval, generate an alert if one of the measurements received is inconsistent with the others, and identify the measure of the failing group corresponding to the inconsistent measure. 6. Failure detection device in the monitoring of an electrical network according to claim 5 characterized in that the group consistency verification tool is further designed to provide a replacement value consistent with the inconsistent measure identified. 7. Failure detection device in the monitoring of an electrical network according to claim 5 or 6 characterized in that the group comprises a number of whole measuring devices m greater than 1, the group consistency verification tool implements at least one dictionary of sets of (m - 1) measures constructed from historical reference measures of the group, and in that it further comprises the means designed to construct at least one set of (m - 1 ) image measurements of the original dated measurement set from said dictionary, generating a coherent dated measurement set from at least one image set, comparing the initial dated measurement set to the coherent dated measurement set. 8. Failure detection device in the monitoring of an electrical network according to claim 7 characterized in that - a dated measurement set is written yt = (yt, i> yt, 2 -yt, m) with the date of measurement t and m measurements yt> 1, yt, 2 -yt, m received; - the group consistency verification tool implements m dictionaries £> ', ie | [l; m], of sets of (τη - 1) measures, constructed from historical reference measures of the group, each dictionary corresponding to one of the m measuring devices of the group; - and in that it further comprises for each i e [1; m] the means designed to create a set of measures xlt = (yt, i, yt, 2 -yt, ii> yt, î + i -yt, m)> build a set df of (τη - 1) measures image of xlt corresponding to the nearest neighbor dj = (dj ^, d} 2 ... dj (i + 1) - dljm) of xlt in the dictionary D ', generate a coherent dated measurement set y) = (% ι,% 2 -ÿt.m) from d) and xi, calculate | yti - ÿ ^ J, compare | yti - to a predetermined threshold, generate an alert if | yti - is greater than a predetermined threshold, identify the device of faulty measurement corresponding to the indexed measurement i if | yt, i - yi, i | is above a predetermined threshold. 9. Device for detecting failure in the monitoring of an electrical network according to claim 8 characterized in that it further comprises for each ie [l; m]] the means designed to propose the value ÿt.i as the value of consistent replacement to measure y, if the latter is identified as inconsistent. 10. Device for detecting failure in the monitoring of an electrical network according to claims 8 or 9 characterized in that: - the historical reference measurements of each group which serve to construct the dictionary D1 correspond to n sets of measurements of the form = (kyi4y2 - faith-i, Hkyi + 1 with ke [1; n] | from n measurement sets with reference dates "y = (* γι, - &amp;% η) each associated with a measurement date tk and to the m measurements received on this date; characterized in that it further comprises for each ie [1; m] the means designed to identify V closest neighbors "x * of the set of measurements dj corresponding to the minima of the distance d (kxl, dj) which is expressed in the form of a product: d (Hkxl, dl) = ee ^ k ~ ^ Gi (kxl, d £) where Θ is a strictly positive predetermined parameter, and is a positive function or zero. 11. Device for detecting a fault in the monitoring of an electrical network according to claim 10 characterized in that the function Gi is written in the following form: where y is a strictly positive predetermined parameter, p (i, q) is an estimate of the correlation coefficient between the measured quantities which appear at the i-th and the q-th position in the dated measurement sets and the functions fq have the purpose of which is to modify the value of the measurements in order to take into account the heterogeneous nature of the measurements. 12. Device for detecting failure in the monitoring of an electrical network according to one of the preceding claims, characterized in that at least one measuring device delivers a measurement of one of the following quantities: voltage in an area of the electrical network , frequency of the voltage in an area of the electrical network, outside temperature, electrical power produced, electrical consumption in the network, wear of a physical part of the electrical network, or any other quantity of interest in the monitoring of an electrical network. 13. Method for detecting failure in the monitoring of an electrical network comprising at least one measuring device, characterized in that it comprises the steps of - measuring from at least one measuring device; - transmission and reception of at least one measurement supplied by a measuring device; - creation of an initial dated measurement set including the measurement date and the measurements received; - application of a consistency check tool to said initial set of measurements; - generation of an alert if a measurement is inconsistent in the initial dated measurement set; - identification of the faulty measuring device corresponding to the inconsistent measurement. 14. The method of claim 13 characterized in that it further comprises a step of proposing a replacement value of the measure identified as inconsistent. 15. The method of claim 13 or 14 characterized in that the electrical network comprises several measuring devices. 16. The method of claim 15 characterized in that at least one measuring device provides a measurement heterogeneous to the measurements of other measuring devices. 17. The method of claim 15 or 16 characterized in that the electrical network measuring devices are divided into several groups and that the method comprises the steps of: - creation of an initial dated measurement set for each group comprising the measurements from the measuring devices of the said group and the measurement date; - application for each group of a consistency check tool to said initial dated set of measures; - generation of an alert for each group if a measurement is inconsistent in the corresponding initial dated measurement set; - identification in each group of the faulty measuring device corresponding to the inconsistent measurement. 18. The method of claim 17 characterized in that it further comprises a step for each group of proposal of a replacement value of the measure identified as inconsistent. 19. The method of claim 18 characterized in that a group comprises an integer number of measuring devices m greater than 1, that the consistency verification tool of said group implements at least one dictionary of games of ( τη -1) measurements constructed from historical reference measurements of the group, and in that it further comprises the steps of: o construction of at least one set of (m - 1) measurements image of the dated measurement set initial from said dictionary; o generation of a consistent dated measurement set from the image set; o comparison of the initial dated measurement set to the coherent dated measurement set. 20. Method according to claim 19 characterized in that: - an initial dated measurement set is written yt = (yt, i, yt, 2 -yt, m) with the measurement date t and τη measurements yt, i, yt, 2 - yt, m received; - the consistency check tool for each group implements τη dictionaries D ', i e constructed from historical reference measurements of the group, each dictionary corresponding to one of the group's measurement devices; - The method further comprises the following steps for each i e [1; mj: o creation of the dated measurement set x [= (yt, i, yt, 2 -yt, z-i, yt, z + i -yt, m)! o construction of the set of (τη - 1) image measures of x [corresponding to the nearest neighbor d \ of x {in the dictionary £> *, dj = (dZ, l "dZ, 2 dZ, (i + l)" · DZ, m)> o generation of a coherent dated measurement set ÿ £ = (ÿt, i.ÿt, 2 from d \ and x [; o calculation of | y £) i - ÿ ^ l; o comparison of | yt> i - yt'É | to a predetermined threshold; o generation of an alert if | yt i - y ^ | is greater than a predetermined threshold; o identification of the faulty measuring device corresponding to the measurement indexed i if | yt> i - ÿt'J is greater than a predetermined threshold. 21. Method according to claim 20 characterized in that the value ÿt ‘f is proposed as a coherent replacement value to replace the measurement y, if the latter is identified as inconsistent. 22. Method according to claim 20 or 21 characterized in that: - the historical reference measurements which serve to construct the dictionary D1 correspond to n sets of measurements of the form = ("yi, kV2 - kyz-1. Feyz + 1 - "y ™) - with ke [1; nor from n sets of measurements dated with reference ky = (.kylt ky2 - ^ νηΐ) each associated with a measurement date tk and with the m measurements received on this date; - The method further comprises the following steps for each i e [1; ml: - the identification of the V nearest neighbors ^ xf of the set of measures dj corresponding to the minima of the distance d (fcX ', dj) which is expressed in the form of a product: dfâx1, d}) = eeltk ~ tlGi (kxl, d}) where Θ is a strictly positive predetermined parameter, and Gi is a positive or zero function. 23. Method according to claim 22 characterized in that the function Gi is written in the following form: Where γ is a strictly positive predetermined parameter, p (i, q) is an estimate of the correlation coefficient between the measured quantities which appear at the i-th and the q-th position in the dated measurement sets and the functions fq have to modify the value of the measures to take into account the heterogeneous nature of the measures. 24. Method according to one of claims 13 to 23 characterized in that at least one measuring device delivers a measurement of one of the following quantities: voltage in an area of the electrical network, frequency of the voltage in an area of the network electrical, outside temperature, electrical power produced, electrical consumption in the network, wear of a physical part of the electrical network, or any other quantity of interest in the monitoring of an electrical network.