Fault location method for primary and secondary fusion circuit breaker based on traveling wave distance measurement

Abstract

The application belongs to the technical field of measuring electric variables, and particularly relates to a primary and secondary fusion circuit breaker fault positioning method based on traveling wave ranging, which comprises the following steps: obtaining a fault transient signal of a primary and secondary fusion circuit breaker; obtaining a time-frequency amplitude matrix by adopting a short-time Fourier transform; constructing a transient time-frequency gradient matrix by a nonlinear square and root operation on the time-frequency amplitude matrix; constructing a non-stationary state perception base on the transient time-frequency gradient matrix; generating a dynamic penalty factor according to the non-stationary state perception base and a pre-extracted background power spectrum benchmark; combining the dynamic penalty factor with an inherent frequency response function to construct an adaptive convolution function, reconstructing the time-frequency amplitude matrix, and obtaining a high-bandwidth traveling wave signal; simultaneously performing wavelet modulus maximum value analysis, extracting a traveling wave head arrival time, and calculating and outputting a fault positioning distance. The application improves the precision of power distribution network traveling wave reconstruction and fault positioning under non-stationary conditions.

Description

Technical Field

[0001] This invention relates to the field of electrical variable measurement technology. More specifically, this invention relates to a method for fault location of primary and secondary integrated circuit breakers based on traveling wave ranging. Background Technology

[0002] As a core node device in modern smart distribution networks, integrated primary and secondary circuit breakers highly integrate the primary high-voltage switch body with the secondary intelligent terminal. In fault location technology for long-distance mixed lines in distribution networks, traveling wave ranging technology based on transient electrical signals has become the preferred technology due to its characteristics of being unaffected by transition resistance and line topology. However, the conventional electronic voltage or current sensors built into integrated primary and secondary circuit breakers are usually designed for power frequency measurement, and their physical characteristics are manifested as a low-pass filter, which leads to the problem of frequency band response limitation of built-in sensors. When a fault traveling wave containing high-frequency components at the megahertz level passes through, the high-frequency wavefront is severely attenuated and smoothed, making it impossible to accurately extract the arrival time of the traveling wave.

[0003] To compensate for the high-frequency loss caused by sensors, the Wiener inversion filtering algorithm is often used for signal reconstruction. This algorithm treats the low-pass characteristics of the sensor as a linear system and uses the inverse filtering process to deconvolve and restore the high-frequency traveling wave edge. However, in practical applications, the traditional Wiener filtering algorithm requires a fixed ratio of signal power spectrum to noise power spectrum in its frequency domain derivation. When a real fault occurs in the distribution network, it is often accompanied by arc discharge. The transient electromagnetic interference generated inside the circuit breaker and the fault signal itself are highly non-stationary and unpredictable. Relying on fixed static prior parameters, when facing non-stationary and drastically changing fault conditions, the algorithm may either over-penalize, resulting in a still flat wavefront, or over-compensate, causing high-frequency electromagnetic noise to be drastically amplified, thus completely drowning out the real traveling wavefront. Summary of the Invention

[0004] To address the technical problem of static parameter failure in the Wiener inversion filtering algorithm under non-stationary conditions, this invention provides a fault location method for integrated primary and secondary circuit breakers based on traveling wave ranging. The method includes: acquiring the transient fault signal collected by the secondary control terminal of the integrated primary and secondary circuit breaker at the instant a line fault is detected; mapping the transient fault signal to a two-dimensional time-frequency domain using short-time Fourier transform to obtain a time-frequency amplitude matrix; calculating the partial derivatives of the time-frequency amplitude matrix in the time and frequency dimensions, and constructing a transient time-frequency gradient matrix through nonlinear sum-of-squares root-finding operations; constructing a non-stationary state sensing basis based on the transient time-frequency gradient matrix using the decay characteristics of the natural exponential function to characterize the consistency of the signal's energy distribution across the entire frequency band; generating a dynamic penalty factor based on the non-stationary state sensing basis and a pre-extracted background power spectrum benchmark; combining the dynamic penalty factor with the inherent frequency response function of the built-in sensor of the integrated primary and secondary circuit breaker to construct an adaptive convolution function, and reconstructing the time-frequency amplitude matrix to restore the high-bandwidth traveling wave signal; performing wavelet modulus maxima analysis on the reconstructed high-bandwidth traveling wave signal to calculate and output the fault location distance.

[0005] Preferably, the window function used in the short-time Fourier transform is the Hanning window.

[0006] Preferably, the time-frequency amplitude matrix is ​​obtained by performing a short-time Fourier transform on the acquired fault transient signal to obtain an initial complex matrix characterizing the signal's changes with time and frequency, and then further calculating the modulus of the initial complex matrix.

[0007] Preferably, the formula for calculating the transient time-frequency gradient matrix is: In the formula, This represents the transient time-frequency gradient matrix at time t and frequency f; Represents a discrete-time index; Represents a discrete frequency index; This represents the time-frequency amplitude matrix obtained by the short-time Fourier transform at time t and frequency f; The sign for partial derivatives; This represents the abrupt change rate of signal energy with time t while keeping the frequency f constant. This represents the spread of signal energy distribution with frequency f when time t remains constant; and These are the time step and frequency step, determined by the system sampling parameters, respectively; It is the system's reference time scale. It is the reference frequency scale.

[0008] This gradient matrix calculation formula organically integrates the instantaneous energy influx rate in the time dimension and the energy distribution expansion rate in the frequency dimension through nonlinear sum of squares root-finding operations. It can characterize the potential transient wavefront edge in two-dimensional space, avoid misjudgment caused by single-dimensional analysis in complex backgrounds, and enhance feature contrast.

[0009] Preferably, the formula for calculating the non-stationary state sensing basis is: In the formula, The cardinality representing the non-stationary state at time t; Represents a discrete-time index; Represents a discrete frequency index; This represents the total number of discrete frequency points along the frequency dimension of the short-time Fourier transform, corresponding to the maximum discrete frequency index. ; Represents a specific frequency at time t The transient time-frequency gradient matrix; This represents the arithmetic mean of the gradients across all frequency bands at time t. The symbol for the natural exponential function; It represents the absolute deviation of the gradient of a specific frequency band from the average level.

[0010] By utilizing the nonlinear decay characteristics of the natural exponential function as a physical filter, the synchronous response consistency of the real traveling wave generated across the entire frequency band can be captured. By suppressing the random discrete arc noise to near zero and amplifying the wavefront signal that conforms to the consistency characteristics to the sensing peak, a high-sensitivity enhanced extraction of the real wavefront is achieved.

[0011] Preferably, the formula for calculating the dynamic penalty factor is: In the formula, This represents the dynamic penalty factor at time t and frequency f; Represents the discrete-time index, indicating the position of the signal on the time axis; Represents a discrete frequency index; This represents the background power spectrum reference pre-extracted and mapped to the frequency domain at frequency f; represent The cardinality of non-stationary states at any given time; Represents the state baseline reference value; This represents the time-frequency amplitude matrix obtained by the short-time Fourier transform at time t and frequency f; This is the average amplitude during steady-state operation; The symbol for the logarithmic function; This is the symbol for the hyperbolic tangent activation function.

[0012] By combining the hyperbolic tangent activation function and the logarithmic function, the penalty factor is given the physical property of automatically contracting and expanding during the transient evolution process. The penalty is greatly reduced at the moment of wavefront impact to maximize the inverse filtering gain, and it recovers rapidly during the noise period to suppress the amplification of high-frequency noise, thus ensuring the robustness and fidelity of the reconstruction algorithm in complex electromagnetic environments.

[0013] Preferably, the adaptive convolution function is calculated as follows: In the formula, This represents the adaptive convolution function at time t and frequency f; Represents a discrete-time index; Represents a discrete frequency index; Represents the natural frequency; This represents the conjugate complex number of the inherent frequency response function; The factory-defined inherent frequency response function of the built-in sensor in the primary and secondary integrated circuit breaker; Represents the squared magnitude of the frequency response function; This represents the dynamic penalty factor at time t and frequency f.

[0014] The adaptive convolution function combines the sensor's inherent frequency response function and its conjugate complex number, which can effectively correct the phase shift and bandwidth attenuation generated when the signal passes through the hardware. It controls the signal-to-noise ratio balance in the inverse filtering process in real time through dynamic adjustment terms, achieving the optimal balance between sharpening the restored wavefront and maintaining the purity of the reconstruction.

[0015] Preferably, the process of extracting the arrival time of the traveling wave front is combined with the high-precision time synchronization information of the double-ended circuit breaker to perform cross-correlation matching correction on the initially detected wave front time, thereby eliminating timing jitter caused by noise disturbance in single-ended detection.

[0016] Preferably, the wavelet modulus maxima analysis uses the Daubechies4 wavelet as the mother wavelet.

[0017] Preferably, the output fault location distance is obtained by adding the total line length to the product of the known wave speed and the time difference between the two ends, and then arithmetically averaging the results.

[0018] The beneficial effects of this invention are as follows:

[0019] The method provided by this invention breaks through the technical bottleneck of traditional Wiener inversion filtering, which must rely on a constant noise-to-signal power spectrum ratio. By deeply mining the gradient consistency characteristics of transient electrical signals inside the primary and secondary integrated circuit breaker in both time and frequency domains, a state basis that can accurately characterize the evolution of non-stationary physical phenomena in the distribution network is constructed. Furthermore, this state basis is converted into a dynamic penalty factor and seamlessly integrated into the core deconvolution algorithm. This achieves automatic removal of high-frequency penalty at the moment of abrupt change in the traveling wave front to maximize bandwidth restoration, while automatically increasing the penalty during electromagnetic interference to suppress random noise amplification. This method does not require the addition of expensive high-frequency traveling wave sensors to the primary and secondary integrated circuit breaker. Through purely data-level physical adaptive reconstruction, it can perfectly restore the steep real traveling wave front under various extremely non-stationary and severe arc interference conditions, improving the fault location and fault finding accuracy of hybrid lines in the distribution network. Attached Figure Description

[0020] Figure 1 This is a flowchart illustrating the primary and secondary fusion circuit breaker fault location method based on traveling wave ranging in this invention. Figure 2 This is a curve showing the correlation between the non-stationary state perception base and the dynamic penalty factor. Figure 3 It is an adaptive gain allocation distribution diagram; Figure 4 This is a comparison chart of the reconstruction effects of the fault traveling wave. Detailed Implementation

[0021] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0022] The specific embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0023] This invention discloses a primary and secondary fusion circuit breaker fault location method based on traveling wave ranging, referring to... Figure 1 This includes steps S1 to S5: S1: Obtain the fault transient signal of the primary and secondary fusion circuit breaker; use short-time Fourier transform to obtain the time-frequency amplitude matrix.

[0024] The secondary control terminal of the integrated primary and secondary circuit breaker acquires the fault transient signal at the moment a line fault is detected. Specifically, the secondary control terminal of the integrated primary and secondary circuit breaker monitors the sudden changes in the line's power frequency current or voltage, the zero-sequence current amplitude, or the sequence component changes of the current / voltage in real time. When the sudden change or sequence component exceeds the preset protection start threshold, it determines that a line fault has occurred and triggers the acquisition and storage of the fault transient signal. The one-dimensional time-domain fault transient signal is mapped to a two-dimensional time-frequency domain space using short-time Fourier transform. In conventional implementations, to ensure that the signal has good resolution in both the time and frequency domains to capture the megahertz-level traveling wave high-frequency components, the window function used in the short-time Fourier transform is preferably the Hanning window.

[0025] Furthermore, the Hanning window length parameter used in the short-time Fourier transform is used to balance the time-domain resolution and frequency-domain resolution, directly affecting the capture accuracy of the high-frequency components of the megahertz-level traveling wave. If the window length is set too short, the frequency-domain resolution is insufficient, making it difficult to accurately separate the traveling wave's main frequency from the background noise, resulting in blurred wavefront characteristics. If the window length is set too long, the time-domain localization capability decreases, making it impossible to accurately locate the moment of the traveling wave's abrupt change, thus weakening the ranging time accuracy. Therefore, the reasonable range for this window length parameter is 256 to 1024 sampling points, and in this embodiment, it is set to 512 sampling points. This setting value is based on balancing the high-frequency component resolution capability and the accuracy of locating time-domain abrupt changes, avoiding both the wavefront delay estimation deviation caused by an excessively wide window and the introduction of spectral leakage interference due to an excessively narrow window, thereby achieving a balance between the integrity of the traveling wave characteristics and time-frequency focusing. In other embodiments, the implementer can adjust the window length parameter within the range according to the line type, fault transient rise time, and sampling system bandwidth.

[0026] It should be noted that after performing a short-time Fourier transform on the acquired one-dimensional fault transient signal, an initial complex matrix representing the signal's variation with time and frequency is obtained. The modulus of this initial complex matrix is ​​further calculated to obtain the time-frequency amplitude matrix. The inherent frequency response function parameters recorded by the sensor during the factory calibration or pre-inspection stage are retrieved in advance through the storage module. The inherent frequency response function is obtained by injecting a sweep frequency excitation signal or a step pulse signal into the sensor before the equipment is put into operation or during maintenance, and simultaneously recording the sensor's response output. Based on the frequency domain relationship between the input and output, the amplitude-frequency characteristics and phase-frequency characteristics of the sensor in the full frequency band are identified, and the inherent frequency response function in the form of a complex function is fitted and stored in the storage module.

[0027] S2: Construct the transient time-frequency gradient matrix by performing a nonlinear sum-of-squares root operation on the time-frequency amplitude matrix.

[0028] After the initial time-frequency mapping is completed, due to the low-pass filtering characteristics of the built-in sensors of the primary and secondary fusion circuit breaker, the energy of the traveling wave front is smoothed in both the time and frequency domains, which manifests as a slow amplitude transition. In order to accurately extract the real transient changes from the weakened signal, this step combines the physical laws of distribution network fault occurrence and enhances the distinguishability of features through nonlinear fusion of multi-dimensional energy gradients.

[0029] Furthermore, when the primary and secondary fusion equipment is subjected to a short-circuit fault transient impact, the energy distribution of current or voltage signals across different frequency bands will undergo drastic and asymmetric abrupt changes. Based on this physical phenomenon, this step evaluates the abrupt change by calculating the partial derivatives of the time-frequency amplitude matrix in the time and frequency dimensions. The partial derivative with respect to time is used to capture the instantaneous influx of energy over time, while the partial derivative with respect to frequency is used to capture the extent of energy diffusion across the entire frequency band. Simultaneously, based on the abrupt change rate of signal energy in the time dimension and the expansion rate of energy distribution in the frequency dimension, the system performs a nonlinear sum-of-squares root-finding operation to construct the transient time-frequency gradient matrix, calculated as follows:

[0030] In the formula, This represents the transient time-frequency gradient matrix at time t and frequency f. The larger the value, the higher the probability that there is a transient wavefront edge at that location. Represents the discrete-time index, indicating the position of the signal on the time axis; Represents a discrete frequency index; This represents the time-frequency amplitude matrix obtained from the short-time Fourier transform at time t and frequency f, recording the signal at each time point. and each frequency point The magnitude of energy on it; The partial derivative sign is used to capture... The degree of drastic change over time or frequency; This represents the abrupt change rate of signal energy with time t while keeping the frequency f constant. This represents the spread of signal energy distribution with frequency f when time t remains constant; and These are the time step and frequency step, determined by the system sampling parameters, respectively. It is the system's reference time scale. This is the reference frequency scale, making the quantities within parentheses dimensionless. The reference time scale is also included. The effective time width of the window function used in the short-time Fourier transform is taken as the reference frequency scale. The width of the frequency domain resolution unit corresponding to the short-time Fourier transform is taken as ; in the discrete implementation, It equals the product of the number of sampling points corresponding to the window function length and the system sampling period. It is equal to the ratio of the system sampling rate to the number of points in the short-time Fourier transform frequency domain.

[0031] It should be noted that, as shown in the above calculation formula, when the transient electrical signal experiences a broadband energy surge in a very short time, the instantaneous rate of change of the time-frequency amplitude matrix in both the time and frequency domains will increase dramatically, leading to a significant increase in the two partial derivative terms. Consequently, the value of the transient time-frequency gradient matrix will increase exponentially. This positive mapping relationship can accurately characterize the potential transient wavefront edge in a two-dimensional matrix, effectively avoiding misjudgments caused by single-dimensional analysis. This provides a high-contrast feature basis for accurately extracting the arrival time of traveling waves in non-stationary arc noise.

[0032] S3: Construct a non-stationary state sensing basis based on the transient time-frequency gradient matrix.

[0033] After extracting the transient time-frequency gradient matrix, the next step is to distinguish between the real traveling wave front and continuous high-frequency arc noise. The real traveling wave front produces a highly consistent energy mapping across the entire frequency band within an extremely short time window, while the frequency band distribution of non-stationary arc noise exhibits extreme randomness and discreteness. The transient time-frequency gradient matrix is ​​evaluated using the decay characteristics of the natural exponential function, and a non-stationary state sensing basis is constructed, calculated as follows:

[0034] In the formula, The non-stationary state sensing basis at time t is represented by its value, which reflects whether the signal in the current sampling window has the physical characteristics of a true traveling wave with full-band consistency. Represents the discrete-time index, indicating the position of the signal on the time axis; Represents a discrete frequency index; This represents the total number of discrete frequency points along the frequency dimension of the short-time Fourier transform, corresponding to the maximum discrete frequency index. ; The transient time-frequency gradient matrix at time t and frequency f represents the intensity of the energy mutation at that frequency band. It represents the arithmetic mean of the gradients of all frequency bands at time t, and serves as a benchmark reference for measuring the global energy transition level at that time. The symbol for the natural exponential function is used to utilize its nonlinear decay characteristics as a physical filter to amplify the uniformity feature and suppress discrete noise.

[0035] Furthermore, the calculation formula utilizes the decay characteristics of the natural exponential function for physical-logical transformation; since the real traveling wave front will cause a full-band synchronous response at the instant of transient impact, the gradient values ​​of each band are close to each other and the overall value is extremely large, making the absolute value of the difference... Approaching zero, the independent variable of the natural exponent term approaches zero, causing the product term of the corresponding frequency band to remain at its maximum value, ultimately outputting an extremely high state-aware peak non-stationary state-aware base, thereby achieving enhanced extraction of the true wavefront; conversely, if the signal is in a period of non-stationary random electromagnetic interference, the gradient distribution of each frequency band exhibits extremely large random differences, leading to an increase in the absolute value of the difference, the independent variable of the exponent term becomes a negative number with a large absolute value, and the exponent term then rapidly decays to approach zero, thereby suppressing the output of the non-stationary state-aware base of the dependent variable.

[0036] S4: Generate a dynamic penalty factor based on the non-stationary state sensing basis and the pre-extracted background power spectrum benchmark.

[0037] In the Wiener inversion filtering algorithm, the setting of the noise prior parameters directly determines the balance between signal reconstruction fidelity and noise suppression. To overcome the failure problem caused by parameter staticization under non-stationary fault conditions, this step utilizes the non-stationary state sensing basis to adjust the intensity of noise suppression in real time. The calculation formula for the dynamic penalty factor is as follows:

[0038] In the formula, The dynamic penalty factor at time t and frequency f is used to adjust the compensation strength of the Wiener filter for high-frequency components online. Its real-time change determines the weight distribution between sharpening and noise reduction of the reconstructed signal. Represents the discrete-time index, indicating the position of the signal on the time axis; Represents a discrete frequency index; The background power spectrum benchmark, which is pre-extracted and mapped to the frequency domain at frequency f, is a basic reference value determined based on the background noise distribution under the steady-state operating environment of the equipment. represent The non-stationary state perception base at any given moment is used to identify in real time whether the current moment is in the physical evolution period of the traveling wave front impact; It represents the state baseline reference value, which is equal to the peak value of the non-stationary state perception baseline calculated by the same algorithm process under standard test fault conditions. This represents the time-frequency amplitude matrix obtained from the short-time Fourier transform at time t and frequency f, recording the signal at each time point. and each frequency point The magnitude of energy on it; The signal energy is standardized to represent the average amplitude during steady-state operation. The symbol for the logarithmic function is used to nonlinearly compress the amplitude to enhance characteristic stability and prevent numerical divergence caused by excessive impact energy. This is the symbol for the hyperbolic tangent activation function, used to map the input state to a normalized gain interval and determine the degree of expansion or contraction of the denominator term.

[0039] Furthermore, this calculation formula endows the penalty factor with the physical property of automatically contracting and expanding during transient evolution through nonlinear mapping; when the signal is at the moment of impact of the traveling wave front, the non-stationary state sensing basis of the input parameters increases sharply, causing the hyperbolic tangent function to... The output value rapidly approaches its extreme value of 1, causing the denominator to expand and approach 2, resulting in a significant contraction of the output dynamic penalty factor, which is much smaller than the inherent background power spectrum reference. In this state, due to the extremely small dynamic penalty factor, the system will execute the maximum inverse filter gain to fully restore the high-frequency edges attenuated by the sensor. Conversely, when the traveling wave front subsides and enters the non-stationary arc noise period, The perceptual base of non-stationary states at any given moment is at a low point. The function output approaches zero, and the denominator shrinks to 1, causing the dynamic penalty factor to rise rapidly and approach the background power spectrum reference, thereby strongly suppressing noise amplification and avoiding severe distortion in wavefront reconstruction.

[0040] For example, Figure 2 The graph shows the linkage between the non-stationary state sensing base and the dynamic penalty factor. This graph reveals the system's real-time sensing and parameter mapping mechanism for transient physical evolution. The non-stationary state sensing base utilizes the decay characteristics of the natural exponential function as a physical filter. At the moment of impact of the traveling wave wavefront, it rapidly generates a pulsed extremum due to the consistency of the full-band synchronous response, accurately locking the traveling wave characteristics. Correspondingly, the dynamic penalty factor automatically shrinks its value when the sensing base spikes through the nonlinear mapping logic of the hyperbolic tangent activation function, allowing the system to remove the restriction on high-frequency gain to restore the wavefront. During the electromagnetic interference period, it rises back to the background power spectrum reference as the sensing base falls, thus strongly suppressing noise amplification and ensuring the robustness of the algorithm in complex electromagnetic environments.

[0041] S5: Combine the dynamic penalty factor with the inherent frequency response function to construct an adaptive convolution function, reconstruct the time-frequency amplitude matrix, and obtain a high-bandwidth traveling wave signal; at the same time, perform wavelet mode maxima analysis to extract the arrival time of the traveling wave front, calculate and output the fault location distance.

[0042] After obtaining the dynamic penalty factor, the system enters the core reconstruction and localization stage. This step aims to utilize a dynamically adjusted gain mechanism to compensate for sensor physical defects while suppressing non-stationary noise, thereby reconstructing a high-bandwidth true traveling wave signal. The formula for the adaptive convolution function is:

[0043] In the formula, The adaptive convolution function at time t and frequency f determines the strength of gain compensation or noise suppression of the original signal at a specific time and frequency point. Represents the discrete-time index, indicating the position of the signal on the time axis; Represents a discrete frequency index; The conjugate complex number representing the inherent frequency is used to correct the phase shift that occurs when the signal passes through the sensor; The factory-defined frequency response function of the built-in sensor of the secondary fusion circuit breaker at frequency f describes the attenuation characteristics of the hardware for signals of different frequencies. The square of the modulus of the frequency response function reflects the energy conversion efficiency of the system in the frequency domain; Representing the dynamic penalty factor, it is a regularization factor after normalization. Its magnitude matches the energy distribution of the sensor's frequency response function and is used to control the signal-to-noise ratio balance in the inverse filtering process in real time.

[0044] Furthermore, the adaptive convolution function modifies the initial time-frequency matrix through pointwise complex multiplication, whereby the pointwise complex multiplication refers to applying the adaptive convolution function... At each time frequency point, the initial complex matrix obtained by the short-time Fourier transform is... The above multiplication yields the corrected time-frequency complex matrix, which is then subjected to a short-time Fourier inverse transform to restore the high-bandwidth traveling wave signal in the time domain. When the dynamic penalty factor dynamically shrinks to a minimum value at the abrupt change in the wavefront driven by the state-aware cardinality, the weights in the denominator used for noise suppression drop sharply, causing the gain of the transfer adaptive convolution function to approach the ideal state of the inverse filter, and fully performing the inverse filtering operation to accurately compensate for the high-frequency attenuation truncated by the sensor. During the electromagnetic interference amplification period, as the value of the dynamic penalty factor rises again, the denominator term increases, which strongly suppresses the amplification gain of the transfer function for high-frequency components, thereby effectively avoiding excessive restoration of background noise and ensuring the purity of the wavefront reconstruction.

[0045] It should be noted that after the signal reconstruction is completed, the system performs wavelet mode maxima analysis on the reconstructed high-bandwidth traveling wave signal to extract the arrival time of the traveling wave front. The extraction process of the arrival time of the traveling wave front is combined with the high-precision time synchronization information of the double-ended circuit breaker to perform cross-correlation matching correction on the initially detected wave front time to eliminate timing jitter caused by noise disturbance in single-ended detection.

[0046] It should be further noted that, in the specific implementation, the Daubechies4 wavelet is preferably used as the mother wavelet for modulus maxima analysis because it has excellent transient change capture capability. To ensure positioning accuracy, a wavelet modulus maxima judgment threshold needs to be set. The reasonable range of this threshold is usually set between [0.20, 0.45], and in this embodiment, it is preferably set to 0.32. The setting of 0.32 is based on the balance between the capture sensitivity of weak traveling wave rising edges and the ability to prevent false triggering. It can filter out the small ripples generated by numerical calculations and ensure the effective identification of the real wavefront under complex working conditions such as high-resistance grounding, thereby achieving a balance between positioning accuracy and system reliability. In other embodiments, the implementer can flexibly adjust the threshold within the range according to the total line length and sampling rate limitations. Finally, by combining the time difference between the two ends and the known wave velocity, the accurate fault location distance is output.

[0047] After extracting the arrival time of the wavefronts of the two-terminal circuit breakers, the system performs physical location using the principle of two-terminal traveling wave ranging. First, it obtains the arrival times of the traveling wave at the beginning and end of the line and calculates the time difference between the two ends. Then, combining the pre-set total line length and the known wave speed of the traveling wave in the line, the time difference is converted into a distance difference. By adding the total line length to the product of the known wave speed and the time difference between the two ends, and then arithmetically averaging the results, the actual physical distance from the fault point to one of the circuit breakers can be accurately calculated, thus completing the high-precision fault location of the mixed distribution network line.

[0048] For example, Figure 3 This figure illustrates the adaptive gain allocation distribution, depicting the intelligent gain allocation strategy of the system's adaptive convolution function during signal reconstruction. The system adjusts the Wiener filter's compensation for high-frequency components online based on a real-time generated dynamic penalty factor. During the extremely short window period of wavefront abrupt changes, the transfer function approaches the ideal inverse filtering state due to the significant collapse of the penalty factor, fully executing gain operation to accurately compensate for the high-frequency attenuation truncated by the hardware sensor. During periods of non-stationary arc noise, the system automatically increases the penalty weight to lock the gain window, strongly preventing excessive restoration of background noise. This achieves an optimal balance between sharpening and noise suppression, ensuring the purity of fault location accuracy.

[0049] For example, Figure 4 This figure shows a comparison of the fault traveling wave reconstruction results. The comparison verifies the superiority of the proposed solution in handling non-stationary conditions of the distribution network. By constructing a dynamic adjustment mechanism, this solution effectively identifies and restores the megahertz-level high-frequency components that are attenuated and smoothed by the low-pass characteristics of the sensor. At the same time, it uses targeted suppression strategies to eliminate the interference of background noise, achieving accurate reconstruction of the steep real traveling wave front. This provides a clear feature basis for the subsequent accurate extraction of the traveling wave arrival time.

Claims

1. A method for fault location of a primary and secondary integrated circuit breaker based on traveling wave ranging, characterized in that, include: Acquire the fault transient signal collected by the secondary control terminal of the primary and secondary integrated circuit breaker at the moment a line fault is detected; The fault transient signal is mapped to a two-dimensional time-frequency domain space using short-time Fourier transform to obtain the time-frequency amplitude matrix; Calculate the partial derivatives of the time-frequency amplitude matrix in the time and frequency dimensions, and construct the transient time-frequency gradient matrix through nonlinear sum-of-squares root-finding operations; Based on the transient time-frequency gradient matrix, a non-stationary state sensing basis is constructed using the decay characteristics of the natural exponential function to characterize the consistency of the signal's energy distribution across the entire frequency band. A dynamic penalty factor is generated based on the non-stationary state sensing baseline and the pre-extracted background power spectrum benchmark. By combining the dynamic penalty factor with the inherent frequency response function of the built-in sensor of the primary and secondary fusion circuit breaker, an adaptive convolution function is constructed, and the time-frequency amplitude matrix is ​​reconstructed to restore the high-bandwidth traveling wave signal. Wavelet mode maxima analysis is performed on the reconstructed high-bandwidth traveling wave signal to extract the arrival time of the traveling wave front, and the fault location distance is calculated and output.

2. The method for fault location of a primary and secondary circuit breaker based on traveling wave ranging according to claim 1, characterized in that, The window function used in the short-time Fourier transform is the Hanning window.

3. The method for fault location of a primary and secondary circuit breaker based on traveling wave ranging according to claim 1, characterized in that, The time-frequency amplitude matrix is ​​obtained by performing a short-time Fourier transform on the acquired fault transient signal to obtain an initial complex matrix characterizing the signal's changes with time and frequency, and then further calculating the modulus of the initial complex matrix.

4. The method for fault location of a primary and secondary circuit breaker based on traveling wave ranging according to claim 1, characterized in that, The formula for calculating the transient time-frequency gradient matrix is: ； In the formula, This represents the transient time-frequency gradient matrix at time t and frequency f; Represents a discrete-time index; Represents a discrete frequency index; This represents the time-frequency amplitude matrix obtained by the short-time Fourier transform at time t and frequency f; The sign for partial derivatives; This represents the abrupt change rate of signal energy with time t while keeping the frequency f constant. This represents the spread of signal energy distribution with frequency f when time t remains constant; and These are the time step and frequency step, determined by the system sampling parameters, respectively; It is the system's reference time scale. It is the reference frequency scale.

5. The method for fault location of a primary and secondary circuit breaker based on traveling wave ranging according to claim 1, characterized in that, The formula for calculating the non-stationary state sensing basis is: ； In the formula, The cardinality representing the non-stationary state at time t; Represents a discrete-time index; Represents a discrete frequency index; This represents the total number of discrete frequency points along the frequency dimension of the short-time Fourier transform, corresponding to the maximum discrete frequency index. ; Represents a specific frequency at time t The transient time-frequency gradient matrix; This represents the arithmetic mean of the gradients across all frequency bands at time t. The symbol for the natural exponential function; It represents the absolute deviation of the gradient of a specific frequency band from the average level.

6. The method for fault location of a primary and secondary circuit breaker based on traveling wave ranging according to claim 1, characterized in that, The formula for calculating the dynamic penalty factor is: In the formula, This represents the dynamic penalty factor at time t and frequency f; Represents the discrete-time index, indicating the position of the signal on the time axis; Represents a discrete frequency index; This represents the background power spectrum reference pre-extracted and mapped to the frequency domain at frequency f; represent The cardinality of non-stationary states at any given time; Represents the state baseline reference value; This represents the time-frequency amplitude matrix obtained by the short-time Fourier transform at time t and frequency f; This is the average amplitude during steady-state operation; The symbol for the logarithmic function; This is the symbol for the hyperbolic tangent activation function.

7. The method for fault location of a primary and secondary circuit breaker based on traveling wave ranging according to claim 1, characterized in that, The formula for calculating the adaptive convolution function is: In the formula, This represents the adaptive convolution function at time t and frequency f; Represents a discrete-time index; Represents a discrete frequency index; This represents the conjugate complex number of the inherent frequency response function; The factory-defined inherent frequency response function of the built-in sensor in the primary and secondary integrated circuit breaker; Represents the squared magnitude of the frequency response function; This represents the dynamic penalty factor at time t and frequency f.

8. The method for fault location of a primary and secondary circuit breaker based on traveling wave ranging according to claim 1, characterized in that, The process of extracting the arrival time of the traveling wave front is combined with the high-precision time synchronization information of the double-ended circuit breaker to perform cross-correlation matching correction on the initially detected wave front time, thereby eliminating timing jitter caused by noise disturbance in single-ended detection.

9. The method for fault location of a primary and secondary circuit breaker based on traveling wave ranging according to claim 1, characterized in that, The wavelet modulus maxima analysis uses the Daubechies4 wavelet as the mother wavelet.

10. The method for fault location of a primary and secondary circuit breaker based on traveling wave ranging according to claim 1, characterized in that, The output fault location distance is obtained by adding the product of the known wave speed and the time difference between the two ends to the total line length, and then taking the arithmetic average of the results.

Images