ELECTRIC ARC DETECTION METHOD AND DEVICE

Abstract

The invention relates to an electric arc detection method and device in which: samples of first, second and third filterings are determined, in parallel, for a signal for a current time window and for another time window,a correlation is determined among: a first correlation between said samples resulting from the first filtering and determined for the current window and samples resulting from the first filtering and determined for the other window;a second correlation between said samples resulting from the second filtering and determined for the current window and samples resulting from the second filtering and determined for the other window;a third correlation between said samples resulting from the third filtering and determined for the current window and samples resulting from the third filtering and determined for the other window; andan electric arc is detected as a function of at least said determined correlation.

Description

The present invention relates to electric arc detection in an electrical installation through a correlation calculation applied to frequency samples. Such detection makes it possible to stop the circulation of electrical current if necessary, in order to prevent the electrical installation from being damaged.

More specifically, the present invention relates to an electric arc detection method in an electrical installation providing an electrical signal, comprising the following steps respectively carried out in a current window of a plurality of time windows for observing the electrical signal:i/ applying, to said electrical signal in parallel with at least one first filtering centered on a first frequency, a second filtering centered on a second frequency and a third filtering centered on a third frequency, said first, second and third frequencies being distinct;ii/ determining, for said window, at least samples resulting from the first filtering, samples resulting from the second filtering and samples resulting from the third filtering;iii/ determining the correlation between said samples determined for the current window and samples determined following the implementation of steps i/ and ii/ for at least one other window from the plurality of time windows for observing the electric signal; andiv/ detecting the presence of an electric arc as a function of at least said determined correlation.

Certain techniques, for example those described in document U.S. Pat. No. 7,345,860, detect an arc by implementing a step for recognizing the type of charges present on the electrical installation, from signals resulting from the filtering of the electrical signal by three bandpass filters arranged in parallel.

Other techniques avoid this charge recognition step by basing themselves on the responses of the three filters. One difficulty is related to the presence of a noise generated by certain charges, such as light dimmers.

Applying a correlation calculation makes it possible to detect the random aspect of this noise and limit false electric arc detections due to such noise. A signal adding the outputs of the three filters over a duration of an observation cycle is determined on the condition that the output of each of the filters is strictly positive, and the correlation between that signal and the signal corresponding to another cycle is calculated. Thus, a correlation, above a given threshold, and obtained for strictly positive values of the output of each of the filters, may be a sign that an arc is present.

The present invention aims to improve electric arc detection.

To that end, according to a first aspect, the invention proposes an electric arc detection method of the aforementioned type, characterized in that step iii/ for determining the correlation comprises determining at least one respective correlation from among:

• • a first correlation between said samples resulting from the first filtering and determined for the current window and samples resulting from the first filtering and determined for the other window;
  • a second correlation between said samples resulting from the second filtering and determined for the current window and samples resulting from the second filtering and determined for the other window;
  • a third correlation between said samples resulting from the third filtering and determined for the current window and samples resulting from the third filtering and determined for the other window; andin that step iv/ for detecting the presence of an arc depends on at least the correlation determined from among said first, second and third correlations.

The invention thus makes it possible to take the correlation of the signal at the output of each filter into account independently and to detect the presence of an arc on that basis. This arrangement makes it possible to reduce the risk of distorting the arc detection relative to a detection technique based on a correlation determined on the sum of the signals provided by the three filters.

Furthermore, the correlation was only calculated when the samples were strictly greater than zero at the output of each of the three filters. However, arc signatures exist according to which only one or two responses are provided by all of the three filters.

In embodiments, the method according to the invention further includes one or more of the following features:

step iv/ for detecting the presence of an arc is a function of the comparison of the determined correlation from among said first, second and third correlations, with a threshold;

step iii/ for determining the correlation comprises determining at least each of said first, second and third respective correlations;

step iv/ for detecting the presence of an arc is a function of at least each of said first, second and third correlations;

the difference between said respective correlation determined among said first, second and third correlations for one window and said respective correlation for another window is calculated, step iv/ for detecting the presence of an arc being a function of at least said calculated difference;

the number of successive samples with a value greater than zero is calculated among the samples resulting from a respective filtering among the first, second and third filterings, for the current window, said number is compared to a threshold, the presence of an electric arc is detected as a function of at least said comparison;

none of said first, second and third frequencies are equal to an integer that is a multiple of another of said frequencies.

According to a second aspect, the present invention proposes a computer program for detecting an electric arc including instructions for implementing the steps of a method according to the first aspect of the invention, when the program is run by computation means.

According to a third aspect, the present invention proposes a device for detecting an electric arc, in an electrical installation providing an electrical signal, comprising a filtering-sampling unit suitable, during a time window for observing the electric signal, for filtering said signal in parallel according to a first filtering centered on a first frequency, according to a second filtering centered on a second frequency, and according to a third filtering centered on a third frequency, said first, second and third frequencies being distinct,

and for determining, for said time window for observing the signal, signal samples resulting from the first filtering, signal samples resulting from the second filtering and signal samples resulting from the third filtering;the detection device further comprising a processing unit suitable for determining a correlation between said samples determined for a current time window for observing the electrical signal and samples determined for at least one other time window for observing the electrical signal, and for detecting the presence of an electric arc as a function of at least said determined correlation;said device being characterized in that the processing unit is suitable for determining said correlation by determining at least one respective correlation from among:

• • a first correlation between said signal samples resulting from the first filtering determined for the current window and signal samples resulting from the first filtering determined for the other window;
  • a second correlation between said signal samples resulting from the second filtering determined for the current window and signal samples resulting from the second filtering determined for the other window;
  • a third correlation between said signal samples resulting from the third filtering determined for the current window and signal samples resulting from the third filtering determined for the other window; andin that the processing unit is suitable for detecting the presence of an arc as a function of at least the correlation determined from among said first, second and third correlations.

These features and advantages of the invention will appear upon reading the following description, provided solely as an example, and done in reference to the appended drawings, in which:

FIG. 1 shows an electrical installation provided with an electric arc detecting device in one embodiment of the invention;

FIG. 2 is a view of functional modules of the electric arc detecting device in one embodiment of the invention;

FIG. 3 is a flowchart of an electric arc detection method in one embodiment of the invention;

FIG. 4 shows the evolution over time, for successive cycles, of the signal s, the samples at the output of the filter F₁ and the value of a corresponding correlation parameter;

FIG. 5 shows the evolution over time, for successive cycles, of the signal s, the correlation of the samples at the output of the filter F₁ and that correlation filtered by a low pass filter;

FIG. 6 shows the evolution over time, for successive cycles, of the signal s and different parameters CorrFilt₁, CorrFiltAC₁ and CorrFiltACFilt₁, obtained from the correlation of the samples at the output of the filter F₁;

FIG. 7 shows the evolution over time, for successive cycles, of the signal s and of different parameters CorrFR₁ and DiffCorrFilt₁ obtained from the correlation of the samples at the output of the filter F₁;

FIG. 8 shows the evolution over time, for successive cycles k=i, i+1 etc., of the signal s, of the 38 samples per cycle k Ech_1k, Ech_2k, Ech_3k, and the parameters count_timeFilt₁, count_timeFilt₂, count_timeFilt₃ computed for those cycles.

In FIG. 1, an electrical installation 1, in one embodiment of the invention, is shown. This electrical installation includes a phase conductor 2 and a neutral conductor 3. It further includes an acquisition system 4, a power supply unit 5, an electric arc detecting device 6, hereinafter referred to as “detection module”, and an actuator 9.

The acquisition system 4 is suitable for acquiring values of at least one electrical property representative of the electrical installation, in particular a current or voltage value.

For example, in the considered case, the acquisition system 4 comprises a current sensor delivering a signal s representative of a line current i circulating in the electrical installation, for example a phase current, using a current measuring toroid. In embodiments, this signal is representative of the time derivative of the current i, di/dt, or the electrical voltage of the power supply.

This electrical signal s is provided at the input of the detection module 6.

The detection module 6 includes a set 10 of bandpass filters, a microprocessor 8 and a database 7.

In the present case, the set 10 of bandpass filters includes three bandpass filters F₁, F₂ and F₃.

The detection module 6 is suitable for performing processing operations from the signal s that is provided to it as input, so as to detect the presence or absence of an electric arc, in the electrical installation 1, and to command activation or non-activation of the actuator 9 as a function of those processing operations.

The actuator 9 is suitable, as a function of commands received from the detection module 6, for interrupting, for example by opening switches 11 positioned on the conductors 2, 3, or allowing the circulation of the current in the electrical installation 1, for example by leaving the switches closed.

The memory 7 of the detection module 6 is suitable for storing the values of various parameters, and of the filtered samples defined below, as well as software instructions defining functions and steps outlined below and implemented by the detection module 6 when the software instructions are executed by the microprocessor 8.

In reference to FIG. 2, the detection module 6 thus includes 3 processing chains 20, 21, 22 acting in parallel from the signal s, and an activation unit 23 to which the results of the processing chains are provided.

The first processing chain 20 includes the filter F₁, followed by a processing module TRAIT₁.

The second processing chain 21 includes the filter F₂, followed by a processing module TRAIT₂.

The third processing chain 22 includes the filter F₃, followed by a processing module TRAIT₃.

Each filter F_n is a bandpass filter centered on the frequency f_n, n=1 to 3.

For example, none of these frequencies f_n is an integer multiple of another of these frequencies f_n. The frequencies f_n are chosen in the range of 10 kHz-100 kHz.

In the considered embodiment, each filter F_n is of the switched capacity type. The output voltage V of the filter is therefore the result of the filtering and sampling of the signal s provided to it as input. In other embodiments, the sampling is done at the input of the processing module.

In the considered case, each filter F_n is suitable for providing, during an i^th considered detection cycle with duration T, for the signal s, N samples with voltage V, representative of the frequency component f_n of the signal s.

In the described embodiment, the sampling frequency is 4 kHz.

Thus, in reference to the flowchart shown in FIG. 3, in the considered case, in a step 100, during the i^th cycle, the filter F_n with n=1 to 3 delivers the samples Ech_ni(j). j=0 to 37, to the processing chain TRAIT_n whereof the processing operations relative to the identification of the presence of noise characteristic of an arc will allow the activation unit 23 to determine the presence or absence of an electric arc.

The processing module TRAIT_n is suitable for carrying out the processing steps indicated below in reference to FIG. 3, with n=1 to 3, so as to obtain the values, for the i^th cycle, of the following parameters, in particular:

Corr_n[i];

CorrFilt_n[i];

CorrFiltAC_n[i];

CorrFiltACFilt_n[i];

DiffCorr_n[i];

DiffCorrFilt_n[i];

count_time_Filt_n[i].

In a step 101, the correlation between distinct cycles, for example the i^th current cycle and the (i−2)^th cycle, for the frequency f_n, is calculated, so as to detect a non-periodic behavior. For example, one first calculates, for j=0 to 37,

|Ech_ni(j)−Ech_n(i−2)(j)|, then one deduces Corr therefrom for the cycle i:

${\mathrm{Corr}}_n\left[i\right]=\sum _{j=0\stackrel{‵}{a}37}^{}{\mathrm{Ech}}_{\mathrm{ni}}\left(j\right)-{\mathrm{Ech}}_{n\left(i-2\right)}\left(j\right).$

As an illustration, the top graph in FIG. 4 shows the signal s obtained between successive cycles k=i−3 to i, as a function of the time t expressed in seconds (s).

The middle graph in FIG. 4 shows the samples Ech_1k(j) with j=0 to 37, obtained between successive cycles k=i−3 to i, as a function of the time t expressed in seconds (s).

The bottom graph of FIG. 4 shows the correlation Corr₁ obtained between successive cycles i−3 to i, as a function of the time t expressed in seconds (s).

In a step 102, in order to ensure a repetitive behavior of the detection and control the activation time to open the actuator 9, a low pass filter is applied to the correlation thus computed, delivering the value of the parameter CorrFilt_n[i] for the cycle i.

A low pass filter is applied to the parameter CorrFilt_n[i] thus computed, delivering the value of the parameter CorrFiltFilt_n[i] for cycle i.

The transfer function of the low pass filter is of the first order:

Y(n+1)=2^−w.X(n)+(1−2^−w). Y(n), where w is a positive integer.

As an illustration, the top graph of FIG. 5 shows the signal s obtained for successive temporal cycles along the time axis expressed in seconds (s).

The middle graph of FIG. 5 shows the correlation values Corr₁ obtained for successive temporal cycles along the time axis expressed in seconds (s), in reference to the low frequency f₁.

The bottom graph of FIG. 5 shows the filtered correlation CorrFilt₁ obtained in reference to the low frequency f₁, for successive temporal cycles along the time axis expressed in seconds (s).

In certain cases, the result of the correlation before an arc is different from zero, and the correlation value increases when the arc appears.

In a step 103, in order to force the correlation to zero before an arc (the correlation being different from zero after the arc), this pre-arc correlation, called continuous component, is eliminated by subtracting the parameter CorrFiltFilt_n[i] from the parameter CorrFilt_n[i]. The result yields the parameter CorrFiltAC_n called, for the considered cycle i: CorrFiltAC_n[i].

In a step 104, additional filtering makes it possible to smooth the obtained signal and avoid frequent passages past zero, delivering the parameter CorrFiltACFilt_n, or for the considered cycle i: CorrFiltACFilt_n[i].

The transfer function of the low pass filter used is of the first order:

Y(n+1)=2^−m.X(n)+(1−2^−m). Y(n), with m a positive integer.

As an illustration, the top graph of FIG. 6 shows the signal s obtained for successive temporal cycles along the time axis expressed in seconds (s).

The middle graph of FIG. 6 shows the filtered correlation values CorrFilt₁ and those obtained after eliminating the continuous component, i.e., CorrFiltAC₁ obtained for successive temporal cycles along the time axis expressed in seconds (s), in reference to the low frequency f₁ (the values of CorrFRAC, are indeed zero in the zone Z).

The bottom graph of FIG. 6 shows the filtered correlation after eliminating the continuous component, i.e., CorrFiltACFilt₁, obtained in reference to the low frequency for successive temporal cycles along the time axis expressed in seconds (s).

In a step 105, so as to be able to differentiate between transitional ratings and the appearance of an arc, the derivative of the correlation is calculated (the correlation being expressed by one of the previously determined correlation parameters Corr_n, CorrFilt_n, CorrFiltAC_n[i] or CorrFiltACFilt_n), which amounts to applying a high pass filter.

One thus calculates DiffCorr_n[i]=|CorrFilt_n[i]−CorrFilt_n[i−1].

Next, in a step 106, a low pass filter is applied to DiffCorr_n to smooth the signal and avoid frequent passages past zero. The signal provided at the output of that filter for the cycle i is DiffCorrFiltn[i].

The transfer function of the low pass filter is of the first order:

Y(n+1)=2^−t.X(n)+(1−2^−t). Y(n), where t is a positive integer.

As an illustration, the top graph of FIG. 7 shows the signal s obtained for successive temporal cycles along the time axis expressed in seconds (s).

The middle graph of FIG. 7 shows the values representative of the correlation (in the considered case, the correlation is represented by the parameter CorrFilt₁) obtained for successive temporal cycles along the time axis expressed in seconds (s), in reference to the low frequency f₁.

The bottom graph of FIG. 7 shows the filtered derivative of the correlation DiffCorrFilt₁, obtained at the end of steps 105 and 106, this time based on the parameter representative of the correlation, CorrFilt₁, in reference to the low frequency f₁, for successive temporal cycles along the time axis expressed in seconds (s).

On this bottom graph, it is then possible to see the zone Z1 corresponding to zero values or values very close to zero of DiffCorrFilt₁, which is a transitional rating, zone Z2 in which the values of DiffCorrFilt₁ are non-zero and much higher than the values in the transitional rating, and which is representative of the presence of an arc.

Furthermore, observing the distribution of the samples delivered by a filter F_n, n=1 to 3 on the one hand, and the number of the samples strictly greater than zero on the other hand, makes it possible to differentiate between a normal operating mode and an arc fault.

In particular, more frequent passages past the zero value of the samples occur when an arc is present.

To that end, in a step 107, for n=1 to 3, the value of the parameter count_time_n[i], which is equal to the largest number of successive samples greater than zero, is determined for the current cycle i during the cycle. This arrangement makes it possible to detect the shape of the distribution of the samples obtained for each frequency f_n.

In a step 108, a filter is next applied to the signal count_time_n to smooth the signal and avoid passages by zero.

The transfer function of the low pass filter is of the first order:

Y(n+1)=2^−v.X(n)+(1−Y(n), where v is a positive integer.

The parameter count_timeFiltn[i] is delivered at the output of that filter.

As an illustration, the top graph of FIG. 8 shows the signal s obtained for successive time cycles along the time axis expressed in seconds (s).

The middle graph in FIG. 8 shows the evolution over time, for successive cycles k=i, i+1 etc., of the signal s, samples Ech_1k at the output of the filter F₁ (in solid lines), Ech_2k at the output of the filter F₂ (in broken lines), Ech_3k at the output of the filter F₃ (in dotted lines).

The bottom graph in FIG. 8 shows the evolution over time, for successive cycles k−i, i+1, etc., of the parameters count_timeFilt₁ (circles), count_timeFilt₂ (squares), count_timeFilt₃ (triangles) computed for those cycles.

As indicated above, the number of successive samples at the output of the filter having a value different from zero may provide discriminatory information regarding the presence of an arc. In particular, a zero value, or on the contrary a very high value, is generally synonymous with the absence of an arc.

In another embodiment, the counting of the successive values different from zero is done over the duration of a cycle portion, for example a half-cycle, in place of the duration of a cycle. This allows the earliest possible detection of the presence of an arc.

The parameters computed for the i^th considered cycle for the three processing chains 20, 21, 22 respectively associated with the low, medium and high frequencies f₁, f₂ and f₃ are delivered to the activating unit 23.

The activating unit is suitable for detecting the presence or absence of an arc as a function of at least some of these parameters computed for at least the cycle i, considered independently for each frequency and/or combined with each other for a given frequency and/or combined with each other for several frequencies, and for sending a command to open the switches 11 to the actuator 9 when an arc is detected.

In one embodiment, this detection is done as a function of rules applied to these parameters.

These rules may apply to the parameters considered independently by frequency f₁, f₂ or f₃, or the parameters may be combined for a given frequency and/or the parameters may be combined for several frequencies.

For example, the rules may comprise the following rules, applied over a single cycle i, or in combination taking their application over several cycles into account:

• • Rule 1: if CorrFiltACFilt₁>SeuilCorr₁ OR if CorrFiltACFilt₂>SeuilCorr₂ OR if CorrFiltACFilt₃>SeuilCorr₃: the presence of an arc is detected (the values of SeuilCorr_n, n=1 to 3 are predetermined thresholds);
  • Rule 2: if DiffCorrFilt₁ OR DiffCorrFilt₂ OR DiffCorrFilt₃ is non-zero: the presence of an arc is detected.
  • Rule 3: if count_time_Filt₁ OR count_time_Filt₂ OR count_time_Filt₃<seuil_count: the presence of an arc is detected;
  • Rule 4: if CorrFiltACFilt1>SeuilCorr1 and if CorrFiltACFilt2>SeuilCorr2 and if CorrFiltACFilt3>SeuilCorr3: the presence of an arc is detected (the values of SeuilCorrn, n=1 to 3, are predetermined thresholds);
  • Rule 5: if CorrFiltACFilt1>SeuilCorr1 Or (if CorrFiltACFilt2>SeuilCorr2 and if CorrFiltACFilt3>SeuilCorr3): the presence of an arc is detected (the values of SeuilCorrn, n=1 to 3, are predetermined thresholds);
  • Rule 6: if (CorrFiltACFilt1>SeuilCorr1 and if CorrFiltACFilt2>SeuilCorr2) Or if CorrFiltACFilt3>SeuilCorr3: the presence of an arc is detected (the values of SeuilCorrn, n=1 to 3 are predetermined thresholds);
  • Rule 7: (if CorrFiltACFilt1>SeuilCorr1 and CorrFiltACFilt3>SeuilCorr3) Or if CorrFiltACFilt2>SeuilCorr2 and if: the presence of an arc is detected (the values of SeuilCorrn, n=1 to 3 are predetermined thresholds);
  • Rule 8: if CorrFiltACFilt1>SeuilCorr1 and if CorrFiltACFilt2>SeuilCorr2 and if DiffCorrFilt3 is non-zero and if count_time_Filt2<seuil_count: the presence of an arc is detected (the values of SeuilCorrn, n=1 to 3 are predetermined thresholds);

These rules are only examples. An infinite number of rules can be determined, combining certain parameters computed above and optionally other criteria.

According to the invention, the value of at least some of the different computed parameters relative to each frequency f₁, f₂, f₃ and, if applicable, the comparison with thresholds, make it possible to detect arcs that were not perceived using the techniques of the prior art.

In the considered embodiment, the electrical installation included a phase conductor and neutral conductor, but the invention may of course be implemented for any type of electrical installation, for example an installation with 3 phase conductors and one neutral conductor.

In the described embodiment, the set of filters included three filters. The number of filters may be 3 or more (it is possible to perform an arc detection with a single filter, but for robustness reasons 3 are used, therefore the number of filters >=3 risks limiting the coverage of the patent), as well as the number of processing chains operating in parallel on the signal upstream from the activation unit 23.

In the embodiment, the bandpass filters have a switched capacity and perform the filtering and sampling. In other embodiments, the filtering and sampling may be done separately.

In embodiments, only some of the steps described above are carried out. It is not necessary for all of the indicated parameters to be calculated, and other parameters may be taken into account for the detection.

Claims

1. An electric arc detection method in an electrical installation providing an electrical signal, comprising the following steps respectively carried out in a current window of a plurality of time windows for observing the electrical signal: i/ applying, to said electrical signal in parallel with at least one first filtering centered on a first frequency, a second filtering centered on a second frequency and a third filtering centered on a third frequency, said first, second and third frequencies being distinct;ii/ determining, for said window, at least samples resulting from the first filtering, samples resulting from the second filtering and samples resulting from the third filtering;iii/ determining the correlation between said samples determined for the current window and samples determined following the implementation of steps i/ and ii/ for at least one other window from the plurality of time windows for observing the electric signal; andiv/ detecting the presence of an electric arc as a function of at least said determined correlation; wherein step iii/ for determining the correlation comprises determining at least one respective correlation from among: a first correlation between said samples resulting from the first filtering and determined for the current window and samples resulting from the first filtering and determined for the other window;a second correlation between said samples resulting from the second filtering and determined for the current window and samples resulting from the second filtering and determined for the other window;a third correlation between said samples resulting from the third filtering and determined for the current window and samples resulting from the third filtering and determined for the other window; and

2. The method according to claim 1, wherein step iv/ for detecting the presence of an arc is a function of comparison of the determined correlation from among said first, second and third correlations, with a threshold.

3. The method according to claim 1, wherein step iii/ for determining the correlation comprises determining at least each of said first, second and third respective correlations.

4. The method according to claim 3, wherein step iv/ for detecting the presence of an arc is a function of at least each of said first, second and third correlations.

5. The method according to claim 1, wherein the difference between said respective correlation determined among said first, second and third correlations for one window and said respective correlation for another window is calculated, step iv/ for detecting the presence of an arc being a function of at least said calculated difference.

6. The method according to claim 1, wherein: the number of successive samples with a value greater than zero is calculated among the samples resulting from a respective filtering among the first, second and third filterings, for the current window;said number is compared to a threshold;the presence of an electric arc is detected as a function of at least said comparison.

7. The method according to claim 1, wherein none of said first, second and third frequencies are equal to an integer that is a multiple of another of said frequencies.

8. A computer program for detecting an electric arc including instructions for implementing the steps of a method according to claim 1, when the program is run by computation means.

9. A device for detecting an electric arc, in an electrical installation providing an electrical signal, comprising a filtering-sampling unit suitable, during a time window for observing the electric signal, for filtering said signal in parallel according to a first filtering centered on a first frequency, according to a second filtering centered on a second frequency, and according to a third filtering centered on a third frequency, said first, second and third frequencies being distinct, and for determining, for said time window for observing the signal, signal samples resulting from the first filtering, signal samples resulting from the second filtering and signal samples resulting from the third filtering;the detection device further comprising a processing unit suitable for determining a correlation between said samples determined for a current time window for observing the electrical signal and samples determined for at least one other time window for observing the electrical signal, and for detecting the presence of an electric arc as a function of at least said determined correlation;wherein the processing unit is suitable for determining said correlation by determining at least one respective correlation from among: a first correlation between said signal samples resulting from the first filtering determined for the current window and signal samples resulting from the first filtering determined for the other window;a second correlation between said signal samples resulting from the second filtering determined for the current window and signal samples resulting from the second filtering determined for the other window;a third correlation between said signal samples resulting from the third filtering determined for the current window and signal samples resulting from the third filtering determined for the other window; and

10. The electric arc detecting device according to claim 9, wherein the processing unit is suitable for detecting the presence of an arc as a function the comparison of the determined correlation from among said first, second and third correlations, with a threshold.

11. The electric arc detecting device according to claim 9, wherein the determination of the correlation by the processing unit comprises determining at least each of said first, second and third respective correlations.

12. The electric arc detecting device according to claim 11, wherein the determination of the correlation by the processing unit comprises determining at least each of said first, second and third respective correlations.

13. The electric arc detecting device according to claim 9, wherein the processing unit is suitable for calculating the difference between said respective correlation determined among said first, second and third correlations for one window and said respective correlation for another window, and for detecting the presence of an arc as a function of at least said calculated difference.

14. The electric arc detecting device according to claim 9, the processing unit being suitable for calculating the number of successive samples with a value greater than zero among the samples resulting from a respective filtering among the first, second and third filterings, for the current window, comparing said number to a threshold, and detecting the presence of an electric arc as a function of at least said comparison.

15. The electric arc detecting device according to claim 9, wherein none of said first, second and third frequencies are equal to an integer that is a multiple of another of said frequencies.