A method and system for quantitatively evaluating water tree branches of XLPE cable based on variable frequency pseudo-trapezoidal wave excitation

Abstract

The present application belongs to the technical field of power equipment insulation state detection, and particularly relates to a XLPE cable water tree branch quantitative evaluation method and system based on variable frequency pseudo-trapezoidal wave excitation, which comprises the following steps: applying pseudo-trapezoidal wave excitation and collecting voltage and current signals; adopting wavelet and adaptive filtering joint denoising; establishing a Wiener-ANN nonlinear dielectric response model, fitting cable response through dynamic linear and static neural network module joint fitting; obtaining frequency domain dielectric spectrum and extracting harmonic characteristic parameters, loss slope, hysteresis angle and time domain nonlinearity; finally outputting water tree aging comprehensive index WTI through a support vector machine regression model to realize quantitative evaluation of aging degree. The system comprises a pseudo-trapezoidal wave excitation, high-voltage amplification, micro-current collection, signal processing and aging evaluation unit. The present application can obtain wide frequency harmonic response in a single test, the test efficiency is improved by more than 75%, the anti-interference is strong, the evaluation precision is high, and the present application is suitable for on-site rapid nondestructive diagnosis.

Description

Technical Field

[0001] This invention belongs to the field of electrical equipment insulation condition detection technology, specifically relating to a quantitative assessment method and system for water treeing in XLPE cables based on frequency conversion pseudo-trapezoidal wave excitation. Background Technology

[0002] Cross-linked polyethylene (XLPE) cables are widely used in medium and high voltage power transmission and distribution systems due to their excellent electrical and mechanical properties. However, during long-term operation, the cables are subjected to a combination of electrical, thermal, mechanical, and environmental stresses, which can easily lead to water treeing aging in the insulation layer. Water treeing, a typical defect in cable insulation aging, can gradually develop into electrical treeing, eventually causing insulation breakdown and triggering power supply accidents.

[0003] Currently, the main detection methods for cable insulation aging include partial discharge detection, time-domain reflectometry (TDR), and frequency-domain dielectric spectroscopy (FDS). Partial discharge detection is sensitive to concentrated defects, but its ability to detect distributed water tree aging is limited, and it is easily affected by electromagnetic interference in the field. The resolution of time-domain reflectometry is limited by the pulse width, making it difficult to identify early and subtle aging. Although frequency-domain dielectric spectroscopy can reflect the insulation state, traditional sinusoidal frequency sweep testing requires multiple excitations, which is time-consuming and inefficient. Furthermore, the limited excitation voltage amplitude leads to a low signal-to-noise ratio and large measurement errors.

[0004] Furthermore, existing assessment methods are mostly based on a single characteristic parameter (such as the tangent of the dielectric loss angle), failing to fully utilize the nonlinear response characteristics introduced by water treeing, resulting in ambiguous assessments of aging degree and insufficient quantitative accuracy. Therefore, there is an urgent need to develop a novel method and system that can achieve high-voltage broadband excitation, efficient anti-interference, and accurate quantitative assessment of water treeing aging based on multidimensional nonlinear characteristics. Summary of the Invention

[0005] The purpose of this invention is to provide a quantitative evaluation method and system for water treeing in XLPE cables based on frequency conversion pseudo-trapezoidal wave excitation. This method can obtain broadband harmonic response in a single test, improves test efficiency by more than 75%, has strong anti-interference capabilities, high evaluation accuracy, and is suitable for rapid non-destructive diagnosis in the field.

[0006] The specific technical solution adopted by this invention is as follows: A method for quantitatively assessing water treeing in XLPE cables includes the following steps: S1: Obtain time-domain test data A pseudo-trapezoidal wave excitation signal is generated and applied to the cable under test, while the excitation voltage is acquired simultaneously. and response current ; S2: Signal Preprocessing An algorithm based on binary wavelet transform and adaptive filtering is used to denoise the acquired signal and eliminate electromagnetic interference. S3: Establish the Wiener-ANN nonlinear dielectric response model Construct a Wiener-ANN nonlinear dielectric response model that includes dynamic linear modules and static nonlinear artificial neural network modules, using a preprocessed excitation voltage. As input, in response to current As the target output, the backpropagation algorithm is used to jointly optimize the model parameters; S4: Obtain the frequency domain dielectric spectrum A virtual pseudo-trapezoidal scanning signal is input into the trained Wiener-ANN nonlinear dielectric response model, and a high-precision frequency domain dielectric spectrum is obtained by calculating the voltage and current phase difference and amplitude ratio at each frequency point. S5: Extract aging characteristic parameters Extracting multiharmonic nonlinear characteristic parameters from frequency domain dielectric spectrum loss spectrum slope ; Extracting the hysteresis circle rotation angle from the time-domain response Time-domain nonlinearity NTD ; S6: Quantitative assessment of aging level Based on the four extracted feature parameters , , , Inputting the data into a pre-trained support vector machine regression model, the model outputs the Water Tree Aging Comprehensive Index (WTI) to achieve a quantitative assessment of the degree of aging.

[0007] Furthermore, in step S3, the static nonlinear artificial neural network module of the Wiener-ANN nonlinear dielectric response model adopts a multilayer perceptron with a single hidden layer, the number of hidden layer nodes is 5-15, and the activation function is the Sigmoid function.

[0008] Furthermore, the dynamic linear module of the Wiener-ANN nonlinear dielectric response model described in step S4 is used to describe the linear dynamic polarization characteristics of the cable insulation, and its transfer function is: ; in, and The expression is a polynomial; the static nonlinear artificial neural network module is implemented using a multilayer perceptron (MLP) artificial neural network. The input to this network is the output of the dynamic linear module, and the output of this network is the final output of the model. The final output formula of the model is as follows: ; in, The model number is represented by the first... The final output in each dimension (predicted dielectric response current) ； Indicates the hidden layer number 1 The nth neuron to the output layer The weight of each node; This represents the bias term of the k-th node in the output layer.

[0009] Furthermore, the parameters of the dynamic linear module and the weights of the static nonlinear artificial neural network module are jointly optimized using the backpropagation algorithm and the particle swarm optimization algorithm to minimize the root mean square error between the simulated output current and the measured response current of the model. During model training, the error performance of the validation set is monitored in real time, and training is automatically terminated when its performance begins to deteriorate, thereby ensuring the generalization ability of the model.

[0010] The multiharmonic nonlinear characteristic parameters The specific method for obtaining the dielectric loss factor curves is as follows: Extract the dielectric loss factor curves for the fundamental, third, fifth, and seventh harmonics from the frequency domain dielectric spectrum; within the frequency range of 0.01 Hz to 1 Hz, calculate the area enclosed by the loss factor curves of each harmonic and the frequency axis, denoted as . The multi-harmonic nonlinear characteristic parameters in step S5 It is defined as the weighted sum of the integral areas of the loss factor curves under each harmonic within a specific frequency band, and the multi-harmonic nonlinear characteristic parameter is... The calculation formula is: ; in, , , , These represent the integral areas of the dielectric loss factors of the fundamental, third, fifth, and seventh harmonics in the 0.01Hz-1Hz frequency band, respectively. , , These are the weighting coefficients related to the voltage amplitude of each harmonic.

[0011] A water tree defect testing system for XLPE cables includes: A pseudo-trapezoidal wave excitation unit is used to generate a pseudo-trapezoidal wave voltage signal with independently adjustable rise time, fall time, and constant duration, wherein the constant peak duration accounts for 20% of the signal period; The high-voltage amplification unit connects the output of the pseudo-trapezoidal wave excitation unit to the input of the high-voltage amplification unit. The output of the high-voltage amplification unit is used to connect to the core of the cable under test. The high-voltage amplification unit can amplify the amplitude of the pseudo-trapezoidal wave voltage signal to a preset voltage amplitude and output it to the cable core, preferably a 5kV test voltage, and the output amplitude fluctuation is less than ±5% in the 10mHz-1kHz frequency band.

[0012] A micro-current acquisition unit is provided, the input end of which is connected to the shielding layer of the cable under test, specifically in series in the grounding circuit of the cable shielding layer, for synchronously acquiring the response current signal of the cable when a pseudo-trapezoidal wave excitation is applied; the micro-current acquisition unit is built based on the I / V conversion principle and can measure the pA-level response current without phase shift.

[0013] The signal processing unit, wherein the output terminal of the micro-current acquisition unit is connected to the input terminal of the signal processing unit, is used to perform noise reduction processing on the acquired excitation voltage and response current signals; An aging assessment unit is included, with the output of the signal processing unit connected to the aging assessment unit. The aging assessment unit incorporates a Wiener-ANN nonlinear dielectric response model, which includes a cascaded dynamic linear module and a static nonlinear artificial neural network module. This model is used to train the preprocessed Wiener-ANN nonlinear dielectric response model, thereby obtaining the frequency domain dielectric spectrum of the defective cable. Aging characteristic parameters, namely the water tree aging comprehensive index or the predicted tree length, are extracted from the time and frequency domain signals to achieve a quantitative assessment of the water tree defect status of the XLPE cable.

[0014] Furthermore, the signal generated by the Wiener-ANN nonlinear dielectric response model is a periodic signal containing a linear rising segment, a flat-top sustaining segment, and a linear falling segment within a single period. The discrete difference equation of the dynamic linear module of the Wiener-ANN nonlinear dielectric response model is as follows: ; in This represents the input signal (pseudo-trapezoidal excitation voltage) at the current time (time n) of the model. This represents the output of the linear module at the current moment (i.e., the input of the nonlinear module). , ,..., Indicates the input signal And the linear coefficients (forward coefficients) of its delay term; , ,..., Indicates the output signal The linear coefficients (feedback coefficients) of the input signal and its delay term; M represents the delay order of the input signal, and N represents the feedback delay order of the output signal; u(n−k) represents the delay value of the input signal at time n−k (k=1,2,...,M); x(n−k) represents the delay value of the output signal at time n−k (k=1,2,...,N).

[0015] The dynamic linear module in the Wiener-ANN nonlinear dielectric response model is implemented using finite impulse response filters of order 10 to 30, which are used to characterize the linear polarization relaxation characteristics of the cable insulation medium.

[0016] The described static nonlinear artificial neural network module employs a multilayer perceptron structure containing a single hidden layer. The number of nodes in the hidden layer is set to 8-12, and the sigmoid function is selected as the activation function to address the complex nonlinear dielectric response introduced by equivalent water tree aging. Each neuron in the hidden layer first processes the input... (i.e., linear module output) The dimensionality of the matrix is ​​expanded linearly, then nonlinearity is introduced through the Sigmoid activation function, as shown in the following formula: ; The Sigmoid activation function maps the linear combination result to the [0,1] interval, and its formula is as follows: in, The input vector of the nonlinear module (i.e., the output of the linear module) The (dimensions); among which Indicates the hidden layer number 1 The output of the nth neuron; represents the output from the i-th node in the input layer to the i-th hidden layer. Weights of each neuron ； This represents the bias term of the j-th neuron in the hidden layer; This represents the Sigmoid activation function, where z is the input to the activation function (i.e., ...). ).

[0017] Furthermore, the aging characteristic parameters extracted by the aging assessment unit include: the rotation angle of the time-domain hysteresis circle ( α ), time-domain nonlinearity ( NTD ), multi-harmonic nonlinear characteristic parameters ( λ m ), low-frequency loss spectrum slope ( K Among them, the multi-harmonic nonlinear characteristic parameters ( λ m The weighted difference is obtained by calculating the integral area of ​​the dielectric loss factor curves of the fundamental, third, fifth, and seventh harmonics in the 0.01Hz-1Hz frequency band.

[0018] The technical effects achieved by this invention are as follows: The present invention provides a quantitative evaluation method and system for water treeing in XLPE cables based on frequency conversion pseudo-trapezoidal wave excitation. The pseudo-trapezoidal wave is used as the excitation source. The dielectric response under multiple harmonic components can be obtained in a single test, avoiding the cumbersome process of multiple frequency sweeps required by traditional methods. The test time is shortened by more than 75%, and the test efficiency is high.

[0019] The present invention provides a quantitative assessment method and system for water treeing in XLPE cables based on frequency conversion pseudo-trapezoidal wave excitation. The output excitation voltage is as high as 5kV, which significantly improves the signal-to-noise ratio. The joint denoising algorithm based on binary wavelet transform and adaptive filtering effectively prevents electromagnetic interference on site and has strong anti-interference ability.

[0020] The present invention provides a quantitative assessment method and system for water treeing in XLPE cables based on frequency conversion pseudo-trapezoidal wave excitation. It establishes a Wiener-ANN nonlinear dielectric response model, which can accurately describe the complex nonlinear characteristics caused by water treeing aging, realize the accurate identification and degree quantification of aging state, and achieve high assessment accuracy. Attached Figure Description

[0021] Figure 1 This is a schematic diagram of the test system of the present invention; Figure 2 This is a comparison diagram of the original signal and the filtered signal of this invention; Figure 3 This is the Wiener-ANN training model of the present invention; Figure 4 This is a comparison diagram of the frequency domain dielectric spectra of healthy and aged cables according to the present invention; Figure 5 This is a graph showing the relationship between the WTI value and the length of water tree branches in this invention. Detailed Implementation

[0022] To make the objectives and advantages of this invention clearer, the invention will be specifically described below with reference to embodiments. It should be understood that the following text is merely used to describe one or more specific embodiments of the invention and does not strictly limit the scope of protection specifically claimed by the invention.

[0023] like Figures 1-5 As shown, the object of the test in this embodiment is a YJLV8.7 / 15kV XLPE power cable that has been in operation for more than 15 years. The cable has shown obvious signs of water treeing aging.

[0024] like Figure 1As shown, an XLPE cable water tree defect testing system includes a pseudo-trapezoidal wave excitation unit, a high-voltage amplification unit, a micro-current acquisition unit, a signal processing unit, and an aging assessment unit. The output of the pseudo-trapezoidal wave excitation unit is connected to the input of the high-voltage amplification unit via a signal cable; the output of the high-voltage amplification unit is connected to the conductor of the cable under test via a high-voltage lead; the input of the micro-current acquisition unit is connected in series with the cable shielding layer grounding circuit via a shielded measurement line; the output of the micro-current acquisition unit is connected to the analog input channel of the signal processing unit; the signal processing unit establishes a bidirectional communication link with the aging assessment unit via a data bus, completing the closed-loop construction of the entire testing system.

[0025] This embodiment describes the detection and evaluation method for water tree defects in XLPE following the steps below: S1: Obtain time-domain test data The pseudo-trapezoidal wave excitation signal was amplified by high voltage and applied between the cable core and the grounding shield. The time-domain waveforms of the excitation voltage U(t) and response current I(t) were simultaneously acquired. The acquisition time was 200 seconds, covering a frequency range of 10 mHz to 1 kHz. The acquired raw signals are shown below. Figure 2 As shown.

[0026] S2: Signal Preprocessing A combined algorithm based on binary wavelet transform (using the db8 wavelet basis) and adaptive filtering (LMS algorithm, step size factor μ=0.01) was used to denoise the acquired signal. The signal-to-noise ratio of the filtered signal was improved from the original 25dB to 52dB, effectively eliminating power frequency interference and high-frequency noise.

[0027] S3: Establish the Wiener-ANN nonlinear dielectric response model A Wiener-ANN nonlinear dielectric response model, comprising dynamic linear modules and static nonlinear artificial neural network modules, is constructed. Among them, the dynamic linear module adopts a 4th-order ARX model structure; Among them, the static nonlinear artificial neural network module adopts a single hidden layer MLP network with 10 hidden layer nodes and the Sigmoid function as the activation function. The preprocessed U(t) is used as the model input, and I(t) is used as the target output. A backpropagation algorithm with momentum term is used for joint training to optimize the model parameters. The training parameters are set as follows: learning rate η = 0.001, momentum factor α = 0.9, and maximum training iterations of 1000. The mean squared error convergence curve of the training process is shown below. Figure 3 As shown, the final training error reached 0.047%.

[0028] S4: Obtain the frequency domain dielectric spectrum A virtual pseudo-trapezoidal scanning signal (frequency range 10mHz-1kHz, 200 frequency points, logarithmically uniformly distributed) is input into the trained Wiener-ANN nonlinear dielectric response model. A high-precision frequency-domain dielectric spectrum is obtained by calculating the phase difference and amplitude ratio of the model's output voltage and output current at each frequency point. The obtained dielectric spectrum is shown below. Figure 4 As shown, it can be seen that the tanδ values ​​under each harmonic are significantly separated in the water tree aged cable.

[0029] S5: Extract aging characteristic parameters The following characteristic parameters are extracted from the frequency domain dielectric spectrum: Multiharmonic nonlinear characteristic parameter λm: Calculate the integral area of ​​the dielectric loss factor curves of the fundamental, third, fifth, and seventh harmonics in the 0.01Hz-1Hz frequency band, according to the formula... Calculate, where the weighting coefficients , , Calculated .

[0030] Loss spectrum slope K: Linear fitting of the fundamental tanδ-f curve in the 0.01Hz-0.1Hz frequency band yields a slope K=-2.45.

[0031] Extract the following feature parameters from the time-domain response: Hysteresis circle rotation angle α: Plot the ui hysteresis circle at 1kHz fundamental frequency and measure the angle α between its major axis and the voltage axis, which is 18.7°.

[0032] Time-domain nonlinearity NTD: The ratio of the root mean square error between the actual response current and the ideal linear response current is calculated as NTD = 3.82.

[0033] S6: Quantitative assessment of aging level The four extracted feature parameters are α=18.7°, NTD=3.82, K=-2.45 was input into a pre-trained support vector machine regression model (using the RBF kernel function, C=100, γ=0.1). This regression model was trained using 50 sets of laboratory cable sample data with known water tree lengths, and the prediction error was less than ±8%.

[0034] The model outputs a water treeing aging comprehensive index (WTI) of 76.5 (range 0-100, with higher values ​​indicating more severe aging), corresponding to a predicted water treeing length of 285 μm. To verify the accuracy of the results, a sample section of the tested cable was observed, and the measured water treeing length was 273 μm, with a relative error of 4.4%, meeting the requirements for engineering applications.

[0035] Analysis of test results shows that the test results in this embodiment are as follows: (1) Using pseudo-trapezoidal wave excitation can obtain rich harmonic response information in a single test, and the test time is reduced from 32 minutes in the traditional method to 8 minutes, improving efficiency by 75%; (2) The Wiener-ANN nonlinear dielectric response model can accurately describe the nonlinear dielectric characteristics of water tree aged cables, with a model fit of 99.2%; (3) The quantitative evaluation method based on multidimensional characteristic parameters can accurately predict the length of water tree, providing a reliable basis for cable condition assessment and remaining life prediction.

[0036] The above description is merely a preferred embodiment of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention. Structures, devices, and operating methods not specifically described or explained in this invention are implemented according to conventional methods in the art unless otherwise specified or limited.

Claims

1. A method for quantitatively assessing water treeing in XLPE cables, characterized in that: Includes the following steps: S1: Obtain time-domain test data A pseudo-trapezoidal wave excitation signal is applied to the cable under test, and the excitation voltage is acquired simultaneously. and response current ; S2: Signal Preprocessing The acquired signals are subjected to noise reduction processing; S3: Establish the Wiener-ANN nonlinear dielectric response model A Wiener-ANN nonlinear dielectric response model, comprising dynamic linear modules and static nonlinear artificial neural network modules, is constructed to excite voltage. As input, in response to current As the target output; S4: Obtain the frequency domain dielectric spectrum A virtual pseudo-trapezoidal scanning signal is input into the trained Wiener-ANN nonlinear dielectric response model, and the frequency domain dielectric spectrum is obtained by calculating the voltage-current phase difference and amplitude ratio at each frequency point. S5: Extract aging characteristic parameters Extracting multiharmonic nonlinear characteristic parameters from frequency domain dielectric spectrum loss spectrum slope ， Extracting the hysteresis circle rotation angle from the time-domain response Time-domain nonlinearity NTD ; Multiharmonic nonlinear characteristic parameters The calculation formula is: ； in, , , , These represent the integral areas of the dielectric loss factors of the fundamental, third, fifth, and seventh harmonics in the 0.01Hz-1Hz frequency band, respectively. , , These are weighting coefficients related to the amplitude of each harmonic voltage. The time-domain nonlinearity (NTD) is obtained by calculating the root mean square error ratio between the actual response current and the ideal linear response current. S6: Quantitative assessment of aging level Based on the four extracted feature parameters , , , Quantitatively assess the aging degree of water tree branches; The S6 extracts four feature parameters , , , The input is fed into a pre-trained support vector machine regression model, and the model outputs a comprehensive index of water tree aging.

2. The method for quantitative assessment of water treeing in XLPE cables according to claim 1, characterized in that: The rise time, fall time and constant duration of the pseudo trapezoidal wave voltage signal generated in step S1 can be adjusted independently, with the constant peak duration accounting for 20% of the signal period.

3. The method for quantitative assessment of water treeing in XLPE cables according to claim 1, characterized in that: In step S2, an algorithm based on a combination of binary wavelet transform and adaptive filtering is used to denoise the acquired signal.

4. The method for quantitative assessment of water treeing in XLPE cables according to claim 1, characterized in that: The discrete difference equation of the dynamic linear module of the Wiener-ANN nonlinear dielectric response model in step S3 is as follows: ； in This represents the input signal of the model at the current moment; This represents the output of the linear module at the current moment. , ,..., Indicates the input signal and the linear coefficients of its delay term; , ,..., Indicates the output signal The linear coefficients of the input signal and its delay term; M represents the delay order of the input signal, and N represents the feedback delay order of the output signal; u(n−k) represents the delay value of the input signal at time n−k, where k=1,2,...,M; x(n−k) represents the delay value of the output signal at time n−k, where k=1,2,...,N; The dynamic linear module in the Wiener-ANN nonlinear dielectric response model is implemented using finite impulse response filters of order 10 to 30, which are used to characterize the linear polarization relaxation characteristics of the cable insulation medium.

5. The method for quantitative assessment of water treeing in XLPE cables according to claim 1, characterized in that: The static nonlinear artificial neural network module of the Wiener-ANN nonlinear dielectric response model described in step S3 adopts a multilayer perceptron with a single hidden layer, the number of hidden layer nodes is 5-15, and the activation function is the Sigmoid function.

6. The method for quantitative assessment of water treeing in XLPE cables according to claim 1, characterized in that: The dynamic linear module of the Wiener-ANN nonlinear dielectric response model mentioned in step S4 is used to describe the linear dynamic polarization characteristics of cable insulation, and its transfer function is: ； in, and It is a polynomial; the static nonlinear artificial neural network module is implemented by a multilayer perceptron artificial neural network. The input of this network is the output of the dynamic linear module, and the output of this network is the final output of the model; the final output formula of the model is as follows: ； in, Indicates the hidden layer number 1 The output of each neuron; The model number is represented by the first... The final output of each dimension ； Indicates the hidden layer number 1 The nth neuron to the output layer The weight of each node; This represents the bias term of the k-th node in the output layer.

7. A water tree defect testing system for XLPE cables, used to implement the method described in any one of claims 1-6, characterized in that: include: The pseudo-trapezoidal wave excitation unit is used to generate a pseudo-trapezoidal wave voltage signal with independently adjustable rise time, fall time, and constant duration. The high-voltage amplification unit has its output terminal connected to the input terminal of the pseudo-trapezoidal wave excitation unit. The output terminal of the high-voltage amplification unit is used to connect to the core of the cable under test. The high-voltage amplification unit is used to amplify the amplitude of the pseudo-trapezoidal wave voltage signal to a preset voltage amplitude and output it to the cable core. A micro-current acquisition unit, the input end of which is connected to the shielding layer of the cable under test, is used to synchronously acquire the response current signal of the cable when a pseudo-trapezoidal wave excitation is applied; The signal processing unit, wherein the output terminal of the micro-current acquisition unit is connected to the input terminal of the signal processing unit, is used to perform noise reduction processing on the acquired excitation voltage and response current signals; An aging assessment unit is included, with the output of the signal processing unit connected to the aging assessment unit. The aging assessment unit incorporates a Wiener-ANN nonlinear dielectric response model to extract characteristic parameters and output aging assessment results.

8. The XLPE cable water tree defect testing system according to claim 7, characterized in that: In the Wiener-ANN model, the order of the dynamic linear module is 10–30, and the number of hidden layer nodes of the static nonlinear module is 8–12.

9. The XLPE cable water tree defect testing system according to claim 7, characterized in that: The high-voltage amplification unit outputs a test voltage of 5kV, and the output amplitude fluctuation is less than ±5% within the 10mHz-1kHz frequency band.

Images