An acoustic diagnosis system and method for power grid equipment fault location

Abstract

The application belongs to the technical field of fault positioning, and discloses a sound wave diagnosis system and method for power grid equipment fault positioning; the method comprises the following steps: collecting sound wave signals of a cable; carrying out denoising processing on the collected sound wave signals; analyzing the sound wave signals after the denoising processing, and judging whether a fault instruction is generated; if the fault instruction is generated, the corresponding fault type is analyzed; according to the fault type, the propagation speed of the sound wave signals in the cable is evaluated; according to the propagation speed of the sound wave signals, the cable is positioned for fault; the application fully considers the propagation speed difference of the sound wave signals under the cable fault condition in the power distribution equipment, quickly and accurately judges the cable fault type and positions the fault position, improves the cable fault positioning precision, and thus improves the intelligent power grid operation and maintenance level and guarantees the safety and reliability of the power supply.

Description

Technical Field

[0001] This invention relates to the field of fault location technology, and more specifically, to an acoustic diagnostic system and method for fault location of power grid equipment. Background Technology

[0002] In recent years, with the development of power systems towards intelligence and informatization, the complexity of power grids and the requirements for power supply reliability have been continuously increasing. The prevention and rapid location of power equipment faults have become crucial to ensuring the safe and stable operation of the power grid. Cables, as important transmission and distribution equipment in the power system, including those used to connect distribution equipment and power supply lines, have always presented a challenge in power grid operation and maintenance in terms of fault diagnosis and location. Especially when connecting various distribution equipment, the connections are complex, and some cable lines are often located in dead zones. This means that when a fault needs to be investigated, the hardware structure of the distribution equipment has often been checked and found to be normal, yet the fault still occurs, at which point cable faults are considered. The existence of dead zones in cable lines makes fault diagnosis difficult. Common cable fault types include short circuits, open circuits, insulation breakdown, partial discharge, and other localized faults. These faults often occur at weak points such as cable joints and terminals, and are highly concealed.

[0003] Traditional cable fault location methods, such as time-domain reflectometry and frequency-domain analysis, are susceptible to environmental interference and have limited detection accuracy, making it difficult to quickly and accurately locate faults. This is especially true in large power grid systems where high-voltage switchgear and metering cabinets are interconnected, with different switchgear units configured according to power distribution needs and linked in series and parallel, as well as connected to metering cabinets, forming a power distribution equipment structure. The accuracy and efficiency of these methods are insufficient to meet the requirements of intelligent power grid operation and maintenance. Acoustic wave detection technology, as an emerging cable fault location method, offers advantages such as remote monitoring and non-destructive testing, providing effective support for cable fault diagnosis and rapid location.

[0004] Chinese Patent Publication No. CN110161377A discloses a method and device for locating cable fault points. The method includes: using a traveling wave monitoring terminal and an acoustic wave monitoring terminal to collect the arrival time of the initial traveling wave wavefront and the time difference between the arrival time of the initial acoustic wavefront after a fault occurs in the cable under test. The traveling wave monitoring terminal and the acoustic wave monitoring terminal are installed at the same location on the cable under test, and their clocks are synchronized. The fault point of the cable under test is located based on the time difference between the initial traveling wave wavefront arrival time and the initial acoustic wavefront arrival time. This invention solves the problem that the dual-end traveling wave positioning method relies on accurate GPS timing.

[0005] While the aforementioned technologies can locate cable faults and solve the problem of relying on precise GPS timing for the two-end traveling wave location method, in reality, different types of faults have different effects on the dielectric constant and mechanical properties of cables, and thus on the propagation speed of sound waves. The aforementioned technologies do not take into account the impact of different types of cable faults on the propagation characteristics of sound waves in cables, which can lead to large deviations in the calculation of time differences, i.e., large deviations in the location results, thereby reducing the accuracy of fault location and affecting the safety and reliability of smart grids.

[0006] In view of this, the present invention proposes an acoustic diagnostic system and method for fault location of power grid equipment to solve the above problems. Summary of the Invention

[0007] To overcome the aforementioned deficiencies of the prior art and achieve the above objectives, the present invention provides the following technical solution: an acoustic diagnostic method for fault location in power grid equipment, comprising:

[0008] Acquire acoustic signals from the cable;

[0009] The acquired acoustic signals are denoised.

[0010] Analyze the denoised acoustic signal to determine whether a fault command should be generated.

[0011] If a fault command is generated, the corresponding fault type is analyzed.

[0012] The propagation speed of acoustic signals in cables is evaluated based on the impact of different fault types on the dielectric constant and mechanical properties of the cables.

[0013] The cable fault is located based on the propagation speed of the sound wave signal.

[0014] Furthermore, the step of denoising the acquired acoustic signal includes:

[0015] Step 1: Set the wavelet basis function and the number of decomposition levels n;

[0016] Step 2: Perform i-level wavelet decomposition on the acoustic signal using wavelet basis functions, decomposing it into a set of approximate terms and i sets of detail terms. ;

[0017] Step 3: Calculate the signal-to-noise ratio for each detail item;

[0018] Step 4: Preset the signal-to-noise threshold, compare the signal-to-noise threshold with the signal-to-noise ratio of each group of approximate items and detail items, remove detail items with a signal-to-noise ratio less than the signal-to-noise threshold, and retain detail items with a signal-to-noise ratio greater than or equal to the signal-to-noise threshold;

[0019] Step 5: Based on the approximation terms and the i-th group of detail terms after thresholding, reconstruct the signal using the same wavelet basis functions as in Step 2;

[0020] Step 6: Calculate the signal-to-noise ratio of the reconstructed signal and compare it with the preset denoising threshold;

[0021] Step 7: If the signal-to-signal ratio of the reconstructed signal is less than the denoising threshold, then let And return to step 2, replacing the acoustic signal with the reconstructed signal; if the signal-to-noise ratio of the reconstructed signal is greater than or equal to the denoising threshold, or Then proceed to step 8;

[0022] Step 8: Output the reconstructed signal, which is the acoustic signal after the noise reduction process.

[0023] Furthermore, the acoustic signal is a mechanical wave caused by the minute vibration of the material inside the cable; a preset coupling interval is used to couple m acoustic sensors onto the cable surface according to the coupling interval;

[0024] The method for determining whether a fault command has been generated includes:

[0025] Extract corresponding feature parameters from each acoustic signal. These feature parameters include amplitude, frequency, and duration. Based on the current flow direction in the cable, assign different digital tags to each of the m acoustic sensors, and label them as sensor tags. The range of sensor tags is... The acoustic wave sensor that acquires the acoustic wave signal is marked as the acquisition sensor; each amplitude is sorted from largest to smallest, and the two acquisition sensors corresponding to the first two amplitudes are marked as the first acquisition sensors;

[0026] Determine if the sensor tags of two initial acquisition sensors are adjacent; if not adjacent, do not generate a faulty sensor; if adjacent, mark the initial acquisition sensor with the larger sensor tag as the first sensor and the initial acquisition sensor with the smaller sensor tag as the second sensor; mark the acquisition sensor with a sensor tag larger than the first sensor as the third sensor and the acquisition sensor with a sensor tag smaller than the second sensor as the fourth sensor; sort the amplitude values ​​corresponding to each third sensor from smallest to largest and generate a first sorting table; sort the amplitude values ​​corresponding to each fourth sensor from largest to smallest and generate a second sorting table.

[0027] If the order of the first sorting table is consistent with the order of the sensor tags corresponding to the third sensor, and the order of the second sorting table is consistent with the order of the sensor tags corresponding to the fourth sensor, then a fault command is generated.

[0028] Furthermore, methods for analyzing fault types include:

[0029] The characteristic parameters corresponding to each acoustic signal are used as analysis data. The analysis data is input into the trained type analysis model, and the corresponding type label is output. The type label is the numerical label corresponding to the fault type. The fault type of the cable is obtained according to the type label.

[0030] The training process for the type analysis model includes:

[0031] Different numerical labels are assigned to different fault types and marked as type labels; g sets of analysis data are collected in advance, and corresponding type labels are assigned to each of the g sets of analysis data, where g is an integer greater than 1;

[0032] The analysis data and corresponding type labels are converted into a set of feature vectors. Each set of feature vectors serves as the input to the type analysis model, which outputs a set of predicted type labels corresponding to each set of analysis data and uses the actual type labels corresponding to each set of analysis data as the prediction target. The actual type labels are the pre-collected type labels corresponding to the analysis data. The training objective is to minimize the sum of prediction errors of all analysis data. The type analysis model is trained until the sum of prediction errors converges, at which point training stops. The type analysis model is a deep neural network model.

[0033] Furthermore, the method for evaluating the propagation speed of acoustic signals in a cable includes:

[0034] The system presets a fault range and a normal propagation speed. The fault range includes a fault amplitude range and a fault frequency range. It subtracts the corresponding minimum value from the maximum value in the fault amplitude range to obtain the amplitude difference; subtracts the corresponding minimum value from the maximum value in the fault frequency range to obtain the frequency difference; compares the amplitude corresponding to the first sensor with the amplitude corresponding to the second sensor, marking the acoustic sensor with the larger amplitude as the near-end sensor and the acoustic sensor with the smaller amplitude as the far-end sensor; obtains the amplitude corresponding to the near-end sensor and marks it as the near-end amplitude; obtains the frequency corresponding to the near-end sensor and marks it as the near-end frequency; subtracts the maximum value in the fault amplitude range from the near-end amplitude to obtain the changed amplitude; subtracts the maximum value in the fault frequency range from the near-end frequency to obtain the changed frequency.

[0035] Divide the change amplitude by the amplitude difference to obtain the amplitude reduction coefficient; divide the change frequency by the frequency difference to obtain the frequency reduction coefficient; use the amplitude reduction coefficient, frequency reduction coefficient, and predicted type label as test data, input the test data into the trained coefficient prediction model, predict the corresponding speed reduction coefficient, and mark it as the single fault reduction coefficient; the training process of the coefficient prediction model is consistent with the training process of the type analysis model, and both are deep neural network models.

[0036] Multiply the single-fault reduction factor by the normal propagation speed to obtain the single-fault propagation speed.

[0037] Further, the times when the near-end sensor and the far-end sensor collect the acoustic wave signals are obtained respectively; the time when the near-end sensor collects the acoustic wave signal is marked as the first time, and the time when the far-end sensor collects the acoustic wave signal is marked as the second time; the second time is subtracted from the first time to obtain the time difference; the time difference is multiplied by the single fault propagation speed to obtain the distance difference; the coupling interval is subtracted from the distance difference and then divided by 2 to obtain the first distance; the first distance is the distance between the near-end sensor and the cable fault; the first distance is added to the distance difference to obtain the second distance, which is the distance between the far-end sensor and the cable fault; the cable fault is located based on the first distance and the second distance.

[0038] Furthermore, the method also includes: if the same acoustic sensor collects multiple acoustic signals, then all collected acoustic signals are classified to obtain P signal sets, each signal set corresponding to a cable fault;

[0039] Methods for obtaining a set of P signals include:

[0040] Obtain the acquisition time of each acoustic signal; sort each acquisition time from morning to night; obtain the sensor tag corresponding to the acoustic sensor for each acoustic signal; replace each acquisition time with the corresponding sensor tag to generate a third sorting table; analyze according to the ascending order of the third sorting table, add the first two sensor tags in the third sorting table to a signal set, and starting from the third sensor tag, classify each sensor tag in turn according to the classification rules, obtaining a total of P signal sets, where no two sensor tags are the same in each signal set, and P is an integer greater than 1.

[0041] Furthermore, the classification rules include:

[0042] Mark the sensor tag currently being classified as the current tag;

[0043] If the current tag is not adjacent to any sensor tag in all signal sets, then add the current tag to a new signal set;

[0044] If the current tag is adjacent to a sensor tag in a signal set that contains only one sensor tag, then the current tag is added to the corresponding signal set;

[0045] If the current tag is adjacent to sensor tags in multiple signal sets, then obtain the acquisition time corresponding to all sensor tags adjacent to the current tag, sort them, and add the current tag to the signal set corresponding to the sensor tag with the earliest acquisition time.

[0046] If the current tag is adjacent to a sensor tag in a signal set, then the current tag is added to the corresponding signal set.

[0047] Furthermore, based on the fault type of each cable fault, the propagation speed of the acoustic signal in the cable is analyzed;

[0048] Based on the third sorting table, the acoustic signal corresponding to the acquisition time of each sensor label in each signal set is obtained, and the corresponding feature parameters are extracted. Then, the type label corresponding to each signal set is obtained. Based on the feature parameters corresponding to the first and second sensor labels in each signal set, the amplitude reduction coefficient and frequency reduction coefficient corresponding to each signal set are calculated. The amplitude reduction coefficient, frequency reduction coefficient, and type label corresponding to each signal set are used as research data. The research data is input into the trained coefficient prediction model to predict the corresponding velocity reduction coefficient, which is marked as the multi-fault reduction coefficient. The multi-fault reduction coefficient is multiplied by the normal propagation speed to obtain the multi-fault propagation speed. The single-fault propagation speed is the propagation speed of the acoustic signal in the cable under multiple cable fault conditions.

[0049] An acoustic diagnostic system for locating faults in power grid equipment, comprising the following steps:

[0050] Acoustic wave acquisition module, used to acquire acoustic wave signals from cables;

[0051] The acoustic wave processing module is used to denoise the acquired acoustic wave signals.

[0052] The fault diagnosis module is used to analyze the noise-reduced acoustic signal and determine whether a fault command should be generated.

[0053] The type analysis module is used to analyze the corresponding fault type if a fault command is generated.

[0054] The speed evaluation module is used to evaluate the propagation speed of acoustic signals in the cable according to the fault type.

[0055] The fault location module is used to locate cable faults based on the propagation speed of sound wave signals.

[0056] The technical effects and advantages of the acoustic diagnostic system and method for fault location in power grid equipment of the present invention are as follows:

[0057] 1. By collecting acoustic signals from cables, wavelet decomposition is used to denoise the signals and extract characteristic parameters to determine whether a fault has occurred in the cable. Deep learning technology is used to analyze and identify the characteristic parameters of the acoustic signals in the faulty cable, thus identifying the cable fault type. Based on the impact of different fault types on the propagation speed of the acoustic signals in the cable, the propagation speed of the acoustic signals is accurately calculated, enabling rapid fault location. By fully considering the differences in the propagation speed of acoustic signals under different cable fault conditions, the method can quickly and accurately determine the cable fault type and locate the fault location, improving the accuracy of cable fault location, thereby enhancing the operation and maintenance level of the smart grid and ensuring the safety and reliability of power supply. 2. Based on the propagation characteristics of acoustic signals, combined with the acquisition time of acoustic signals and the positional relationship of acoustic sensors, each acoustic signal is reasonably classified to accurately acquire the acoustic signal caused by each cable fault. Furthermore, through deep learning technology, the propagation speed of acoustic signals under the combined influence of multiple cable faults is analyzed, providing an accurate basis for subsequent fault location. This improves the applicability of the location method in complex fault scenarios, achieves accurate fault location in cases of multiple cable faults, effectively solves the location difficulties in multiple fault situations, further improves fault location accuracy, and thus enhances the safety and reliability of the smart grid. Attached Figure Description

[0058] Figure 1 A schematic diagram showing the installation of an acoustic sensor in a capacitor bank of a power distribution system;

[0059] Figure 2 A schematic diagram showing the installation of acoustic sensors in the isolation cabinet of a power distribution equipment.

[0060] Figure 3 A schematic diagram showing the installation of acoustic sensors in the distribution cabinet of a power distribution system;

[0061] Figure 4 This is a schematic diagram of an acoustic diagnostic system for locating faults in power grid equipment according to Embodiment 1 of the present invention;

[0062] Figure 5 This is a schematic diagram of the acoustic sensor markings in Embodiment 1 of the present invention.

[0063] Figure 6 This is a schematic diagram of an acoustic diagnostic system for locating faults in power grid equipment according to Embodiment 2 of the present invention;

[0064] Figure 7 This is a schematic diagram of multiple cable faults in Embodiment 2 of the present invention;

[0065] Figure 8 This is a flowchart of an acoustic diagnostic method for locating faults in power grid equipment according to Embodiment 3 of the present invention. Reference numerals: 1, acoustic sensor. Detailed Implementation

[0066] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0067] Example 1:

[0068] Please see Figures 1 to 4 As shown in this embodiment, an acoustic diagnostic system for locating faults in power grid equipment is used to locate and diagnose cable line faults between power distribution equipment. It includes an acoustic acquisition module, an acoustic processing module, a fault judgment module, a type analysis module, a speed evaluation module, and a fault location module. The modules are connected by wired and / or wireless means to realize data transmission between the modules.

[0069] The acoustic wave acquisition module is used to collect acoustic wave signals from cables. The acoustic wave acquisition module is installed in the power distribution equipment, and can be found in [reference needed]. Figures 1 to 3 As shown, the location of the acoustic sensor 1 varies depending on the different power distribution equipment. At least two acoustic sensors can be installed at each corresponding wiring interface to cover the corresponding location.

[0070] The acoustic signal is a mechanical wave caused by the minute vibration of the material inside the cable; a preset coupling interval is used to couple m acoustic wave sensors to the cable surface according to the coupling interval, and the acoustic wave signal of the cable is collected by the acoustic wave sensors; the acoustic wave sensors are, for example, piezoelectric acoustic wave sensors, fiber optic acoustic wave sensors, etc.; the coupling interval is preset by those skilled in the art according to the actual length and structure of the cable.

[0071] It should be noted that under normal operating conditions, the acoustic signal energy generated by the cable is very small, usually between 20Hz and 500Hz, with a low amplitude. The cable's structure and materials have a certain effect on sound absorption and blocking, so the generated acoustic signal is difficult to be collected by the acoustic sensor. However, when the cable malfunctions, it will generate a significant high-frequency acoustic signal that can be collected by the acoustic sensor.

[0072] The acoustic wave processing module is used to denoise the acquired acoustic wave signals.

[0073] The steps for denoising the acquired acoustic signals include:

[0074] Step 1: Set the wavelet basis function and the number of decomposition levels n;

[0075] Wavelet basis functions include Daubechies wavelet, Symlet wavelet, Coiflet wavelet, etc.; the number of decomposition layers n is determined by those skilled in the art through multiple acquisitions of corresponding acoustic signals during historical cable faults, based on the average frequency of multiple acoustic signals.

[0076] Step 2: Perform i-level wavelet decomposition on the acoustic signal using wavelet basis functions, decomposing it into a set of approximate terms and i sets of detail terms. When performing wavelet decomposition on the acoustic signal for the first time, i is 1.

[0077] Step 3: Calculate the signal-to-noise ratio for each detail item;

[0078] The method for calculating the signal-to-noise ratio is existing technology and will not be elaborated on here. The approximation term mainly reflects the low-frequency trend component of the signal and basically does not contain noise. Therefore, only the signal-to-noise ratio needs to be calculated for the detail term in order to carry out subsequent denoising processing.

[0079] Step 4: Preset the signal-to-noise threshold, compare the signal-to-noise threshold with the signal-to-noise ratio of each group of approximate items and detail items, remove detail items with a signal-to-noise ratio less than the signal-to-noise threshold, and retain detail items with a signal-to-noise ratio greater than or equal to the signal-to-noise threshold;

[0080] The signal-to-noise threshold is determined by a person skilled in the art through multiple acquisitions of environmental noise signals during the normal operation of the historical cable, calculation of the corresponding signal-to-noise ratios, and the average of multiple signal-to-noise ratios as the signal-to-noise threshold.

[0081] Step 5: Based on the approximation terms and the i-th group of detail terms after thresholding, reconstruct the signal using the same wavelet basis functions as in Step 2.

[0082] Step 6: Calculate the signal-to-noise ratio of the reconstructed signal and compare it with the preset denoising threshold;

[0083] The noise reduction threshold is determined by a person skilled in the art by selecting a test site with low environmental noise, collecting acoustic signals from cable faults multiple times, calculating the corresponding signal-to-noise ratio, and using the average of multiple signal-to-noise ratios as the noise reduction threshold.

[0084] Step 7: If the signal-to-signal ratio of the reconstructed signal is less than the denoising threshold, then let And return to step 2, replacing the acoustic signal with the reconstructed signal; if the signal-to-noise ratio of the reconstructed signal is greater than or equal to the denoising threshold, or Then proceed to step 8.

[0085] Step 8: Output the reconstructed signal, which is the acoustic signal after the noise reduction process.

[0086] It should be noted that the purpose of denoising the acquired acoustic signals is to remove environmental noise from the acoustic signals, thereby improving the signal-to-noise ratio of the acoustic signals, highlighting the essential characteristics of the acoustic signals, and thus improving the quality of the acoustic signals and the accuracy and reliability of subsequent cable fault diagnosis.

[0087] The fault diagnosis module is used to analyze the denoised acoustic signal and determine whether a fault command should be generated.

[0088] Methods for determining whether a fault command has been generated include:

[0089] The corresponding feature parameters are extracted from each sound wave signal. The feature parameters include amplitude, frequency and duration. Amplitude represents the strength of the sound wave signal, i.e., sound pressure level, which is used to measure the power of the sound source. Frequency represents the spectral components of the sound wave signal. Duration is the length of time from the start to the end of the sound wave signal, reflecting the time domain characteristics of the sound wave. Extracting the corresponding feature parameters from the sound wave signal is an existing technology and will not be elaborated on here.

[0090] Based on the current flow direction in the cable, different digital tags are sequentially assigned to m acoustic wave sensors, and these tags are designated as sensor tags. The range of sensor tags is... The acoustic wave sensor that acquires the acoustic wave signal is marked as the acquisition sensor; each amplitude is sorted from largest to smallest, and the two acquisition sensors corresponding to the first two amplitudes are marked as the first acquisition sensors;

[0091] Determine if the sensor tags of two initial sampling sensors are adjacent; if they are not adjacent, do not generate a faulty sensor; if they are adjacent, proceed as follows. Figure 5 As shown, the sensor with the larger sensor label is marked as the first sensor, and the sensor with the smaller sensor label is marked as the second sensor; the sensor with a larger sensor label than the first sensor is marked as the third sensor, and the sensor with a smaller sensor label than the second sensor is marked as the fourth sensor; the amplitude values ​​corresponding to each third sensor are sorted from smallest to largest to generate a first sorting table; the amplitude values ​​corresponding to each fourth sensor are sorted from largest to smallest to generate a second sorting table.

[0092] If the order of the first sorting table is consistent with the order of the sensor tags corresponding to the third sensor, and the order of the second sorting table is consistent with the order of the sensor tags corresponding to the fourth sensor, then a fault command is generated.

[0093] For example, acoustic wave sensors labeled 4, 5, 6, 7, 8, and 9 collected acoustic wave signals. Among them, the acoustic wave sensors labeled 6 and 7 had the largest amplitudes. Therefore, the acoustic wave sensor labeled 7 is the first sensor, the acoustic wave sensor labeled 6 is the second sensor, the acoustic wave sensors labeled 8 and 9 are the third sensors, and the acoustic wave sensors labeled 4 and 5 are the fourth sensors. At this time, the amplitude of the acoustic wave sensor labeled 4 is less than that of the acoustic wave sensor labeled 5, and the amplitude of the acoustic wave sensor labeled 9 is greater than that of the acoustic wave sensor labeled 8. Therefore, the order in the first sorting table is 8 and 9, and the order in the second sorting table is 4 and 5. Since the order of the sensor labels corresponding to the third sensor is also 8 and 9, and the order of the sensor labels corresponding to the fourth sensor is also 4 and 5, a fault command is generated.

[0094] It's important to note that the reason for using sensor tags to determine if a fault is caused is that when a cable fault occurs, it generates a significant high-frequency sound wave signal that can be collected by sound wave sensors. In addition, due to complex geological conditions, cables also generate sound wave signals during terrain changes, such as earthquakes, road construction, and landslides. The difference lies in the fact that when the sound wave signal is generated by a cable fault, since the fault occurs at only one point or multiple points at different times, the sound wave signal propagates from the fault point to both sides, gradually weakening with the propagation distance. The two sensors with the largest amplitude are the closest sensors on either side of the fault point, and the amplitudes of other sensors decrease as the distance between the corresponding sensors and the fault point increases. Furthermore, each sensor receives the sound wave signal at a different time. However, when the sound wave signal is generated by terrain changes, the amplitude corresponding to each sensor is irregular, and each sensor receives the sound wave signal at approximately the same time. Therefore, by analyzing the amplitude variation pattern of the sound wave signal through sensor tags, it is possible to determine whether the sound wave signal is caused by a cable fault.

[0095] The type analysis module is used to analyze the corresponding fault type if a fault command is generated.

[0096] Methods for analyzing fault types include:

[0097] The characteristic parameters corresponding to each acoustic signal are used as analysis data. The analysis data is input into the trained type analysis model, and the corresponding type label is output. The type label is a numerical label corresponding to the fault type. The fault type of the cable is obtained according to the type label. Fault types include short circuit, open circuit, insulation breakdown, partial discharge, etc.

[0098] The specific training process of the type analysis model includes:

[0099] Different numerical labels are assigned to different fault types and marked as type labels; g sets of analysis data are collected in advance, and corresponding type labels are assigned to each of the g sets of analysis data, where g is an integer greater than 1; the type labels corresponding to the analysis data are assigned by those skilled in the art during the historical cable fault diagnosis process, by collecting g sets of analysis data, diagnosing the corresponding fault types in sequence under the conditions of each set of analysis data, and assigning type labels to the corresponding diagnosed fault types for each set of analysis data; the corresponding type labels are assigned to the g sets of analysis data in sequence.

[0100] The analysis data and corresponding type labels are converted into a set of feature vectors. Each set of feature vectors serves as input to the type analysis model, which outputs a set of predicted type labels corresponding to each set of analysis data and uses the actual type labels corresponding to each set of analysis data as the prediction target. The actual type labels are the pre-collected type labels corresponding to the analysis data. The training objective is to minimize the sum of prediction errors for all analysis data. The formula for calculating the prediction error is as follows: ,in The prediction error is represented by K, where K is the group number of the feature vector corresponding to the analyzed data. The predicted type label for the Kth group of analyzed data. Let K be the actual type label corresponding to the Kth group of analyzed data; train the type analysis model until the sum of prediction errors converges and then stop training.

[0101] The aforementioned analysis model is specifically a deep neural network model; it includes an input layer, hidden layers, and an output layer; each hidden layer contains multiple neurons, and each neuron is connected to the neurons in the next layer. The connections contain weights that determine the importance and influence of data transmission in the neural network; each neuron between the hidden layer and the output layer applies an activation function, which introduces non-linearity, allowing the network to learn more complex patterns and features.

[0102] It should be noted that the reason for determining the type of cable fault based on characteristic parameters is that different types of cable faults correspond to different characteristic parameters. When the cable fault is a short circuit, the cable will generate a large current, triggering a strong arc discharge. The resulting acoustic signal has a high amplitude, a short duration, and a high frequency component in the spectrum, exhibiting broadband characteristics. When the cable fault is an open circuit, the conductor inside the cable breaks, generating a brief arc discharge. The resulting acoustic signal has a smaller amplitude and a shorter duration than that of a short circuit fault, and the spectrum of the acoustic signal is concentrated in the low-frequency band, exhibiting narrow-frequency characteristics. When the cable fault is an insulation breakdown fault, partial discharge will occur inside the cable. The acoustic signal has a longer duration, a spectrum concentrated in the mid-to-low frequency band, and a low amplitude. When the cable fault is a partial discharge fault, a weak partial discharge acoustic signal will be generated, with a short duration, a low amplitude, and a spectrum concentrated in the mid-to-high frequency band.

[0103] The speed assessment module is used to assess the propagation speed of acoustic signals in the cable based on the fault type.

[0104] Methods for assessing the propagation speed of sound signals in cables include:

[0105] The fault range and normal propagation speed are preset. The fault range includes the fault amplitude range and the fault frequency range. The fault range is obtained by those skilled in the art through multiple acquisitions of acoustic signals emitted by the cable during historical cable faults, and the extraction of corresponding characteristic parameters. The maximum and minimum values ​​among multiple amplitudes are taken as the fault amplitude range, and the maximum and minimum values ​​among multiple frequencies are taken as the fault frequency range. The normal propagation speed is obtained by those skilled in the art through the pulse reflection method during the normal operation of the cable in history.

[0106] The amplitude difference is obtained by subtracting the corresponding minimum value from the maximum value in the fault amplitude range; the frequency difference is obtained by subtracting the corresponding minimum value from the maximum value in the fault frequency range; the amplitude corresponding to the first sensor is compared with the amplitude corresponding to the second sensor, and the acoustic sensor with the larger amplitude is marked as the near-end sensor, and the acoustic sensor with the smaller amplitude is marked as the far-end sensor; the amplitude corresponding to the near-end sensor is obtained and marked as the near-end amplitude; the frequency corresponding to the near-end sensor is obtained and marked as the near-end frequency; the amplitude variation is obtained by subtracting the maximum value in the fault amplitude range from the near-end amplitude; the frequency variation is obtained by subtracting the maximum value in the fault frequency range from the near-end frequency.

[0107] Divide the change amplitude by the amplitude difference to obtain the amplitude reduction coefficient; divide the change frequency by the frequency difference to obtain the frequency reduction coefficient; use the amplitude reduction coefficient, frequency reduction coefficient, and predicted type label as test data, input the test data into the trained coefficient prediction model, predict the corresponding speed reduction coefficient, and mark it as the single fault reduction coefficient; the specific training process of the coefficient prediction model is consistent with the training process of the type analysis model, and both are deep neural network models.

[0108] Multiply the single-fault reduction factor by the normal propagation speed to obtain the single-fault propagation speed; the single-fault propagation speed is the propagation speed of the sound wave signal in the cable under a single cable fault condition.

[0109] It should be noted that different types of faults affect the dielectric constant and mechanical properties of cables, thereby affecting the propagation speed of acoustic signals within the cable. For example, a short-circuit fault causes the insulation in the cable to break down, leading to a decrease in the dielectric constant, while the conductor in the cable will melt and deform. An open-circuit fault causes the insulation in the cable to lose its charge, leading to an increase in the dielectric constant, while the insulation layer in the cable may also crack or detach. Changes in the dielectric constant affect the electric field distribution inside the cable, thus altering the propagation speed of acoustic signals within the cable. Changes in mechanical properties alter the stiffness and density of the cable, thereby affecting the propagation speed of acoustic signals within the cable.

[0110] The fault location module is used to locate cable faults based on the propagation speed of sound wave signals.

[0111] The times when the near-end sensor and the far-end sensor collect the acoustic wave signals are obtained respectively; the time when the near-end sensor collects the acoustic wave signal is marked as the first time, and the time when the far-end sensor collects the acoustic wave signal is marked as the second time; the first time and the second time are obtained from the timestamps recorded when the corresponding acoustic wave sensors collect the acoustic wave signals; the second time is subtracted from the first time to obtain the time difference; the time difference is multiplied by the single fault propagation speed to obtain the distance difference; the coupling interval is subtracted from the distance difference and then divided by 2 to obtain the first distance; the first distance is the distance between the near-end sensor and the cable fault; the first distance is added to the distance difference to obtain the second distance, which is the distance between the far-end sensor and the cable fault; the cable fault is located based on the first distance and the second distance.

[0112] This embodiment collects acoustic signals from cables, performs denoising processing using wavelet decomposition, and extracts characteristic parameters of the acoustic signals to determine whether a fault has occurred in the cable. Deep learning technology is used to analyze and identify the characteristic parameters of the acoustic signals in the faulty cable, identifying the cable fault type. Furthermore, based on the impact of different fault types on the propagation speed of the acoustic signals in the cable, the propagation speed of the acoustic signals is accurately calculated, enabling rapid fault location. By fully considering the differences in the propagation speed of acoustic signals under different cable fault conditions, the embodiment quickly and accurately determines the cable fault type and locates the fault position, improving the accuracy of cable fault location and thus enhancing the operation and maintenance level of the smart grid and ensuring the safety and reliability of power supply.

[0113] Example 2:

[0114] Please see Figure 6 As shown, this embodiment further improves upon the design of Embodiment 1. The method in Embodiment 1 is for fault location when a fault occurs in a cable. When multiple faults occur in a cable in a short period of time, the acoustic signal will propagate in the cable, resulting in the same acoustic sensor collecting multiple acoustic signals in a short time. In this case, the method in Embodiment 1 cannot effectively distinguish which fault caused the acoustic signal collected by the acoustic sensor, thus making it impossible to accurately locate the fault. Therefore, this embodiment provides an acoustic diagnostic system for fault location of power grid equipment, which also includes a signal classification module and a velocity analysis module.

[0115] The signal classification module is used to classify all the collected sound wave signals if the same sound wave sensor collects multiple sound wave signals, and obtain P signal sets, each signal set corresponding to a cable fault.

[0116] Methods for obtaining a set of P signals include:

[0117] Obtain the acquisition time of each acoustic signal; sort each acquisition time from morning to night; obtain the sensor tag corresponding to the acoustic sensor for each acoustic signal; replace each acquisition time with the corresponding sensor tag to generate a third sorting table; analyze according to the ascending order of the third sorting table, add the first two sensor tags in the third sorting table to a signal set, and starting from the third sensor tag, classify each sensor tag in turn according to the classification rules, obtaining a total of P signal sets, where no two sensor tags are the same in each signal set, and P is an integer greater than 1.

[0118] The classification rules include:

[0119] Mark the sensor tag currently being classified as the current tag;

[0120] If the current tag is not adjacent to any sensor tag in all signal sets, then add the current tag to a new signal set;

[0121] If the current tag is adjacent to a sensor tag in a signal set that contains only one sensor tag, then the current tag is added to the corresponding signal set;

[0122] If the current tag is adjacent to sensor tags in multiple signal sets, then obtain the acquisition time corresponding to all sensor tags adjacent to the current tag, sort them, and add the current tag to the signal set corresponding to the sensor tag with the earliest acquisition time.

[0123] If the current tag is adjacent to a sensor tag in a signal set, then the current tag is added to the corresponding signal set.

[0124] For example, there are acoustic sensors with sensor labels 1-8 on the cable. Cable faults occur between sensor labels 2 and 3, and between sensor labels 5 and 6. The cable fault between sensor labels 2 and 3 is marked as the first fault, and the cable fault between sensor labels 5 and 6 is marked as the second fault. The occurrence time of the first fault is earlier than the occurrence time of the second fault. The order of the third sorting table is 5, 6, 3, 2, 4, 7, 4, 1, 3, 8. 5 and 6 are added to a signal set and marked as the first set. Since 3 is not adjacent to either 5 or 6 in the first set, 3 is added to a new signal set and marked as the second set. Since 2 is adjacent to 3 in the second set, and there is only one 3 in the second set, 2 is added to the second set. Since 4 is adjacent to both 5 in the first set and 3 in the second set, but 5 was collected earlier than 3, 4 is added to the first set. Since 7 is adjacent to 6 in the first set, 7 is added to the first set. Since 4 is adjacent to both 5 in the first set and 3 in the second set, but 4 is already in the first set, 4 is added to the second set. Since 1 is adjacent to 2 in the second set, 1 is added to the second set. Since 3 is already in the second set, and 3 is adjacent to 4 in the first set, 3 is added to the first set. Since 8 is adjacent to 7 in the first set, 8 is added to the first set. The first set includes 5, 6, 4, 7, 3, and 8, and the second set includes 3, 2, 4, and 1, all conforming to the propagation laws of sound waves. A detailed diagram is shown below. Figure 7 As shown in the figure, the dashed line represents the propagation range of the acoustic signal generated by the first fault, and the thick solid line represents the propagation range of the acoustic signal generated by the second fault.

[0125] The velocity analysis module is used to analyze the propagation speed of acoustic signals in the cable based on the fault type of each cable fault.

[0126] Based on the third sorting table, the acoustic signal corresponding to the acquisition time of each sensor label in each signal set is obtained, and the corresponding feature parameters are extracted. Then, the type label corresponding to each signal set is obtained. Based on the feature parameters corresponding to the first and second sensor labels in each signal set, the amplitude reduction coefficient and frequency reduction coefficient corresponding to each signal set are calculated. The amplitude reduction coefficient, frequency reduction coefficient, and type label corresponding to each signal set are used as research data. The research data is input into the trained coefficient prediction model to predict the corresponding velocity reduction coefficient, which is marked as the multi-fault reduction coefficient. The multi-fault reduction coefficient is multiplied by the normal propagation speed to obtain the multi-fault propagation speed. The single-fault propagation speed is the propagation speed of the acoustic signal in the cable under multiple cable fault conditions.

[0127] It should be noted that when multiple identical faults occur in a cable, the subsequent fault location method is the same as when a single fault occurs in the cable, that is, the P signal sets are treated as P cables with single faults for fault location.

[0128] This embodiment, based on the propagation characteristics of acoustic signals and the relationship between the acquisition time and the position of the acoustic sensors, rationally classifies each acoustic signal to accurately acquire the acoustic signal caused by each cable fault. Furthermore, through deep learning technology, it analyzes the propagation speed of acoustic signals under the combined influence of multiple cable faults, providing accurate data for subsequent fault location. This improves the applicability of the location method in complex fault scenarios, achieves accurate fault location in cases of multiple cable faults, effectively solves the location difficulties in multi-fault situations, further improves fault location accuracy, and thus enhances the safety and reliability of the smart grid.

[0129] Example 3:

[0130] Please see Figure 8 As shown, parts not described in detail in this embodiment are described in Embodiments 1 and 2. An acoustic diagnostic method for locating faults in power grid equipment is provided, the method comprising:

[0131] Acquire acoustic signals from the cable;

[0132] The acquired acoustic signals are denoised.

[0133] Analyze the denoised acoustic signal to determine whether a fault command should be generated.

[0134] If a fault command is generated, the corresponding fault type is analyzed.

[0135] Based on the type of fault, assess the propagation speed of the acoustic signal in the cable;

[0136] The cable fault is located based on the propagation speed of the sound wave signal.

[0137] Furthermore, the step of denoising the acquired acoustic signal includes:

[0138] Step 1: Set the wavelet basis function and the number of decomposition levels n;

[0139] Step 2: Perform i-level wavelet decomposition on the acoustic signal using wavelet basis functions, decomposing it into a set of approximate terms and i sets of detail terms. ;

[0140] Step 3: Calculate the signal-to-noise ratio for each detail item;

[0141] Step 4: Preset the signal-to-noise threshold, compare the signal-to-noise threshold with the signal-to-noise ratio of each group of approximate items and detail items, remove detail items with a signal-to-noise ratio less than the signal-to-noise threshold, and retain detail items with a signal-to-noise ratio greater than or equal to the signal-to-noise threshold;

[0142] Step 5: Based on the approximation terms and the i-th group of detail terms after thresholding, reconstruct the signal using the same wavelet basis functions as in Step 2;

[0143] Step 6: Calculate the signal-to-noise ratio of the reconstructed signal and compare it with the preset denoising threshold;

[0144] Step 7: If the signal-to-signal ratio of the reconstructed signal is less than the denoising threshold, then let And return to step 2, replacing the acoustic signal with the reconstructed signal; if the signal-to-noise ratio of the reconstructed signal is greater than or equal to the denoising threshold, or Then proceed to step 8;

[0145] Step 8: Output the reconstructed signal, which is the acoustic signal after the noise reduction process.

[0146] Furthermore, the acoustic signal is a mechanical wave caused by the minute vibration of the material inside the cable; a preset coupling interval is used to couple m acoustic sensors onto the cable surface according to the coupling interval;

[0147] The method for determining whether a fault command has been generated includes:

[0148] Extract corresponding feature parameters from each acoustic signal. These feature parameters include amplitude, frequency, and duration. Based on the current flow direction in the cable, assign different digital tags to each of the m acoustic sensors, and label them as sensor tags. The range of sensor tags is... The acoustic wave sensor that acquires the acoustic wave signal is marked as the acquisition sensor; each amplitude is sorted from largest to smallest, and the two acquisition sensors corresponding to the first two amplitudes are marked as the first acquisition sensors;

[0149] Determine if the sensor tags of two initial acquisition sensors are adjacent; if not adjacent, do not generate a faulty sensor; if adjacent, mark the initial acquisition sensor with the larger sensor tag as the first sensor and the initial acquisition sensor with the smaller sensor tag as the second sensor; mark the acquisition sensor with a sensor tag larger than the first sensor as the third sensor and the acquisition sensor with a sensor tag smaller than the second sensor as the fourth sensor; sort the amplitude values ​​corresponding to each third sensor from smallest to largest and generate a first sorting table; sort the amplitude values ​​corresponding to each fourth sensor from largest to smallest and generate a second sorting table.

[0150] If the order of the first sorting table is consistent with the order of the sensor tags corresponding to the third sensor, and the order of the second sorting table is consistent with the order of the sensor tags corresponding to the fourth sensor, then a fault command is generated.

[0151] Furthermore, methods for analyzing fault types include:

[0152] The characteristic parameters corresponding to each acoustic signal are used as analysis data. The analysis data is input into the trained type analysis model, and the corresponding type label is output. The type label is the numerical label corresponding to the fault type. The fault type of the cable is obtained according to the type label.

[0153] The training process for the type analysis model includes:

[0154] Different numerical labels are assigned to different fault types and marked as type labels; g sets of analysis data are collected in advance, and corresponding type labels are assigned to each of the g sets of analysis data, where g is an integer greater than 1;

[0155] The analysis data and corresponding type labels are converted into a set of feature vectors. Each set of feature vectors serves as the input to the type analysis model, which outputs a set of predicted type labels corresponding to each set of analysis data and uses the actual type labels corresponding to each set of analysis data as the prediction target. The actual type labels are the pre-collected type labels corresponding to the analysis data. The training objective is to minimize the sum of prediction errors of all analysis data. The type analysis model is trained until the sum of prediction errors converges, at which point training stops. The type analysis model is a deep neural network model.

[0156] Furthermore, the method for evaluating the propagation speed of acoustic signals in a cable includes:

[0157] The system presets a fault range and a normal propagation speed. The fault range includes a fault amplitude range and a fault frequency range. It subtracts the corresponding minimum value from the maximum value in the fault amplitude range to obtain the amplitude difference; subtracts the corresponding minimum value from the maximum value in the fault frequency range to obtain the frequency difference; compares the amplitude corresponding to the first sensor with the amplitude corresponding to the second sensor, marking the acoustic sensor with the larger amplitude as the near-end sensor and the acoustic sensor with the smaller amplitude as the far-end sensor; obtains the amplitude corresponding to the near-end sensor and marks it as the near-end amplitude; obtains the frequency corresponding to the near-end sensor and marks it as the near-end frequency; subtracts the maximum value in the fault amplitude range from the near-end amplitude to obtain the changed amplitude; subtracts the maximum value in the fault frequency range from the near-end frequency to obtain the changed frequency.

[0158] Divide the change amplitude by the amplitude difference to obtain the amplitude reduction coefficient; divide the change frequency by the frequency difference to obtain the frequency reduction coefficient; use the amplitude reduction coefficient, frequency reduction coefficient, and predicted type label as test data, input the test data into the trained coefficient prediction model, predict the corresponding speed reduction coefficient, and mark it as the single fault reduction coefficient; the training process of the coefficient prediction model is consistent with the training process of the type analysis model, and both are deep neural network models.

[0159] Multiply the single-fault reduction factor by the normal propagation speed to obtain the single-fault propagation speed.

[0160] Further, the times when the near-end sensor and the far-end sensor collect the acoustic wave signals are obtained respectively; the time when the near-end sensor collects the acoustic wave signal is marked as the first time, and the time when the far-end sensor collects the acoustic wave signal is marked as the second time; the second time is subtracted from the first time to obtain the time difference; the time difference is multiplied by the single fault propagation speed to obtain the distance difference; the coupling interval is subtracted from the distance difference and then divided by 2 to obtain the first distance; the first distance is the distance between the near-end sensor and the cable fault; the first distance is added to the distance difference to obtain the second distance, which is the distance between the far-end sensor and the cable fault; the cable fault is located based on the first distance and the second distance.

[0161] Furthermore, the method also includes: if the same acoustic sensor collects multiple acoustic signals, then all collected acoustic signals are classified to obtain P signal sets, each signal set corresponding to a cable fault;

[0162] Methods for obtaining a set of P signals include:

[0163] Obtain the acquisition time of each acoustic signal; sort each acquisition time from morning to night; obtain the sensor tag corresponding to the acoustic sensor for each acoustic signal; replace each acquisition time with the corresponding sensor tag to generate a third sorting table; analyze according to the ascending order of the third sorting table, add the first two sensor tags in the third sorting table to a signal set, and starting from the third sensor tag, classify each sensor tag in turn according to the classification rules, obtaining a total of P signal sets, where no two sensor tags are the same in each signal set, and P is an integer greater than 1.

[0164] Furthermore, the classification rules include:

[0165] Mark the sensor tag currently being classified as the current tag;

[0166] If the current tag is not adjacent to any sensor tag in all signal sets, then add the current tag to a new signal set;

[0167] If the current tag is adjacent to a sensor tag in a signal set that contains only one sensor tag, then the current tag is added to the corresponding signal set;

[0168] If the current tag is adjacent to sensor tags in multiple signal sets, then obtain the acquisition time corresponding to all sensor tags adjacent to the current tag, sort them, and add the current tag to the signal set corresponding to the sensor tag with the earliest acquisition time.

[0169] If the current tag is adjacent to a sensor tag in a signal set, then the current tag is added to the corresponding signal set.

[0170] Furthermore, based on the fault type of each cable fault, the propagation speed of the acoustic signal in the cable is analyzed;

[0171] Based on the third sorting table, the acoustic signal corresponding to the acquisition time of each sensor label in each signal set is obtained, and the corresponding feature parameters are extracted. Then, the type label corresponding to each signal set is obtained. Based on the feature parameters corresponding to the first and second sensor labels in each signal set, the amplitude reduction coefficient and frequency reduction coefficient corresponding to each signal set are calculated. The amplitude reduction coefficient, frequency reduction coefficient, and type label corresponding to each signal set are used as research data. The research data is input into the trained coefficient prediction model to predict the corresponding velocity reduction coefficient, which is marked as the multi-fault reduction coefficient. The multi-fault reduction coefficient is multiplied by the normal propagation speed to obtain the multi-fault propagation speed. The single-fault propagation speed is the propagation speed of the acoustic signal in the cable under multiple cable fault conditions.

[0172] Furthermore, according to embodiments of this application, the processes described above with reference to the flowchart can be implemented as computer software programs. For example, this application provides a non-transitory machine-readable storage medium storing machine-readable instructions that can be executed by a processor to perform instructions corresponding to the method steps provided in this application, such as an acoustic diagnostic method for locating faults in power grid equipment. When this computer program is executed by a central processing unit (CPU), it performs the functions defined in the method of this application.

[0173] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

[0174] In conclusion, the above description is only a preferred embodiment of the present invention and is 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 within the protection scope of the present invention.

Claims

1. A method of acoustic diagnosis for power grid equipment fault location, characterized in that, The method comprises the following steps: Collecting acoustic wave signals of the cable; De-noising the collected acoustic wave signals; Analyzing the de-noised acoustic wave signals to determine whether to generate a fault instruction; If a fault instruction is generated, analyze the corresponding fault type; According to different fault types, evaluate the propagation speed of the acoustic wave signal in the cable according to the influence of the dielectric constant and mechanical properties of the cable; According to the propagation speed of the acoustic wave signal, the fault of the cable is located; The acoustic wave signal is a mechanical wave caused by the slight vibration of the internal substance of the cable; A preset coupling interval is provided, and m acoustic wave sensors are coupled on the surface of the cable according to the coupling interval; The method for determining whether to generate a fault instruction comprises the following steps: Corresponding feature parameters are extracted from each acoustic wave signal, the feature parameters including amplitude, frequency and duration; different digital labels are sequentially set for the m acoustic wave sensors according to the current flow direction in the cable, and the sensors are marked as sensor labels, the sensor label range is ; the acoustic wave sensors collecting the acoustic wave signals are marked as collection sensors; the amplitudes of each are sorted from large to small, and the two collection sensors corresponding to the two amplitudes in the front are marked as first collection sensors; Determine whether the sensor tags of the two first acquisition sensors are adjacent; if not, do not generate a fault sensor; if adjacent, mark the first acquisition sensor with a larger sensor tag as a first sensor, and mark the first acquisition sensor with a smaller sensor tag as a second sensor; mark the acquisition sensor with a sensor tag larger than the first sensor as a third sensor, and mark the acquisition sensor with a sensor tag smaller than the second sensor as a fourth sensor; sort the amplitude of each third sensor from small to large and generate a first sorting table; sort the amplitude of each fourth sensor from large to small and generate a second sorting table; If the order of the first sorting table is consistent with the order of the corresponding sensor tag of the third sensor, and the order of the second sorting table is consistent with the order of the corresponding sensor tag of the fourth sensor, a fault instruction is generated; The method for analyzing the fault type comprises the following steps: Each feature parameter corresponding to the acoustic wave signal is used as analysis data, the analysis data is input into a trained type analysis model, and a corresponding type tag is output, the type tag is a digital tag corresponding to the fault type, and the fault type of the cable is obtained according to the type tag.

2. The acoustic diagnostic method for power grid equipment fault location according to claim 1, characterized in that, The step of de-noising the collected acoustic wave signals comprises the following steps: Step 1: set the wavelet base function and the decomposition layer number n; Step 2: i-layer wavelet decomposition of the acoustic signal using wavelet basis functions into a set of approximations and i sets of details, ; Step 3: calculate the signal-to-noise ratio of each group of detail items; Step 4: preset a signal-to-noise threshold, compare the signal-to-noise threshold with the signal-to-noise ratio of each group of approximation items and detail items, remove the detail items with a signal-to-noise ratio less than the signal-to-noise threshold, and retain the detail items with a signal-to-noise ratio greater than or equal to the signal-to-noise threshold; Step 5: according to the approximation items and the i-th group of detail items after threshold processing, the same wavelet base function as in step 2 is used for signal reconstruction; Step 6: calculate the signal-to-noise ratio of the reconstructed signal and compare it with the preset de-noising threshold; Step 7: If the signal ratio of the reconstructed signal is less than the denoising threshold, let and return to Step 2, replacing the acoustic signal with the reconstructed signal; if the signal ratio of the reconstructed signal is greater than or equal to the denoising threshold, or then proceed to Step 8. Step 8: output the reconstructed signal, which is the acoustic wave signal after de-noising.

3. The acoustic wave diagnostic method for fault positioning of the power grid equipment according to claim 2, wherein The training process of the type analysis model comprises the following steps: Different digital tags are set for different fault types and are marked as type tags; g groups of analysis data are collected in advance, and each group of analysis data is set with a corresponding type tag, g is an integer greater than 1; The analysis data and the corresponding type label are converted into a corresponding set of feature vectors; each set of feature vectors is taken as an input of a type analysis model, the type analysis model takes a set of predicted type labels corresponding to each set of analysis data as an output, the actual type label corresponding to each set of analysis data is taken as a prediction target, and the actual type label is a type label collected in advance and corresponding to the analysis data; a training target is to minimize the sum of prediction errors of all analysis data; the type analysis model is trained until the sum of prediction errors converges, and the training is stopped; and the type analysis model is a deep neural network model.

4. The acoustic diagnostic method for power grid equipment fault location according to claim 3, characterized in that, The method for evaluating the propagation speed of the sound wave signal in the cable comprises: A fault range and a normal propagation speed are preset, the fault range comprises a fault amplitude range and a fault frequency range; a maximum value in the fault amplitude range is subtracted by a corresponding minimum value to obtain an amplitude difference; a maximum value in the fault frequency range is subtracted by a corresponding minimum value to obtain a frequency difference; a value corresponding to the amplitude of the first sensor is compared with a value corresponding to the amplitude of the second sensor, the sound wave sensor corresponding to the larger amplitude is marked as a proximal sensor, and the sound wave sensor corresponding to the smaller amplitude is marked as a distal sensor; the amplitude corresponding to the proximal sensor is obtained and marked as a proximal amplitude; the frequency corresponding to the proximal sensor is obtained and marked as a proximal frequency; the proximal amplitude is subtracted by the maximum value in the fault amplitude range to obtain a change amplitude; the proximal frequency is subtracted by the maximum value in the fault frequency range to obtain a change frequency; The change amplitude is divided by the amplitude difference to obtain an amplitude reduction coefficient; the change frequency is divided by the frequency difference to obtain a frequency reduction coefficient; the amplitude reduction coefficient, the frequency reduction coefficient and the predicted type label are taken as test data, the test data is input into the trained coefficient prediction model, a corresponding speed reduction coefficient is predicted and marked as a single-fault reduction coefficient; the training process of the coefficient prediction model is consistent with the training process of the type analysis model, and both are deep neural network models; The single-fault reduction coefficient is multiplied by the normal propagation speed to obtain a single-fault propagation speed.

5. The acoustic diagnostic method for power grid equipment fault location according to claim 4, characterized in that, The time at which the proximal sensor collects the sound wave signal is marked as a first time, and the time at which the distal sensor collects the sound wave signal is marked as a second time; the second time is subtracted by the first time to obtain a time difference; the time difference is multiplied by the single-fault propagation speed to obtain a distance difference; the coupling interval is subtracted by the distance difference and then divided by 2 to obtain a first distance; the first distance is the distance between the proximal sensor and the cable fault; The first distance is added to the distance difference to obtain a second distance, and the second distance is the distance between the distal sensor and the cable fault; the cable fault is located according to the first distance and the second distance.

6. The acoustic diagnostic method for power grid equipment fault location according to claim 5, characterized in that, The method further comprises: if a same sound wave sensor collects multiple sound wave signals, classifying all collected sound wave signals to obtain P signal sets, each signal set corresponding to one cable fault; The method for obtaining the P signal sets comprises: Acquisition time of each acoustic signal is obtained; each acquisition time is sorted from early to late; a sensor tag corresponding to an acoustic sensor corresponding to each acoustic signal is obtained; each acquisition time is replaced by the corresponding sensor tag to generate a third sorting table; according to the third sorting table, the first two sensor tags in the third sorting table are added to a signal set in a positive sequence, and each sensor tag is classified according to a classification rule starting from the third sensor tag, so that P signal sets are obtained, and there is no same sensor tag in each signal set, and P is an integer greater than 1.

7. The acoustic diagnostic method for power grid equipment fault location according to claim 6, characterized in that, The classification rule comprises: a current sensor tag being classified is marked as a current tag; if the current tag is not adjacent to each sensor tag in all signal sets, the current tag is added to a new signal set; if the current tag is adjacent to a sensor tag in a signal set in which there is only one sensor tag, the current tag is added to the corresponding signal set; if the current tag is adjacent to sensor tags in multiple signal sets respectively, acquisition times corresponding to all sensor tags adjacent to the current tag are obtained and sorted, and the current tag is added to a signal set corresponding to a sensor tag with the earliest acquisition time; if the current tag is adjacent to a sensor tag in a signal set, the current tag is added to the corresponding signal set.

8. The acoustic diagnostic method for power grid equipment fault location according to claim 7, characterized in that, According to the fault type of each cable fault, the propagation speed of the acoustic signal in the cable is analyzed; According to the third sorting table, the acoustic signal corresponding to the acquisition time corresponding to each sensor tag in each signal set is obtained, and the corresponding characteristic parameters are extracted, and then the type tag corresponding to each signal set is obtained; according to the characteristic parameters corresponding to the first sensor tag and the second sensor tag in each signal set, the amplitude reduction coefficient and the frequency reduction coefficient corresponding to each signal set are calculated; the amplitude reduction coefficient, the frequency reduction coefficient and the type tag corresponding to each signal set are taken as research data, and the research data is input into a trained coefficient prediction model to predict the corresponding speed reduction coefficient and mark it as a multi-fault reduction coefficient; The multi-fault reduction coefficient is multiplied by the normal propagation speed to obtain a multi-fault propagation speed; The single-fault propagation speed is the propagation speed of the acoustic signal in the cable under the condition of multiple cable faults.

9. An acoustic diagnostic system for locating faults in power grid equipment, implementing the method of any one of claims 1 to 8, characterized in that, It comprises: an acoustic signal acquisition module for acquiring acoustic signals of a cable; an acoustic signal processing module for denoising the acquired acoustic signals; a fault judgment module for analyzing the denoised acoustic signals to determine whether to generate a fault instruction; a type analysis module for analyzing the corresponding fault type if the fault instruction is generated; a speed evaluation module for evaluating the propagation speed of the acoustic signal in the cable according to the fault type; a fault positioning module for positioning the cable according to the propagation speed of the acoustic signal.

Images