Switch cabinet discharge detection method and system based on spectral characteristics, and computer equipment

Abstract

The invention belongs to the technical field of power equipment fault diagnosis, and particularly relates to a switch cabinet discharge detection method based on spectral characteristics, which comprises the following steps: collecting partial discharge multispectral signals collected by a partial discharge multispectral sensor at different positions in a switch cabinet; performing feature extraction on the partial discharge multispectral signal to obtain a multispectral partial discharge feature quantity; the multispectral partial discharge characteristic quantity comprises a discharge pulse number, an average pulse amplitude, a maximum discharge pulse amplitude and an average discharge pulse amplitude after windowing; inputting the multispectral partial discharge characteristic quantity into a partial discharge diagnosis model, and carrying out partial discharge type identification and partial discharge severity division; the partial discharge diagnosis model adopts a k-means clustering algorithm used for dividing the severity of partial discharge and a deep neural network algorithm used for identifying the type of partial discharge. Multi-spectral information is fully mined, and reliable diagnosis of partial discharge can be realized without depending on phases.

Description

Technical Field

[0001] The present invention belongs to the technical field of power equipment fault diagnosis, and in particular relates to a switch cabinet discharge detection method, system and computer equipment based on spectral characteristics. Background Art

[0002] In power equipment, switchgear is one of the main components, and its operation has an important impact on the safety and stability of the distribution cabinet. From the analysis of switchgear failure, it can be seen that the insulation failure of the switchgear is a process of gradual development and gradual deterioration. In the long-term operation stage of the switchgear, it often grows from the early insulation degradation state to the flashover breakdown fault, which greatly reduces the ability of the winding to resist the impact of short-circuit current. Partial discharge becomes one of the important factors that endanger power equipment. With the further expansion of the application field of switchgear, the problem of switchgear insulation state identification needs to be solved urgently. At present, various partial discharge detection methods have been proposed one after another: such as pulse current method based on charge transfer; ultrasonic method based on sound wave propagation; ultra-high frequency method based on electromagnetic wave propagation and optical measurement method based on optical radiation. Among them, the optical measurement method measures the optical radiation signal generated by partial discharge, which has the characteristics of strong anti-interference and high sensitivity. Since the optical radiation signal is generated during the whole process of partial discharge, the optical measurement method can be used to conduct in-depth analysis of partial discharge characteristics and insulation aging.

[0003] However, the application of traditional optical measurement is mainly to detect discharge light pulses, and the spectral information contained in the optical signal that can reflect the intrinsic characteristics of the discharge is not fully utilized; the traditional optical measurement method uses the light radiation intensity as the analysis basis, and the signal attenuation during the light propagation process affects the light radiation intensity, making it impossible to evaluate the discharge energy. Therefore, it is of great significance to mine the discharge spectrum information and apply multi-spectral analysis technology to the field of partial discharge diagnosis. Summary of the invention

[0004] The purpose of the present invention is to provide a switch cabinet discharge detection method based on spectral characteristics, which solves the problem that the traditional optical measurement method cannot evaluate the discharge energy.

[0005] The present invention is achieved through the following technical solutions:

[0006] A switch cabinet discharge detection method based on spectral characteristics comprises the following steps:

[0007] S1, collecting partial discharge multi-spectral signals collected by the partial discharge multi-spectral sensor at different positions in the switch cabinet;

[0008] S2, extracting features from the partial discharge multi-spectrum signal to obtain multi-spectrum partial discharge feature quantities;

[0009] The multi-spectral local discharge characteristic quantities include the number of discharge pulses, the average pulse amplitude, the maximum discharge pulse amplitude and the average discharge pulse amplitude after windowing;

[0010] S3, inputting the multi-spectral partial discharge feature quantity into the partial discharge diagnosis model to identify the partial discharge type and classify the partial discharge severity;

[0011] The partial discharge diagnosis model adopts a k-means clustering algorithm and a deep neural network algorithm;

[0012] The deep neural network algorithm is used to identify the type of partial discharge, and the k-means clustering algorithm is used to classify the severity of partial discharge.

[0013] Furthermore, in S3, the construction process of the partial discharge diagnosis model is as follows:

[0014] A large number of partial discharge tests were conducted on typical discharge defect models. The discharge pulse characteristics in multiple bands were studied through multi-spectral pulse detection. The collected multi-spectral signals were synchronously measured and analyzed to extract multi-spectral partial discharge feature quantities, and a feature database for partial discharge diagnosis was established.

[0015] S2. The extracted multi-spectral partial discharge feature quantities were trained using the k-means clustering algorithm and the deep neural network algorithm. For the identification of partial discharge types, the recognition effect of the deep neural network algorithm was higher than that of the k-means clustering algorithm. For the severity of discharge, the recognition effect of the k-means clustering algorithm was higher than that of the deep neural network algorithm.

[0016] The partial discharge diagnosis model is constructed using the k-means clustering algorithm and deep neural network algorithm.

[0017] Furthermore, typical discharge defect models include tip discharge model, surface discharge model and suspension discharge model.

[0018] Furthermore, the spectral range of spectral pulse detection includes full band, 350nm, 400nm, 450nm, 500nm, 550nm, 600nm and 650nm.

[0019] Furthermore, in S2, the calculation formula for the number of discharge pulses is:

[0020]

[0021] The calculation formula for the proportion of the discharge pulse number component is:

[0022]

[0023] Where N iis the total number of discharge pulses collected for the i-th time under each voltage group; N is the average number of discharge pulses measured under each voltage group; n i % is the proportion of discharge pulse number in the required spectral band; N t is the number of discharge pulses in each spectral band; k is the number of spectral bands, and n is the total number of acquisitions.

[0024] Furthermore, in S2, the calculation formula for the average pulse amplitude is:

[0025]

[0026] Where, L m The discharge pulse amplitude obtained during the mth acquisition; L m,i is the i-th discharge pulse amplitude obtained during the m-th acquisition; L is the average discharge pulse amplitude obtained by multiple measurements under each voltage group; L i is the average discharge pulse amplitude in the target spectral band; n is the total number of acquisitions; L i % is the proportion of the average discharge pulse amplitude component in the target spectral band; k is the number of spectral bands.

[0027] Furthermore, in S2, the calculation formula for the maximum discharge pulse amplitude is:

[0028] A i =max{L i,1 ,L i,2 ,…,L i,n}

[0029]

[0030] In the formula, A i is the maximum value of the discharge pulse amplitude collected for the i-th time under each voltage group; L m,n is the discharge pulse amplitude collected for the mth time under each voltage group; a i % is the proportion of the maximum discharge pulse amplitude component in the required spectral band; A i is the maximum number of discharge pulses; k is the number of spectral bands. t is the discharge pulse amplitude obtained by the tth acquisition.

[0031] Furthermore, in S2, the calculation formula for the average discharge pulse amplitude after the window is opened is:

[0032]

[0033] In the formula, is the phase angle; t is the time of discharge pulse acquisition; T is the power frequency period; L k is the discharge pulse amplitude in each window, k is the number of spectral bands;

[0034] L is the average discharge pulse amplitude obtained by multiple measurements within the next cycle T of each voltage group;

[0035] In the calculation process, the collected discharge pulses are first superimposed in the same power frequency period T in the order of collection time t, and then the average discharge pulse amplitude in each phase window interval is calculated.

[0036] The present invention also discloses a switch cabinet insulation detection system based on spectral characteristics, comprising:

[0037] A data acquisition module, used to collect the partial discharge multi-spectral signals collected by the partial discharge multi-spectral sensor at different positions in the switch cabinet;

[0038] A feature extraction module is used to extract features from the partial discharge multi-spectrum signal to obtain multi-spectrum partial discharge feature quantities;

[0039] The multi-spectral local discharge characteristic quantities include the number of discharge pulses, the average pulse amplitude, the maximum discharge pulse amplitude and the average discharge pulse amplitude after windowing;

[0040] A partial discharge diagnosis module is used to input multi-spectral partial discharge feature quantities into a partial discharge diagnosis model to identify partial discharge types and classify partial discharge severity;

[0041] The partial discharge diagnosis model adopts a k-means clustering algorithm and a deep neural network algorithm;

[0042] The deep neural network algorithm is used to identify the type of partial discharge, and the k-means clustering algorithm is used to classify the severity of partial discharge.

[0043] The present invention also discloses a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the steps of the switch cabinet discharge detection method based on spectral features are implemented.

[0044] Compared with the prior art, the present invention has the following beneficial technical effects:

[0045] The present invention discloses a partial discharge diagnosis method based on spectral features, which extracts features from partial discharge multispectral signals to obtain multispectral partial discharge feature quantities, which include discharge pulse number, average pulse amplitude, maximum discharge pulse amplitude and average discharge pulse amplitude after windowing; these four feature quantities are used as input quantities of a partial discharge diagnosis model, and the partial discharge diagnosis model adopts a k-means clustering algorithm and a deep neural network algorithm; the deep neural network algorithm is used to identify the type of partial discharge, and the k-means clustering algorithm is used to divide the severity of partial discharge. When k-means is used as the clustering algorithm, its diagnostic effect is significantly better than the diagnostic effect based on the PRPD operator under the traditional full-band; and for the deep neural network algorithm, the diagnostic effects of the two methods are roughly the same, but compared with the traditional partial discharge diagnosis method, the partial discharge diagnosis method based on multispectral analysis fully mines the multispectral information contained in the discharge light radiation signal, and does not need to rely on the phase, and only extracts the multispectral discharge pulse number, discharge pulse amplitude and other information to achieve reliable diagnosis of partial discharge, and has good application prospects. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 is a flow chart of the construction process of the partial discharge diagnostic model of the present invention;

[0047] Figure 2a It is the average discharge pulse amplitude ratio after windowing in each band during the slight discharge stage during surface discharge;

[0048] Figure 2b It is the average discharge pulse amplitude ratio after windowing in each band during the moderate discharge stage during surface discharge;

[0049] Figure 2c It is the average discharge pulse amplitude ratio after windowing in each band during the intense discharge stage during surface discharge;

[0050] Figure 3a It is the average discharge pulse amplitude ratio after windowing in each band during the slight discharge stage during suspension discharge;

[0051] Figure 3b It is the average discharge pulse amplitude ratio after windowing in each band during the moderate discharge stage during suspension discharge;

[0052] Figure 3c It is the average discharge pulse amplitude ratio after windowing in each band during the intense discharge stage during suspension discharge;

[0053] Figure 4a It is the average discharge pulse amplitude ratio after windowing in each band during the slight discharge stage during tip discharge;

[0054] Figure 4bIt is the average discharge pulse amplitude ratio after windowing in each band during the moderate discharge stage during tip discharge;

[0055] Figure 4c It is the average discharge pulse amplitude ratio after windowing in each band during the intense discharge stage of tip discharge;

[0056] Figure 5 It is a flow chart of a switch cabinet discharge detection method based on spectral characteristics of the present invention;

[0057] Figure 6 The schematic diagram of the switch cabinet insulation detection system based on spectral characteristics of the present invention is shown in FIG. DETAILED DESCRIPTION

[0058] In order to make the purpose, technical solution and advantages of the present invention more clear, the following is further described in detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention, that is, the embodiments described are only part of the embodiments of the present invention, not all embodiments.

[0059] The detailed description of the embodiments of the present invention provided in the following drawings is not intended to limit the scope of the invention claimed for protection, but merely represents a selected embodiment of the present invention. Based on the drawings and embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.

[0060] Example 1

[0061] like Figure 5 As shown, the present invention discloses a switch cabinet discharge detection method based on spectral characteristics, comprising the following steps:

[0062] S1, collecting partial discharge multi-spectral signals collected by the partial discharge multi-spectral sensor at different positions in the switch cabinet;

[0063] S2, extracting features from the partial discharge multi-spectrum signal to obtain multi-spectrum partial discharge feature quantities;

[0064] The multi-spectral local discharge characteristic quantities include the number of discharge pulses, the average pulse amplitude, the maximum discharge pulse amplitude and the average discharge pulse amplitude after windowing;

[0065] S3, inputting the multi-spectral partial discharge feature quantity into the partial discharge diagnosis model to identify the partial discharge type and classify the partial discharge severity;

[0066] The partial discharge diagnosis model adopts a k-means clustering algorithm and a deep neural network algorithm;

[0067] The deep neural network algorithm is used to identify the type of partial discharge, and the k-means clustering algorithm is used to classify the severity of partial discharge.

[0068] Example 2

[0069] See also Figure 1 , based on the embodiment, the construction process of the partial discharge diagnosis model is described, specifically:

[0070] 1.1. Carry out a large number of partial discharge tests on typical discharge defect models, and study the discharge pulse characteristics in multiple bands through multi-spectral detection. The spectrum range includes full band, 350nm, 400nm, 450nm, 500nm, 550nm, 600nm and 650nm, and collect partial discharge multi-spectral signals;

[0071] The collected multi-spectral signals of partial discharge are synchronously measured and analyzed, the multi-spectral partial discharge characteristics are extracted, and a characteristic database for partial discharge diagnosis is established.

[0072] Typical discharge defect models generally include tip discharge model, surface discharge model and suspension discharge model.

[0073] The multi-spectral local discharge characteristic quantities include the number of discharge pulses, the average pulse amplitude, the maximum discharge pulse amplitude and the average discharge pulse amplitude after windowing.

[0074] Among them, the calculation formula for the number of discharge pulses and their proportion is:

[0075]

[0076] Where N i is the total number of discharge pulses collected for the i-th time under each voltage group; N is the average number of discharge pulses measured under each voltage group; n i % is the proportion of discharge pulse number in the required spectral band; N t is the number of discharge pulses in each spectral band; k is the number of spectral bands, and n is the total number of acquisitions.

[0077] The calculation formula for the average pulse amplitude and its proportion is:

[0078]

[0079]

[0080] Where, L m The discharge pulse amplitude obtained during the mth acquisition; L m,i is the i-th discharge pulse amplitude obtained during the m-th acquisition; L is the average discharge pulse amplitude obtained by multiple measurements under each voltage group; L iis the average discharge pulse amplitude in the target spectral band; n is the total number of acquisitions; L i % is the proportion of the average discharge pulse amplitude component in the target spectral band; k is the number of spectral bands.

[0081] The calculation formula for the maximum discharge pulse amplitude and its proportion is:

[0082] A i =max{L i,1 ,L i,2 ,…,L i,n}

[0083]

[0084] In the formula, A i is the maximum value of the discharge pulse amplitude collected for the i-th time under each voltage group; L m,n is the discharge pulse amplitude collected for the mth time under each voltage group; a i % is the proportion of the maximum discharge pulse amplitude component in the required spectral band; A i is the maximum number of discharge pulses; k is the number of spectral bands. t is the discharge pulse amplitude obtained by the tth acquisition.

[0085] The calculation formula of the average discharge pulse amplitude after window opening is:

[0086]

[0087] In the formula, is the phase angle; t is the time of discharge pulse acquisition; T is the power frequency period; L k is the discharge pulse amplitude in each window, k is the number of spectral bands. L is the average discharge pulse amplitude obtained by multiple measurements within the next cycle T of each voltage group. In the calculation process, the collected discharge pulses are first superimposed in the same power frequency cycle T in the order of collection time t, and then the average discharge pulse amplitude in each phase window interval is calculated.

[0088] The preliminary analysis of the eigenvalues ​​mainly leads to the following conclusions:

[0089] 1) If Figure 2a-2c As shown in the figure, for surface discharge, the number of discharge pulses and the amplitude of discharge pulses in the 450nm band account for the largest proportion, and with the increase of the external applied voltage, the proportion of each parameter gradually decreases. In the early stage of discharge development, the phase distribution corresponding to the discharge pulse is relatively symmetrical. As the discharge develops to the intense discharge stage, "abnormal pulses" begin to appear, and their discharge pulse amplitudes are much larger than other pulses or the corresponding phases are dispersed in the phase intervals corresponding to other pulses.

[0090] 2) If Figure 3a-3cAs shown in the figure, for suspended discharge, the amplitude ratio of all discharge pulses under each spectrum is relatively stable, and the discharge amplitude of the positive half cycle is slightly smaller than that of the negative half cycle. With the development of discharge, the corresponding phase distribution range of the positive and negative half cycle discharge pulses gradually increases, and the phase distribution of the positive and negative half cycles tends to be symmetrical, and the difference between the amplitudes of the positive and negative half cycle discharge pulses gradually increases.

[0091] 3) If Figure 4a-4c As shown in the figure, for tip discharge, the proportion of discharge pulses in the 450nm band is much greater than that in other bands, and its proportion only drops rapidly when the applied voltage approaches the breakdown voltage. As the discharge progresses, the proportion of discharge pulse amplitude in the 450nm band drops significantly, while the proportion of the 500nm band component increases, the phase distribution range corresponding to the discharge pulse increases, and the discharge pulses gradually become denser.

[0092] 4) The multi-spectral characteristics of partial discharge show that: ① In the same discharge stage, the proportion of the discharge pulse amplitude component in each spectral band is relatively stable, and will only change accordingly when the discharge severity changes, which provides a basis for the classification of discharge severity; ② For different discharge defects, the proportion of the discharge pulse number, maximum discharge pulse amplitude and average pulse amplitude in each band is different, and the trend of each parameter changing with voltage is also different. Therefore, these characteristic quantities can be used to classify the partial discharge type.

[0093] 1.2. The k-means clustering algorithm and deep neural network algorithm (DNN algorithm) are used to train and test the extracted multi-spectral partial discharge feature quantities. The preliminary analysis of the multi-spectral characteristic parameters of partial discharge shows that for different discharge defects, the number of discharge pulses, the maximum discharge pulse amplitude and the average discharge pulse amplitude characteristics in each band are different. Therefore, these feature quantities can be used to classify the partial discharge type. In addition, for the same discharge defect, the proportion of the discharge pulse amplitude components in each spectral band at the same voltage is relatively stable and hardly changes with the change of phase. Only when the discharge severity changes, the proportion of the pulse amplitude in each band will change accordingly. Therefore, the discharge severity can be divided based on this. The following conclusions are mainly drawn:

[0094] 1) For the local discharge type recognition part, in the recognition method based on the k-means clustering algorithm, the three characteristic quantities of discharge pulse number, average pulse amplitude and maximum discharge pulse amplitude are used to identify the discharge type and accuracy. The accuracy of the discharge severity judgment using the multi-spectral discharge pulse amplitude after windowing is above 90%. Among them, the clustering effect with the average pulse amplitude in the multi-spectrum as the characteristic quantity is the best, with an accuracy rate of up to 98.189%.

[0095] This is due to the characteristics of the k-means clustering algorithm. The k-means clustering algorithm compares similarities by calculating the distances between each data point. Therefore, when the number of feature types increases, the clustering effect can be effectively improved compared to the feature values ​​with relatively low recognition accuracy, but it cannot exceed the optimal recognition effect.

[0096] For the discharge type recognition of three types of multi-spectral local discharge feature quantities based on the DNN algorithm, the accuracy rate is 100%, and the learning curve, verification curve and loss curve converge fastest when the three types of multi-spectral local discharge feature quantities, namely the number of discharge pulses, average pulse amplitude and maximum discharge pulse amplitude, are used for DNN model training at the same time. This is because the DNN algorithm learns the internal features of the data set. The more types of feature quantities used, the more comprehensive the description of the target characteristics, and the better the training effect. Overall, for the recognition of local discharge types, the recognition effect of the DNN algorithm is better than that of the k-means clustering algorithm.

[0097] 2) For the discharge severity discrimination part, based on k-means clustering, the accuracy of severity classification under surface discharge and suspension discharge is above 95%, while the accuracy of discharge stage classification under tip discharge is relatively low. For the discharge severity classification based on the DNN algorithm, the accuracy of severity assessment of the three discharge defects is about 80%.

[0098] In general, the training accuracy based on the DNN algorithm is not as high as that based on the k-means algorithm, but the discharge severity assessment under the three discharge types can also be considered to have a good classification effect. Based on the k-means algorithm and the DNN algorithm, the discharge severity assessment accuracy under the tip discharge is lower than that of the surface discharge and the suspended discharge. This may be because the discharge characteristics change more slowly from the initial stage to the development stage of the discharge under the tip discharge, or the discharge stage divided according to the trend of the discharge amount changing with the voltage under the full band is not accurate enough, resulting in a relatively low accuracy of the severity assessment under this discharge type.

[0099] The partial discharge diagnosis is performed based on the traditional PRPD statistical calculator in the full band as the feature quantity. The diagnosis results are compared with those based on multi-spectral discharge feature quantities, and the following conclusions are drawn:

[0100] When the diagnosis algorithm is the k-means clustering algorithm, the accuracy of partial discharge diagnosis using multispectral features is significantly higher; when the diagnosis algorithm is the DNN algorithm, the diagnosis effect of the DNN algorithm and the PRPD statistical algorithm is equivalent. Under the same diagnostic effect, the diagnosis method based on multispectral analysis does not need to rely on the voltage phase, and can adapt to the discharge diagnosis of on-site phase-free information and DC equipment, which is more advantageous than traditional diagnostic methods.

[0101] When insulation detection is required for the switch cabinet to be tested, the partial discharge multi-spectral signals collected by the partial discharge multi-spectral sensor at the corresponding positions in the switch cabinet are first collected, and then the feature quantity in the signal is extracted according to the method of step S1, and the signal is input into the established partial discharge diagnosis model based on the k-means clustering algorithm or the DNN algorithm to identify the partial discharge type and classify the partial discharge severity.

[0102] Example 3

[0103] like Figure 6 As shown, the present invention also discloses a switch cabinet insulation detection system based on spectral characteristics, comprising:

[0104] A data acquisition module, used to collect the partial discharge multi-spectral signals collected by the partial discharge multi-spectral sensor at different positions in the switch cabinet;

[0105] A feature extraction module is used to extract features from the partial discharge multi-spectrum signal to obtain multi-spectrum partial discharge feature quantities;

[0106] The multi-spectral local discharge characteristic quantities include the number of discharge pulses, the average pulse amplitude, the maximum discharge pulse amplitude and the average discharge pulse amplitude after windowing;

[0107] A partial discharge diagnosis module is used to input multi-spectral partial discharge feature quantities into a partial discharge diagnosis model to identify partial discharge types and classify partial discharge severity;

[0108] The partial discharge diagnosis model adopts a k-means clustering algorithm and a deep neural network algorithm;

[0109] The deep neural network algorithm is used to identify the type of partial discharge, and the k-means clustering algorithm is used to classify the severity of partial discharge.

[0110] Example 4

[0111] The switch cabinet discharge detection method based on spectral characteristics of the present invention can be implemented in the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention can be implemented in the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program codes.

[0112] If the switch cabinet discharge detection method based on spectral characteristics of the present invention is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the present invention implements all or part of the processes in the above-mentioned embodiment method, and can also be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium. When the computer program is executed by the processor, the steps of the above-mentioned various method embodiments can be implemented. Among them, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form. Computer-readable storage media include permanent and non-permanent, removable and non-removable media, and information storage can be realized by any method or technology. Information can be computer-readable instructions, data structures, program modules or other data. It should be noted that the content contained in the computer-readable medium can be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable media do not include electrical carrier signals and telecommunication signals. Among them, the computer storage medium can be any available medium or data storage device that can be accessed by the computer, including but not limited to magnetic storage (such as floppy disks, hard disks, magnetic tapes, magneto-optical disks (MO), etc.), optical storage (such as CD, DVD, BD, HVD, etc.), and semiconductor storage (such as ROM, EPROM, EEPROM, non-volatile memory (NANDFLASH), solid-state drive (SSD)), etc.

[0113] Example 5

[0114] In an exemplary embodiment, a computer device is also provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, and the processor implements the steps of the switch cabinet discharge detection method based on spectral features when executing the computer program. The processor may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc.

[0115] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the relevant field should understand that the specific implementation methods of the present invention can still be modified or replaced by equivalents. Any modification or equivalent replacement that does not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A switch cabinet discharge detection method based on spectral characteristics, characterized in that: The following steps are involved: S1, collecting partial discharge multi-spectral signals collected by the partial discharge multi-spectral sensor at different positions in the switch cabinet; S2, extracting features from the partial discharge multi-spectrum signal to obtain multi-spectrum partial discharge feature quantities; The multi-spectral local discharge characteristic quantities include the number of discharge pulses, the average pulse amplitude, the maximum discharge pulse amplitude and the average discharge pulse amplitude after windowing; S3, inputting the multi-spectral partial discharge feature quantity into the partial discharge diagnosis model to identify the partial discharge type and classify the partial discharge severity; The partial discharge diagnosis model adopts a k-means clustering algorithm and a deep neural network algorithm; The deep neural network algorithm is used to identify the type of partial discharge, and the k-means clustering algorithm is used to classify the severity of partial discharge.

2. A switch cabinet discharge detection method based on spectral characteristics according to claim 1, characterized in that: In S3, the construction process of the partial discharge diagnosis model is as follows: A large number of partial discharge tests were conducted on typical discharge defect models. The discharge pulse characteristics in multiple bands were studied through multi-spectral pulse detection. The collected multi-spectral signals were synchronously measured and analyzed to extract multi-spectral partial discharge feature quantities, and a feature database for partial discharge diagnosis was established. S2. The extracted multi-spectral partial discharge feature quantities were trained using the k-means clustering algorithm and the deep neural network algorithm. For the identification of partial discharge types, the recognition effect of the deep neural network algorithm was higher than that of the k-means clustering algorithm. For the severity of discharge, the recognition effect of the k-means clustering algorithm was higher than that of the deep neural network algorithm. The partial discharge diagnosis model is constructed using the k-means clustering algorithm and deep neural network algorithm.

3. The switch cabinet discharge detection method based on spectral characteristics according to claim 2 is characterized in that: Typical discharge defect models include tip discharge model, surface discharge model and suspension discharge model.

4. The switch cabinet discharge detection method based on spectral characteristics according to claim 2 is characterized in that: The spectral range of spectral pulse detection includes full band, 350nm, 400nm, 450nm, 500nm, 550nm, 600nm and 650nm.

5. The switch cabinet discharge detection method based on spectral characteristics according to claim 1 is characterized in that: In S2, the calculation formula for the number of discharge pulses is: The calculation formula for the proportion of the discharge pulse number component is: Where N i is the total number of discharge pulses collected for the i-th time under each voltage group; N is the average number of discharge pulses measured under each voltage group; n i % is the proportion of discharge pulse number in the required spectral band; N t is the number of discharge pulses in each spectral band; k is the number of spectral bands, and n is the total number of acquisitions.

6. The switch cabinet discharge detection method based on spectral characteristics according to claim 1 is characterized in that: In S2, the calculation formula for the average pulse amplitude is: Where, L m The discharge pulse amplitude obtained during the mth acquisition; L m,i is the i-th discharge pulse amplitude obtained during the m-th acquisition; L is the average discharge pulse amplitude obtained by multiple measurements under each voltage group; L i is the average discharge pulse amplitude in the target spectral band; n is the total number of collections; L i % is the proportion of the average discharge pulse amplitude component in the target spectral band; k is the number of spectral bands.

7. The switch cabinet discharge detection method based on spectral characteristics according to claim 1 is characterized in that: In S2, the calculation formula for the maximum discharge pulse amplitude is: HAS i =max{L i,1 ,L i,2 ,…,L i,n } In the formula, A i is the maximum value of the discharge pulse amplitude collected for the i-th time under each voltage group; L m,n is the discharge pulse amplitude collected for the mth time under each voltage group; a i % is the proportion of the maximum discharge pulse amplitude component in the required spectral band; A i is the maximum number of discharge pulses; k is the number of spectral bands; A t is the discharge pulse amplitude obtained by the tth acquisition.

8. The switch cabinet discharge detection method based on spectral characteristics according to claim 1 is characterized in that: In S2, the calculation formula for the average discharge pulse amplitude after window opening is: In the formula, is the phase angle; t is the time of discharge pulse acquisition; T is the power frequency period; L k is the discharge pulse amplitude in each window, k is the number of spectral bands; L is the average discharge pulse amplitude obtained by multiple measurements within the next cycle T of each voltage group; In the calculation process, the collected discharge pulses are first superimposed in the same power frequency period T in the order of collection time t, and then the average discharge pulse amplitude in each phase window interval is calculated.

9. A switch cabinet insulation detection system based on spectral characteristics for implementing the switch cabinet discharge detection method according to any one of claims 1 to 8, characterized in that: include: A data acquisition module, used to collect the partial discharge multi-spectral signals collected by the partial discharge multi-spectral sensor at different positions in the switch cabinet; A feature extraction module is used to extract features from the partial discharge multi-spectrum signal to obtain multi-spectrum partial discharge feature quantities; The multi-spectral local discharge characteristic quantities include the number of discharge pulses, the average pulse amplitude, the maximum discharge pulse amplitude and the average discharge pulse amplitude after windowing; A partial discharge diagnosis module is used to input multi-spectral partial discharge feature quantities into a partial discharge diagnosis model to identify partial discharge types and classify partial discharge severity; The partial discharge diagnosis model adopts a k-means clustering algorithm and a deep neural network algorithm; The deep neural network algorithm is used to identify the type of partial discharge, and the k-means clustering algorithm is used to classify the severity of partial discharge.

10. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the steps of the switch cabinet discharge detection method based on spectral features as claimed in any one of claims 1 to 8 are implemented.

Images