Substation partial discharge uhf signal feature compression and reconstruction method and system

Abstract

The application discloses a substation partial discharge UHF signal feature compression and reconstruction method and system, and the method comprises the following steps: constructing a partial discharge UHF signal generation model, representing an original signal as a superposition form of multi-source pulse excitation and propagation channel response, and representing multi-path reflection and attenuation characteristics in a GIS metal enclosed structure; pre-processing the original signal and extracting time-frequency features to construct a time-frequency joint feature representation matrix; performing sparse coding on the features based on an over-complete dictionary to obtain a sparse coefficient matrix; introducing pulse structure constraints and energy constraints in the sparse coefficient space, constructing a nonlinear feature compression mapping, determining a feature compression rate according to wireless communication parameters in the substation, and realizing feature dimension reduction; constructing a feature weight function according to pulse energy distribution, screening low-dimensional feature expression, and forming compressed data; after the compressed data is transmitted through a wireless link, feature recovery and signal reconstruction are realized at a receiving end based on an inverse mapping and the partial discharge UHF signal generation model. The application can reduce data transmission burden while maintaining the integrity of key features of the partial discharge signal.

Description

Technical Field

[0001] This invention relates to the field of partial discharge monitoring technology for GIS equipment, specifically to a method and system for compressing and reconstructing the characteristics of UHF partial discharge signals in substations. Background Technology

[0002] Gas-insulated metal-enclosed switchgear (GIS) in substations is a critical piece of equipment in power systems. Operating long-term in high-voltage and strong electromagnetic environments, its internal insulation defects are prone to causing partial discharges, affecting the safe operation of the equipment. Effective monitoring of partial discharges in GIS equipment is beneficial for understanding the insulation condition and preventing faults.

[0003] Detection methods based on ultra-high frequency (UHF) signals are widely used due to their strong anti-interference capabilities. However, partial discharge UHF signals have characteristics such as strong pulse characteristics, wide bandwidth, and multi-path propagation. In GIS closed structures and complex electromagnetic environments, they are prone to reflection superposition and noise interference, making signal feature extraction difficult.

[0004] Existing methods mostly employ time-frequency analysis or data-driven signal processing, but they suffer from problems such as sensitivity to noise and unstable feature representation. Meanwhile, raw UHF signals have large data volumes, placing high demands on bandwidth and real-time performance in wireless transmission. Summary of the Invention

[0005] The technical problem to be solved by this invention is to achieve effective feature compression and reconstruction of partial discharge UHF signals, so as to reduce the transmission burden while ensuring the integrity of key information and improve wireless communication efficiency and monitoring capabilities.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution: A method for compressing and reconstructing the characteristics of UHF signals from partial discharge in substations includes: A partial discharge UHF signal generation model is constructed, which represents the original signal as a superposition of multi-source pulse excitation and propagation channel response, and characterizes the multipath reflection and attenuation characteristics in GIS metal enclosed structures. The original signal is preprocessed and time-frequency features are extracted to construct a joint time-frequency feature representation matrix; the features are sparsely encoded based on an overcomplete dictionary to obtain a sparse coefficient matrix. Pulse structure constraints and energy constraints are introduced into the sparse coefficient space to construct a nonlinear feature compression mapping, and the feature compression rate is determined according to the wireless communication parameters in the substation to achieve feature dimensionality reduction. A feature weighting function is constructed based on the pulse energy distribution to filter low-dimensional feature representations and form compressed data; After the compressed data is transmitted via a wireless link, feature recovery and signal reconstruction are achieved at the receiving end based on the inverse mapping and partial discharge UHF signal generation model.

[0007] Furthermore, a time-frequency joint feature representation matrix is ​​constructed, including: performing bandpass filtering, normalization, and time window segmentation on the original signal to obtain a discrete signal sequence, and performing short-time Fourier transform and wavelet multi-scale decomposition on it to construct a time-frequency joint feature representation matrix to describe the energy spread and spectral discrete distribution characteristics of the partial discharge pulse in the UHF band.

[0008] Furthermore, obtaining the sparse coefficient matrix involves: sparsely encoding and solving for the sparse coefficient matrix based on the low duty cycle and spatial dispersion of the partial discharge UHF signal in the time domain. .

[0009] Furthermore, the sparse coding process satisfies the following optimization model: ; in, This represents the joint time-frequency characteristic representation matrix. Represents an overcomplete dictionary matrix. Indicates the first A sparse component For a sparse component set, This represents the first-order difference operator with respect to the time variable. and Represents the regularization coefficient. This represents the square value of the Euclidean norm. Describing the L1 norm, This represents the sparse coefficient matrix that minimizes the objective function.

[0010] Furthermore, a nonlinear feature compression mapping is constructed, and the feature compression ratio is determined based on the wireless communication parameters within the substation. This includes: calculating the feature compression ratio based on the available bandwidth of the wireless communication link and the original signal transmission rate, and performing dimensionality reduction compression on the sparse coefficient matrix based on the feature compression ratio to obtain a low-dimensional feature expression that meets the constraints of the substation's wireless communication link bandwidth and anti-interference performance.

[0011] Furthermore, the feature compression ratio satisfies the following relationship: ; in, For compression ratio, For the compressed dimension, To compress the dimension before, For wireless communication link bandwidth, For the original data rate, Let V be the variance of the signal within the current time window. for The UHF partial discharge signal received at all times.

[0012] Furthermore, a characteristic weighting function is constructed based on the pulse energy distribution, expressed by the following formula: ; in, Indicates the first Each pulse corresponds to a feature weight, and Δ is the width of the pulse neighborhood window. for UHF partial discharge signals received at all times , The first , At the time of pulse occurrence, This represents the total number of pulses.

[0013] Furthermore, the low-dimensional feature representations are filtered and compressed data is formed, including: retaining features with feature weights greater than a preset threshold to participate in compression encoding and constructing compressed transmission data packets.

[0014] Furthermore, the processes of feature recovery and signal reconstruction satisfy the following relationship: ; ; ; ; ; in, To reconstruct the sparse coefficient matrix, for Reconstructed signal at time, For the reconstruction of the first Amplitude of a partial discharge pulse To reconstruct the pulse timing, It is a nonlinear feature compression mapping model. It is the Sigmoid activation function. This is a constraint term for anti-interference performance. In the sparse component set Select the maximum value from the list. To reconstruct the sparse coefficient matrix The Middle The column vectors corresponding to the sparse components The square value of the Euclidean norm. For the first Each sparse component corresponds to a time center position. For the mapping matrix, For bias terms, The total number of pulses. For the first Each pulse corresponds to a feature weight. In order to be in The propagation channel impulse response function that consistently satisfies the frequency-selective attenuation characteristics within the UHF signal band. These are low-dimensional feature vectors.

[0015] The present invention also provides a system for applying the above-described method for compressing and reconstructing the characteristics of UHF signals from partial discharge in substations, comprising: The signal modeling module is used to construct a partial discharge UHF signal generation model, which represents the original signal as a superposition of multi-source pulse excitation and propagation channel response, and characterizes the multipath reflection and attenuation characteristics in GIS metal enclosure structures. The feature extraction module is used to preprocess the original signal and extract time-frequency features, and construct a joint time-frequency feature representation matrix; based on an overcomplete dictionary, the features are sparsely encoded to obtain a sparse coefficient matrix; The sparse coding module is used to introduce pulse structure constraints and energy constraints in the sparse coefficient space, construct a nonlinear feature compression mapping, and determine the feature compression rate based on the wireless communication parameters in the substation to achieve feature dimensionality reduction. The feature compression module is used to construct a feature weighting function based on the pulse energy distribution, filter low-dimensional feature representations, and form compressed data. The wireless transmission module and signal reconstruction module are used to compress data through the wireless link. Compared with the prior art, the beneficial effects of the present invention are: This invention constructs a partial discharge UHF signal generation model and a feature compression and reconstruction structure. The structure includes time-frequency joint feature representation, overcomplete dictionary sparse coding, nonlinear feature compression mapping, and an inverse reconstruction mechanism. Based on actual acquired partial discharge UHF signals from GIS equipment, under conditions of complex signal propagation paths and unknown multi-source pulse superposition, a signal generation model is used to achieve a structured representation of the original signal. Sparse feature parameters are obtained through time-frequency joint analysis and sparse coding, and pulse structure constraints and energy constraints are introduced into the sparse coefficient space to achieve stable characterization of key discharge features. Furthermore, a feature weight function is constructed to filter sparse features, retaining only features with weights greater than a set threshold for compression coding, thereby reducing data dimensionality while ensuring the integrity of the main discharge pulse information, achieving feature-level compression. At the receiving end, sparse coefficients are recovered based on the inverse mapping model, and signal reconstruction is achieved in conjunction with the signal generation model.

[0016] Compared to the direct transmission of raw data, this invention can significantly reduce the amount of data transmitted and improve data transmission efficiency under the same communication bandwidth conditions. Through sparse modeling and feature weighting screening mechanisms, it effectively suppresses the impact of noise on feature extraction and improves the accuracy and stability of key discharge pulse identification. In the signal reconstruction stage, by introducing structural constraints, reconstruction errors can be reduced and the consistency of feature information can be guaranteed.

[0017] The proposed solution has high compression efficiency and strong anti-interference capability. It can maintain stable feature expression performance even in complex electromagnetic environments. While ensuring the consistency of the main features of the reconstructed signal with the original signal, it reduces the wireless communication load and improves the real-time performance of the system, thus providing an efficient and reliable technical means for online status monitoring of substation GIS equipment.

[0018] This invention integrates the physical propagation characteristics of GIS partial discharge UHF signals, the structure and energy characteristics of partial discharge pulses, and the wireless communication link conditions of substations into a single feature compression and reconstruction process. Conventional signal compression methods typically do not perform characteristic analysis and modeling of the signal source object; they only extract features and reduce dimensionality for the signal itself. Compression parameters are relatively fixed, and there is a lack of direct connection between the signal processing process and subsequent wireless transmission conditions. For the specific application scenario of GIS partial discharge UHF signals, conventional methods also typically do not jointly analyze wireless communication parameters with the object itself.

[0019] This invention first considers the multipath reflection and attenuation characteristics of UHF signals in GIS metal enclosed structures to establish a signal generation model. Based on this, a sparse representation is performed, and the compression process is constrained by the structure and energy characteristics of partial discharge pulses. At the same time, the degree of compression is dynamically determined according to the available bandwidth of the wireless communication link, the original signal transmission rate, and the current signal state. Then, based on the pulse energy distribution, more important features representing the partial discharge state are selected for transmission. Finally, at the receiving end, the signal is reconstructed by combining the aforementioned signal generation model.

[0020] This invention can be summarized as establishing a correlation mechanism between the physical characteristics, feature compression, and wireless communication resources of GIS partial discharge UHF signals, while reducing the amount of wireless transmission data and communication burden, and maintaining the key pulse characteristics of partial discharge and its state characterization capabilities as much as possible. Attached Figure Description

[0021] Figure 1 This is a flowchart of a method for compressing and reconstructing the characteristics of a UHF signal with partial discharge in a substation, according to an embodiment of the present invention.

[0022] Figure 2 This is a schematic diagram comparing the original signal and the reconstructed signal in an embodiment of the present invention.

[0023] Figure 3 This is a schematic diagram comparing the original transmission and the communication performance of the method of the present invention in an embodiment of the present invention.

[0024] Figure 4 This is a schematic diagram of the time-frequency energy spectrum distribution of the original signal and the reconstructed signal in an embodiment of the present invention. Detailed Implementation

[0025] To facilitate understanding of the technical solution of the present invention by those skilled in the art, the technical solution of the present invention will now be further described in conjunction with the accompanying drawings.

[0026] The terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Therefore, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this application, "multiple" means two or more, unless otherwise explicitly specified.

[0027] See Figure 1 This invention provides a method for compressing and reconstructing the characteristics of UHF signals from partial discharge in substations, comprising: S10. Construct a partial discharge UHF signal generation model, representing the original signal as a superposition of multi-source pulse excitation and propagation channel response, characterizing the multipath reflection and attenuation characteristics in GIS metal enclosed structures.

[0028] In one embodiment of the present invention, a model for generating partial discharge UHF signals caused by dielectric defects inside substation GIS equipment is constructed. The multipath reflection and attenuation characteristics of electromagnetic waves inside the GIS shell in the metal enclosed structure are represented by the propagation channel impulse response function. The original signal is represented as a convolution superposition of multi-source pulse excitation and propagation channel response. Based on this model, signal structure constraints are constructed.

[0029] In this embodiment, the partial discharge UHF signal generation model is represented as follows: ; ; in, express UHF partial discharge signals received at all times Indicates the first Amplitude of a partial discharge pulse, Indicates the first At the time of pulse occurrence, This represents the propagation channel impulse response function that satisfies the frequency-selective attenuation characteristics within the UHF band. This represents the total number of effective transmission paths. Indicates the first The amplitude attenuation coefficient of the path, Denotes the Dirac impulse function. Indicates the first The propagation delay of each path, This indicates the propagation type mechanism, such as direct, reflected, and leaky. It is an exponentially decaying term. The first part represents the first part, which consists of conductors, dielectrics, and surface currents. Path decay factor This indicates noise items including thermal noise and electromagnetic interference from substations. Indicates the total number of pulses. Indicates the noise variance. The model is a zero-mean Gaussian white noise model.

[0030] S20: Preprocess the original signal and extract time-frequency features to construct a joint time-frequency feature representation matrix; perform sparse coding on the features based on an overcomplete dictionary to obtain a sparse coefficient matrix.

[0031] In one embodiment of the present invention, the acquired raw signal is subjected to bandpass filtering, normalization and time window segmentation to obtain a discrete signal sequence, and then subjected to short-time Fourier transform and wavelet multi-scale decomposition to construct a time-frequency joint feature representation matrix to describe the energy spread and spectral discrete distribution characteristics of the partial discharge pulse in the UHF band.

[0032] In this embodiment, the time-frequency joint feature representation matrix The structure is as follows: ; in, Indicates the signal in time Time-frequency energy distribution at that location This represents the sliding window function. Indicates frequency, Represents the Fourier basis functions. Represents a time-shifted variable. This represents the time variance of the Gaussian sliding window function.

[0033] In one embodiment of the present invention, an overcomplete dictionary matrix is ​​constructed based on the partial discharge UHF signal generation model and the time-frequency joint feature representation matrix, and the time-frequency joint feature representation matrix is ​​sparsely encoded to obtain a sparse coefficient matrix. In this embodiment, sparse coding is performed to solve for the sparse coefficient matrix, addressing the low duty cycle and spatial discreteness of the partial discharge UHF signal in the time domain. Its sparse coding process satisfies the following optimization model: ; in, This represents the time-frequency joint feature representation matrix. Represents an overcomplete dictionary matrix. Indicates the first A sparse component For a sparse component set, This represents the first-order difference operator with respect to the time variable. and Represents the regularization coefficient. This represents the square value of the Euclidean norm. Describing the L1 norm, This represents the sparse coefficient matrix that minimizes the objective function. .

[0034] S30 introduces pulse structure constraints and energy constraints in the sparse coefficient space to construct a nonlinear feature compression mapping, and determines the feature compression rate based on the wireless communication parameters in the substation to achieve feature dimensionality reduction.

[0035] In one embodiment of the present invention, a nonlinear feature compression mapping is constructed, and the feature compression ratio is determined according to the wireless communication parameters in the substation. This includes: calculating the feature compression ratio based on the available bandwidth of the wireless communication link and the original signal transmission rate, and performing dimensionality reduction compression on the sparse coefficient matrix based on the feature compression ratio to obtain a low-dimensional feature expression that meets the constraints of the wireless communication link bandwidth and anti-interference performance of the substation.

[0036] In this embodiment, the nonlinear mapping is used to preserve the transient characteristics and energy concentration of the partial discharge pulse while suppressing the influence of background electromagnetic noise on the feature representation. Its mapping satisfies the following relationship: ; in, It is a low-dimensional feature vector. For the mapping matrix, For bias terms, It is a non-linear activation function. This is an anti-interference performance constraint (signal structure constraint) item.

[0037] The anti-interference performance constraint is expressed as follows: ; in, and Represent the original and the result of the sparse coefficient matrix, respectively. Estimated pulse position, and Represent the original and the result of the sparse coefficient matrix, respectively. SThe estimated pulse energy, pulse location, and energy difference are used to constrain the consistency of the discharge path and the stability of the discharge intensity of the partial discharge source within the GIS. and These are the weighting coefficients.

[0038] In this embodiment, during the dimensionality reduction and compression of the sparse coefficient matrix, the feature compression ratio is dynamically determined based on the wireless communication link conditions. The feature compression ratio satisfies the following relationship: ; in, For compression ratio, For the compressed dimension, To compress the dimension before, For wireless communication link bandwidth, For the original data rate, The variance of the signal within the current time window The compression ratio constraint simultaneously meets the transmission requirements of the substation online monitoring system and the electromagnetic interference immunity requirements of the wireless link. for The UHF partial discharge signal received at all times.

[0039] S40: Construct a feature weighting function based on pulse energy distribution, filter low-dimensional feature representations, and form compressed data.

[0040] In one embodiment of the present invention, a feature weighting function is constructed based on the temporal distribution characteristics and energy accumulation characteristics of partial discharge pulses. Low-dimensional feature expressions are screened, and features with feature weights greater than a preset threshold are retained to participate in compression encoding and construct compressed transmission data packets.

[0041] In this embodiment, the feature weight function is expressed by the following formula: ; in, Indicates the first Each pulse corresponds to a feature weight, and Δ is the width of the pulse neighborhood window. for UHF partial discharge signals received at all times , The first , At the time of pulse occurrence, This represents the total number of pulses.

[0042] S50: After compressed data is transmitted via a wireless link, feature recovery and signal reconstruction are achieved at the receiving end based on the inverse mapping and partial discharge UHF signal generation model.

[0043] In one embodiment of the present invention, the compressed data packet is sent through a wireless communication link, and the sparse coefficient matrix is ​​recovered at the receiving end based on the inverse mapping model, and the original signal or key features are reconstructed by combining the partial discharge UHF signal generation model.

[0044] In this embodiment, the processes of feature recovery and signal reconstruction satisfy the following relationship: ; ; ; ; ; in, To reconstruct the sparse coefficient matrix, for Reconstructed signal at time, For the reconstruction of the first Amplitude of a partial discharge pulse, To reconstruct the pulse timing, It is a nonlinear feature compression mapping model. It is the Sigmoid activation function. This is a constraint term for anti-interference performance. In the sparse component set Select the maximum value from the list. To reconstruct the sparse coefficient matrix The Middle The column vectors corresponding to the sparse components The square value of the Euclidean norm. For the first Each sparse component corresponds to a time center position. For the mapping matrix, For bias terms, The total number of pulses. For the first Each pulse corresponds to a feature weight. In order to be in The propagation channel impulse response function that consistently satisfies the frequency-selective attenuation characteristics within the UHF signal band. These are low-dimensional feature vectors.

[0045] The method for compressing and reconstructing the UHF wireless signal characteristics of partial discharge in substation GIS equipment proposed in this invention was used for field verification of a 220kV substation GIS partial discharge online monitoring system. Figure 1As shown. The UHF signal is acquired by sensors deployed inside the GIS bay, preprocessed by the acquisition nodes, and then transmitted to the monitoring terminal via a wireless link. Primary equipment within the station includes GIS switchgear, busbars, and circuit breakers. Communication nodes are distributed near the GIS bay, in the cable trench, and in the relay protection room. UHF signals under different operating conditions are acquired at a sampling frequency of 500MHz, with a single sampling duration of 20ms. 200 sets of data are collected for each operating condition, of which 160 sets are used for model building and 40 sets are used for testing. First, the signal undergoes bandpass filtering and normalization from 300MHz to 1.5GHz to construct a time-frequency feature matrix, and sparse coding is performed based on an overcomplete dictionary. Key pulses are selected using a feature weight function, retaining only features with weights greater than 0.05 for compression to form a low-dimensional representation.

[0046] Table 1 shows a performance comparison between the present invention and other methods. Under the conditions of 5MHz bandwidth and the same transmission power, the original data transmission volume is about 2.4MB, while it is reduced to 0.38MB after using the method of the present invention, with a compression rate of 84.2% and an average transmission latency reduced from 120ms to 35ms. NMSE Normalized mean square error STFT Feature Compression

[0047] Table 1 Performance Comparison Results of Different Methods

[0048] The signal reconstruction result is as follows Figure 2 As shown, the reconstructed signal maintains consistency with the original signal in pulse position and energy distribution, with a normalized mean square error of 0.036. Based on support vector machines, the identification accuracy of the normal state and three types of defect states is improved from 94.5% for the original signal to 97.8% for the method of this invention, with the identification accuracy of each type of defect exceeding 96%. Regarding communication stability... like Figure 3 As shown, during 30 minutes of continuous operation, the original transmission method had a data packet loss rate of 6.2% and an average retransmission rate of 2.8 times / minute, while the method of this invention reduced these to 2.1% and 0.9 times / minute, respectively. The time-frequency distributions of the original signal and the reconstructed signal are shown below. Figure 4As shown, the overall structure remains consistent, with good preservation of the main frequency energy and pulse timing structure. Only slight differences exist in local high-frequency details and energy distribution, indicating that the proposed method effectively preserves the key features of the partial discharge signal while achieving data compression. In summary, under the complex electromagnetic environment of a 220kV substation, the method of this invention achieves high-ratio data compression while ensuring the integrity of key partial discharge features, significantly reducing communication burden, and simultaneously improving signal reconstruction accuracy and recognition performance, demonstrating good feasibility for engineering applications.

[0049] This invention also provides a system for compressing and reconstructing the characteristics of UHF signals from partial discharge in substations, comprising: The signal modeling module is used to construct a partial discharge UHF signal generation model, representing the original signal as a superposition of multi-source pulse excitation and propagation channel response, and characterizing the multipath reflection and attenuation characteristics in GIS metal enclosed structures.

[0050] The feature extraction module is used to preprocess the original signal and extract time-frequency features to construct a joint time-frequency feature representation matrix; based on an overcomplete dictionary, the features are sparsely encoded to obtain a sparse coefficient matrix.

[0051] The sparse coding module is used to introduce pulse structure constraints and energy constraints in the sparse coefficient space, construct a nonlinear feature compression mapping, and determine the feature compression rate based on the wireless communication parameters in the substation to achieve feature dimensionality reduction.

[0052] The feature compression module is used to construct a feature weighting function based on the pulse energy distribution, filter low-dimensional feature representations, and form compressed data.

[0053] The wireless transmission module and signal reconstruction module are used to perform feature recovery and signal reconstruction at the receiving end based on the inverse mapping and partial discharge UHF signal generation model after compressed data is transmitted through the wireless link.

[0054] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the invention can be implemented in other specific forms without departing from its spirit or essential characteristics. Therefore, the embodiments should be considered illustrative and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be included within the present invention, and no reference numerals in the claims should be construed as limiting the scope of the claims.

[0055] The above embodiments are merely examples of implementation methods of the invention. The scope of protection of the present invention is not limited to the above embodiments. For those skilled in the art, several modifications and improvements can be made without departing from the concept of the present invention, and these all fall within the scope of protection of the present invention.

Claims

1. A method for compressing and reconstructing the characteristics of UHF signals with partial discharge in substations, characterized in that, include: A partial discharge UHF signal generation model is constructed, which represents the original signal as a superposition of multi-source pulse excitation and propagation channel response, and characterizes the multipath reflection and attenuation characteristics in GIS metal enclosed structures. The original signal is preprocessed and time-frequency features are extracted to construct a joint time-frequency feature representation matrix; Sparse coding of features is performed based on an overcomplete dictionary to obtain a sparse coefficient matrix; Pulse structure constraints and energy constraints are introduced into the sparse coefficient space to construct a nonlinear feature compression mapping, and the feature compression rate is determined according to the wireless communication parameters in the substation to achieve feature dimensionality reduction. A feature weighting function is constructed based on the pulse energy distribution to filter low-dimensional feature representations and form compressed data; After the compressed data is transmitted via a wireless link, feature recovery and signal reconstruction are achieved at the receiving end based on the inverse mapping and partial discharge UHF signal generation model.

2. The method for compressing and reconstructing the characteristics of UHF signals with partial discharge in substations according to claim 1, characterized in that, The construction of the time-frequency joint feature representation matrix includes: performing bandpass filtering, normalization, and time window segmentation on the original signal to obtain a discrete signal sequence, and performing short-time Fourier transform and wavelet multi-scale decomposition on it to construct the time-frequency joint feature representation matrix to describe the energy spread and spectral discrete distribution characteristics of the partial discharge pulse in the UHF band.

3. The method for compressing and reconstructing the characteristics of UHF signals with partial discharge in substations according to claim 1, characterized in that, Obtaining the sparse coefficient matrix involves: sparsely encoding and solving for the sparse coefficient matrix to account for the low duty cycle and spatial discreteness of the partial discharge UHF signal in the time domain. .

4. The method for compressing and reconstructing the characteristics of UHF signals with partial discharge in substations according to claim 3, characterized in that, The sparse coding process satisfies the following optimization model: ； in, This represents the joint time-frequency characteristic representation matrix. Represents an overcomplete dictionary matrix. Indicates the first A sparse component For a sparse component set, This represents the first-order difference operator with respect to the time variable. and Represents the regularization coefficient. This represents the square value of the Euclidean norm. Describing the L1 norm, This represents the sparse coefficient matrix that minimizes the objective function.

5. The method for compressing and reconstructing the characteristics of UHF signals with partial discharge in substations according to claim 1, characterized in that, A nonlinear feature compression mapping is constructed, and the feature compression ratio is determined based on the wireless communication parameters in the substation. This includes: calculating the feature compression ratio based on the available bandwidth of the wireless communication link and the original signal transmission rate, and performing dimensionality reduction compression on the sparse coefficient matrix based on the feature compression ratio to obtain a low-dimensional feature expression that meets the constraints of the wireless communication link bandwidth and anti-interference performance of the substation.

6. The method for compressing and reconstructing the characteristics of UHF partial discharge signals in substations according to claim 1, characterized in that, Feature compression ratio satisfies the following relationship: ； in, For compression ratio, For the compressed dimension, To compress the dimension before, For wireless communication link bandwidth, For the original data rate, Let V be the variance of the signal within the current time window. for The UHF partial discharge signal received at all times.

7. The method for compressing and reconstructing the characteristics of UHF signals with partial discharge in substations according to claim 1, characterized in that, The feature weighting function is constructed based on the pulse energy distribution and is expressed by the following formula: ； in, Indicates the first Each pulse corresponds to a feature weight, and Δ is the width of the pulse neighborhood window. for UHF partial discharge signals received at all times , The first , At the time of pulse occurrence, This represents the total number of pulses.

8. The method for compressing and reconstructing the characteristics of UHF signals with partial discharge in substations according to claim 7, characterized in that, The low-dimensional feature representations are filtered and compressed data is generated, including: retaining features with feature weights greater than a preset threshold to participate in compression encoding and constructing compressed transmission data packets.

9. The method for compressing and reconstructing the characteristics of UHF signals with partial discharge in substations according to claim 1, characterized in that, The processes of feature recovery and signal reconstruction satisfy the following relationship: ； ； ； ； ； in, To reconstruct the sparse coefficient matrix, for Reconstructed signal at time, For the reconstruction of the first Amplitude of a partial discharge pulse To reconstruct the pulse timing, It is a nonlinear feature compression mapping model. It is the Sigmoid activation function. This is a constraint term for anti-interference performance. In the sparse component set Select the maximum value from the list. To reconstruct the sparse coefficient matrix The Middle The column vectors corresponding to the sparse components The square value of the Euclidean norm. For the first Each sparse component corresponds to a time center position. For the mapping matrix, For bias terms, The total number of pulses. For the first Each pulse corresponds to a feature weight. In order to be in The propagation channel impulse response function that consistently satisfies the frequency-selective attenuation characteristics within the UHF signal band. These are low-dimensional feature vectors.

10. A system for compressing and reconstructing the characteristics of a substation partial discharge UHF signal according to any one of claims 1-9, characterized in that, include: The signal modeling module is used to construct a partial discharge UHF signal generation model, which represents the original signal as a superposition of multi-source pulse excitation and propagation channel response, and characterizes the multipath reflection and attenuation characteristics in GIS metal enclosure structures. The feature extraction module is used to preprocess the original signal and extract time-frequency features to construct a joint time-frequency feature representation matrix; Sparse coding of features is performed based on an overcomplete dictionary to obtain a sparse coefficient matrix; The sparse coding module is used to introduce pulse structure constraints and energy constraints in the sparse coefficient space, construct a nonlinear feature compression mapping, and determine the feature compression rate based on the wireless communication parameters in the substation to achieve feature dimensionality reduction. The feature compression module is used to construct a feature weighting function based on the pulse energy distribution, filter low-dimensional feature representations, and form compressed data. The wireless transmission module and signal reconstruction module are used to perform feature recovery and signal reconstruction at the receiving end based on the inverse mapping and partial discharge UHF signal generation model after compressed data is transmitted through the wireless link.

Images