JP3852097B2 - Insulation diagnostic system - Google Patents

Description

  The present invention relates to an insulation diagnosis system for diagnosing an insulation state of electrical equipment provided in a factory, a building, or the like.

Conventionally, when diagnosing insulation degradation that occurs in electrical equipment such as power cables and electric motors for power transfer, attention is focused on partial discharge that occurs with insulation degradation. The characteristic amount used for quantifying the partial discharge characteristics generally includes a pulse current generated by the partial discharge, a discharge generation phase φ with respect to an applied voltage obtained based on the pulse current, a charge amount q of the partial discharge, Occurrence frequency n and the like. These discharge generation phase φ, charge amount q, and generation frequency n are obtained based on a pulse current generated by partial discharge.
The conventional insulation diagnosis system applies a neural network to the distribution pattern of discharge generation phase φ−charge amount q−generation frequency n formed by the discharge generation phase φ, the charge amount q, and the generation frequency n. A paper on this (see Non-Patent Document 1) has been published.

However, the neural network used in the above conventional insulation diagnosis system expresses a complex non-linear function by superimposing sigmoid functions (deterministic). It is not possible to include a temporal causal relationship between the feature quantities. In other words, the neural network is suitable for expressing a deterministic signal that does not depend on the passage of time, but it changes probabilistically with the passage of time, such as a feature value obtained based on the pulse current generated by partial discharge. There is a problem that the diagnostic accuracy of the insulation state is low.
Trend of Partial Discharge Signal Processing Technology (Author Hitoda), Proceedings of 2003 Annual Conference of the Institute of Electrical Engineers of Japan [1] Fundamentals 1-S3 (10)

  Therefore, in the present invention, it is a problem to be solved to provide an insulation diagnosis system that can improve the diagnosis accuracy of the insulation state of electrical equipment by applying a hidden Markov model that is a stochastic statistical pattern recognition means. To do.

The above problem can be solved by the insulation diagnosis system described in the claims.
According to the insulation diagnosis system according to claim 1, the learning means includes a hidden Markov model representing each feature obtained by patterning a feature quantity related to insulation detected by the feature quantity detection means or data obtained based on the feature quantity. Since each parameter is stored in advance as an insulation state-corresponding parameter, the diagnostic means can be obtained based on the feature value related to insulation detected in the operating state of the electrical equipment or the feature value. The output probability of the pattern obtained by patterning the obtained data is calculated based on each parameter stored in advance in the learning means, and the insulation state corresponding to the parameter having the maximum output probability is calculated as the insulation state of the electrical equipment. It is diagnosed and judged.
Thus, by applying the hidden Markov model that is a probability statistical pattern recognition means, the diagnostic accuracy of the insulation state of the electrical equipment can be increased.

  In addition, according to the insulation deterioration diagnosis system according to the second aspect, since parameters stored in advance in the learning means can be updated or exchanged, versatility can be improved.

Here, the concept of the hidden Markov model will be described with reference to the schematic diagram shown in FIG.
In FIG. 1, a11, a12, a22, a23, a33, a34 indicate state transition probabilities, and aij is the probability of transition from state Si to state Sj. Further, bi (x) indicates the probability of outputting the feature quantity x in the state Si. S1 indicates an initial state, and S4 indicates a final state. Thus, the hidden Markov model has a state transition probability, an output probability, and an initial state probability πi (probability that the initial state is Si. In the example of FIG. 1, since only S1 is the initial state, π1 = 1, π2 = π3 = ... = 0) as parameters, and each parameter is stored for each normal state of the insulator and each insulation deterioration state to be recognized, which corresponds to the function of the learning means described above. To do.

  In the present invention, as described above, each parameter is stored for each normal state of the insulator or each insulation deterioration state that is desired to be recognized. For the data pattern obtained from the quantity, the probability of generating the pattern is calculated based on the respective parameters stored in the learning means, and the electric power corresponding to the parameter having the maximum probability is calculated. In addition to diagnosing the insulation state of the equipment, that is, whether the insulation of the electrical equipment is normal or deteriorated, the contents are determined when it is determined that the insulation is deteriorated.

  Note that the pattern recognition method using the hidden Markov model can reduce the number of data compared to the pattern recognition method that simply compares the distribution waveform of the feature amount. As a result, the storage capacity for storing data can be reduced, and the amount of calculation until the diagnosis result is calculated can be greatly reduced.

  ADVANTAGE OF THE INVENTION According to this invention, the diagnostic accuracy of the insulation state of an electrical installation can be improved by applying the hidden Markov model which is a stochastic statistical pattern recognition means.

Next, an embodiment of the present invention will be described.
FIG. 2 is a system diagram conceptually showing the configuration of the insulation diagnosis system 1. For example, an insulator 2 used for a power cable or the like is sandwiched between an electrode plate 3 and an electrode plate 4, and the electrode plate 3 is By applying a voltage and connecting a detection impedance Z used as a sensor between the electrode plate 4 and the ground, a discharge pulse current generated due to insulation deterioration of the insulator 2 is detected and detected by the detection impedance Z A detection signal of the discharge pulse current is input to the partial discharge measuring device 5.

  In the above case, the discharge pulse current is detected as the characteristic amount related to the insulation of the insulator 2, but as the characteristic amount, in addition to this, the maximum discharge charge amount or the average within the discharge pulse current obtained based on the characteristic amount is used. The charge amount q such as the discharge charge amount, the number n of discharge occurrences, the discharge generation phase φ, the frequency f, the vibration v, the sound, the electromagnetic wave, and the like can also be used as feature amounts. In addition, a pattern distribution obtained by combining any of the above feature quantities, for example, a time series pattern of discharge generation phase φ−charge amount q, discharge generation phase φ−charge amount q−number of discharge occurrences n, etc. It is included as a feature vector as a set. Furthermore, a frequency spectrum and a time series pattern of the frequency spectrum are included as feature quantities. Or the signal produced by the computer provided in the inside of the partial discharge measuring apparatus 5 may be made into a feature-value based on the signal detected by the sensor.

  FIG. 3 shows a general waveform of a pulse current generated during partial discharge. FIG. 4 shows a distribution pattern of discharge generation phase φ−charge amount q−generation frequency n obtained based on the partial discharge pulse current detected as the feature amount. Here, since the discharge generation phase φ represents the phase of the applied voltage, it can be regarded as a quantity representing a kind of time. That is, the distribution pattern of discharge generation phase φ−charge amount q−generation frequency n can also be considered as a time series distribution pattern of charge amount q−generation frequency n per unit time. Thus, the above-described hidden Markov model can be applied to the distribution pattern itself of the discharge generation phase φ−charge amount q−generation frequency n. However, the discharge generation phase φ is obtained with reference to the phase of the applied voltage.

  Also, even if the pattern to be recognized is not composed of a single feature quantity but is composed of multi-feature quantities (multi-dimensions), the time-series patterns of those multi-dimensional distributions are time-sequential. Hidden Markov models can be applied as patterns. Furthermore, by using a principal component analysis or a self-organizing map, the present invention can be applied to a time series for a multi-characteristic (multi-dimensional) reduced to one dimension.

  In addition to the detection impedance Z described above, there are a CT (current transformer) sensor, an AE sensor, a UHF sensor, a Rogowski coil, and the like as a sensor for detecting a characteristic amount related to insulation deterioration. Application examples of these sensors will be described later.

  FIG. 5 is a block diagram showing the overall configuration of the insulation diagnosis system 1. As shown in FIG. 5, since the feature amount detected by the sensor 10 that collectively refers to the various sensors is generally an analog signal, each analog signal is converted into a digital signal by the A / D conversion circuit 11. Then, it is input to the learning unit 12 or the diagnosis unit 13.

  The learning unit 12 inputs various feature amounts detected by the sensor 10, and hides a Markov model of each feature amount pattern corresponding to a predetermined normal state of the insulating portion and a deteriorated state of the insulating portion. Is parameterized on the basis of the above and each parameter is stored.

  The diagnosis unit 13 calculates the output probability of the feature amount pattern detected in the operating state of the electrical equipment based on the respective parameters stored in advance in the learning unit 12, and the output probability is maximum. By determining the following parameters, the insulation state of the insulation part of the electrical equipment is diagnosed and determined.

  The display unit 14 displays the insulation state of the insulating unit of the electrical equipment determined by the diagnostic unit 13 and is a display unit of a computer that constitutes the learning unit 12 and the diagnostic unit 13. The insulation state of the insulation part of the electrical equipment determined by the diagnosis unit 13 is displayed on the computer display as described above, and can be transmitted to a central monitoring device of a factory or building, for example, by communication means.

  The A / D conversion circuit 11, the learning unit 12, the diagnosis unit 13, and the display unit 14 described above constitute a partial discharge measuring device 5 described below.

FIG. 6 is a system explanatory diagram showing the overall configuration of the insulation diagnostic system 1 for diagnosing the state of the insulator 22 of the power cable 20 to which a high voltage is applied to the conductor 21.
As shown in FIG. 6, one end of a cable shielding ground wire 24 is connected to the shielding tape 23, and the other end of the cable shielding ground wire 24 is grounded. In this state, the CT sensor 25 is clamped to the cable shielding ground line 24, the partial discharge pulse current generated with the deterioration of the insulator 22 is detected, and the detection signal is input to the partial discharge measuring device 5. Has been.

  The partial discharge measuring device 5 is configured as shown in FIG. 5 described above, and applies the hidden Markov model to the feature quantity obtained based on the partial discharge pulse current detection signal output from the CT sensor 25. By doing so, the insulation state of the insulator 22 is diagnosed. When a partial discharge occurs due to a traumatic defect of the power cable 20, the feature pattern has characteristics similar to those when a partial discharge occurs due to a sealed void. This is useful not only for the sealed void of the cable 20 but also when a partial discharge occurs due to a traumatic defect.

  FIG. 7 is a system explanatory diagram showing the overall configuration of the insulation diagnosis system 1 for diagnosing the insulation state in the insulation connection portion 31 to which the power cables 20a and 20b are connected. When the insulating connecting portion 31 is formed of a corrosion-proof vinyl sheath, as shown in FIG. 7, metal foil electrodes 33a and 33b are attached to both sides of the sheath insulation tube 32, and the metal foil electrodes 33a and 33b are sandwiched. By connecting the detection impedance Zd, even if partial discharge occurs in either of the power cables 20a and 20b or on both sides of the insulation connection portion 31 with the sheath insulation tube 32 interposed therebetween, the detection impedance Zd The partial discharge pulse current can be detected. With this configuration, the partial discharge pulse current detection signal detected by the detection impedance Zd is input to the partial discharge measurement circuit 5.

  The partial discharge measuring apparatus 5 configured as shown in FIG. 5 described above has a feature amount pattern obtained based on the partial discharge pulse current detection signal output from the detection impedance Zd, for example, the discharge generation phase φ−. By applying a hidden Markov model to the distribution pattern of the charge amount q−the occurrence frequency n, the insulation state of the insulation connection portion 31 is diagnosed.

  FIG. 8 is a system explanatory diagram showing the overall configuration of the insulation deterioration diagnosis system 1 when the insulating connection portion 31 in FIG. 7 is formed of a steel tube, and the metal foil electrodes 33a, 33b is not used.

  FIG. 9 shows the arrangement of sensors for detecting a partial discharge caused by a defect in the gas-insulated switchgear (GIS) 40. The external UHF sensor 41 shown here detects an electromagnetic wave signal in the UHF band where noise is considered to be low, and is often installed when a gas insulated switchgear (GIS) 40 is already installed. Similarly, the internal UHF sensor 42 detects an electromagnetic wave signal in the UHF band, but is installed when a gas insulated switchgear (GIS) 40 is newly installed or updated. Furthermore, the AE sensor 43 detects minute vibrations of the tank 44 when a minute pressure wave is generated and propagates in the gas when a partial discharge occurs due to a defect in the gas insulated switchgear (GIS) 40. is there. Since the detection signals output from the external UHF sensor 41, the internal UHF sensor 42, and the AE sensor 43 are input to a partial discharge measuring device (not shown), the partial discharge measuring device detects each detection signal. The insulation state of the gas-insulated switchgear (GIS) 40 is diagnosed by applying a hidden Markov model to a pattern obtained based on the signal (feature value), for example, a φ-qn distribution.

  FIG. 10 shows that a plurality of AE sensors 51 are arranged outside the transformer 50 in order to detect partial discharges generated inside the transformer 50, and the detection signals output from the respective AE sensors 51 are partially displayed. FIG. 2 is a system explanatory diagram showing an overall configuration of an insulation diagnosis system 1 that diagnoses an insulation state of a transformer 50 by inputting to the discharge measuring device 5.

  FIG. 11 is a system explanatory diagram showing the overall configuration of the insulation diagnosis system 1 for diagnosing the insulation state of the electric motor 60. As shown in FIG. 11, high-frequency CT sensors 62R, 62S, and 62T are attached to the electric wires 61R, 61S, and 61T connected to the respective phases of the electric motor 60 to detect partial discharge pulse currents. The detection signals output from 62S and 62T are input to the partial discharge measuring device 5 to diagnose the insulation state of the electric motor 60. In this case, the high-frequency CT sensors 62R, 62S, and 62T are attached to the electric wires 61R, 61S, and 61T, respectively, but the partial discharge pulse current is collectively collected by one high-frequency CT sensor for the three-phase electric wires 61R, 61S, and 61T. May be detected. Alternatively, a discharge pulse current flowing in a ground line (not shown) may be detected by a high frequency CT sensor.

  The insulation diagnostic system 1 described above includes a gas circuit breaker (GCB), in addition to the power cable 20, the insulation connection part 31, the gas insulation switchgear (GIS) 40, the transformer 50, and the electric motor 60 that are the diagnosis targets of the insulation state. Gas insulation transformers, C-GIS, pipeline air transmission lines, capacitors, and the like can be targeted for insulation diagnosis.

In addition, each insulation state corresponding | compatible parameter memorize | stored in the memory | storage device of the learning part 12 of the insulation diagnostic system 1 can be updated or exchanged. In this case, it is possible to update and exchange the insulation state corresponding parameter using a computer provided in the learning unit 12 or to exchange the storage device itself. When the computer provided in the learning unit 12 is used to update and exchange the insulation state corresponding parameter, the computer again creates a parameter corresponding to each insulation state and stores it in the storage device.
In addition, when replacing the storage device itself, an external computer having a processing speed higher than that of the computer provided in the learning unit 12 can be used, so that the above parameters can be created in a short time. The storage device used here is a memory module, a memory stick, a memory card, or the like. Thus, since each insulation state corresponding | compatible parameter memorize | stored in the memory | storage device of the learning part 12 of the insulation diagnostic system 1 can be updated or exchanged, the versatility of the insulation diagnostic system 1 can be improved.

It is a conceptual explanatory drawing of a hidden Markov model. 1 is a system diagram conceptually showing the configuration of an insulation diagnosis system 1. FIG. The general waveform of the pulse current generated at the time of partial discharge is shown. It is a distribution pattern of discharge generation phase φ−charge amount q−generation frequency n. 1 is a block diagram showing an overall configuration of an insulation diagnosis system 1. FIG. It is a system explanatory view showing the composition of insulation diagnostic system 1 for diagnosing the insulation state of a power cable. It is a system explanatory view showing the composition of insulation diagnostic system 1 for diagnosing the insulation state of an insulation connection part. It is a system explanatory view showing the composition of insulation diagnostic system 1 for diagnosing the insulation state of an insulation connection part. It is a sensor layout for detecting an insulation state of a gas insulated switchgear. It is system explanatory drawing which showed the structure of the insulation diagnostic system 1 for diagnosing the insulation state of a transformer. It is a system explanatory view showing the composition of insulation diagnostic system 1 for diagnosing the insulation state of an electric motor.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Insulation diagnostic system 2 Insulator 3 Electrode plate 4 Electrode plate 5 Partial discharge measuring device 10 Sensor 11 A / D conversion circuit 12 Learning part 13 Diagnosis part 14 Display part 20 Power cable 25 CT sensor 31 Insulation connection part Z, Zd Detection impedance 50 Transformer 60 Electric motor

Claims (2)

  Feature quantity detection means for detecting feature quantities relating to insulation of electrical equipment, and feature quantities relating to insulation detected by the feature quantity detection means or respective patterns obtained by patterning data obtained based on the feature quantities are hidden Markov Based on the model, the learning means for storing each parameter in advance as an insulation state-corresponding parameter, and the feature value related to insulation detected in the operating state of the electrical equipment or the feature value The output probability of the pattern obtained by patterning the obtained data is calculated based on the respective parameters stored in advance in the learning means, and the insulation state corresponding to the parameter having the maximum output probability is calculated. Insulation diagnostic system comprising diagnostic means for determining whether or not Temu.   2. The insulation diagnosis system according to claim 1, wherein the parameters stored in the learning means can be updated or exchanged.

Images