FR2936319A1 - Current line ground fault directional-detection device for three-phase network in ground protection relay, has calculation device to calculate average and standard deviation, and comparison device to compare average and deviation - Google Patents

Abstract

The device (20) has a signal processing module (30) comprising a calculation device (34) for calculating standardized correlation coefficients (rA-rC) between a signal representative of a homopolar current and signals representing phase currents (I-A-I-C) in a predetermined duration. The module includes another calculation device (36) for calculating an arithmetic average (mu) and a standard deviation (sigma) between absolute values of the calculated correlation coefficients. An interpretation module (40) comprises a comparison device (38) for comparing the average and the standard deviation. Independent claims are also included for the following: (1) a ground fault passage indicator comprising current sensors (2) a ground protection relay comprising an activation unit (3) a method for directional-detection of a ground fault in a multiphase electrical network (4) a method for protecting a current line during appearance of a ground fault.

Description

The invention relates to the directional detection of a ground fault without measuring the voltage of the line. In particular, the invention relates to a method for detecting a ground fault in a network which also makes it possible to determine whether the earth fault is located upstream or downstream of the detection point. The method according to the invention is based on the only signals representative of the currents of each phase of the network, whose processing results in parameters allowing the directional localization.

In another aspect, the invention relates to a detection device adapted to implement the above method. In particular, the directional earth fault detection device comprises means for calculating parameters from the current signals of each phase, the interpretation of said parameters giving the relative location of the fault without using values representative of the voltage. between phases, nor of values representative of simple voltages.

Finally, the invention relates to a fault signaling device and a tripping relay comprising current sensors associated with each phase of the network and providing the preceding detection device with the signals enabling the signaling, for example by light, or the triggering of a cut-off device of the network.

STATE OF THE ART

Ground fault detection devices are particularly used in medium-voltage three-phase electrical distribution networks. FIG. 1 thus represents a diagram of an electrical distribution network 1 which comprises a three-phase translragmator 2 whose secondary is connected to a main distribution line 3; the secondary further comprises a neutral common conductor 4 generally connected to the earth by an impedance. The main line 3 feeds the start lines 5, 5 ', 5 ", some of which may have a circuit breaker or other cut-off device 6 protecting them at the beginning of the line 5. The starting lines 5', or line sections 5, may comprise a device 7 for detecting an earth fault The device 7 can serve as a fault passing indicator, for example lighting a warning light 8, a device 7 can also be associated with or integrated with a protective relay 9 capable of to order Opening of circuit breaker contacts 6.

Ideally, the network l is balanced, that is to say that the homopolar current Io circulating therein is zero; by homopolar current Io, (or zero sequence current according to the English terminology) is meant, to a possible factor three, the vector summation of the different phase currents, or the current corresponding to the instantaneous resultant of the phase currents , sometimes called residual current, which possibly corresponds to the earth fault current (gr) and default current according to the English terminology) or to the leakage current. If one of the phases is accidentally grounded, this equilibrium disappears. of the defect 10: the detection of a defect can thus consist, as illustrated in FIG. 2A, in the measurement of the homopolar current Io, for example by a torus 11 surrounding the line 5, and the comparison of the signal representative of said current Io with a threshold Sd, the exceeding of the threshold being understood as the detection D of a fault 10. Alternatively, the earth fault current Io can be obtained by summing the current signals of each phase, a current sensor 12 being then located on each conductor of the line 5 (see Figures 2B and 2C).

However, and particularly if the starting ligaments are buried cables, high capacitance values may occur between the line conductors 5A, 5B, 5c and the ground.

Thus, in the case of the presence of a fault 10 on the ground on a line 5, capacitors 13 are at the origin of the circulation of large homopolar currents Io on the other starting lines 5 ', 5 "(which does not have no earth fault), or downstream of the fault 10. These (relatively) strong fault currents Io can be the cause of false detections by the detection devices 7 present on the neighboring lines 5 'not faulty.

It is therefore important to distinguish a fault 10 located downstream of a detection device 7 ;, a fault 10 upstream of the device 7; + - which can detect de facto a failure of a neighbor departure by capacitive link; this differentiation is not feasible by the device shown diagrammatically in FIG. 2A, especially in the case of a compensated neutral network 1 (that is to say that the neutral 4 is grounded by a compensation coil) which generates a residual current Io of insufficient value, in particular in the case of buried cables.

For such directional fault detections, the detection devices 7, as illustrated more precisely in FIG. 2B, are based on the measurement of the current but also of the voltage of the line 5. The detection system thus comprises current 12A, 12B, 12C supplying the signals representative of currents IA, IB, IC flowing in each of the phases of line 5 and of voltage sensors 14A, 14B, 14C, in particular in the form of voltage transformers, which provide signals representative of the voltages VA, VB, VC of each phase conductor. The detection device 7 comprises means 15 which condition the data of said sensors 12, 14 and a processing module 16 which, on the basis of the signals representative of the homopolar current Io and the homopolar voltage Vo thus obtained, makes it possible to detect at the times the presence D and the relative location L of a fault 10.

However, voltage transformers 14 are both bulky and expensive; moreover, they are not always necessarily adapted for installation on lines already existing. For an economically viable implementation, it is preferable for the fault detection devices 7 to be devoid of measuring means 14 or of processing the voltages of the network 1.

Thus, as illustrated in FIG. 2C, earth fault devices 17 have been developed in which the voltage is not used. For example, document EP 1 475 874 discloses a directional earth fault detection device for which the processing module 18 uses, in addition to the signal representative of the residual current I o, the signal representative of the current reversing each of the signals being compared to a threshold for signaling a ground fault downstream of the current sensors 12. The detection of an upstream fault, however, requires a second processing unit, and the transition to complex components further increases the detection device 17.

Another approach has been developed in EP 1 890 165 in which the processing module 19 compares the shape of the homopolar current Io with the phase currents IA, IB, the to determine whether the detected earth fault is downstream or not. However, this comparison is based on a neural network 19, a complex system whose learning base strongly conditions the diagnosis of detection D and direction L (a training base in inadequacy with the final network 1 may lead to erroneous detection results). Moreover, the learning base requires fixed coefficients (weight and bias factors of the neural network), which correspond to a specific operating mode, such as the neutral system of the network, the trigger level of the detector: any modification of adjustment or operating parameters of the detection device 17 may then require a new learning, complex to achieve, in particular by the teams in charge of the installation or maintenance of such equipment.

It thus appears that the existing directional earth fault detection devices 7, 17 are not optimized for wide implementation because of their complexity, that this is due to the number of sensors 12, 14 to be used. place or system 18, 19 signal processing.

SUMMARY OF THE INVENTION

Among other advantages, the invention aims to overcome the disadvantages of devices and methods for directional detection of existing earth fault. In particular, the directionality principle used is based on the analysis of correlation coefficients between the signal representative of the homopolar current and the signals representative of the phase currents, without using the different voltages of the network.

In one of its aspects, the invention relates to a directional earth fault detection method in a multiphase network, preferably three-phase, comprising a first fault detection stage by comparison of a signal representative of the circulating homopolar current. the section of line monitored at a detection threshold. The signal representative of the homopolar current can be obtained directly, or by calculation from the signals representative of the currents of each phase conductor of said section.

If the first stage detects the presence of a ground fault in said section, the second stage of the method according to a preferred embodiment of the invention is triggered. The second stage is based on the processing of the signals representative of the currents of each phase of said section and of the fault current of the same section, these signals being obtained over a sufficient predetermined duration, for example an integer number of half-periods of the network. Advantageously, the signals representative of the phase currents are filtered, in particular analogically, and / or sampled, preferably at a frequency making it possible to obtain at least twenty points over the predetermined duration, for example of the order of 1 kHz for a network at 50 Hz and half a period. The signal representative of the homopolar current can here also be obtained directly, or by calculation from the signals representative of the currents of each phase conductor of said section, before or after filtering and / or sampling.

After the acquisition of signals representative of different currents, the second stage of the process continues with a processing of said signals to allow interpretation of whether the fault detected in the first stage is upstream or downstream of the measurement point of the currents of phase. The signal processing comprises, according to the invention, the calculation of normalized linear correlation coefficients between the signal representative of the homopolar current of the section and each of the signals representative of the phase currents; preferably, the Pearson-Bravais formula and / or unsigned coefficients are used. The dispersion of said coefficients is then analyzed, in particular by calculating their mean and their standard deviation.

If the dispersion of the unsigned linear correlation coefficients is high, for example with a standard deviation greater than 0.3 and an average greater than 0.5, then the defect is located downstream of the detection apparatus. Conversely, for a lesser dispersion, in particular an average of less than 0.5 and a standard deviation of less than 0.2, then the fault is located upstream of the detection apparatus.

Preferably, the dispersion of the coefficients is achieved by formulating a comparative relationship. In particular, if the difference between the standard deviation multiplied by root of three and the difference between the mean and one is positive, then the defect is upstream of the detection; any other comparison based on the inequality 3.6 '> (1 - p) 2 can be used.

According to another option, the comparison is made graphically with respect to a line of equation y = 1u x: according to whether the mean and standard deviation point is positioned or not in the half-plane containing the origin, the defect is upstream or downstream.

According to a preferred embodiment of the method according to the invention, said directional detection method is associated with an actuation of a cut-off device for isolating the section from the point downstream of which a fault has been detected.

In another aspect, the invention relates to a directional detection device for a ground fault of a line in a multiphase network, preferably three-phase, adapted for the above method. The directional detection device according to the invention can be associated with current sensors of each of the phases of the line, for example detection cores, which provide the signals representative of said currents. The directional detection device may furthermore be part of a fault-passing indicator, for example by activating light-type warning means if a defect downstream to the sensors is detected. In a particularly preferred embodiment, the directional detection device according to the invention is associated with a protection relay of the line, the warning means causing the actuation of a cutoff device of the line to isolate the section on which a fault has been detected.

In particular, the directional detection device of a ground fault according to the invention comprises first means for receiving signals representative of currents of each phase of the line to be monitored as well as second means for receiving a signal representative of the current homopolar of said line, these second means preferably comprising means for deriving signals representative of the phase currents, in particular by summation, said signal representative of the homopolar current. Advantageously, the first signal reception means representative of the phase currents are associated with means for filtering said signals, for example an analog filter; preferably, the first means comprise sampling means to obtain a sufficient number of discrete values, for example at a frequency of 1 kHz.

The device according to the invention comprises representative signal processing means obtained associated with activation means of said processing means, the activation means being triggered by the detection of the occurrence of a ground fault. Preferably, the detection of the occurrence of a ground fault actuating the activation means is carried out by the device according to the invention which comprises suitable means, in particular means for comparing the signal representative of the homopolar current with a detection limit.

The signal processing means of the device according to the invention comprise means for calculating the normalized, preferably unsigned linear correlation coefficients, preferably according to the Bravais-Pearson formula, between the signal representative of the zero sequence current and each of the representative signals. phase currents. Preferably, the calculation means are associated with timing means for acquiring the signals for a period corresponding to an integer number of half-periods of the network. The calculated data is transmitted to a module for calculating the arithmetic mean and the standard deviation, coupled to output means of interpretation comparing the results to determine whether the defect is upstream or downstream of the point of obtaining signals representative of the phase currents; the means of interpretation may be of the graphical or computational type.

BRIEF DESCRIPTION OF THE FIGURES Other advantages and features will emerge more clearly from the following description of particular embodiments of the invention, given by way of illustration and in no way limiting, represented in the appended figures.

Figure 1, already described, shows an electrical network in which earth fault detection devices can be used.

FIGS. 2A, 2B and 2C, already described, show block diagrams of ground fault detection devices according to the prior art.

FIG. 3 shows schematically and filtered signals representative of the phase and homopolar currents at the occurrence of a ground fault on a phase respectively upstream and downstream of the detection device. Figure 4 illustrates the detection method according to a preferred embodiment of the invention.

Fig. 5 shows a block diagram of a ground fault detecting device according to a preferred embodiment of the invention.

DETAILED DESCRIPTION OF A PREFERRED EMBODIMENT

The directional fault detection device 20 according to the invention can be used in any multiphase network 1, such as that described in FIG. 1, in place of the existing devices 7, 17. In the illustrated embodiment and preferred embodiment of the invention, the line 5 on which the device 20 is put in place comprises three phase conductors 5A, 5B, 5c and the network is balanced, that is to say say that the homopolar current Io is zero in the absence of fault. However, it is possible to deviate from this ideal situation, and the network may comprise another number of phases.

Upon the occurrence of a ground fault on one of the phases A, the current of said phase IA becomes (ideally) zero downstream of the fault 10, and its amplitude increase upstream. As illustrated in FIG. 3, a current sensor 12.A on the faulty conductor 5A thus provides a signal representative of the current IA marking a clear break at the fault 10, and different depending on whether the device according to the invention is located upstream 20i + i or downstream 20i of said sensor 12. Figure 3 also shows that the current is punctually altered in the other two phases B and C, but remains almost constant. FIG. 3 finally illustrates the homopolar current I o detected on line 5, either by a suitable sensor 11 (see FIG. 2A), or by calculation from the three phase currents IA, IB, IC obtained by the sensors 12.

According to the invention, the correlation coefficients between the different curves are used to determine whether the fault 10 is located upstream or downstream of the sensors 12 providing signals representative of the currents of the line 5, preferably filtered. In fact, it is found that, after occurrence of the earth fault, in the case of a fault 10 occurring downstream of the device 20, the resemblance between the zero sequence current Io and the current IA of the phase in error is important, unlike the resemblance, mean, between the homopolar current Io and currents IB, the other phases; thus, the normalized correlation coefficients with the homopolar current Io should be, in absolute value, close to unity for phase A and average, for example close to 0.5, for the other phases.

Conversely, in the case of a fault 10 occurring upstream of the device 20 + i, after its occurrence, the resemblance of the homopolar current Io with the current IA of the faulty phase is small while the resemblance between the zero sequence current Io and IB currents, the other phases remain average; thus, the normalized correlation coefficients not signed with the homopolar current Io should remain average, for example close to 0.5, for the phases B and C, and be close to zero for the phase A. Preferably, the signals are analyzed. over a sufficient period Tacq, including acquisition of signals greater than half a period of the network, for example a period or any integer number of half-periods.

The normalized linear correlation coefficient used in the process according to the invention is preferably obtained by the Bravais-Pearson formula and unsigned: in fact, the sensors 12 provide an alternating current signal flowing in each of the phases 5A, 5B, It is preferred to sample to obtain a defined number of discrete values representative of the current. Advantageously, the number N of values provided by the sampling is adapted to the acquisition duration Tacq to optimize the reproducibility and reliability of the results of the Bravais-Pearson formula, that is to say preferably more than 20. values in the case of an acquisition period Tacq equal to one half-grating period, with for example a sampling frequency of the order of one kilohertz for a three-phase network 1 at 50 Hz. Thus, in the method according to 4, once the detected fault D, for example by a method similar to that described with reference to FIG. 2A, the signals representative of the phase currents, preferably filtered, are acquired over a period of acquisition. Tacq, then sampled; in parallel, the homopolar current Io is calculated for the same duration. The correlation coefficients, here unsigned, between the signals representative of the sampled filtered phase currents IAf *, IBf *, Icf * and the signal representative of the homopolar current Io are calculated by the formula of Bravais-Pearson according to the equation (1 ), in which rxy denotes the unsigned linear correlation index between the two variables X, Y, of which respectively N point values x, y are known. ## EQU1 ## where ## EQU1 ## where ## EQU1 ## is obtained thus giving rise to three unsigned normalized linear correlation coefficients. rA, rB, rc. As indicated above, depending on whether the defect 10 is upstream or downstream, the coefficients rA are closer to 1 or 0. According to the invention, to evaluate this proximity, the arithmetic mean and the difference between type a of the distribution of the correlation coefficients are used. In fact, a high average mean / standard deviation (, a) (for example> 0.5 and a> 0.3) corresponds to a downstream defect, unlike a mean / standard deviation (, a) with values low (for example <0.5 and a <0.2) which corresponds to an upstream fault. (1) rxy = 25 More particularly, it can be seen that, for a fault 10 downstream of the detection 20, the following relation (2) is satisfied, whereas for an upstream fault 10, 20i + i, the inverse is true, ie: 3.62 1 - 02. (2) 'x6> (u, u) In a preferred method according to the invention, after calculating the mean and standard deviation, the comparison according to the relation (2), or any relation derived directly from it, is performed and, depending on the result, the directional interpretation L as to the position of the defect is given, either signaled externally or transmitted to a relay 9 for cut a circuit breaker 6 upstream, or any other use.

To lighten this calculation step of the method according to the invention, it is possible graphically to determine the upstream or downstream relative location of the defect 10 by positioning the point of coordinates (, 6) on a graph (x, y) where is plotted a equation right (3): if the point is in the half plane containing the origin (0,0), it is an upstream fault. The method according to the invention can be implemented in a protection relay 9, in a fault indicator with warning system 8, by implementation in a suitable directional earth fault detection device 20. A device 20 according to a preferred embodiment of the invention is shown schematically in FIG. 5. It comprises means 22 making it possible to obtain signals representative of the phase currents supplied by suitable sensors 12, for example detection cores, advantageously with filtering means 24 adapted as an analog filter. In the preferred embodiment, the filtered signals IA, 1g ,, IcF are further conditioned by sampling and the means for obtaining the representative signals 22 comprise a sampling module 26, advantageously operating at more than 1 kHz. thus providing filtered sampled signals to process IAF *, IBF *, I, F *. The device 20 according to the invention further comprises means 28 for obtaining a signal representative of the homopolar current Io. These means can be coupled directly to a current sensing toroid surrounding the line 5 (not shown); preferably, the means 28 relating to the value of the homopolar current I0 determine said current by processing the signals relating to the phase currents IA, IB, If. (not shown), preferably filtered I, IB, ICf (not shown), optionally sampled IAf *, IBf *, I ,, f * (Figure 5).

In addition to entering the determination means 28 of the signal representative of the zero sequence current Io, the signals representative of the filtered sampled phase currents IAf *, IBf *, Icf * are transmitted to a processing module 30 which includes a fourth input for the signal representative of the homopolar current Io, and which is activated according to the detection D of a ground fault 10. For this purpose, the processing module 30 is connected to any fault detection device 32, preferably here a homopolar current comparison module I ,, with a detection threshold Sd: if the threshold is exceeded, then a fault D is detected and the processing module 30 is activated.

The processing module 30 comprises successively a computing device 34 of unsigned normalized correlation coefficients r from the four input signals IAf *, IBf *, Icf *, h, a device 36 for calculating the arithmetic mean and the standard deviation a of the three data inputs rA, r8, rc, a comparison device 38 with two inputs 4, 6 and which is connected to interpretation means 40 whose output is a directional fault detection signal L earth downstream or upstream of the sensors 12 according to the result of the interpretation.

The means for calculating unsigned normalized correlation coefficients 34 preferably use the Bravais-Pearson formula (1) and are associated with delay means to ensure that the signals representative of the phase currents IAf *, IBf * , Icf * were acquired over a sufficient period Tacq, for example a half-period or a period of network 1, or more.

The comparison means 38 supplying the value to the interpretation module 40 may use different functions. In particular, the comparison means 38 may comprise a graphical comparison by two values entered with respect to a half-plane delimited by the line of equation (3); the means 38 can determine the difference between the triple of the square of the standard deviation and the square of the difference between one and the average, to compare it to zero; any other option is possible.

The device 20 of FIG. 5 may advantageously be associated with a protective relay 9 for electrical networks, or with a fault-passing indicator for underground medium voltage lines connected in a network 1, the output of the triggering interpretation module. breaking a circuit breaker 6, lighting a warning light 8 or any other means of security and / or warning.

Thus, according to the invention, a method and a device for directionally detecting a ground fault of a line 5 of a multiphase network 1 have been produced without voltage measurement, which lightens the devices as much as their implementation.

Although the invention has been described with reference to a three-phase distribution network in which the neutral is grounded by compensated impedance, it is not limited thereto: other types of multiphase networks may be concerned by the invention; in particular, any neutral scheme is appropriate. Furthermore, although described with determination and treatment of the instantaneous homopolar current Io, the method according to the invention can use the variation of said current Io with respect to its value determined over a prior period: this variant is particularly interesting in the case networks with a slight imbalance between phases, whose homopolar current Io is therefore not zero in non-default situation.

In fact, the various circuits, modules and functions presented in the context of the preferred embodiment of the invention can be made of analog, digital or programmable components operating with microcontrollers or microprocessors, and the representative signals described can having forms of electrical or electronic signals, data or information values in memories or registers, optical signals that can be displayed in particular on LEDs or screens, or mechanical signals acting with actuators. Similarly, the current sensors may be different from the described transformers, such as Hall effect sensors or magnetoresistors.

Claims (16)

REVENDICATIONS1. Apparatus (20) for directionally detecting a ground fault (10) in a multiphase network (1) comprising: first means (22) for receiving signals representative of the current (IA, IB, Ic) of each of the phases ; second means (28) for receiving a signal representative of the homopolar current (Io) of all the phases; signal processing means (30, I1, IB, Ic) representative of current; means for activating said processing means (30) as a function of a signal (D) for detecting the occurrence of a ground fault in the network (1); means (40) for interpreting the results of the signal processing to determine whether the fault is upstream or downstream of the device; characterized in that the signal processing means (30) comprises: means (34) for calculating the normalized correlation coefficients (rA, rB, rc) between the signal representative of the homopolar current (Io) and each of the signals representative of phase currents (IA, IB, Ic) over a predetermined period of time (Tacq); means (36) for calculating the mean (p) and the standard deviation (a) between the absolute value of the calculated correlation coefficients (rA, rB, rc), the interpretation means (40) comprising means for comparing said average (p) and said standard deviation (a). 2. directional detection device (20) according to claim 1 wherein the second means (28) comprises means for summing the signals representative of each of the phase currents (IA, IB, Ic) to provide a signal representative of the zero sequence current (Io). 3. Device according to one of claims 1 or 2 further comprising means (32) for detecting the occurrence of a ground fault in the network connected to the activation means of the signal processing means (30). ) comprising a comparator of the signal representative of the homopolar current (Io) at a detection threshold (Sd). 15 4. directional detection device (20) according to one of claims 1 to 3 wherein the means for calculating the correlation coefficients (34) are associated with signal acquisition means for a period (Tacq) corresponding to a integer number of half-periods of the network (1). Directional detection device (20) according to one of Claims 1 to 4, in which the first signal reception means (22) representative of the currents of each phase (IA, IB, Ic) comprise sampling means ( 26) adapted to provide a sufficient number of values for the predetermined duration (Tacq). Directional detection device (20) according to one of claims 1 to 5, comprising an analog filter (24) connectable at the input to the phase conductors of the network (5A, 5B, 5c), said filter (24) being coupled to the first signal receiving means (22). 7. A ground fault crossing indicator comprising current sensors (12A, 12B, 12c) arranged on each phase conductor (5A, 5B, 5c) of an electrical network (1) to be monitored and comprising a control device. directional fault detection (20) according to one of claims 1 to 6 connected to said current sensors (12A, 12B, 12c) for receiving the phase current representative signals (IA, IB, Ic). 8. Earth protection relay (9) comprising at least one fault indicator according to claim 7 and means for actuating a cutoff device (6) according to the results of the interpretation means (40) of the directional detection device (20) of the indicator. 9. A method for directionally detecting (D, L) a ground fault (10) in a multiphase network (1) comprising triggering, after obtaining a signal (D) indicating the presence of said defect to the earth (10), the directional determination (L) of the defect (10), said directional determination comprising the successive steps of: obtaining signals representative of each of the phase currents (IA, IB, Ic) over a period of time ( Tacq) of at least half a period of the network; obtaining a signal representative of the homopolar current (Io) circulating in the network over the same duration (Tacq); processing signals representative of the phase and ground fault currents (Io, IA, IB, Ic) interpreting the signal processing results to indicate whether the detected fault (D) is downstream or upstream of the location where the signals representative of the phase currents (IA, IB, Ic) have been obtained; characterized in that said signal processing step comprises: calculating the unsigned normalized correlation coefficients (rA, rB, rc) between the zero sequence current signals (Io) and each of the phase currents (IA, IB, Ic ) over the predetermined duration (Tacq); calculating the arithmetic mean (g) and the standard deviation (6) of said correlation coefficients (rA, rB, rc); said interpretation being performed by a comparison using said average (g) and said standard deviation (a). 10. Directional detection method according to claim 9 wherein the signal (D) indicating the presence of a ground fault (10) is obtained by receiving a signal representative of the homopolar current (Io) flowing in the network (1) and the result of a comparison of the zero sequence current signal (Io) with a fault detection threshold (Sd). 11. The method of directional detection according to one of claims 9 or 10 wherein obtaining a signal representative of the hornopolar current (Io) flowing in the network (1) comprises calculating said current as a function of the signals representative of each phase currents (IA, IB, Ic). 12. Directional detection method according to one of claims 9 to 11 wherein the step of interpretation comprises the comparison with zero of the difference between letriple of the square of the standard deviation (a) and the square of the difference between the mean (p) and one. 13. Directional detection method according to one of Claims 9 to 11, in which the interpretation step comprises the graphical comparison of the mean (p) and the standard deviation (a) with an equation line. y = 1 x 14. Directional detection method according to one of claims 9 to 13 wherein the step of providing the signals representative of the phase currents (IA, IB, Ic) 10 comprises a sampling of the current at a frequency greater than 1 kHz. 15. directional detection method according to one of claims 9 to 14 wherein the calculation of normalized correlation coefficients (rA, rB, rc) is carried out according to the formula of Bravais-Pearson. 16. A method of protecting a current line (5) during the occurrence of a ground fault (10) comprising actuating a cut-off device (6) of said line (5) if a Earth fault (10) has been detected by a method according to one of claims 9 to 15 downstream of said cut-off device (6). 15 20