Multi-source feature extraction method and system for cable signal
By improving the empirical mode decomposition algorithm and autoencoder feature extraction, the problem of poor cable signal decomposition results is solved, improving the accuracy and stability of cable fault diagnosis and adapting to complex electromagnetic environments.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- JIANGMEN POWER SUPPLY BUREAU OF GUANGDONG POWER GRID CO LTD
- Filing Date
- 2025-12-15
- Publication Date
- 2026-04-21
AI Technical Summary
In existing technologies, EMD decomposition algorithms for cable signals are susceptible to interference from complex electromagnetic environments, resulting in poor decomposition quality, difficulty in accurately locating fault characteristics, and impacting cable operation and maintenance efficiency.
By using an improved empirical mode decomposition algorithm, iterative decomposition is performed using denoising statistics to dynamically optimize the noise coefficient, thereby enhancing the stability of signal decomposition. In addition, feature extraction is performed by combining an autoencoder to improve the quality of signal decomposition.
It improves the quality and stability of cable signal decomposition, enhances adaptability to complex interference environments, provides reliable fault characteristic data support, and improves the accuracy of cable fault diagnosis.
Smart Images
Figure CN121901718A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of signal analysis, and in particular to a method and system for extracting multi-source features from cable signals. Background Technology
[0002] As a key carrier for power transmission and information transmission, the operating status of cables directly affects power grid security and communication quality. In actual operation, cable signals are often exposed to complex electromagnetic environments and are affected by a combination of various interference sources.
[0003] Currently, the analysis of cable signals often relies on the traditional EMD (Empirical Mode Decomposition) algorithm, which is easily affected by the complex environment in which the cable is actually located, resulting in poor quality of the decomposed components. Ultimately, it is difficult to accurately locate fault features, which affects the efficiency of cable operation and maintenance.
[0004] Therefore, how to effectively analyze cable signals and improve the reliability of subsequent power operation and maintenance has become a technical problem that urgently needs to be solved by those skilled in the art. Summary of the Invention
[0005] This invention provides a method and system for multi-source feature extraction of cable signals, addressing how to improve the quality of signal decomposition by improving the ensemble empirical mode decomposition algorithm.
[0006] To address the aforementioned technical problems, embodiments of the present invention provide a method for multi-source feature extraction of cable signals, comprising: Acquire the cable fault signal of the target cable and perform iterative empirical mode decomposition on the cable fault signal; In each iteration, the denoising statistics of the cable fault signal are obtained; The initial noise coefficient introduced in each iteration is dynamically optimized in conjunction with the aforementioned denoising statistical index to obtain the corresponding optimal noise coefficient; Using the optimal noise figure, extract the signal component information generated in each iteration from the cable fault signal; When the iteration terminates, the target decomposition result of the cable fault signal is determined based on the signal component information; Feature extraction is performed on the target decomposition results to obtain cable fault features.
[0007] Furthermore, the method of dynamically optimizing the initial noise coefficient introduced in each iteration in conjunction with the denoising statistical index to obtain the corresponding optimal noise coefficient includes: A weighted analysis is performed on the denoising statistical indicators to construct an objective function, and gradient descent iteration is performed on the initial noise coefficient with the goal of maximizing the output of the objective function. The iteration terminates when the difference between the noise figure values calculated in two consecutive gradient descent iterations meets the preset convergence condition, and the optimal noise figure is determined.
[0008] Furthermore, the step of extracting the signal component information generated in each iteration from the cable fault signal using the optimal noise figure includes: White noise, adjusted by the optimal noise figure, is injected into the signal to be decomposed in each iteration to obtain an enhanced decomposed signal; the signal to be decomposed reflects the residual generated in the previous iteration. Empirical mode decomposition is performed on the enhanced decomposed signal to generate IMF component information and residual signal.
[0009] Furthermore, performing empirical mode decomposition on the enhanced decomposed signal further includes: Perform first-order empirical mode decomposition on the white noise corresponding to the current iteration to obtain the corresponding first-order IMF components; The first-order IMF component is used as the original white noise injected into the signal to be decomposed in the next iteration.
[0010] Furthermore, the feature extraction from the target decomposition result to obtain cable fault features includes: Extract time-frequency feature sets from the target decomposition results, and input the time-frequency feature sets into a preset autoencoder; A loss function is introduced to optimize the training process of the autoencoder, driving the autoencoder to output the cable fault features.
[0011] Furthermore, the iteration termination condition includes: Calculate the energy ratio between the residual signal at the current iteration and the cable fault signal before empirical mode decomposition; The iteration terminates when the energy ratio meets the threshold condition.
[0012] Furthermore, after extracting features from the target decomposition results to obtain cable fault features, the process further includes: The cable fault characteristics are input into a pre-trained fault classification model, and the fault identification result of the target cable is output. The fault identification results are sent to the cable maintenance terminal.
[0013] Another embodiment of the present invention provides a multi-source feature extraction system for cable signals, comprising: The signal decomposition module is used to acquire the cable fault signal of the target cable and perform iterative empirical mode decomposition on the cable fault signal. The indicator acquisition module is used to acquire the denoising statistical indicators generated by the cable fault signal during each iteration. The noise optimization module is used to work with the denoising statistics to dynamically optimize the initial noise coefficient introduced in each iteration, so as to obtain the corresponding optimal noise coefficient. The component extraction module is used to extract signal component information generated in each iteration from the cable fault signal using the optimal noise figure. The decomposition termination module is used to determine the target decomposition result of the cable fault signal based on the signal component information when the iteration terminates. The feature extraction module is used to extract features from the target decomposition results to obtain cable fault features.
[0014] Another embodiment of the present invention provides a computer device including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor executes the computer program to implement the multi-source feature extraction method for cable signals as described above.
[0015] In another embodiment of the present invention, a computer-readable storage medium is provided, the computer-readable storage medium storing a computer program, wherein when the device containing the computer-readable storage medium executes the computer program, the multi-source feature extraction method for cable signals as described above is implemented.
[0016] Compared with the prior art, the beneficial effects of the embodiments of the present invention are at least one of the following: This invention utilizes a collaboratively improved empirical mode decomposition algorithm to decompose cable fault signals. In each iterative decomposition process, an objective function is constructed based on multiple indicators reflecting the denoising effect to achieve adaptive optimization of the noise figure, thus enabling adaptive adjustment of noise intensity and improving the adaptability of the decomposition process to complex and variable interference environments. By adjusting the injected signal noise using the optimal noise figure, the stability of the signal during the decomposition process is significantly enhanced, effectively suppressing mode aliasing and increasing the interpretability of each mode component. Furthermore, by extracting fault features from the generated component data, reliable data support is provided for the accurate diagnosis of subsequent cable faults. Attached Figure Description
[0017] Figure 1 This is a schematic flowchart of a multi-source feature extraction method for cable signals in one embodiment of the present invention; Figure 2 This is a schematic diagram of the multi-source feature extraction system for cable signals in one embodiment of the present invention; Figure 3 This is a structural block diagram of a preferred embodiment of a computer device provided by 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. The purpose of providing these embodiments is to make the disclosure of the present invention more thorough and comprehensive. 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] In the description of this application, the terms "first," "second," "third," etc., are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Therefore, a feature defined with "first," "second," "third," etc., may explicitly or implicitly include one or more of that feature. In the description of this application, unless otherwise stated, "a plurality of" means two or more.
[0020] In the description of this application, it should be noted that, unless otherwise expressly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to fixed connections, detachable connections, or integral connections; they can refer to mechanical connections or electrical connections; they can refer to direct connections or indirect connections through an intermediate medium; and they can refer to the internal communication between two components. The terms "vertical," "horizontal," "left," "right," "upper," "lower," and similar expressions used herein are for illustrative purposes only and do not indicate or imply that the device or component referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as limiting the invention. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items. Those skilled in the art can understand the specific meaning of the above terms in this application based on the specific circumstances.
[0021] In the description of this application, it should be noted that, unless otherwise defined, all technical and scientific terms used in this invention have the same meaning as commonly understood by one of ordinary skill in the art. The terminology used in this specification is for the purpose of describing specific embodiments only and is not intended to limit the invention. Those skilled in the art can understand the specific meaning of the above terms in this application based on the specific circumstances.
[0022] One embodiment of the present invention provides a method for multi-source feature extraction of cable signals. For details, please refer to [link to relevant documentation]. Figure 1 , Figure 1 The diagram shown is a flowchart of a multi-source feature extraction method for cable signals according to one embodiment of the present invention, including the following steps: S1. Obtain the cable fault signal of the target cable and perform iterative empirical mode decomposition on the cable fault signal.
[0023] In some embodiments of the present invention, cable fault signals generated when a target cable in the target power area experiences a fault are collected. Preferably, a Hall current sensor is used to simultaneously collect fault current signals of cables with different degrees of insulation degradation, such as partial discharges caused by insulation defects. The sampling frequency must meet the high-frequency component capture requirements (typical value ≥2MHz), and the cable signal length must cover the entire discharge process (≥10ms) to ensure effective capture of the cable fault signal.
[0024] After data acquisition, the raw cable fault signals are initialized. Specifically, the acquired cable fault signals are denoted as x(t), and zero-mean, unit-variance Gaussian white noise is added. Where i = 1, 2, ..., I, and I is the noise order. The resulting mixed cable fault signal after adding this white noise is expressed as: .in, The noise figure is the first decomposition. It determines the noise intensity added to the cable signal. It is a continuous parameter. If it is too small, it will cause the IMF (Intrinsic Mode Components) to be too smooth. If it is too large, it will cause the IMF to exhibit instability.
[0025] In this embodiment, the algorithm used for signal decomposition is the EEMD algorithm (Ensemble Empirical Mode Decomposition). To address the residual noise problem that occurs during the decomposition process, a noise figure optimization mechanism is designed. It should be understood that the empirical mode decomposition of the cable fault signal x(t) involves multiple iterative decomposition processes.
[0026] S2~S3. During each iteration, the denoising statistics of the cable fault signal are obtained. The initial noise coefficient introduced in each iteration is dynamically optimized based on the denoising statistics to obtain the corresponding optimal noise coefficient.
[0027] During the multiple iterative decomposition processes of EEMD, the following three statistical indicators are used for evaluation: Signal-to-noise ratio (SNR) is expressed as follows: SNR is the ratio of the signal power to the noise power in a cable fault, usually expressed in decibels (dB). An increase in SNR after denoising indicates that the noise has been effectively removed. For cable fault signal power, This represents the noise power. The higher the value, the better the noise reduction effect.
[0028] Mean Square Error (MSE) is expressed as follows: MSE is the average of the squares of the differences in cable fault signals before and after denoising, which measures the degree of distortion of the cable fault signal after denoising. It is the original cable fault signal. This is the noise-reduced cable fault signal. This is the length of the cable fault signal. The smaller the value, the better the noise reduction effect.
[0029] The correlation coefficient is expressed as follows: Among them, the correlation coefficient The linear relationship between the original cable fault signal and the denoised cable fault signal was measured. and These are the average values of the original cable fault signal and the noise-reduced cable fault signal, respectively. The higher the value, the better the noise reduction effect.
[0030] Based on the above three indicators, a weighted analysis is performed on the three denoising statistical indicators to construct an objective function. It is expressed as follows: in, These are the weights of the signal-to-noise ratio, mean square error, and correlation coefficient, respectively, and the sum of the three is 1.
[0031] The initial noise coefficient is optimized through gradient descent iterations with the objective function output as the goal. Specifically: the objective function is calculated. about The gradient of (where β refers to the noise coefficient in each iteration of decomposition) is expressed as: Gradient descent iterations using this formula represent only the iteration process of the noise figure in a single EMD decomposition (Empirical Mode Decomposition). Let... ( i = 1,2,… I ) is the first The noise figure of the second EMD decomposition is used to obtain optimal value Further iterative optimization calculations are then performed. The iterative process of gradient descent is represented as follows: in, For the first In the first EMD decomposition Second-rate( The noise figure during gradient descent iterations; For the ( +1) The noise figure value obtained after the gradient descent iteration update; The learning rate is set to 0.001, which is used to control the step size of gradient descent.
[0032] Based on this, the iteration terminates when the difference between the noise figure values calculated in two adjacent gradient descent iterations satisfies the preset convergence condition, and the solution is obtained. Optimal noise figure of the second EMD decomposition , means as follows: in, This indicates the range of values for the given noise figure.
[0033] The termination condition for gradient solving is expressed as: This step provides an example of the initial decomposition: For The whole was decomposed into EMD and obtained The first IMF component is IMF1(t). During the decomposition process, an objective function is established based on the aforementioned denoising statistical indices, and the collaborative gradient solution method is used for... The solution is performed to find its optimal value. And this optimal value... In the current decomposition stage, the target mode can be separated as cleanly as possible to ensure the "balance" of the modes.
[0034] S4. Extract the signal component information generated in each iteration from the cable fault signal with the optimal noise figure.
[0035] After finding the optimal noise coefficient for each iterative decomposition stage, the noise of the injected signal is optimized using this coefficient, and then the decomposition of the IMF components and residual calculation are performed.
[0036] Specifically: white noise, adjusted by the optimal noise figure, is injected into the signal to be decomposed in each iteration to obtain an enhanced decomposed signal. For example, in the first iteration (i.e., the initial decomposition), white noise adjusted by the optimal noise figure is injected into the signal to be decomposed (the original cable fault signal). The optimal noise figure determined during the first iteration of decomposition is injected. The enhanced decomposition signal is obtained: .
[0037] Empirical Mode Decomposition (EMD) is performed on the enhanced decomposed signal to generate IMF component information and residual signals, such as the first IMF component IMF1(t) generated during the initial decomposition, and the residual... The calculation process is as follows: In order to inject white noise into the signal to be decomposed in the next iteration, this embodiment performs first-order empirical mode decomposition on the white noise corresponding to the current iteration to obtain the corresponding first-order IMF component. Taking the initial decomposition process as an example: White noise Perform first-order EMD decomposition to obtain First-order IMF components And record Then, this first-order IMF component is used as the original white noise injected into the signal to be decomposed in the next iteration.
[0038] To elaborate on the above iterative decomposition and coefficient optimization process, the following is a specific example of the second EMD iterative decomposition. First, in this iteration, the noise figure is denoted as... ,right EMD decomposition was performed to obtain the second IMF component. During the decomposition process, obtain optimal value It can be seen that in each iteration, the residual generated in the previous iteration and the first-order IMF component will participate in the decomposition of the next iteration. In this embodiment, the signal to be decomposed injected with white noise in each iteration reflects the residual from the previous iteration.
[0039] Then, white noise Perform first-order EMD decomposition to obtain First-order IMF components And record .
[0040] Among them, the residual in this iteration The calculation process is as follows: In the third iteration, it will target Decompose it.
[0041] S5. When the iteration terminates, determine the target decomposition result of the cable fault signal based on the signal component information.
[0042] In this embodiment, the iteration process is repeated until the residual energy percentage meets the condition, at which point the iteration terminates. Therefore, for the ... The next iteration, specifically, is as follows: First, regarding the residuals Perform EMD decomposition to obtain No. IMF components During this process, the solution is obtained through gradient descent iteration of the objective function. optimal value .
[0043] Then, white noise Perform first-order EMD decomposition to obtain First-order IMF components And record .
[0044] No. The residual of the next iteration decomposition It can be expressed by the following formula: It should be understood that, in this embodiment, the residual energy ratio serves as a limitation on the number of iterations, thus balancing "decomposition sufficiency" and "IMF effectiveness." Specifically, the energy ratio between the residual signal at the current iteration and the cable fault signal before EMF decomposition is calculated, i.e. When the energy ratio meets the threshold condition (2% in this embodiment based on historical experience data), that is, when the energy ratio is ≤2%, the iteration terminates.
[0045] The cable fault signal underwent multiple iterative decompositions. By acquiring information from several IMF components, first-order IMF components, and residuals, the cable fault signal was reconstructed, yielding a target decomposition result suitable for feature extraction. These IMF components, having eliminated much noise and unimportant components, carry the main fault information from the original cable fault signal.
[0046] S6. Extract features from the target decomposition results to obtain cable fault features.
[0047] For the reconstructed signal and IMF components, multidimensional cable fault features are further extracted. Specifically, time-frequency feature sets are extracted from these target decomposition results and input into a preset autoencoder.
[0048] The time-frequency feature set includes: time-domain features such as mean, standard deviation, root mean square value, peak value, peak-to-peak value, peak factor, skewness, kurtosis, impulse factor, waveform factor, and margin factor; frequency-domain features such as center frequency, mean square frequency, root mean square frequency, frequency variance, frequency standard deviation, mean spectral kurtosis, standard deviation spectral kurtosis, skewness spectral kurtosis, and kurtosis; and time-frequency domain features such as energy entropy, permutation entropy, singular entropy, mean instantaneous frequency, and peak instantaneous amplitude.
[0049] After extraction, these feature data need to be cleaned and standardized to ensure data quality. Ultimately, a high-dimensional time-frequency feature set characterizing cable insulation faults is obtained. , n This represents the number of samples with different insulation conditions. m It refers to the number of feature indicators selected.
[0050] The self-encoder designed in this embodiment includes, respectively, components... L An encoder and decoder with 1 hidden layer, wherein each hidden layer of the encoder uses a non-linear activation function. f l Mapping high-dimensional data to a low-dimensional latent space can be represented as follows: In the formula, superscript l The first hidden layer of the encoder represents the... l layer, l =1, 2,…, L , L Indicates the number of hidden layers in the encoder. This represents the input sample data, where 0 indicates the input layer. This represents the encoder output. They represent the first l The weights and bias vectors of the layer encoder.
[0051] Each hidden layer of the decoder uses a non-linear activation function. g l This is used to reconstruct the data from the aforementioned low-dimensional latent space back into the original data space: In the formula, superscript l The hidden layer of the decoder represents the first l layer, l =1, 2,…, L , L Indicates the number of hidden layers in the decoder. They represent the first l Weights and biases of the layer decoder; Indicates the input for reconstruction; This represents the final reconstructed data.
[0052] To effectively measure the difference between the original sample data and the reconstructed sample data, a loss function is introduced to optimize the training process of the autoencoder, driving the autoencoder to output the cable fault features. In this embodiment, to prevent overfitting, a regularization term is added to the loss function, as follows: In the formula, It is the Frobenius norm; λ is the regularization coefficient; This is the initial loss function. It is the loss function of the regularization term. It is the reconstructed sample, that is, the output data of the decoder.
[0053] Train the autoencoder to minimize the loss function, let The value of has converged and no longer decreases with the increase of the number of hidden layer nodes.
[0054] In some embodiments of the present invention, t-SNE (t-distributed random neighborhood embedding) is also used to visualize and evaluate the dimensionality reduction effect, as shown below: in, It is a low-dimensional feature vector of the cable insulation fault data extracted by the encoder. It's data with reduced dimensionality.
[0055] t-SNE can be used to visualize the characteristics of different insulation degradation levels, enabling the visualization of dimensionality-reduced features. Y Mapping to 3D space allows samples with different degrees of insulation degradation to form distinct clusters.
[0056] These cable fault features can be used for subsequent cable operation and maintenance and diagnosis. For example, cable fault features can be input into a pre-trained fault classification model, which outputs the fault identification result for the target cable. The classification model can be a support vector machine, random forest, gradient boosting decision tree, or deep neural network. The data used for model training consists of feature vectors extracted from historical cases using the same process and their corresponding known fault labels (e.g., short circuit, open circuit, high resistance fault, low resistance fault, partial discharge, insulation dampness, etc.). The identification results are given in the form of a probability distribution or a defined type. Then, the fault identification results are sent to the cable operation and maintenance terminal, such as automatically generating standardized inspection work orders for use by operation and maintenance personnel.
[0057] In summary, this embodiment performs iterative empirical mode decomposition on cable fault signals. In each iteration, an objective function is constructed using the denoising statistics generated by the decomposition, and optimization algorithms such as gradient descent are used to dynamically optimize the noise coefficient introduced during the iteration process. The optimal noise coefficient obtained by this optimization is used to inject white noise into the signal to be decomposed to enhance the decomposition stability. Iterative empirical mode decomposition is performed to obtain intrinsic mode components and residual signals. Finally, deep feature mining is performed on the decomposed mode component data, which improves the robustness and accuracy of cable fault feature extraction under multi-source interference.
[0058] One embodiment of the present invention provides a multi-source feature extraction system for cable signals. For details, please refer to [link to documentation]. Figure 2 , Figure 2 The diagram shown illustrates the structure of a multi-source feature extraction system for cable signals according to one embodiment of the present invention, comprising: The signal decomposition module M1 is used to acquire the cable fault signal of the target cable and perform iterative empirical mode decomposition on the cable fault signal. The indicator acquisition module M2 is used to acquire the denoising statistical indicators generated by the cable fault signal during each iteration. The noise optimization module M3 is used to work with the denoising statistics to dynamically optimize the initial noise coefficient introduced in each iteration, so as to obtain the corresponding optimal noise coefficient. The component extraction module M4 is used to extract the signal component information generated in each iteration from the cable fault signal with the optimal noise figure. The decomposition termination module M5 is used to determine the target decomposition result of the cable fault signal based on the signal component information when the iteration terminates. The feature extraction module M6 is used to extract features from the target decomposition results to obtain cable fault features.
[0059] like Figure 3 As shown, this embodiment of the invention also provides a computer device. Figure 3 This is a structural block diagram of a preferred embodiment of a computer device provided by the present invention. The computer device includes a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, it implements the method described above.
[0060] Preferably, the computer program can be divided into one or more modules / units (such as computer program 1, computer program 2, ...), and the one or more modules / units are stored in the memory and executed by the processor to complete the present invention. The one or more modules / units can be a series of computer program instruction segments capable of performing specific functions, and the instruction segments are used to describe the execution process of the computer program in the computer device.
[0061] The processor can be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor, or the processor can be any conventional processor. The processor is the control center of the terminal device, connecting various parts of the terminal device through various interfaces and lines.
[0062] The memory mainly includes a program storage area and a data storage area. The program storage area can store the operating system, applications required for at least one function, etc., while the data storage area can store related data, etc. Furthermore, the memory can be a high-speed random access memory, or a non-volatile memory, such as a plug-in hard drive, a SmartMedia Card (SMC), a Secure Digital (SD) card, and a Flash Card, or other volatile solid-state storage devices.
[0063] It should be noted that the aforementioned terminal devices may include, but are not limited to, processors and memory, as will be understood by those skilled in the art. Figure 3 The structural block diagram is merely an example of a terminal device and does not constitute a limitation on the terminal device. It may include more or fewer components than shown, or combine certain components, or use different components. Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. The storage medium may be a magnetic disk, optical disk, read-only memory (ROM), or random access memory (RAM), etc.
[0064] Accordingly, embodiments of the present invention provide a computer-readable storage medium, the computer-readable storage medium including a stored computer program, wherein, when the computer program is executed, it controls the device where the computer-readable storage medium is located to perform the steps in the method of the above embodiments, for example... Figure 1 Steps S1 to S6 as described above.
[0065] The technical features and effects of the multi-source feature extraction system for cable signals proposed in this embodiment of the invention are the same as those of the multi-source feature extraction method for cable signals proposed in this embodiment of the invention, and will not be repeated here.
[0066] The embodiments described above are merely illustrative of several implementations of the present invention, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of the present invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these modifications and improvements all fall within the scope of protection of the present invention. Therefore, the scope of protection of this patent should be determined by the appended claims.
Claims
1. A method for multi-source feature extraction of cable signals, characterized in that, include: Acquire the cable fault signal of the target cable and perform iterative empirical mode decomposition on the cable fault signal; In each iteration, the denoising statistics of the cable fault signal are obtained; The initial noise coefficient introduced in each iteration is dynamically optimized in conjunction with the aforementioned denoising statistical index to obtain the corresponding optimal noise coefficient; Using the optimal noise figure, extract the signal component information generated in each iteration from the cable fault signal; When the iteration terminates, the target decomposition result of the cable fault signal is determined based on the signal component information; Feature extraction is performed on the target decomposition results to obtain cable fault features.
2. The multi-source feature extraction method for cable signals as described in claim 1, characterized in that, The method of dynamically optimizing the initial noise coefficient introduced in each iteration in conjunction with the denoising statistical index to obtain the corresponding optimal noise coefficient includes: A weighted analysis is performed on the denoising statistical indicators to construct an objective function, and gradient descent iteration is performed on the initial noise coefficient with the goal of maximizing the output of the objective function. The iteration terminates when the difference between the noise figure values calculated in two consecutive gradient descent iterations meets the preset convergence condition, and the optimal noise figure is determined.
3. The multi-source feature extraction method for cable signals as described in claim 1, characterized in that, The step of extracting signal component information generated in each iteration from the cable fault signal using the optimal noise figure includes: White noise, adjusted by the optimal noise figure, is injected into the signal to be decomposed in each iteration to obtain an enhanced decomposed signal; the signal to be decomposed reflects the residual generated in the previous iteration. Empirical mode decomposition is performed on the enhanced decomposed signal to generate IMF component information and residual signal corresponding to the current iteration.
4. The multi-source feature extraction method for cable signals as described in claim 3, characterized in that, The step of performing empirical mode decomposition on the enhanced decomposed signal further includes: Perform first-order empirical mode decomposition on the white noise corresponding to the current iteration to obtain the corresponding first-order IMF components; The first-order IMF component is used as the original white noise injected into the signal to be decomposed in the next iteration.
5. The multi-source feature extraction method for cable signals as described in claim 1, characterized in that, The feature extraction from the target decomposition result to obtain cable fault features includes: Extract time-frequency feature sets from the target decomposition results, and input the time-frequency feature sets into a preset autoencoder; A loss function is introduced to optimize the training process of the autoencoder, driving the autoencoder to output the cable fault features.
6. The multi-source feature extraction method for cable signals as described in claim 3, characterized in that, The iteration termination condition includes: Calculate the energy ratio between the residual signal at the current iteration and the cable fault signal before empirical mode decomposition; The iteration terminates when the energy ratio meets the threshold condition.
7. The multi-source feature extraction method for cable signals as described in claim 1, characterized in that, After extracting features from the target decomposition results to obtain cable fault features, the process further includes: The cable fault characteristics are input into a pre-trained fault classification model, and the fault identification result of the target cable is output. The fault identification results are sent to the cable maintenance terminal.
8. A multi-source feature extraction system for cable signals, characterized in that, include: The signal decomposition module is used to acquire the cable fault signal of the target cable and perform iterative empirical mode decomposition on the cable fault signal. The indicator acquisition module is used to acquire the denoising statistical indicators generated by the cable fault signal during each iteration. The noise optimization module is used to work with the denoising statistics to dynamically optimize the initial noise coefficient introduced in each iteration, so as to obtain the corresponding optimal noise coefficient. The component extraction module is used to extract signal component information generated in each iteration from the cable fault signal using the optimal noise figure. The decomposition termination module is used to determine the target decomposition result of the cable fault signal based on the signal component information when the iteration terminates. The feature extraction module is used to extract features from the target decomposition results to obtain cable fault features.
9. A computer device, characterized in that, The method includes a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor, when executing the computer program, implements the multi-source feature extraction method for cable signals as described in any one of claims 1 to 7.
10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program, wherein when the device containing the computer-readable storage medium executes the computer program, it implements the multi-source feature extraction method for cable signals as described in any one of claims 1 to 7.