Power distribution network optical fiber fault locating method and system based on time domain reflection

Abstract

The application provides a power distribution network optical fiber fault positioning method and system based on time domain reflection, relates to the technical field of power distribution networks, and comprises the following steps: obtaining optical fiber basic parameter information, generating and phase-modulating an optical carrier signal, dividing the modulated signal into chip sequences and performing orthogonal coding and time domain interleaving modulation to form an enhanced detection signal with space-time diversity characteristics; injecting the signal into an optical fiber and receiving a reflected signal, extracting fault characteristic fingerprint information by using multi-scale wavelet transform and adaptive threshold screening, realizing fault type identification and accurate positioning, and effectively improving fault detection accuracy and positioning precision.

Description

TECHNICAL FIELD

[0001] The present application relates to power distribution network technology, and in particular to a power distribution network optical fiber fault location method and system based on time domain reflection. BACKGROUND

[0002] As an important part of the power system, the safe and stable operation of the power distribution network is of great significance to the protection of social and economic development and people's life. With the advancement of smart grid construction, power distribution network optical fiber communication systems are widely used in the monitoring, protection and control of power systems. Optical fiber communication has become the main way of power distribution network communication due to its high bandwidth, anti-electromagnetic interference and information security advantages. However, due to external environmental factors, human damage and equipment aging, etc., various faults such as fiber breakage, bending, and extrusion may occur in the power distribution network optical fiber, which will directly affect the communication quality and operation reliability of the power distribution network. Therefore, quickly and accurately locating the fault point of the optical fiber is of great significance to shorten the fault repair time and improve the reliability of the power distribution network.

[0003] Time domain reflection technology is a common method for optical fiber fault location, which basically works by transmitting a probe signal into the optical fiber and analyzing the returned reflection signal to determine the fault location. Although traditional optical time domain reflection technology has been widely used in the field of optical fiber fault location, it still has some problems and challenges in the complex environment of the power distribution network.

[0004] The existing power distribution network optical fiber fault location technology has the following defects and deficiencies: First, the traditional probe signal design lacks spatial and temporal diversity characteristics, and has limited anti-interference ability in complex noise environments, resulting in low fault location accuracy and high false alarm rate in the power distribution network optical fiber system under strong electromagnetic interference and complex environment. Second, the conventional signal processing method has insufficient detection sensitivity for weak fault signals, making it difficult to identify early minor faults or complex situations where multiple faults exist simultaneously, and unable to accurately identify fault types and assess fault severity. Finally, the existing fault feature extraction algorithm fails to fully consider the feature differences of different fault types in the multi-scale domain, lacks an adaptive feature selection mechanism, and is difficult to maintain stable and reliable positioning performance in the dynamically changing power distribution network environment, especially in long-distance optical fibers, where the positioning error increases significantly with distance. SUMMARY

[0005] The embodiments of the present application provide a power distribution network optical fiber fault location method and system based on time domain reflection, which can solve the problems in the prior art.

[0006] In a first aspect of the embodiments of the present application, a power distribution network optical fiber fault location method based on time domain reflection is provided, comprising:

[0007] Obtaining basic parameter information of the power distribution network optical fiber, generating an initial optical carrier signal according to the basic parameter information, and phase modulating the initial optical carrier signal to obtain a modulated probe signal;

[0008] Dividing the modulated probe signal into a plurality of chip sequences, encoding each chip sequence using an orthogonal code sequence, and interleaving and modulating the encoded plurality of chip sequences in the time domain to generate an enhanced probe signal with space-time diversity characteristics;

[0009] Injecting the enhanced probe signal into the power distribution network optical fiber, receiving a reflected signal returned by the power distribution network optical fiber, and demodulating the reflected signal to extract phase information and amplitude information from the reflected signal;

[0010] Performing fault feature decomposition on the phase information and the amplitude information based on multi-scale wavelet transform, setting an adaptive threshold at each scale level for feature screening, mapping the screened features to a high-dimensional feature space through a nonlinear activation function for modeling, and extracting fault feature fingerprint information;

[0011] Performing fault type recognition and location positioning according to the fault feature fingerprint information to obtain fault type determination results and fault location coordinates, correcting the fault location according to the fault type determination results and the fault location coordinates in combination with the basic parameter information to obtain final fault positioning results, and generating a fault positioning report based on the final fault positioning results.

[0012] Generating an initial optical carrier signal according to the basic parameter information, and phase modulating the initial optical carrier signal to obtain a modulated probe signal includes:

[0013] Analyzing the optical fiber transmission characteristics on the time scale corresponding to the basic parameter information, extracting a principal singular value component reflecting the dispersion characteristics of the optical fiber, performing nonlinear compensation on the optical carrier signal according to the principal singular value component to generate an initial optical carrier signal, calculating the phase modulation depth based on the principal singular value component, and phase modulating the initial optical carrier signal to obtain a modulated probe signal.

[0014] Dividing the modulated probe signal into a plurality of chip sequences, encoding each chip sequence using an orthogonal code sequence, and interleaving and modulating the encoded plurality of chip sequences in the time domain to generate an enhanced probe signal with space-time diversity characteristics includes:

[0015] Multiplying the modulated probe signal by a transform kernel function and performing fractional order transform processing to obtain a transform processing result;

[0016] extracting a signal feature component of the transform processing result in a transform domain, determining an energy distribution threshold based on the signal feature component, determining a signal division position according to the energy distribution threshold, and dividing the modulated probe signal into a plurality of chip sequences based on the signal division position;

[0017] extracting a feature frequency point from the transform processing result, performing feature matching on the feature frequency point and the transform kernel function, and generating a group of orthogonal code sequences based on a result of the feature matching;

[0018] encoding each orthogonal code sequence in the group of orthogonal code sequences with a corresponding chip sequence to obtain a group of encoded chip sequences, and determining an interleaving parameter based on a feature enhancement effect of the transform processing result;

[0019] performing time sequence rearrangement on the group of encoded chip sequences according to the interleaving parameter, and performing interleaving modulation on the rearranged chip sequences to generate an enhanced probe signal with space-time diversity characteristics.

[0020] performing fault feature decomposition on the phase information and the amplitude information based on multi-scale wavelet transform, setting an adaptive threshold at each scale level for feature screening, mapping the screened features to a high-dimensional feature space through a nonlinear activation function for modeling, and extracting fault feature fingerprint information, including:

[0021] performing multi-scale wavelet transform on the phase information and the amplitude information to obtain feature components at a plurality of scale levels, calculating feature weights using a nonlinear activation function with a learnable shape parameter according to time-frequency energy distribution of the feature components at each scale level;

[0022] performing dynamic quantization coding on the feature components at each scale level based on the feature weights, calculating a mapping relationship between the coded feature components using the nonlinear activation function, quantitatively reconstructing the feature components according to the mapping relationship, and obtaining multi-scale reconstructed features;

[0023] improving feature discrimination using the learnable shape parameter of the nonlinear activation function on the multi-scale reconstructed features, determining an adaptive threshold based on the distribution of the mapped features, and dividing the multi-scale reconstructed features into main features and auxiliary features;

[0024] mapping the main features and the auxiliary features using the nonlinear activation function respectively, and iteratively optimizing the learnable shape parameter to obtain mapped high-dimensional features;

[0025] mapping the mapped high-dimensional features to different feature expression spaces using the nonlinear activation function, adaptively fusing outputs of different feature expression spaces, and generating final fault feature fingerprint information.

[0026] The feature components of each scale level are dynamically quantization coded based on the feature weights, a mapping relationship between the coded feature components is calculated using the nonlinear activation function, the feature components are quantization reconstructed according to the mapping relationship, and multi-scale reconstructed features are obtained, including:

[0027] A dynamic quantization encoder is constructed according to feature weights, a quantization step is calculated based on the feature weights, and quantization feature values are generated by wavelet decomposition based on the quantization step;

[0028] The encoding parameters of the dynamic quantization encoder are determined based on the precision of the quantization feature values, the quantization feature values of different scale levels are wavelet packet transformed by the dynamic quantization encoder, and an encoded feature group is generated, the quantization precision of each encoded feature in the encoded feature group being determined by the encoding depth and the encoding base;

[0029] The encoded feature group and the quantization feature values are wavelet reconstructed according to the encoding depth and the encoding base to obtain reconstructed features, and the encoding parameters of the dynamic quantization encoder are adjusted based on the reconstructed features;

[0030] The reconstructed features of multiple scale levels are encoded using the updated encoding parameters to generate a feature encoding sequence, a fusion weight is generated based on the feature encoding sequence, and the fusion weight and the reconstructed features are jointly reconstructed to obtain final multi-scale fusion features.

[0031] Fault type recognition and location positioning are performed according to the fault feature fingerprint information, and fault type determination results and fault position coordinates are obtained, including:

[0032] A classification model is constructed according to the fault feature fingerprint information, an adaptive wavelet decomposition scale is set for the fault feature fingerprint information, a feature energy value is calculated at each wavelet decomposition scale, and a feature energy distribution is obtained;

[0033] The feature energy distribution is input into the classification model, a probability value of each fault type is calculated, the type with the largest probability value is determined as the fault type determination result, a corresponding wavelet base function is selected according to the fault type determination result, the fault feature fingerprint information is analyzed in time and frequency using the wavelet base function, and an energy distribution sequence on a time-frequency plane is obtained;

[0034] A spatial mapping curve is established according to the energy distribution sequence, an energy extreme point is searched on the spatial mapping curve, and a spatial position corresponding to the energy extreme point is determined as the fault position coordinates.

[0035] According to the fault type determination result and the fault position coordinates, the basic parameter information is combined to correct the fault position, and a final fault positioning result is obtained, and based on the final fault positioning result, a fault positioning report is generated, including:

[0036] According to the fault type determination result and the fault position coordinates, the basic parameter information is combined to construct a chaotic attractor state space, the fault position coordinates are mapped to the chaotic attractor state space to obtain a chaotic orbit, and the state evolution sequence of the chaotic orbit is calculated based on the evolution equation of the chaotic attractor state space;

[0037] The state mapping matrix is used to iteratively correct the fault position coordinates, the step of the iterative correction is related to the singular value of the state mapping matrix, when the singular value change amount of the state mapping matrix is less than a preset change threshold, the iteration is stopped, the current fault position coordinates are determined as the final fault positioning result, and a fault positioning report is generated based on the stability analysis of the chaotic orbit and the characteristic change point.

[0038] The second aspect of the embodiment of the application provides a power distribution network optical fiber fault positioning system based on time domain reflection, including:

[0039] The first unit is used for acquiring basic parameter information of the power distribution network optical fiber, generating an initial optical carrier signal according to the basic parameter information, and performing phase modulation on the initial optical carrier signal to obtain a modulated probe signal;

[0040] The second unit is used for dividing the modulated probe signal into a plurality of chip sequences, encoding each chip sequence using an orthogonal code sequence, and interlacing and modulating the encoded plurality of chip sequences in the time domain to generate an enhanced probe signal with space-time diversity characteristics;

[0041] The third unit is used for injecting the enhanced probe signal into the power distribution network optical fiber, receiving a reflected signal returned by the power distribution network optical fiber, and performing demodulation processing on the reflected signal to extract phase information and amplitude information in the reflected signal;

[0042] The fourth unit is used for performing fault feature decomposition on the phase information and the amplitude information based on multi-scale wavelet transform, setting an adaptive threshold on each scale level for feature screening, mapping the screened features to a high-dimensional feature space for modeling through a nonlinear activation function, and extracting fault feature fingerprint information;

[0043] A fifth unit is configured to identify a fault type and locate a fault position according to the fault feature fingerprint information, to obtain a fault type determination result and a fault position coordinate, to correct the fault position according to the fault type determination result and the fault position coordinate, to obtain a final fault positioning result, and to generate a fault positioning report based on the final fault positioning result.

[0044] In a third aspect, the embodiment of the present application provides an electronic device, comprising:

[0045] a processor;

[0046] a memory for storing processor-executable instructions;

[0047] The processor is configured to invoke the instructions stored in the memory to execute the method described above.

[0048] In a fourth aspect, the embodiment of the present application provides a computer-readable storage medium, which stores computer program instructions, and the computer program instructions are executed by a processor to implement the method described above.

[0049] The present application has the following beneficial effects:

[0050] The present application generates a detection signal with enhanced anti-interference capability by phase-modulating an initial optical carrier signal and using a quadrature code sequence for space-time diversity interleaving modulation, effectively improves the fault detection sensitivity and accuracy in a complex power distribution network environment, and solves the technical problem of inaccurate positioning of traditional detection methods under strong noise interference.

[0051] The present application uses multi-scale wavelet transform to perform feature decomposition on the reflected signal, sets an adaptive threshold at each scale level for feature screening, and maps the screened features to a high-dimensional feature space through a nonlinear activation function, which can accurately capture the subtle feature differences of different types of faults and greatly improve the accuracy of fault type identification and the precision of fault position positioning.

[0052] The present application corrects the fault position in combination with the basic parameter information of the power distribution network optical fiber and generates a detailed fault positioning report, realizes rapid and accurate positioning of the power distribution network optical fiber fault, reduces the operation and maintenance cost, improves the reliability and safety of the power distribution network, and has significant economic and social benefits. BRIEF DESCRIPTION OF DRAWINGS

[0053] Figure 1 FIG. 1 is a flowchart of a power distribution network optical fiber fault positioning method based on time domain reflection according to an embodiment of the present application;

[0054] Figure 2 FIG. 4 is a detection signal enhancement processing flowchart according to an embodiment of the present application.

[0055] Figure 3 Flow chart of chaos mapping for fault location of embodiments of the present application. DETAILED DESCRIPTION

[0056] In order to make the objects, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by a person of ordinary skill in the art without creative work fall within the protection scope of the present application.

[0057] The technical solutions of the present application will be described in detail below with specific embodiments. The following specific embodiments can be combined with each other, and the same or similar concepts or processes can not be described in some embodiments.

[0058] Figure 1 Flow chart of the power distribution network optical fiber fault location method based on time domain reflection of embodiments of the present application, as shown in Figure 1 The method comprises:

[0059] Obtaining basic parameter information of the power distribution network optical fiber, generating an initial optical carrier signal according to the basic parameter information, and performing phase modulation on the initial optical carrier signal to obtain a modulated probe signal;

[0060] Dividing the modulated probe signal into a plurality of chip sequences, encoding each chip sequence using an orthogonal code sequence, and interleaving and modulating the encoded plurality of chip sequences in the time domain to generate an enhanced probe signal with space-time diversity characteristics;

[0061] Injecting the enhanced probe signal into the power distribution network optical fiber, receiving a reflected signal returned by the power distribution network optical fiber, and performing demodulation processing on the reflected signal to extract phase information and amplitude information in the reflected signal;

[0062] Performing fault feature decomposition on the phase information and the amplitude information based on multi-scale wavelet transform, setting an adaptive threshold at each scale level for feature screening, mapping the screened features to a high-dimensional feature space through a nonlinear activation function for modeling, and extracting fault feature fingerprint information;

[0063] According to the fault feature fingerprint information, fault type recognition and location positioning are performed to obtain a fault type determination result and a fault position coordinate. According to the fault type determination result and the fault position coordinate, the fault position is corrected in combination with the basic parameter information to obtain a final fault positioning result. Based on the final fault positioning result, a fault positioning report is generated.

[0064] In an optional implementation, an initial optical carrier signal is generated according to the basic parameter information, and the initial optical carrier signal is phase-modulated to obtain a modulated probe signal.

[0065] The optical fiber transmission characteristics are analyzed on a time scale corresponding to the basic parameter information, and a main singular value component reflecting the dispersion characteristics of the optical fiber is extracted. The optical carrier signal is nonlinearly compensated according to the main singular value component to generate an initial optical carrier signal. The phase modulation depth is calculated based on the main singular value component, and the initial optical carrier signal is phase-modulated to obtain a modulated probe signal.

[0066] The basic parameter information is acquired in an optical fiber transmission system, including an optical fiber length of 50 kilometers, a dispersion coefficient of 17 picoseconds / (nanometer·kilometer), a nonlinear coefficient of 1.3 watts·kilometer -1 ·kilometer -1 , an optical fiber loss of 0.2 decibels / kilometer, an input optical power of 5 milliwatts, a center wavelength of 1550 nanometers, and a signal bandwidth of 10 gigahertz. These parameters constitute the basis for analyzing the optical fiber transmission characteristics.

[0067] When analyzing the optical fiber transmission characteristics on a time scale corresponding to the acquired basic parameter information, a singular value decomposition method is used. Specifically, the optical fiber transmission equation is discretized into a matrix form, and the time sampling points are set to 1024, and the time resolution is 5 picoseconds. The optical fiber transmission matrix includes a dispersion term and a nonlinear term, wherein the dispersion term is realized by Fourier transform, and the nonlinear term is calculated by step-by-step iteration. The singular value decomposition is performed on the constructed optical fiber transmission matrix to obtain a series of singular values and corresponding left and right singular vectors.

[0068] From the decomposition result, the first three singular values with the largest contribution are selected as the main singular value components, and the corresponding values are 0.923, 0.756 and 0.421 respectively. The left and right singular vectors corresponding to the first main singular value 0.923 are denoted as u1 and v1 respectively, which contain the main information of the fiber dispersion characteristics. According to the actual calculation, the first ten elements of the u1 vector are [0.0312, 0.0345, 0.0378, 0.0456, 0.0534, 0.0623, 0.0712, 0.0801, 0.0845, 0.0889], and the first ten elements of the v1 vector are [0.0301, 0.0332, 0.0363, 0.0425, 0.0487, 0.0549, 0.0611, 0.0673, 0.0704, 0.0735].

[0069] According to the extracted main singular value components, the nonlinear compensation is performed on the optical carrier signal to generate an initial optical carrier signal. In the compensation process, an inverse transmission operator is constructed using the main singular value components, which can pre-compensate the nonlinear distortion in the fiber transmission process. In a specific implementation, the complex amplitude of the initial optical carrier signal is denoted as A(t), where t is the time variable. Let the power distribution function P(t) = |A(t)| 2 obeys a Gaussian distribution, with a peak power of 5 milliwatts and a pulse width of 100 picoseconds. The signal is convolved with the main singular value component u1 to obtain the pre-compensated signal A'(t).

[0070] In the pre-compensation calculation, the initial Gaussian pulse signal is discretized into 1024 points, and then convolved with the discretized u1 vector. The convolution result is multiplied by a compensation coefficient 0.87 (which is determined by optimization experiments). After this processing, the generated initial optical carrier signal shows a slightly compressed Gaussian shape in power, with a peak power of 4.35 milliwatts, but the pulse edges have small amplitude oscillation characteristics, which are designed to compensate for the nonlinear effects in the fiber transmission.

[0071] Based on the main singular value components, the phase modulation depth is calculated, and the initial optical carrier signal is phase modulated to obtain the modulated probe signal. The calculation of the phase modulation depth φ is related to the main singular value 0.923 and the corresponding singular vector v1. According to experimental optimization, φ = π × (1-0.923) × K, where K is the modulation coefficient, and its value is 2.5. The calculated phase modulation depth φ is about 0.1205π.

[0072] The phase modulation is realized by an electro-optical modulator, and the input is the initial optical carrier signal A'(t) generated above, and the modulation signal is cos(ω0t), where ω0 is the modulation frequency, which is set to 1 / 5 of the signal bandwidth, i.e., 2 GHz. The modulated probe signal is represented as A'(t)·exp(jφ·cos(ω0t)), where j is the imaginary unit. In actual implementation, the amplitude of the modulation signal is controlled at 0.8 volts, corresponding to the calculated phase modulation depth.

[0073] Through the above processing, the obtained modulated probe signal presents a quasi-Gaussian pulse with phase modulation in the time domain, and the power envelope is the same as the initial optical carrier signal, but the phase changes with time according to the cosine function. This modulation structure can effectively detect the nonlinear and dispersion characteristics in the optical fiber. In the frequency domain, the modulated signal presents a main lobe near the center frequency and a side lobe structure with a modulation frequency interval, and the side lobe power is about 0.01 times that of the main lobe.

[0074] This embodiment extracts the optical fiber transmission characteristics through singular value decomposition, realizes accurate pre-compensation and phase modulation of the optical carrier signal, and provides technical support for high-quality optical fiber detection. By adjusting the modulation parameters and pre-compensation coefficients, the detection requirements of different optical fiber systems can be met.

[0075] In an optional embodiment, the modulated probe signal is divided into a plurality of chip sequences, each of the chip sequences is encoded using an orthogonal code sequence, and the encoded plurality of chip sequences are interleaved and modulated in the time domain to generate an enhanced probe signal with space-time diversity characteristics, including:

[0076] The modulated probe signal is multiplied by a transform kernel function and subjected to fractional order transform processing to obtain a transform processing result;

[0077] Signal characteristic components of the transform processing result are extracted in the transform domain, an energy distribution threshold is determined based on the signal characteristic components, a signal division position is determined according to the energy distribution threshold, and the modulated probe signal is divided into a plurality of chip sequences based on the signal division position;

[0078] Feature frequency points are extracted from the transform processing result, the feature frequency points are subjected to feature matching with the transform kernel function, and a group of orthogonal code sequences is generated based on the feature matching result;

[0079] Each orthogonal code sequence in the group of orthogonal code sequences is encoded with a corresponding chip sequence to obtain a group of encoded chip sequences, and an interleaving parameter is determined based on the feature enhancement effect of the transform processing result;

[0080] The interleaving parameter is used to rearrange the time sequence of the coded chip sequence group, and the rearranged chip sequence is modulated to generate an enhanced probe signal with space-time diversity characteristics.

[0081] As shown in Figure 2 The method comprises:

[0082] The radar system generates a raw probe signal, uses a linear frequency modulation signal as a basic waveform, the signal bandwidth is 500 MHz, the pulse width is 10 μs, and the sampling rate is 1 GHz. The raw probe signal is generated by a digital waveform generator, and the signal-to-noise ratio is set to 15 dB. The system modulates the raw probe signal by quadrature amplitude modulation, and the modulation parameters include a carrier frequency of 10 GHz, a modulation order of 16, and a symbol rate of 50 MHz. The modulation uses a 16QAM modulation mode, and the bit stream is mapped to 16 points on the constellation diagram, each point representing 4 bits of information. The modulated probe signal has a data amount of 10000 sampling points and a signal duration of 10 μs.

[0083] The modulated probe signal is multiplied by a transform kernel function and subjected to fractional order transform processing to obtain a transform processing result. The transform kernel function is a fractional Fourier transform kernel, and the transform order α is set to 0.75. The selection of this transform order is based on signal characteristic analysis, and when α = 0.75, the linear frequency modulation signal has the best energy concentration characteristics. The specific form of the transform kernel function is the product of a complex exponential function and a non-integer power of time, and the kernel function length matches the signal length, which is 10000 points. The system multiplies the modulated probe signal point by point with the transform kernel function to obtain a weighted signal. The weighted signal is transformed by a discrete fractional Fourier transform processor, which is implemented using a fast algorithm with a computational complexity of O(N log N), where N is the signal length. The transform processing result is a complex number sequence containing 10000 complex number points, each containing a real part and an imaginary part.

[0084] In the transform domain, the signal characteristic components of the transform processing result are extracted, and the signal characteristic component extraction is based on energy distribution analysis. The system calculates the energy spectrum of the transform processing result, which is equal to the square of the modulus of the complex number sequence. For a 10000-point transform result, the energy spectrum also contains 10000 points. The system performs smoothing processing on the energy spectrum, and convolves a Hanning window with a length of 51 points to reduce the influence of noise. The smoothed energy spectrum presents a multi-peak distribution characteristic, containing multiple energy concentration regions, which correspond to different frequency components of the signal. The system calculates statistical characteristics based on the smoothed energy spectrum, including mean, standard deviation, kurtosis, and skewness. For the signal in this example, the energy spectrum mean is 0.025, the standard deviation is 0.018, the kurtosis is 3.75, and the skewness is 0.82.

[0085] The energy distribution threshold is determined based on the signal characteristic component, and the energy distribution threshold is set by an adaptive method. The calculation formula is the mean value plus the standard deviation multiplied by the adjustment coefficient. The adjustment coefficient is dynamically adjusted according to the kurtosis and skewness. For a distribution with high kurtosis, the adjustment coefficient increases; for a distribution with large skewness, the adjustment coefficient decreases. In this example, the adjustment coefficient is calculated as 2.1, and the energy distribution threshold is determined as 0.063. The system marks the points greater than the threshold in the energy spectrum as characteristic points, and marks the points less than the threshold as non-characteristic points. The characteristic points account for 28% of the total number of points, mainly distributed in a specific area of the transform domain, representing the main energy concentration area of the signal.

[0086] The signal division position is determined according to the energy distribution threshold. The signal division is based on the distribution characteristics of the characteristic points in the transform domain. The system combines the continuous characteristic points into characteristic regions, and a total of 4 main characteristic regions are identified, which are located at different positions in the transform domain. The system calculates the center position and span for each characteristic region. The center position corresponds to a specific time in the time domain signal, and the span corresponds to the duration of the time domain signal. Based on these parameters, the system determines the division position of the time domain signal, and divides the 10 μs signal into 4 chip sequences, each with a duration of about 2.5 μs and a sample point number of about 2500 points. The division position is specifically the 0th, 2500th, 5000th, 7500th and 10000th sample points of the original signal.

[0087] Based on the signal division position, the modulated probe signal is divided into multiple chip sequences. The original 10000-point signal is divided into 4 chip sequences according to the division position, each sequence containing about 2500 sample points. The first chip sequence contains the 0th to 2499th points of the original signal, the second chip sequence contains the 2500th to 4999th points, the third chip sequence contains the 5000th to 7499th points, and the fourth chip sequence contains the 7500th to 9999th points. The system performs energy normalization processing on each chip sequence to ensure that the energy levels of the sequences are consistent, facilitating subsequent processing. After normalization processing, the energy mean of the four chip sequences is 1, and the standard deviation is less than 0.05.

[0088] The characteristic frequency points are extracted from the transform processing result. The characteristic frequency points are the points corresponding to the energy spectrum peaks in the transform domain, representing the main frequency components of the signal. The system searches for local maximum points in the transform domain, sets the search window to 51 points, and when the energy value of the center point is greater than all other points in the window and greater than the energy distribution threshold, it is marked as a characteristic frequency point. The system identifies a total of 16 characteristic frequency points, located at different positions in the transform domain, with a frequency range covering 250MHz to 450MHz. The position index and energy value of each characteristic frequency point are recorded to form a characteristic frequency point set. The system arranges the characteristic frequency points in descending order of energy value and selects the top 8 points with the highest energy as key characteristic frequency points. The position indexes of these points are 1245, 3782, 5631, 6230, 7112, 7865, 8524, and 9130.

[0089] The characteristic frequency points are matched with the transform kernel function, and a set of orthogonal code sequences is generated based on the characteristic matching result. The characteristic matching process is based on the values of the transform kernel function at the characteristic frequency points. The system calculates the values of the transform kernel function at the 8 key characteristic frequency points to obtain 8 complex numbers. The system uses these complex numbers to construct a Walsh-Hadamard matrix, with a matrix order of 8 and each matrix element value of +1 or -1. The construction method is to quantize the phase of the complex number, with a phase between 0 and π quantized as +1 and a phase between π and 2π quantized as -1. The generated 8x8 matrix has orthogonality, with an inner product of 0 between any two rows or columns. The system extracts 4 rows from the matrix as a set of orthogonal code sequences, with each sequence length of 8 and sequence elements of +1 or -1. The selected 4 sequences are the 1st, 3rd, 5th, and 7th rows of the matrix, which are mutually orthogonal and suitable for encoding different chip sequences.

[0090] Each orthogonal code sequence in the set of orthogonal code sequences is encoded with the corresponding chip sequence to obtain a set of encoded chip sequences. The encoding process uses direct sequence spread spectrum technology to multiply each sample point of each chip sequence with the corresponding element in the orthogonal code sequence. Since the length of the orthogonal code sequence is 8 and the length of the chip sequence is about 2500, the system repeats the orthogonal code sequence periodically to expand it to the same length as the chip sequence. After encoding, the original information in each chip sequence is expanded to a wider frequency band, with anti-interference capability. After encoding of the four chip sequences with the four orthogonal code sequences, a set of encoded chip sequences is formed, each sequence still containing about 2500 sample points.

[0091] The interleaving parameters are determined based on the feature enhancement effect of the transformation processing result. The feature enhancement effect is evaluated by calculating the cross-correlation coefficient of the original signal and the signal after transformation processing. The cross-correlation coefficient is 0.78, indicating that the transformation processing has good feature enhancement effect. The interleaving parameters include two key parameters: interleaving depth and interleaving mode. The interleaving depth represents the number of units participating in interleaving, which is set to 32; the interleaving mode represents the interleaving rule, which adopts a pseudo-random interleaving mode. The system generates a pseudo-random sequence with a length of 32, and the sequence element value range is 0 to 31, with each element being unique. The pseudo-random sequence generation adopts a linear feedback shift register method, and the characteristic polynomial is selected as an 8-order irreducible polynomial, and the initial state is set to 0x5A. The generated pseudo-random sequence is [7, 15, 23, 31, 3, 11, 19, 27, 6, 14, 22, 30, 2, 10, 18, 26, 5, 13, 21, 29, 1, 9, 17, 25, 4, 12, 20, 28, 0, 8, 16, 24].

[0092] According to the interleaving parameters, the timing rearrangement is performed on the coded chip sequence group, and four coded chip sequence groups are combined into a long sequence with a length of about 10000 points. The system divides the long sequence into 32 equal-length data blocks, and each data block has a length of about 312 points. The 32 data blocks are rearranged according to the order specified by the pseudo-random sequence, for example, the 0th data block of the original sequence is moved to the 28th position of the new sequence, the 1st data block of the original sequence is moved to the 20th position of the new sequence, and so on. After rearrangement, the originally adjacent data blocks are scattered to different positions in the sequence, enhancing the anti-interference ability of the signal. After timing rearrangement, the system obtains a new long sequence with the same length as the original sequence, about 10000 points.

[0093] The rearranged chip sequence is interleaved and modulated to generate an enhanced probe signal with space-time diversity characteristics. The interleaved modulation adopts orthogonal frequency division multiplexing technology. The system divides the rearranged long sequence into 64 subcarriers, and each subcarrier modulates 312 / 64≈5 data points. The modulation adopts QPSK mode, which maps the data points to four points on the complex plane. The system generates 64 orthogonal subcarriers with a frequency interval of 7.8MHz, covering a bandwidth of 500MHz. The system modulates the mapped data points onto the corresponding subcarriers, and then adds all the subcarrier signals to obtain the enhanced probe signal in the time domain. The enhanced probe signal has time and frequency diversity characteristics. The time diversity comes from timing rearrangement, and the frequency diversity comes from orthogonal frequency division multiplexing. The final generated enhanced probe signal has a length of 10000 points, a duration of 10μs, and a bandwidth of 500MHz.

[0094] In practical applications, the system tests the enhanced detection signal in performance, and the results show that the signal has excellent performance in multipath environment. In a two-path environment with a reflection coefficient of 0.5, the distance resolution of the enhanced detection signal is 0.3 meters, which is 20% higher than that of the original detection signal; the sidelobe ratio of the waveform ambiguity function is -28dB, which is 5dB lower than that of the original signal; in an environment with a signal-to-noise ratio of 5dB, the target detection probability reaches 92%, which is 15 percentage points higher than that of the original signal. These performance improvements prove the effectiveness of the proposed signal processing method, and the enhanced detection signal generated by the method is suitable for target detection tasks in complex electromagnetic environments.

[0095] In an optional implementation, the phase information and the amplitude information are subjected to fault feature decomposition based on a multi-scale wavelet transform, an adaptive threshold is set on each scale level for feature screening, the screened features are mapped to a high-dimensional feature space through a nonlinear activation function for modeling, and the extraction of the fault feature fingerprint information includes:

[0096] The phase information and the amplitude information are subjected to multi-scale wavelet transform to obtain feature components at multiple scale levels, and a nonlinear activation function with a learnable shape parameter is used to calculate feature weights according to the time-frequency energy distribution of the feature components at each scale level;

[0097] The feature components at each scale level are subjected to dynamic quantization coding based on the feature weights, the mapping relationship between the coded feature components is calculated using the nonlinear activation function, the feature components are quantitatively reconstructed according to the mapping relationship, and multi-scale reconstructed features are obtained;

[0098] The learnable shape parameter of the nonlinear activation function is used to improve the feature discrimination of the multi-scale reconstructed features, and an adaptive threshold is determined based on the distribution of the mapped features, and the multi-scale reconstructed features are divided into main features and auxiliary features;

[0099] The main features and the auxiliary features are respectively mapped using the nonlinear activation function, and the learnable shape parameter is iteratively optimized to obtain the mapped high-dimensional features;

[0100] The mapped high-dimensional features are mapped to different feature expression spaces using the nonlinear activation function, and the outputs of different feature expression spaces are adaptively fused to generate the final fault feature fingerprint information.

[0101] Phase information and amplitude information of the power equipment are acquired, the phase information includes phase angle data of three-phase voltage and current, and the amplitude information includes amplitude data of three-phase voltage and current. The acquisition of the phase information and the amplitude information is completed by a power monitoring device, a sampling frequency is set to 12.8 kHz, 256 points are sampled per cycle, and a collection time length is 4 cycles before fault occurrence to 16 cycles after fault occurrence, a total of 20 cycle data, forming a 5120-point sampling sequence. The original data collected is preprocessed by a digital filter to remove power frequency interference and high-frequency noise, the filter adopts a Butterworth bandpass design, and a passband range is 10 Hz to 5 kHz. The phase information and the amplitude information after preprocessing are stored according to three phases respectively, forming six basic signal sequences.

[0102] Multi-scale wavelet transform is performed on the phase information and the amplitude information to obtain feature components of multiple scale levels, discrete wavelet transform is used for multi-scale decomposition, db4 wavelet basis function is selected, and 5-layer decomposition is performed. For the phase information and the amplitude information of each phase, the system obtains five levels of detail coefficients and one approximate coefficient of the coarsest scale through discrete wavelet transform. The detail coefficient represents the high-frequency component of the signal, and the approximate coefficient represents the low-frequency component of the signal. Taking the A-phase voltage amplitude signal as an example, the original signal contains 5120 sampling points, after 5-layer wavelet decomposition, the lengths of the detail coefficients of each layer are 2560, 1280, 640, 320 and 160 points respectively, and the length of the approximate coefficient of the coarsest scale is 160 points. These coefficients jointly constitute the multi-scale feature components of the signal.

[0103] According to the time-frequency energy distribution of the feature components of each scale level, a nonlinear activation function with a learnable shape parameter is used to calculate the feature weight. The time-frequency energy distribution is obtained by calculating the energy value of each level feature component, and the energy value is equal to the sum of squares of each coefficient in the feature component. For the 5-layer decomposition of the A-phase voltage amplitude signal, the energy percentage of each layer of detail coefficients is 5%, 8%, 15%, 25% and 40% respectively, and the energy percentage of the approximate coefficient of the coarsest scale is 7%. The system designs a nonlinear activation function with a learnable shape parameter, and the function form is a variant of the Swish function with a learnable shape parameter. The learnable shape parameter includes a slope parameter a and an offset parameter b, and the initial values are set to 1.0 and 0.0 respectively. The system inputs the energy percentage of each level into the nonlinear activation function to obtain the corresponding feature weight. After mapping by the activation function, the feature weights of the 5-layer detail coefficients are 0.15, 0.22, 0.35, 0.62 and 0.85 respectively, and the feature weight of the approximate coefficient of the coarsest scale is 0.18. The feature weight reflects the contribution degree of each level feature component to the fault feature.

[0104] The feature components of each scale level are dynamically quantized and encoded based on the feature weights, and a dynamic quantization encoder is constructed. The encoder adaptively sets the quantization parameter according to the feature weight. The quantization step is inversely proportional to the feature weight. The higher the feature weight, the smaller the quantization step, and the higher the quantization precision. The system sets the basic quantization step to 0.05, and calculates the actual quantization step of each level according to the feature weight. For the 5th layer detail coefficient with a feature weight of 0.85, the quantization step is 0.059; for the 1st layer detail coefficient with a feature weight of 0.15, the quantization step is 0.333. The system performs wavelet packet transform on the feature components based on the quantization step to generate quantized feature values. Wavelet packet transform is an extension of wavelet transform, which further subdivides the feature components of each scale level to provide finer frequency division. The system sets the wavelet packet decomposition level according to the feature weight. The higher the feature weight, the deeper the decomposition level. For the 4th and 5th layer detail coefficients with a feature weight greater than 0.6, the wavelet packet decomposition level is set to 3; for the 2nd and 3rd layer detail coefficients with a feature weight between 0.2 and 0.6, the decomposition level is set to 2; for the 1st layer detail coefficient and the approximation coefficient with a feature weight less than 0.2, the decomposition level is set to 1.

[0105] The mapping relationship between the encoded feature components is calculated using a nonlinear activation function. For each subband coefficient after wavelet packet transform, the system applies the aforementioned nonlinear activation function to obtain the activated feature expression. The shape parameter of the activation function is optimized through the backpropagation algorithm, with the goal of maximizing the feature discrimination between different fault types. The system calculates the mutual information between different subbands to construct a feature correlation graph, where nodes represent subbands and edges represent the correlation strength between subbands. Based on the feature correlation graph, the system analyzes the mapping relationship between feature components and identifies highly correlated feature groups and complementary feature groups. For the A-phase voltage amplitude signal, the system finds that the 3rd and 4th layer detail coefficients have strong complementarity, while the 5th layer detail coefficient has high correlation with the coarsest scale approximation coefficient.

[0106] According to the mapping relationship, the feature components are quantitatively reconstructed to obtain multi-scale reconstructed features, which involves two stages: inverse wavelet packet transform and inverse wavelet transform. The system first applies inverse wavelet packet transform to each subband coefficient to obtain the reconstructed feature components of each scale level. Then, the reconstructed feature components are weighted and combined according to the feature weights to obtain the complete multi-scale reconstructed features. During the reconstruction process, the system considers the mapping relationship between feature components, adopts a common reconstruction strategy for highly correlated feature groups, and adopts a separate reconstruction and fusion strategy for complementary feature groups. For the A-phase voltage amplitude signal, the average relative error between the reconstructed signal and the original signal is 3.5%, and the structural similarity is 0.96, which retains the key feature information of the original signal.

[0107] The learnable shape parameters of the nonlinear activation function for the multiscale reconstruction features improve the feature discrimination, and the shape parameters of the nonlinear activation function are optimized for different types of fault signals. For single-phase ground fault, the slope parameter a is optimized to 1.5 and the offset parameter b is optimized to 0.2; for two-phase short-circuit fault, a is optimized to 1.8 and b is optimized to -0.1; for three-phase short-circuit fault, a is optimized to 2.0 and b is optimized to 0.0. The system applies the optimized nonlinear activation function to the multiscale reconstruction features to enhance the discrimination of fault features. The transformed features form a more dispersed distribution in the feature space, which is beneficial to subsequent fault recognition.

[0108] The adaptive threshold is determined based on the distribution of the mapped features, and the multiscale reconstruction features are divided into main features and auxiliary features. The statistical distribution of the transformed features is analyzed, and the mean, standard deviation, and kurtosis are calculated. The adaptive threshold is set to the mean plus 1.5 times the standard deviation. The part of the feature value greater than the threshold is determined as the main feature, and the part of the feature value less than or equal to the threshold is determined as the auxiliary feature. For the A-phase voltage amplitude signal, after threshold division, the main features account for 25% of the total features and are mainly distributed in the 4th and 5th layer detail coefficients, representing the high-frequency transient features during fault occurrence; the auxiliary features account for 75% of the total features and are mainly distributed in the 1st to 3rd layer detail coefficients and the approximation coefficient, representing the low-frequency characteristics and persistent impact of the fault.

[0109] The nonlinear activation function is used to map the main features and auxiliary features, and the learnable shape parameters are iteratively optimized to obtain the mapped high-dimensional features. For the main features, the system uses a steeper activation function, with the slope parameter a increased by 20% and the offset parameter b reduced by 50%, to enhance the nonlinearity of the feature expression. For the auxiliary features, the system uses a gentler activation function, with the slope parameter a reduced by 10% and the offset parameter b increased by 100%, to maintain the smooth transition of the features. The system designs an iterative optimization algorithm to continuously adjust the shape parameters by minimizing the intra-fault sample distance and maximizing the inter-fault sample distance objective function. The gradient descent method is used in the optimization process, with the learning rate initially set to 0.01 and attenuated to 0.9 times every 50 iterations. After 200 iterations of optimization, the shape parameters of the main features and auxiliary features converge to the optimal value, obtaining stable high-dimensional feature expression. The high-dimensional feature dimension is 3 times that of the original feature, containing more rich fault feature information.

[0110] The mapped high-dimensional features are mapped to different feature expression spaces using a nonlinear activation function. The system designs three parallel feature expression spaces: time-domain feature space, frequency-domain feature space, and time-frequency joint feature space. For the time-domain feature space, the system maintains the time sequence structure of the high-dimensional features and applies one-dimensional convolution operation to extract time-domain patterns. For the frequency-domain feature space, the system applies Fourier transform to the high-dimensional features to extract frequency spectrum features. For the time-frequency joint feature space, the system applies wavelet packet transform to extract time-frequency joint features. The three feature spaces extract different aspects of fault features, forming complementary feature expressions.

[0111] The outputs of different feature expression spaces are adaptively fused to generate the final fault feature fingerprint information. The adaptive fusion is based on the attention mechanism. The system calculates the importance weights of the outputs of the three feature spaces. The weight calculation considers three indicators: information entropy, sparsity, and discriminability. For the feature space with high information entropy, moderate sparsity, and high discriminability, a higher fusion weight is assigned. The system weights and sums the outputs of the three feature spaces according to the fusion weights to obtain the preliminary fusion features. A nonlinear activation function is applied to the preliminary fusion features for final transformation to generate the fault feature fingerprint information. The fault feature fingerprint information is represented as a 256-dimensional vector, containing key features and identification marks of the fault.

[0112] In practical applications, a power monitoring system of a certain 110kV substation collects phase and amplitude data of a single-phase ground fault. After multi-scale wavelet transform, 5 layers of feature components are obtained. Through time-frequency energy analysis, the feature weights of each layer are calculated as 0.12, 0.18, 0.31, 0.67, and 0.89, and the approximation coefficient weight is 0.15. The system performs dynamic quantization coding based on the feature weights, with quantization steps ranging from 0.056 to 0.417. After nonlinear activation function mapping and quantization reconstruction, the structural similarity between the obtained multi-scale reconstruction features and the original signal reaches 0.97. The system divides the features into main features and auxiliary features using an adaptive threshold of 0.42, with the main features accounting for 22%. After mapping and iterative optimization, a 768-dimensional high-dimensional feature expression is obtained. The system maps the high-dimensional features to three feature expression spaces, with fusion weights of 0.45, 0.30, and 0.25. The final generated fault feature fingerprint information successfully identifies the fault type as A-phase single-phase ground fault, with a positioning accuracy of 98.5%, providing a reliable basis for fault diagnosis and handling.

[0113] In an alternative embodiment, the feature components of each scale level are dynamically quantized and coded based on the feature weights, the mapping relationship between the coded feature components is calculated using the nonlinear activation function, and the feature components are quantitatively reconstructed according to the mapping relationship to obtain multi-scale reconstruction features, including:

[0114] According to the feature weight, a dynamic quantization encoder is constructed, a quantization step is calculated based on the feature weight, and a wavelet decomposition is performed on the feature component based on the quantization step to generate a quantized feature value;

[0115] Based on the accuracy of the quantized feature value, the encoding parameters of the dynamic quantization encoder are determined, and a wavelet packet transform is performed on the quantized feature value of different scale levels by the dynamic quantization encoder to generate an encoded feature group, and the quantization accuracy of each encoded feature in the encoded feature group is determined by the encoding depth and the encoding base;

[0116] According to the encoding depth and the encoding base, the encoded feature group and the quantized feature value are wavelet reconstructed to obtain a reconstructed feature, and the encoding parameters of the dynamic quantization encoder are adjusted based on the reconstructed feature;

[0117] The reconstructed features of multiple scale levels are encoded by using the updated encoding parameters to generate a feature encoding sequence, a fusion weight is generated based on the feature encoding sequence, and a final multi-scale fusion feature is obtained by joint reconstruction of the fusion weight and the reconstructed feature.

[0118] The original feature data is obtained, which can be an image, audio or a signal sequence collected by a sensor. Taking image processing as an example, the system collects a high-resolution image of 1024x1024 pixels, each pixel contains three channels of RGB, and the pixel value ranges from 0 to 255. The original feature data is processed through a preprocessing link, including normalization processing to map the pixel value to the range of 0-1, and denoising processing to reduce random noise interference. The preprocessed feature data is decomposed into feature components of multiple scale levels. The system uses discrete wavelet transform for multi-scale decomposition, selects db4 wavelet base function, and performs 5-layer decomposition to obtain feature components at different scale levels. After each layer of decomposition, a low-frequency approximation component and a high-frequency detail component are obtained, and 5-layer decomposition generates 5 high-frequency detail components and 1 low-frequency approximation component, which constitute a feature component set.

[0119] The feature weight of each scale level feature component is calculated, and the feature weight calculation is based on information entropy and local contrast. For the i-th layer feature component, the system divides it into 16x16 local blocks, calculates the information entropy value of each block, and the higher the information entropy value, the more information the block contains. At the same time, the standard deviation of the pixel value in the block is calculated, and the higher the standard deviation, the higher the local contrast. The system sums the information entropy value and the standard deviation to obtain the block importance index, and the weight coefficients are 0.6 and 0.4 respectively. The importance indexes of all blocks are sorted, and the top 30% blocks are taken as the key blocks, and the average importance index of the key blocks is taken as the feature weight of the scale level feature component. The feature weight values of the 5-layer decomposition are 0.85, 0.72, 0.63, 0.51 and 0.42, and the weight of the low-frequency approximation component is 0.9. The feature weight value reflects the importance of the feature component of different scale levels in the final reconstruction.

[0120] A dynamic quantization encoder is constructed according to the feature weight, and the dynamic quantization encoder is composed of multiple quantization units, and each quantization unit is responsible for processing a scale level feature component. The quantization unit internally includes a quantization step calculation module, a nonlinear activation function module and an encoding parameter adjustment module. The system calculates the quantization step based on the feature weight, and the quantization step is inversely proportional to the feature weight. The higher the feature weight, the smaller the quantization step, and the higher the quantization precision. The specific calculation method is: the quantization step is equal to the basic step divided by the square of the feature weight, and the basic step is set to 0.05. For the first layer feature component with a feature weight of 0.85, the calculated quantization step is 0.069; for the fifth layer feature component with a feature weight of 0.42, the quantization step is 0.283.

[0121] Based on the quantization step, the wavelet decomposition of the feature component is generated to generate the quantized feature value. For each scale level feature component, the system uses the corresponding quantization step for uniform quantization. In the quantization process, the continuous value of the feature component is mapped to the discrete quantization level. The number of quantization levels is determined by the quantization step, and the smaller the quantization step, the more the quantization levels. For the first layer feature component, the quantization step is 0.069, and the number of quantization levels is 15; for the fifth layer feature component, the quantization step is 0.283, and the number of quantization levels is 4. The quantized feature value is called the quantized feature value, which retains the main information of the original feature component and realizes data compression at the same time.

[0122] The encoding parameters of the dynamic quantization encoder are determined based on the precision of the quantized feature values, including two key parameters: encoding depth and encoding basis. The encoding depth represents the number of wavelet packet transform layers, and the encoding basis represents the type of wavelet basis function. The system dynamically adjusts these two parameters according to the precision of the quantized feature values. The higher the precision of the quantized feature values, the deeper the encoding depth and the more complex the encoding basis. The precision calculation method is the logarithmic value of the number of quantization levels, and the precision value ranges from 0.6 to 1.2. For quantized feature values with a precision value greater than 1.0, the encoding depth is set to 3 and the encoding basis is selected as db6; for quantized feature values with a precision value between 0.8 and 1.0, the encoding depth is set to 2 and the encoding basis is selected as db4; for quantized feature values with a precision value less than 0.8, the encoding depth is set to 1 and the encoding basis is selected as haar.

[0123] The dynamic quantization encoder performs wavelet packet transform on the quantized feature values of different scale levels to generate an encoding feature group. Wavelet packet transform is an extension of wavelet transform, which not only decomposes the low-frequency part but also decomposes the high-frequency part, providing more detailed frequency division. The system performs wavelet packet transform on the quantized feature values of each scale level according to the previously determined encoding depth and encoding basis. For the first layer of quantized feature values with an encoding depth of 3, wavelet packet transform produces 8 subbands; for the second and third layers of quantized feature values with an encoding depth of 2, it produces 4 subbands; for the fourth and fifth layers of quantized feature values with an encoding depth of 1, it produces 2 subbands. These subbands form the encoding feature group, and the quantization precision of each encoding feature in the encoding feature group is determined by the encoding depth and the encoding basis.

[0124] According to the encoding depth and the encoding basis, the wavelet reconstruction is performed on the encoding feature group and the quantized feature values to obtain the reconstructed features. Wavelet reconstruction is the inverse process of wavelet decomposition, which recombines the components of different frequency bands into a complete signal. In the reconstruction process, the system first applies inverse wavelet packet transform to each subband in the encoding feature group to obtain the preliminary reconstructed quantized feature values. Then, the inverse quantization operation is applied to the reconstructed quantized feature values to map the discrete quantization levels back to the continuous value domain. The inverse quantization operation considers the influence of the quantization step size and the nonlinear activation function, making the reconstructed values closer to the original feature components. The nonlinear activation function uses the rectified linear unit function, which truncates negative values to zero and preserves positive values unchanged, enhancing the sparsity of the reconstructed features. For the first layer of feature components, the reconstruction error is controlled within 3%; for the fifth layer of feature components, the reconstruction error is about 12%.

[0125] The encoding parameters of the dynamic quantization encoder are adjusted based on the reconstructed features. The error between the original feature components and the reconstructed features is calculated. The larger the error, the less optimized the current encoding parameters. The system uses the gradient descent method to update the encoding parameters, with the goal of minimizing the reconstruction error. During the update process, the encoding depth can be increased or decreased by 1 layer, and the encoding basis can be switched between the five wavelet bases: haar, db2, db4, db6, and sym4. The system evaluates different parameter combinations and selects the one with the smallest reconstruction error as the updated encoding parameters. In the actual optimization process, the encoding depth of the first layer feature component increases from 3 to 4, and the encoding basis changes from db6 to sym4. The encoding depth of the fifth layer feature component remains 1, and the encoding basis changes from haar to db2.

[0126] The reconstructed features of multiple scale levels are encoded using the updated encoding parameters to generate a feature encoding sequence. During the encoding process, the system applies the optimized wavelet packet transform to each scale level's reconstructed feature to obtain a more refined sub-band division. For each sub-band, the system applies entropy encoding techniques for further compression. The entropy encoding uses the Huffman encoding method, which assigns variable-length codes based on the frequency of values within the sub-band. The encoded data forms the feature encoding sequence, which contains both the encoded data and the encoding parameters. The encoding parameters record the encoding depth, encoding basis, and quantization step size used for each scale level, which are used for subsequent decoding and reconstruction.

[0127] Based on the feature encoding sequence, fusion weights are generated, which determine the contribution proportion of reconstructed features at different scale levels in the final multi-scale fusion. The system analyzes the statistical properties of the feature encoding sequence, including information entropy, encoding length, and encoding efficiency. The scale levels with higher information entropy, shorter encoding length, and higher encoding efficiency are assigned higher fusion weights. The calculated fusion weights for the first to fifth layer feature components are 0.32, 0.25, 0.18, 0.15, and 0.10, respectively. The system further considers the influence of the original feature weights and takes a weighted average of the fusion weights and feature weights to obtain the final fusion weights of 0.45, 0.28, 0.20, 0.12, and 0.05.

[0128] The final multi-scale fusion features are obtained by jointly reconstructing the fusion weights and the reconstructed features. The joint reconstruction process uses a weighted summation method to combine the reconstructed features of different scale levels according to the fusion weights. To maintain the structural integrity of the features, the system performs spatial alignment on the reconstructed features of each scale level before weighted summation, ensuring that the features of different scale levels are superimposed in the same coordinate space. The system also applies an edge-preserving filter to enhance the feature boundaries. The filter is designed using a bilateral filter with a spatial domain radius of 3 and a value domain radius of 0.1. The final multi-scale fusion features retain the main structures and detail information of the original features while achieving data compression and feature enhancement.

[0129] In one specific image processing case, a medical image of 1024x1024 pixels is processed. The original image contains important detailed structures and background information, which needs to be accurately preserved while compressing the data volume. The system calculates the feature weight of five scale levels as 0.88, 0.75, 0.62, 0.48 and 0.39 respectively. The quantization step calculated based on these weight values is 0.065, 0.089, 0.130, 0.217 and 0.328 respectively. The encoded feature set generated after wavelet packet transform contains 21 subbands, and the total data volume is 35% of the original image. After parameter optimization and encoding reconstruction, the data volume of the final multi-scale fused feature is only 28% of the original image, while maintaining 97% of the structural similarity and 45dB of the peak signal-to-noise ratio, meeting the dual requirements of high-quality compression and feature preservation.

[0130] In an alternative embodiment, the fault type recognition and location positioning according to the fault feature fingerprint information obtain a fault type determination result and a fault location coordinate, comprising:

[0131] According to the fault feature fingerprint information, a classification model is constructed, the adaptive wavelet decomposition scale is set for the fault feature fingerprint information, the feature energy value is calculated at each wavelet decomposition scale, and the feature energy distribution is obtained;

[0132] The feature energy distribution is input into the classification model, the probability value of each fault type is calculated, the type with the largest probability value is determined as the fault type determination result, the corresponding wavelet basis function is selected according to the fault type determination result, the time-frequency analysis of the fault feature fingerprint information is carried out using the wavelet basis function, and the energy distribution sequence on the time-frequency plane is obtained;

[0133] According to the energy distribution sequence, a space mapping curve is established, an energy extreme point is searched on the space mapping curve, and the space position corresponding to the energy extreme point is determined as the fault location coordinate.

[0134] The fault feature fingerprint information in the power system is obtained by a fault detection device, including three-phase voltage and current waveform data, the sampling frequency is set to 12.8 kHz, 256 points are sampled per cycle, and each phase data is converted into a digital signal by a 16-bit A / D converter and then stored. The fault feature fingerprint information also includes basic data such as load current, voltage level and system impedance parameters before the fault occurs. The waveform data collection time is from 4 cycles before the fault occurs to 16 cycles after the fault occurs, a total of 20 cycles of data, forming a 5120-point sampling sequence. The original data collected is preprocessed by a digital filter to remove power frequency interference and high-frequency noise, and the filter is designed as a band-pass filter with a passband range of 10 Hz to 5 kHz.

[0135] A classification model is constructed for the preprocessed fault feature fingerprint information. The classification model adopts a support vector machine structure, and considering the common fault types of power systems, the faults are classified into 10 categories: single-phase ground fault, two-phase short-circuit fault, two-phase ground fault, three-phase short-circuit fault, three-phase ground fault, broken line fault, high resistance fault, lightning fault, intermittent fault and equipment fault. The training of the classification model is based on the historical fault database, which contains 5000 groups of labeled fault cases. The training process adopts a cross-validation method, and the data set is divided into a training set and a validation set in a ratio of 8:2, and the training accuracy reaches 97.5%. The kernel function of the classification model is selected as a radial basis function, the penalty factor C is set to 100, and the kernel parameter γ is set to 0.01.

[0136] An adaptive wavelet decomposition scale is set for the fault feature fingerprint information, and the number of wavelet decomposition layers J is automatically adjusted according to the frequency characteristics of the fault waveform. The calculation method is an adaptive algorithm based on Shannon entropy. In specific implementation, the system first performs wavelet decomposition on the fault waveform for 1 to 10 layers, calculates the Shannon entropy value of each layer after decomposition, and determines the optimal decomposition layer number when the entropy value change rate is less than the preset threshold 0.05. For faults with obvious high-frequency characteristics such as lightning faults, the decomposition layer number is usually 7 to 8 layers; for faults with low-frequency characteristics as the main, such as broken line faults, the decomposition layer number is usually 4 to 5 layers. The system uses db4 wavelet as the initial decomposition wavelet basis to obtain the wavelet coefficients of each layer.

[0137] The feature energy value is calculated at each wavelet decomposition scale to obtain the feature energy distribution. For the jth layer of wavelet decomposition coefficients, the energy value Ej is calculated, and the energy value calculation method is the sum of the squares of all wavelet coefficients in this layer. For a 5120-point sampling sequence, the number of coefficients obtained after decomposition is 2560, 1280, 640, 320, 160, 80, 40 and 20 points in turn. Taking 8-layer decomposition as an example, the energy distribution vector [E1, E2, E3, E4, E5, E6, E7, E8] is formed. The system normalizes the energy distribution vector so that the sum of the energy values of each layer is 1, forming the feature energy distribution. For a single-phase ground fault of phase A, the normalized feature energy distribution is [0.05, 0.08, 0.12, 0.25, 0.30, 0.15, 0.03, 0.02], indicating that the energy is mainly concentrated in the 4th and 5th layers.

[0138] The characteristic energy distribution is input into the classification model, the probability values of each fault type are calculated, the output of the classification model is converted into the probability values of each fault type using a softmax function, a probability distribution vector P = [P1, P2, …, P10] is formed, where Pi represents the probability of belonging to the ith fault type. The system identifies the type with the highest probability as the fault type determination result. For the above case, the calculated probability distribution is [0.92, 0.03, 0.02, 0.01, 0.01, 0.01, 0.00, 0.00, 0.00, 0.00], and the single-phase ground fault is determined with a confidence of 92%.

[0139] According to the fault type determination result, the corresponding wavelet basis function is selected, and the mapping relationship between the fault type and the optimal wavelet basis is established: the single-phase ground fault corresponds to db4 wavelet, the two-phase short-circuit fault corresponds to db6 wavelet, the two-phase ground fault corresponds to db8 wavelet, the three-phase short-circuit fault corresponds to sym4 wavelet, the three-phase ground fault corresponds to sym6 wavelet, the broken line fault corresponds to coif3 wavelet, the high-impedance fault corresponds to bior3.5 wavelet, the lightning strike fault corresponds to rbio2.8 wavelet, the intermittent fault corresponds to dmey wavelet, and the equipment fault corresponds to haar wavelet. The selection of the optimal wavelet basis is based on a large number of fault case analyses to ensure the best representation of the time-frequency characteristics of a specific type of fault. For the case determined as a single-phase ground fault, the system selects db4 wavelet as the wavelet basis function for subsequent analysis.

[0140] The selected wavelet basis function is used for time-frequency analysis of the fault characteristic fingerprint information to obtain the energy distribution sequence on the time-frequency plane. The continuous wavelet transform method is used to transform the fault waveform at different scales, and the scale parameter a has a value range of 1 to 64, with 32 discrete scale points set. The value range of the translation parameter b covers the entire time window, and the step size is set to 2 times the sampling interval. For each time-frequency point (a, b), the wavelet coefficient W(a, b) is calculated, and the square of the coefficient represents the energy value E(a, b) of the point. The system constructs a time-frequency energy distribution matrix with a size of 32 × 2560, where 32 represents the number of scales and 2560 represents the number of time points. For each column in the matrix, the maximum energy value and its corresponding scale are extracted to form an energy distribution sequence {(t1, E1, a1), (t2, E2, a2), …, (t2560, E2560, a2560)}.

[0141] A spatial mapping curve is established according to the energy distribution sequence, mapping the time t to the spatial distance d, and the mapping relationship is based on the propagation speed v of electromagnetic waves in the line. For overhead lines, v takes a value of 98% of the speed of light, about 2.94 × 10 8 m / s; for cable lines, v takes a value of 60% of the speed of light, about 1.8 × 10 8m / s. The conversion formula between time t and distance d is d = v x t. For the two-end measurement data, the system adopts the time difference positioning principle, considering the reflection and refraction effects of the wave. The system establishes a three-dimensional space mapping curve with coordinates (d, E, a), where d is the distance coordinate, E is the energy value, and a is the scale parameter. For the single-phase grounding fault case, the space mapping curve shows a clear energy concentration phenomenon in the distance coordinate interval of 35 km to 40 km.

[0142] On the space mapping curve, the energy extreme point is searched, and a search algorithm based on particle swarm optimization is designed. 50 search particles are initialized and distributed in the fault area. The fitness function of each particle is defined as the energy value E at that position. The particle iteratively updates its position, with a maximum of 100 iterations and a convergence condition of a change of less than 0.1% in the optimal value for 10 consecutive iterations. During the search process, the system considers both the energy value E and the scale parameter a, and preferentially selects areas with high energy and moderate scale parameters. For complex line structures, the system introduces topological constraints to ensure that the search range conforms to the actual line layout. Through iterative search, the system finds the energy extreme point (d*, E*, a*), where d* represents the fault distance.

[0143] The space position corresponding to the energy extreme point is determined as the fault location coordinate, and the distance d* is converted into the actual physical coordinate (x, y, z), based on the geographic information data of the line. The system queries the line tower database to determine the section and adjacent tower number where the fault point is located. The system also calculates the confidence interval of fault location, which is to extract all points with energy values not less than 80% of the extreme value near the energy extreme point, and to calculate the distance distribution of these points to obtain the 95% confidence interval of the fault location. For the single-phase grounding fault case, the system finally determines the fault location coordinate as 37.8 km away from the substation, corresponding to the section between tower No. 72 and tower No. 73, with a 95% confidence interval of [37.5 km, 38.1 km].

[0144] To verify the effectiveness of the method, an application test was conducted on a 110 kV transmission line, which is 85 kilometers long, has a total of 156 towers, and uses LGJ-240 / 30 conductors. The fault point was simulated to be located 37.6 kilometers away from the substation, and the fault type was an A-phase metallic ground fault. After the fault detection device collected the fault waveform, the system processed the fault characteristic fingerprint information. Adaptive wavelet decomposition determined that the optimal decomposition level was 6 layers, and the characteristic energy distribution was [0.04, 0.07, 0.13, 0.26, 0.32, 0.18]. The classification model determined that the fault type was a single-phase ground fault, with a confidence level of 94.6%. The system selected db4 wavelet for time-frequency analysis and obtained the time-frequency energy distribution matrix. After establishing the space mapping curve, the system searched for the energy extreme point corresponding to the distance of 37.8 kilometers. The error between the final fault location result and the actual fault point was 0.2 kilometers, with a relative error of 0.24%, meeting the requirements of engineering applications.

[0145] The method also has adaptive adjustment capability, which can dynamically adjust the wavelet decomposition parameters and search strategy according to the line parameters and operating state. For different types of lines and faults, the system can automatically select the most suitable processing parameters, improving the accuracy and adaptability of fault location. Actual application shows that the average positioning error of this method for various faults is less than 0.5% of the line length, greatly improving the fault handling efficiency and power system reliability.

[0146] In an alternative embodiment, according to the fault type determination result and the fault location coordinates, the basic parameter information is combined to correct the fault location, and the final fault location result is obtained. Based on the final fault location result, a fault location report is generated, including:

[0147] According to the fault type determination result and the fault location coordinates, the basic parameter information is combined to construct a chaotic attractor state space. The fault location coordinates are mapped to the chaotic attractor state space to obtain a chaotic orbit. The state evolution sequence of the chaotic orbit is calculated based on the evolution equation of the chaotic attractor state space. The characteristic change point is determined according to the time derivative of the state evolution sequence. The state mapping matrix is constructed based on the characteristic change point.

[0148] The state mapping matrix is used to iteratively correct the fault location coordinates. The step size of the iterative correction is related to the singular value of the state mapping matrix. When the singular value of the state mapping matrix changes by less than a preset change threshold, the iteration is stopped, and the current fault location coordinates are determined as the final fault location result. Based on the stability analysis of the chaotic orbit and the characteristic change point, a fault location report is generated.

[0149] As shown in Figure 3 the method comprises:

[0150] The basic parameter information of the power line is acquired, including line length, tower position coordinates, conductor type, line impedance parameters, line load conditions, and meteorological conditions. The parameter collection method is to extract static parameters from the line management database and obtain dynamic operation parameters through the SCADA system. The basic parameter information is stored in a structured data table, and each record contains a parameter ID, a parameter name, a parameter value, a unit, and a timestamp.

[0151] The fault detection device collects voltage and current waveforms during line faults, with a sampling frequency of 12.8 kHz and 256 samples per cycle. The waveform data is digitized by a 16-bit A / D converter and transmitted to the processing unit. The system pre-processes the collected waveform data, including denoising, baseline calibration, and data normalization. The denoising uses the wavelet threshold method, with db4 wavelet and 5 layers of decomposition, and the soft threshold value is 3.5 times the standard deviation.

[0152] The pre-processed waveform data is decomposed by db6 wavelet for 8 layers, and the energy features, phase features, and polarity features of each frequency band are extracted. The extracted feature vector contains 24 elements, corresponding to the energy percentage, phase difference, and polarity identification of each frequency band. The system inputs the feature vector into the fault recognition model, which uses the support vector machine algorithm, with radial basis function as the kernel function, penalty factor C set to 100, and kernel parameter γ set to 0.01. The model outputs the fault type judgment result, including fault type code and confidence.

[0153] The initial fault location coordinates are calculated based on the traveling wave method, and the arrival time of the traveling wave head in the fault waveform is detected. The wavelet transform modulus maximum method is used to determine the arrival time of the traveling wave, with a time resolution of 5 microseconds. For single-end data, the system uses the reflected wave method to calculate the fault distance, with a wave speed of 2.9×10^8 meters / second. For double-end data, the system uses the time difference method to calculate the fault distance. The initial fault location coordinates are represented as (d, 0, h), where d is the distance from the fault point to the monitoring end, and h is the conductor height.

[0154] The chaos attractor state space is constructed for fault location correction, and the appropriate chaos system model is selected according to the fault type. For single-phase ground fault, the Lorenz chaos system is selected; for phase-to-phase short circuit fault, the Rössler chaos system is selected; for broken line fault, the Chen chaos system is selected. Taking the Lorenz system as an example, the control parameters are set to σ=10, ρ=28, and β=8 / 3.

[0155] The chaos system parameters are adjusted in combination with the basic parameter information, and the line length affects the parameter σ, with the adjustment formula σ'=σ×(L / 100) 0.5; Line load level impact parameter p, adjustment formula p'=p x (P / Prated) 0.3 ; Conductor type impact parameter b, adjustment formula b'=b x (S / 240) 0.2 Where S is the conductor cross-sectional area. For 110 kV lines, the typical adjustment range of parameters is p'=832, b'=2.5~3.0.

[0156] Map the initial fault location coordinates to the chaotic attractor state space, the mapping relationship is x=d / L, y=0, z=h / hmax, where L is the total length of the line, and hmax is the maximum conductor height. For example, for a fault point distance of 45.7 km and a conductor height of 25 m, the mapped initial point is (0.457, 0, 0.833), assuming the total length of the line is 100 km and the maximum conductor height is 30 m.

[0157] Calculate the chaotic orbit based on the chaotic attractor evolution equation, and use the fourth-order Runge-Kutta method for evolution calculation, with a time step of 0.01 and an iteration number of 2000. Calculate the state point (xi, yi, zi) at each iteration to form a state sequence. The system stores the calculation results in a three-dimensional array, each element containing the time point ti and the corresponding state value. After the orbit calculation is completed, the system checks whether the orbit fills the characteristic region of the chaotic attractor and whether the number of orbit points meets the statistical analysis requirements (>1000 points).

[0158] Calculate the time derivative of the state evolution sequence and identify the characteristic change points. For each state point (xi, yi, zi), calculate its derivative value (dxi / dt, dyi / dt, dzi / dt). The derivative calculation uses the central difference method, and the calculation formula is dxi / dt=(xi+1-xi-1) / (2Δt). The system sets the derivative threshold vector (Tx, Ty, Tz), and when the absolute value of any dimension derivative exceeds the corresponding threshold, the point is marked as a characteristic change point. The threshold setting is related to the system parameters, and the typical value is Tx=15, Ty=20, Tz=25.

[0159] Perform clustering analysis on the identified characteristic change points and remove outliers. Use the density-based clustering algorithm DBSCAN with parameters set to ε=0.05, MinPts=4. After clustering, retain the characteristic points in the main cluster to form a simplified set of characteristic change points {(xj, yj, zj)|j=1,2,...,k}. The system calculates the distribution characteristics of the characteristic points in each dimension, including mean, standard deviation, and skewness. These statistics are used to assess the uncertainty of the fault location.

[0160] Based on the feature change point, a state mapping matrix M is constructed, and the feature change point coordinates are organized into a k x 3 matrix, with each row corresponding to the three-dimensional coordinates of a feature point. The matrix is centered by subtracting the mean of each column. Singular value decomposition is performed on the centered matrix to obtain M = UΣV T , where Σ is a diagonal matrix with singular values s1≥s2≥s3≥0 on the diagonal. The system extracts the singular values (s1, s2, s3) and the right singular vector matrix V for subsequent iterative correction.

[0161] Based on the state mapping matrix, the fault location is iteratively corrected, and the iteration step size is set to α = 0.1 / s1 to ensure stable convergence in the correction process. The coordinate update formula for each iteration is (xnew, ynew, znew) = (xold, yold, zold) + α·V·(s1, 0, 0) T , where V is the right singular vector matrix representing the main change direction. After each iteration, the system recalculates the chaotic orbit and the state mapping matrix to obtain new singular values (s1', s2', s3').

[0162] The singular value change amount Δs = |s1'-s1|+|s2'-s2|+|s3'-s3| is calculated and compared with the preset change threshold δ = 0.01. When the singular value change amount of three consecutive iterations is less than the threshold δ, the system stops the iteration process. At this time, the fault location coordinates are determined as the final fault positioning result. The system reflects the state space coordinates back to the physical space to obtain the actual fault point distance d' = x·L and height h' = z·hmax.

[0163] The chaotic orbit is analyzed for stability to evaluate the reliability of the positioning result. The maximum Lyapunov exponent λ is calculated, and the small perturbation method is used with an initial perturbation size of 10 -10 and an evolution time of 10 time units. The average value is calculated from 20 initial points. When λ < 0.5, the chaotic orbit has good stability and the positioning result has high reliability; when 0.5 ≤ λ < 1.0, the positioning result has medium reliability; and when λ ≥ 1.0, the positioning result has low reliability and needs to be verified by other methods.

[0164] The spatial distribution of feature change points is analyzed to determine the fault point range. The standard deviation σx of feature points in the x dimension is calculated, and the 95% confidence interval of the fault point is [d'-1.96·σx·L, d'+1.96·σx·L]. The system optimizes the fault range representation by considering the actual situation of the line, such as tower location, terrain features, etc., and generates a patrol recommendation section.

[0165] Generate a fault location report, including the following: basic fault information (time, duration, affected equipment); fault type and confidence level; initial fault location coordinates and positioning method; chaos system parameter setting and adjustment basis; number of characteristic change points and distribution characteristics; iteration correction process record, including coordinate change and singular value change of each iteration; final fault location result, including distance value, range and confidence level; physical location description corresponding to the fault point, such as tower number, GPS coordinates, etc.; chaos orbit stability analysis results; recommended inspection range and route; preliminary analysis of fault causes and treatment suggestions.

[0166] Send the fault location report to the operation and maintenance center and the mobile terminal on site through the data communication network. The report format supports both PDF and HTML formats, including text description, data table and visual chart. The mobile terminal application can display the location of the fault point on the electronic map and provide navigation function to guide the maintenance personnel to the fault location. The system records the difference between the fault location result and the actual fault point for subsequent optimization of chaos system parameters and improvement of positioning accuracy.

[0167] In a practical case, a single-phase grounding fault occurred on a 110kV transmission line during thunderstorm weather conditions. The system obtained the line parameters: total length 87.5km, conductor type LGJ-240 / 30, a total of 154 towers. The fault detection device collected the fault waveform and determined it as a single-phase grounding fault of phase A with a confidence level of 98.7%. The preliminary fault location result was 45.7km from the substation. The system constructed a Lorenz chaotic attractor with parameters set to σ=9.35, ρ=29.4, β=2.78. The characteristic change point identified 27 effective points, mainly distributed in the interval x=0.45~0.47. After 9 iterations of correction, the final fault location was determined to be 46.2km from the substation, corresponding to the interval between 84# and 85# towers. The 95% confidence interval of the fault point was [45.8km, 46.6km]. The maximum Lyapunov exponent λ=0.38, indicating high reliability of the positioning result. The maintenance personnel checked according to the recommended inspection range and found a tree branch touching the conductor at 46.3km, with an error of only 0.1km from the system positioning result.

[0168] In a second aspect of the embodiments of the present application, a power distribution network optical fiber fault location system based on time domain reflection is provided, comprising:

[0169] A first unit is configured to obtain basic parameter information of the power distribution network optical fiber, generate an initial optical carrier signal based on the basic parameter information, and perform phase modulation on the initial optical carrier signal to obtain a modulated detection signal.

[0170] The second unit is configured to divide the modulated probe signal into a plurality of chip sequences, encode each chip sequence using an orthogonal code sequence, and interleave and modulate the encoded chip sequences in the time domain to generate an enhanced probe signal with space-time diversity characteristics.

[0171] The third unit is configured to inject the enhanced probe signal into a power distribution network optical fiber, receive a reflected signal returned by the power distribution network optical fiber, and demodulate the reflected signal to extract phase information and amplitude information from the reflected signal.

[0172] The fourth unit is configured to perform fault feature decomposition on the phase information and the amplitude information based on a multi-scale wavelet transform, set an adaptive threshold at each scale level to perform feature screening, map the screened features to a high-dimensional feature space through a nonlinear activation function for modeling, and extract fault feature fingerprint information.

[0173] The fifth unit is configured to perform fault type identification and location positioning based on the fault feature fingerprint information, obtain a fault type determination result and a fault location coordinate, correct the fault location based on the fault type determination result and the fault location coordinate in combination with the basic parameter information, obtain a final fault positioning result, and generate a fault positioning report based on the final fault positioning result.

[0174] In a third aspect, an electronic device is provided, including:

[0175] a processor;

[0176] a memory for storing processor-executable instructions;

[0177] The processor is configured to invoke the instructions stored in the memory to execute the method described above.

[0178] In a fourth aspect, a computer-readable storage medium is provided, which stores computer program instructions, and the computer program instructions are executed by a processor to implement the method described above.

[0179] The present application can be a method, device, system and / or computer program product. The computer program product can include a computer readable storage medium having stored thereon computer readable program instructions that, when executed by a computer, cause the computer to carry out various aspects of the present application.

[0180] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, and are not intended to limit the present application; although the present application has been described in detail with reference to the above embodiments, those skilled in the art should understand that the technical solutions recorded in the above embodiments can be modified, or some or all of the technical features can be replaced by equivalents; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the present application.

Claims

1. A time domain reflectometry based method for locating fiber faults in a power distribution network, characterized in that, The method comprises the following steps: obtaining basic parameter information of the power distribution network optical fiber, generating an initial optical carrier signal according to the basic parameter information, and phase modulating the initial optical carrier signal to obtain a modulated probe signal; dividing the modulated probe signal into multiple chip sequences, encoding each chip sequence using an orthogonal code sequence, and interlacing the encoded multiple chip sequences in the time domain to generate an enhanced probe signal with space-time diversity characteristics, comprising: multiplying the modulated probe signal by a transform kernel function and performing fractional order transform processing to obtain a transform processing result; extracting signal characteristic components of the transform processing result in the transform domain, determining an energy distribution threshold based on the signal characteristic components, determining a signal division position according to the energy distribution threshold, and dividing the modulated probe signal into multiple chip sequences based on the signal division position; extracting feature frequency points from the transform processing result, performing feature matching between the feature frequency points and the transform kernel function, and generating an orthogonal code sequence group based on the feature matching result; encoding each orthogonal code sequence in the orthogonal code sequence group with the corresponding chip sequence to obtain an encoded chip sequence group, and determining an interlacing parameter based on the feature enhancement effect of the transform processing result; rearranging the encoded chip sequence group in time sequence according to the interlacing parameter, interlacing the rearranged chip sequence to generate an enhanced probe signal with space-time diversity characteristics; injecting the enhanced probe signal into the power distribution network optical fiber, receiving a reflected signal returned by the power distribution network optical fiber, and demodulating the reflected signal to extract phase information and amplitude information from the reflected signal; performing fault feature decomposition on the phase information and the amplitude information based on multi-scale wavelet transform, setting an adaptive threshold at each scale level for feature screening, mapping the screened features to a high-dimensional feature space through a nonlinear activation function for modeling, and extracting fault feature fingerprint information, comprising: performing multi-scale wavelet transform on the phase information and the amplitude information to obtain feature components at multiple scale levels, calculating feature weights using a nonlinear activation function with a learnable shape parameter according to the time-frequency energy distribution of the feature components at each scale level; dynamically quantizing and encoding the feature components at each scale level based on the feature weights, calculating the mapping relationship between the encoded feature components using the nonlinear activation function, quantizing and reconstructing the feature components according to the mapping relationship, and obtaining multi-scale reconstructed features; using the learnable shape parameter of the nonlinear activation function to improve the feature discrimination of the multi-scale reconstructed features, determining an adaptive threshold based on the distribution of the mapped features, and dividing the multi-scale reconstructed features into main features and auxiliary features; mapping the main features and the auxiliary features using the nonlinear activation function, respectively, and iteratively optimizing the learnable shape parameter to obtain the mapped high-dimensional features. The mapped high-dimensional features are mapped to different feature expression spaces by using the nonlinear activation function, the outputs of the different feature expression spaces are adaptively fused, and final fault feature fingerprint information is generated; According to the fault feature fingerprint information, fault type recognition and location positioning are performed to obtain fault type determination results and fault position coordinates, the fault position is corrected according to the fault type determination results and the fault position coordinates in combination with the basic parameter information, and finally, fault positioning results are obtained, and a fault positioning report is generated based on the final fault positioning results.

2. The method of claim 1, wherein, According to the basic parameter information, an initial optical carrier signal is generated, and the initial optical carrier signal is phase-modulated to obtain a modulated probe signal, including: On the time scale corresponding to the basic parameter information, the optical fiber transmission characteristics are analyzed, and the main singular value component reflecting the optical fiber dispersion characteristics is extracted; the main singular value component is used to perform nonlinear compensation on the optical carrier signal to generate an initial optical carrier signal; and the main singular value component is used to calculate the phase modulation depth, and the initial optical carrier signal is phase-modulated to obtain a modulated probe signal.

3. The method of claim 1, wherein, Based on the feature weight, the feature components of each scale level are dynamically quantized and encoded, the mapping relationship between the encoded feature components is calculated by using the nonlinear activation function, the feature components are quantitatively reconstructed according to the mapping relationship, and multi-scale reconstructed features are obtained, including: A dynamic quantization encoder is constructed according to the feature weight, a quantization step is calculated based on the feature weight, and the feature components are wavelet-decomposed based on the quantization step to generate quantized feature values; Based on the accuracy of the quantized feature values, the encoding parameters of the dynamic quantization encoder are determined, the dynamic quantization encoder is used to perform wavelet packet transform on the quantized feature values of different scale levels to generate an encoded feature group, and the quantization accuracy of each encoded feature in the encoded feature group is determined by the encoding depth and the encoding base; The encoded feature group and the quantized feature values are wavelet-reconstructed according to the encoding depth and the encoding base to obtain reconstructed features, and the encoding parameters of the dynamic quantization encoder are adjusted based on the reconstructed features; The reconstructed features of multiple scale levels are encoded by using the updated encoding parameters to generate a feature code sequence, a fusion weight is generated based on the feature code sequence, and the final multi-scale fusion features are obtained by jointly reconstructing the fusion weight and the reconstructed features.

4. The method of claim 1, wherein, According to the fault feature fingerprint information, fault type recognition and location positioning are performed to obtain fault type determination results and fault position coordinates, including: A classification model is constructed according to the fault feature fingerprint information, adaptive wavelet decomposition scales are set for the fault feature fingerprint information, feature energy values are calculated at each wavelet decomposition scale, and feature energy distribution is obtained. The feature energy distribution is input into the classification model, a probability value of each fault type is calculated, and a type with the maximum probability value is determined as a fault type determination result; a corresponding wavelet basis function is selected according to the fault type determination result, time-frequency analysis is performed on the fault feature fingerprint information by using the wavelet basis function, and an energy distribution sequence on a time-frequency plane is obtained; A spatial mapping curve is established according to the energy distribution sequence, an energy extreme point is searched on the spatial mapping curve, and a spatial position corresponding to the energy extreme point is determined as a fault position coordinate.

5. The method of claim 1, wherein, According to the fault type determination result and the fault position coordinate, the fault position is corrected in combination with the basic parameter information to obtain a final fault positioning result, and a fault positioning report is generated based on the final fault positioning result, including: According to the fault type determination result and the fault position coordinate, a chaotic attractor state space is constructed in combination with the basic parameter information; the fault position coordinate is mapped to the chaotic attractor state space to obtain a chaotic orbit, a state evolution sequence of the chaotic orbit is calculated based on an evolution equation of the chaotic attractor state space; a feature change point is determined according to a time derivative of the state evolution sequence, and a state mapping matrix is constructed based on the feature change point; The state mapping matrix is used to iteratively correct the fault position coordinate, a step of the iterative correction is related to a singular value of the state mapping matrix, and when a change amount of the singular value of the state mapping matrix is less than a preset change threshold, the iteration is stopped, the current fault position coordinate is determined as a final fault positioning result, and a fault positioning report is generated based on stability analysis of the chaotic orbit and the feature change point.

6. Time-domain reflection based power distribution network fiber fault location system for implementing the method of any of the preceding claims 1-5, characterized in that, Including: The first unit is configured to obtain basic parameter information of the power distribution network optical fiber, generate an initial optical carrier signal based on the basic parameter information, and perform phase modulation on the initial optical carrier signal to obtain a modulated probe signal; The second unit is configured to divide the modulated probe signal into a plurality of chip sequences, encode each chip sequence by using an orthogonal code sequence, and interleave and modulate the encoded plurality of chip sequences in the time domain to generate an enhanced probe signal with space-time diversity characteristics; The third unit is configured to inject the enhanced probe signal into the power distribution network optical fiber, receive a reflected signal returned by the power distribution network optical fiber, and perform demodulation processing on the reflected signal to extract phase information and amplitude information in the reflected signal; The fourth unit is configured to perform fault feature decomposition on the phase information and the amplitude information based on multi-scale wavelet transform, set an adaptive threshold at each scale level to perform feature screening, map the screened features to a high-dimensional feature space by using a nonlinear activation function for modeling, and extract fault feature fingerprint information. A fifth unit is configured to identify a fault type and locate a fault position according to the fault feature fingerprint information, to obtain a fault type determination result and a fault position coordinate, to correct the fault position according to the fault type determination result and the fault position coordinate in combination with the basic parameter information, to obtain a final fault positioning result, and to generate a fault positioning report based on the final fault positioning result.

7. An electronic device, comprising: Comprise: a processor; a memory for storing processor-executable instructions; wherein the processor is configured to invoke the instructions stored in the memory to execute the method of any one of claims 1 to 5.

8. A computer-readable storage medium having stored thereon computer program instructions, wherein, The computer program instructions, when executed by the processor, implement the method of any one of claims 1 to 5.

Images