A cable accessory aging detection method based on multi-parameter fusion wavelet transform

Abstract

The application discloses a kind of cable accessory aging detection methods based on multi-parameter fusion wavelet transform, comprising: establishing cable accessory ac steady-state equivalent circuit model, obtains the current and voltage dynamic data of target cable accessory sample at test temperature;Signal is decomposed using wavelet transform, and dynamic threshold is set according to the amplitude difference of wavelet coefficient under each scale, and noise and target signal are separated;Through signal reconstruction, the dielectric property change data of high signal-to-noise ratio are extracted, and the aging state of cable accessory is analyzed in combination with dielectric loss angle.The application solves the problem that the detection period of traditional method is long and easy to be disturbed, significantly improves the accuracy and real-time performance of cable accessory aging degradation monitoring, and can effectively prevent power safety accidents.

Description

Technical Field

[0001] This invention relates to power cable condition monitoring technology, specifically to an aging detection method for cable accessories based on multi-parameter fusion wavelet transform. Background Technology

[0002] The aging and deterioration of cable accessory insulation materials is a key problem that the power system has long faced. It directly leads to a decline in electrical performance, a reduction in mechanical strength, and even serious accidents such as short circuits and fires. At present, traditional detection methods such as partial discharge monitoring and infrared thermometry have the following limitations: (1) Complex operation: It requires complex equipment and manual intervention, making it difficult to achieve automated monitoring; (2) Long detection cycle: Offline detection cannot meet the real-time requirements; (3) Susceptible to environmental interference: Noise interference causes signal distortion, affecting the accuracy of aging assessment.

[0003] In the existing technology, although detection methods based on changes in dielectric properties can indirectly reflect the degree of insulation aging, they are severely affected by noise interference and lack efficient noise reduction methods. Summary of the Invention

[0004] Purpose of the invention: The purpose of this invention is to provide a cable accessory aging detection method based on multi-parameter fusion wavelet transform, which can solve the problems of long detection cycle and weak anti-noise interference capability of traditional technology, and realize accurate judgment of the aging and deterioration degree of cable accessories.

[0005] Technical solution: The present invention provides a cable accessory aging detection method based on multi-parameter fusion wavelet transform, comprising:

[0006] An AC steady-state equivalent circuit model of the cable accessory is established. Dynamic data of current and voltage of the target cable accessory sample as a function of test temperature and test time are obtained from the AC steady-state equivalent circuit model of the cable accessory. The calculation formula of the dielectric loss angle is derived through the sinusoidal steady-state voltage-current relationship matrix. The dielectric loss angle of the cable accessory is calculated using the calculation formula of the dielectric loss angle.

[0007] Wavelet transform decomposition was performed on the dynamic data of current and voltage to obtain multi-scale wavelet coefficients;

[0008] Based on the normal distribution characteristics of noise signals and the multi-parameter fusion mechanism, a dynamic threshold is set by combining the amplitude differences of wavelet coefficients at different scales; the wavelet coefficients of the target signal are selected from the wavelet coefficients at multiple scales using the dynamic threshold, and noise in the target signal is filtered out.

[0009] The filtered wavelet coefficients are reconstructed to extract dielectric property change data from the reconstructed signal.

[0010] By combining data on changes in dielectric properties with dielectric loss angle, the degree of aging and deterioration of the insulation material of cable accessories can be assessed.

[0011] Furthermore, the process of establishing the AC steady-state equivalent circuit model of the cable accessory is as follows:

[0012] When the cable accessory is operating normally, a small segment dx is taken at a distance X from the beginning. The AC steady-state equivalent circuit model of the cable accessory includes three sets of series-connected small segments with line resistance R0dx and line inductance L0dx. The output terminal of the first set, the input and output terminals of the second set, and the input terminal of the third set are all connected to a set of parallel small segment conductance to ground G0dx and small segment capacitance to ground C0dx.

[0013] Furthermore, the db4 wavelet basis function is used to perform wavelet transform decomposition on the current and voltage dynamic data to obtain multi-scale wavelet coefficients.

[0014] Furthermore, based on the normal distribution characteristics of the noise signal and the multi-parameter fusion mechanism, a dynamic threshold is set by combining the amplitude differences of wavelet coefficients at various scales, including:

[0015] The standard deviation σ of the noise wavelet coefficients at the nth layer is calculated based on the statistical distribution characteristics of the noise wavelet coefficients. n This is used to characterize the intensity fluctuation of noise at this scale;

[0016] Calculate the signal energy E of the nth layer s,n With the noise energy E of the nth layer n,n The ratio is used to reflect the difference in energy distribution between the target signal and noise at a specific scale;

[0017] Through formula Determine the adaptive dynamic threshold ζ for the nth layer n Where υ is the basic attenuation factor and α is the adaptive adjustment factor;

[0018] The modulus maxima of the wavelet coefficients of noise decrease with increasing scale, while the modulus maxima of the wavelet coefficients of the target signal increase with increasing scale. The adaptive dynamic threshold ζ of the nth layer... n It decays with scale.

[0019] Furthermore, the standard deviation σ of the noise wavelet coefficients at the nth layer is calculated using the statistical distribution characteristics of the noise wavelet coefficients. n Characterized by noise intensity fluctuations at this scale, including:

[0020] In the initial decomposition layer, assuming noise is dominant, the wavelet coefficients D of the initial decomposition layer are extracted. n,k k is the coefficient index;

[0021] The standard deviation σ of the nth layer noise wavelet coefficients was calculated using statistical methods. n :

[0022]

[0023] in, N is the total number of wavelet coefficients;

[0024] For subsequent scales n>1, the noise standard deviation decreases according to an empirical formula:

[0025] σ n =σ n-1 ·ν

[0026] Here, ν is the basic attenuation factor, which reflects the attenuation characteristics of noise energy as the scale increases.

[0027] Furthermore, the E of the signal energy of the nth layer s,n The expression is as follows:

[0028]

[0029] in, The wavelet coefficients retained after denoising; N is the total number of wavelet coefficients;

[0030] The noise energy E at layer n n,n The expression is as follows:

[0031]

[0032] Among them, D n,k These are the wavelet coefficients of the nth layer before denoising.

[0033] Furthermore, the adaptive adjustment factor α is used to balance scale decay and energy weight, and its value ranges from 0.1 to 0.3.

[0034] Furthermore, dynamic thresholding is used to select wavelet coefficients of the target signal from multi-scale wavelet coefficients, and noise in the target signal is filtered out, including:

[0035] If the wavelet coefficients at the current scale are greater than the adaptive dynamic threshold ζ n If the wavelet coefficients at the current scale are less than the adaptive dynamic threshold ζ, then the target signal is retained. n It is then identified as noise and filtered out, where the adaptive dynamic threshold ζ n Dynamically adjusted through a multi-parameter fusion mechanism.

[0036] Furthermore, the dielectric performance change data includes the dielectric loss factor and capacitance change rate, which are used to quantitatively analyze the degree of degradation of the insulation material of cable accessories.

[0037] Furthermore, the calculation formula for the dielectric loss angle is derived through the sinusoidal steady-state voltage-current relationship matrix. Using this formula, the dielectric loss angle of the cable accessories is calculated, including:

[0038] Considering that the voltage and current waveforms of the cable accessories are sinusoidal steady-state waveforms, the voltage and current relationship matrix across the cable accessory line is established as follows:

[0039]

[0040] In the formula, γ is the propagation coefficient of the line; Z0 is the wave impedance of the line; e is the natural logarithm; and the expressions for the propagation coefficient γ and the wave impedance Z0 of the line are as follows:

[0041]

[0042] The voltage-current relationship matrix at both ends of the cable accessory line is obtained as follows:

[0043]

[0044] The dielectric loss angle of cable accessories satisfies the following formula:

[0045]

[0046] From the above formula, we can obtain:

[0047]

[0048] As can be seen from the above formula, the ratio of the current difference between the two ends of the cable accessory to the average voltage at the two ends will not change due to the load current flowing through the cable accessory; the complementary angle of the phase angle difference between the current difference between the two ends of the cable accessory and the average voltage at the two ends is taken as the dielectric loss angle of the cable accessory.

[0049] Beneficial effects: Compared with the prior art, the significant technical effects of the present invention are as follows: (1) Multi-parameter fusion dynamic threshold technology: By introducing the noise standard deviation σ n Quantize noise intensity, combined with signal-to-noise energy ratio E s,n / E n,nReflecting the differences in energy distribution between signal and noise at different scales, and using an adaptive factor α (range 0.1 to 0.3) to dynamically adjust the threshold weight, the traditional denoising method relies on manual experience to adjust the threshold and relies solely on scale attenuation characteristics, resulting in difficulty in balancing noise suppression and signal preservation, weak anti-interference ability, and low detection accuracy. It achieves accurate separation of noise and target signal, improves signal-to-noise ratio, reduces the calculation error of dielectric loss angle, and significantly enhances the accuracy of aging assessment; (2) Adaptive adjustment mechanism: Based on experimental calibration of the adaptive factor α (0.1 to 0.3), combined with the signal characteristics under different noise environments, the threshold attenuation rate and energy weight allocation are dynamically optimized: This solves the problem that the traditional method has poor adaptability in complex noise scenarios, requires repeated manual adjustment of parameters, and is cumbersome and inefficient. To achieve adaptability to high-noise (such as industrial environment) and low-noise (such as laboratory) scenarios, improve the adaptive matching rate of denoising parameters, reduce manual intervention, and enhance the practicality and universality of the method; (3) Simplify the operation process and achieve automated detection through the following technical means: Automated data acquisition: Use high-precision sensors to collect current and voltage dynamic data in real time, and automatically calculate the dielectric loss angle through equivalent circuit model to reduce manual intervention; Intelligent signal processing: Based on multi-parameter fusion dynamic threshold algorithm, automatically separate noise and target signal without relying on manual experience to adjust denoising parameters; One-click aging assessment: Real-time online assessment of aging status is achieved through automatic extraction and formulaic analysis of dielectric performance data. It solves the operational complexity problems of traditional methods, which are mainly reflected in data acquisition relying on manual operation, denoising parameters needing experience adjustment, and aging assessment requiring manual calculation of dielectric parameters. Attached Figure Description

[0050] Figure 1 This is a schematic diagram of the process of the present invention;

[0051] Figure 2 A schematic diagram of the AC steady-state equivalent circuit model for cable accessories;

[0052] Figure 3 A schematic diagram of the process for generating the target signal using a multi-parameter fusion threshold. Detailed Implementation

[0053] The technical solution of the present invention will now be described in detail with reference to specific embodiments and accompanying drawings.

[0054] To address the problem of frequent power safety accidents caused by the inability of existing technologies to accurately determine the degree of aging and deterioration of cable accessories, a cable accessory aging detection method based on multi-parameter fusion wavelet transform is proposed. Figure 1 As shown, the cable accessory aging detection method based on multi-parameter fusion wavelet transform of the present invention includes the following steps:

[0055] S1. Establish an AC steady-state equivalent circuit model for the cable accessory, and obtain dynamic data of current and voltage of the target cable accessory sample as a function of test temperature and test time from the AC steady-state equivalent circuit model for the cable accessory; derive the calculation formula of the dielectric loss angle through the sinusoidal steady-state voltage-current relationship matrix, and calculate the dielectric loss angle of the cable accessory using the calculation formula of the dielectric loss angle.

[0056] When the cable accessories are operating normally, a small segment dx is taken at a distance X from the beginning, and a structure is constructed as follows: Figure 2 The AC steady-state equivalent circuit model of the cable accessory is shown. This model includes three sets of series-connected micro-segments with along-line resistance R0dx and along-line inductance L0dx. The output of the first set, the input and output of the second set, and the input of the third set are all connected to a set of parallel micro-segment conductance to ground G0dx and capacitance to ground C0dx. In the model, U1 and I1 represent the voltage and current at the beginning of the cable accessory, respectively, and U2 and I2 represent the voltage and current at the end of the cable accessory, respectively. This AC steady-state equivalent circuit model of the cable accessory covers parameters such as along-line resistance, inductance, conductance to ground, and capacitance to ground of the micro-segment. These parameters can be obtained from experimental measurements, cable specification manuals, or historical data. For example, the along-line resistance (R0dx) of the micro-segment can be calculated based on the resistivity, cross-sectional area, and length of the cable accessory; the capacitance to ground (C0dx) can be calculated based on the geometry of the cable accessory and the dielectric constant of the insulating material.

[0057] Under the test temperature conditions, current and voltage change data of the target cable accessory samples were collected during the test time.

[0058] In this embodiment, the calculation formula for the dielectric loss angle is derived from the sinusoidal steady-state voltage-current relationship matrix, and the dielectric loss angle of the cable accessory is calculated using this formula. Specifically:

[0059] For high-voltage cable accessories, the voltage-current relationship per unit length is shown in the following formula:

[0060]

[0061] Considering that the voltage and current waveforms of the cable accessories are sinusoidal steady-state waveforms, the voltage and current relationship matrix across the cable accessory line is established as follows:

[0062]

[0063] In the formula, γ is the propagation coefficient of the line; Z0 is the wave impedance of the line; e is the natural logarithm; and the expressions for the propagation coefficient γ and the wave impedance Z0 of the line are as follows:

[0064]

[0065] The voltage-current relationship matrix at both ends of the cable accessory line is obtained as follows:

[0066]

[0067] The dielectric loss angle of cable accessories satisfies the following formula:

[0068]

[0069] From the above formula, we can obtain:

[0070]

[0071] As can be seen from the above formula, the ratio of the current difference between the two ends of the cable accessory to the average voltage at the two ends will not change as the load current flows through the cable accessory; the complementary angle of the phase angle difference between the current difference between the two ends of the cable accessory and the average voltage at the two ends can be used as the dielectric loss angle of the cable accessory, which can effectively determine its insulation aging degree.

[0072] S2. Perform wavelet transform decomposition on the current and voltage dynamic data to obtain multi-scale wavelet coefficients.

[0073] Wavelet transform decomposition was performed on dynamic current and voltage data containing noise. After wavelet transform, wavelet coefficients at different scales can be obtained, and the target signal and noise signal show significant amplitude differences at most scales. The target signal amplitude is relatively high at most scales, while the noise signal amplitude decays rapidly with scale, approaching zero. The first step of the decomposition is to select appropriate wavelet basis functions so that the signal energy can be distributed on a few basis functions. This property of wavelets mainly depends on the regularity of the signal function, the order of the vanishing moments of the wavelet function, and the size of the support. Daubechies wavelets (dbN) are widely used in signal processing due to their excellent regularity and sparse basis properties. In this invention, after multiple experiments and analyses, the db4 wavelet basis function was selected to perform wavelet transform decomposition on the dynamic current and voltage data. The advantages of choosing the db4 wavelet basis function are as follows: (1) The db4 wavelet basis function can maintain high smoothness during signal reconstruction and reduce signal distortion; (2) The vanishing moment order of the db4 wavelet basis function is 4, which is moderate and can effectively extract the feature information of the signal while avoiding overfitting; (3) The support size of the db4 wavelet basis function is 8, which can reduce the computational complexity while maintaining the local features of the signal. The number of decomposition layers was verified by experiments and determined to be 5 layers. The db4 wavelet basis function was used to decompose the signal into 5 layers to ensure that the features of the signal are fully extracted at different scales.

[0074] S3. Based on the normal distribution characteristics of the noise signal and the multi-parameter fusion mechanism, a dynamic threshold is set by combining the amplitude differences of wavelet coefficients at each scale; the wavelet coefficients of the target signal are selected from the wavelet coefficients of the multi-scale wavelet coefficients using the dynamic threshold, and the noise in the target signal is filtered out.

[0075] When using wavelet transform to remove noise interference, signals and noise exhibit different characteristics at different scales. For continuous signal functions, wavelet transform coefficients increase with scale; however, wavelet transform coefficients for noise decrease with scale. Based on this characteristic, in wavelet space, the two can be distinguished by the different propagation characteristics of the wavelet coefficients of signals and noise as they change with scale. Wavelet threshold denoising is an effective denoising method. It sets a threshold and removes only the details exceeding that threshold, thereby using wavelets to perform threshold denoising on the signal and extracting the fundamental signal.

[0076] like Figure 3 As shown, the specific implementation process of step S3 is as follows:

[0077] S3.1. Based on the normal distribution characteristics of the noise signal and the multi-parameter fusion mechanism, a dynamic threshold is set by combining the amplitude differences of wavelet coefficients at various scales. The specific steps are as follows:

[0078] S3.1.1 Calculate the standard deviation σ of the nth layer noise wavelet coefficients using the statistical distribution characteristics of the noise wavelet coefficients. n This is used to characterize the intensity fluctuation of noise at this scale; the specific steps are as follows:

[0079] S3.1.1.1 In the initial decomposition layer (e.g., the highest frequency layer n=1), assuming noise is dominant, extract the wavelet coefficients D of the initial decomposition layer. n,k k is the coefficient index.

[0080] S3.1.1.2 Calculate the standard deviation σ of the nth layer noise wavelet coefficients using statistical methods. n :

[0081]

[0082] in, N is the total number of wavelet coefficients;

[0083] S3.1.1.3 For subsequent scales n>1, the noise standard deviation decreases according to an empirical formula:

[0084] σ n =σ n-1 ·v

[0085] Among them, ν is the basic attenuation factor (with a value of 0.5 to 0.8), which reflects the attenuation characteristics of noise energy as the scale increases.

[0086] The smaller the noise standard deviation, the faster the noise energy decays, and the more stringent the threshold setting needs to be to suppress residual noise.

[0087] S3.1.2 Calculate the signal energy E of the nth layer. s,n With the noise energy E of the nth layer n,n The ratio is used to reflect the difference in energy distribution between the target signal and noise at a specific scale.

[0088] In this embodiment, the signal energy E of the nth layer is calculated from the residual signal after preliminary threshold denoising (such as the hard threshold method). s,n The signal energy E of the nth layer s,n The expression is as follows:

[0089]

[0090] in, N represents the wavelet coefficients retained after denoising, and N is the total number of wavelet coefficients.

[0091] In this embodiment, the noise energy E of the nth layer is calculated using the wavelet coefficients of the denoised signal. n,n The noise energy E of the nth layer n,n The expression is as follows:

[0092]

[0093] in, D represents the wavelet coefficients retained after denoising. n,k Let N be the wavelet coefficients of the nth layer before denoising, where N is the total number of wavelet coefficients.

[0094] A higher energy ratio indicates that the signal dominates at that scale, requiring a higher threshold to retain the effective signal; conversely, a lower energy ratio requires a lower threshold to enhance denoising capabilities.

[0095] S3.1.3, through formula Determine the adaptive dynamic threshold ζ for the nth layer n , where υ is the basic attenuation factor (ranging from 0.5 to 0.8), and α is the adaptive adjustment factor.

[0096] S3.1.4. The modulus maxima of the wavelet coefficients of noise decrease with increasing scale, while the modulus maxima of the wavelet coefficients of the target signal increase with increasing scale. The adaptive dynamic threshold ζ of the nth layer... n It decays with scale.

[0097] In this embodiment, the adaptive adjustment factor α is used to balance scale attenuation and energy weighting. Specifically, the adaptive adjustment factor α is the weighting coefficient of the signal-to-noise energy ratio contribution to the threshold. The larger the value of α, the more significant the influence of the energy ratio on the threshold, which is suitable for environments with weak signals and complex noise; conversely, it emphasizes scale attenuation characteristics.

[0098] In this embodiment, the adaptive adjustment factor α is determined as follows:

[0099] (1) Select typical cable accessory samples (such as new cables and severely aged cables) and apply different noise intensity interference to them respectively.

[0100] (2) Through comparative experiments, the optimal α value was determined using signal-to-noise ratio (SNR) and dielectric loss angle error as evaluation indicators.

[0101] (3) Experimental results show that the denoising effect is best when the adaptive adjustment factor α is in the range of 0.1 to 0.3.

[0102] S3.2. Use dynamic thresholding to select wavelet coefficients of the target signal from multi-scale wavelet coefficients and filter out noise in the target signal.

[0103] If the wavelet coefficients at the current scale are greater than the adaptive dynamic threshold ζ n If the wavelet coefficients at the current scale are less than the adaptive dynamic threshold ζ, then the target signal is retained. n It is then identified as noise and filtered out, where the adaptive dynamic threshold ζ n Dynamically adjusted through a multi-parameter fusion mechanism.

[0104] As the scale increases, the threshold gradually decreases to ensure that noise signals are effectively filtered out while the target signal is preserved. Setting different thresholds at different scales allows noise signals to be attenuated and filtered out as much as possible while the target signal is better preserved and recorded, resulting in a higher signal-to-noise ratio and increasing the accuracy of the measurement signal.

[0105] In step S3, appropriate threshold values ​​are set based on the wavelet coefficients of the signal at various scales, and these thresholds are used to separate and filter the target signal from the noise signal. The dynamic threshold setting method is as follows: based on the normal distribution characteristics of the noise signal, threshold values ​​are set at different scales. Specifically, the maximum modulus of the wavelet coefficients of the noise decreases with increasing scale, while the maximum modulus of the wavelet coefficients of the target signal increases with increasing scale. By analyzing the normal distribution characteristics of the noise signal, the threshold at each scale can be determined. This invention utilizes a multi-parameter fusion mechanism, introducing a signal-to-noise energy ratio on top of the original denoising method. This ratio reflects the energy proportion of the signal and noise at a specific scale. In the low-frequency layer (signal-dominated), the energy ratio increases significantly, and the threshold is raised to retain more effective signals; in the high-frequency layer (noise-dominated), the energy ratio decreases, and the threshold is lowered to enhance denoising capabilities. Simultaneously, an adaptive adjustment factor is introduced to flexibly adjust the contribution of the energy ratio to the threshold, adapting to different noise environments.

[0106] In step S3, by introducing the noise standard deviation, signal-to-noise energy ratio and adaptive factor, the dynamic threshold setting is more in line with the actual signal characteristics, which solves the limitation of the traditional method that relies solely on scale attenuation and significantly improves the denoising accuracy and signal fidelity.

[0107] S4. Reconstruct the wavelet coefficients after screening and extract the dielectric property change data from the reconstructed signal.

[0108] The new wavelet coefficient sequence after analysis and processing is rearranged to finally obtain the complete target signal after noise removal.

[0109] Dielectric property change data, including dielectric loss factor and capacitance change rate, are used to quantitatively analyze the degree of degradation of cable accessory insulation materials.

[0110] S5. Combine dielectric performance change data with dielectric loss angle to assess the degree of aging and deterioration of cable accessory insulation materials.

[0111] This invention constructs an AC steady-state equivalent circuit model for cable accessories, combines wavelet transform signal decomposition, adaptive dynamic threshold screening, and dielectric performance analysis to achieve real-time and accurate assessment of the aging state of cable accessory insulation materials. It is particularly suitable for the deterioration detection and power safety early warning of XLPE cable accessories.

[0112] The wavelet transform-based dynamic threshold denoising technology proposed in this invention significantly improves signal fidelity and the accuracy of aging assessment by using a multi-parameter fusion mechanism (noise standard deviation, signal-to-noise energy ratio, and adaptive adjustment factor) combined with dielectric loss angle analysis, providing an innovative solution for real-time status monitoring of cable accessories.

[0113] This invention solves the problems of long detection cycles and susceptibility to interference in traditional methods, significantly improving the accuracy and real-time performance of monitoring cable accessory aging and deterioration, and effectively preventing power safety accidents.

[0114] The following specific embodiments verify the technical effectiveness of the cable accessory aging detection method based on multi-parameter fusion wavelet transform of the present invention.

[0115] This embodiment aims to demonstrate how to use a wavelet transform-based method for detecting the aging and deterioration of cable accessory insulation materials to assess the aging of actual cable accessories and select representative cable accessory samples for detailed testing and analysis.

[0116] I. Sample Selection

[0117] From the cable accessory samples included in the State Grid's report on the operational performance analysis of aging power cables, two cable accessories were selected for this test: a 110kV high-voltage cable accessory (25 years of service) and a 35kV medium-voltage cable accessory (0 years of service, representing a new cable accessory). These two cable accessories were chosen because their significantly different service lives help to clearly demonstrate the application effects and differences of the testing methods on cable accessories with different aging levels.

[0118] II. Establishing an AC steady-state equivalent circuit model for cable accessories

[0119] Under normal operating conditions, a small segment dx is taken at a distance X from the beginning of the cable accessory to construct an AC steady-state equivalent circuit model of the cable accessory. For a 110kV high-voltage cable accessory and a 35kV cable accessory, the key parameters in the model are determined: resistance (R0dx), inductance (L0dx), conductance to ground (G0dx), and capacitance to ground (C0dx) along the small segment. These parameters are determined based on the specific specifications, material characteristics, and actual operating environment of the cable accessory, obtained by consulting relevant technical documents, referring to historical data of similar cables, or conducting on-site measurements. For example, for a 110kV high-voltage cable that has been in operation for 25 years, after detailed review of its cable accessory design documents, reference to parameters of similar cable accessories under similar operating conditions, and actual on-site measurements, its R0dx = 0.01Ω, L0dx = 0.1mH, and G0dx = 10Ω were determined. -6 S, C0dx = 0.1μF; For a 35kV medium-voltage cable accessory, referring to the standard parameters of similar new cable accessories and making minor adjustments based on actual conditions, its R0dx = 0.008Ω, L0dx = 0.08mH, and G0dx = 8×10 -7 S, C0dx = 0.08 μF.

[0120] The dielectric loss angle is derived using a sinusoidal steady-state voltage-current relationship matrix. Considering that the voltage and current waveforms of cable accessories exhibit sinusoidal steady-state waveforms, a voltage-current relationship matrix is ​​constructed between the two ends of the cable accessory line. Through calculations using relevant formulas, including parameters such as propagation coefficient (γ) and wave impedance (Z0), the dielectric loss angle is gradually obtained. For a certain 110kV high-voltage cable accessory, its propagation coefficient γ = 0.01 + j0.05 and wave impedance Z0 = 50 + j20 are obtained after precise calculation. According to the standard calculation formula:

[0121]

[0122] Where ε' is the relative permittivity; ε” is the loss index; ω is the angular frequency; ε0 is the vacuum permittivity; ε r ρ is the relative permittivity; g is the equivalent conductivity of the dielectric relaxation polarization loss; I p I represents the active component of the total dielectric current under an alternating electric field. q This represents the reactive component loss of the total dielectric current under an alternating electric field. The calculated dielectric loss angle is 3°, with a tangent of approximately 0.05. For a 35kV medium-voltage cable accessory, similar rigorous calculations yielded a dielectric loss angle of 1.5° and a tangent of approximately 0.02.

[0123] III. Acquiring Dynamic Current and Voltage Data

[0124] Under a set test temperature of 25℃, an AC voltage of 50Hz and 10kV was applied to two selected cable accessory samples to simulate the actual operating conditions of the cable accessories. High-precision current and voltage sensors were used to continuously collect dynamic data of current and voltage changes over 100 seconds at 0.01s intervals. During data acquisition, the precise installation position of the sensors was strictly ensured, and effective shielding measures were taken to avoid external interference, thereby guaranteeing the high accuracy and reliability of the collected data. The collected data was initially processed and recorded; some data is shown below:

[0125]

[0126] IV. Wavelet Transform Decomposition of Signals

[0127] Based on the description of wavelet basis function selection in the patent, and combined with actual signal characteristics and processing experience, the db4 wavelet basis function was selected to perform wavelet transform decomposition on the acquired current and voltage dynamic data. During the decomposition process, after multiple experiments and analyses, a suitable decomposition level of 5 levels was determined, which allows for the full extraction of signal feature information. Through wavelet transform, the original signal was decomposed into different frequency sub-bands, causing the target signal and noise signal to exhibit significantly different characteristics at different scales, laying a solid foundation for subsequent denoising and feature extraction.

[0128] V. Setting Dynamic Thresholds and Noise Reduction

[0129] Based on the normal distribution characteristics of the noise signal and the new dynamic threshold selection equation, thresholds are set at different scales. The maximum modulus of the wavelet transform coefficients of the noise signal at the first scale is calculated to be ζ1. For a 110kV high-voltage cable, σ1 = 0.5, and the signal energy E after the first layer of denoising is... s,1 =120, noise energy E n,1 =60, then the energy ratio E s，1 / E n,1 =2, ν=0.6, the adaptive adjustment factor α was calibrated through multiple experiments and finally selected as α=0.2 to balance the weight of signal preservation and noise suppression. Then ζ1=0.6×0.5×(1+0.2×2)=0.42; for a new 35kV cable, ζ1=0.3, and the initial threshold is calculated with reference to the relevant formula in the patent. As the scale increases, the threshold is reasonably adjusted according to the principle of attenuation of the wavelet coefficient modulus maxima of noise and enhancement of the wavelet coefficient modulus maxima of the target signal. Taking the third layer as an example, after detailed analysis and calculation, taking a 110kV high-voltage cable accessory as an example, σ3=0.2, E s,3 / E n,3 =5, then ζ3 = 0.6 × 0.2 × (1 + 0.2 × 5) = 0.24. Similarly, the threshold for the new 35kV cable is set to 0.15. During the denoising process, by comparing thresholds, when the wavelet coefficient at a certain scale is greater than the threshold, it is retained as the target signal; when the wavelet coefficient is less than the threshold, it is determined to be noise and filtered out. For example, in the wavelet coefficients of the third layer of a high-voltage cable accessory, wavelet coefficients with a value of 0.3 (greater than the threshold 0.24) are retained, while wavelet coefficients with a value of 0.1 (less than the threshold 0.24) are filtered out; the new 35kV cable is processed according to the same rule.

[0130] VI. Signal Reconstruction and Dielectric Property Data Extraction

[0131] The filtered wavelet coefficients are reconstructed to obtain a clean signal after noise removal. Dielectric performance variation data, including key indicators such as dielectric loss factor and capacitance change rate, are accurately extracted from the reconstructed signal. After a series of rigorous calculations and analyses, the dielectric loss factor of the 110kV high-voltage cable accessory is 0.05, and the capacitance change rate is 5%; the dielectric loss factor of the 35kV new cable accessory is 0.02, and the capacitance change rate is 1%. These data are consistent with the above precise calculations, accurately reflecting the difference in dielectric performance between the two cable accessories.

[0132] VII. Assess the degree of aging and deterioration of cable insulation materials

[0133] Based on the analysis results of reconstructed signals and dielectric loss angles, the aging and deterioration degree of the insulation materials of cable accessories was assessed. Comparing the dielectric performance data and dielectric loss angles of 110kV high-voltage cable accessories and 35kV new cable accessories, it is evident that the dielectric loss factor (0.05) of a certain high-voltage cable accessory is significantly higher than that of the new cable accessory (0.02), and its capacitance change rate (5%) is also much higher than that of the new cable accessory (1%). Furthermore, the dielectric loss angle of the high-voltage cable accessory (3°) is greater than that of the new cable (1.5°). Combined with the research conclusions on cable accessory aging in the State Grid report, it can be concluded that the insulation material of the certain high-voltage cable accessory has a high degree of aging, indicating that there may be many defects and aging products inside, which seriously affect the electrical performance of the cable accessory. In contrast, the insulation performance of the 35kV new cable accessory is relatively good, with very little sign of aging. Based on comprehensive assessment, the insulation material of a certain 110kV high-voltage cable accessory has undergone severe aging and deterioration, requiring enhanced monitoring or timely maintenance. The new 35kV cable accessory can continue to operate normally, but it still needs to be inspected regularly as required to ensure its long-term stable operation.

Claims

1. A method for aging detection of cable accessories based on multi-parameter fusion wavelet transform, characterized in that, include: An AC steady-state equivalent circuit model of the cable accessory is established, and dynamic data of current and voltage of the target cable accessory sample as a function of test temperature and test time are obtained from the AC steady-state equivalent circuit model of the cable accessory. The formula for calculating the dielectric loss angle is derived from the sinusoidal steady-state voltage-current relationship matrix. Using this formula, the dielectric loss angle of cable accessories is calculated, including: Considering that the voltage and current waveforms of the cable accessories are sinusoidal steady-state waveforms, the voltage and current relationship matrix across the cable accessory line is established as follows: ； In the formula, The propagation coefficient of the line; The wave impedance of the line; Let be the natural logarithm; where is the propagation coefficient of the line. and the wave impedance of the line The expression is as follows: ； ； The voltage-current relationship matrix at both ends of the cable accessory line is obtained as follows: ； The dielectric loss angle of cable accessories satisfies the following formula: ； From the above formula, we can obtain: ； As can be seen from the above formula, the ratio of the current difference between the two ends of the cable accessory to the average voltage between the two ends will not change as the load current flows through the cable accessory; the complementary angle of the phase angle difference between the current difference between the two ends of the cable accessory and the average voltage between the two ends is taken as the dielectric loss angle of the cable accessory. Wavelet transform decomposition was performed on the dynamic data of current and voltage to obtain multi-scale wavelet coefficients; Based on the normal distribution characteristics of noise signals and a multi-parameter fusion mechanism, a dynamic threshold is set by combining the amplitude differences of wavelet coefficients at various scales, including: Calculate the first by using the statistical distribution characteristics of the noise wavelet coefficients. Standard deviation of layer noise wavelet coefficients This is used to characterize the intensity fluctuation of noise at this scale; Calculate the first Layer signal energy With the Layer noise energy The ratio is used to reflect the difference in energy distribution between the target signal and noise at a specific scale; Through formula Determine the first Adaptive dynamic threshold of the layer ,in, Based on the attenuation factor, This is an adaptive adjustment factor; The modulus maxima of the wavelet coefficients of noise decrease with increasing scale, while the modulus maxima of the wavelet coefficients of the target signal increase with increasing scale. Adaptive dynamic threshold of the layer Decrease with scale; The wavelet coefficients of the target signal are selected from the multi-scale wavelet coefficients using a dynamic threshold, and noise in the target signal is filtered out. The filtered wavelet coefficients are reconstructed to extract dielectric property change data from the reconstructed signal. By combining data on changes in dielectric properties with dielectric loss angle, the degree of aging and deterioration of the insulation material of cable accessories can be assessed.

2. The cable accessory aging detection method based on multi-parameter fusion wavelet transform according to claim 1, characterized in that, The process of establishing the AC steady-state equivalent circuit model of the cable accessory is as follows: When the cable accessories are operating normally, a small segment is cut off at a distance X from the beginning. The AC steady-state equivalent circuit model of the cable accessories includes the line resistance of three sets of series-connected infinitesimal segments. and the inductance along the line of the micro-element segment Furthermore, the output terminals of the first group, the input and output terminals of the second group, and the input terminals of the third group are all connected to the ground conductivity of a set of parallel micro-element segments. and the capacitance to ground of the micro-element segment .

3. The cable accessory aging detection method based on multi-parameter fusion wavelet transform according to claim 1, characterized in that, The current and voltage dynamic data were decomposed by wavelet transform using the db4 wavelet basis function to obtain multi-scale wavelet coefficients.

4. The cable accessory aging detection method based on multi-parameter fusion wavelet transform according to claim 1, characterized in that, Calculate the first by using the statistical distribution characteristics of the noise wavelet coefficients. Standard deviation of layer noise wavelet coefficients Characterized by noise intensity fluctuations at this scale, including: In the initial decomposition layer, assuming noise is dominant, wavelet coefficients of the initial decomposition layer are extracted. , For coefficient index; The first was calculated using statistical methods. Standard deviation of layer noise wavelet coefficients : ； in, , This represents the total number of wavelet coefficients. For subsequent scales n>1, the noise standard deviation decreases according to an empirical formula: ； in, It is the basic attenuation factor, which reflects the attenuation characteristics of noise energy as the scale increases.

5. The cable accessory aging detection method based on multi-parameter fusion wavelet transform according to claim 1, characterized in that, The first Layer signal energy The expression is as follows: ； in, The wavelet coefficients retained after denoising; This represents the total number of wavelet coefficients. No. Layer noise energy The expression is as follows: ； in, For noise reduction Wavelet coefficients of the layer.

6. The cable accessory aging detection method based on multi-parameter fusion wavelet transform according to claim 1, characterized in that, The adaptive adjustment factor It is used to balance scale decay and energy weight, and its value ranges from 0.1 to 0.

3.

7. The cable accessory aging detection method based on multi-parameter fusion wavelet transform according to claim 1, characterized in that, The wavelet coefficients of the target signal are selected from multi-scale wavelet coefficients using dynamic thresholding, and noise in the target signal is filtered out, including: If the wavelet coefficients at the current scale are greater than the adaptive dynamic threshold If the wavelet coefficients at the current scale are less than the adaptive dynamic threshold, then the target signal is retained. It is then identified as noise and filtered out, where the adaptive dynamic threshold is used. Dynamically adjusted through a multi-parameter fusion mechanism.

8. The cable accessory aging detection method based on multi-parameter fusion wavelet transform according to claim 1, characterized in that, The dielectric performance change data includes the dielectric loss factor and capacitance change rate, which are used to quantitatively analyze the degree of degradation of the insulation material of cable accessories.

Images