A transient traveling wave positioning method for broken strand fault of overhead insulated conductor

Abstract

This invention belongs to the field of positioning technology, specifically relating to a transient traveling wave positioning method for broken strand faults in overhead insulated conductors. The method includes: acquiring a single-end transient current signal of the line, obtaining a reconstructed signal through bandpass filtering, and performing an initial time-frequency domain transformation to obtain an initial time-frequency distribution; calculating the power spectrum and constructing a parallel time-varying parameter oscillator array to obtain multiple narrowband enhanced signals; performing a second time-frequency domain transformation on the enhanced signals to construct a fused time-frequency energy matrix; extracting the modulus maxima propagation path from the matrix; determining the arrival times of the initial traveling wave and the reverse polarity fault reflection wave based on the moment when the singularity characteristic index first enters the abrupt change threshold interval; and calculating the fault distance based on the calibrated traveling wave velocity and the single-end traveling wave ranging principle. This method effectively suppresses noise and enhances weak features, significantly improving the accuracy of wavefront identification and the reliability of single-end positioning.

Description

Technical Field

[0001] This invention relates to the field of positioning technology. More specifically, this invention relates to a transient traveling wave positioning method for a broken strand fault in an overhead insulated conductor. Background Technology

[0002] Strand breakage faults in overhead insulated conductors are a type of hidden initial fault. If not detected and properly handled in a timely manner, they can easily evolve into serious accidents, threatening the safe and stable operation of the power grid. The transient signals generated when such faults occur are relatively weak, with a low signal-to-noise ratio, relatively complex frequency components, and are easily submerged by strong interference signals.

[0003] Currently, the existing traveling wave ranging method has the advantages of fast positioning speed and high positioning accuracy. It mainly analyzes the signal at different scales to detect the instantaneous characteristics of the signal, and identifies the arrival time of the traveling wave front by tracing the propagation path of the time-frequency transformation modulus maxima at different scales.

[0004] However, when existing traveling wave localization methods based on traditional time-frequency transforms, such as continuous wavelet transforms, are applied to weak faults like broken strands, the energy of the broken strand fault signal is extremely weak, and its modulus maxima features in the time-frequency domain are often not prominent enough. A low signal-to-noise ratio can easily lead to blurred or even interrupted modulus maxima paths, making it difficult to effectively identify the wavefronts of the initial and reflected traveling waves. Furthermore, background noise and various electromagnetic interferences can easily generate pseudo-modulus maxima points in the time-frequency spectrum, which can interfere with the identification of the true wavefront, leading to significant errors in the localization results or even the risk of localization failure. Especially for the even weaker reflected waves from the fault point, reliable extraction of the wavefront of the reflected wave is crucial for the effective application of single-ended traveling wave localization technology; however, traditional methods still have limitations in their feature extraction capabilities in this regard. Summary of the Invention

[0005] To address the technical problem of weak and easily interfered signals from strand breakage faults, making reliable extraction of reflected wavefronts difficult, this invention proposes a transient traveling wave localization method for strand breakage faults in overhead insulated conductors. The method includes: acquiring a single-end transient current signal of the line; filtering the transient current signal using a digital bandpass filter based on the frequency band characteristics of the strand breakage fault to extract the target frequency band signal, thus obtaining a reconstructed signal; performing an initial time-frequency domain transformation on the reconstructed signal to obtain an initial time-frequency distribution; calculating the power spectrum of the input signal, identifying the frequencies corresponding to N prominent spectral peaks in the power spectrum, and constructing an array of N parallel time-varying parameter oscillators; feeding the input signal into the oscillator array to obtain N narrowband enhanced signals, wherein the center of each oscillator... The frequency is consistent with the corresponding peak frequency of the spectrum, and the time-varying damping coefficient is determined based on the corresponding coefficient magnitude in the initial time-frequency distribution. N narrowband enhancement signals are each subjected to a second time-frequency domain transformation to obtain N enhanced time-frequency distributions, which are then combined with the initial time-frequency distribution to construct a fused time-frequency energy matrix. The propagation path of the modulus maxima is extracted from the fused time-frequency energy matrix. The arrival time of the initial traveling wavefront is determined based on the first entry of the index reflecting singularity characteristics on the path into the preset abrupt change threshold interval. The arrival time of the reflected wavefront at the fault point is determined based on the index re-entering the preset abrupt change threshold interval with the wavefront polarity reversed. The fault distance is calculated based on the pre-calibrated propagation speed of the traveling wave in the conductor, according to the single-end traveling wave ranging principle.

[0006] This invention effectively suppresses background noise and interference from irrelevant frequency components through high-precision digital bandpass filtering and initial time-frequency domain transformation, thereby improving the signal-to-noise ratio of the processed signal. Simultaneously, by utilizing a parallel time-varying parameter oscillator array constructed based on the fault signal power spectrum, selective narrowband enhancement of fault characteristic frequency components is performed, significantly improving the identification of weak transient fault signals. A fused time-frequency energy matrix containing rich fault details is constructed by combining initial and enhanced time-frequency information. Furthermore, singularity analysis of the modulus maxima path is performed using indices reflecting singularity characteristics, such as the Lipschitz exponent, enabling relatively accurate detection of the arrival times of the initial and reflected traveling waves' wavefronts. This comprehensively improves the extraction capability of weak fault features and the accuracy of traveling wavefront identification, thus achieving reliable single-end localization of broken strand faults in overhead insulated conductors.

[0007] Preferably, the step of using a digital bandpass filter to filter the transient current signal and extract the target frequency band signal based on the frequency band characteristics of the strand breakage fault includes: setting the frequency band characteristics of the strand breakage fault to a frequency range of 50kHz to 500kHz; configuring a digital bandpass filter so that its passband range corresponds to the frequency range of 50kHz to 500kHz; inputting the single-end transient current signal of the line into the digital bandpass filter to perform filtering processing; and using the output signal as the reconstructed signal.

[0008] This invention sets the frequency band characteristics of strand breakage faults to a specific high-frequency range and directly uses a mature and efficient digital bandpass filter to accurately extract the corresponding target frequency band. It can selectively retain the frequency band information that embodies the characteristics of strand breakage faults, further filter out interference from low-frequency power frequency components and high-frequency white noise, and avoid truncation errors and computational losses caused by complex transform domain reconstruction, thus providing a cleaner input signal for subsequent signal analysis.

[0009] Preferably, the calculation of the power spectrum of the input signal and the identification of the frequencies corresponding to the N prominent spectral peaks in the power spectrum include: calculating the power spectral density of the input signal using a fast Fourier transform algorithm; setting a peak search threshold of 10% of the maximum amplitude of the power spectrum; searching for all local maxima points in the power spectrum whose amplitudes exceed the threshold as spectral peaks; arranging all identified spectral peaks in descending order of amplitude, selecting the N spectral peaks with relatively large amplitudes, and using the frequencies corresponding to these N spectral peaks as the N prominent spectral peak frequencies.

[0010] This invention employs a fast Fourier transform algorithm to calculate the power spectral density and sets a reasonable peak search threshold to filter and sort local maxima in descending order. This effectively identifies prominent characteristic frequency components in the signal, providing a reliable frequency parameter basis for the subsequent construction of a precisely matched time-varying parameter oscillator array.

[0011] Preferably, the construction of an oscillator array consisting of N parallel time-varying parameter oscillators specifically includes: configuring each time-varying parameter oscillator as a second-order linear narrowband resonant filter system; feeding the reconstructed input signal into each oscillator for narrowband resonant enhancement processing, wherein the natural angular frequency of each oscillator is determined by its corresponding spectral peak frequency, and each oscillator is configured with a time-varying damping coefficient that is dynamically adjusted with time; and the output dynamic response of each oscillator is the corresponding narrowband enhancement signal.

[0012] This invention sets the response behavior of time-varying parameter oscillators to a quasi-second-order linear resonance process, and can set a suitable natural angular frequency for each oscillator according to the identified spectral peak frequency. This enables the oscillator array to produce a good resonance response to the weak characteristic frequency components in the input signal, thereby achieving effective enhancement of signals in a specific frequency band.

[0013] Preferably, for any oscillator, the time-varying damping coefficient The calculation method is as follows:

[0014] in, For the initial time-frequency distribution corresponding to the first The center frequency and scale of the oscillator At the present moment The magnitude of the coefficient, It represents the maximum value of the magnitudes of all coefficients in the initial time-frequency distribution. This is the set damping adjustment normal value.

[0015] This invention dynamically calculates the time-varying damping coefficient using the corresponding coefficient magnitude in the initial time-frequency distribution. When the transient signal has strong energy near the characteristic frequency, the damping coefficient decreases accordingly, causing the oscillator to approach the resonance state and enhance the characteristic output. When the signal energy is weak, the damping coefficient increases, causing the oscillator to decay rapidly and suppress noise fluctuations. This achieves adaptive enhancement and noise reduction of fault signal characteristics.

[0016] Preferably, the step of constructing the fused time-frequency energy matrix by combining the initial time-frequency distribution includes: the value of the fused time-frequency energy matrix at any coordinate point is the product of the coefficient magnitude of the initial time-frequency distribution at that coordinate point and the sum of the coefficient magnitudes of all N enhanced time-frequency distributions at the same coordinate point; the calculation method is as follows:

[0017] in, To fuse the time-frequency energy matrix in terms of scale With time The values ​​at the coordinate points formed. The initial time-frequency distribution at scale With time The modulus of the coefficients at the same coordinate point constituted. For the first The magnitude of the coefficients of the enhanced time-frequency distribution at this coordinate point.

[0018] This invention constructs a fused time-frequency energy matrix by multiplying the coefficient magnitude of the initial time-frequency distribution with the sum of the coefficient magnitudes of each enhanced time-frequency distribution. This ensures that the energy value is significantly prominent only when a certain time-frequency point has energy characteristics in the original signal and can be effectively enhanced by a narrowband oscillator. This method can effectively improve the signal-to-noise ratio of the traveling wavefront on the time-frequency diagram and effectively suppress interference signals unrelated to faults.

[0019] Preferably, determining the initial arrival time of the traveling wavefront based on the first entry of an index reflecting singularity characteristics along the path into a preset abrupt change threshold interval includes: calculating the Lipschitz exponent at each point as the index reflecting singularity characteristics; and using an adaptive interval reflecting the true attenuation and abrupt change characteristics of the wavefront as the preset abrupt change threshold interval, setting this interval as... After extracting the propagation path of the modulus maxima from the fused time-frequency energy matrix, the modulus values ​​of the coefficients at corresponding positions are taken from the initial time-frequency coefficient matrix according to the scale and time of each point on the path, and the slope is obtained by linear fitting in logarithmic coordinates. Then, the Lipschitz exponent was calculated. When the index value falls into the preset mutation threshold range for the first time, it is determined that a fault traveling wave has been detected.

[0020] This invention obtains the slope by performing linear fitting in a logarithmic coordinate system, thereby deriving the Lipschitz exponent, and sets an adaptive fault singularity threshold range with generalization ability. It can sensitively and stably capture the singularity characteristics exhibited when the traveling wave front arrives, thus accurately determining the arrival time of the initial traveling wave front.

[0021] Preferably, the pre-calibrated propagation speed of the traveling wave in the conductor is determined by theoretical formulas based on the conductor type, material, and installation environment parameters. Calculate the determined theoretical propagation speed, where, For the theoretical propagation speed, Let be the inductance per unit length of the conductor. The capacitance per unit length of the conductor; or by injecting a standard pulse signal at the beginning of the line, using a synchronous clock to record the time difference between the emitted and received round-trip reflected signals, and through... The actual propagation speed is calculated by inversely using the principle; among which, Given the total length of the route, To calculate the actual propagation speed, This is the round-trip propagation time difference between the transmission and reception of the pulse signal.

[0022] This invention provides a more accurate reference value for the traveling wave propagation speed of the single-end ranging formula by calculating the theoretical propagation speed by combining conductor parameters or by injecting a standard pulse signal to calculate the actual propagation speed. This helps to reduce positioning errors and improve the accuracy of the final calculated fault distance.

[0023] Preferably, in the step of acquiring the single-end transient current signal of the line and reconstructing the signal and then performing the initial time-frequency domain transformation: a high-speed data acquisition card is used to acquire the transient current quantity after the fault occurs at a sampling frequency of not less than 1MHz to obtain a discrete time series; in the time-frequency domain transformation, a short-time Fourier transform window function or a wavelet basis in the form of complex Morlet wavelets is selected as the analysis function, and the initial time-frequency domain transformation operation is performed on the filtered discrete time series.

[0024] Preferably, determining the arrival time of the reflected wavefront of the fault point based on the indicator re-entering the preset abrupt change threshold interval and the opposite wavefront polarity specifically includes: continuing to search for the modulus maxima propagation path; when the indicator re-enters the preset abrupt change threshold interval, comparing the real part signs of the time-frequency transformation coefficients of the modulus maxima path corresponding to the initial traveling wavefront arrival time and the current time at a certain reference scale; if the real part signs are opposite, that is, the original signal polarities are opposite, then confirming the current time as the arrival time of the reflected wavefront of the fault point.

[0025] The beneficial effects of this invention are as follows: This invention effectively suppresses background noise and interference from irrelevant frequency components through initial wavelet transform and signal reconstruction, thereby improving the signal-to-noise ratio of the processed signal. At the same time, by using a parallel time-varying parameter oscillator array constructed based on the power spectrum of the fault signal, selective narrowband enhancement is performed on the fault characteristic frequency components, which significantly improves the identification of weak transient fault signals.

[0026] Furthermore, by fusing initial and enhanced time-frequency information, this invention constructs a fused time-frequency energy matrix containing rich fault details. By combining the Lipschitz exponent to perform singularity analysis on the modulus maxima path, it can more accurately detect the arrival time of the wavefronts of the initial traveling wave and the reflected traveling wave. This comprehensively improves the ability to extract weak fault features and the accuracy of traveling wavefront identification, thereby achieving reliable single-end location of broken strand faults in overhead insulated conductors. Attached Figure Description

[0027] Figure 1 This is a flowchart of a transient traveling wave localization method for a broken strand fault in an overhead insulated conductor according to the present invention; Figure 2 This is a schematic diagram illustrating the variation of the damping coefficient with time in this invention; Figure 3 This is a comparison chart of experimental results in this invention. Detailed Implementation

[0028] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are some embodiments of the present invention, but not all embodiments.

[0029] This invention discloses a transient traveling wave localization method for strand breakage faults in overhead insulated conductors, referring to... Figure 1 This includes steps S1-S3: S1. Obtain the single-end transient current signal of the line; based on the frequency band characteristics of the broken strand fault, use a digital bandpass filter to filter the transient current signal to obtain the reconstructed signal; perform an initial time-frequency domain transformation on the reconstructed signal to obtain the initial time-frequency distribution.

[0030] In an optional embodiment, a high-speed data acquisition card is used to acquire the transient current after the fault occurs at a sampling frequency of not less than 1 MHz, obtaining a discrete-time series. Based on prior knowledge, the high-frequency traveling wave signal generated by the strand breakage fault is mainly concentrated in the 50 kHz to 500 kHz frequency band. Therefore, a digital bandpass filter, such as an FIR or IIR filter, is directly designed and applied, with its passband set to 50 kHz to 500 kHz, to digitally filter the acquired transient discrete series. After filtering, a reconstructed input signal is obtained that has removed low-frequency power frequency components and high-frequency white noise interference, and centrally reflects the characteristics of the strand breakage fault.

[0031] Subsequently, time-frequency analysis library functions are called, such as PyWavelets in the Python scientific computing library or WaveletToolbox in MATLAB. The analysis window function in the form of complex Morlet wavelets is selected to perform the initial time-frequency domain transformation, such as the initial continuous wavelet transform, on the filtered and reconstructed discrete time series, and a complex matrix is ​​obtained. This matrix is ​​the initial time-frequency distribution.

[0032] S2. Calculate the power spectrum of the input signal, identify the frequencies corresponding to the N prominent spectral peaks in the power spectrum, and construct an oscillator array consisting of N parallel time-varying parameter oscillators. Feed the input signal into the oscillator array to obtain N narrowband enhanced signals, where the center frequency of each oscillator is consistent with the corresponding spectral peak frequency, and the time-varying damping coefficient is determined based on the corresponding coefficient magnitude in the initial time-frequency distribution. Perform a second time-frequency domain transformation on each of the N narrowband enhanced signals to obtain N enhanced time-frequency distributions, and construct a fused time-frequency energy matrix by combining the initial time-frequency distributions.

[0033] In an optional embodiment, firstly, the Welch average periodogram (Welch algorithm) is used to calculate the power spectral density of the reconstructed input signal. Then, a peak detection algorithm is applied to the obtained power spectral curve. By setting an energy threshold and a minimum inter-peak distance, N prominent spectral peaks are identified, and their corresponding frequencies are recorded. .

[0034] Subsequently, a system of N parallel second-order linear time-varying oscillators is constructed. The technical logic of its narrowband resonant filtering of the input signal is manifested as the system's dynamic response driven by the input signal. As a specific implementation of dynamic behavior, each oscillator can be described by the differential equation:

[0035] in, The reconstructed input signal, For the first The output response of each oscillator is the narrowband enhanced signal. For the first The natural angular frequency of the oscillator The damping coefficient is time-varying. It is a time variable.

[0036] Time-varying damping coefficient Set as a function related to the initial time-frequency distribution magnitude, refer to Figure 2 For example, set as:

[0037] in, The frequency in the initial time-frequency distribution Corresponding wavelet scale exist The magnitude of the coefficient at time t, and All are preset positive real numbers.

[0038] Using the input signal as the excitation term, and considering that the acquired fault transient sequence is a data segment within a finite time window with a small data volume, existing conventional microprocessors can meet the computational requirements without special hardware support. Numerical solutions such as the fourth-order Runge-Kutta algorithm (RK45) are used to numerically solve these N differential equations, and the obtained solution results... , ... This results in N narrowband enhanced signals. The same quadratic time-frequency domain transformation as described above is then performed on each of these N narrowband enhanced signals to obtain N enhanced time-frequency distributions.

[0039] Next, the magnitude of the initial time-frequency distribution is multiplied point-by-point by the sum of the magnitudes of the N enhanced time-frequency distributions to construct the fused time-frequency energy matrix, the calculation formula of which is:

[0040] in, To fuse the time-frequency energy matrix at scale With time The values ​​at the coordinate points formed. The magnitude of the initial time-frequency distribution at this point. For the first The modulus of the enhanced time-frequency distribution at this point.

[0041] Furthermore, the power spectrum of the input signal is calculated, and the frequencies corresponding to the N prominent spectral peaks in the power spectrum are identified. Specifically, the power spectral density of the input signal is calculated using the Fast Fourier Transform algorithm; the peak search threshold is set to 10% of the maximum amplitude of the power spectrum; all local maxima points in the power spectrum with amplitudes exceeding this threshold are searched as spectral peaks; all identified spectral peaks are sorted in descending order of amplitude, and the N spectral peaks with relatively large amplitudes are selected, and the frequencies corresponding to these N spectral peaks are taken as the N prominent spectral peak frequencies.

[0042] After obtaining the reconstructed input signal, a Fast Fourier Transform (FFT) algorithm is applied to obtain its frequency domain representation. The square of the magnitude of this frequency domain representation is calculated to obtain the power spectral density (PSD) curve of the signal. The global maximum amplitude is then searched on the obtained PSD curve. And set a peak search threshold. Traverse the entire power spectrum and identify all those that satisfy... and and The local maxima of the given conditions; where, For frequency The power spectral density amplitude at that point, This refers to the frequency resolution.

[0043] For example, N is set to 5, and the amplitudes and corresponding frequencies of all found local maxima are recorded. The recorded results are sorted in descending order of amplitude, and the top 5 peaks are selected. The frequencies corresponding to these 5 peaks, such as 75kHz, 120kHz, 185kHz, 250kHz, and 310kHz, are determined as the center frequencies for constructing the oscillator array.

[0044] The aforementioned exponential function is merely an example of a basic theoretical model for calculating the damping coefficient. In a preferred embodiment of the invention, the center frequency of each oscillator corresponds to its spectral peak frequency, and the time-varying damping coefficient is determined based on the corresponding coefficient modulus in the initial time-frequency distribution. That is, for any oscillator, the preferred method for calculating the time-varying damping coefficient is as follows:

[0045] in, For the first An oscillator in time The time-varying damping coefficient at the location, For the initial time-frequency distribution corresponding to the first The center frequency and scale of the oscillator At the present moment The magnitude of the coefficient, It represents the maximum value of the magnitudes of all coefficients in the initial time-frequency distribution. This is the set damping adjustment normal value.

[0046] For constructing an array of N parallel time-varying parameter oscillators, specifically, each time-varying parameter oscillator is configured as a second-order linear narrowband resonant filter system, the input of which is the reconstructed input signal. The output is a narrowband enhanced signal from the oscillator. Its dynamic response model is realized using a second-order linear ordinary differential equation. The natural angular frequency is determined by the peak frequency of the spectrum corresponding to the oscillator, and the time-varying damping coefficient is calculated using the above method.

[0047] For the selected N, which consists of 5 prominent frequencies ,in The range of values ​​is Five parallel second-order linear oscillators were constructed. The dynamic behavior of each oscillator was also determined by... Indicated; the natural angular frequency is calculated as .

[0048] in, The reconstructed input signal, For the first The output response of each oscillator is the narrowband enhanced signal. For the first The natural angular frequency of the oscillator The damping coefficient is time-varying. It is a time variable.

[0049] Calculating the time-varying damping coefficient At that time, record the maximum value of the magnitude of all coefficients in the entire time-frequency distribution obtained from the first continuous wavelet transform. For the first An oscillator, at any given moment Find the center frequency The wavelet scale corresponding to the initial time-frequency distribution that has been pre-calculated and stored in the aforementioned steps. And directly read the coefficient modulus of that point. The normalized coefficient modulus at this point is... At the same time, adjust the damping to normal value. Set it to 0.5. This parameter setting allows the transient signal to operate at a frequency of... The nearby energy is relatively strong, i.e., the coefficient modulus is high. When it is large, the time-varying damping coefficient The corresponding decrease prompts the oscillator to approach its resonant state, thereby effectively enhancing the output response of that frequency component; conversely, when the energy near that frequency is weaker, i.e., the coefficient magnitude is lower... When it is small, the time-varying damping coefficient The increase in amplitude causes the oscillator to decay rapidly, thus suppressing noise. By numerically solving these five differential equations using the fourth-order Runge-Kutta method, five narrowband enhancement signals can be obtained. to .

[0050] For constructing a fused time-frequency energy matrix by combining the initial time-frequency distribution, the value of the fused time-frequency energy matrix at any coordinate point is the product of the coefficient magnitude of the initial time-frequency distribution at that coordinate point and the sum of the coefficient magnitudes of all N enhanced time-frequency distributions at the same coordinate point.

[0051] Specifically, the N obtained from the solution represents 5 narrowband enhancement signals. Perform two consecutive wavelet transforms, using the same mother wavelet (e.g., complex Morlet wavelet) and the same scale range as the first transform, to obtain five enhanced time-frequency distributions. Next, a fused time-frequency energy matrix with the exact same dimensions as the initial time-frequency distribution is created. The method for calculating the value of any coordinate point in this matrix is ​​as follows:

[0052] in, To fuse the time-frequency energy matrix in terms of scale With time The values ​​at the coordinate points formed. The initial time-frequency distribution at scale With time The modulus of the coefficients at the same coordinate point constituted. For the first The magnitude of the coefficients of the enhanced time-frequency distribution at this coordinate point. To indicate the first An index corresponding to a frequency.

[0053] This calculation process is physically equivalent to a logical AND operation. Only when a certain time-frequency point has a prominent energy characteristic in the original signal, and at least one narrowband oscillator exists that can detect and enhance this energy characteristic, will the energy value of that point in the fusion matrix be significantly prominent. This calculation method effectively improves the signal-to-noise ratio of the traveling wavefront on the time-frequency diagram and provides good suppression of interference signals and background noise unrelated to the strand breakage fault.

[0054] S3. Extract the propagation path of the modulus maxima from the fused time-frequency energy matrix. Determine the arrival time of the initial traveling wave front based on the first entry of the index reflecting the singularity characteristics on the path into the preset mutation threshold interval. Determine the arrival time of the reflected wave front at the fault point based on the index re-entering the preset mutation threshold interval with the opposite polarity of the wave front. Calculate the fault distance based on the pre-calibrated propagation speed of the traveling wave in the conductor and the single-end traveling wave ranging principle.

[0055] In an optional embodiment, in the fused time-frequency energy matrix, starting from the finest scale (i.e., the smallest scale), all local maxima are searched along the time axis. From each maxima, the search continues to the nearest coarser scale (i.e., a larger scale) to find the point with the highest energy and connects them, iterating this process until the largest scale is reached, thus obtaining multiple modulus maxima propagation paths spanning all scales. For each path, at each time point... At this point, extract the data along the path at different scales. The time-frequency coefficient magnitude under In a double logarithmic coordinate system, Set as the horizontal axis Let be the vertical axis, and perform linear least squares fitting on the above points. Let the slope of the fitted line be denoted as . Then at that moment The Lipschitz index can be used as an indicator to reflect singularity characteristics and is calculated as follows: To avoid detection failure due to electromagnetic interference in actual power grid operating conditions and wavefront distortion caused by enhancement algorithms, it is necessary to preset an adaptive abrupt change threshold range, such as... ,from Begin searching along the timeline, recording indicators such as the Lipschitz index. The moment of first entry into this interval This moment marks the arrival of the initial traveling wavefront. Continuing the search, we find... The moment of re-entering the interval At the same time, comparison and The sign of the real part of the time-frequency transform coefficients of the path corresponding to the modulus maxima at a certain reference scale; if the signs are opposite, then it is confirmed. This represents the arrival time of the reflected wavefront.

[0056] Furthermore, the arrival time of the initial traveling wave front is determined based on the first entry of the indicators reflecting singularity characteristics along the path into a preset abrupt change threshold interval; that is, the preset abrupt change threshold interval is set as... After extracting the propagation path of the modulus maxima from the fused time-frequency energy matrix, the modulus values ​​of the coefficients at corresponding positions are taken from the initial time-frequency coefficient matrix according to the scale and time of each point on the path, and the slope is obtained by linear fitting in logarithmic coordinates. According to the aforementioned slope, it is expressed as The theoretical equation can be used to calculate the Lipschitz exponent by algebraic rearrangement. When the index value falls into the preset mutation threshold range for the first time, it is determined that a fault traveling wave has been detected.

[0057] In the generated fusion time-frequency energy matrix In the middle, along the timeline Each point on the scale axis Search for local maxima of energy and connect these maxima to obtain the propagation path of the modulus maxima. Record the scale of each point on each path. and time Then locate the coefficient magnitude at the same coordinate position in the initial time-frequency coefficient matrix. The arriving wavefront exhibits singular characteristics, and theoretically, the modulus maxima are directly proportional to the scale. ,in, This is the Lipschitz exponent. In logarithmic coordinates, and The relationship is linear, and the slope is expressed as Therefore, at each point along the propagation path of the modulus maxima... Nearby, calculated through linear fitting right Local slope ,in The scale for this data point. For the time period of this data point, the local Lipschitz exponent at that point is: Since the ideal traveling wavefront is a step-change signal, its theoretical Lipschitz exponent is 0. Considering the combined effects of conductor dispersion causing a gradual change in the wavefront and noise-induced pseudo-changes, the threshold interval for the change in the faulty traveling wavefront is adaptively set to a small neighborhood centered at 0. ,from Start monitoring and calculating along the time axis Value, when Record the time when you first enter this interval. As the arrival time of the initial traveling wavefront; continue monitoring, when The value re-enters the interval, and through analysis of the traveling wave... and The polarity of the original signal near the time, such as confirmation. The time is the positive peak and Record the time after the moment reaches a negative peak (i.e., the polarity is reversed). The arrival time of the wavefront reflected from the fault point.

[0058] The pre-defined propagation speed of traveling waves in a conductor is a specific value determined through theoretical calculations or field measurements based on the conductor type, material, and installation environment parameters. Its range is typically [value missing] per second. meters per second rice.

[0059] Traveling wave propagation speed Calibration is fundamental to accurate positioning. During implementation, the specific model of the overhead insulated conductor must be obtained, for example, a 10kV overhead insulated aluminum stranded wire of model JKLGY-10-150 / 25. The inductance and capacitance parameters per unit length should be consulted from the technical manual for this model. The theoretical propagation speed can be calculated using the formula... Calculated; where, For the theoretical propagation speed, Let be the inductance per unit length of the conductor. This represents the capacitance per unit length of the conductor. This calculation process also needs to consider the environmental factors such as the conductor's installation height, phase-to-phase distance, and ground conductivity, which affect the inductance and capacitance parameters.

[0060] For example, the theoretical speed, obtained through calculation and correction, is [value] per second. As an alternative or verification method, during power outage maintenance, a standard pulse signal can be injected at the measurement point at the beginning of the line, while simultaneously short-circuiting the signal at the end of the line. A fault recording device equipped with a high-precision synchronous clock can then be used to record the time of pulse signal transmission and the time of receiving a first reflected signal. The actual propagation speed is calculated by inversely using the principle; among which, Given the total length of the route, To calculate the actual propagation speed, This represents the round-trip propagation time difference from the transmission to reception of the pulse signal. Substitute the calibrated velocity value into the single-end ranging formula. In this way, the distance between the measurement point and the point of strand breakage can be calculated.

[0061] To verify the effectiveness of this scheme, comparative tests were conducted. The experimental conditions were set as follows: a 10 kV, 10 km overhead insulated conductor model was built using PSCAD; the line type was JKLGY-150 / 25; and the traveling wave velocity was calibrated to [value missing] per second. Meters. Strand failure faults are introduced at different locations along the line, such as 2 km, 5 km, and 8 km, and Gaussian white noise of varying intensities is superimposed to achieve signal-to-noise ratios of 20 dB, 10 dB, and 5 dB, respectively. Comparison Scheme 1 represents the complete process proposed in this invention; Scheme 2 removes the oscillator array enhancement and time-frequency energy fusion steps, performing singularity feature analysis only on the time-frequency distribution after the initial time-frequency domain transformation, using wavelet transform and Lipschitz exponent as examples; Scheme 3 is the control scheme, a traditional traveling wave detection method that directly extracts the modulus maxima from the original signal through time-frequency domain transformation.

[0062] Reference Figure 3Under a signal-to-noise ratio (SNR) of 20 dB, Scheme 1 achieved 100% fault detection accuracy with an average positioning error of 0.45%; Scheme 2's accuracy dropped to 98.6%, with the average positioning error increasing to 0.95%; and Scheme 3's accuracy was only 94.1%, with an average positioning error as high as 2.56%. When the SNR decreased to 10 dB, Scheme 1 maintained an accuracy of 99.2% with an error of 0.86%; Scheme 2's accuracy decreased to 91.3%, with the error increasing to 1.84%; and Scheme 3's accuracy plummeted to 78.5%, with an error of 4.88%. In a strong noise environment of 5 dB, Scheme 1 still achieved an accuracy of 97.5% and a positioning error of 1.12%, while the performance of Schemes 2 and 3 showed a significant decline, with accuracies dropping to 82% and 65.3%, respectively.

[0063] The performance improvement of Scheme 1 compared to Scheme 2 is mainly attributed to the synergistic effect of the oscillator array and the fused time-frequency energy matrix. This combined structure can perform narrowband enhancement on the key frequency band containing fault information based on the signal's own spectral characteristics, and through product fusion, it highlights the traveling wave front energy characteristics that exist simultaneously in the original and enhanced signals, effectively suppressing random noise and background interference. The improvement of Scheme 2 compared to Scheme 3 is due to the reasonable determination of the target frequency band of the digital bandpass filter. That is, the direct filtering effect of the digital bandpass filter effectively filters out most of the power frequency and high-frequency noise, providing a cleaner signal for subsequent analysis. Therefore, through the synergistic effect of digital bandpass filtering, oscillator enhancement, and energy fusion, this invention can achieve the extraction and detection of weak strand breakage fault traveling wave signals at a high signal-to-noise ratio.

[0064] It should be noted that those skilled in the art can make various modifications and improvements without departing from the inventive concept, and these all fall within the scope of protection of this invention. Therefore, the scope of protection of this patent should be determined by the appended claims.

Claims

1. A transient traveling wave localization method for broken strand faults in overhead insulated conductors, characterized in that, include: The transient current signal at one end of the line is acquired. Based on the frequency band characteristics of the broken strand fault, a digital bandpass filter is used to filter the transient current signal to extract the target frequency band signal and obtain the reconstructed signal. The reconstructed signal is then subjected to an initial time-frequency domain transformation to obtain the initial time-frequency distribution. Calculate the power spectrum of the input signal, identify the frequencies corresponding to the N prominent spectral peaks in the power spectrum, and construct an oscillator array consisting of N parallel time-varying parameter oscillators. Feed the input signal into the oscillator array to obtain N narrowband enhanced signals, wherein the center frequency of each oscillator is consistent with the corresponding spectral peak frequency, and the time-varying damping coefficient is determined based on the corresponding coefficient magnitude in the initial time-frequency distribution. Each of the N narrowband enhancement signals is subjected to a second time-frequency domain transformation to obtain N enhancement time-frequency distributions, and a fused time-frequency energy matrix is ​​constructed by combining the initial time-frequency distributions. The propagation path of the modulus maxima is extracted from the fused time-frequency energy matrix. The arrival time of the initial traveling wave front is determined based on the index reflecting the singularity characteristics on the path entering the preset mutation threshold range for the first time. The arrival time of the reflected wave front at the fault point is determined based on the index entering the preset mutation threshold range again and the wave front polarity being reversed. The fault distance is calculated based on the pre-calibrated propagation speed of the traveling wave in the conductor and the single-end traveling wave ranging principle.

2. The transient traveling wave localization method for a broken strand fault in an overhead insulated conductor according to claim 1, characterized in that, The step of using a digital bandpass filter to filter the transient current signal and extract the target frequency band signal based on the frequency band characteristics of the strand breakage fault includes: setting the frequency band characteristics of the strand breakage fault to a frequency range of 50kHz to 500kHz; configuring a digital bandpass filter so that its passband range corresponds to the frequency range of 50kHz to 500kHz; inputting the single-end transient current signal of the line into the digital bandpass filter to perform filtering processing; and using the output signal as the reconstructed signal.

3. The transient traveling wave localization method for a broken strand fault in an overhead insulated conductor according to claim 1, characterized in that, The calculation of the power spectrum of the input signal and the identification of the frequencies corresponding to the N prominent spectral peaks in the power spectrum include: calculating the power spectral density of the input signal using a fast Fourier transform algorithm; setting a peak search threshold of 10% of the maximum amplitude of the power spectrum; searching for all local maxima points in the power spectrum whose amplitudes exceed the threshold as spectral peaks; arranging all identified spectral peaks in descending order of amplitude, selecting the N spectral peaks with relatively large amplitudes, and identifying the frequencies corresponding to these N spectral peaks as the N prominent spectral peak frequencies.

4. The transient traveling wave localization method for a broken strand fault in an overhead insulated conductor according to claim 1, characterized in that, The construction of the oscillator array consisting of N parallel time-varying parameter oscillators specifically includes: configuring each time-varying parameter oscillator as a second-order linear narrowband resonant filter system; feeding the reconstructed input signal into each oscillator for narrowband resonant enhancement processing; the natural angular frequency of each oscillator is determined by its corresponding spectral peak frequency; and each oscillator is configured with a time-varying damping coefficient that dynamically adjusts with time; the output dynamic response of each oscillator is the corresponding narrowband enhancement signal.

5. The transient traveling wave localization method for a broken strand fault in an overhead insulated conductor according to claim 4, characterized in that, For any oscillator, the time-varying damping coefficient The calculation method is as follows: in, For the initial time-frequency distribution corresponding to the first The center frequency and scale of the oscillator At the present moment The magnitude of the coefficient, It represents the maximum value of the magnitudes of all coefficients in the initial time-frequency distribution. This is the set damping adjustment normal value.

6. The transient traveling wave localization method for a broken strand fault in an overhead insulated conductor according to claim 1, characterized in that, The construction of the fused time-frequency energy matrix by combining the initial time-frequency distribution includes: The value of the fused time-frequency energy matrix at any coordinate point is the product of the coefficient magnitude of the initial time-frequency distribution at that coordinate point and the sum of the coefficient magnitudes of the N enhanced time-frequency distributions at the same coordinate point; The calculation method is as follows: in, To fuse the time-frequency energy matrix in terms of scale With time The values ​​at the coordinate points formed. The initial time-frequency distribution at scale With time The modulus of the coefficients at the same coordinate point constituted. For the first The magnitude of the coefficients of the enhanced time-frequency distribution at this coordinate point.

7. The transient traveling wave localization method for a broken strand fault in an overhead insulated conductor according to claim 1, characterized in that, The step of determining the initial arrival time of the traveling wavefront based on the first entry of an index reflecting singularity characteristics along the path into a preset abrupt change threshold interval includes: calculating the Lipschitz exponent at each point as the index reflecting singularity characteristics; and using an adaptive interval reflecting the true attenuation and abrupt change characteristics of the wavefront as the preset abrupt change threshold interval, setting this interval as... After extracting the propagation path of the modulus maxima from the fused time-frequency energy matrix, the modulus values ​​of the coefficients at corresponding positions are taken from the initial time-frequency coefficient matrix according to the scale and time of each point on the path, and the slope is obtained by linear fitting in logarithmic coordinates. Then, the Lipschitz exponent was calculated. When the index value falls into the preset mutation threshold range for the first time, it is determined that a fault traveling wave has been detected.

8. The transient traveling wave localization method for a broken strand fault in an overhead insulated conductor according to claim 1, characterized in that, The pre-calibrated propagation speed of the traveling wave in the conductor is determined by theoretical formulas based on the conductor type, material, and installation environment parameters. Calculate the determined theoretical propagation speed, where, For the theoretical propagation speed, Let be the inductance per unit length of the conductor. The capacitance per unit length of the conductor; or by injecting a standard pulse signal at the beginning of the line, using a synchronous clock to record the time difference between the emitted and received round-trip reflected signals, and through... The actual propagation speed is calculated by inversely using the principle; among which, Given the total length of the route, To calculate the actual propagation speed, This is the round-trip propagation time difference between the transmission and reception of the pulse signal.

9. The transient traveling wave localization method for a broken strand fault in an overhead insulated conductor according to claim 1, characterized in that, In the step of acquiring the single-end transient current signal of the line and reconstructing the signal, and then performing the first time-frequency domain transformation: a high-speed data acquisition card is used to acquire the current transient quantity after the fault occurs at a sampling frequency of not less than 1MHz to obtain a discrete time series; in the time-frequency domain transformation, a short-time Fourier transform window function or a wavelet basis in the form of complex Morlet wavelets is selected as the analysis function, and the first time-frequency domain transformation operation is performed on the filtered discrete time series.

10. The transient traveling wave localization method for a broken strand fault in an overhead insulated conductor according to claim 1, characterized in that, The step of determining the arrival time of the reflected wavefront of the fault point based on the indicator re-entering the preset mutation threshold interval and having the opposite wavefront polarity specifically includes: continuing to search for the modulus maxima propagation path; when the indicator re-enters the preset mutation threshold interval, comparing the real part signs of the time-frequency transformation coefficients of the modulus maxima path corresponding to the initial arrival time of the traveling wavefront and the current time at a certain reference scale; if the real part signs are opposite, that is, the polarity of the original signal is opposite, then confirming the current time as the arrival time of the reflected wavefront of the fault point.

Images