Partial discharge signal diagnosis method and equipment based on variable frequency division and narrowband phase fusion analysis

Abstract

The present invention discloses a partial discharge signal diagnosis method and device based on variable frequency division and narrowband phase fusion analysis, which is used to diagnose partial discharge signals in power systems. The present invention receives a broadband electromagnetic wave signal on a broadband frequency band including ultra-high frequency, divides the broadband electromagnetic wave signal to obtain narrowband electromagnetic wave signals of multiple different frequency bands, down-converts each narrowband high-frequency electromagnetic wave signal after frequency division to obtain a low-frequency electromagnetic wave signal corresponding to each divided frequency band, and performs partial discharge power frequency phase analysis on each low-frequency electromagnetic wave signal. Based on the phase analysis results of each narrowband low-frequency electromagnetic wave signal, it is determined whether there is a noise signal in the low-frequency electromagnetic wave signal of each divided frequency band. Finally, by processing the narrowband low-frequency electromagnetic wave signal that is determined to be free of noise signals, the relevant situation of the partial discharge signal in the broadband electromagnetic wave signal is determined. The present invention has the advantages of low cost, small noise impact, high signal detection sensitivity, and strong reliability.

Description

Technical Field

[0001] The present invention belongs to the technical field of fault detection of power systems, and specifically relates to a diagnosis technology for partial discharge signals of power systems, in particular a method and corresponding equipment for diagnosing partial discharge signals based on variable frequency division and narrowband phase fusion analysis. Background Art

[0002] As a core method for assessing the insulation condition of power equipment, partial discharge detection technology plays a key role in preventing equipment failure and extending its service life. Partial discharge (PD) is an early sign of insulation degradation, and its cumulative effects can cause insulation breakdown, threatening the stable operation of power systems.

[0003] Existing partial discharge detection technologies include ultra-high frequency (UHF), ultrasonic (AE), and infrared thermography. UHF uses a frequency band of 300 MHz to 3 GHz to acquire electromagnetic wave signals, capable of detecting weak discharges as low as 10 pC. This method is suitable for online monitoring of enclosed equipment such as GIS and transformers. However, due to the electromagnetic shielding effect of metal structures, it lacks sensitivity to non-discharge defects (such as mechanical looseness). Furthermore, using traditional high-speed ADCs to acquire raw data presents challenges such as high ADC costs, high data storage requirements, high data transmission costs, and high data computation costs. The broadband UHF signal acquisition process contains a significant amount of noise, which seriously affects measurement accuracy.

[0004] Ultrasonic detection uses piezoelectric sensors to receive 20-200kHz acoustic signals, offering excellent resistance to electromagnetic interference and making it particularly suitable for physical location in open equipment such as switchgear. However, it is sensitive to the attenuation of acoustic propagation media (such as oil-immersion environments) and susceptible to interference from mechanical vibration and noise. Infrared thermal imaging, which identifies abnormal discharge areas through temperature field distribution, offers the advantage of non-contact detection, but is significantly affected by the equipment's heat dissipation conditions and lacks quantitative analysis capabilities.

[0005] Multi-sensor fusion has become a mainstream approach to improving detection reliability. For example, UHF-AE combined detection systems can simultaneously acquire electromagnetic and acoustic signals, increasing the accuracy of defect identification in GIS equipment. Intelligent data processing technologies (such as improved EMD noise reduction algorithms and BP neural network positioning models) have significantly enhanced signal analysis capabilities under complex working conditions. Summary of the Invention

[0006] (1) Technical issues to be resolved

[0007] Existing ultra-high frequency (UHF) partial discharge (PD) signal detection collects broadband signals and directly uses them for PD analysis. Broadband signals contain a large amount of noise signals in different frequency bands, which seriously affects the monitoring sensitivity and accuracy of PD signals.

[0008] The main technical problem solved by the present invention is the high cost, insufficient detection sensitivity and susceptibility to interference by environmental noise signals in existing ultra-high frequency electromagnetic signal partial discharge detection.

[0009] (2) Technical solution

[0010] To address the above technical issues, the first aspect of the present invention provides a method for diagnosing partial discharge signals in a power system. The method comprises the following steps: receiving a broadband electromagnetic wave signal over a broadband frequency band including an ultra-high frequency band; frequency-dividing the broadband electromagnetic wave signal to generate multiple narrowband high-frequency electromagnetic wave signals in different frequency bands; downconverting each of the divided narrowband high-frequency electromagnetic wave signals to generate multiple narrowband low-frequency electromagnetic wave signals; performing phase analysis on each narrowband low-frequency electromagnetic wave signal, and determining whether noise signals and partial discharge signals are present in each narrowband low-frequency electromagnetic wave signal based on the phase analysis results of each narrowband low-frequency electromagnetic wave signal; and processing each narrowband low-frequency electromagnetic wave signal determined to be free of noise signals to determine whether a partial discharge signal is present in the broadband electromagnetic wave signal.

[0011] According to a preferred embodiment of the present invention, the phase analysis refers to performing a phase correlation analysis on each narrowband low-frequency electromagnetic wave signal and the power frequency signal of the power system.

[0012] According to a preferred embodiment of the present invention, the step of determining whether noise signals and partial discharge signals exist in each narrowband low-frequency electromagnetic wave signal includes: analyzing the power frequency phase-radius spectrum characteristics of each narrowband low-frequency electromagnetic wave signal, and determining whether noise signals and partial discharge signals exist based on the distribution of each phase-radius spectrum characteristics in each narrowband low-frequency electromagnetic wave signal.

[0013] According to a preferred embodiment of the present invention, the step of determining whether there is a noise signal and a partial discharge signal based on the distribution of each phase-radius spectrum feature in each narrow-band low-frequency electromagnetic wave signal includes: analyzing the common phase-radius spectrum feature and the partial discharge power frequency phase-radius spectrum feature in each narrow-band low-frequency electromagnetic wave signal; when the amplitude of a narrow-band low-frequency electromagnetic wave signal is greater than a first threshold, but the narrow-band low-frequency electromagnetic wave signal does not have an obvious partial discharge power frequency phase-radius spectrum feature, it is determined that the narrow-band low-frequency electromagnetic wave signal contains a noise signal; when a narrow-band low-frequency electromagnetic wave signal When the amplitude is less than the second threshold, it is determined that the narrowband low-frequency electromagnetic wave signal has no noise signal and partial discharge signal, and the second threshold is less than the first threshold; when the amplitude of a narrowband low-frequency electromagnetic wave signal is greater than the first threshold, and the narrowband low-frequency electromagnetic wave signal has obvious partial discharge power frequency phase amplitude spectrum characteristics, it is determined that the narrowband low-frequency electromagnetic wave signal contains a partial discharge signal; when multiple narrowband low-frequency electromagnetic wave signals have obvious partial discharge power frequency phase amplitude spectrum characteristics, and the partial discharge power frequency phase amplitude spectrum characteristics are consistent, it is determined that the broadband electromagnetic wave signal has a partial discharge signal.

[0014] According to a preferred embodiment of the present invention, the step of determining whether there are noise signals and partial discharge signals based on the distribution of each phase-radius spectrum feature in each narrowband low-frequency electromagnetic wave signal includes: establishing a noise feature judgment model based on machine learning, and using a training data set to train the noise feature judgment model, the training data set including noise power frequency phase-radius spectrum feature data; inputting the power frequency phase-radius spectrum feature data in each narrowband low-frequency electromagnetic wave signal into the noise feature judgment model, calculating the probability of each narrowband low-frequency electromagnetic wave signal containing the noise power frequency phase-radius spectrum feature, and determining whether there is a noise signal in each narrowband low-frequency electromagnetic wave signal based on the probability.

[0015] According to a preferred embodiment of the present invention, the step of processing each narrowband low-frequency electromagnetic wave signal judged as not containing a noise signal to determine whether a partial discharge signal exists in the broadband electromagnetic wave signal includes: establishing a partial discharge feature judgment model based on machine learning, and using a training data set to train the partial discharge feature judgment model, the training data set including partial discharge power frequency phase amplitude spectrum feature data; inputting the power frequency phase amplitude spectrum feature data of each narrowband low-frequency electromagnetic wave signal judged as not containing a noise signal into the partial discharge feature judgment model, calculating the probability that each narrowband low-frequency electromagnetic wave signal contains a partial discharge power frequency phase amplitude spectrum feature, and determining whether there is a partial discharge signal in each narrowband low-frequency electromagnetic wave signal based on the probability.

[0016] According to a preferred embodiment of the present invention, the partial discharge power frequency phase amplitude spectrum characteristics include partial discharge type characteristics, and the partial discharge characteristic judgment model is also used to calculate the probability of the partial discharge power frequency phase amplitude spectrum characteristics of the partial discharge signals of each partial discharge type contained in each narrowband low-frequency electromagnetic wave signal.

[0017] The second aspect of the present invention proposes a partial discharge signal diagnostic device for diagnosing partial discharge signals in an electric power system, comprising: a signal acquisition unit for receiving a broadband electromagnetic wave signal on a broadband frequency band including an ultra-high frequency; a frequency division unit for dividing the broadband electromagnetic wave signal to obtain a plurality of narrowband high-frequency electromagnetic wave signals in different frequency bands; a frequency conversion unit for downconverting each narrowband high-frequency electromagnetic wave signal after frequency division to generate a plurality of narrowband low-frequency electromagnetic wave signals; a noise analysis unit for performing power frequency phase analysis on each narrowband low-frequency electromagnetic wave signal after frequency division and frequency reduction, and judging whether there are noise signals and partial discharge signals in each narrowband low-frequency electromagnetic wave signal based on the phase analysis results of each narrowband low-frequency electromagnetic wave signal; a partial discharge diagnostic unit for processing each narrowband low-frequency electromagnetic wave signal whose amplitude of the noise signal is less than a predetermined value based on the data analysis results of the noise analysis unit to judge whether there are partial discharge signals in the broadband electromagnetic wave signal.

[0018] According to a preferred embodiment of the present invention, the noise analysis unit is used to perform phase correlation analysis on the narrowband low-frequency electromagnetic wave signals after frequency division and frequency reduction and the power frequency signal of the power system.

[0019] According to a preferred embodiment of the present invention, the noise analysis unit is used to analyze the phase-radius spectrum characteristics of each narrowband low-frequency electromagnetic wave signal based on the phase of the industrial frequency signal, and determine whether there are noise signals and partial discharge signals according to the distribution of each phase-radius spectrum characteristics in each narrowband low-frequency electromagnetic wave signal.

[0020] According to a preferred embodiment of the present invention, the noise analysis unit is used for common phase-amplitude spectrum characteristics and partial discharge power frequency phase-amplitude spectrum characteristics in each narrowband low-frequency electromagnetic wave signal, and: when the amplitude of a narrowband low-frequency electromagnetic wave signal is greater than a first threshold, but the narrowband low-frequency electromagnetic wave signal does not have obvious partial discharge power frequency phase-amplitude spectrum characteristics, it is determined that the narrowband low-frequency electromagnetic wave signal contains a noise signal; when the amplitude of a narrowband low-frequency electromagnetic wave signal is less than a second threshold, it is determined that the narrowband low-frequency electromagnetic wave signal has no noise signal and partial discharge signal, and the second threshold is less than the first threshold; when the amplitude of a narrowband low-frequency electromagnetic wave signal is greater than the first threshold, and the narrowband low-frequency electromagnetic wave signal has obvious partial discharge power frequency phase-amplitude spectrum characteristics, it is determined that the narrowband low-frequency electromagnetic wave signal contains a partial discharge signal; when multiple narrowband low-frequency electromagnetic wave signals have obvious partial discharge power frequency phase-amplitude spectrum characteristics, and the partial discharge power frequency phase-amplitude spectrum characteristics are consistent, it is determined that the broadband electromagnetic wave signal has a partial discharge signal.

[0021] According to a preferred embodiment of the present invention, the noise analysis unit is used to: establish a noise feature judgment model based on machine learning, and use a training data set to train the noise feature judgment model, wherein the training data set includes noise power frequency phase amplitude spectrum feature data; input each phase amplitude spectrum feature data in each narrowband low-frequency electromagnetic wave signal into the noise feature judgment model, calculate the probability that each narrowband low-frequency electromagnetic wave signal contains the noise power frequency phase amplitude spectrum feature, and determine whether there is a noise signal in each narrowband low-frequency electromagnetic wave signal based on the probability.

[0022] According to a preferred embodiment of the present invention, the partial discharge diagnostic unit is used to: establish a partial discharge feature judgment model based on machine learning, and use a training data set to train the partial discharge feature judgment model, the training data set including partial discharge power frequency phase amplitude spectrum feature data; input the power frequency phase amplitude spectrum feature data of each narrow-band low-frequency electromagnetic wave signal judged to be free of noise signals into the partial discharge feature judgment model, calculate the probability that each narrow-band low-frequency electromagnetic wave signal contains the partial discharge power frequency phase amplitude spectrum feature, and determine whether there is a partial discharge signal in each narrow-band low-frequency electromagnetic wave signal based on the probability.

[0023] According to a preferred embodiment of the present invention, the partial discharge power frequency phase amplitude spectrum characteristics include partial discharge type characteristics, and the partial discharge characteristic judgment model is also used to calculate the probability of the partial discharge power frequency phase amplitude spectrum characteristics of the partial discharge signals of each partial discharge type contained in each narrowband low-frequency electromagnetic wave signal.

[0024] (3) Beneficial effects

[0025] The partial discharge signal diagnosis method and device based on variable frequency division and narrowband phase fusion analysis proposed in the present invention have the advantages of low cost, little influence by noise, high detection sensitivity and strong reliability. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 It is an overall flow chart of the partial discharge signal diagnosis method based on variable frequency division and narrowband phase fusion analysis of the present invention.

[0027] Figure 2 The figure is a schematic diagram of the steps of a partial discharge signal diagnosis process based on variable frequency division and narrowband phase fusion analysis according to an embodiment of the present invention.

[0028] Figures 3 to 12 This is a PRPD spectrum of another embodiment of the present invention for performing phase analysis on a low-frequency electromagnetic wave signal after frequency division and conversion.

[0029] Figure 13 It is a functional unit structure diagram of a corresponding diagnostic device embodiment based on the above-mentioned partial discharge signal diagnosis method of the present invention. DETAILED DESCRIPTION

[0030] A partial discharge signal is a periodic but non-stationary pulse signal characterized by high amplitude and sparse occurrence, while background noise is typically complex but evenly distributed. Because the partial discharge signal is broadband and has a strong phase correlation with the 50Hz power frequency signal, the phase-amplitude spectra of each narrowband frequency component of the partial discharge signal exhibit a strong correlation with the power frequency signal. However, the energy values ​​of different frequency bands vary, resulting in different energy values, but the phase distribution remains relatively uniform.

[0031] In the present invention, the performance of a signal in the phase-radius spectrum is referred to as a "phase-radius spectrum feature", and the common performance of different signals in the phase-radius spectrum is referred to as a "common phase-radius spectrum feature". Thus, the performance of a partial discharge signal in the phase-radius spectrum is referred to as a "partial discharge phase-radius spectrum feature", and the performance of a noise signal in the phase-radius spectrum is referred to as a "noise phase-radius spectrum feature". In particular, the "power frequency phase-radius spectrum feature" refers to the phase-radius spectrum feature of different signals based on the power frequency signal. Thus, in the present invention, the phase-radius spectrum features of the partial discharge signal and the noise signal based on the phase of the power frequency signal are referred to as "partial discharge power frequency phase-radius spectrum feature" and "noise power frequency phase-radius spectrum feature", respectively.

[0032] A PRPD (Phase Resolved Partial Discharge) spectrum is a typical phase-amplitude spectrum characteristic of a partial discharge signal. The following description of the embodiments of the present invention will take the PRPD as an example.

[0033] Based on this, the present invention proposes a method for diagnosing partial discharge signals, which effectively eliminates the influence of noise signals and enables rapid and accurate diagnosis of partial discharge signals in power systems. This method first uses a broadband electromagnetic wave sensor to receive broadband electromagnetic wave signals over a wide frequency band, and then performs subsequent operations. The broadband frequency band refers to a frequency band that includes ultra-high frequencies (UHF), typically between 300 MHz and 3 GHz.

[0034] Figure 1 It is an overall flow chart of the partial discharge signal diagnosis method based on variable frequency division and narrowband phase fusion analysis of the present invention.

[0035] like Figure 1As shown, considering that the use of traditional high-speed ADC to collect original broadband electromagnetic wave signals will face problems such as high ADC cost, high data storage requirements, high data transmission cost, and high data calculation cost, the present invention also proposes to convert the ultra-high frequency signal into a relatively low-frequency signal through step S2 "frequency division" and step S3 "frequency conversion" to facilitate data collection and analysis. For example, the frequency band of 1400MHz to 1500Mhz after frequency division is converted into 0Mhz to 100MHz through "frequency conversion". In other words, the present invention divides and reduces the frequency of the received broadband electromagnetic wave signal. For the convenience of description, the narrowband electromagnetic wave signals generated after frequency division and frequency reduction are collectively referred to as "narrowband low-frequency electromagnetic wave signals" in this article. The present invention does not limit the specific frequency band of the narrowband low-frequency electromagnetic wave signal, but it is preferably between 0MHz and 100MHz.

[0036] Furthermore, it is worth noting that the present invention takes into account the prominent role of PRPD spectrum characteristics based on the power frequency phase in determining partial discharge and noise signals. Therefore, frequency division is performed before the signal is reduced to a low-frequency electromagnetic wave signal, thereby generating multiple narrowband low-frequency electromagnetic wave signals in different frequency bands. Generally speaking, the number of frequency bands should be at least four, and preferably 5 to 16. The greater the number of narrowband frequency bands, the higher the computational complexity of subsequent processing, but also the higher the accuracy.

[0037] Unlike the prior art, the present invention proposes performing phase analysis on each narrowband low-frequency electromagnetic wave signal after frequency division and downconversion, and based on the phase analysis results of each narrowband low-frequency electromagnetic wave signal, determining the presence of a noise signal in each narrowband low-frequency electromagnetic wave signal (step S4). The phase analysis here preferably involves performing a phase correlation analysis between each narrowband low-frequency electromagnetic wave signal and the power frequency signal to generate a PRPD spectrum based on the power frequency phase. Furthermore, unlike the single-band PRPD spectrum of the prior art, the present invention analyzes the PRPD spectrum characteristics based on the power frequency phase of each narrowband low-frequency electromagnetic wave signal after frequency division and downconversion, and determines the presence of a noise signal based on the distribution of each PRPD spectrum characteristic based on the power frequency phase within each narrowband low-frequency electromagnetic wave signal.

[0038] As a preferred embodiment of the present invention, when determining whether a noise signal exists, the common PRPD spectrum characteristics based on the power frequency phase in each narrow-band low-frequency electromagnetic wave signal can be first analyzed, and when it is determined that the narrow-band low-frequency electromagnetic wave signal does not contain the common PRPD spectrum characteristics based on the power frequency phase, or contains PRPD spectrum characteristics of other signals, it is determined that a noise signal exists in the frequency band of the narrow-band low-frequency electromagnetic wave signal.

[0039] For example, when the amplitude of a narrowband low-frequency electromagnetic wave signal is greater than a first threshold, but the narrowband low-frequency electromagnetic wave signal does not have obvious PRPD spectrum characteristics, the narrowband low-frequency electromagnetic wave signal is determined to contain a noise signal. When the amplitude of a narrowband low-frequency electromagnetic wave signal is less than a second threshold (less than the first threshold), the narrowband low-frequency electromagnetic wave signal is determined to have no noise signal or partial discharge signal. When the amplitude of a narrowband low-frequency electromagnetic wave signal is greater than the first threshold, and the narrowband low-frequency electromagnetic wave signal has obvious PRPD spectrum characteristics based on the power frequency, the narrowband low-frequency electromagnetic wave signal is determined to contain a partial discharge signal. When multiple narrowband low-frequency electromagnetic wave signals have obvious PRPD spectrum characteristics based on the power frequency, and the PRPD spectrum characteristics based on the power frequency are consistent, the broadband electromagnetic wave signal is determined to have a partial discharge signal.

[0040] Another more preferred implementation method is to adopt a machine learning method, which requires establishing a noise feature judgment model and using a training data set to train the noise feature judgment model. Therefore, after the model training is completed, the PRPD spectrum feature data in each narrow-band low-frequency electromagnetic wave signal is input into the noise feature judgment model. The model can calculate the probability that each narrow-band low-frequency electromagnetic wave signal contains a PRPD spectrum feature related to the noise signal. The user can subsequently determine whether there is a noise signal based on the probability and certain regulations.

[0041] The training dataset typically contains a large amount of training data. The training dataset in the present invention should include PRPD spectrum feature data related to the noise signal. It should be noted that the present invention is not limited to PRPD spectrum data; any phase-amplitude spectrum that can reflect the phase of the power frequency signal can achieve similar results.

[0042] Through the above-mentioned means, the present invention can ultimately determine that narrowband low-frequency electromagnetic wave signals with relatively few noise signals are not noise signals. Each narrowband low-frequency electromagnetic wave signal determined to be noise-free is processed so that the collected partial discharge signal contains very little noise signal, thereby diagnosing the received broadband electromagnetic wave signal to determine whether a partial discharge signal is present in the original received broadband electromagnetic wave signal, or even to determine the type of partial discharge signal. When diagnosing partial discharge signals, the present invention also preferably uses machine learning methods to establish a partial discharge feature judgment model, calculates the probability of containing a partial discharge feature, or the probability of a specific type of partial discharge feature, in each narrowband low-frequency electromagnetic wave signal determined to be noise-free (in fact, relatively few noise signals may be present), and based on this, determines whether a partial discharge signal exists, or even the type of partial discharge signal.

[0043] Types of partial discharge include, for example, internal discharge, surface discharge, and suspended discharge. Internal discharge occurs inside solid insulation, such as small bubbles or impurities in transformer winding insulation or cable insulation. Surface discharge occurs on the surface of solid insulation, and local air or surface contamination between electrodes can trigger discharge. Suspended discharge refers to the introduction of irregular discharge when the conductor insulation layer is not grounded or has a suspended potential. Of course, partial discharge signals can also be divided according to other division rules, which all fall within the scope of protection of the present invention.

[0044] Similarly, the present invention requires the use of a training data set to train the PD feature judgment model. The training data set should include a large amount of characteristic data related to PD signals for each narrowband low-frequency electromagnetic wave signal. PD features include time-domain features, frequency-domain features, and time-frequency domain features. Time-domain features include signal amplitude, pulse width, repetition frequency (PD frequency), time interval statistical features, and frequency-domain features. Frequency-domain features include signal power spectrum density and energy distribution in specific frequency bands. Examples of time-frequency domain features include time-frequency analysis methods based on wavelet transforms, short-time Fourier transform (STFT) extraction of signal time-frequency distribution characteristics, and multi-band narrowband time-domain PD statistical features supported by broadband power spectrum features of broadband frequency-domain signals.

[0045] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with specific embodiments and with reference to the accompanying drawings.

[0046] Figure 2 FIG1 is a schematic diagram of the steps of the partial discharge signal diagnosis process based on variable frequency division and narrowband phase fusion analysis according to an embodiment of the present invention. Figure 2 As shown, in this embodiment, assuming that the broadband signal received by the broadband electromagnetic wave sensor is 300Mhz~2000Mhz, that is, an ultra-high frequency electromagnetic wave signal, which contains a local discharge signal and an environmental interference signal, the signal is subjected to frequency division and frequency conversion. During frequency division, the broadband signal of 300~2000MHz is divided into multiple narrowband high-frequency bands, such as 6 frequency bands of 300~500MHz, 500~800MHz, 800~1200MHz, 1200~1500MHz, 1500~1800MHz and 1800~2000MHz. After frequency division, the signal of each narrowband high-frequency band is converted into a low-frequency signal through "frequency conversion". In this embodiment, they are all converted into low-frequency signals of 0~100MHZ, which is convenient for subsequent processing. The narrowband low-frequency signal after frequency division and frequency conversion is Figure 2 They are called narrowband A, narrowband B, narrowband C, narrowband D, narrowband E and narrowband F respectively.

[0047] Then, a phase correlation analysis of each narrowband with the power frequency signal was performed, revealing that narrowband C and narrowband E have low correlations with the PD signal, indicating the presence of significant noise signals in narrowband C and narrowband E. On the other hand, since PD signals are assumed to also be present in the broadband signal, this step may also reveal that narrowband A, narrowband B, narrowband D, and narrowband F have strong correlations with the PD signal.

[0048] After obtaining the above analysis results, narrowband partial discharge monitoring and narrowband frequency selection are performed, that is, only the signals of narrowband A, narrowband B, narrowband D and narrowband F, which are judged to have less noise signals, are used for partial discharge signal diagnosis.

[0049] It should be noted that, in other embodiments, the range of the frequency bands, the number of frequency bands, and the interval between the frequency bands can all be adjusted according to the actual power system.

[0050] Figures 3 to 12 This is a PRPD spectrum of another embodiment of the present invention for performing phase analysis on a narrowband low-frequency electromagnetic wave signal after frequency division and conversion.

[0051] In this embodiment, Figure 2 The illustrated embodiment is similar, but the frequency division process divides the 300-2000 MHz broadband signal into eight high-frequency bands: 300-500 MHz, 500-700 MHz, 700-900 MHz, 900-1100 MHz, 1100-1300 MHz, 1300-1500 MHz, 1500-1800 MHz, and 1800-2000 MHz. This embodiment then performs phase analysis on each narrowband low-frequency electromagnetic wave signal, plotting a PRPD spectrum based on the power frequency phase for analysis.

[0052] For comparison, Figure 3 The figure shows the result of the prior art of collecting broadband electromagnetic wave signals from 300MHz to 2000MHz and then performing correlation analysis with the power frequency signal (50Hz). Figure 3 As shown in the figure, the acquired broadband signal contains both partial discharge (PD) and interference signals. Each dot in the figure represents the distribution of signal strength at a specific point in time along the power frequency signal phase axis. Red represents the PD signal, while yellow and green represent the noise signal. The noise signal, represented by the yellow dot, is comparable in strength to the PD signal and can significantly interfere with the distribution of the PD signal in the figure and subsequent diagnostic analysis. Consequently, existing conventional methods are prone to misjudgment.

[0053] Figures 4 to 11They are the PRPD spectra of the low-frequency electromagnetic wave signals after frequency downconversion of the original electromagnetic wave signals in the frequency bands of 300~500 MHz, 500~700 MHz, 700~900 MHz, 900~1100 MHz, 1100~1300 MHz, 1300~1500 MHz, 1500~1800 MHz and 1800~2000 MHz.

[0054] From the spectrum of multiple frequency divisions, we can see that since the partial discharge is a broadband signal, the partial discharge has a strong correlation with the power frequency signal (50Hz sinusoidal signal) in the spectrum after each frequency division. However, the energy in different frequency bands is different, so the corresponding energy amplitude will be different, but the phase distribution is relatively uniform. Figure 6 、 Figure 8 、 Figure 9 The PRPD data displayed for the 700-900MHz, 1100-1300MHz, and 1300-1500MHz frequency bands do not show consistent correlation with other frequency ranges. This indicates that there is a large noise signal in these frequency bands, so measurements in these frequency bands should be avoided.

[0055] Figure 12 This is the PRPD spectrum after removing the frequency band of noise signal interference. It can be clearly determined from the figure that there is a partial discharge signal.

[0056] Figure 13 FIG. 1 is a functional unit structure diagram of a corresponding diagnostic device embodiment based on the above-mentioned partial discharge signal diagnostic method of the present invention. Figure 13 As shown, the device includes a signal acquisition unit, a frequency division unit, a frequency conversion unit, a noise analysis unit and a partial discharge diagnosis unit.

[0057] The signal acquisition unit is, for example, a broadband electromagnetic wave sensor device, which is used to receive broadband electromagnetic wave signals on a broadband frequency band including ultra-high frequency. The unit can also perform pre-processing such as filtering and gain control on the acquired broadband electromagnetic wave signals.

[0058] like Figure 13 As shown, the frequency division unit and the frequency conversion unit can work independently or be combined into a frequency division and frequency conversion unit 1. As mentioned above, the frequency division unit divides the broadband electromagnetic wave signal to generate a plurality of narrowband high-frequency electromagnetic wave signals in different frequency bands, and the frequency conversion unit downconverts the narrowband high-frequency electromagnetic wave signals in each frequency band after frequency division to generate a plurality of narrowband low-frequency electromagnetic wave signals.

[0059] The noise analysis unit and partial discharge diagnosis unit can be implemented by devices with data processing capabilities, such as an MCU, CPU, DSP, or FPGA, or by a single device with digital signal processing capabilities, forming the main control unit 1. The noise analysis unit performs phase analysis on each narrowband low-frequency electromagnetic wave signal and, based on the phase analysis results of each narrowband low-frequency electromagnetic wave signal, determines whether a noise signal exists in each narrowband low-frequency electromagnetic wave signal. The partial discharge diagnosis unit fuses multiple low-frequency electromagnetic wave signals with relatively low noise signals to determine whether a partial discharge signal exists in the broadband electromagnetic wave signal. In addition, the device may also include a frequency selection unit to select narrowband low-frequency electromagnetic wave signals with relatively low noise signals for partial discharge diagnosis. The frequency selection unit can be controlled by the main control unit 2.

[0060] like Figure 13 As shown, as a more preferred embodiment, the diagnostic device of this embodiment further includes a signal generating unit. The signal generating unit is used to generate a simulated UHF high-frequency electromagnetic wave signal, and the simulated UHF high-frequency electromagnetic wave signal is used to perform active calibration on the diagnostic device.

[0061] The steps performed by each functional unit have been described in detail in the previous embodiment of the partial discharge signal diagnosis method and will not be repeated here. Furthermore, other components required to implement the diagnostic device, such as memory, power supply, switches, circuits and circuit boards, and housing, can be designed and installed by those skilled in the art based on practical needs and will not be described here.

[0062] The specific embodiments described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above are only specific embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A partial discharge signal diagnosis method based on variable frequency division and narrowband phase fusion analysis, used to diagnose partial discharge signals in power systems, characterized in that: The steps include: Receiving broadband electromagnetic wave signals over a broadband frequency band including ultra-high frequency; Frequency-dividing the broadband electromagnetic wave signal to generate a plurality of narrowband high-frequency electromagnetic wave signals in different frequency bands; Down-converting each narrow-band high-frequency electromagnetic wave signal after frequency division to generate multiple narrow-band low-frequency electromagnetic wave signals; Performing phase correlation analysis on each narrowband low-frequency electromagnetic wave signal and the power frequency signal of the power system, and analyzing the power frequency phase-radiation spectrum characteristics of each narrowband low-frequency electromagnetic wave signal based on the phase analysis results of each narrowband low-frequency electromagnetic wave signal, and judging whether there are noise signals and partial discharge signals in each narrowband low-frequency electromagnetic wave signal based on the distribution of each phase-radiation spectrum characteristic in each narrowband low-frequency electromagnetic wave signal; Establishing a partial discharge feature judgment model based on machine learning, and training the partial discharge feature judgment model using a training data set, wherein the training data set includes partial discharge power frequency phase amplitude spectrum feature data; The power frequency phase amplitude spectrum feature data of each narrowband low-frequency electromagnetic wave signal judged to be free of noise signals is input into the partial discharge feature judgment model, the probability of each narrowband low-frequency electromagnetic wave signal containing the power frequency phase amplitude spectrum feature of partial discharge is calculated, and based on the probability, it is determined whether there is a partial discharge signal in each narrowband low-frequency electromagnetic wave signal.

2. The method for diagnosing partial discharge signals based on variable frequency division and narrowband phase fusion analysis according to claim 1, characterized in that: The step of determining whether there are noise signals and partial discharge signals according to the distribution of each phase amplitude spectrum feature in each narrowband low-frequency electromagnetic wave signal includes: Analyze the common phase-radiation spectrum characteristics and partial discharge power frequency phase-radiation spectrum characteristics in each narrowband low-frequency electromagnetic wave signal; When the amplitude of a narrowband low-frequency electromagnetic wave signal is greater than a first threshold, but the narrowband low-frequency electromagnetic wave signal does not have obvious partial discharge power frequency phase amplitude spectrum characteristics, it is determined that the narrowband low-frequency electromagnetic wave signal contains a noise signal; when the amplitude of a narrowband low-frequency electromagnetic wave signal is less than a second threshold, it is determined that the narrowband low-frequency electromagnetic wave signal has no noise signal and partial discharge signal, and the second threshold is less than the first threshold; when the amplitude of a narrowband low-frequency electromagnetic wave signal is greater than the first threshold, and the narrowband low-frequency electromagnetic wave signal has obvious partial discharge power frequency phase amplitude spectrum characteristics, it is determined that the narrowband low-frequency electromagnetic wave signal contains a partial discharge signal; When multiple narrowband low-frequency electromagnetic wave signals have obvious partial discharge power frequency phase amplitude spectrum characteristics, and the partial discharge power frequency phase amplitude spectrum characteristics are consistent, it is determined that the broadband electromagnetic wave signal has a partial discharge signal.

3. The method for diagnosing partial discharge signals based on variable frequency division and narrowband phase fusion analysis according to claim 1, characterized in that: The step of determining whether there are noise signals and partial discharge signals according to the distribution of each phase amplitude spectrum feature in each narrowband low-frequency electromagnetic wave signal includes: Establishing a noise feature judgment model based on machine learning, and training the noise feature judgment model using a training data set, wherein the training data set includes noise power frequency phase amplitude spectrum feature data; The power frequency phase amplitude spectrum feature data in each narrowband low-frequency electromagnetic wave signal is input into the noise feature judgment model, and the probability of each narrowband low-frequency electromagnetic wave signal containing the noise power frequency phase amplitude spectrum feature is calculated respectively, and based on the probability, it is determined whether there is a noise signal in each narrowband low-frequency electromagnetic wave signal.

4. The method for diagnosing partial discharge signals based on variable frequency division and narrowband phase fusion analysis according to claim 1, characterized in that: The partial discharge power frequency phase amplitude spectrum characteristics include partial discharge type characteristics, and the partial discharge characteristic judgment model is also used to calculate the probability of the partial discharge power frequency phase amplitude spectrum characteristics of the partial discharge signals of each partial discharge type contained in each narrowband low-frequency electromagnetic wave signal.

5. A partial discharge signal diagnostic device based on variable frequency division and narrowband phase fusion analysis, used for diagnosing partial discharge signals in power systems, characterized in that: include: a signal acquisition unit for receiving broadband electromagnetic wave signals over a broadband frequency band including ultra-high frequency; A frequency division unit, which divides the broadband electromagnetic wave signal to obtain a plurality of narrowband high-frequency electromagnetic wave signals in different frequency bands; A frequency conversion unit, configured to reduce the frequency of each narrow-band high-frequency electromagnetic wave signal after frequency division to generate a plurality of narrow-band low-frequency electromagnetic wave signals; a noise analysis unit for performing phase correlation analysis on each narrowband low-frequency electromagnetic wave signal after frequency division and frequency reduction and the power frequency signal of the power system, and analyzing, based on the phase analysis results of each narrowband low-frequency electromagnetic wave signal, a phase-radius spectrum feature of each narrowband low-frequency electromagnetic wave signal based on the phase of the power frequency signal, and judging whether there is a noise signal and a partial discharge signal in each narrowband low-frequency electromagnetic wave signal based on the distribution of each phase-radius spectrum feature in each narrowband low-frequency electromagnetic wave signal; A partial discharge diagnosis unit is used to establish a partial discharge characteristic judgment model based on machine learning, and use a training data set to train the partial discharge characteristic judgment model, wherein the training data set includes partial discharge power frequency phase and amplitude spectrum characteristic data; the power frequency phase and amplitude spectrum characteristic data of each narrow-band low-frequency electromagnetic wave signal judged to be free of noise signals are input into the partial discharge characteristic judgment model, the probability of each narrow-band low-frequency electromagnetic wave signal containing the partial discharge power frequency phase and amplitude spectrum characteristic is calculated, and based on the probability, it is determined whether there is a partial discharge signal in each narrow-band low-frequency electromagnetic wave signal.

6. The partial discharge signal diagnostic device based on variable frequency division and narrowband phase fusion analysis according to claim 5, characterized in that: The noise analysis unit is used to analyze the common phase-radiation spectrum characteristics and the partial discharge power frequency phase-radiation spectrum characteristics in each narrowband low-frequency electromagnetic wave signal, and: When the amplitude of a narrowband low-frequency electromagnetic wave signal is greater than a first threshold, but the narrowband low-frequency electromagnetic wave signal does not have obvious partial discharge power frequency phase amplitude spectrum characteristics, it is determined that the narrowband low-frequency electromagnetic wave signal contains a noise signal; when the amplitude of a narrowband low-frequency electromagnetic wave signal is less than a second threshold, it is determined that the narrowband low-frequency electromagnetic wave signal has no noise signal and partial discharge signal, and the second threshold is less than the first threshold; when the amplitude of a narrowband low-frequency electromagnetic wave signal is greater than the first threshold, and the narrowband low-frequency electromagnetic wave signal has obvious partial discharge power frequency phase amplitude spectrum characteristics, it is determined that the narrowband low-frequency electromagnetic wave signal contains a partial discharge signal; When multiple narrowband low-frequency electromagnetic wave signals have obvious partial discharge power frequency phase amplitude spectrum characteristics, and the partial discharge power frequency phase amplitude spectrum characteristics are consistent, it is determined that the broadband electromagnetic wave signal has a partial discharge signal.

7. The partial discharge signal diagnostic device based on variable frequency division and narrowband phase fusion analysis according to claim 5, characterized in that: The noise analysis unit is used for: Establishing a noise feature judgment model based on machine learning, and training the noise feature judgment model using a training data set, wherein the training data set includes noise power frequency phase amplitude spectrum feature data; The phase-radius spectrum feature data in each narrowband low-frequency electromagnetic wave signal is input into the noise feature judgment model, the probability that each narrowband low-frequency electromagnetic wave signal contains the noise power frequency phase-radius spectrum feature is calculated, and based on the probability, it is determined whether there is a noise signal in each narrowband low-frequency electromagnetic wave signal.

8. The partial discharge signal diagnostic device based on variable frequency division and narrowband phase fusion analysis according to claim 5, characterized in that: The partial discharge power frequency phase amplitude spectrum characteristics include partial discharge type characteristics, and the partial discharge characteristic judgment model is also used to calculate the probability of the partial discharge power frequency phase amplitude spectrum characteristics of the partial discharge signals of each partial discharge type contained in each narrowband low-frequency electromagnetic wave signal.

Images