Fault positioning method and system based on multi-modal acousto-optic-electric signal cooperative diagnosis

Abstract

The application provides a fault positioning method and system based on multi-modal acousto-optoelectric signal cooperative diagnosis, which first collects sound signal sequences, light signal sequences and electric signal sequences during the operation of high-voltage electrical equipment, then performs cross-modal signal hierarchical correlation processing on the multi-modal signals to generate a multi-modal signal correlation hierarchical chain, extracts a dynamic correlation feature set based on the multi-modal signal correlation hierarchical chain, inputs the dynamic correlation feature set into a fault feature evolution model to generate a fault feature identification sequence and a feature space diffusion trajectory, and finally determines the fault starting area and the fault influence boundary range of the equipment, thereby effectively improving the accuracy and reliability of high-voltage electrical equipment fault positioning.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of power operation and maintenance, in particular to a fault positioning method and system based on multi-modal acoustic-optical-electric signal collaborative diagnosis. BACKGROUND

[0002] In the process of continuous development of the power industry, high-voltage electrical equipment, as a core component of the power grid, its stability is directly related to the safety and reliability of the entire power system. With the growth of equipment operation time and the complexity of the operating environment, the risk of failure is increasing.

[0003] At present, there are many limitations in the fault positioning method of high-voltage electrical equipment. The traditional fault diagnosis method often only relies on a single type of signal, such as detecting the abnormality of the equipment by analyzing only the electrical signal, but the above method is easily affected by external interference and the complex characteristics of the equipment itself, resulting in inaccurate fault positioning. Some methods try to combine multiple signals, but lack of mining the deep relationship between different modal signals, and fail to fully consider the collaborative mechanism of acoustic, optical and electrical signals in the process of equipment failure. Different signals have a specific trigger response relationship in the process of fault triggering and propagation, and the existing technology fails to effectively establish the relationship model, making it difficult to accurately determine the starting area and impact boundary range of the fault when facing complex faults. SUMMARY

[0004] In view of the above-mentioned problems, in combination with the first aspect of the present application, the embodiments of the present application provide a fault positioning method based on multi-modal acoustic-optical-electric signal collaborative diagnosis, which comprises:

[0005] Collecting acoustic signal sequences, optical signal sequences and electrical signal sequences of high-voltage electrical equipment in the running state, the acoustic signal sequences containing acoustic waveform change information generated by the vibration of the internal components of the equipment, the optical signal sequences containing light radiation intensity change information generated by the partial discharge of the equipment surface, and the electrical signal sequences containing instantaneous fluctuation information of current and voltage in the equipment loop;

[0006] Performing cross-modal signal hierarchical correlation processing on the acoustic signal sequences, optical signal sequences and electrical signal sequences, establishing the trigger response hierarchical relationship between different signal sequences, and generating a multi-modal signal correlation hierarchical chain;

[0007] Extracting a dynamic correlation feature set of each signal sequence in the hierarchical response process based on the multi-modal signal correlation hierarchical chain, the dynamic correlation feature set containing amplitude linkage parameters, phase synchronization parameters and waveform distortion correlation parameters of the trigger signal and the hierarchical response signal;

[0008] input the dynamic correlation feature set into a preset fault feature evolution model to generate a fault feature identification sequence and a feature space diffusion trajectory;

[0009] determine a fault starting area and a fault influence boundary range of the high-voltage electrical equipment according to the fault feature identification sequence and the feature space diffusion trajectory.

[0010] In still another aspect, an embodiment of the present application also provides a fault positioning system based on collaborative diagnosis of multi-modal acoustic-optical-electrical signals, which comprises a processor and a machine readable storage medium, the machine readable storage medium is connected with the processor, the machine readable storage medium is used for storing programs, instructions or codes, and the processor is used for executing the programs, instructions or codes in the machine readable storage medium to realize the above method.

[0011] Based on the above aspects, an embodiment of the present application acquires multi-dimensional information such as internal component vibration, surface partial discharge and loop current voltage fluctuation of the equipment by collecting acoustic, optical and electrical signal sequences under the running state of the high-voltage electrical equipment, performs cross-modal signal hierarchical correlation processing on the multi-modal signals, establishes a trigger response hierarchical relationship between different signal sequences, generates a multi-modal signal correlation hierarchical chain, deeply reveals the internal relationship of acoustic, optical and electrical signals in the fault occurrence process, extracts a dynamic correlation feature set based on the correlation hierarchical chain, accurately captures key parameters such as amplitude linkage, phase synchronization and waveform distortion correlation between the trigger signals and the hierarchical response signals, further describes the dynamic features of the fault signals, inputs these features into a preset fault feature evolution model to generate a fault feature identification sequence and a feature space diffusion trajectory, can intuitively present the development and change process and the space propagation path of the fault features, and finally determines the fault starting area and the influence boundary range according to these results, thereby improving the accuracy and reliability of fault positioning of the high-voltage electrical equipment. BRIEF DESCRIPTION OF DRAWINGS

[0012] Figure 1 is an execution flow schematic diagram of the fault positioning method based on collaborative diagnosis of multi-modal acoustic-optical-electrical signals provided by an embodiment of the present application.

[0013] Figure 2 is a schematic diagram of exemplary hardware and software components of the fault positioning system based on collaborative diagnosis of multi-modal acoustic-optical-electrical signals provided by an embodiment of the present application. DETAILED DESCRIPTION

[0014] The present application will be described in detail below with reference to the accompanying drawings, Figure 1 is a flow schematic diagram of the fault positioning method based on collaborative diagnosis of multi-modal acoustic-optical-electrical signals provided by an embodiment of the present application, and the fault positioning method based on collaborative diagnosis of multi-modal acoustic-optical-electrical signals will be described in detail below.

[0015] Step S110: Collecting the sound signal sequence, the light signal sequence and the electric signal sequence of the high-voltage electrical equipment in the running state, the sound signal sequence contains the sound wave waveform change information generated by the vibration of the internal components of the equipment, the light signal sequence contains the light radiation intensity change information generated by the partial discharge of the surface of the equipment, and the electric signal sequence contains the instantaneous fluctuation information of the current and voltage in the loop of the equipment.

[0016] In the embodiment, a high-voltage circuit breaker in operation is selected as a specific object of the high-voltage electrical equipment. In order to comprehensively capture the multi-modal signals of the equipment in operation, corresponding sensing devices need to be deployed at key positions of the circuit breaker. For the collection of the sound signal sequence, sound sensors are installed near the components prone to vibration, such as the operating mechanism and the arc extinguishing chamber of the circuit breaker. When the circuit breaker is operated to open or close, the mechanical movement of the operating mechanism and the arc extinguishing process inside the arc extinguishing chamber will generate sound waves of different frequencies and amplitudes, which together constitute the sound signal sequence, and the waveform of the sound signal sequence will change with the operation process and the state of the equipment.

[0017] The collection of the light signal sequence is aimed at the positions of the circuit breaker where partial discharge may occur, such as the insulating pull rod and the surrounding of the moving and static contacts. When partial discharge occurs in these positions due to insulation defects or poor contact, light radiation will occur, and the light sensor can sense the subtle changes in the light radiation intensity, thereby forming the light signal sequence. For example, when there is an oxide layer on the surface of the moving and static contacts, causing an increase in contact resistance, weak discharge may occur locally, and the corresponding light intensity fluctuation will appear in the light signal sequence.

[0018] The collection of the electric signal sequence is achieved by setting current transformers and voltage transformers in the incoming and outgoing line loops of the circuit breaker. The current transformer is used to monitor the instantaneous changes of the current in the loop, and the voltage transformer is responsible for capturing the instantaneous fluctuations of the voltage. When the circuit breaker is in normal operation, the current and voltage are in a relatively stable fluctuation state; when faults such as contact welding and insulation breakdown occur, the current or voltage will exhibit abnormal instantaneous changes, and the above change information will be recorded in the electric signal sequence.

[0019] Step S120: Performing cross-modal signal hierarchical correlation processing on the sound signal sequence, the light signal sequence and the electric signal sequence, establishing the trigger-response hierarchical relationship between different signal sequences, and generating a multi-modal signal correlation hierarchical chain.

[0020] After obtaining the sound signal sequence, the light signal sequence and the electric signal sequence of the high-voltage circuit breaker, cross-modal signal hierarchical correlation processing is needed to sort out the trigger-response relationship between different signals and construct a hierarchical chain that can reflect their internal relationship.

[0021] Step S121: intercepting a signal segment of the sound signal sequence, the light signal sequence and the electric signal sequence in the same monitoring period, the signal segment containing continuous signal fluctuation units and signal stable units, wherein the signal fluctuation unit is a signal segment with fluctuation in amplitude or frequency, and the signal stable unit is a signal segment with stable amplitude and frequency.

[0022] In this embodiment, a fixed monitoring period is set. The signal segment in the monitoring period is intercepted from the sound signal sequence, which contains multiple signal fluctuation units and signal stable units. For example, when the circuit breaker is tripped, the action of the operating mechanism will cause the amplitude and frequency of the sound signal to fluctuate significantly, generating a signal fluctuation unit; when the circuit breaker is in a stable closed state, the amplitude and frequency of the sound signal are relatively stable, generating a signal stable unit.

[0023] Similarly, the signal segment of the same monitoring period is intercepted from the light signal sequence. If a short-term partial discharge occurs in the insulating pull rod of the circuit breaker during the monitoring period, the intensity of the light signal will fluctuate, generating a signal fluctuation unit; during the period without partial discharge, the light signal intensity remains stable, generating a signal stable unit.

[0024] For the electric signal sequence, the signal segment of the same monitoring period is intercepted. When the circuit breaker is normally carrying current, the amplitudes and frequencies of the current and voltage are relatively stable, generating a signal stable unit; when the circuit breaker is closed, the current will change suddenly at the moment of loop connection, generating a signal fluctuation unit.

[0025] Step S122: identifying a sound wave mutation segment in the signal segment of the sound signal sequence, the amplitude change rate of which exceeds a preset change rate threshold, recording the starting time mark of the sound wave mutation segment as the starting point of the sound trigger level, and the amplitude change rate being the change amount of amplitude per unit time.

[0026] After obtaining the sound signal segment, it needs to be analyzed to identify the sound wave mutation segment. First, according to the amplitude change characteristics of the sound signal when the circuit breaker is normally running, a suitable amplitude change rate threshold is set.

[0027] Step S1221: traversing each signal sampling point in the signal segment of the sound signal sequence, calculating the amplitude change amount and change time interval of the current sampling point and the previous sampling point, the amplitude change amount being the difference between the amplitude of the current sampling point and the amplitude of the previous sampling point, and the change time interval being the time difference between the two sampling points.

[0028] Traverse all sampling points in the sound signal segment. For each current sampling point, subtract the amplitude of the previous sampling point from the amplitude of the current sampling point to obtain the amplitude change amount. The time difference between two adjacent sampling points is the change time interval, which is determined by the sampling frequency.

[0029] Step S1222: calculating a magnitude change rate according to the magnitude change amount and the change time interval, the magnitude change rate being a ratio of the magnitude change amount to the change time interval, the ratio being positive when indicating an increase in the magnitude and being negative when indicating a decrease in the magnitude.

[0030] The calculated magnitude change amount is divided by the change time interval to obtain the magnitude change rate. If the ratio is positive, it indicates that the magnitude of the acoustic signal is increasing; if it is negative, it indicates that the magnitude is decreasing.

[0031] Step S1223: when the magnitude change rate of a plurality of continuous sampling points exceeds a preset change rate threshold, marking a signal segment composed of the plurality of continuous sampling points as a candidate acoustic wave mutation segment.

[0032] In the traversal process, if it is found that the magnitude change rate of a plurality of continuous sampling points exceeds a preset change rate threshold, a signal segment composed of the plurality of continuous sampling points is marked as a candidate acoustic wave mutation segment. For example, when the operating mechanism of a circuit breaker has a jamming fault, the magnitude change rate of the acoustic wave generated in the movement process of the operating mechanism will exceed the threshold for a plurality of continuous sampling points, thereby forming a candidate acoustic wave mutation segment.

[0033] Step S1224: positioning a starting sampling point of the candidate acoustic wave mutation segment and extracting a time coordinate of the starting sampling point in the acoustic signal segment.

[0034] For each candidate acoustic wave mutation segment, a first sampling point, i.e., a starting sampling point, is determined. According to the position of the sampling point in the acoustic signal segment, a corresponding time coordinate is determined.

[0035] Step S1225: taking the time coordinate as an acoustic trigger level starting point, repeating the above operation on all candidate acoustic wave mutation segments in the signal segment of the acoustic signal sequence that meet the conditions, and generating an acoustic trigger level starting point set, in which the starting points are arranged in time sequence and are attached with corresponding signal magnitude information.

[0036] The time coordinate of the starting sampling point is taken as an acoustic trigger level starting point, and all candidate acoustic wave mutation segments in the signal segment that meet the conditions are processed in the same way to obtain a plurality of acoustic trigger level starting points. These starting points are arranged in time sequence, and each starting point is attached with corresponding signal magnitude information to generate an acoustic trigger level starting point set.

[0037] Step S123: identifying a light radiation mutation segment in which a light intensity change rate exceeds a preset change rate threshold in the signal segment of the light signal sequence, recording a starting time mark of the light radiation mutation segment as a light trigger level starting point, and the light intensity change rate being a change amount of light radiation intensity per unit time.

[0038] When processing the optical signal segment, first, according to the change of the optical intensity of the optical signal when the circuit breaker is in normal operation, a preset threshold of the optical intensity change rate is set.

[0039] Each sampling point in the optical signal segment is traversed, and the optical intensity change amount (optical intensity of the current sampling point minus optical intensity of the previous sampling point) and the change time interval of the current sampling point and the previous sampling point are calculated. Then, the optical intensity change rate, that is, the ratio of the optical intensity change amount to the change time interval, is calculated. When the optical intensity change rate of continuous multiple sampling points exceeds the preset threshold, the signal segment is marked as a candidate optical radiation mutation segment.

[0040] The starting sampling point of the candidate optical radiation mutation segment is located, and the time coordinate thereof is extracted as the optical trigger level starting point. All candidate optical radiation mutation segments meeting the condition are processed to generate a set of optical trigger level starting points arranged in time sequence and attached with optical intensity information. For example, when local discharge occurs due to poor contact between the moving contact and the static contact of the circuit breaker, the optical intensity change rate of the optical signal exceeds the threshold, and the optical trigger level starting point is generated.

[0041] Step S124: identifying an electrical parameter mutation segment in which the current-voltage change rate exceeds a preset change rate threshold in the signal segment of the electrical signal sequence, recording the starting time mark of the electrical parameter mutation segment as an electrical trigger level starting point, and the current-voltage change rate is the change amount of the current or voltage per unit time.

[0042] When processing the electrical signal segment, a corresponding change rate preset threshold is set for the current and the voltage respectively.

[0043] For the current signal, each sampling point is traversed, the current change amount and the change time interval of the current sampling point and the previous sampling point are calculated, and then the current change rate is obtained. When the current change rate of continuous multiple sampling points exceeds the preset threshold, it is marked as a candidate current mutation segment.

[0044] For the voltage signal, the same method is used to calculate the voltage change rate, and when the voltage change rate of continuous multiple sampling points exceeds the preset threshold, it is marked as a candidate voltage mutation segment.

[0045] The candidate current mutation segment and the candidate voltage mutation segment are collectively referred to as an electrical parameter mutation segment, the starting sampling point thereof is located, and the time coordinate is extracted as an electrical trigger level starting point. All electrical parameter mutation segments meeting the condition are processed to generate a set of electrical trigger level starting points arranged in time sequence and attached with current or voltage information. For example, when a contact welding fault occurs in the circuit breaker, the current will abnormally change, the current change rate will exceed the threshold, and the electrical trigger level starting point will be generated.

[0046] Step S125: arranging the acoustic trigger level starting points, the optical trigger level starting points, and the electrical trigger level starting points in time sequence to generate a trigger level sequence, and each starting point in the trigger level sequence is attached with a corresponding signal type identifier.

[0047] All the trigger level starting points obtained in the collecting steps S122, S123, S124 are sorted according to their time coordinates to generate a trigger level sequence. Each starting point is marked with its corresponding signal type, i.e. sound signal, light signal or electric signal, in the sequence. For example, in the sequence, an electric trigger level starting point (due to current mutation) can be followed by a sound trigger level starting point (due to abnormal vibration of the operating mechanism) and then a light trigger level starting point (due to partial discharge), which are arranged in time sequence.

[0048] Step S126: When the signal change corresponding to the previous trigger level starting point triggers the signal change corresponding to the next trigger level starting point, a level response connection between the adjacent trigger level starting points is established.

[0049] For two adjacent trigger level starting points in the trigger level sequence, it is necessary to determine whether there is a signal response relationship between them.

[0050] Step S1261: Two adjacent trigger level starting points are selected from the trigger level sequence and marked as a higher-level trigger starting point and a lower-level trigger starting point, respectively.

[0051] In the trigger level sequence, two adjacent starting points are selected in sequence, the former being the higher-level trigger starting point and the latter being the lower-level trigger starting point.

[0052] Step S1262: The amplitude fluctuation curve of the signal segment corresponding to the higher-level trigger starting point before and after triggering is extracted, and the peak interval of amplitude fluctuation and the fluctuation duration period are determined. The peak interval is the time range in which the amplitude reaches the maximum value, and the fluctuation duration period is the time range from the start of signal change to the recovery to stability.

[0053] The amplitude fluctuation curve of the signal segment corresponding to the higher-level trigger starting point before and after triggering is extracted. By analyzing the curve, the time range in which the amplitude reaches the maximum value (peak interval) and the time range from the start of signal change to the recovery to stability (fluctuation duration period) are determined.

[0054] Step S1263: The amplitude fluctuation curve of the signal segment corresponding to the lower-level trigger starting point before and after triggering is extracted, and the peak interval of amplitude fluctuation and the fluctuation duration period are determined.

[0055] The amplitude fluctuation curve of the signal segment corresponding to the lower-level trigger starting point before and after triggering is extracted by the same method as step S1262, and the peak interval and the fluctuation duration period are determined.

[0056] Step S1264: Comparing the peak value interval of the upper-level signal with the peak value interval of the lower-level signal, calculating the overlap time length ratio of the overlap time length to the total time length of the peak value interval of the lower-level signal, the overlap time length being the intersection part of the two peak value intervals, and the total time length being the time span of the peak value interval of the lower-level signal.

[0057] Comparing the peak value intervals of the upper-level signal and the lower-level signal, finding the overlap part, calculating the overlap time length, and then dividing the overlap time length by the total time length of the peak value interval of the lower-level signal to obtain the overlap time length ratio.

[0058] Step S1265: Calculating the time interval between the upper-level trigger starting point and the lower-level trigger starting point, and determining the reasonable response time range in combination with the signal conduction characteristics of the device components.

[0059] Calculating the time interval between the upper-level trigger starting point and the lower-level trigger starting point, and then determining the reasonable response time range according to the signal conduction characteristics of each component inside the circuit breaker, such as the propagation characteristics of sound waves in air and solid, the conduction speed of electrical signals in conductors, etc.

[0060] Step S1266: When the overlap time length ratio exceeds the preset ratio threshold and the time interval is within the reasonable response time range, it is determined that the lower-level signal change is caused by the upper-level signal change.

[0061] A ratio threshold is set. If the overlap time length ratio exceeds the threshold and the time interval is within the reasonable response time range, it is determined that the lower-level signal change is caused by the upper-level signal change.

[0062] Step S1267: Establishing a hierarchical response connection between the upper-level trigger starting point and the lower-level trigger starting point, recording the response intensity coefficient of the connection and the signal conduction path identifier, the response intensity coefficient being positively correlated with the overlap time length ratio, and the signal conduction path identifier reflecting the propagation path of the signal inside the device.

[0063] When it is determined that there is a response relationship, a hierarchical response connection is established between the upper-level and lower-level trigger starting points. The response intensity coefficient is determined according to the overlap time length ratio. The higher the ratio, the greater the coefficient. The signal conduction path identifier is determined according to the signal type and the device structure, such as the sound wave conduction path from the operating mechanism to the arc extinguishing chamber.

[0064] Step S127: Forming a multi-modal signal correlation hierarchical chain containing a trigger hierarchical sequence, a hierarchical response delay parameter, and an amplitude linkage coefficient through a plurality of hierarchical response connections, the trigger hierarchical sequence reflecting the logical order of signal triggering, the hierarchical response delay parameter being the time difference between adjacent trigger starting points, and the amplitude linkage coefficient reflecting the correlation degree of the amplitude changes of the front and rear signals.

[0065] Integrate all levels of response connection, generate a multi-modal signal correlation hierarchy chain. The multi-modal signal correlation hierarchy chain records the trigger hierarchy sequence, the hierarchy response delay parameter (the time difference between adjacent trigger starting points) and the amplitude linkage coefficient (calculated by the ratio of the signal amplitude change of the upper level and the lower level, reflecting the correlation degree of the amplitude change of the two).

[0066] Step S130: Based on the multi-modal signal correlation hierarchy chain, extract the dynamic correlation feature set of each signal sequence in the hierarchy response process, which includes the amplitude linkage parameter, the phase synchronization parameter and the waveform distortion correlation parameter of the trigger signal and the hierarchy response signal.

[0067] From the multi-modal signal correlation hierarchy chain, extract the feature set that can reflect the dynamic correlation of each signal sequence in the hierarchy response process. The above features will be used for subsequent fault diagnosis.

[0068] Step S131: Extract the trigger signal segment and the hierarchy response signal segment corresponding to all levels of response connection from the multi-modal signal correlation hierarchy chain.

[0069] Step S1311: Analyze the structure of the multi-modal signal correlation hierarchy chain to determine the upper node and the lower node of each hierarchy response connection, the upper node corresponding to the trigger signal and the lower node corresponding to the hierarchy response signal.

[0070] Analyze the structure of the multi-modal signal correlation hierarchy chain to determine the upper node and the lower node in each hierarchy response connection, the upper node corresponding to the trigger signal and the lower node corresponding to the hierarchy response signal.

[0071] Step S1312: According to the time mark of the upper node, intercept the signal segment of the preset time length before and after the time mark as the trigger signal segment, which includes the stable segment before the trigger, the mutation segment at the trigger and the decay segment after the trigger.

[0072] According to the time mark of the upper node, intercept the signal of the preset time length before and after the time mark as the trigger signal segment, which includes the stable segment before the trigger, the mutation segment at the trigger and the decay segment after the trigger.

[0073] Step S1313: According to the time mark of the lower node, intercept the signal segment of the preset time length before and after the time mark as the hierarchy response signal segment, which includes the preparation segment before the response, the mutation segment at the response and the stable segment after the response.

[0074] According to the time mark of the lower node, intercept the signal of the preset time length before and after the time mark as the hierarchy response signal segment, which includes the preparation segment before the response, the mutation segment at the response and the stable segment after the response.

[0075] Step S1314: After the signal integrity check of the extracted trigger signal segment and the hierarchical response signal segment, the trigger signal segment and the hierarchical response signal segment belonging to the same hierarchical response connection are stored in association to generate a set of hierarchical associated signal segment pairs.

[0076] The extracted trigger signal segment and the hierarchical response signal segment are subjected to integrity check to ensure that the signals are not missing or damaged. Then the two signal segments belonging to the same hierarchical response connection are stored in association to generate a set of hierarchical associated signal segment pairs.

[0077] Step S1315: A corresponding hierarchical response connection identifier is added to each hierarchical associated signal segment pair, and the information of the hierarchical response connection identifier includes the superior node ID, the subordinate node ID and the connection serial number.

[0078] An identifier is added to each hierarchical associated signal segment pair, which contains the superior node ID, the subordinate node ID and the connection serial number, so as to be identified and processed subsequently.

[0079] Step S132: The ratio of the amplitude peak value of the trigger signal segment to the amplitude peak value of the hierarchical response signal segment in each hierarchical response connection is calculated, and the ratio is taken as the amplitude linkage parameter. The amplitude peak value is the maximum value of the amplitude in the signal segment.

[0080] For each hierarchical response connection, the maximum amplitude value (amplitude peak value) in the trigger signal segment and the maximum amplitude value in the hierarchical response signal segment are found, and the ratio of the two is calculated, which is taken as the amplitude linkage parameter.

[0081] Step S133: The trigger signal segment and the hierarchical response signal segment are subjected to frequency spectrum analysis, the phase spectrum curves of the trigger signal segment and the hierarchical response signal segment are extracted, and the phase difference absolute value of the phase spectrum curves in the same frequency interval is calculated.

[0082] The trigger signal segment and the hierarchical response signal segment are subjected to frequency spectrum analysis to obtain their respective phase spectrum curves. In the same frequency interval, the phase difference absolute value of the two phase spectrum curves is calculated.

[0083] Step S134: The proportion of the frequency interval with the phase difference absolute value less than the preset phase difference threshold value in the total analysis frequency interval is calculated, and the proportion is taken as the phase synchronization parameter.

[0084] A phase difference threshold value is set, and the proportion of the frequency interval with the phase difference absolute value less than the threshold value in the total analysis frequency interval is calculated, and the proportion is taken as the phase synchronization parameter.

[0085] Step S135: Comparing the waveform features of the trigger signal segment and the hierarchical response signal segment, extracting the number and degree of waveform distortion points, calculating the ratio of the number of distortion points and the correlation coefficient of the degree of distortion, and taking them as the waveform distortion correlation parameters. The waveform distortion point is a position deviating from the normal waveform trend.

[0086] Step S1351: Extracting the waveform features of the trigger signal segment, identifying the distortion points in the waveform, the distortion point is a sampling point deviating from the normal waveform trend, and the normal waveform trend is obtained by fitting the smooth part in the signal segment.

[0087] Extracting the waveform features of the trigger signal segment, obtaining the normal waveform trend by fitting the smooth part in the signal segment, and identifying the sampling points deviating from the trend as distortion points.

[0088] Step S1352: Counting the total number of distortion points in the trigger signal segment, and calculating the proportion of the distortion points in the total length of the trigger signal segment as the trigger distortion proportion, the total length of the trigger signal segment is the total number of sampling points contained in the trigger signal segment.

[0089] In the trigger signal segment, count all the identified distortion points to get the total number of distortion points. At the same time, count the total number of sampling points contained in the trigger signal segment, and divide the total number of distortion points by the total number of sampling points to get the trigger distortion proportion. The trigger distortion proportion can reflect the overall degree of waveform distortion in the trigger signal segment. The higher the proportion, the more serious the waveform distortion of the trigger signal segment. For example, if the trigger signal segment has a certain number of sampling points, and a certain number of them are identified as distortion points, then the trigger distortion proportion is the ratio of the two numbers.

[0090] Step S1353: Extracting the waveform features of the hierarchical response signal segment, identifying the distortion points in the waveform, and counting the total number of distortion points in the hierarchical response signal segment.

[0091] The same method as step S1351 is used to process the hierarchical response signal segment. First, fit the normal waveform trend of the smooth part in the hierarchical response signal segment, then identify the sampling points deviating from the trend as distortion points, and finally count the total number of distortion points in the hierarchical response signal segment. This process can accurately capture the abnormal changes of the waveform in the hierarchical response signal segment.

[0092] Step S1354: Calculate the ratio of the total number of distortion points in the hierarchical response signal segment and the total number of distortion points in the trigger signal segment as the ratio of the number of distortion points. The ratio greater than 1 indicates that the response signal distortion is more serious, and the ratio less than 1 indicates that the trigger signal distortion is more serious.

[0093] The total number of distortion points in the hierarchical response signal segment obtained in step S1353 is divided by the total number of distortion points in the trigger signal segment in step S1352 to obtain a ratio of the number of distortion points. The ratio can be used to intuitively compare the degree of waveform distortion of the trigger signal segment and the hierarchical response signal segment. When the ratio is greater than 1, it indicates that the hierarchical response signal segment has more distortion points, i.e., the response signal is more severely distorted. When the ratio is less than 1, it indicates that the trigger signal segment is more severely distorted. For example, if the total number of distortion points in the hierarchical response signal segment is several and the total number of distortion points in the trigger signal segment is several, the ratio of the two can clearly reflect which signal segment is more severely distorted.

[0094] Step S1355: Perform pairwise analysis on corresponding distortion points in the trigger signal segment and the hierarchical response signal segment, calculate the distortion degree difference value of each pair of distortion points, and the corresponding distortion points are distortion points that have a correlation in time.

[0095] In the trigger signal segment and the hierarchical response signal segment, find out the distortion points that have a correlation in time, i.e., the corresponding distortion points. For example, a distortion point that occurs at a certain time point in the trigger signal segment and a distortion point that occurs within a short time after the time point in the hierarchical response signal segment can be considered as a pair of corresponding distortion points. For each pair of corresponding distortion points, their distortion degrees are calculated respectively, and then the distortion degree of the distortion point in the hierarchical response signal segment is subtracted from the distortion degree of the corresponding distortion point in the trigger signal segment to obtain the distortion degree difference value of each pair of distortion points. The distortion degree can be determined according to the amplitude of deviation from the normal waveform trend, and the greater the deviation amplitude, the higher the distortion degree.

[0096] Step S1356: Calculate the correlation coefficient of the overall distortion degree according to the distortion degree difference value, and the closer the correlation coefficient is to the preset reference value, the more significant the correlation of the distortion degree is.

[0097] Collect the distortion degree difference values of all pairs of distortion points and calculate the correlation coefficient of the overall distortion degree using statistical analysis methods. For example, the correlation coefficient can be obtained by calculating the variance, covariance, etc. of these difference values. A reference value is preset, and the closer the correlation coefficient is to the reference value, the more significant the correlation of the distortion degree of the trigger signal segment and the hierarchical response signal segment is, i.e., the waveform distortion of the two has a strong correlation in degree; otherwise, the correlation is weak.

[0098] Step S136: Integrate the amplitude linkage parameter, the phase synchronization parameter, and the waveform distortion correlation parameter to generate a dynamic correlation feature set, and the integration needs to be grouped according to the hierarchical response connection, and each group of features corresponds to one connection.

[0099] The amplitude linkage parameter obtained in step S132, the phase synchronization parameter obtained in step S134, and the waveform distortion correlation parameter (including the ratio of the number of distortion points and the correlation coefficient of the distortion degree) obtained in step S135 are integrated. In the integration process, grouping is performed according to the hierarchical response connection, and each hierarchical response connection corresponds to a group of the above-mentioned parameters, thereby forming a dynamic correlation feature set. The purpose of this grouping is to ensure that the features of each hierarchical response connection can be clearly and independently presented, facilitating subsequent input into the fault feature evolution model for processing. For example, a group of features corresponding to a certain hierarchical response connection includes the amplitude linkage parameter, the phase synchronization parameter, the ratio of the number of distortion points, and the correlation coefficient of the distortion degree of the connection.

[0100] Step S140: inputting the dynamic correlation feature set into a preset fault feature evolution model to generate a fault feature identification sequence and a feature space diffusion trajectory.

[0101] After obtaining the dynamic correlation feature set, it is input into the pre-constructed fault feature evolution model, and the fault feature identification sequence reflecting the change process of the fault feature and the feature space diffusion trajectory of the propagation path of the fault feature in the device space are generated through the processing of the model.

[0102] Step S141: converting the dynamic correlation feature set into a feature matrix meeting the input requirements of the fault feature evolution model, wherein the row dimension of the feature matrix corresponds to different hierarchical response connections, the column dimension corresponds to different dynamic correlation features, and the matrix elements are quantized values of the feature parameters.

[0103] The parameters in the dynamic correlation feature set are quantized and converted into numerical form. Then, according to the requirements of the fault feature evolution model for the input data format, a feature matrix is constructed. The rows of the feature matrix represent different hierarchical response connections, each row corresponds to all the features of a hierarchical response connection; the columns represent different dynamic correlation features, such as amplitude linkage parameters, phase synchronization parameters, etc.; and each element in the matrix is the quantized value of the corresponding hierarchical response connection on the corresponding dynamic correlation feature. Through the above conversion, the dynamic correlation feature set can be input in a form recognizable by the model. For example, assuming that there are a number of hierarchical response connections and a number of dynamic correlation features, the feature matrix will be a matrix with the number of hierarchical response connections as the row number and the number of dynamic correlation features as the column number, and each element in the matrix is a specific quantized value.

[0104] Step S142: calling the feature mapping module of the fault feature evolution model, mapping and matching the feature matrix with a plurality of fault evolution mode templates stored in the fault feature evolution model, calculating the mode matching degree of the feature matrix with each fault evolution mode template, and selecting the fault evolution mode template with the highest mode matching degree as the target evolution template.

[0105] The fault feature evolution model includes a feature mapping module, which compares the input feature matrix with multiple pre-stored fault evolution pattern templates in the model. Each fault evolution pattern template corresponds to a specific fault evolution process and contains the typical change patterns of the dynamic correlation features during that process. The feature mapping module measures the similarity between the feature matrix and each template by calculating the pattern matching degree between the two. The pattern matching degree can be calculated based on similarity algorithms between features, such as cosine similarity and Euclidean distance. After the calculation is completed, the fault evolution pattern template with the highest pattern matching degree is selected as the target evolution template. This target evolution template best reflects the fault evolution corresponding to the current dynamic correlation feature set. For example, the model stores multiple evolution pattern templates for contact overheating faults and insulation aging faults. By calculating the matching degree between the feature matrix and these templates, the template with the highest matching degree with the contact overheating fault evolution pattern template is selected as the target evolution template.

[0106] Step S143: extracting the fault type identification sequence corresponding to the target evolution template and using it as a fault feature identification sequence, wherein the fault feature identification sequence reflects the characteristics of the fault at different stages from the initial stage to the development stage.

[0107] Each fault evolution pattern template corresponds to a fault type identification sequence, which consists of a series of identifiers, each of which represents the characteristics of the fault from its initial occurrence to a certain stage of gradual development. For example, for the target evolution template of a contact overheating fault, its corresponding fault characteristic identification sequence may include identifiers such as "initial abnormal increase in contact resistance," "slow increase in local temperature," "heat accumulation causing slight aging of surrounding insulation materials," "partial discharge," and "sharp temperature rise causing obvious faults." The fault type identification sequence corresponding to the target evolution template is extracted and used as the fault characteristic identification sequence for the current fault. This fault characteristic identification sequence can clearly show the characteristic changes of the fault at different stages.

[0108] Step S144: Based on the spatial distribution information contained in the feature matrix and in combination with the spatial diffusion model of the target evolution template, a diffusion trajectory of the fault feature in the equipment structure space is generated, and the spatial distribution information is associated with the signal acquisition position.

[0109] The spatial distribution information contained in the feature matrix is derived from the signal collection positions, i.e., the signals corresponding to the responses of different levels are collected from which positions of the device, and the above position information reflects the spatial positions where the fault features are likely to appear. The spatial diffusion model of the target evolution template describes the general propagation mode and path of the features of this type of fault in the spatial structure of the device. In combination with the information of the two aspects, the diffusion trajectory of the fault features in the spatial structure of the device is generated. For example, if the spatial distribution information shows that the initial signal collection position is at the moving and static contacts of the circuit breaker, and the spatial diffusion model of the target evolution template shows that the features of this type of fault will spread from the contacts to the surrounding insulation components, then the generated feature spatial diffusion trajectory will show a path starting from the moving and static contacts and gradually spreading to the nearby insulation pull rods, arc extinguishing chambers and other components.

[0110] Step S145: correlation verification is performed on the fault feature identification sequence and the feature spatial diffusion trajectory, so that each identification in the identification sequence forms a one-to-one correspondence with the corresponding position on the diffusion trajectory.

[0111] In order to ensure that the fault feature identification sequence and the feature spatial diffusion trajectory can accurately reflect the evolution process of the fault, correlation verification needs to be performed on the two. In the verification process, according to the time sequence of fault development and the spatial propagation law, each identification in the fault feature identification sequence is associated with a specific position on the feature spatial diffusion trajectory, to generate a one-to-one correspondence. For example, the identification of “initial abnormal increase of contact resistance” corresponds to the initial position of the moving and static contacts on the diffusion trajectory; the identification of “slow local temperature rise” corresponds to the position slightly diffused from the initial position on the trajectory; the identification of “local discharge phenomenon appears” corresponds to the position diffused to the vicinity of the insulation component on the trajectory. Through the above correlation verification, the time evolution features and the spatial propagation features of the fault can be verified with each other, and the reliability of the results is improved.

[0112] Step S150: determining the fault starting area and the fault influence boundary range of the high-voltage electrical device according to the fault feature identification sequence and the feature spatial diffusion trajectory.

[0113] The fault starting area and the boundary range affected by the fault of the high-voltage electrical device are comprehensively analyzed and determined by using the generated fault feature identification sequence and the feature spatial diffusion trajectory.

[0114] Step S151: analyzing the initial identification in the fault feature identification sequence to obtain the fault initial feature information corresponding to the initial identification, the initial identification being the earliest appearing identification in the fault feature identification sequence, reflecting the initial state of the fault occurrence.

[0115] The first identification appearing in the sequence of fault feature identifications is the initial identification, and analyzing the initial identification can obtain the feature information at the initial time of the fault occurrence. The above information includes the signal feature at the initial time of the fault, the possible abnormal state of the involved component, and the like. For example, the initial identification is "initial local electric field distortion of the insulating pull rod", and the corresponding initial feature information of the fault can include the abnormal condition of the electric field distribution near the insulating pull rod at the initial time, the weak fluctuation feature of the related acousto-optic-electric signal, and the like.

[0116] Step S152: In combination with the starting point coordinates of the feature space diffusion trajectory and the initial direction of the trajectory, the spatial coordinates of the fault starting point are located. The starting point coordinates are the first point of the feature space diffusion trajectory, and the initial direction of the trajectory is the tangent direction of the starting point of the trajectory.

[0117] The starting point coordinates of the feature space diffusion trajectory correspond to the spatial position where the fault feature first appears, and the initial direction of the trajectory indicates the initial diffusion direction of the fault feature. In combination with the initial feature information of the fault corresponding to the initial identification and the starting point coordinates and the initial direction of the trajectory, the specific spatial coordinates of the fault starting point are determined through spatial coordinate conversion and positioning algorithms. For example, the starting point coordinates of the trajectory are located at a specific position of the insulating pull rod, the initial direction of the trajectory points to the connection part of the pull rod and the contact, and in combination with the local electric field distortion information of the insulating pull rod reflected by the initial identification, it can be determined that the spatial coordinates of the fault starting point are at the specific position of the insulating pull rod.

[0118] Step S153: According to the fault feature identification intensity corresponding to each trajectory point on the feature space diffusion trajectory, the core area of the fault influence is determined. The core area is the area where the identification intensity is in a preset high intensity interval, and the identification intensity reflects the degree of prominence of the fault feature.

[0119] Each trajectory point on the feature space diffusion trajectory corresponds to a fault feature identification, and each identification has a corresponding identification intensity. The identification intensity reflects the degree of prominence of the fault feature at the trajectory point. The higher the intensity, the more obvious the fault feature. A high intensity interval is preset, and the area formed by the trajectory points on the trajectory whose identification intensity is in the interval is determined as the core area of the fault influence. For example, if the preset high intensity interval is a certain range, the points on the trajectory whose identification intensity is in the range are concentrated in the connection part of the insulating pull rod and the contact and the nearby area, and then the area is the core area of the fault influence.

[0120] Step S154: The identification intensity attenuation law in the trajectory extension process is analyzed, and when the identification intensity decreases to a preset boundary intensity, the corresponding coordinate point forms the fault influence boundary. The identification intensity attenuation law is obtained by fitting the intensity variation trend.

[0121] With the extension of the trajectory, the fault feature identification intensity will show a certain attenuation trend. By analyzing the identification intensity of each point on the trajectory, the attenuation law of the identification intensity is fitted, such as linear attenuation, exponential attenuation, etc. A boundary intensity value is preset, and when the identification intensity attenuates to the boundary intensity along the extension direction of the trajectory, the contour formed by the corresponding trajectory points is the fault influence boundary. For example, the identification intensity presents an exponential attenuation law, and when it attenuates to the preset boundary intensity, the closed curve formed by connecting the corresponding coordinate points constitutes the fault influence boundary.

[0122] Step S155: Calculate the geometric center coordinates of the fault starting area and the maximum extension distance of the influence boundary, and generate the fault location information containing the center coordinates, boundary contour and range size. The geometric center coordinates are the barycentric position of the core area.

[0123] By geometric calculation method, the geometric center coordinates of the fault starting area are determined, which are the barycentric position of the fault core area, which can be obtained by calculating the average value of all trajectory point coordinates in the core area. At the same time, the distance from each point on the fault influence boundary to the geometric center coordinates is calculated, and the maximum distance is the maximum extension distance of the influence boundary. The geometric center coordinates, the contour shape of the fault influence boundary and the range size (such as the maximum extension distance, the area of the region surrounded by the boundary, etc.) are integrated to generate complete fault location information. For example, the geometric center coordinates of the fault starting area are a certain specific coordinate value, the maximum extension distance of the influence boundary is a certain length, and the boundary contour is an irregular polygon. The above information collectively constitutes the fault location information.

[0124] Step S156: Spatially map the fault location information with the three-dimensional structure model of the high-voltage electrical equipment to obtain the specific location of the fault in the equipment entity structure and the involved component information, and the spatial mapping is based on the coordinate system conversion of the equipment.

[0125] The high-voltage electrical equipment has a corresponding three-dimensional structure model, which contains the detailed structure and spatial coordinate information of each component of the equipment. Spatial mapping of the generated fault location information with the three-dimensional structure model, i.e. through the coordinate system conversion of the equipment, converts the coordinates of the fault location information into the coordinates in the three-dimensional structure model. Through the above mapping, the specific location of the fault in the equipment entity structure and the equipment components involved in the location can be determined. For example, after spatial mapping, it is found that the fault starting area corresponds to the middle segment position of the insulating pull rod in the three-dimensional structure model, and the involved components include the insulating pull rod body and the contact component connected thereto.

[0126] In order to enable the fault feature evolution model to accurately process the dynamic associated feature set and generate reliable results, the fault feature evolution model needs to be trained in advance.

[0127] Step S211: Collecting multi-modal signal data of high-voltage electrical equipment under different fault states and corresponding fault diagnosis results to construct a training data set.

[0128] A wide range of sound signal sequences, light signal sequences, and electrical signal sequences of high-voltage electrical equipment under various fault states (such as contact welding, insulation aging, and operating mechanism jamming) are collected, and detailed diagnosis results corresponding to these faults are collected, including fault type, fault development stage, fault location, and other information. These multi-modal signal data and corresponding diagnosis results are sorted together to construct a training data set. During the collection process, the diversity and representativeness of the data need to be ensured, covering different models, different operating periods of equipment under different fault conditions, to improve the generalization ability of the model.

[0129] Step S212: Processing the multi-modal signal data in the training data set according to the method of steps S120 to S136 to generate a dynamic correlation feature set for training.

[0130] The same method as processing actual fault signals is used to perform cross-modal signal hierarchical correlation processing and dynamic correlation feature extraction on the multi-modal signal data in the training data set to generate a dynamic correlation feature set for training. These training features have the same data format and feature type as the dynamic correlation feature set generated in actual application, so as to be used for model training.

[0131] Step S213: According to the fault diagnosis results in the training data set, constructing corresponding fault feature identification sequences and feature space diffusion trajectories as training labels.

[0132] According to the diagnosis results of each fault case in the training data set, artificial or automatic labeling is used to construct corresponding fault feature identification sequences and feature space diffusion trajectories as training labels. The training labels need to accurately reflect the feature evolution process and spatial diffusion of the fault case, and correspond to the dynamic correlation feature set for training.

[0133] Step S214: Inputting the dynamic correlation feature set for training and the corresponding training labels into the initial fault feature evolution model, adjusting the model parameters through the backpropagation algorithm, so that the fault feature identification sequence and the feature space diffusion trajectory output by the model are within the preset range of the training labels.

[0134] The initial fault feature evolution model is an untrained model with initial parameter settings. The training dynamic correlation feature set is input into the initial fault feature evolution model, and the fault feature evolution model outputs the corresponding fault feature identification sequence and feature space diffusion trajectory. These output results are compared with the corresponding training labels, and the error between them is calculated. Then, using the backpropagation algorithm, the parameters of the fault feature evolution model (such as the weight coefficients in the feature mapping module, the parameters in the space diffusion model, etc.) are adjusted according to the error size. Repeat this process until the error between the model output and the training label is reduced to within the preset range. At this time, the fault feature evolution model training is completed, and it can better process the input feature set and generate results that meet the actual situation.

[0135] During the model parameter adjustment process, fine-tuning is required on a module-by-module basis. For the feature mapping module, it contains multiple sub-layers for feature matching, and each sub-layer has corresponding weight coefficients. When the calculated error is large, start from the output layer and reverse-propagate the error to determine the degree of influence of each layer's weight coefficient. For example, if the matching degree between the feature matrix and the fault evolution pattern template is low, resulting in an error, the weight of the sub-layer responsible for pattern comparison in the feature mapping module can be adjusted to enhance the weight proportion of high-matching-degree templates and reduce the influence of low-matching-degree templates.

[0136] For the space diffusion model, its parameters include factors that describe fault diffusion speed, direction preference, etc. When the feature space diffusion trajectory output by the model deviates significantly from the actual trajectory in the training label, the backpropagation algorithm is used to analyze whether the deviation is mainly due to estimation error of diffusion speed or deviation of direction judgment, and then adjust the corresponding parameters. For example, if the trajectory extends too slowly, the diffusion speed factor can be increased; if the trajectory direction deviates from the actual direction, the direction preference parameter can be corrected to make the model more consistent with the actual fault diffusion law.

[0137] During the entire training process, the appropriate number of iterations needs to be set. Each iteration will use all the training dynamic correlation feature sets for a complete forward calculation and backward parameter adjustment. At the same time, to avoid overfitting of the model, the training data can be divided into a training set and a validation set. The model performance is verified using the validation set after each iteration. If the error on the training set continues to decrease but the error on the validation set starts to rise, stop the current parameter adjustment and use the model parameters when the validation set performance is best to ensure that the fault feature evolution model also maintains good generalization ability on new dynamic correlation feature sets.

[0138] When the model training is completed, it also needs to be tested for stability. A part of the dynamic correlation feature set not involved in the training is selected as the test set, which is input into the trained fault feature evolution model to observe the stability of the fault feature identification sequence and feature space diffusion trajectory output by the fault feature evolution model. If the fluctuation of the model output result is within the preset fluctuation range after multiple inputs of the same test set, it indicates that the fault feature evolution model training is stable and can be used for subsequent fault diagnosis process.

[0139] Thus, the trained fault feature evolution model is encapsulated and stored for calling in the actual diagnosis process.

[0140] The trained fault feature evolution model needs to be encapsulated to generate a module that can be independently called. During the encapsulation process, the input and output interfaces of the fault feature evolution model can be specified. The input interface needs to match the format of the dynamic correlation feature set in the actual diagnosis process, and the output interface specifies the output format of the fault feature identification sequence and feature space diffusion trajectory. At the same time, necessary metadata is added to the model, including model training time, training data size, model performance indicators (such as error range, accuracy, etc.), for subsequent maintenance and version management.

[0141] The encapsulated fault feature evolution model is stored in a special model library, using a distributed storage method to ensure that it can respond quickly when called by multiple diagnosis terminals. During storage, the model file is encrypted, and only authorized diagnosis systems can access and call it to prevent the model from being illegally tampered with or stolen. In addition, a model version update mechanism is established. When new fault cases and corresponding dynamic correlation feature sets are collected, incremental training can be performed based on the original model to generate a new model version and replace the old version, so that the model always maintains the diagnosis ability for new faults.

[0142] Figure 2 A schematic diagram of exemplary hardware and software components of a fault location system 100 based on multi-modal acoustic-optical-electrical signal collaborative diagnosis that can implement the idea of the present application is shown. For example, the processor 120 can be used in the fault location system 100 based on multi-modal acoustic-optical-electrical signal collaborative diagnosis, and is used to execute the functions in the present application.

[0143] The fault location system 100 based on multi-modal acoustic-optical-electrical signal collaborative diagnosis can be a general server or a special-purpose server, both of which can be used to implement the fault location method based on multi-modal acoustic-optical-electrical signal collaborative diagnosis of the present application. Although only one server is shown in the present application, for the sake of convenience, the functions described in the present application can be implemented in a distributed manner on multiple similar platforms to balance the processing load.

[0144] For example, the multi-modal acoustic-optical-electrical signal collaborative diagnosis based fault locating system 100 can include a network port 110 connected to a network, one or more processors 120 for executing program instructions, a communication bus 130, and different forms of storage media 140, such as a disk, a ROM, or a RAM, or any combination thereof. Exemplarily, the multi-modal acoustic-optical-electrical signal collaborative diagnosis based fault locating system 100 can also include program instructions stored in a ROM, a RAM, or other types of non-transitory storage media, or any combination thereof. The method of the present application can be implemented according to these program instructions. The multi-modal acoustic-optical-electrical signal collaborative diagnosis based fault locating system 100 also includes an I / O interface 150 between the computer and other input and output devices.

[0145] For the convenience of illustration, only one processor is described in the multi-modal acoustic-optical-electrical signal collaborative diagnosis based fault locating system 100. However, it should be noted that the multi-modal acoustic-optical-electrical signal collaborative diagnosis based fault locating system 100 in the present application can also include multiple processors, and therefore the steps performed by one processor described in the present application can also be jointly performed or individually performed by multiple processors. For example, if the processor of the multi-modal acoustic-optical-electrical signal collaborative diagnosis based fault locating system 100 performs steps A and B, it should be understood that steps A and B can also be jointly performed by two different processors or individually performed in one processor. For example, a first processor performs step A, a second processor performs step B, or the first processor and the second processor jointly perform steps A and B.

[0146] In addition, the present application also provides a readable storage medium, in which computer executable instructions are pre-set, and when the processor executes the computer executable instructions, the multi-modal acoustic-optical-electrical signal collaborative diagnosis based fault locating method is realized.

[0147] It should be noted that, in order to simplify the expression of the present application and to help the understanding of one or more embodiments of the present application, in the foregoing description of the embodiments of the present application, various features are sometimes combined into one embodiment, figure or description thereof.

Claims

1. A fault location method based on multimodal acoustic, optical and electrical signal collaborative diagnosis, characterized in that: The method comprises: Acquiring acoustic signal sequences, optical signal sequences, and electrical signal sequences from high-voltage electrical equipment in operation. The acoustic signal sequence includes information on changes in acoustic waveforms caused by vibrations in internal components of the equipment. The optical signal sequence includes information on changes in optical radiation intensity caused by partial discharges on the equipment surface. The electrical signal sequence includes information on instantaneous fluctuations in current and voltage in the equipment circuit. Performing cross-modal signal hierarchical association processing on the acoustic signal sequence, optical signal sequence, and electrical signal sequence, establishing a trigger response hierarchical relationship between different signal sequences, and generating a multimodal signal association hierarchical chain; Extracting a dynamic correlation feature set of each signal sequence in a hierarchical response process based on the multimodal signal correlation hierarchical chain, wherein the dynamic correlation feature set includes amplitude linkage parameters, phase synchronization parameters, and waveform distortion correlation parameters of the trigger signal and the hierarchical response signal; Inputting the dynamic correlation feature set into a preset fault feature evolution model to generate a fault feature identification sequence and a feature space diffusion trajectory; Determine the fault starting area and fault impact boundary range of the high-voltage electrical equipment based on the fault feature identification sequence and feature space diffusion trajectory; Determining the fault starting area and fault impact boundary range of the high-voltage electrical equipment based on the fault feature identification sequence and feature space diffusion trajectory includes: Parsing the initial identifier in the fault feature identifier sequence to obtain the initial fault feature information corresponding to the initial identifier, where the initial identifier is the earliest identifier in the fault feature identifier sequence and reflects the initial state of the fault; The spatial coordinates of the fault start are located by combining the starting point coordinates and the initial direction of the feature space diffusion trajectory, where the starting point coordinates are the first point of the feature space diffusion trajectory, and the initial direction of the trajectory is the tangent direction of the starting point of the trajectory; Determine the core area affected by the fault based on the fault feature identification strength corresponding to each trajectory point on the feature space diffusion trajectory, where the core area is an area where the identification strength is within a preset high intensity range, and the identification strength reflects the significance of the fault feature; Analyze the attenuation law of the marker strength during the trajectory extension process. When the marker strength drops to the preset boundary strength, the corresponding coordinate point constitutes the fault impact boundary. The marker strength attenuation law is obtained by fitting the strength change trend. Calculate the geometric center coordinates of the fault starting area and the maximum extension distance of the affected boundary, and generate fault location information including the center coordinates, boundary outline, and range size. The geometric center coordinates are the center of gravity of the core area. The fault location information is spatially mapped with the three-dimensional structural model of the high-voltage electrical equipment to obtain the specific location of the fault in the physical structure of the equipment and the information of the components involved. The spatial mapping is achieved based on the coordinate system conversion of the equipment.

2. The fault location method based on multimodal acoustic, optical and electrical signal collaborative diagnosis according to claim 1, characterized in that: The performing cross-modal signal hierarchical association processing on the acoustic signal sequence, the optical signal sequence, and the electrical signal sequence, establishing a trigger response hierarchical relationship between different signal sequences, and generating a multimodal signal association hierarchical chain includes: Intercepting signal segments of the acoustic signal sequence, optical signal sequence, and electrical signal sequence within the same monitoring period, wherein the signal segments include continuous signal fluctuation units and signal stable units, wherein the signal fluctuation units are signal segments with fluctuations in amplitude or frequency, and the signal stable units are signal segments with stable amplitude and frequency; Identifying a sudden acoustic wave segment in a signal segment of the acoustic signal sequence whose amplitude change rate exceeds a preset change rate threshold, and recording a starting time mark of the sudden acoustic wave segment as the starting point of the acoustic trigger level, wherein the amplitude change rate is the amount of change in amplitude per unit time; Identify a light radiation mutation segment in a signal segment of the light signal sequence, in which the light intensity change rate exceeds a preset change rate threshold, and record a start time mark of the light radiation mutation segment as a starting point of the light trigger level, wherein the light intensity change rate is a change in light radiation intensity per unit time; Identifying an electrical parameter mutation segment in which the current or voltage change rate exceeds a preset change rate threshold in a signal segment of the electrical signal sequence, and recording a start time mark of the electrical parameter mutation segment as a starting point of the electrical triggering level, wherein the current or voltage change rate is a change in current or voltage per unit time; Arranging the sound trigger level starting point, the light trigger level starting point, and the electrical trigger level starting point in chronological order to generate a trigger level sequence, wherein each starting point in the trigger level sequence is accompanied by a corresponding signal type identifier; Analyze the signal response relationship between adjacent trigger level starting points in the trigger level sequence, and establish a level response connection between the adjacent trigger level starting points when the signal change corresponding to the previous trigger level starting point triggers the signal change corresponding to the next trigger level starting point; A multimodal signal association hierarchical chain is formed through multiple hierarchical response connections, which includes a trigger hierarchical sequence, a hierarchical response delay parameter, and an amplitude linkage coefficient. The trigger hierarchical sequence reflects the sequence logic of each signal trigger, the hierarchical response delay parameter is the time difference between adjacent trigger starting points, and the amplitude linkage coefficient reflects the degree of correlation between the amplitude changes of the previous and subsequent signals.

3. The fault location method based on multimodal acoustic, optical and electrical signal collaborative diagnosis according to claim 2, characterized in that: The step of identifying a sound wave mutation segment in a signal segment of the sound signal sequence in which the amplitude change rate exceeds a preset change rate threshold, and recording a start time mark of the sound wave mutation segment as a sound trigger level starting point, includes: Traversing each signal sampling point in the signal segment of the acoustic signal sequence, calculating an amplitude change and a change time interval between a current sampling point and a previous sampling point, wherein the amplitude change is the difference between the amplitude of the current sampling point and the amplitude of the previous sampling point, and the change time interval is the time difference between the two sampling points; Calculating an amplitude change rate based on the amplitude change amount and the change time interval, wherein the amplitude change rate is a ratio of the amplitude change amount to the change time interval, wherein a positive value of the ratio indicates an increase in the amplitude, and a negative value indicates a decrease in the amplitude; When the amplitude change rate of a plurality of consecutive sampling points exceeds a preset change rate threshold, marking the signal segment consisting of the plurality of consecutive sampling points as a candidate acoustic wave mutation segment; Locating the starting sampling point of the candidate sound wave mutation segment and extracting the time coordinate of the starting sampling point in the sound signal segment; Taking the time coordinate as the sound triggering level starting point, repeat the above operation for all qualified candidate sound wave mutation segments in the signal segment of the sound signal sequence to generate a sound triggering level starting point set, in which each starting point is arranged in chronological order and is accompanied by corresponding signal amplitude information.

4. The fault location method based on multimodal acoustic, optical and electrical signal collaborative diagnosis according to claim 2, characterized in that: The analyzing the signal response relationship between adjacent trigger level starting points in the trigger level sequence, and establishing a level response connection between adjacent trigger level starting points when a signal change corresponding to a previous trigger level starting point triggers a signal change corresponding to a subsequent trigger level starting point, includes: Select two adjacent trigger level starting points from the trigger level sequence and mark them as the upper trigger starting point and the lower trigger starting point respectively; Extract the amplitude fluctuation curve of the signal segment corresponding to the upper trigger starting point before and after the trigger, and determine the peak interval and fluctuation duration of the amplitude fluctuation. The peak interval is the time range when the amplitude reaches the maximum value, and the fluctuation duration is the time range from the beginning of the signal change to the restoration of stability; Extract the amplitude fluctuation curve of the signal segment corresponding to the lower-level trigger starting point before and after the trigger, and determine the peak range and duration of the amplitude fluctuation; Compare the peak interval of the superior signal with the peak interval of the subordinate signal, and calculate the ratio of the overlapping duration to the total duration of the peak interval of the subordinate signal, where the overlapping duration is the intersection of the two peak intervals, and the total duration is the time span of the peak interval of the subordinate signal; Calculate the time interval between the upper trigger start point and the lower trigger start point, and determine a reasonable response time range based on the signal conduction characteristics of the device components; When the overlap duration ratio exceeds a preset ratio threshold and the time interval is within a reasonable response time range, it is determined that the change in the lower-level signal is caused by the change in the upper-level signal; A hierarchical response connection is established between the upper trigger starting point and the lower trigger starting point, and the response strength coefficient and signal conduction path identifier of the connection are recorded. The response strength coefficient is positively correlated with the overlap duration ratio, and the signal conduction path identifier reflects the propagation path of the signal inside the device.

5. The fault location method based on multimodal acoustic, optical and electrical signal collaborative diagnosis according to claim 1, characterized in that: The extracting of a dynamic correlation feature set of each signal sequence in a hierarchical response process based on the multimodal signal correlation hierarchical chain includes: Extracting trigger signal segments and hierarchical response signal segments corresponding to all hierarchical response connections from the multimodal signal association hierarchical chain; Calculate the ratio of the peak amplitude of the trigger signal segment to the peak amplitude of the level response signal segment in each level response connection, and use the ratio as the amplitude linkage parameter, where the peak amplitude is the maximum amplitude in the signal segment; Perform spectrum analysis on the trigger signal segment and the hierarchical response signal segment, extract the phase spectrum curves of the trigger signal segment and the hierarchical response signal segment, and calculate the absolute value of the phase difference of the phase spectrum curves in the same frequency range; Counting the proportion of frequency intervals where the absolute value of the phase difference is less than a preset phase difference threshold to the total analyzed frequency intervals, and using the proportion as a phase synchronization parameter; Comparing the waveform features of the trigger signal segment and the hierarchical response signal segment, extracting the number and degree of waveform distortion points, calculating the ratio of the number of distortion points and the correlation coefficient of the distortion degree, and using it as the waveform distortion correlation parameter; the waveform distortion point is the position that deviates from the normal waveform trend; The amplitude linkage parameters, phase synchronization parameters and waveform distortion correlation parameters are integrated to generate a dynamic correlation feature set. During the integration, the features are grouped according to the hierarchical response connections, and each set of features corresponds to one connection.

6. The fault location method based on multimodal acoustic, optical and electrical signal collaborative diagnosis according to claim 5, characterized in that: The extracting the trigger signal segments and the hierarchical response signal segments corresponding to all hierarchical response connections from the multimodal signal association hierarchical chain includes: Parsing the structure of the multimodal signal association hierarchical chain to determine the upper node and lower node of each hierarchical response connection, wherein the upper node corresponds to the trigger signal and the lower node corresponds to the hierarchical response signal; According to the time stamp of the upper-level node, a signal segment of a preset length before and after the time stamp is intercepted as a trigger signal segment, wherein the trigger signal segment includes a stable segment before the trigger, a sudden change segment during the trigger, and a decay segment after the trigger; According to the time stamp of the lower-level node, a signal segment of a preset length before and after the time stamp is intercepted as a hierarchical response signal segment, wherein the hierarchical response signal segment includes a preparatory segment before the response, a sudden change segment during the response, and a stable segment after the response; After performing a signal integrity check on the extracted trigger signal segments and hierarchical response signal segments, the trigger signal segments and hierarchical response signal segments belonging to the same hierarchical response connection are associated and stored to generate a hierarchical associated signal segment pair set; A corresponding hierarchical response connection identifier is added for each hierarchical association signal segment pair, wherein information of the hierarchical response connection identifier includes an upper-level node ID, a lower-level node ID, and a connection sequence number.

7. The fault location method based on multimodal acoustic, optical and electrical signal collaborative diagnosis according to claim 5, characterized in that: The comparing the waveform features of the trigger signal segment and the hierarchical response signal segment, extracting the number of waveform distortion points and the degree of distortion, and calculating the correlation coefficient of the ratio of the number of distortion points and the degree of distortion, includes: Extract waveform features from the trigger signal segment to identify distortion points in the waveform. The distortion points are sampling points that deviate from the normal waveform trend, which is obtained by fitting the stable portion of the signal segment. Counting the total number of distortion points in the trigger signal segment and calculating the ratio of the distortion points to the total length of the trigger signal segment as the trigger distortion ratio, where the total length of the trigger signal segment is the total number of sampling points included in the trigger signal segment; Extract waveform features of the hierarchical response signal segment, identify distortion points in the waveform, and count the total number of distortion points in the hierarchical response signal segment; Calculating a ratio of the total number of distortion points in the hierarchical response signal segment to the total number of distortion points in the trigger signal segment as the ratio of the number of distortion points, wherein a ratio greater than 1 indicates that the response signal is more severely distorted, and a ratio less than 1 indicates that the trigger signal is more severely distorted; Perform pairwise analysis on the corresponding distortion points in the trigger signal segment and the hierarchical response signal segment, and calculate the difference in distortion degree of each pair of distortion points. The corresponding distortion points are the distortion points that are temporally correlated. The correlation coefficient of the overall distortion degree is calculated according to the distortion degree difference. The closer the correlation coefficient is to a preset reference value, the more significant the distortion degree correlation is.

8. The fault location method based on multimodal acoustic, optical and electrical signal collaborative diagnosis according to claim 1, characterized in that: Inputting the dynamic correlation feature set into a preset fault feature evolution model to generate a fault feature identification sequence and a feature space diffusion trajectory includes: Converting the dynamic correlation feature set into a feature matrix that meets the input requirements of the fault feature evolution model, wherein the row dimension of the feature matrix corresponds to different hierarchical response connections, the column dimension corresponds to different dynamic correlation features, and the matrix elements are the quantized values ​​of the feature parameters; Calling the feature mapping module of the fault feature evolution model, mapping and matching the feature matrix with multiple fault evolution pattern templates stored in the fault feature evolution model, calculating the pattern matching degree between the feature matrix and each fault evolution pattern template, and selecting the fault evolution pattern template with the highest pattern matching degree as the target evolution template; Extracting the fault type identification sequence corresponding to the target evolution template and using it as a fault feature identification sequence, wherein the fault feature identification sequence reflects the characteristics of the fault at different stages from initialization to development; Based on the spatial distribution information contained in the feature matrix and combined with the spatial diffusion model of the target evolution template, a diffusion trajectory of the fault feature in the device structure space is generated, and the spatial distribution information is associated with the signal acquisition position; An association check is performed on the fault feature identifier sequence and the feature space diffusion trajectory, so that each fault feature identifier in the fault feature identifier sequence forms a one-to-one correspondence with a corresponding position on the feature space diffusion trajectory.

9. A fault location system based on multimodal acoustic, optical and electrical signal collaborative diagnosis, characterized in that: It includes a processor and a memory, the memory is connected to the processor, the memory is used to store programs, instructions or codes, and the processor is used to execute the programs, instructions or codes in the memory to implement the fault location method based on multimodal acoustic, optical and electrical signal collaborative diagnosis as described in any one of claims 1 to 8.

Images