JP6553303B2 - Partial discharge monitoring system - Google Patents

Description

  Embodiments described herein relate generally to a partial discharge monitoring system inside a gas-insulated device.

  In power facilities such as substations, gas-insulated switchgears such as GIS, gas-insulated buses such as gas-insulated buses, gas-insulated transformers, etc. are used. The gas-insulated device is a device in which a high voltage conductor is housed in a closed metal container filled with an insulating gas and supported by an insulator. In this gas-insulated device, it is known that when a defective portion such as contact failure or metal foreign matter contamination is generated inside the metal container, a partial discharge is generated from the defective portion.

  If partial discharges in gas-insulated equipment are left unchecked, they may eventually lead to dielectric breakdown and develop into a serious accident. Therefore, it is important to take some measures such as early detection of partial discharges and repair of defective parts to prevent serious accidents.

  As a preventive maintenance technique for insulation diagnosis of gas-insulated equipment, there has been proposed a system that makes a determination, including whether or not a partial discharge has occurred, using a neural network. This system includes a plurality of partial discharge detection units that detect partial discharges inside the gas-insulated equipment, and a data processing device that monitors partial discharge signals from each partial discharge detection unit via a ring-shaped network. These conventional techniques are described in Patent Documents 1 to 3, for example.

Japanese Patent Application Laid-Open No. 07-181218 Unexamined-Japanese-Patent No. 11-237432 JP, 2015-78882, A

  An example of the learning method of the neural network in the prior art is as follows. (1) A sample having a simulated defect, for example, a void in an insulator, is charged to generate a partial discharge. Teacher data such as the magnitude (amplitude) of the partial discharge signal detected by the partial discharge sensor, the partial discharge occurrence phase with respect to the applied voltage, and the occurrence frequency of the partial discharge are input to the neural network. Thereafter, when teacher data is input, learning of the neural network is performed so that a predetermined value is output from the output layer.

(2) A partial discharge is generated by applying a voltage to a sample having another simulated defect, for example, a protrusion having a high voltage conductor, and teacher data on the simulated defect is collected by the partial discharge sensor. Thereafter, when teacher data of another simulated defect is input, learning of the neural network is performed so that a predetermined value is output from the output layer.

(3) Gather teacher data when partial discharge is not generated using a sound sample, and when sound teacher data is input, learn neural network so that the corresponding value is output. Do.

(4) After performing learning a sufficient number of times as described in (1) to (3) above, it is determined whether a new sample is a cause of partial discharge or presence or absence.

  In general, partial discharge sensors used for learning of neural networks are installed at one place for gas insulation equipment of a three-phase collective configuration. However, when the partial discharge monitoring system is applied to a three-phase gas insulation device, the AC phase decomposition type pattern corresponding to the representative voltage signal of the main circuit indicates which phase of the three phases depends on the physical position of the cause of the partial discharge. It depends on whether the voltage is applied.

  For this reason, in order to collect teacher data for learning diagnostic software such as a neural network, it is necessary to collect data using materials having simulated defects, or to collect data by changing the applied voltage phase.

  Further, in the conventional system, the time interval for acquiring, sampling and digitizing the signal from the partial discharge sensor is, for example, an interval in which one cycle is divided into 64. Therefore, in acquiring teacher data for neural network learning corresponding to the phase shift of the three-phase voltage, it is necessary to acquire according to the position of the simulated defect with respect to the high voltage conductor, and it takes time to acquire teacher data. . The same problem occurs in the generation of teacher data even in diagnostic processing units other than neural networks.

  An object of an embodiment of the present invention is to obtain a signal from a partial discharge sensor and appropriately select a time interval for sampling and digitizing, thereby reducing the time required for acquiring teacher data. It is in providing a discharge monitoring system.

The system of the present embodiment for achieving the above object has the following configuration.
(1) A partial discharge sensor provided in a three-phase collective type electric device.
(2) A conversion device that samples the partial discharge signal at a multiple of 12 for each cycle of the AC voltage phase in synchronization with the main circuit voltage phase of the electrical device.
(3) A data processing device connected to the conversion device.
(4) A pattern generation unit which converts the sampled partial discharge signal into a pattern for the AC voltage phase of the generated partial discharge.
(5) A diagnostic processing unit that determines the cause of the partial discharge or the presence or absence of the partial discharge based on the pattern.
(6) In the process of generating diagnostic knowledge by learning based on teacher data, the diagnostic processing unit provides a simulated defect between each phase or between each phase and the ground, and a partial discharge signal generated by the simulated defect Based on the AC phase decomposition pattern obtained from the above, generating the basic teacher data, and by shifting the phase of the basic teacher data in units of the sampling interval, between each phase not caused by the simulated defect or Generate teacher data between each phase and ground.

Block diagram showing the configuration of the partial discharge monitoring system of the embodiment The block diagram which shows the structure of the data processing part in the system of FIG. Cross-sectional view of a three-phase collective gas insulation device with a partial discharge detector Acquisition data in the partial discharge monitoring system of the embodiment Explanatory diagram for explaining the configuration and function of the neural network Explanatory drawing which shows the alternating current phase decomposition type pattern of the partial discharge signal in embodiment Explanatory drawing which shows the case where the reference voltage is changed about the alternating current phase-resolved type | mold pattern of the partial discharge signal in other embodiment.

[1. First embodiment]
[1-1. Constitution]
Hereinafter, a first embodiment of the present invention will be described with reference to the drawings.
(1) Overall System Configuration FIG. 1 is an explanatory diagram of a partial discharge monitoring system that performs partial discharge identification. A partial discharge sensor 1 installed in a three-phase collective gas-insulated electrical apparatus is connected to a converter 3 by a coaxial cable 2. The conversion device 3 is connected to the data interface device 7 via the data transmission path 4, the HUB 5 and the data transmission path 6. The data interface unit 7 is connected to the data processing unit 9 via the data transmission line 8.

  The data interface device 7 receives a representative voltage signal of the main circuit of the electric device, for example, an A-phase voltage, and transmits a reference signal synchronized with the phase of the main circuit voltage to the conversion device 3. The converter 3 samples the partial discharge signal in synchronization with the main circuit voltage phase. That is, the conversion device 3 detects the signal of the sensor 1 for each time-divided region on the basis of the zero cross timing of the phase of the main circuit voltage signal detected by the data interface device 7.

  The data processing device 9 is processed by a data transmission / reception unit 91 that transmits / receives data via the data transmission path 8 and a data processing unit 92 that performs conversion of received data, creation of teacher data, and partial discharge determination processing. And a storage unit 93 for storing the stored data. The data processing device 9 includes a display device 94 that displays a generated pattern of the processed partial discharge signal, and an input / output unit 95 including an input unit such as a keyboard and a mouse, and an output unit such as a contact signal.

  As shown in FIG. 2, the data processing unit 92 converts the partial discharge signal received from the data transmission / reception unit 91 into a pattern (generally called an AC phase decomposition type pattern) with respect to the AC voltage phase of the generated partial discharge. It has a pattern generation unit 921. On the output side of the pattern generation unit 921, a diagnosis processing unit 922 such as a neural network is mounted.

  In the process of generating diagnostic knowledge by learning based on teacher data, the diagnostic processing device 922 provides a simulated defect either between each phase or between each phase and the ground, and is obtained from the partial discharge signal generated by this simulated defect. Based on the AC phase-resolved pattern, the basic teaching data is generated, and the phase of the basic teaching data is shifted with the sampling interval as a unit, so that each phase or between each phase and the ground which is not caused by a simulated defect Generate teacher data for The diagnosis processing device 922 performs learning according to the teacher data generated in this way, and diagnoses what partial discharge the signal input from the pattern generation unit 921 is.

  The storage device 93 stores the magnitude (amplitude) of the partial discharge signal, the partial discharge occurrence phase with respect to the applied voltage, and the occurrence frequency of the partial discharge when the partial discharge is detected. At the time of actual operation of the neural network, these pieces of information are input to the diagnosis processing unit 922, and the determination result of the presence or absence of partial discharge is displayed on the display device 94. In the display device 94, the generation pattern of the partial discharge signal is displayed in a graph in which the horizontal axis is an electrical angle of 0 to 360 °.

(2) Detection of Partial Discharge As shown in FIG. 3, the three-phase batch type gas-insulated electrical apparatus includes the A-phase conductor 15A, the B-phase conductor 15B and the C-phase conductor 15C disposed in the tank 14, and the tank 14 A partial discharge sensor 1 provided at a part of In the gas-insulated electrical equipment with a three-phase collective configuration, partial discharge is performed between A phase to ground, B phase to ground, C phase to ground, A phase to B phase, B phase to C phase, C phase to A phase These partial discharges are detected by the sensor 1, and the diagnosis processing unit 922 specifies the generation location based on the AC phase decomposition pattern.

(3) Configuration of AC Phase-Resolved Pattern FIG. 4 shows the configuration of the AC phase-resolved pattern converted in the data processing unit 92 based on the partial discharge signal received from the partial discharge sensor. In FIG. 4, the partial discharge signal corresponding to one cycle of the system voltage is displayed on the horizontal axis with an electrical angle of 0 to 360 °. Further, the magnitude of the partial discharge signal in the time domain of every 15 ° obtained by dividing the electrical angle 0 to 360 ° into 24 is displayed on the vertical axis. This AC phase-resolved pattern is used as teacher data and partial discharge determination data of the diagnosis processing unit 922.

(4) Structure and Function of Neural Network FIG. 5 is an explanatory diagram for explaining the structure and function of the diagnosis processing unit 922. As shown in FIG. The diagnostic processing unit 922 has a three-layer structure of an input layer 11, an intermediate layer 12, and an output layer 13 as an example. The cells 111 to 113 of the input layer 11 are connected to the cells 121 to 123 of the intermediate layer 12, and the cells 121 to 123 of the intermediate layer 12 are cells 131 (with partial discharge) and 132 (without partial discharge) of the output layer 13 ).

  The cells 111 to 113 of the input layer 11 correspond to a specific phase range (for example, 0 ° to 18 °) and an amplitude range (for example, 10 to 20 pC) among the information stored in the partial discharge measuring device 7. The number of signals satisfying the range becomes the input value of each cell. When a value is input to each cell of the input layer 11, evaluation is performed in the intermediate layer 12 and the output layer 13, and evaluation values are obtained in the cells 131 and 132 of the output layer 13. The cause of the partial discharge or the presence or absence is determined based on the magnitude relationship of the evaluation values.

(5) Learning Method of Neural Network The learning method of the diagnosis processing unit 922 will be described. First, a sample having a simulated defect 1, for example, a void in an insulator, is charged to generate a partial discharge. The output of the cell 131 of the output layer 13 when the magnitude (amplitude) of the partial discharge signal detected by the partial discharge detection sensor, the partial discharge occurrence phase with respect to the applied voltage, and the occurrence frequency of partial discharge are input to the diagnosis processing unit 922 The neural network is trained so that the value is “1”, and the output values of the cell 132 and the cell 133 are “0”.

  A sample having a simulated defect 2, for example, a protrusion having a high-voltage conductor is applied to generate a partial discharge. The output value of the cell 132 of the output layer 13 is “1”, and the output values of the cell 131 and the cell 133 are The neural network is learned so that it becomes “0”. Further, when a partial discharge is not generated using a healthy sample, the neural network is set so that the output value of the cell 133 of the output layer 13 is “1” and the output values of the cell 131 and the cell 132 are “0”. To learn. Then, after learning has been performed a sufficient number of times, it is determined whether or not the cause of partial discharge or the presence or absence of a new sample.

[1-2. Action]
Hereinafter, the operation of the present embodiment will be described. FIGS. 6A and 6B show AC phase-resolved patterns of partial discharge signals detected.

(1) AC phase-resolved pattern FIG. 6A is an AC phase-resolved pattern when there is a void defect in the insulator between the A-phase conductor and the ground (tank 14). The horizontal axis is displayed as an electrical angle of 0 to 360 degrees of the A phase-ground voltage. In FIG. 6A, partial discharge occurs in the voltage phase of the voltage increase region 0 to 90 ° and 165 to 285 ° with respect to the A-phase applied voltage.

  FIG. 6B shows an AC phase-resolved pattern in the case where there is a void defect in the insulator between the A phase and the B phase. In FIG. 6B, since partial discharge occurs according to the voltage phase between the A and B phases, the partial discharge occurs at the voltage phases of −30 ° to 60 ° and 135 ° to 355 ° of the applied voltage of the A phase. Has occurred.

(2) Creation of basic teacher data as a basis and learning of diagnosis processing unit 922 As described in (1), the AC phase-resolved pattern of the partial discharge signal in the gas-insulated electrical apparatus of the three-phase batch configuration is a physical cause of the partial discharge. As it differs depending on the target position, in order to collect teacher data for learning diagnostic software such as a neural network, it is necessary to collect data by using materials having simulated defects different in physical position.

  On the other hand, in the present embodiment, the sampling number of the electrical angle 0 to 360 ° of the voltage is 24 which is a multiple of 12. Therefore, the AC phase-resolved pattern data of the partial discharge signal corresponding to FIG. 6B can be realized by shifting the data in FIG. 6A by, for example, two sampling data with a sampling interval as a unit. It becomes.

  In gas-insulated electrical equipment with a three-phase collective configuration, it occurs between A phase to ground, B phase to ground, C phase to ground, A phase to B phase, B phase to C phase, and C phase to A phase These partial discharges have the same distance between the conductors of each phase and the tank, and the same distance between the conductors of each phase. It is almost identical. Since the distances between the conductors 15A to 15C of each phase and the sensor 1 are not exactly equal, there is a slight difference, but when viewed as teacher data of the diagnostic processing unit 922, it can be ignored.

  Therefore, in the present embodiment, first, a partial discharge signal generated due to a simulated defect between A phase and ground is converted into AC phase-resolved pattern data by pattern generation unit 921, and teacher data creation unit 922 converts A phase to ground. Teacher data is created, and the diagnosis processing unit 922 performs learning. After that, the diagnostic processing unit 922 performs learning even in the case where partial discharge using other simulated defects and a healthy sample is not generated. In this way, the AC phase decomposition pattern corresponding to the cause of partial discharge and the physical position of the cause is acquired for the phase A to the ground, and the diagnosis processing unit 922 is made to learn this data.

  The A-phase to ground teacher data created in this manner, that is, the AC phase-resolved pattern of FIG. 6A and the partial discharge determination results obtained thereby are stored in the storage unit 93, and It is the basis of teacher data. The learning result of the diagnosis processing unit 922 is also stored in the storage unit 93.

(3) Preparation of Teacher Data of Other Portion and Learning of Diagnosis Processing Unit 922 After learning of the diagnosis processing unit 922 is completed regarding the partial discharge between the A phase and the ground, the A phase stored in the storage unit 93- By reading out the AC phase-resolved pattern data between the ground and shifting the phase by −30 °, the AC phase-resolved pattern generated between the A and B phases in FIG. Thereafter, an AC phase-resolved pattern whose phase is shifted is input to the diagnosis processing unit 922 as teacher data for partial discharge detection generated between the A and B phases, and the diagnosis processing unit 922 performs learning.

  Similarly, for partial discharges other than A-phase to ground and A-phase to B-phase, teacher data is created by shifting the phase of the AC phase-resolved pattern between A-phase to ground, and diagnostic processing is performed. The part 922 is studied. In this case, how much the phase is shifted with respect to which partial discharge is determined by giving the same simulated defect to each part, detecting the AC phase-resolved pattern, and comparing the phases.

(4) Judgment at the time of operation After learning about partial discharge of each part is completed in the diagnosis processing unit 922, the partial discharge signal from the sensor 1 is taken into the data processing unit 92 as in the conventional partial discharge monitoring system. . The pattern generation unit 921 of the data processing unit 92 converts the partial discharge signal into an AC phase-resolved pattern, and processes this pattern by the diagnostic processing unit 922 to generate any partial discharge at any point. It is determined whether or not.
[1-3. effect]
As described above, in the present embodiment, when collecting teacher data for learning of the diagnosis processing unit 922, data of another physical position is acquired by collecting data of one partial discharge occurrence factor and physical position. Can be easily obtained. In particular, in the gas-insulated electrical equipment having a three-phase collective configuration, the simulated defects are as follows: A-phase to ground, B-phase to ground, C-phase to ground, A-phase to B-phase, B-phase to C-phase, C-phase It was necessary to acquire the teacher data by collecting the partial discharges generated between the phases A and arranged by the sensors arranged in the respective phases. In this embodiment, since the number of samplings per AC voltage phase cycle is a multiple of 12, for example, it is possible to shift sampling data acquired under the condition between A phase and ground in units of 30 °, and so on. It becomes possible to easily create teacher data of conditions. Therefore, it is possible to provide an inexpensive and easy-to-use partial discharge monitoring system.

[2. Other embodiments]
Although several embodiments of the present invention have been described, these embodiments are presented by way of example and are not intended to limit the scope of the invention. These embodiments can be implemented in other various forms, and various omissions, replacements, and modifications can be made without departing from the scope of the invention. These embodiments and modifications thereof are included in the scope of the present invention and the gist thereof, and are also included in the invention described in the claims and the equivalent scope thereof. The following is an example.

(1) FIG. 7A shows a partial discharge appearance displayed on the display unit 94 in which the reference voltage of the voltage phase axis is the A phase voltage. FIG. 7 (b) shows the same data as FIG. 7 (a), and a partial discharge mode in which the reference voltage of the voltage phase axis is the voltage between A phase and B phase. In this modification, the display unit 94 can be provided with a function of changing the reference voltage of the voltage phase axis from the A phase voltage to the A phase-B phase voltage for display. As a result, it is possible to display the phenomenon in which the causes of partial discharge occurrence are the same as the same visual image, and it is possible to provide a convenient partial discharge monitoring system.

(2) In the illustrated embodiment, the number of samplings at an electrical angle of 0 to 360 ° is set to 24, but the number of samplings is not limited to 24 and may be a multiple of 12. In order to shift the phase of the AC phase-resolved pattern to create other teacher data, it is sufficient if the phase can be shifted according to the number of samplings.

(3) The embodiment can be applied to a partial discharge monitoring system having a diagnosis processing unit that requires teacher data other than the diagnosis processing unit 922.

1 ... Partial Discharge Sensor 2 ... Coaxial Cable 3 ... Converter 4 ... Data Transmission Line 5 ... HUB
6 Data transmission path 7 Data interface device 8 Data transmission path 9 Data processing device 11 Input layer 12 Intermediate layer 13 Output layer 14 Tank 15A Phase A conductor 15B Phase B conductor 15C Phase C Conductor 91 ... Data transmission / reception unit 92 ... Data processing unit 921 ... Pattern generation unit 922 ... Diagnostic processing unit 93 ... Storage unit 94 ... Display unit 95 ... Input / output unit

Claims (2)

A partial discharge sensor provided in a three-phase collective type electric device;
A converter that samples the partial discharge signal in a multiple of 12 for each cycle of the AC voltage phase in synchronization with the main circuit voltage phase of the electrical device;
A data processor connected to the converter;
The data processing device includes:
A pattern generation unit for converting the sampled partial discharge signal into a pattern for the AC voltage phase of the generated partial discharge;
And a diagnostic processing unit that determines the cause of the partial discharge or the presence or absence of the partial discharge based on the pattern .
The diagnostic processing unit generates diagnostic knowledge by learning based on teacher data.
A simulated defect is provided between each phase or between each phase and the ground, and based on the AC phase-resolved pattern obtained from the partial discharge signal generated by the simulated defect, basic teaching data is generated,
A partial discharge monitoring system that generates teacher data between phases or between each phase and ground that is not caused by a simulated defect by shifting the phase of the basic teacher data in units of the sampling interval . The data processing device is provided with a display unit,
The display unit, a reference voltage of the voltage phase axis that are displayed, partial discharge monitoring system of claim 1 wherein is for displaying is changed to the phase voltage and each phase voltage.

Images