FR2926370A1 - Interior arc detecting method for e.g. gas-insulated metal-clad substation, involves locating arc with help of information obtained by recognition of light signal and by recognition of pressure waves - Google Patents

Abstract

The method involves positioning a pressure sensor and an optical sensor in each compartment of a high voltage device, and detecting an arc by recognition of a light signal emitted by the arc, with the help of the optical sensor. The arc is detected with the help of the pressure sensor by recognition of pressure waves emitted by the arc in gas e.g. sulfur hexafluoride. The arc is located with the help of information obtained by the recognition of the light signal and by the recognition of the pressure waves, where the information are transmitted to a processing unit. An independent claim is also included for a device for detecting an interior arc in a high voltage device.

Description

DEVICE AND METHOD FOR INNER ARC DETECTION BY COMBINED PRESSURE AND LIGHT MEASUREMENT

TECHNICAL FIELD AND PRIOR ART The invention relates to the field of medium or high voltage equipment. In this type of equipment, faults can occur. It relates more particularly to a method and a device for the detection of an electrical fault, and more specifically of an internal arc, in an electrical connection or in an electrical station, at medium or high voltage, in a metal envelope (also called PSEM or GIS) or in a gas-insulated line (also called LIG or GIL). The detection of the internal fault arc (as opposed to a cut-off arc) inside high voltage equipment is necessary for safety, maintenance and operation. An existing detection method uses the currents: the measurement of the incoming and outgoing currents of a portion of GIS (for example a busbar) allows by summation (which is zero in normal time) to reliably and quickly detect a defect arc, but without being able to indicate precisely which compartment is in fault in the zone. This principle is quite suitable for protection systems (differential protection), thus answering security questions thanks to short detection times (less than 20 ms), without, however, satisfying maintenance requirements. As, on the one hand, the HT equipment is most often imposing and limited access and that, on the other hand, the intervention time must be short for a quick restart, we try to locate the compartment or the zone in default. This operation is then possibly possible only by means of a portable gas analyzer, to move from one valve to the other. As the number of valves of a given section can be large (several tens), this operation is tedious. Such a system therefore meets the need for security, but is expensive because dedicated to this function.

It requires the presence of current transformers in order to perform the measurements. Another type of known system operates on an optical principle. It is shown schematically in Figure 10 which shows an application of this type of system to a metal shell of a GIS. In this figure 10 is shown a GIS which comprises a metal shell 2 compartmentalized and a central conductor 4, isolated by a pressurized gas, for example nitrogen and / or sulfur hexafluoride SF6. The compartments are delimited by insulating supports 6, which, in addition, maintain the electrical conductor 4. On each compartment is mounted an optical sensor CL1, CL2, CL3, ..., individually connected to an acquisition unit 12 by a set of optical fibers FOI, FO2, FO3, Various disadvantages are inherent in the implementation of this optical method: the defects are fugitive, there is no signal memory and this method additionally requires additional sensors and judiciously placed in the installation to be monitored, it therefore results in a very high cost. This principle of optical detection is therefore widely used but has limitations.

A considerable part of HV equipment faults occur during the operation of the devices. On the basis of light, detection can be fast. But, this information is fleeting. Since the HV devices generate light during the maneuvers because of the cut-off arcs (which are not fault arcs), the arc detection system must then be obscured. The detection time of such a system is generally short (less than 20 ms), except in the case of device operation where it is blind. Another disadvantage is the difficulty of crossing insulators or active parts or elbows HT material that can lead to shadows. It is not easy to place optical fibers (which are usually used to bring back the light emitted by the arc at the acquisition unit) over long distances and under difficult environmental conditions. GIS are most often in building, on the other hand 30 GIL are outside. However, even in buildings, placing optical fibers between the GIS switchgear and the cabinets containing the acquisition system is not easy because the journey is tortuous. This technique is also unsuitable for measurements on busbars or LIGs of several tens of meters to several hundred meters. The parallel architecture of the light sensors leads to a multiplication of the number of optical cables (as can be seen in FIG. 10) and to acquisition units 12 having many inputs and therefore at very low costs. On the other hand, the optical sensors being inert (it is a simple optical conductor of 5 to 10mm section, extending the fiber inside the compartment and ending with a light collector, the light can not This is problematic because of the attenuation of the light signal inside the optical fibers.

In addition, such a system is particularly exposed to failures in the fibers FOI, FO2, FO3, .... due to their number, their limited radius of curvature, and their exposure in the maintenance areas of the earth. HT equipment.

This is problematic because of the impossibility for the system to test itself. Therefore, close and complicated maintenance campaigns should be planned because of the need to manually generate a light at the end of each fiber and to test the reaction of the acquisition unit, this having an impact on the continuity of operation of the line. In addition, this system is also expensive.

Another principle already used is based on the measurement of gas pressure inside the envelopes. The document EP 1 004 884 describes a method making it possible to detect, following the creation of such an arc, a rising edge of an internal overpressure wave created by this arc, which waves propagate in the envelope. According to this technique, an internal arc will generate heat that can be measured through an increase in pressure in the hermetically sealed volume of an envelope. As the temperature decreases relatively slowly, the pressure variation is a persistent phenomenon which makes it possible to establish a reliable diagnosis by a large number of measurement points. However, the time of transmission of the pressure wave in the gas (typically 136 m / s) leads to a delay in the signal measurement depending on the relative position of the arc and the pressure sensor. This delay may for example be greater than 20 ms in a shell of 4 linear m.

In addition, the additional filling of a compartment, which also results in an increase in pressure, can lead to a false diagnosis. Another problem, the internal arc recognition based on the use of analog sensors is, for its part, not satisfactory, because the analog measurements are sensitive to the electromagnetic disturbances generated by the devices HT, including those issued by the internal arc itself. The electromagnetic disturbances emitted during an internal arc result in the circulation of currents flowing in the envelopes of the high-voltage devices and in the earth circuits, but also in the screens of the sensor cables. These currents transmit by coupling disturbances on the wires of the cable. In the case of analog communication sensors, the measurement is disturbed. There is therefore the problem of finding a new method and a new device making it possible to carry out fault identifications more effectively than by currently known techniques. In particular, faster identification of a defect and faster intervention following this identification is desirable. Preferably such a novel method or device is adaptable to the identification of various types of defects. DISCLOSURE OF THE INVENTION The invention relates to a method for detecting an internal arc in a high-voltage device, comprising a metal envelope, at least one central conductor, a pressurized gas acting as a dielectric insulator, and a plurality of compartments traversed by this conductor, this method comprising: - a detection of the arc by recognition of a light signal emitted by the arc, - a detection of the arc by recognition of the pressure waves emitted by the arc in the gas, - the location of the arc using information obtained on the one hand by recognition of the light signal, on the other hand by recognition of pressure waves. A method according to the invention allows a precise location of the arc in the HT device, in addition to the detection of the area of this device in which the defect appeared. For example, the measurement of the time difference between light and pressure wave is converted into distance. A method according to the invention also allows rapid detection by optical means, but while enjoying the advantages of detection by pressure sensor. A method according to the invention can be implemented. either by a device, such as a transmitter, integrating the two sensors and the two detection functions, or by two separate sensors, one being an ordinary pressure sensor, the other an optical sensor, preferably all two connected to a single unit of acquisition and processing. According to one embodiment, the pressure wave detection is carried out only if a light signal is first detected. An alarm can be generated as soon as a light signal has been detected but before detection by pressure waves.

In a variant, light or a pressure wave is detected, the light and pressure data are memorized, then a signature is generated from these light and pressure data, and the signature thus obtained is compared with a signature. reference. In any case, an alarm can be triggered when an arc is detected from the light signal data and pressure data.

A method according to the invention offers the following advantages. It is operational during the operation of HT devices. Due to the conjunction of two physical phenomena (light and pressure detection), the diagnosis of presence of internal fault arcs is reliable, compared to the detection of light or measurement of pressure taken in isolation.

A method according to the invention makes it possible to reduce the cost of fault internal arc detection, in particular on busbar-type or LIG gas-insulated lines, by means of a local diagnosis (in a single transmitter) and the fusion of the communication means with the pressure sensors in the form of a fieldbus. There is also a reduction in the cost of the arc localization function and differential protection backup solution thanks to an accelerated and reliable diagnosis established on different physical phenomena.

The invention also relates to a device for detecting an internal arc in a high voltage device, comprising: optical means, arc detection by recognition of a light signal emitted by the arc, sensor means of pressure, arc detection by recognition of the pressure waves emitted by the arc in the gas, - calculation means for locating an arc using the information obtained on the one hand by recognition of the light signal, on the other hand, by recognition of pressure waves. The invention also relates to a high voltage device, comprising a metal casing, at least one central conductor and a plurality of compartments traversed by this conductor, and a device according to the invention, in particular as explained above. The invention also relates to a high voltage device, comprising a metal casing, at least one central conductor and a plurality of compartments traversed by this conductor, and pairs of sensors, each comprising a pressure sensor and an optical sensor, for implement a method according to the invention, especially as explained above. In a device according to the invention, the data processing means may comprise a switching power regulator.

A device according to the invention may further comprise means capable of measuring a temperature representative of the temperature of the metal casing. Each pressure sensor may be of capacity variation type or piezoelectric type. In a method and device according to the invention, pressure data and optical data characteristic of a defect can be transmitted to data processing means.

One or more of the pressure sensors may be of the digital type, provided with means for sampling pressure data, and for identifying pressure data characteristic of said defect. In this case, the acquisition of measurement data, and their processing to identify the appearance of a defect, are performed by own means, associated and dedicated to the corresponding sensor or, better still, to each sensor. Such a digital pressure sensor, equipped with its processing means and storage means, can therefore permanently perform an analysis of the evolution of the pressure. This can be measured according to a fast observation cycle. Due to the existence of several sensors, a parallel processing architecture is used, taking advantage of the processing or calculation means available in each sensor. The synthesis of the data thus obtained can be transmitted to an acquisition unit in a single communication frame, for example of a duration of the order of a few milliseconds, which will mainly depend on the communication speed of the sensor. Thus, a duration of less than 1 msec per sensor could be obtained in the context of the present invention. The storage by each sensor of the pressure disturbance over a sliding time window makes it possible to carry out an analysis leading to a very reliable diagnosis of the presence of a defect, for example an internal arc. The processing means of each sensor are able to self-diagnose the presence of a defect, and spontaneously warn the acquisition unit, which has to deal with only the alarm message of a single sensor. For this purpose, the processing means may comprise one or more algorithms embedded in each sensor. This results in a shorter diagnostic time than that obtained with existing devices. The diagnostic time can be divided by a factor of about 50 to 15.

According to one embodiment, a trace of the evolution of the pressure, over a certain time window, is stored in a memory, such as a rotating memory, which precedes the measurement instant.

An analysis, using the data processing means associated locally with each sensor, enables it to detect almost instantaneously, or very quickly, the presence of an anomaly, for example an overpressure anomaly, such as that generated. by an internal arc.

The sensor is therefore capable of transmitting, in a single frame, an alarm message containing a summary of the recent evolution of the pressure values.

The sensor also has the capacity to transmit, for a detailed analysis by the acquisition unit, the measured unit values within a time window which is determined either by the acquisition unit. either by the sensor according to its own analysis. The response time of each sensor is preferably less than 1.5 msec. Pressure variation data characteristic of a defect can be identified using a pressure variation calculation AP for a certain duration AT, (OP / AT), and / or using a frequency of pressure oscillations and / or measurement of peak to peak pressure amplitude and / or identification of a damped pressure wave and / or a persistent pressure gradient . The AP / AT ratio and / or the frequency of pressure oscillations and / or the peak-to-peak amplitude and / or the shape of the pressure wave can be measured or identified at regular intervals. In the case of the detection of an arc, pressure variation data, characteristic of an arc, can be identified by means of a calculation of pressure variation AP for a certain duration AT, AP / AT, which can be done at regular intervals.

It may be interesting to measure in addition a temperature representative of the temperature of the metal casing. The sensors and the processing unit can operate in autonomous mode, which enables the sensor to inform, in an immediate manner, the processing unit of the detection of a phenomenon. But the sensor and the processing unit are preferably in master-slave relationship, each sensor can transmit data or information only if it has received processing means a request for data transmission or information. Preferably, each sensor can transmit data or information, after receiving a request for data transmission or information, for a predefined and limited time. A method according to the invention may comprise, between step a) and step b), a step a ') of transmitting, to the data processing means, a fault appearance information. After step a '), an abnormality signal may be produced, for example by the data processing means.

This anomaly signal can be produced before step b). Preferably, step b) is performed on request of the data processing means to the sensor having identified a defect.

The defect is for example the appearance of an internal arc, or an external shock on the envelope, or an insulating gas leak, or an expansion of the envelope. It is also possible to reconfigure a sensor, programmed for example to identify arcs, to identify another type of defect, for example a dilation. The invention also makes it possible to solve the problem related to the influence of electromagnetic disturbances on the measurements. Thus, it may happen that information, for example transmitted in the form of a frame, is affected by a disturbance (for example a byte becomes unreadable), and in this case the information, for example the frame in the example. chosen, can be reissued. This is especially the case for digital communication sensors, for which there are at least two pairs of wires. One of the two pairs is dedicated to the supply of the sensor (for example: 24VDC nominal supply voltage), the other being dedicated to the communication, for example in the form of frame exchange. A transmission control can be realized (often called CRC). In addition, thanks to the memory of the sensor, the data transmission can be delayed, especially after the electromagnetic disturbances have disappeared (so after the fault that caused the disturbances). This can be particularly applied in the case where all the measurement points are downloaded for a posterior human analysis, after the sensor has detected and reported a defect in near real time. The sensor supply leads can also restore disturbances to the sensor, which operates during the fault. This is why filters, preferably EMC, can be provided at the sensor input. Other characteristics can be implemented for this type of situation: for example, a sufficient reserve of energy can be provided in the event of a voltage dip and / or an extended voltage operating range. In a device and a method according to the invention, each sensor and the corresponding data processing means may be contained in a probe body to be inserted into the metal casing of an HT device. In a device and a method according to the invention, data processing means of a sensor can be made in the form of a printed circuit housed in the probe body. The invention also relates to a high-voltage device, comprising a metal casing, at least one central conductor and a plurality of compartments traversed by this conductor, and possibly digital sensors for measuring pressure to implement a method according to the invention. invention, according to any one of the embodiments described above.

BRIEF DESCRIPTION OF THE DRAWINGS FIG. 1 represents a first embodiment of a method according to the invention; FIG. 2 represents a second embodiment of a method according to the invention; FIG. embodiments of a method according to the invention, - Figures 4A and 4B each represent a sensor that can be used in the context of the invention, - Figures 5A and 5B are schematic representations, on the one hand of a sensor which can be used in the context of the invention and, on the other hand, of its equivalent circuit, FIG. 6 is a schematic representation of a temperature probe that can be used in a sensor according to the invention, FIG. is a schematic representation of an electronic card that can be used in a sensor according to the invention, - Figures 8A and 8B show an application of the invention, with a combined pressure-optical sensor, respect In a metal casing of a GIL and a GIS, FIG. 8C shows an application of the invention, with a non-combined pressure-optical sensor, to a metal casing of a GIS, FIGS. 9C represent temporal evolutions of pressure data measured by a sensor according to the invention; FIG. 10 represents an optical system of known type, in an application to a metal envelope of a GIL. FIGS. 11A-11C represent measurements made according to the invention. DETAILED DESCRIPTION OF PARTICULAR EMBODIMENTS Examples of implementations of the invention are illustrated in FIGS. 8A and 8B, respectively in GIS or GIL architectures. The invention can be implemented in GIS or GIL type architectures, but cut-off or isolation devices, such as one or more circuit-breakers or disconnectors, can be integrated in this type of architecture. The invention can also be implemented in an interrupting chamber, for example at SF6, an unshielded circuit breaker or in an unshielded disconnector. In particular, it is then possible to detect the presence of light and a pressure signal, for example a pressure variation AP, before the opening of a circuit breaker, which makes it possible to distinguish a defect arc from a breaking arc. FIG. 8A shows a GIS comprising a compartmentalized metal casing 2 and a central conductor 4, isolated by a gas under pressure, for example nitrogen and / or sulfur hexafluoride SF6. The compartments are delimited by insulating supports 6, which, in addition, maintain the electrical conductor 4. The invention also applies to three-phase high-voltage equipment comprising three conductors 4 in the same envelope. On each compartment is mounted a pressure sensor MP and an optical sensor DL, these two sensors being associated with a local unit denoted S1, S2, S3, etc.). The different pairs of sensors (MP, DL) are connected to an acquisition unit 12 by a digital communication bus 14. Such acquisition means 12 are for example microcomputer-type and comprise for example a central unit or a microprocessor, which will make it possible to carry out processing on the data received from the sensors, and to store said data as well as calculations made from of these data. A display unit makes it possible to display data, for example in the form of graphs such as those of FIGS. 11A-11C, or calculated data.

The means 12 may be programmed to implement the method steps described in the present application. According to the invention, it is not process by successive requests on each sensor and sending all the data to the unit 12, which would introduce a cycle time incompatible with a fast sampling on each sensor. On the contrary, each unit S1, S2, etc. has means for sampling, preferably at high speed, each signal.

Each unit S1, S2, ... furthermore comprises storage means, for example a rotating memory, which can temporarily store at least a certain number of data which have been taken during a time interval of duration AT. , elapsed, preceding the time tO of measurement (this is the interval [tO-AT, t0]). Such an interval is represented for example in FIG. 9A, which gives the evolution of the pressure as a function of time, as measured by a pressure sensor ML. On the curve of FIG. 9A is further shown a pressure jump AP, which corresponds to the variation between the initial pressure and the final pressure, as detected by the pressure sensor. This Figure 9A and Figures 9B, 9C are discussed further. Each local unit S1, S2,... Further comprises, locally, data processing means making it possible to process the data stored in the storage means. From this sampling and processing, an anomaly can be qualified.

For example, an internal arc can be detected from a high frequency of the movements of the pressure generated in the gas, or from a significant peak-to-peak pressure difference. Pressure measurements during an electric arc in SF6 made it possible to characterize frequencies and amplitudes as well as attenuations during propagation in the gas. The frequencies are greater than those of a mechanical noise, thus making it possible to distinguish with the case of a mechanical noise.

The measurement of a peak-to-peak pressure difference is particularly well suited in busbar or LIG-type compartments, where it is difficult to detect small, low-energy arcs, and where, therefore, variations in AP pressure is often low due to a particularly large volume of compartments. For example again, the pressure rise over a certain time interval, such as the AT interval above, is calculated. The data processing means associated with the sensor can therefore calculate an average value of the AP / AT ratio, AT being a time window immediately preceding the measurement instant. This value can be compared to a threshold value, below which there is no anomaly, and beyond which the anomaly is characterized, that is to say corresponds to an arc. This confirms the internal arc diagnosis after light detection. Indeed, the mere presence of light is not automatically significant of the presence of a defect arc. This is the case with corona phenomena. The information relating to the events detected by each sensor or each pair of sensors is available in the corresponding unit S1, S2. With a device and a method according to the invention, such events can be detected very rapidly, and allow the setting up a security measure of the installation, to prevent the situation, which results from the event in question, from taking on a very serious character.

Such events, or the corresponding information, can be stored or archived in a database, which can then be analyzed, for example in terms of aging and material wear. Figures 9A-9C provide examples of form for various situations and types of events. If the anomaly is characterized, then there may be backup of at least all or part of the sensor data identifying the anomaly. The presence of the anomaly is reported to the acquisition unit 12. The latter can then immediately decide on an action (for example triggering a signaling or an alarm), then download all of these data. saved, for example by querying the sensor that reported the anomaly. As soon as the reception of the data by the acquisition unit 12 is good, the sensor 10 can resume the normal fast interrogation cycle. The acquisition unit 12 therefore first has information relating to the occurrence of an incident (production of an arc or other anomaly), which saves time to trigger an alarm, rather than to wait for the reception of all the data, those which identify the anomaly and possibly other data. The transfer of the measuring points, which justify the triggering of the alarm, can then be realized.

In other words, according to the invention, each sensor and the means which are locally associated with it, has the ability to detect an anomaly in a very short time, to report this anomaly to the unit 12, which will first act for that the situation resulting from the appearance of this anomaly is controlled. Examples of methods according to the invention, and of operation of a sensor according to the invention, will be given in conjunction with FIGS.

First, in Fig. 1, process steps are referred to as SO - S5. In this example, the method is implemented to diagnose the presence of an internal fault arc during the operation of HV disconnectors.

In a first step SO, the detection system is powered up. The sensors (pressure and optical) are made active (step Si). Step S2 corresponds to the presence detection of light, which determines a time T0. As long as such detection is not performed, the method loops on step S1. Once light has been detected, a rapid sampling of the pressure as a function of time is carried out (step S3). The sampled values are stored. A trend calculation makes it possible to determine an increase, or a signature, of pressure. As long as a pressure signature characteristic of a defect arc is not detected (step S4), the method loops to step S1.

As soon as a pressure signature, characteristic of a defect arc, is detected (step S4), the start of the pressure rise is dated (instant Tl).

An alarm can then be generated (step S5) indicating the presence of a fault arc. The mere presence of light is not necessarily significant of a defect in the device, for example in a disconnector compartment or in an earthing switch. It is the presence of light (step S2) which makes it possible to trigger the rapid pressure sampling (step S3). In the case of a disconnector, the sampling time is close to the operating time of the disconnector, ie a few seconds. A variant of this method will be given in connection with FIG. 2, in which the process steps are designated by the references S10-S17. Steps S10 and S11 are identical to steps S0 and S1 already described above. Then (step S12) is carried out a rapid sampling of the pressure measurement and the optical detection, which makes it possible to identify the presence or absence of emitted light.

The data is stored in a rotating memory. If light is detected or if a pressure signature characteristic of an arc of a defect (step S13) is identified, sampling is continued (step S14) during the maneuvering period and the data is set. memory.

The signature can be developed as described above. This signature is compared with a reference signature (step S15), in order to identify a possible anomaly. An alarm can then be generated (step S16) indicating the presence of a fault arc. During steps S13 and S15, if no abnormality is detected, the method loops back to step S10. FIG. 3 represents the detection of the presence of an internal fault arc and the location of the fault within compartments, particularly advantageous for long busbar or LIG busbar compartments. In this figure, the process steps are designated S20 - S26. The steps S20, S21, S22, S23 and S24 are identical to the steps S0, S1, S2, S3 and S4 already described above. At the end of step S22, an alarm is generated, indicating the presence of light (step S'22). During step S25, a calculation of the location of the fault is performed, based on the difference T1-T0. An alarm can then be generated, (step S26), confirming the presence of a fault arc. The distance of the fault from the sensor is indicated. In this embodiment the presence of light can be considered as preliminary detection.

FIG. 4A represents an integrated probe that can be used in the context of the invention. This probe comprises on the one hand pressure sensor means, and on the other hand means for collecting optical information. In the illustrated embodiment, this probe is of substantially cylindrical shape about an axis AA '. It comprises a body, here constituted in two parts 150, 152, fixed on a base 153. The latter is traversed by a cylindrical passage 113, connecting the atmosphere of the compartment to the sensor. At the end of this passage is disposed the sensor 110. The latter can be insulated, held in a ceramic part 140, itself held by the base 153, and in the base 141 of which an orifice 115 leaves the free passage to the ambient atmosphere from the conduit 113 and whose pressure is to be measured. The diameter of this orifice 115, as well as the diameter of the portion of the duct 113 located between the base 141 and the sensor 110 can be adjusted to form a low-pass mechanical filter, which makes it possible to filter the noises. The outer tube 152, which forms the lower part of the body of the device, is preferably made of stainless steel and is fixed on the base 153. During operation, the temperature of this tube 152 is that of the casing 2. cover 150 which closes the device may be a molded part, for example treated polyurethane, and which holds the temperature over a range substantially between -40 ° C and + 110 ° C. In the extension of the sensor 110, and substantially on the axis AA 'of the device is arranged a printed circuit 130, comprising processing means and data storage means measured from the sensor 110, as well as communication means with an external acquisition unit 12. The sensor 110 is connected from its rear face to the circuit 130 by connections 111. The circuit 130 is held in the tube 152 by a filling material 156, for example based on silicone. A second level of filling, or external filling, 154 is for example polyurethane foam. The circuit 130 is itself connected to a connector 170, on which connecting means 175 can be fixed. These means 175 are extended by a cable 180.

The assembly comprising the sensor 110, the circuit 130 and the connector 170 can be held by a tube 155, itself held in the tube 152 by the filling material 154. Preferably, a starting distance d, calculated for a certain voltage value, for example 30 kV, separates the ring 171 from the connector 170 of the outer tube 152. This device further comprises a lateral optical input provided with an optical connector 190, into which a sheath 196 containing two optical fibers 192, 194. Each of these fibers comprises, at its end located in the body of the device, respectively: a photodiode Del 202 test, which will allow to transmit, via the fiber 192, the test light to the set of the optical detection chain, and a photodetector, such as a photoresistor 204 which will convert an optical signal received by the fiber 194 into an electrical signal which can be transmitted to the electrical circuit In fact, the photoresistor 204 may itself be integrated into the circuit 130. The LED is preferably also integrated in the circuit 130.

Other means than a photo resistance can be used: for example a photodiode or a photo transistor. If a photoresistor is used, it may be from the Perkinelmer VT series, for example reference VT33CT. The output of this component is proportional to the amount of light received. The test diode of the fiber can emit in visible light or the near infrared, for example at 880 nm for the Perkin-Elmer VTE1261 diode.

A relatively narrow target angle in the fiber (for example +/- 10 degrees) allows the light to be channeled into the fiber, thus avoiding direct illumination of the light sensor. The daylight, which may possibly enter the device, for example by one or more portholes, may optionally be filtered (for example the Perkin-Elmer CR50DE-DLF photodiode makes it possible to perform this filtering). Note that it is also possible to produce a device with a single receiving fiber 194, without test fiber or test photodiode. By its other end, the sheath 196 can be introduced by means of an optical connector 200 into an opening in the wall 206, 206 'of a device under pressure (this wall is for example the envelope of a GIS or GIL under pressure). At the end of the fibers located on this side is disposed an optical sensor 210 (or light collector). The reference 220 schematically represents an internal arc which produces light designated by the arrows 221, a light of which a portion is picked up by the sensor 210. The test device constituted by the diode 202 and the test fiber 192 allows the automatic test. of the photoresistor 204, but also the fiber 194 and the light collector 210. In this test, the emitting diode 202 sends a flash of light to the light collector 210 via the fiber 192. A fraction of the light emitted is then reflected at the end of the collector 210 to the return fiber 194, thereby testing the entire chain. In current optical devices, a test can be performed manually and fiber by fiber by disconnecting the end of the fiber. This test is long. It does not include the light collector because it can not be dismantled. During this test operation, which lasts several hours, the arc detection system is inoperative, leading to a shutdown of the monitored system, for example GIS or GIL. There is also a risk when the end of the fiber is put back into service, so reassembled on the envelope of the device. On the contrary, with a device comprising test means, the sensor according to the invention does not require an operating stop. It can take place in 10ms and automatically. The device described above in connection with FIG. 4A integrates means making it possible to collect both pressure data and optical data.

Another type of device may be used in the context of the present invention: it is a pressure sensor which is not coupled with optical means, as illustrated in FIG. 4B, on which numerical references identical to those of FIG. 4A designate identical or similar elements therein. The structure of this sensor is therefore very similar to that of FIG. 4A, but does not include a lateral opening which makes it possible to insert an optical connector and means such as optical fibers. One or more optical fibers will therefore be placed in parallel with this sensor (as indicated very schematically in FIG. 4B), the optical signals that it takes directly to an acquisition unit 12 (as in FIG. below).

In the configuration of two separate sensors, as in FIG. 4B, it is possible to use an intermediate means 181, of the housing type, uniting the data of the pressure sensor and the data of the optical sensor, and then making it possible to transmit the all of the data via a single link to the unit 12. It is in this intermediate box that the electronic processing and communication means can be hosted.

Figure 5A schematically shows the sensor 110 itself. This pressure sensor 110 operates on the principle of a capacity variation caused by the bending of a ceramic membrane 112 (Al2O3) metallized on its internal face. This membrane is sealed on a base 114 itself ceramic and on which three electrodes 115, 117 and 119 have also been previously screen printed. This seal is made by serigraphy of a thin glass thickness 116 in the pasty state around the periphery of the membrane 112 and on the base 114, then, after superposition. Under the influence of pressure, the center of the diaphragm 112 approaches the base 114 and causes the increase of the measuring capacitance Cm constituted by the electrode 113 located in the center of the base and the membrane. The peripheral electrode serves as reference capacitor Cr. In one embodiment, the equivalent circuit 30 of this device is shown in FIG. 5B, in which the measurement and reference capacities are respectively represented. The measurement capacities Cm and reference Cr are of the order of 30pF each. The variation of the measurement capacitance (central capacitance) is of the order of 1 pF to 2 pF for the full scale of measurement. The conditioning of the signal is carried out on the principle of charge transfer from a capacitor to both capacitors of the sensor. The voltage obtained across the two capacitors is then filtered to give an analog signal. In order to determine the temperature of the casing 2 of the disconnector, temperature sensor means may be arranged in the measuring device, preferably with the electronic circuit 130. For example, these means 109 comprise two temperature sensors RTH1 and RTH2. , arranged as shown in Figure 6, are used. Each of these two sensors can be made of platinum, in the form of a resistive chip. The nominal resistance of each sensor, at 0 ° C, is 1000 Ohm. The two sensors are arranged in series, with a common terminal. Tcap pres designates the temperature of the pressurized capsule, Telectronique designates the temperature of the electronic circuit 130. A calculation method for determining the temperature of the envelope is for example the following: Tnew = Tcap pres - ((Telectronique - Tcap pres) / RTH2) * (RTHI + RTH2).

This type of calculation gives a very correct result insofar as the power (energy) dissipated by the electronic means is constant and where the isolation of the heart of the transmitter is correctly ensured (from the thermal point of view) with respect to the outside. . In this case, the flow of calories dissipated by the electronic card will be essentially to the envelope 2 of the device. This temperature-related information can then be exploited to determine the density of the isolation gas or the evolution of this density as a function of time. The information concerning the density, and / or the pressure and temperature information, can be sent to the acquisition unit, either spontaneously, in the case of an autonomous mode of operation, or at the request of the unit. in the case of a master-slave operating mode. According to a preferred embodiment, this sensor is precisely supplied and regulated under a DC voltage of 24 V. Its operation is permanent, so the heat flow is constant. With such a sensor it is possible to precisely measure the temperature of the end (screwed tip 153) of the sensor 10, which is fairly representative of the temperature of the casing 2. The measurement accuracy is 1 ° C. or even lower. The validity of the measuring principle is in particular ensured by the good thermal conductivity of the ceramic electrical insulation 140, which is about twice as good as that of the stainless steel used for the construction of the end 153 of the sensor. FIG. 7 represents an electronic card of a pressure sensor according to the invention. Downstream of this electronic card are the means, an embodiment of which has been described in more detail above, which can collect pressure data, and radiation data: the pressure sensor 110, the photo resistor 204, as well as diode 202, and the means, respectively 122, 222, respectively conditioning the pressure signal and the optical signal. In use, these measuring means are associated with a metal casing 2, for example a GIS or a GIL. In this figure, are also schematically the means 109 which can collect temperature data. The transmitter electronics is structured around a microcontroller 132 which incorporates features such as signal acquisition and communication. In fact, this electronics comprises means or a card, for example in the form of printed circuits, central processor 132 with its power supply 134 and an interface 136, for example of the RS485 type. The means 132 comprise means for conditioning the signal of the capacitive pressure sensor.

The means 130 may comprise means, or a filtering card, in particular to provide EMI and EMC protection. In the structure of FIG. 4 these means can be placed on the circuit 130, immediately behind the base 171. Preferably, the microcontroller means 132 make it possible to carry out an acquisition of analog signals (ADC) with a multiplexer with 4 inputs and 16 bits of minimum resolution. The acquisition time is preferably less than 10 ps. These means 132 make it possible to calculate the effective value of the pressure and of the temperature in a computation time lower than 0.5 ms (including thermal compensation, linearity, scaling, etc.).

The RS485 serial link is of UART type with FIFO type buffers. These means are equipped with a "programmable" FLASH memory in situ (minimum 16Kbyte) They also include a RAM sufficient to contain all the working variables as well as the communication buffers and the measurement recordings. capacity 8Kbyte, or more, These means also include an EEPROM memory, for storing various information such as, for example, calibration parameters and / or the references and / or the serial number of the sensor.These microcontroller means 132 comprise in addition, an integrated clock system or means for generating a real-time clock signal, these clocking means will also make it possible to manage the synchronization of the selected protocol, such as the Modbus protocol. for this purpose 3 counters with 16 bits.

The microcontroller 132 preferably has an internal RC oscillator. For example its frequency is 8 MHz and its accuracy is of the order of 2%. It also incorporates a PLL that can multiply the internal clock frequency by 4, 8 or 16. The instruction cycle is composed of 4 clock cycles. The execution power can be programmed to get 8 MIPS (system clock = internal oscillator x 4), it can be increased to 16MIPS.

Normally, the accuracy of the internal oscillator is sufficient to ensure proper data transmission on the serial link, but an 8 MHz ceramic resonator (frequency accuracy> 0.5%) can be implemented on the electronic board, to remedy a problem. possible problem. Finally, the microcontroller will enable the management of interrupts with a table of programmable vectors. For the specific use that is made of it, this microcontroller preferably has a low consumption, for example about 22 mA at 5 volts in 8 Mips. It operates over a wide operating voltage range, for example from 2.5V to 5.5V, and over a temperature range of -40 ° C to 85 ° C (the latter corresponds to a functional guarantee for the measurement ). But, it is desirable that the resistance of the component to the temperature is in fact about 140 ° C, that is to say that for any temperature below 140 ° C there is neither destruction nor irreversible damage. An example of a microcontroller that can be used is dsPIC30F3012 or dsPIC30F3013 from MICROCHIP. The device 130 includes means 134 for feeding. Preferably, and in order to maintain a constant temperature gradient on the electronic board with respect to its environment, and whatever the supply voltage, this power supply implements the switching energy conversion technology 15 . On the microcontroller side, the consumption is almost constant, as long as it operates at a constant frequency. Regardless of the supply voltage, the power consumed by the electronic board is approximately constant, only the current varies. Given the operating conditions of the entire system, the power supply is the element that is most directly affected by any disturbances generated by the environment. Therefore, it is preferred to choose a robust power supply, which can provide protection for the microcontroller; it is for example of the type qualified for embedded applications in the field of the automobile. Preferably, the main characteristics of a switching power regulator 134 for this application are the following: - a low consumption (operating current <1 mA), - a high switching frequency (for example of the order 600 kHz) in order to minimize the value of the energy transfer choke, - a high voltage integrated power MOS technology, to allow a wide input range, - an integrated voltage reference, - a protection against current peaks and short circuits.

A suitable switching regulator is the National Semiconductor LM5008 controller. Its built-in 100V MOS supports a power supply voltage range of 10 to 90Vdc (for a nominal supply voltage of 24Vdc). The switching frequency varies, depending on the current delivered by the controller, between 100 and 600 KHz. The value of the self is between 100 pH and 220 pH. If the microcontroller previously selected (dsP1C30F3012) is programmed to provide a calculation power of 4 MIPS, its theoretical power consumption at 5 Vdc is 22 mA, which gives an overall consumption of (5V / 24V x 22 mA) + 1 mA <6mA for a 24 Vdc board, considering that the microcontroller is the main current consumer. The stall level of this power supply is 8 volts, and a consumption of only 6 mA is satisfactory, far from the imposed limits. The RS485 communication interface 136 provides data transmission at 115 kbps in duplex 5-duplex mode over a twisted pair. Since the RS485 interface driver is the most exposed component to the disturbances generated by the environment (in particular the driven voltage peaks), the latter is selected according to its performances with respect to overvoltages and short circuits. An ESD protection, for example of +/- 15 KW, is therefore integrated into the component. In order not to generate itself conducted or radiated disturbances, the RS485 driver is selected in the speed range 15 between 500 Kbauds and 2.5 Mbauds. The choice of a fast component makes it possible to limit current calls during switching operations. Since the current is drawn from the sensor supply, this may result in noise on the measurements. The current calls also depend on the load represented by the communication bus between the sensor and the acquisition unit. Preferably, the consumption of the driver in operation does not exceed 1 to 2 mA. Several drivers have been tested. The first is LinearTechnology's LTCI785CS8. This driver is the most efficient in terms of protection against external disturbances, it provides protection against surges up to 60V, against short circuits, and it supports ESD up to + 15 KV without damage or latchup (IEC-61000-4-2, level 4, which identifies electrostatic discharges) as well as transient, fast, burst discharges of + 4 KV (IEC-61000-4-4 level 4, which identifies type of electrostatic discharge). The disadvantage of this component is its power consumption (42 mA in 5 volts, 9 mA in 24 volts). This brings the sensor consumption to 15 mA. This is a disadvantage for the sensor because excessive power consumption can lead to internal heating complicating the temperature measurement. In addition, the cost of this component is high. Another component, Texas SN65176B, gave during the same tests results superior to those obtained with the LINEAR component. The TTLx technology used in this Texas component, compared to the LINEAR CMOS technique, is more immune to noise because of its much faster switching times.

Another component, the MAX3O85E MAXIM, is protected against disturbances: it supports ESD + 15 KV, and provides protection against overvoltages (direct or reverse) of + 13 V. The fronts of the transmitted signal are limited, so to reduce the disturbances emitted. Its consumption in operational mode is of the order of 0.4 mA. An operation of a data processing and exploitation software implemented in means 132 of an individual sensor 10 is as follows.

Such software makes it possible to implement all or part of the functions or operations already described above. On power-up, the microcontroller 132 itself manages, according to the programmed configuration (specific memory area of the means 132)), its Reset phase and the activation of its oscillator. All inputs / outputs (I / O) are preferably configured as input: in a first step, the software configures the inputs / outputs of the microcontroller 132 as well as the various associated functions (for example: the analog inputs for the sensor signals , UART (universal asynchronous transceiver) for the serial link that drives the RS485 driver, ....). The interruptions (IT) in real time make it possible to ensure a good time flow, perfectly clocked.

In a first phase, the real-time ITs are generated at a frequency that can be adjusted, for example every 200 ps. We seek a compromise between processing capacity in IT in real time and background tasks. In general, IT processing should not exceed 30% of the overall capacity. The tasks performed in the processing of the IT in real time are essentially the acquisition of signals from the sensors and the incrementation of various counters such as those relating to the management of the temporal progress of the communication protocol between the sensors 10 and the unit 12. One or more acquisitions of each of the sensor signals are performed at each interruption in real time. Each signal value thus obtained is added to the previous ones in a buffer register. All 8 ITs in real time, the buffer is transferred to initiate the pressure and temperature calculations, then reset for the next acquisition sequence. Calculations are supported by a background task manager (excluding IT). Thus, it is possible to obtain an updated pressure and temperature value every 1 ms. The treatment on each sample packet, for example 16 samples, leads to an average after elimination of the atypical values and the extreme values, by the use of an absolute and relative coherence filter. The transmission and reception of information is done under IT. The microcontroller 132 is provided with buffers (in transmission and reception) which ensure good transmission, without risk of loss of characters, and without the risk of breaks in the rate of transmission of the characters of the frame (maximum time between 2 characters = 1.5 characters). The main background tasks enabled in the measurement mode are as follows. measurement management: the values of the pressure and the temperature are calculated as a function of the acquisition values and the calibration parameters (thermal compensation, linearity, scaling, etc.). - Frames management received frames are analyzed. There is no action if the frame is intended for another sensor or processing the command in the opposite case, with in this case loading the response buffer to the request of the master. When the command is processed and the buffer loaded, a flag is set to initiate the transmission of characters under IT UART. - Monitoring of the functioning of the transmitter: this function is ensured thanks to mechanisms of validity of the accesses in memories for the program and the data.

Secondary tasks include, for example, ancillary procedures that the main program occasionally uses to perform particular and exceptional functions, such as: a function that makes it possible to program or load or download a new version of the software without physically opening the product . This function is used to receive the new version of the program via the RS485 serial link. For this purpose a specific function of the protocol is reserved, and ensures the implementation in memory "FLASH". Possibly, this function makes it possible to implement another version of the data processing if one changes the type of fault to be characterized.

Calibration, which is a specific function that the main program uses only during the calibrator phase of the transmitter, in production or during subsequent calibrations. The transmitter is fully assembled prior to calibration, which is performed via the RS485 serial link and a specific control of the reserved protocol. This function includes the recording of the traceability and identification data, which data are stored in a memory (for example of the EEPROM type) of the transmitter.

The offset and gain parameters can be adjusted beforehand, the calibration on site then being limited to a comparison with a standard. The data of the sensors 10 are transmitted to the calculation unit 12. To do this, a dialogue protocol between these different elements has been developed. These elements are in master-slave relationship: the sensors (slaves) can only transmit if they have been invited by the unit 12 (the master). Each exchange of data is initiated by it. When it interrogates a sensor, it has a maximum duration (for example 1 ms or even 500 ps), to process the command and respond; a message sent out of this time window is considered invalid. Likewise after the execution of a command, or after receiving a frame of which it is not addressed, the transmitter 10 has a maximum duration of time, for example 500 ps, in order to get back into tune. The physical link is provided by an interface RS485 which transmits a differential signal in half duplex (the different devices successively access bus 14 (FIG. 1) This link uses a twisted pair for the transmission of the data, and one pair for the power supply (24 Vdc nominal).

The format of the data transmitted on the bus is as follows: - 1 Start bit, - 8 data bits (least significant bit first), - 1 Stop bit. If the transmission speed is 115200 baud, the transmission time of a character is therefore of the order of 100 ps. When transmitting a frame, the characters are transmitted successively, without interruption. The transmission protocol can specify that the maximum duration between 2 characters is less than 1.5 characters, so less than 150 ps. An interruption of the transmission of characters, for a duration greater than 3 characters, (thus greater than 300 ps) is interpreted as a frame end. The format of a frame transmitted by the master is as follows: Station address Code Function Data (optional) _1 CRC16 H 1 CRC16 L The station address is that of the relevant sensor. An address, for example 250, can be reserved for "BROADCAST" commands, that is to say commands addressed to all the sensors 10 simultaneously. These commands do not require any response from the sensors. The other addresses, for example 1 to 249, can be used for the transmitters present on the bus 14. The function code determines the command to be executed by the station (slave). The function code is for example coded on 7 bits, the eighth bit (MSB) being always equal to 0.

The parameters or data are the information allowing the execution of the function. The 2 bytes CRC16 constitute a key of control of the whole frame and make it possible to guarantee the integrity of the received data. The calculation of CRC16 is subject to a standard described in the Modbus standard (modbus.org). A response sent by a transmitter responds to a specific format: Station address X or Code Function n Data (optional) _1 CRC16 H 1 CRC16 L At the top of its response, the transmitter returns its own address to confirm that it is he who performed the function. If the transmitter has correctly executed the function, it returns the same function code as that sent by the master. If any error has occurred, it returns the function code by forcing the eighth bit (MSB) to 1. In this case, the following data is an exceptional error code.

Data data is present if the function requires feedback.

For the bytes CRC16, one will refer to the explanations above. In general, the length of a reply sent by a station comprises between 5 and 10 characters.

Finally, exceptional errors can be programmed, such as transmission errors. Each transmitter can implement the following functions. - Reading of the calibration parameters and technical specifications. - Writing of the calibration parameters and technical specifications. - Initialization of the devices; device identification given back. - Programming the bus address. -Reading the serial number. - Reading of real pressure and temperature values in floating point format. - Reading actual pressure and temperature values in whole format. For each of these functions, is defined. a request from the master, according to a predefined format, for example that indicated above, a response from the transmitter, according to a predefined format, for example that indicated above, one or more exceptional errors A function for reading the values measured in "cascade", also called "fast BUS", can be implemented. It consists in successively obtaining the response of several transmitters following a single request from the master, it is specific and contrary to the principle of the Modbus protocol, but makes it possible to improve the efficiency of a method according to the invention. If several devices are connected to the bus, only the first is interrogated by the master; the others answer each in turn, provided that the addresses are in ascending and consecutive order.

Otherwise, the acquisition unit polls the sensors one by one. Each pressure value measured, for example every millisecond, is stored in a 512 value revolving stack.

This value is expressed in millibar. 2 bytes are used per value, a stack of 1024 bytes. This stack can be stored in a RAM of the DSP, using half of the RAM. The remaining 1024 bytes are sufficient for the purposes of the application. The application has a pressure event detection algorithm, based on a threshold of AP / AT (configurable). The calculation is done at regular intervals, for example each msec, by calculating the pressure rise over a certain time interval, for example an interval of 10 msec. If the threshold is exceeded, a confirmation based on the slope obtained by linear regression can be performed on, for example, 10 values.

In addition, if exceeded, the stack is frozen with a certain value taken as the original value of the record, for example the value N-20. The following values (for example: 512 values) are saved and kept available to the master for, for example, one minute. The trigger triggers a millisecond counter for time marking of the event, when read by the master.

In the case of an event, during the response (whatever the question, pressure or temperature), a specific code is returned to the master. The latter having been informed of an event, a function code makes it possible to load (Dump) the stack of these values as well as the time counter. The master returns an acknowledgment of receipt, or reiterates his request. On receipt of this acknowledgment, or at the latest after a predefined period, for example 64 seconds, the operation of the rotating battery is re-initialized, and the operation of the counter is stopped. Preferably, the response time of a sensor is less than 1.5 msec. This makes it possible to ensure a duration of a Question / Answer (Q / R) cycle, during an exchange of data or information with the unit 12, less than 2.5 msec. More preferably, the time between the end of a master's question and the beginning of the sensor's response is less than 500 microseconds.

The initialization time of the sensor is preferably less than 10 msec, and preferably less than 3 complete cycles of Q / R under pressure, ie at most 7 msec. In an exemplary embodiment, the dimensions of the electronic card are 1 = 16 (+ 0.1) mm, L = 30 + 10 mm. The use of the latest generation of SMD components ensures miniaturization and resistance to electromagnetic interference. The SMD resistors and capacitors are not subject to power requirements or voltage withstand, are in 0402 housing (which is a miniature housing, preferable except in case of heating of the component). Non-limiting examples of embodiment details are as follows.

The operational amplifiers for the conditioning of the signals coming from the pressure or temperature sensors are in an MSOP8 package. The printed circuit is of 4-layer class-5B or class-6B technology. The 2 central layers will be reserved mainly for the grounding of the microcontroller and the positive supply voltage of the microcontroller (for example +5 Volt). The material is FR4 with a thickness of about 1.6mm.

The thickness of the conductors (copper) is standard: +/- 35 microns, the protective coatings of drivers are in LPI - 2 x green. Given the extreme level of miniaturization, there is no silkscreen for the identification of components, it has more disadvantages for the good welding of components that real interest in the development or testing of cards . The minimum width of the conductors (tracks) is 0.2mm, and the hole diameter is 0.35mm 5 minimum.

FIGS. 8A and 8B show an application of the invention, with a combined pressure-optical sensor, respectively to a metal casing 2 of a GIL and a GIS. In its figures, as well as in Figure 8C, described below, there are references which have already been described above, in the introduction to the present application, in conjunction with Figure 10. 15 The pressure sensors are designated in FIGS. 8A-8C by the letters MP and the light sensors by the letters DL. The references S1, S2, S3,... Each denote either an integrated type sensor, for example 20 such as that of FIG. 4A (this is the case in FIGS. 8A and 8B), or a pressure sensor. , associated with a radiation sensor but not integrated with it, for example such as that of Figure 4B (this is the case in Figure 8C). It can be seen from both FIGS. 8A and 8B that the sensors are connected in a simple manner to the acquisition unit 12 by a digital supply and communication bus 14. The latter may consist of a first pair of wires dedicated to power, and a second pair dedicated to digital communication. In both cases, only one bus cable is implemented and there is no signal loss related to the distance. In addition, the structure of the sensor of FIG. 4A makes it possible, as already explained above, to carry out an in situ test, with means that are themselves integrated in the sensor. A data processing can furthermore be carried out in the sensor itself, using means 130. Such a device makes it possible to reinforce the fault diagnosis of an arc. In both cases, each sensor S1, S2, S3 .... is a combined sensor, for example of the type described above in connection with FIG. 4A. In addition, each combined sensor is connected to processing and communication means of the type described above in connection with FIG. 7. These means will make it possible to communicate and implement a communication protocol (described below) with the acquisition unit 12. FIG. 8C represents an application of the invention, with a non-combined pressure-optical sensor, for example of the type described above in connection with FIG. 4B, with a metallic envelope 2 of a GIS. The connection with the acquisition unit 12 is this time more complex, it implements, on the one hand, a bus 14 for supply and digital communication with the pressure sensors, and, on the other hand, a 100 set of optical fibers that will allow to transmit the optical measurement data to the same acquisition unit 12. The bus 14 may consist of a first pair of son dedicated to the power supply, and a second dedicated pair to digital communication. To each of the compartments C1, C2, C3, .... of the metal casing is associated a pair consisting of a pressure sensor MP and an optical fiber or optical means DL for detecting the presence of a radiation emitted in the corresponding compartment. This time, it is the pressure sensor means that will be associated with processing and communication means of the type described above in connection with FIG. 7. These means will make it possible to communicate and implement a communication protocol. (described later) with the acquisition unit 12. The configuration described in connection with Figure 8C is less advantageous than that of Figures 8A and 8B, but it is nevertheless within the scope of the invention.

FIGS. 9A-9C show the evolutions of pressure signals as a function of time in the context of measurements made with a device according to the invention. In all 3 cases, an event occurs in the envelope 2 of a high-voltage device for a duration Tarc. Pressure variations are observed during this time interval, and the question arises whether these pressure variations effectively characterize the presence or absence of an electric arc. The pressure data are obtained, in all three cases, according to one of the algorithms which have been described above in connection with FIGS. 1 to 3. The first case (FIG. 9A) and that of a small volume compartment , where of a major defect, which releases a lot of energy. In this case, the amplitude AP, that is to say the difference between the pressure after defect and the pressure before defect is significant, that is to say greater than the measurement error of the sensor. pressure.

The second case (Figure 9B) and that of a large volume compartment, or a small defect, which releases little energy. In this case, it is preferable to identify the frequency of the pressure variations. If this frequency is high, and if the oscillations occur over a sufficient period of time, a fault can be characterized. The third case (Figure 9C) and that of a large volume compartment, or a small defect, with a sensor located relatively close to the defect. In this case, it is preferable to identify the peak-to-peak amplitude AcP; a high peak to peak peak amplitude confirms the presence of an arc. FIGS. 11A-11C represent measurements, of the type shown schematically in FIG. 9A, for a 990 A arc (duration of 210 ms), performed with a self-triggered recording sensor from Bourdon. Pressure values can be identified for every millisecond. On these 3 graphs, the abscissa axis is in ms and the ordinate axis in mbar.

The sensor is located at right of the arc for Figure 11A and at 3.45 m from the arc for Figure 11B. For FIG. 11C, two curves are represented, one (curve I) with the sensor located at the right of the arc and the other (curve II) with the sensor located at 3.45 m from the arc. The measured AP is 24.5 mbar. The insulating gas is SF6 in all 3 cases. The light signal can be taken into account in various ways on these examples of curves, and in general on any representation of an event detected by the invention. Thus, in the case of FIG. 1, the light signal can be represented in all or nothing form as a slot of duration equal to the arcing time Tarc, with a high state in the presence of light, and positioned in advance on the pressure wave with a start at a time T0. There may be overlap between this niche and the pressure disruption. The slot can also be subdivided into several parts, the light energy is not constant, and depends, for example pulses at 100 Hz (positive pulse + negative pulse, at 50Hz) in case of short circuit (less light in passing by zero of the voltage). On the other hand, the light energy of the arc can be obscured by clouds formed by decomposition products of the gas, for example SF6, during the arc. It is therefore wise to use the front of the first pulse (whose light intensity is not impaired) as a trigger.

However it is also possible to measure the amount of light received. In this case the slot is transformed into a curve (the slot is from the curve on which a threshold of detection is applied). With the light detection means, in particular with photodetector, it is possible to measure the amount of light received, which, measured over the same cycle time as the pressure measurements, allows a superposition of the curves. This does not prevent a triggered operation, thanks to a detection threshold. An example of advantage related to the invention is as follows. When operating the disconnectors or earthing switches of a GIS station, while the charging current has previously been cut by the circuit breaker of the bay, arcs are nevertheless generated between the contacts. It is not about switching arcs having electrical In this defect, but of trapped charge arcs, the one capacitance, and the lines of the inductances and capacitances. Detection based on light generated by the shielded hardware is like a case where only systems will detect a fault, which is why these disconnected during maneuvers. This is a problem because, statistically, fault arcs occur for a good part during maneuvers, precisely on breaking a metal part for example. In this case, the invention makes it possible to maintain the functional arc detection, the detected arc diagnosis being confirmed only if a rise in abnormal pressure is detected.

Another advantage is a more precise location of the defect. The pressure sensors are installed automatically in the equipment to perform the gas monitoring and are at least one per compartment to monitor. The optical systems of the prior art are equipment added to standard equipment. They can only be installed in limited numbers on available connections and only on certain envelopes. An optical sensor used according to known techniques can therefore cover an area comprising several compartments. As a result, the location is much less precise than with a device and a method according to the invention (not to mention the cost aspect already discussed).

Claims (13)

1. A method of detecting an internal arc in a high voltage device, comprising a metal envelope (2), at least one central conductor (4) and a plurality of compartments traversed by this conductor, said method comprising: a) arranging, in each compartment, a pressure sensor (MP, 110) and an optical sensor (204), b) detecting an arc by recognizing, with the aid of the optical sensor, a light signal emitted by the arc, c) detecting the arc, by the pressure sensor, by recognizing the pressure waves emitted by the arc in the gas, d) locating the arc using the information obtained on the one hand by recognition of the light signal, d on the other hand by recognition of pressure waves. 2. Method according to claim 1, wherein the measurement of the time difference, between a light detection time and a pressure wave detection time, is converted into a distance. 3. Method according to one of claims 1 or 2, wherein it is carried out the pressure wave detection if a light signal is first detected. 4. Method according to claim 5, comprising an alarm generation step (S'22) or the sending of an anomaly signal, after step b) and before step c), or after step c) and before step d). 5. Method according to one of claims 1 to 4, wherein, if one detects light or a pressure wave, the light and pressure data are stored, a signature is drawn from these light data and pressure, and then compare the resulting signature with a reference signature. 6. Method according to one of claims 1 to 5, wherein the information obtained on the one hand by recognition of the light signal, on the other hand by recognition of pressure waves, are transmitted to a processing unit, the sensors and the processing unit: operating in autonomous mode, or being in master-slave relationship, each sensor being able to transmit data or information only if it has received processing means a request for data transmission or data transmission; 'information. 7. Method according to one of claims 1 to 6, wherein: - pressure variation characteristic of an arc data are identified using a pressure variation calculation AP for a certain duration AT, ( AP / AT), and / or using a pressure oscillation frequency and / or peak-to-peak amplitude measurement, -AP / AT and / or frequency of oscillation. oscillations of the pressure and / or peak-peak amplitude and / or the shape of the pressure wave being measured or identified at regular intervals. 8. Device for detecting an internal arc in a high-voltage device, comprising: - an optical sensor (204), an arc detection device by recognition of a light signal emitted by the arc, - a pressure sensor ( MP, 110), for the arc detection by recognizing the pressure waves emitted by the arc in the gas, - data processing means (12) for locating an arc using the information obtained from a on the other hand by recognition of the light signal, on the other hand by recognition of pressure waves. 9. Device according to claim 8, wherein said means (12) for calculating can convert a distance time difference between a light detection time and a detection time of a pressure wave. 10. Device according to claim 8 or 9, integrating at least one pressure sensor and at least one light signal sensor. 11. Device according to one of claims 8 to 10, at least one pressure sensor being a digital sensor. 12. Device according to one of claims 8 to 11, comprising a plurality of pairs of sensors, each pair comprising a pressure sensor and an optical sensor: - each pair being positioned inside an envelope (150, 152 ), in which a pressurized insulating gas enters via a conduit (113), or only the pressure sensor being positioned inside an envelope (150, 152), in which a pressurized insulating gas enters through a conduit (113). High-voltage device, comprising a metal casing (2), at least one central conductor (4) and a plurality of compartments through which this conductor passes, and a device according to claim 12, each pair of sensors being positioned in connection with a compartments and connected to the data processing means.