Partial discharge signal multi-source cooperative detection method and system based on time-frequency fusion analysis

Abstract

The invention discloses a partial discharge signal multi-source cooperative detection method and system based on time-frequency fusion analysis, and relates to the technical field of multi-source cooperative detection, and the method comprises the following steps: obtaining partial discharge original signals of an electric signal channel and an ultrahigh frequency channel, respectively carrying out the time-frequency transformation, and generating a multi-source time-frequency characteristic spectrum; extracting high-frequency noise modes of different channels under a non-partial discharge condition based on the multi-source time-frequency characteristic spectrum, and constructing a cross-source interference fingerprint database; based on the cross-source interference fingerprint database, carrying out frequency-band-by-band difference mapping calculation on actually acquired multi-source time-frequency characteristics to generate a difference characteristic matrix; and introducing reverse consistency constraint into the differential feature matrix, carrying out weighted fusion to obtain multi-source coupling time-frequency feature mapping, and outputting a detection result of the partial discharge signal. According to the method, the multi-source time-frequency characteristic spectrum is constructed to carry out band-by-band differential mapping on the actual signal, so that cross-source noise interference is inhibited, and the accuracy and robustness of partial discharge detection in a strong noise environment are improved.

Description

Technical Field

[0001] This invention relates to the field of multi-source collaborative detection technology, and more specifically, to a method and system for multi-source collaborative detection of partial discharge signals based on time-frequency fusion analysis. Background Technology

[0002] Currently, partial discharge detection mainly relies on multi-channel acquisition of electrical and ultra-high frequency signals, extracting signal features through time-frequency analysis methods such as Fourier transform and wavelet transform, and then performing multi-channel fusion to achieve detection. However, when strong noise exists simultaneously in multiple channels, different noise sources may exhibit similar patterns in the high-frequency band, causing feature cross-contamination. Traditional feature extraction and fusion methods cannot distinguish these cross-source noise similarities, easily amplifying errors and leading to unreliable detection results. Furthermore, existing technologies lack processing methods for the synergistic effects of multi-source noise, making it difficult to maintain high-precision identification of partial discharge signals in complex noise environments.

[0003] The above-disclosed technical solutions have at least the following technical problems: noise in different channels may exhibit similar patterns in the high-frequency band (such as overlap of radio interference and partial discharge UHF signals), resulting in feature mixing.

[0004] To address the above problems, this invention proposes a solution. Summary of the Invention

[0005] To overcome the aforementioned deficiencies of the prior art, embodiments of the present invention provide a multi-source collaborative detection method and system for partial discharge signals based on time-frequency fusion analysis. By performing frequency-band differential mapping on the actual acquired signals, multi-channel collaborative feature enhancement and cross-source noise suppression are achieved, thereby solving the problems of easy interference, low detection accuracy, and insufficient robustness of partial discharge signals in strong noise environments in the prior art.

[0006] To achieve the above objectives, the present invention provides the following technical solution: On the one hand, the multi-source collaborative detection method for partial discharge signals based on time-frequency fusion analysis includes the following steps: acquiring the original partial discharge signals of the electrical signal channel and the ultra-high frequency channel, performing time-frequency transformation on each, and generating multi-source time-frequency feature maps; extracting high-frequency noise patterns of different channels under non-partial discharge conditions based on the multi-source time-frequency feature maps, and constructing a cross-source interference fingerprint database; performing band-by-band differential mapping calculation on the actually acquired multi-source time-frequency features based on the cross-source interference fingerprint database, and generating a differential feature matrix; introducing reverse consistency constraints into the differential feature matrix, and performing weighted fusion to obtain the multi-source coupled time-frequency feature mapping, and outputting the detection results of the partial discharge signal.

[0007] In a preferred embodiment, the acquisition of the original partial discharge signals of the electrical signal channel and the ultra-high frequency channel, and the subsequent time-frequency transformation to generate a multi-source time-frequency feature map, specifically involves: acquiring the original partial discharge signals of the electrical signal channel and the ultra-high frequency channel, and performing differential operations with pre-stored partial discharge-free background signals to obtain the residual signals corresponding to each channel; segmenting and windowing the residual signals, and generating an initial time spectrum through short-time Fourier transform; applying a differential enhancement operator to the initial time spectrum to highlight the burst characteristics of partial discharge; refining the frequency band of the differentially enhanced time spectrum through multi-scale wavelet packet decomposition, and performing energy normalization processing on the time-frequency results of different channels to generate a multi-source time-frequency feature map.

[0008] In a preferred embodiment, the step of extracting high-frequency noise patterns of different channels under non-partial discharge conditions based on multi-source time-frequency feature maps specifically involves: selecting high-frequency intervals based on multi-source time-frequency feature maps, calculating the energy coupling degree function of different channels in each high-frequency band, identifying frequency bands that are simultaneously enhanced in multiple channels as cross-source noise resonance bands; performing local binary pattern texture encoding on the time-frequency spectrum within the resonance bands to extract the morphological distribution features of the noise; and performing robust statistics on the morphological distribution features based on the median absolute deviation method to form stable high-frequency noise patterns.

[0009] In a preferred embodiment, the step of selecting high-frequency intervals based on multi-source time-frequency feature maps specifically involves: calculating the energy difference sequence between the electrical signal channel and the ultra-high frequency channel in the frequency dimension based on the multi-source time-frequency feature maps, and selecting frequency bands with energy differences less than a preset threshold as first candidate intervals; calculating the cross-source phase consistency time variance of the first candidate intervals, and marking frequency bands with unstable phases and continuous existence within a preset time as second candidate intervals; taking the intersection of the first candidate intervals and the second candidate intervals, and obtaining the high-frequency intervals through frequency domain morphological closing operations.

[0010] In a preferred embodiment, the construction of the cross-source interference fingerprint database specifically involves: based on the extracted high-frequency noise patterns, marking the corresponding frequency bands as candidate interference intervals on the time-frequency feature maps of each channel; extracting local statistical features of each channel within each candidate interference interval, including average energy, energy variance, time-frequency texture descriptor, and local gradient trend; fusing the features of each channel by frequency band to form a frequency band-level cross-source interference fingerprint vector; storing the interference fingerprint vector and its boundary, bandwidth, and duration information in the fingerprint database, and establishing an index to support fast matching and retrieval; and dynamically iteratively optimizing the fingerprint database by updating the statistical features or adding new frequency band fingerprints when new candidate high-frequency interference interval data is available, through matching with the fingerprint database.

[0011] In a preferred embodiment, the step of performing band-by-band differential mapping calculation on the actually collected multi-source time-frequency features based on the cross-source interference fingerprint database to generate a differential feature matrix specifically involves: mapping the actually collected multi-source time-frequency features to candidate frequency band regions based on the high-frequency noise mode frequency bands in the cross-source interference fingerprint database; standardizing the time-frequency energy of each channel within the candidate frequency band, using the statistical mean and variance of the fingerprint database as a benchmark; calculating the difference between the standardized signal and the fingerprint database features to form a channel-level differential vector; and integrating the channel differential vectors into a differential feature matrix according to frequency bands.

[0012] In a preferred embodiment, the step of introducing a reverse consistency constraint into the differential feature matrix and performing weighted fusion to obtain a multi-source coupled time-frequency feature map specifically involves: calculating the channel-to-channel reverse consistency index for each frequency band and time point of the differential feature matrix and mapping it as a weight; performing weighted fusion of the differential feature matrix according to the weights to generate a fused multi-source coupled feature matrix; and combining the results of fusing all frequency bands and time points into a matrix form to form a multi-source coupled time-frequency feature map.

[0013] In a preferred embodiment, the detection result of the output partial discharge signal specifically involves: setting a discrimination threshold for each frequency band and time point of the multi-source coupled time-frequency feature mapping matrix, judging and marking them as partial discharge candidate points; performing connectivity analysis on the marked candidate points in the frequency and time dimensions, retaining continuous regions that meet the minimum bandwidth and minimum duration, and eliminating isolated points or transient noise; verifying the consistency of continuous regions on each channel, retaining regions with enhanced cross-source coupling, and improving the reliability and noise resistance of detection; and outputting the finally retained continuous regions as partial discharge events, including start and end times, frequency band range, and corresponding intensity indicators, forming detection results that can be used for subsequent alarms or automatic control.

[0014] On the other hand, the multi-source collaborative detection system for partial discharge signals based on time-frequency fusion analysis includes the following modules: a time-frequency spectrum construction module, used to acquire the original partial discharge signals of the electrical signal channel and the ultra-high frequency channel, perform time-frequency transformation respectively, and generate a multi-source time-frequency feature spectrum; a cross-source interference fingerprint extraction module, used to extract the high-frequency noise patterns of different channels under non-partial discharge conditions based on the multi-source time-frequency feature spectrum, and construct a cross-source interference fingerprint library; a differential feature mapping generation module, used to perform band-by-band differential mapping calculation on the actually acquired multi-source time-frequency features based on the cross-source interference fingerprint library, and generate a differential feature matrix; and a fusion mapping module, used to introduce reverse consistency constraints into the differential feature matrix, perform weighted fusion, obtain a multi-source coupled time-frequency feature map, and output the detection results of the partial discharge signal.

[0015] The technical effects and advantages of the multi-source collaborative detection method and system for partial discharge signals based on time-frequency fusion analysis of this invention are as follows: 1. This invention constructs a multi-source time-frequency feature spectrum, extracts high-frequency noise patterns, and establishes a cross-source interference fingerprint database. It performs frequency-band differential mapping on actual signals and introduces reverse consistency constraints and weighted fusion to achieve multi-channel collaborative feature enhancement, thereby effectively suppressing cross-source noise interference and improving the accuracy and robustness of partial discharge detection in strong noise environments.

[0016] 2. This invention performs candidate point discrimination, continuity analysis, and cross-source consistency verification based on multi-source coupled time-frequency feature mapping, outputting structured partial discharge events, including time intervals, frequency band ranges, and intensity indicators, to achieve end-to-end closed-loop detection, support automatic alarms and control, and improve the safety and reliability of power distribution systems or equipment. Attached Figure Description

[0017] Figure 1 This is a flowchart illustrating the multi-source collaborative detection method for partial discharge signals based on time-frequency fusion analysis according to the present invention. Figure 2 This is a schematic diagram of the multi-source collaborative detection system for partial discharge signals based on time-frequency fusion analysis according to the present invention. Detailed Implementation

[0018] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0019] Example 1, Figure 1 The present invention provides a multi-source collaborative detection method for partial discharge signals based on time-frequency fusion analysis, comprising the following steps: S1, acquire the original partial discharge signals from the electrical signal channel and the ultra-high frequency channel, perform time-frequency transformation on each, and generate a multi-source time-frequency feature map, specifically: The original partial discharge signals of the electrical signal channel and the ultra-high frequency channel are acquired, and differential operations are performed with the pre-stored partial discharge-free background signal to obtain the residual signal corresponding to each channel. The residual signal is segmented and windowed, and the initial time spectrum is generated by short-time Fourier transform. A differential enhancement operator is applied to the initial frequency spectrum to highlight the burst characteristics of partial discharge; The time-frequency spectrum after differential enhancement is refined by multi-scale wavelet packet decomposition, and the time-frequency results of different channels are normalized by energy, thereby generating a multi-source time-frequency feature map with cross-source comparability. The residual signal is used to characterize the difference components between the original signal and the background noise, thereby suppressing the common interference of cross-source background noise in the map generation stage and highlighting the unique abnormal characteristics of partial discharge.

[0020] The difference enhancement operator is specifically:

[0021]

[0022]

[0023] in, For difference enhancement operators, For the first difference in the time direction, For the first-order difference in the frequency direction, , These are the weighting coefficients (adjusted according to the noise environment). Let f be the time-frequency spectrum energy at time t. The span of adjacent time windows, It represents the span between adjacent frequency bands.

[0024] S2, based on the multi-source time-frequency feature map, extract the high-frequency noise patterns of different channels under non-partial discharge conditions, and construct a cross-source interference fingerprint database; In this embodiment, high-frequency noise patterns of different channels under non-partial discharge conditions are extracted based on multi-source time-frequency feature maps, specifically as follows: Based on the multi-source time-frequency feature map, high-frequency intervals are selected, and the energy coupling function of different channels in each high-frequency band is calculated. The frequency band regions that are simultaneously enhanced in multiple channels (when the correlation between the energy sequences of two channels is greater than the preset correlation threshold) are identified as cross-source noise resonance bands. Local binary pattern texture encoding is performed on the time-frequency spectrum within the resonance band to extract the morphological distribution features of noise; Robust statistics on morphological distribution characteristics are performed based on the median absolute deviation method to form a stable high-frequency noise pattern.

[0025] The selection of high-frequency intervals based on multi-source time-frequency feature maps specifically involves: The energy difference sequence between the electrical signal channel and the ultra-high frequency channel in the frequency dimension is calculated based on the multi-source time-frequency feature map, and the frequency band with energy difference less than a preset threshold is selected as the first candidate interval. Calculate the cross-source phase consistency time variance of the first candidate interval, and mark the frequency bands that are unstable in phase and persist within a preset time as the second candidate interval; The intersection of the first and second candidate intervals is obtained, and the high-frequency interval is obtained through frequency domain morphological closing operations.

[0026] The energy difference sequence is specifically as follows:

[0027] The cross-source phase consistency time variance is specifically as follows:

[0028] in, It is an energy difference sequence. The set time window length, The time-frequency energy of the electrical signal channel. This refers to the time-frequency energy of the ultra-high frequency channel. Let f be the variance of the phase difference between the electrical signal channel and the ultra-high frequency channel at frequency f. To calculate the variance of a time series, The phase angle of the electrical signal channel at frequency f and time t. The phase angle of the ultra-high frequency channel at frequency f and time t.

[0029] The energy coupling degree function is specifically as follows:

[0030] ,

[0031] ,

[0032] in, The correlation between the energy sequences of the two channels. , These are the normalized mean energy values, , These are the average energy values ​​of the electrical signal channel and the ultra-high frequency channel, respectively.

[0033] The construction of the cross-source interference fingerprint database specifically involves: Based on the extracted high-frequency noise patterns, the corresponding frequency bands are marked as candidate interference intervals on the time-frequency feature maps of each channel; Local statistical features of each channel are extracted within each candidate interference interval, including average energy, energy variance, time-frequency texture descriptor, and local gradient trend. The features of each channel are fused according to frequency band to form a frequency band-level cross-source interference fingerprint vector; The interference fingerprint vector and its boundary, bandwidth, and duration information are stored in the fingerprint database, and an index is created to support fast matching and retrieval. When new candidate high-frequency interference interval data are available, the fingerprint database can be dynamically iteratively optimized by updating statistical features or adding new frequency band fingerprints through matching with the fingerprint database.

[0034] S3, based on the cross-source interference fingerprint database, performs frequency band differential mapping calculation on the actual collected multi-source time-frequency features to generate a differential feature matrix; In this embodiment, based on the cross-source interference fingerprint database, a frequency band-by-frequency differential mapping calculation is performed on the actually collected multi-source time-frequency features to generate a differential feature matrix, specifically: Based on the high-frequency noise pattern frequency band in the cross-source interference fingerprint database, the actual collected multi-source time-frequency features are mapped to the candidate frequency band region; The time-frequency energy of each channel in the candidate frequency band is standardized, based on the statistical mean and variance of the fingerprint database. Calculate the difference between the standardized signal and the fingerprint database features to form a channel-level difference vector; The differential vectors of each channel are integrated into a differential feature matrix according to the frequency band.

[0035] The differential feature matrix reflects the deviation of the actual signal from the cross-source interference fingerprint and is used for partial discharge discrimination or interference suppression.

[0036] The difference feature matrix is ​​specifically as follows:

[0037]

[0038]

[0039] in, It is the difference characteristic matrix. The differential characteristics of the electrical signal channel, Differential characteristics of the ultra-high frequency channel. The difference between the actual signal and the reference features in the fingerprint database. For the standardized time-frequency energy, This is the cross-source interference feature vector for the corresponding frequency band of this channel in the fingerprint database. For the actual multi-source time-frequency features collected, , These represent the statistical mean and standard deviation of the corresponding frequency band in the fingerprint database, respectively.

[0040] S4 introduces a reverse consistency constraint into the differential feature matrix and performs weighted fusion to obtain a multi-source coupled time-frequency feature map, and outputs the detection result of the partial discharge signal.

[0041] In this embodiment, a reverse consistency constraint is introduced into the difference feature matrix, and weighted fusion is performed to obtain a multi-source coupled time-frequency feature map, specifically: Calculate the channel-to-channel reverse consistency index for each frequency band and time point of the differential feature matrix. And mapped to weights ,in For the Sigmoid function; The difference feature matrices are weighted and fused according to their weights to generate a fused multi-source coupled feature matrix; The results of fusing all frequency bands and time points are combined into a matrix form to form a multi-source coupled time-frequency feature map.

[0042] The reverse consistency metric is specifically:

[0043] The multi-source coupling feature matrix is ​​specifically as follows:

[0044] in, As a reverse consistency indicator, The differential characteristics of the electrical signal channel, Differential characteristics of the ultra-high frequency channel. To avoid tiny constants with a denominator of zero, This is the multi-source coupling characteristic matrix.

[0045] The detection result of the output partial discharge signal is as follows: For each frequency band and time point of the multi-source coupled time-frequency feature mapping matrix, a discrimination threshold is set, and the points are judged and marked as candidate points for partial discharge. The labeled candidate points are subjected to connectivity analysis in the frequency and time dimensions. Continuous regions that meet the minimum bandwidth and minimum duration are retained, while isolated points or transient noise are removed. Verify the consistency of continuous regions across channels, preserve regions with enhanced cross-source coupling, and improve detection reliability and noise resistance; The final retained continuous area is output as a partial discharge event, including start and end time, frequency band range and corresponding intensity index, forming a detection result that can be used for subsequent alarm or automatic control.

[0046] Example 2, Figure 2 The present invention provides a multi-source collaborative detection system for partial discharge signals based on time-frequency fusion analysis, comprising the following modules: Time-frequency spectrum construction module: used to acquire the original partial discharge signals of the electrical signal channel and the ultra-high frequency channel, perform time-frequency transformation on them respectively, and generate multi-source time-frequency feature spectrum; Cross-source interference fingerprint extraction module: used to extract high-frequency noise patterns of different channels under non-partial discharge conditions based on multi-source time-frequency feature maps, and to construct a cross-source interference fingerprint database; Differential Feature Mapping Generation Module: Used to perform band-by-band differential mapping calculation on the actual collected multi-source time-frequency features based on the cross-source interference fingerprint database, and generate a differential feature matrix; The fusion mapping module is used to introduce reverse consistency constraints into the differential feature matrix, perform weighted fusion, obtain multi-source coupled time-frequency feature mapping, and output the detection results of partial discharge signals.

[0047] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0048] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, in the form of a computer program product.

[0049] Those skilled in the art will recognize that the modules and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0050] In addition, the functional modules in the various embodiments of this application can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module.

[0051] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0052] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A multi-source collaborative detection method for partial discharge signals based on time-frequency fusion analysis, characterized in that, Includes the following steps: The original partial discharge signals of the electrical signal channel and the ultra-high frequency channel are acquired, and time-frequency transformation is performed on them respectively to generate multi-source time-frequency feature maps; High-frequency noise patterns of different channels under non-partial discharge conditions are extracted based on multi-source time-frequency feature maps, and a cross-source interference fingerprint database is constructed. Based on the cross-source interference fingerprint database, frequency band differential mapping calculation is performed on the actual collected multi-source time-frequency features to generate a differential feature matrix; A reverse consistency constraint is introduced into the differential feature matrix, and weighted fusion is performed to obtain the multi-source coupled time-frequency feature map, and the detection result of the partial discharge signal is output.

2. The multi-source collaborative detection method for partial discharge signals based on time-frequency fusion analysis according to claim 1, characterized in that, The process involves acquiring the original partial discharge signals from the electrical signal channel and the ultra-high frequency channel, performing time-frequency transformations on each, and generating multi-source time-frequency feature maps, specifically as follows: The original partial discharge signals of the electrical signal channel and the ultra-high frequency channel are acquired, and differential operations are performed with the pre-stored partial discharge-free background signal to obtain the residual signal corresponding to each channel. The residual signal is segmented and windowed, and the initial time spectrum is generated by short-time Fourier transform. A differential enhancement operator is applied to the initial frequency spectrum to highlight the burst characteristics of partial discharge; The time-frequency spectrum after differential enhancement is refined by multi-scale wavelet packet decomposition, and the time-frequency results of different channels are normalized by energy to generate a multi-source time-frequency feature map.

3. The multi-source collaborative detection method for partial discharge signals based on time-frequency fusion analysis according to claim 2, characterized in that, The extraction of high-frequency noise patterns for different channels under non-partial discharge conditions based on multi-source time-frequency feature maps is specifically as follows: Based on the multi-source time-frequency feature map, high-frequency intervals are selected, and the energy coupling degree function of different channels in each high-frequency band is calculated to identify the frequency band regions that are simultaneously enhanced in multiple channels, which are taken as cross-source noise resonance bands. Local binary pattern texture encoding is performed on the time-frequency spectrum within the resonance band to extract the morphological distribution features of noise; Robust statistics on morphological distribution characteristics are performed based on the median absolute deviation method to form a stable high-frequency noise pattern.

4. The multi-source collaborative detection method for partial discharge signals based on time-frequency fusion analysis according to claim 3, characterized in that, The selection of high-frequency intervals based on multi-source time-frequency feature maps specifically involves: The energy difference sequence between the electrical signal channel and the ultra-high frequency channel in the frequency dimension is calculated based on the multi-source time-frequency feature map, and the frequency band with energy difference less than a preset threshold is selected as the first candidate interval. Calculate the cross-source phase consistency time variance of the first candidate interval, and mark the frequency bands that are unstable in phase and persist within a preset time as the second candidate interval; The intersection of the first and second candidate intervals is obtained, and the high-frequency interval is obtained through frequency domain morphological closing operations.

5. The multi-source collaborative detection method for partial discharge signals based on time-frequency fusion analysis according to claim 4, characterized in that, The construction of the cross-source interference fingerprint database specifically involves: Based on the extracted high-frequency noise patterns, the corresponding frequency bands are marked as candidate interference intervals on the time-frequency feature maps of each channel; Local statistical features of each channel are extracted within each candidate interference interval, including average energy, energy variance, time-frequency texture descriptor, and local gradient trend. The features of each channel are fused according to frequency band to form a frequency band-level cross-source interference fingerprint vector; The interference fingerprint vector and its boundary, bandwidth, and duration information are stored in the fingerprint database, and an index is created to support fast matching and retrieval. When new candidate high-frequency interference interval data are available, the fingerprint database can be dynamically iteratively optimized by updating statistical features or adding new frequency band fingerprints through matching with the fingerprint database.

6. The multi-source collaborative detection method for partial discharge signals based on time-frequency fusion analysis according to claim 5, characterized in that, The method based on the cross-source interference fingerprint database performs band-by-band differential mapping calculation on the actually collected multi-source time-frequency features to generate a differential feature matrix, specifically as follows: Based on the high-frequency noise pattern frequency band in the cross-source interference fingerprint database, the actual collected multi-source time-frequency features are mapped to the candidate frequency band region; The time-frequency energy of each channel in the candidate frequency band is standardized, based on the statistical mean and variance of the fingerprint database. Calculate the difference between the standardized signal and the fingerprint database features to form a channel-level difference vector; The differential vectors of each channel are integrated into a differential feature matrix according to the frequency band.

7. The multi-source collaborative detection method for partial discharge signals based on time-frequency fusion analysis according to claim 6, characterized in that, A reverse consistency constraint is introduced into the difference feature matrix, and weighted fusion is performed to obtain the multi-source coupled time-frequency feature map, specifically: For each frequency band and time point of the differential feature matrix, calculate the channel-to-channel reverse consistency index and map it to weights; The difference feature matrices are weighted and fused according to their weights to generate a fused multi-source coupled feature matrix; The results of fusing all frequency bands and time points are combined into a matrix form to form a multi-source coupled time-frequency feature map.

8. The multi-source collaborative detection method for partial discharge signals based on time-frequency fusion analysis according to claim 7, characterized in that, The detection result of the output partial discharge signal is as follows: For each frequency band and time point of the multi-source coupled time-frequency feature mapping matrix, a discrimination threshold is set, and the points are judged and marked as candidate points for partial discharge. The labeled candidate points are subjected to connectivity analysis in the frequency and time dimensions. Continuous regions that meet the minimum bandwidth and minimum duration are retained, while isolated points or transient noise are removed. Verify the consistency of continuous regions across channels, preserve regions with enhanced cross-source coupling, and improve detection reliability and noise resistance; The final retained continuous area is output as a partial discharge event, including start and end time, frequency band range and corresponding intensity index, forming a detection result that can be used for subsequent alarm or automatic control.

9. A system using a multi-source collaborative detection method for partial discharge signals based on time-frequency fusion analysis as described in any one of claims 1-8, characterized in that, Includes the following modules: Time-frequency spectrum construction module: used to acquire the original partial discharge signals of the electrical signal channel and the ultra-high frequency channel, perform time-frequency transformation on them respectively, and generate multi-source time-frequency feature spectrum; Cross-source interference fingerprint extraction module: used to extract high-frequency noise patterns of different channels under non-partial discharge conditions based on multi-source time-frequency feature maps, and to construct a cross-source interference fingerprint database; Differential Feature Mapping Generation Module: Used to perform band-by-band differential mapping calculation on the actual collected multi-source time-frequency features based on the cross-source interference fingerprint database, and generate a differential feature matrix; The fusion mapping module is used to introduce reverse consistency constraints into the differential feature matrix, perform weighted fusion, obtain multi-source coupled time-frequency feature mapping, and output the detection results of partial discharge signals.

Images