A power device voiceprint defect fault identification method of a smart grid

Abstract

This invention discloses a method for identifying acoustic fingerprint defects in power equipment in a smart grid, specifically relating to the field of acoustic fingerprint defect identification technology. This invention collects historical acoustic fingerprint signals from power equipment under normal and different fault states, extracts time-domain, frequency-domain, and time-frequency-domain parameters, and constructs a set of time-domain, frequency-domain, and time-frequency-domain ratios by calculating the ratios to normal state parameters. This set of ratios is then transformed into a closed graph to form a defect alarm graph set. Based on the graph area and weight, the defect alarm value is calculated, achieving standardized processing of historical fault characteristics. After real-time acquisition of acoustic fingerprint signals, a real-time alarm graph set and a real-time alarm value are generated according to the same logic. Through multi-dimensional comparison of the two, the existence of a fault is selectively determined, and a comprehensive evaluation value is further calculated to accurately match the fault type. This solves the problem in existing technologies where real-time and historical acoustic fingerprint signals of power equipment cannot be combined for analysis.

Description

Technical Field

[0001] This invention relates to the field of voiceprint defect recognition technology, and more specifically, to a method for identifying voiceprint defects in power equipment in a smart grid. Background Technology

[0002] A smart grid is a new type of power grid that integrates information technology, communication technology, power technology and other technologies. Its stable operation is crucial to social production and people's lives. As the core component of a smart grid, the operating status of power equipment directly affects the safety and reliability of the power grid. During the long-term operation of power equipment, various defects and faults are prone to occur due to mechanical wear, electrical aging, environmental factors and other factors. If they are not detected and dealt with in time, they may cause serious power grid accidents.

[0003] With the development of sensor technology and artificial intelligence technology, fault identification methods based on voiceprint signals have gradually attracted attention. Voiceprint signals are sound signals generated during the operation of power equipment, which contain information about the equipment's operating status. When a fault occurs in the equipment, the voiceprint signal will change accordingly. However, existing fault identification methods based on voiceprint signals still have the following shortcomings in practical applications:

[0004] On the one hand, the fault identification process ignores the historical acoustic signature signals of power equipment under different fault states, relies on a single signal source, and cannot combine and analyze the real-time acoustic signature signals and historical acoustic signature signals of power equipment to determine whether a fault exists and further locate the cause of the fault. The level of intelligence and data utilization of fault identification is low, and the accuracy cannot be guaranteed.

[0005] Furthermore, existing technologies for fault identification focus primarily on faults that have already occurred, lacking effective monitoring of potential faults that have not yet reached the fault threshold but exhibit abnormalities, and the reliability of early warning results is ambiguous.

[0006] Therefore, a method for identifying acoustic defects in power equipment in smart grids is proposed. Summary of the Invention

[0007] To overcome the aforementioned deficiencies of the prior art, embodiments of the present invention provide a method for identifying acoustic fingerprint defects in power equipment in a smart grid.

[0008] To achieve the above objectives, the present invention provides the following technical solution:

[0009] A method for identifying acoustic fingerprint defects in power equipment in a smart grid includes:

[0010] Equipment data processing: Collect historical acoustic fingerprint signals of each power equipment under normal operation and different fault conditions as training samples. After evaluating and processing the historical acoustic fingerprint signals of each power equipment under normal operation and different fault conditions using the set data processing logic, determine the defect alarm graphic set and defect alarm value of each power equipment under different fault conditions.

[0011] The specific process of assessment and treatment is as follows:

[0012] Time-domain parameters, frequency-domain parameters, and time-frequency-domain parameters are extracted from historical acoustic signature signals corresponding to fault states of power equipment. The time-domain parameters include time-domain peak value, time-domain mean, time-domain variance, and time-domain sheathness, denoted as […]. Frequency domain parameters include peak frequency, mean frequency, variance frequency, and bandwidth, denoted as . The time-frequency domain parameters include the proportion of high-energy regions, the migration distance of energy centers, and the wavelet energy entropy, denoted as... ;

[0013] Extract the time-domain parameters, frequency-domain parameters, and time-frequency-domain parameters of the power equipment under normal operating conditions, and denot them as follows: , , The ratios of any value of the time-domain parameter, frequency-domain parameter, and time-frequency-domain parameter of the power equipment under the fault state to the corresponding value under the normal operating state are calculated to obtain the time-domain ratio set a, the frequency-domain ratio set b, and the time-frequency-domain ratio set c of the power equipment under the fault state; that is, expressed as: ;

[0014] Voiceprint signal acquisition: The voiceprint signals acquired in real time from each power device are evaluated and processed in combination with the historical voiceprint signals under normal operating conditions to determine the real-time alarm graphic set and real-time alarm value of each power device.

[0015] Fault identification and processing: Based on the real-time alarm values ​​of each power device and the defect alarm values ​​under different fault states, it is determined whether there is a fault or potential fault. If the determination result shows that a power device has a fault, the corresponding steps are executed to locate the possible cause of the fault. If the determination result shows that a power device has a potential fault, the corresponding steps are executed to generate a warning confidence level and send it to the monitoring center of the smart grid simultaneously. The warning confidence level includes low confidence level, medium confidence level and high confidence level.

[0016] Specifically, determining the defect alarm graphic set and defect alarm value for each power device under different fault states involves:

[0017] Starting from the origin of the Cartesian coordinate system, the number of ratios within the set is counted and denoted as m. m is used as the number of extending rays, and the extension direction of each ray is determined according to a set angular distribution, where the set angle passes through... The calculation shows that m is the number of ratios in each set; each ray represents a set of ratios in the set, and the rays extend from the direction of each ray according to the size of the ratios until the length of the ray is equal to the size of the corresponding ratio. The endpoints of the extended rays are determined and connected sequentially to form a closed figure. The closed figures constructed by different sets under the corresponding fault conditions of power equipment are marked as time-domain figures, frequency-domain figures, and time-frequency-domain figures, and integrated as a set of defect early warning figures for the corresponding fault conditions of power equipment.

[0018] Extract the area of ​​the time-domain graph, frequency-domain graph, and time-frequency-domain graph corresponding to the fault state of the power equipment, multiply them by the preset time-domain weight, frequency-domain weight, and time-frequency-domain weight respectively, and then sum them to obtain the defect alarm value of the power equipment corresponding to the fault state.

[0019] Specifically, determining the real-time alarm graphic set and real-time alarm value for each power device involves:

[0020] Time-domain parameters, frequency-domain parameters, and time-frequency-domain parameters are extracted from the real-time acoustic fingerprint signals of the power equipment. The ratio of any value of the time-domain parameters, frequency-domain parameters, and time-frequency-domain parameters extracted from the real-time acoustic fingerprint signals of the power equipment to the corresponding value under normal operating conditions is calculated to obtain the time-domain ratio set, frequency-domain ratio set, and time-frequency-domain ratio set of the power equipment under real-time conditions.

[0021] After constructing a closed graph based on the process of constructing a set of early warning graphs for different sets of reasonable defects, the closed graphs constructed by different sets under the real-time state of the power equipment are used as the set of real-time alarm graphs under the real-time state of the power equipment.

[0022] Extract the area of ​​each group of closed shapes constructed under the real-time state of the power equipment, multiply it by the preset time domain weight, frequency domain weight and time-frequency domain weight respectively, and then sum them to obtain the real-time alarm value of the power equipment under the real-time state.

[0023] Specifically, the determination of whether a fault or potential fault exists based on the real-time alarm values ​​of each power device and the defect alarm values ​​under different fault conditions involves:

[0024] The real-time alarm value of the power equipment is extracted and compared with the preset alarm threshold and the defect alarm value under different fault conditions. If it is higher than the preset alarm threshold or the defect alarm value under any fault condition, a fault is determined and a deep fault inspection signal is triggered. Otherwise, the lower value is extracted from the defect alarm value under different fault conditions, and the pre-distance value is obtained by subtracting the real-time alarm value from the lower value. If the pre-distance value is less than the preset pre-distance reference value, a potential fault is determined and a pre-inspection signal is triggered.

[0025] Specifically, if the determination result indicates that a certain power equipment has a fault, then corresponding steps are executed to locate the possible cause of the power equipment fault, specifically:

[0026] If a fault deep inspection signaling is triggered, the defect alarm graphic set corresponding to different fault states of the power equipment is extracted. The three sets of graphics in the defect alarm graphic set under different fault states are numbered by z, x, and c, respectively. The three sets of graphics in the real-time alarm graphic set of the power equipment are represented by e, r, and t, respectively. Among them, z corresponds to e, x to r, and c to t.

[0027] Analyze the area depth detection values ​​between the defect alarm graphic set and the real-time alarm graphic set. X-ray depth detection value and contour depth detection value The area depth detection value of the defect alarm graphic set and the real-time alarm graphic set. X-ray depth detection value and contour depth detection value After normalization, substitute into the formula A weighted calculation is performed to obtain the comprehensive evaluation value P between the defect alarm graphic set and the real-time alarm graphic set; where These represent the area depth detection values, respectively. X-ray depth detection value and contour depth detection value The corresponding area weights, ray weights, and contour weights;

[0028] Under the triggering of the fault deep inspection signaling, the comprehensive evaluation value P obtained by analyzing the defect alarm graphic set and the current real-time alarm graphic set under different fault states is selected as the possible fault cause of the power equipment.

[0029] Specifically, the area depth detection value and X-ray depth detection value The specific analysis process is as follows:

[0030] Substitute the areas Qz, Qx, and Qc of the three sets of graphics in the defect alarm graphic set with the areas Qe, Qr, and Qt of the three sets of graphics in the real-time alarm graphic set into the formula. Calculations were performed to obtain the area depth detection value between the defect alarm graphic set and the real-time alarm graphic set. ;

[0031] Extract the lengths of each ray from the defect alarm graphic set numbered z. Extract the lengths of each ray from the real-time alarm graphic set numbered e. Calculate the absolute difference between the lengths of each ray in the z graphic and the lengths of rays in the same direction in the e graphic. Sum the absolute differences of each group to obtain the ray deviation value between the z and e graphics. Similarly, calculate the ray deviation values ​​between the x and r graphics, and between the c and t graphics. Accumulate the ray deviation values ​​of each group and divide by an integer three to obtain the ray depth detection value. .

[0032] Specifically, the contour depth detection value The specific analysis process is as follows:

[0033] Edge detection is performed on the z-shaped and e-shaped graphics. The closed boundary contours of the graphics are extracted by a threshold segmentation algorithm to obtain the z-shaped contour sequence and the e-shaped contour sequence composed of continuous pixels. An 8-direction Freeman chain code is used to map the directions of adjacent contour points in the z-shaped contour sequence and the e-shaped contour sequence to integers within a set range. That is, different directions correspond to a set of integers within a set range.

[0034] Using the lowest point on the left side of the z-shaped contour as a reference, traverse all contour points in a clockwise direction, record the integer values ​​converted from the directions of adjacent points as direction values, and use each set of direction values ​​as code values ​​to form a chain code sequence; similarly, form the chain code sequence of the e-shaped contour.

[0035] For the chain code sequence of the z-shape and the chain code sequence of the e-shape, calculate the code value difference at corresponding positions; sum the code value differences of each group and denote them as j; use the formula D=j / (n×4) to calculate the average difference rate D at each position between the z-shape and the e-shape; where n represents the chain code length and 4 is the maximum possible code value difference;

[0036] The contour similarity between the z-shape and the e-shape is obtained through 1-D. Similarly, the contour similarity between the x-shape and the r-shape, and between the c-shape and the t-shape are calculated. The contour similarities of each group are summed and divided by an integer three to obtain the contour depth detection value. .

[0037] Specifically, the triggering of the pre-inspection signaling involves the following steps:

[0038] Under the triggering of the pre-inspection signal, the comprehensive evaluation value P of the fault status defect alarm graphic set corresponding to the pre-distance value and the current real-time alarm graphic set is analyzed. Three sets of comprehensive evaluation value intervals corresponding to the comprehensive evaluation value P are preset, and each set of comprehensive evaluation value intervals corresponds to a warning confidence level. The analyzed comprehensive evaluation value P is matched with the corresponding comprehensive evaluation value interval to determine the warning confidence level of the current potential fault.

[0039] The technical effects and advantages of this invention are as follows:

[0040] (1) This invention collects historical acoustic signals of power equipment under normal and different fault states, extracts time domain, frequency domain, and time-frequency domain parameters, and constructs a set of time domain, frequency domain, and time-frequency domain ratios by calculating the ratio with the normal state parameters. The set of ratios is transformed into a closed graph to form a set of defect alarm graphs. The defect alarm value is calculated based on the area and weight of the graph, realizing the standardized processing of historical fault characteristics. After real-time acquisition of acoustic signals, a real-time alarm graph set and a real-time alarm value are generated according to the same logic. The existence of a fault is selectively determined by multi-dimensional comparison of the two, and the comprehensive evaluation value is further calculated to accurately match the fault type. This solves the problem in the prior art that ignores the historical acoustic signals of power equipment under different fault states, relies only on a single signal source, and cannot combine and analyze the real-time acoustic signals and historical acoustic signals of power equipment.

[0041] (2) In this invention, if the real-time alarm value is higher than the preset threshold or the defect alarm value, it is determined to be a fault and triggers the deep inspection signaling. The possible cause of the fault is located by the comprehensive evaluation value. If the real-time alarm value does not reach the threshold, the pre-distance value is calculated. When the pre-distance value is less than the reference value, it is determined to be a fault hazard and triggers the pre-inspection signaling to determine the confidence level of the early warning. This realizes full-cycle monitoring from fault detection to hazard early warning.

[0042] (3) In the process of analyzing the comprehensive evaluation value, the present invention comprehensively evaluates the area depth detection value, ray depth detection value and contour depth detection value of the defect alarm graphic set and the real-time alarm graphic set, which fully reflects the degree of fault similarity between the two, avoids the limitation of a single feature, and improves the accuracy of fault judgment. Attached Figure Description

[0043] Figure 1 This is a flowchart of a method for identifying acoustic defects in power equipment in a smart grid according to the present invention. Detailed Implementation

[0044] 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.

[0045] Example

[0046] like Figure 1 As shown, a method for identifying acoustic fingerprint defects in power equipment in a smart grid includes:

[0047] Equipment data processing: Collect historical acoustic fingerprint signals of each power equipment under normal operation and different fault conditions as training samples. After preprocessing the training samples, use the set data processing logic to evaluate and process the historical acoustic fingerprint signals of each power equipment under normal operation and different fault conditions, and determine the defect alarm graphic set and defect alarm value of each power equipment under different fault conditions.

[0048] For example, if the power equipment is a circuit breaker, historical voiceprint signals under conditions such as normal operation, contact overheating, insulation damage, and mechanical failure are collected as training samples.

[0049] Specifically:

[0050] Time-domain parameters, frequency-domain parameters, and time-frequency-domain parameters are extracted from historical acoustic signature signals corresponding to fault states of power equipment. The time-domain parameters include time-domain peak value, time-domain mean, time-domain variance, and time-domain sheathness, denoted as […]. Frequency domain parameters include peak frequency, mean frequency, variance frequency, and bandwidth, denoted as . The time-frequency domain parameters include the proportion of high-energy regions, the migration distance of energy centers, and the wavelet energy entropy, denoted as... ;

[0051] Additional explanation,

[0052] Peak value: The maximum amplitude of the acoustic signal over a period of time. When electrical equipment experiences mechanical wear or other faults, abnormal collisions between components may cause the peak value to increase significantly. For example, when the circuit breaker contacts overheat, the arc generated by poor contact will increase the peak value of the acoustic signal.

[0053] Mean: The average value of the acoustic signal amplitude. When the equipment is running normally, the mean is usually in a relatively stable range. Under fault conditions, such as wear of motor bearings, the increased vibration will cause the mean to shift significantly.

[0054] Variance: Reflects the degree of dispersion of the signal amplitude relative to the mean. When the equipment is running unstable, such as when the transformer windings are loose, the fluctuation of the acoustic signal will increase, and the variance will increase accordingly.

[0055] Kurtosis: Used to describe the steepness of the signal amplitude distribution; for faults containing impulsive components, such as the acoustic signal generated by a broken gear tooth, the kurtosis value will increase significantly;

[0056] Peak frequency: The amplitude corresponding to the frequency point with the highest energy in the spectrum diagram; different fault types often correspond to specific frequency peaks, for example, motor rotor faults will show a peak at a specific frequency;

[0057] Spectrum mean: The average level of spectrum energy; when the equipment is running normally, the spectrum mean is relatively stable, but when problems such as core failure occur, the spectrum mean will change;

[0058] Spectral variance: reflects the dispersion of spectral energy distribution; aging of equipment components and other faults may cause the spectral energy distribution to become more dispersed, increasing the spectral variance;

[0059] Bandwidth: The frequency range occupied by the main frequency components of a signal; for example, when a transformer experiences insulation damage, the bandwidth of its acoustic signal will widen.

[0060] High-frequency energy proportion: The proportion of high-frequency energy increases significantly under fault conditions and continues to increase as the fault intensifies;

[0061] Energy center of gravity migration distance: Faults can cause the energy center of gravity to shift towards higher or lower frequencies, increasing the migration distance;

[0062] Wavelet energy entropy: The energy entropy under fault conditions exhibits characteristic changes with the fault type. The entropy value rises to 0.5-0.6 in early faults and reaches 0.7-0.9 in severe faults.

[0063] Extract the time-domain parameters, frequency-domain parameters, and time-frequency-domain parameters of the power equipment under normal operating conditions, and denot them as follows: , , The ratios of any value of the time-domain parameter, frequency-domain parameter, and time-frequency-domain parameter of the power equipment under the fault state to the corresponding value under the normal operating state are calculated to obtain the time-domain ratio set a, the frequency-domain ratio set b, and the time-frequency-domain ratio set c of the power equipment under the fault state; that is, expressed as: Starting from the origin of the Cartesian coordinate system, the number of ratios within the set is counted and denoted as m. m is used as the number of extending rays, and the extension direction of each ray is determined according to a set angular distribution, where the set angle passes through... The calculation shows that m is the number of ratios within each set;

[0064] Each ray represents a set of ratios within the set. Starting from the direction of each ray's extension according to the magnitude of the ratio, the ray continues until its length equals the magnitude of the corresponding ratio. The endpoints of the extended rays are then determined and connected sequentially to form a closed figure.

[0065] Additional explanation: If a = {2.4, 1.2, 3.0, 2.5} under a certain fault condition, then the 0° ray will be extended by 2.4 units, the 90° ray by 1.2 units, the 180° ray by 3.0 units, and the 270° ray by 2.5 units; the unit length can be customized, such as 1 unit = 1 cm;

[0066] The closed graphs constructed from different sets corresponding to the fault states of power equipment are labeled as time-domain graphs, frequency-domain graphs, and time-frequency-domain graphs, and integrated as a set of defect early warning graphs for the corresponding fault states of power equipment.

[0067] To elaborate further, the closed figure constructed by the time-domain ratio set a and the frequency-domain ratio set b is a quadrilateral, while the closed figure constructed by the time-frequency-domain ratio set c is a triangle.

[0068] Extract the area of ​​the time-domain graph, frequency-domain graph, and time-frequency-domain graph corresponding to the fault state of the power equipment, multiply them by the preset time-domain weight, frequency-domain weight, and time-frequency-domain weight respectively, and then sum them to obtain the defect alarm value of the power equipment corresponding to the fault state.

[0069] Voiceprint signal acquisition: The voiceprint acquisition module is used to acquire the voiceprint signals of each power device in the smart grid in real time. After preprocessing, the voiceprint signals acquired in real time by each power device are evaluated and processed in combination with the historical voiceprint signals under normal operation to determine the real-time alarm graphic set and real-time alarm value of each power device.

[0070] The voiceprint acquisition module includes multiple sound sensors, which are evenly distributed around the power equipment to collect voiceprint signals from multiple directions, ensuring that the collected voiceprint signals are more comprehensive and accurate.

[0071] Preprocessing includes noise reduction, filtering, and normalization, specifically:

[0072] The noise reduction process employs a wavelet thresholding algorithm to decompose the original voiceprint signal into wavelet coefficients at different scales. A threshold is determined and the wavelet coefficients are then processed. The processed wavelet coefficients are then reconstructed using wavelet to obtain the denoised voiceprint signal. The wavelet thresholding algorithm effectively eliminates noise in the voiceprint signal and improves the signal-to-noise ratio.

[0073] The filtering process uses an adaptive filter that automatically adjusts the filtering parameters according to the frequency characteristics of the voiceprint signal, which can better filter out interference signals of different frequencies and retain useful voiceprint information.

[0074] Normalization is a process that adjusts the amplitude of the voiceprint signal to a preset range to eliminate the impact of signal amplitude differences on subsequent feature extraction and fault identification.

[0075] That is, extract time-domain parameters, frequency-domain parameters, and time-frequency-domain parameters from the real-time acoustic fingerprint signals collected by the power equipment, and calculate the ratio of any value of the time-domain parameters, frequency-domain parameters, and time-frequency-domain parameters extracted from the real-time acoustic fingerprint signals of the power equipment with the corresponding value under normal operating conditions to obtain the time-domain ratio set, frequency-domain ratio set, and time-frequency-domain ratio set of the power equipment under real-time conditions;

[0076] After constructing closed graphs based on different sets, the closed graphs constructed by different sets under the real-time state of the power equipment are used as the real-time alarm graph set under the real-time state of the power equipment.

[0077] Extract the area of ​​each group of closed shapes constructed under the real-time state of the power equipment, multiply it by the preset time domain weight, frequency domain weight and time-frequency domain weight respectively, and then sum them to obtain the real-time alarm value of the power equipment under the real-time state.

[0078] Fault identification and handling: Based on the real-time alarm values ​​of each power device and the defect alarm values ​​under different fault states, it is determined whether a fault or potential fault exists. If the determination result indicates that a power device has a fault, the corresponding steps are executed to locate the possible cause of the fault. If the determination result indicates that a power device has a potential fault, the corresponding steps are executed to generate a warning confidence level, which is simultaneously sent to the monitoring center of the smart grid so that staff can take appropriate measures in a timely manner. The information sent also includes the determination result, the location of the device, the possible cause of the fault, and the warning confidence level. The warning confidence level includes low confidence level, medium confidence level, and high confidence level.

[0079] Specifically:

[0080] The real-time alarm value of the power equipment is extracted and compared with the preset alarm threshold and the defect alarm value under different fault conditions. If it is higher than the preset alarm threshold or the defect alarm value under any fault condition, a fault is determined and a deep fault inspection signal is triggered. Otherwise, the lower value is extracted from the defect alarm value under different fault conditions, and the lower value is subtracted from the real-time alarm value to obtain the pre-distance value. If the pre-distance value is less than the preset pre-distance reference value, a potential fault is determined and a pre-inspection signal is triggered.

[0081] If a fault deep detection signal is triggered, the defect alarm graphic set corresponding to different fault states of the power equipment is extracted. The three sets of graphics in the defect alarm graphic set under different fault states are numbered by z, x, and c respectively. The three sets of graphics in the real-time alarm graphic set of the power equipment are represented by e, r, and t respectively. Among them, z and e, x and r, and c and t are constructed by the same type of parameters, namely time domain parameters, frequency domain parameters, and time-frequency domain parameters.

[0082] Substitute the areas Qz, Qx, and Qc of the three sets of graphics in the defect alarm graphic set with the areas Qe, Qr, and Qt of the three sets of graphics in the real-time alarm graphic set into the formula. Calculations were performed to obtain the area depth detection value between the defect alarm graphic set and the real-time alarm graphic set. ;

[0083] The ratio of the real-time graphic area to the defect graphic area is measured to be 1. For example, the ratio of the real-time frequency domain graphic area of ​​4.2 cm² to the mechanical fault frequency domain graphic area of ​​4.0 cm² is 1.05, which is a deviation of 5%.

[0084] Extract the lengths of each ray from the defect alarm graphic set numbered z. Extract the lengths of each ray from the real-time alarm graphic set numbered e. Calculate the absolute difference between the lengths of each ray in the z graphic and the lengths of rays in the same direction in the e graphic. Sum the absolute differences of each group to obtain the ray deviation value between the z and e graphics. Similarly, calculate the ray deviation values ​​between the x and r graphics, and between the c and t graphics. Accumulate the ray deviation values ​​of each group and divide by an integer three to obtain the ray depth detection value. ;

[0085] For example, the deviation between the 0° ray length of 2.3 in the real-time time-domain graph and the 0° ray length of 2.4 in the contact overheating time-domain graph is 0.1.

[0086] Edge detection is performed on the z-shaped and e-shaped graphics. The closed boundary contours of the graphics are extracted by a threshold segmentation algorithm (such as the Otsu algorithm) to obtain the contour sequences of the z-shaped and e-shaped graphics, which are composed of continuous pixels. If the number of contour points of the two graphics is different, the chain code length is unified by interpolation or sampling.

[0087] An 8-direction Freeman chain code is used to map the directions of adjacent contour points in the z-shaped and e-shaped contour sequences to integers within a set range. That is, different directions correspond to a set of integers within a set range. The set range can be set to 0-7, for example, 0 represents right, 1 represents upper right, 2 represents up, 3 represents upper left, 4 represents left, 5 represents lower left, 6 represents down, and 7 represents lower right.

[0088] Using the lowest point on the left side of the z-shaped contour as a reference, traverse all contour points in a clockwise direction, record the integer values ​​converted from the directions of adjacent points as direction values, and use each set of direction values ​​as code values ​​to form a chain code sequence; similarly, form the chain code sequence of the e-shaped contour.

[0089] After the chain code sequence is formed, the following processing is performed:

[0090] Starting point normalization: Since the starting point of the contour may be offset, the chain code is cyclically shifted to match the minimum prefix of the two chain codes. For example, the real-time chain code [1, 2, 3, 1, 2] and the defect chain code [2, 3, 1, 2, 1] can be unified into [1, 2, 3, 1, 2] after cyclic shifting, eliminating the influence of the difference in the starting point position;

[0091] Rotation normalization: Calculate the direction histogram of the chain code sequence, find the direction with the highest frequency as the main direction, and rotate the two chain codes to be consistent with the main direction; for example, if the main direction of the real-time graphic chain code is 2 (top) and the main direction of the defect graphic is 3 (top left), then rotate the defect chain code by -1 direction (i.e., reduce all code values ​​by 1 and take the remainder modulo 8) to achieve rotation invariance.

[0092] Scaling normalization: By calculating the ratio of the side length of the bounding rectangle of the contour, the pixel distance corresponding to the chain code is proportionally adjusted to ensure that the scaling ratio of the two graphics is consistent. For example, if the aspect ratio of the bounding rectangle of the real-time graphic is 2:1 and that of the defect graphic is 4:1, then the lateral distance of the defect chain code is compressed to 1 / 2 to ensure scaling independence.

[0093] Calculate the code value difference at corresponding positions between the chain code sequence of the z-shape and the chain code sequence of the e-shape; if the code value difference is greater than 4, then subtract the code value difference from 8, since the reverse difference is the smallest among the 8-direction chain codes, which is 4.

[0094] Sum the differences of each group of code values ​​and denote them as j. Use the formula D=j / (n×4) to calculate the average difference rate D between each position of the z-shaped and e-shaped graphs; where n represents the chain code length and 4 is the maximum possible code value difference; the value of D ranges from 0 to 1, and the smaller the value, the higher the chain code matching degree.

[0095] The contour similarity between the z-shape and the e-shape is obtained through 1-D. Similarly, the contour similarity between the x-shape and the r-shape, and between the c-shape and the t-shape are calculated. The contour similarities of each group are summed and divided by an integer three to obtain the contour depth detection value. ;

[0096] Area depth detection values ​​for defect alarm graphic sets and real-time alarm graphic sets X-ray depth detection value and contour depth detection value After normalization, substitute into the formula A weighted calculation is performed to obtain the comprehensive evaluation value P between the defect alarm graphic set and the real-time alarm graphic set; where These represent the area depth detection values, respectively. X-ray depth detection value and contour depth detection value The corresponding area weights, ray weights, and contour weights;

[0097] To elaborate further, by using three dimensions—area depth detection value, ray depth detection value, and contour depth detection value—the differences between real-time graphics and defect graphics are comprehensively captured, avoiding the limitations of single feature judgment. Area deviation reflects overall scale changes, ray deviation focuses on differences in local parameter ratios, and contour similarity reflects the degree of shape feature matching. The combination of the three makes fault type matching more accurate.

[0098] Under the triggering of fault deep inspection signaling, for the comprehensive evaluation value P obtained by analyzing the defect alarm graphic set and the current real-time alarm graphic set under different fault states, the fault state with the higher comprehensive evaluation value P is selected as the possible fault cause of the power equipment.

[0099] Under the triggering of the pre-inspection signal, the comprehensive evaluation value P of the fault status defect alarm graphic set corresponding to the pre-distance value and the current real-time alarm graphic set is analyzed. Three sets of comprehensive evaluation value intervals corresponding to the comprehensive evaluation value P are preset, and each set of comprehensive evaluation value intervals corresponds to a warning confidence level. The larger the comprehensive evaluation value P, the higher the probability of matching a high confidence level.

[0100] The comprehensive evaluation value P is matched with the corresponding comprehensive evaluation value range to determine the confidence level of the early warning of the current potential fault.

[0101] In addition, by calculating the "pre-distance value" (the difference between the lower defect alarm value and the real-time alarm value) and combining it with the preset pre-distance reference value, early abnormalities in the equipment status (such as parameters slowly deviating but not reaching the fault standard) can be accurately captured before a fault occurs, and pre-inspection signaling can be triggered to achieve preventive maintenance and significantly reduce the risk of sudden faults.

[0102] In the early warning of potential faults, the warning results are divided into three levels: "low confidence, medium confidence, and high confidence" by setting the range of the comprehensive evaluation value P. This hierarchical mechanism makes the warning information more decision-making and guides the work, avoids staff from over-responding to ineffective warnings, and improves operation and maintenance efficiency.

[0103] The above formulas are all dimensionless calculations. Dimensionless calculations can be performed using various methods such as standardization, which will not be elaborated here. The formulas are derived from software simulations based on a large amount of collected data, and the preset parameters in the formulas can be set by those skilled in the art according to the actual situation.

[0104] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more sets of available media. The available medium can be a magnetic medium (e.g., floppy disk, ATA hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium. The semiconductor medium can be a solid-state ATA hard disk.

[0105] It should be understood that in the various embodiments of this application, the order of the above-mentioned processes does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0106] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0107] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0108] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment, depending on actual needs.

[0109] In addition, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0110] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable ATA hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0111] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for identifying acoustic fingerprint defects in power equipment in a smart grid, characterized in that, include: Equipment data processing: Collect historical acoustic fingerprint signals of each power equipment under normal operation and different fault conditions as training samples. After evaluating and processing the historical acoustic fingerprint signals of each power equipment under normal operation and different fault conditions using the set data processing logic, determine the defect alarm graphic set and defect alarm value of each power equipment under different fault conditions. The specific process of assessment and treatment is as follows: Time-domain parameters, frequency-domain parameters, and time-frequency-domain parameters are extracted from historical acoustic signature signals corresponding to fault states of power equipment. The time-domain parameters include time-domain peak value, time-domain mean, time-domain variance, and time-domain sheathness, denoted as […]. Frequency domain parameters include peak frequency, mean frequency, variance frequency, and bandwidth, denoted as . The time-frequency domain parameters include the proportion of high-energy regions, the migration distance of energy centers, and the wavelet energy entropy, denoted as... ; Extract the time-domain parameters, frequency-domain parameters, and time-frequency-domain parameters of the power equipment under normal operating conditions, and denot them as follows: , , The ratios of any value of the time-domain parameter, frequency-domain parameter, and time-frequency-domain parameter of the power equipment under the fault state to the corresponding value under the normal operating state are calculated to obtain the time-domain ratio set a, the frequency-domain ratio set b, and the time-frequency-domain ratio set c of the power equipment under the fault state; that is, expressed as: ; Voiceprint signal acquisition: The voiceprint signals acquired in real time from each power device are evaluated and processed in combination with the historical voiceprint signals under normal operating conditions to determine the real-time alarm graphic set and real-time alarm value of each power device. Fault identification and processing: Based on the real-time alarm values ​​of each power equipment and the defect alarm values ​​under different fault states, determine whether there is a fault or potential fault. If the determination result shows that a power equipment has a fault, then execute the corresponding steps to locate the possible cause of the power equipment fault. If the determination result shows that a power equipment has a potential fault, then execute the corresponding steps to generate an early warning confidence level and send it to the monitoring center of the smart grid simultaneously. The reliability levels of the early warning system include low reliability, medium reliability, and high reliability. Based on the real-time alarm values ​​of each power device and the defect alarm values ​​under different fault conditions, it is determined whether a fault or potential fault exists, specifically as follows: The real-time alarm value of the power equipment is extracted and compared with the preset alarm threshold and the defect alarm value under different fault conditions. If it is higher than the preset alarm threshold or the defect alarm value under any fault condition, a fault is determined and a deep fault inspection signal is triggered. Otherwise, the low value is extracted from the defect alarm value under different fault conditions, and the pre-distance value is obtained by subtracting the real-time alarm value from the low value. If the pre-distance value is less than the preset pre-distance reference value, a potential fault is determined and a pre-inspection signal is triggered.

2. The method for identifying acoustic defect faults in power equipment in a smart grid according to claim 1, characterized in that, The determination of the defect alarm graphic set and defect alarm value for each power equipment under different fault states specifically involves: Starting from the origin of the Cartesian coordinate system, the number of ratios within the set is counted and denoted as m. m is used as the number of extending rays, and the extension direction of each ray is determined according to a set angular distribution, where the set angle passes through... The calculation shows that m is the number of ratios within each set; Each ray represents a set of ratios within the set. The rays extend from their respective directions according to the ratios until the ray length equals the corresponding ratio. The endpoints of the extended rays are determined and connected sequentially to form a closed figure. The closed figures constructed by different sets corresponding to the fault states of power equipment are marked as time-domain figures, frequency-domain figures, and time-frequency-domain figures, and integrated as a set of defect early warning figures for the corresponding fault states of power equipment. Extract the area of ​​the time-domain graph, frequency-domain graph, and time-frequency-domain graph corresponding to the fault state of the power equipment, multiply them by the preset time-domain weight, frequency-domain weight, and time-frequency-domain weight respectively, and then sum them to obtain the defect alarm value of the power equipment corresponding to the fault state.

3. The method for identifying acoustic fingerprint defects in power equipment in a smart grid according to claim 2, characterized in that, The determination of the real-time alarm graphic set and real-time alarm value for each power device specifically involves: Time-domain parameters, frequency-domain parameters, and time-frequency-domain parameters are extracted from the real-time acoustic fingerprint signals of the power equipment. The ratio of any value of the time-domain parameters, frequency-domain parameters, and time-frequency-domain parameters extracted from the real-time acoustic fingerprint signals of the power equipment to the corresponding value under normal operating conditions is calculated to obtain the time-domain ratio set, frequency-domain ratio set, and time-frequency-domain ratio set of the power equipment under real-time conditions. After constructing a closed graph based on the process of constructing a set of early warning graphs for different sets of reasonable defects, the closed graphs constructed by different sets under the real-time state of the power equipment are used as the set of real-time alarm graphs under the real-time state of the power equipment. Extract the area of ​​each group of closed shapes constructed under the real-time state of the power equipment, multiply it by the preset time domain weight, frequency domain weight and time-frequency domain weight respectively, and then sum them to obtain the real-time alarm value of the power equipment under the real-time state.

4. The method for identifying acoustic defect faults in power equipment in a smart grid according to claim 1, characterized in that, If the determination result indicates that a certain power equipment has a fault, then the corresponding steps are executed to locate the possible causes of the power equipment fault, specifically: If a fault deep inspection signaling is triggered, the defect alarm graphic set corresponding to different fault states of the power equipment is extracted. The three sets of graphics in the defect alarm graphic set under different fault states are numbered by z, x, and c, respectively. The three sets of graphics in the real-time alarm graphic set of the power equipment are represented by e, r, and t, respectively. Among them, z corresponds to e, x to r, and c to t. Analyze the area depth detection values ​​between the defect alarm graphic set and the real-time alarm graphic set. X-ray depth detection value and contour depth detection value The area depth detection value of the defect alarm graphic set and the real-time alarm graphic set. X-ray depth detection value and contour depth detection value After normalization, substitute into the formula A weighted calculation is performed to obtain the comprehensive evaluation value P between the defect alarm graphic set and the real-time alarm graphic set; where These represent the area depth detection values, respectively. X-ray depth detection value and contour depth detection value The corresponding area weights, ray weights, and contour weights; Under the triggering of the fault deep inspection signaling, for the comprehensive evaluation value P obtained by analyzing the defect alarm graphic set and the current real-time alarm graphic set under different fault states, the fault state with the higher comprehensive evaluation value P is selected as the possible fault cause of the power equipment.

5. The method for identifying acoustic fingerprint defects in power equipment in a smart grid according to claim 4, characterized in that, The area depth detection value and X-ray depth detection value The specific analysis process is as follows: Substitute the areas Qz, Qx, and Qc of the three sets of graphics in the defect alarm graphic set with the areas Qe, Qr, and Qt of the three sets of graphics in the real-time alarm graphic set into the formula. Calculations were performed to obtain the area depth detection value between the defect alarm graphic set and the real-time alarm graphic set. ; Extract the lengths of each ray from the defect alarm graphic set numbered z. Extract the lengths of each ray from the real-time alarm graphic set numbered e. Calculate the absolute difference between the lengths of each ray in the z graphic and the lengths of rays in the same direction in the e graphic. Sum the absolute differences of each group to obtain the ray deviation value between the z and e graphics. Similarly, calculate the ray deviation values ​​between the x and r graphics, and between the c and t graphics. Accumulate the ray deviation values ​​of each group and divide by an integer three to obtain the ray depth detection value. .

6. The method for identifying acoustic fingerprint defects in power equipment in a smart grid according to claim 4, characterized in that, The contour depth detection value The specific analysis process is as follows: Edge detection is performed on the z-shaped and e-shaped graphics. The closed boundary contours of the graphics are extracted by a threshold segmentation algorithm to obtain the z-shaped contour sequence and the e-shaped contour sequence composed of continuous pixels. An 8-direction Freeman chain code is used to map the directions of adjacent contour points in the z-shaped contour sequence and the e-shaped contour sequence to integers within a set range. That is, different directions correspond to a set of integers within a set range. Using the lowest point on the left side of the z-shaped contour as a reference, traverse all contour points in a clockwise direction, record the integer values ​​converted from the directions of adjacent points as direction values, and use each set of direction values ​​as code values ​​to form a chain code sequence; similarly, form the chain code sequence of the e-shaped contour. Calculate the code value difference at corresponding positions between the chain code sequence of the z-shape and the chain code sequence of the e-shape; Sum the differences of each group of code values ​​and denote them as j. Use the formula D=j / (n×4) to calculate the average difference rate D between each position of the z-shaped and e-shaped graphs; where n represents the chain code length and 4 is the maximum possible code value difference. The contour similarity between the z-shape and the e-shape is obtained through 1-D. Similarly, the contour similarity between the x-shape and the r-shape, and between the c-shape and the t-shape are calculated. The contour similarities of each group are summed and divided by an integer three to obtain the contour depth detection value. .

7. A method for identifying acoustic fingerprint defects in power equipment in a smart grid according to claim 4, characterized in that, The triggering of the pre-inspection signaling executes the corresponding steps, specifically: Under the triggering of the pre-inspection signal, the comprehensive evaluation value P of the fault status defect alarm graphic set corresponding to the pre-distance value and the current real-time alarm graphic set is analyzed. Three sets of comprehensive evaluation value intervals corresponding to the comprehensive evaluation value P are preset, and each set of comprehensive evaluation value intervals corresponds to a pre-warning confidence level. The comprehensive evaluation value P is matched with the corresponding comprehensive evaluation value range to determine the confidence level of the early warning of the current potential fault.

Images