Photovoltaic cable fault positioning method based on multi-source data fusion

Abstract

The invention relates to the technical field of photovoltaic cable fault positioning, in particular to a photovoltaic cable fault positioning method based on multi-source data fusion, which comprises the following steps of: deploying fault recording units at two ends of a combiner box and an inverter side of a photovoltaic string, acquiring a voltage waveform and a current waveform when a fault occurs, and performing band-pass filtering processing to obtain a fault recording unit; and a double-end synchronous transient traveling wave signal is obtained. In the photovoltaic cable fault positioning process, the purity and synchronism of input signals are guaranteed by collecting voltage waveforms and current traveling waveforms at the two ends and conducting filtering processing, wavelet decomposition is conducted on the basis of synchronous transient traveling wave signals, time points corresponding to modulus maximum points are extracted, energy values are calculated, and the fault positioning accuracy is improved. Double characterization of the time domain and the energy domain is achieved, and the time characteristic and the energy characteristic of traveling wave propagation can be synchronously analyzed.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of photovoltaic cable fault location, and particularly relates to a photovoltaic cable fault location method based on multi-source data fusion. BACKGROUND

[0002] The technical field of cable fault location mainly relates to fault detection, fault analysis and accurate positioning methods in the operation and maintenance of power systems.

[0003] Existing cable fault location relies on single time domain analysis or energy calculation based on fixed features, and is easily affected by noise interference or signal distortion during operation. For example, in the long-distance photovoltaic cable laying scene, single time domain calculation will cause deviation accumulation due to propagation speed error, resulting in a large deviation of the positioning result. The energy extraction method of the fixed time window lacks the adaptability to dynamic signal features, thereby causing energy distortion and reducing the reliability of the result. Therefore, improvement is needed. SUMMARY

[0004] The purpose of the present application is to solve the problems existing in the prior art, and a photovoltaic cable fault location method based on multi-source data fusion is proposed.

[0005] In order to achieve the above-mentioned purpose, the technical scheme adopted by the present application is as follows: the photovoltaic cable fault location method based on multi-source data fusion comprises the following steps: Deploying fault recording units at both ends of the busbar box and the inverter side of the photovoltaic string, collecting voltage waveforms and current waveforms when the fault occurs, and performing band-pass filtering to obtain double-end synchronous transient traveling wave signals; Based on the double-end synchronous transient traveling wave signals, wavelet decomposition is performed on the double-end synchronous transient traveling wave signals, the time point corresponding to the modulus maximum value point in the target high-frequency sub-band is extracted as the traveling wave arrival timestamp, and the traveling wave arrival timestamp and the sub-band energy value are calculated and obtained, the traveling wave arrival timestamps from both ends are combined, and a time-energy joint feature vector is established; Based on the time-energy joint feature vector, the total length of the photovoltaic cable and the traveling wave propagation speed parameters are called, the double-end traveling wave arrival timestamps in the time-energy joint feature vector are combined for time difference operation, time domain positioning coordinates are obtained, the cable attenuation coefficient parameters are called, and the double-end sub-band energy values in the time-energy joint feature vector are combined for operation, energy domain positioning coordinates are obtained, and a double-domain fault position estimation set is obtained; Based on the double-end synchronous transient traveling wave signals and the double-domain fault position estimation set, a sliding time window is set on the double-end synchronous transient traveling wave signals, a local symmetry breaking factor sequence is obtained, and based on the local symmetry breaking factor sequence, a weighted fusion photovoltaic cable fault point is calculated.

[0006] Preferably, the step of acquiring the dual-ended synchronous transient traveling wave signal is as follows: Fault recording units are deployed at both ends of the combiner box and the inverter side. A uniform sampling rate and trigger threshold are set. GPS synchronization is enabled to write a timestamp for each sampling point. The fault trigger interval is located and the voltage waveform value and current waveform value of the corresponding time period are extracted according to the timestamp to obtain the voltage waveform value and current waveform value of the dual-end synchronous acquisition. Based on the voltage waveform and current waveform values ​​acquired synchronously at both ends, the lower cutoff frequency parameter and the upper cutoff frequency parameter are set, bandpass filtering is performed, and forward and backward merging is performed. The first and last samples are removed according to the filter order parameter, and the fault trigger interval segment is retained to obtain the voltage waveform and current waveform values ​​after dual-end bandpass filtering. Based on the voltage waveform and current waveform values ​​after dual-end bandpass filtering, the sampling order at both ends is checked according to the absolute timestamp of GPS synchronization and resampled on a unified time axis. Only the abrupt change components in the fault-triggered interval segment are retained and the stable segment is discarded to obtain the dual-end synchronous transient traveling wave signal.

[0007] Preferably, the steps for obtaining the traveling wave arrival time stamp and sub-band energy value are as follows: Based on the dual-end synchronous transient traveling wave signal, multi-resolution wavelet decomposition is performed and modulus maxima are detected in the target high-frequency sub-band. The absolute time index corresponding to the first modulus maxima is extracted according to the sample order and defined as the initial traveling wave arrival time stamp, thus forming the left initial traveling wave arrival time stamp and the right initial traveling wave arrival time stamp. Based on the initial traveling wave arrival timestamps at the left and right ends, an adaptive time window is established on the target high-frequency sub-band with the traveling wave arrival timestamp as the starting point, and amplitude samples within the window are extracted. The sub-band energy value is calculated to obtain the traveling wave arrival timestamp and the sub-band energy value.

[0008] Preferably, the step of obtaining the time-energy joint feature vector is as follows: Based on the arrival timestamp of the traveling wave and the subband energy value, the initial arrival timestamp of the traveling wave at the left end, the initial arrival timestamp of the traveling wave at the right end, the subband energy value at the left end, and the subband energy value at the right end are obtained. They are then matched according to the port number and the time and energy parameters are combined in a unified structure to form a time-energy joint feature vector.

[0009] Preferably, the steps for obtaining the dual-domain fault location estimation set are as follows: Based on the time-energy joint feature vector, the arrival timestamps of the traveling wave at the left and right ends are extracted. The absolute value of the difference between the arrival timestamps of the traveling wave at the left and right ends is calculated and the sign of the difference is recorded. The total length of the photovoltaic cable and the traveling wave propagation speed parameters are called to convert the absolute value of the difference into length and determine the direction according to the sign of the difference to obtain the time-domain positioning coordinates. Based on the time-domain positioning coordinates, the energy values ​​of the left and right terminals in the time-energy joint feature vector are read, the logarithmic ratio of the energy values ​​of the left and right terminals is calculated, and the cable attenuation coefficient parameter is called to map the logarithmic ratio to the length offset. The length offset is corrected in the same direction as the time-domain positioning coordinates and the value is limited to the total length of the photovoltaic cable to obtain the energy-domain positioning coordinates. Based on the energy domain positioning coordinates, the time domain positioning coordinates and energy domain positioning coordinates are checked for boundaries within the total length of the photovoltaic cable, and out-of-bounds values ​​are eliminated. The source is recorded according to the port number to form a dual-domain fault location estimation set.

[0010] Preferably, the step of obtaining the local symmetry breaking factor sequence is as follows: Based on the dual-end synchronous transient traveling wave signal and the dual-domain fault location estimation set, an equally spaced starting point is set on the time axis of the dual-end synchronous transient traveling wave signal, and continuous segments are cut out according to a fixed sample length. Each segment is divided into a first half signal and a second half signal while keeping the index consistent, and a sliding time window is generated. According to the sliding time window, the amplitude of the first half of the signal in each sliding time window is recorded point by point and stored as a sequence. The amplitude of the second half of the signal in each sliding time window is recorded point by point after being reversed according to the time index and stored as a sequence. The first half signal sequence and the reversed second half signal sequence are paired according to a unified index to generate a pairing sequence of the first half signal and the reversed second half signal. Based on the paired sequence of the first half signal and the reversed second half signal, the difference ratio of the amplitudes of the two sets of sequences within each sliding time window is calculated and normalized. The normalization results of all sliding time windows are arranged in order and kept in correspondence with the time axis to form a local symmetry breaking factor sequence.

[0011] Preferably, the step of obtaining the fault point of the weighted fusion photovoltaic cable is as follows: Based on the local symmetry breaking factor sequence, the amplitudes of adjacent samples are compared point by point according to the sample index, local maxima are filtered out, and finally the local maximum with the largest amplitude is selected and the corresponding sample index is recorded as the peak position. At the same time, the corresponding amplitude is recorded as the peak value, thus obtaining the peak value and peak position. Based on the peak value and peak position, a core energy window and a supporting energy window centered on the peak position are established in the local symmetry breaking factor sequence, and the energy concentration coefficient is calculated.

[0012] Preferably, the step of obtaining the weighted fusion photovoltaic cable fault point further includes: based on the energy concentration coefficient, calling the time domain positioning coordinates and energy domain positioning coordinates in the dual-domain fault location estimation set, combining the peak value and the energy concentration coefficient in proportion, performing weighted fusion to obtain a single fault location and keeping the result within the total length of the photovoltaic cable, thereby generating the weighted fusion photovoltaic cable fault point.

[0013] Compared with the prior art, the advantages and positive effects of the present invention are as follows: In this invention, during the fault location process of photovoltaic cables, voltage and current waveforms are collected at both ends and filtered to ensure the purity and synchronization of the input signals. Wavelet decomposition is performed based on the synchronous transient traveling wave signal to extract the time points corresponding to the modulus maxima and calculate the energy values, achieving dual representation in both the time and energy domains. This allows the time and energy characteristics of the traveling wave propagation to be analyzed simultaneously. Furthermore, by combining the total cable length, traveling wave propagation speed, and attenuation coefficient parameters, the time difference and energy ratio are transformed into location parameters, forming a dual-domain fault location estimation set. Complementary verification based on different physical quantities avoids the drawbacks of accumulated errors in calculations from a single time or energy domain. On this basis, a local symmetry breaking factor sequence is constructed by setting a sliding time window, and weights are built based on the sequence peak value and sharpness. The location results from the two domains are then fused, improving the accuracy and robustness of fault location, shortening the location time, reducing the investigation scope, and maintaining stable and reliable location results even in complex environments. Attached Figure Description

[0014] Figure 1 This is a schematic diagram of the steps of the present invention. Detailed Implementation

[0015] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0016] Please see Figure 1 This invention provides a technical solution for photovoltaic cable fault location based on multi-source data fusion, comprising the following steps: Fault recording units are deployed at both ends of the photovoltaic string's combiner box and inverter side to collect the voltage waveform and current waveform when the fault occurs, and perform bandpass filtering to obtain a dual-end synchronous transient traveling wave signal. Based on the dual-end synchronous transient traveling wave signal, wavelet decomposition is performed on the dual-end synchronous transient traveling wave signal. The time point corresponding to the modulus maxima point in the target high-frequency sub-band is extracted as the arrival time stamp of the traveling wave. The arrival time stamp of the traveling wave and the sub-band energy value are calculated and obtained. The arrival time stamp of the traveling wave from both ends and the sub-band energy value are combined to establish a time-energy joint feature vector. Based on the time-energy joint feature vector, the total length of the photovoltaic cable and the traveling wave propagation speed parameters are called, and the time difference is calculated by combining the arrival timestamps of the traveling waves at both ends in the time-energy joint feature vector to obtain the time domain positioning coordinates. The cable attenuation coefficient parameter is called, and the energy values ​​of the two terminals in the time-energy joint feature vector are calculated to obtain the energy domain positioning coordinates and obtain the dual-domain fault location estimation set. Based on the dual-end synchronous transient traveling wave signal and the dual-domain fault location estimation set, a sliding time window is set on the dual-end synchronous transient traveling wave signal to obtain the local symmetry breaking factor sequence. Based on the local symmetry breaking factor sequence, the weighted fusion photovoltaic cable fault point is calculated.

[0017] The steps for obtaining a dual-ended synchronous transient traveling wave signal are as follows: Fault recording units are deployed at both ends of the combiner box and the inverter side. A uniform sampling rate and trigger threshold are set. GPS synchronization is enabled to write a timestamp for each sampling point. The fault trigger interval is located and the voltage waveform value and current waveform value of the corresponding time period are extracted according to the timestamp to obtain the voltage waveform value and current waveform value of the dual-end synchronous acquisition. Based on the voltage waveform and current waveform values ​​acquired synchronously at both ends, the lower cutoff frequency parameter and the upper cutoff frequency parameter are set, bandpass filtering is performed, and forward and backward merging is performed. The first and last samples are removed according to the filter order parameter, and the fault triggering interval segment is retained to obtain the voltage waveform and current waveform values ​​after dual-end bandpass filtering. Based on the voltage waveform and current waveform values ​​after dual-end bandpass filtering, the sampling order at both ends is checked according to the absolute timestamp of GPS synchronization and resampled on a unified time axis. Only the abrupt change components in the fault-triggered interval segment are retained and the stable segment is discarded to obtain the dual-end synchronous transient traveling wave signal.

[0018] Specifically, the waveform recorders deployed at both ends of the combiner box and inverter side are initially set to a sampling rate of 1MHz. This sampling rate is sufficient to capture the key characteristics of the high-frequency transient traveling wave signal generated during photovoltaic cable faults. Then, a fault trigger threshold is set. The process for setting this threshold is as follows: The system first collects background noise voltage and current data under normal operating conditions for at least 24 hours. The average value and standard deviation of the voltage waveform and current traveling wave values ​​during this period are calculated. For example, if the standard deviation of the voltage background noise is calculated to be 0.2V, then the voltage trigger threshold is set to the average value plus three times the standard deviation. That is, if the average value is 0V, then the threshold is 0.6V. Similarly, the current trigger threshold is set using the same method for the current traveling wave values ​​to ensure that the trigger logic is sensitive to actual fault disturbances but not to normal operating waveforms. If the system is not sensitive to changes in speed, it then enables GPS synchronization and writes an absolute timestamp to each collected voltage and current sample point with microsecond-level accuracy. When the signal amplitude detected by the fault recording unit at either end instantaneously exceeds the set trigger threshold, the system immediately locates that moment as the fault trigger point. Based on the timestamp of this trigger point, it automatically extracts data within a time window. This time window includes 10 milliseconds of pre-fault data before the trigger point and 50 milliseconds of post-fault data after the trigger point. Based on the start and end absolute timestamps of this 60-millisecond time window, the system accurately extracts all voltage waveform values ​​and current waveform values ​​within the corresponding time period from the continuous data stream of the fault recording units deployed at both ends, thus obtaining the voltage waveform values ​​and current waveform values ​​collected synchronously at both ends.

[0019] Based on the voltage and current waveforms acquired synchronously from both ends, the parameters of the bandpass filter were first set, including the lower cutoff frequency, upper cutoff frequency, and filter order. The lower cutoff frequency was set to 3kHz to filter out the 50Hz power frequency component and its low-order harmonic interference. The upper cutoff frequency was set to 200kHz to filter out high-frequency white noise introduced by the sensor and acquisition system, while retaining the main energy frequency band of the traveling wave front. The filter order was set to 6th order to obtain a sufficiently steep attenuation characteristic in the frequency response transition band. A Butterworth filter was selected for the design. Subsequently, bandpass filtering was performed on the voltage and current waveforms acquired synchronously from both ends. This processing adopted zero-phase filtering technology, specifically by first sequentially... A sixth-order Butterworth bandpass filter is applied sequentially, and then the filtered result sequence is time-reversed. The same sixth-order Butterworth bandpass filter is then applied for a second filtering, and finally the result of the second filtering is time-reversed again to obtain the final filtered sequence. This backward merging process can eliminate the phase delay introduced by the filter and ensure that the time relationship of each frequency component in the signal is not distorted. After the zero-phase filtering is completed, the filtering result needs to be trimmed because the filtering process will produce transient response edge effects at the beginning and end of the data sequence. According to the selected sixth-order filter order parameter, 18 sample points at the beginning and end of the data sequence (usually three times the filter order) are removed, and only the effective segment in the fault trigger interval is retained, thus obtaining the voltage waveform value and current waveform value after double-ended bandpass filtering.

[0020] Based on the voltage and current waveforms after dual-end bandpass filtering, the data from both ends is first aligned using the absolute timestamps from GPS synchronization included in the data. Specifically, the earlier start timestamp from both ends is selected as the starting point of a unified time axis, and a common, discretized time series is constructed using a time step of 1 microsecond corresponding to a 1MHz sampling rate. Then, a cubic spline interpolation algorithm is used to resample the bandpass-filtered data from the combiner box side and the inverter side onto this newly created unified time axis, ensuring that each sampling point of the signals from both ends corresponds strictly in time. After time alignment, the core abrupt change component needs to be separated from the complete fault-triggered interval segment. To this end, the system performs resampling on the resampled signals. The system calculates the first-order difference of the sequence, which involves calculating the amplitude change between adjacent sampling points. It also calculates the standard deviation of the first-order difference sequence within the pre-fault data segment (e.g., 5 milliseconds before the fault trigger). This standard deviation is multiplied by a coefficient of 5 to obtain a dynamic change threshold. The system checks each point starting from the beginning of the signal. The point where the absolute value of the first-order difference exceeds the dynamic change threshold is defined as the starting point of the abrupt change component. The point where the absolute value of the first-order difference remains below the dynamic change threshold after the signal recovers and is defined as the ending point of the abrupt change component. Finally, based on the calculated start and end point indices, the system extracts this core segment containing the initial fault information from the aligned signal and discards the stable segment data before and after it, thus obtaining a dual-end synchronous transient traveling wave signal.

[0021] The steps for obtaining the arrival time stamp of the traveling wave and the sub-band energy value are as follows: Based on the dual-end synchronous transient traveling wave signal, multi-resolution wavelet decomposition is performed and modulus maxima are detected in the target high-frequency sub-band. The absolute time index corresponding to the first modulus maxima is extracted according to the sample order and defined as the initial traveling wave arrival time stamp, forming the left initial traveling wave arrival time stamp and the right initial traveling wave arrival time stamp. Based on the initial arrival timestamps of the traveling wave at the left and right ends, an adaptive time window is established on the target high-frequency sub-band, starting from the arrival timestamps of the traveling wave. Amplitude samples are extracted within the window, and the sub-band energy value is calculated. The formula for calculating the arrival timestamps of the traveling wave and the sub-band energy value is as follows: ; in, Let c be the subband energy value of the c-th end within the adaptive time window. Let be the amplitude of the i-th sample in the c-th high-frequency subband of the target. This is the sample index corresponding to the arrival time stamp of the initial traveling wave at end c. To analyze window length, Take the left and right ends of the port number. For summation index variables.

[0022] Specifically, based on the dual-end synchronous transient traveling wave signal, the "db4" (Daubechies 4) wavelet, which has tight support and orthogonality, is first selected as the mother wavelet. Five-level multi-resolution wavelet decomposition is performed on the dual-end synchronous transient traveling wave signals from the combiner box side and the inverter side. The number of decomposition levels is chosen based on the system's 1MHz sampling rate. After decomposition, the frequency ranges corresponding to each detail sub-band are d1 (250-500kHz), d2 (125-250kHz), d3 (62.5-125kHz), d4 (31.25-62.5kHz), and d5 (15.625-31.25kHz). According to the spectral characteristics of the photovoltaic cable fault traveling wave signal, the d3 sub-band (62.5-125kHz) is selected as the target high-frequency sub-band for subsequent analysis. On the coefficient sequence of this target high-frequency sub-band, the principle of the Canny edge detection algorithm is used to identify modulus maxima, i.e., to find the point with the largest local amplitude in the wavelet transform coefficient sequence. This is done to filter out noise caused by background noise. To identify pseudo-mode maxima, a threshold needs to be set. This threshold is set by extracting the coefficient sequence of the d3 subband within 5 milliseconds before the fault occurs, calculating the standard deviation of the amplitude of all coefficients in the sequence. For example, if the calculated standard deviation is 0.05, the identification threshold is set to three times this standard deviation, i.e., 0.15. Subsequently, starting from the beginning of the target high-frequency subband coefficient sequence, the system scans point by point in chronological order, detecting all modulus maxima with amplitudes exceeding 0.15, and extracting the first modulus maxima. The sample index number of this point in the original signal sampling sequence is recorded, and combined with the GPS absolute timestamp information, its corresponding precise time is calculated. This time is defined as the initial traveling wave arrival timetamp at this end. The above process is repeated for the signals on the combiner box side and the inverter side, ultimately forming the initial traveling wave arrival timetamps at the left and right ends.

[0023] formula: middle, Port numbers are used to distinguish whether the data source is the combiner box side or the inverter side. Specifically, the combiner box side is fixedly numbered 1, i.e. The inverter side is fixedly numbered as 2, that is This identifier is uniquely determined by the configuration information of the fault recording unit during the data acquisition phase and is bound to the acquired waveform data. In the data processing flow, the system reads this identifier to distinguish and process the data from both ends separately, ensuring that subsequent timestamps and energy values ​​can be correctly matched. For example, when processing data from the combiner box, all subscripts in the formula... All were replaced with 1, thus calculating , and Conversely, when processing inverter-side data, it is replaced with 2.

[0024] Let be the amplitude of the i-th sample in the c-th target high-frequency sub-band. This parameter is the coefficient sequence of the target high-frequency sub-band (d3 sub-band) obtained after wavelet decomposition. Each value in this sequence represents the intensity component of the original signal at the corresponding time point and in the corresponding frequency band (62.5-125kHz). These coefficient values ​​are directly obtained from the previous step "perform multi-resolution wavelet decomposition," and their magnitude directly reflects the energy distribution of the transient traveling wave signal in that frequency band. For example, for the combiner box side ( The signal, after wavelet decomposition, yields a time series of d3 subband coefficients. .

[0025] The sample index corresponding to the arrival time stamp of the initial traveling wave at end c is used. This parameter indicates the starting position for calculating the traveling wave front energy. Its value is determined by the previous step of "extracting the absolute time index corresponding to the first modulus maximum point". It represents the number of sampling points traversed from the start of data acquisition to the detection of the first valid traveling wave signal. The accuracy of this index directly affects the effectiveness of the energy calculation because it ensures that the calculation window can accurately cover the initial abrupt change of the traveling wave. For example, if the left end ( The initial arrival time stamp of the traveling wave is determined at the 20150th sampling point, then .

[0026] The analysis window length, representing the number of sample points used to calculate energy, needs to be determined to fully cover the main part of the traveling wave front while avoiding excessive subsequent reflections or noise. Its value is based on statistical analysis of numerous photovoltaic cable fault waveforms. Statistical results show that at a 1MHz sampling rate, the rising edge and peak region of the fault traveling wave front typically last 20 to 40 microseconds. To ensure the stability and representativeness of the calculation, a fixed intermediate value is selected as the window length, here set to 30 sampling points. .

[0027] For summation index variables, it is an index variable that increments from 0 to 1 during the summation operation. An integer counter is used to iterate through each sample point within the analysis window.

[0028] Calculation process: To calculate the sub-band energy value on the combiner box side (left end) For example, First, obtain the parameter from the previous steps: port number. Sample index corresponding to the initial traveling wave arrival time stamp Analysis window length , At the same time, obtain coefficient sequence of time d3 subband And extract 30 sample amplitudes starting from index 20150, for example: , , , ,... , , Substitute the above parameters into the formula: ; Calculation process: ; ; ; Assuming the sum of the squares of the subsequent 26 values ​​is 75.2, the final result is: ; The result indicates that the energy value of the traveling wavefront measured on the combiner box side (left end) within the target high-frequency subband is 83.18. This value will be compared with the energy value calculated on the inverter side (right end) for subsequent energy domain positioning calculations. It is 45.62, because This allows us to initially determine that the fault location is closer to the junction box side.

[0029] The steps for obtaining the time-energy joint feature vector are as follows: Based on the arrival timestamp of the traveling wave and the subband energy value, the initial arrival timestamp of the traveling wave at the left end, the initial arrival timestamp of the traveling wave at the right end, the subband energy value at the left end, and the subband energy value at the right end are obtained. They are then matched according to the port number and the time and energy parameters are combined in a unified structure to form a time-energy joint feature vector.

[0030] Specifically, based on the traveling wave arrival timestamp and subband energy value, the system first retrieves four core data points calculated and stored in the first two steps from memory: the initial traveling wave arrival timestamp on the left (e.g., timestamp value 1627883251.000020 seconds), the initial traveling wave arrival timestamp on the right (e.g., timestamp value 1627883251.000023 seconds), the subband energy value on the left (e.g., value 83.18), and the subband energy value on the right (e.g., value 45.62). Then, the system performs a strict pairing operation based on the data source identifier (i.e., port number, 1 for the left end and 2 for the right end). The system associates the timestamp on the left with the energy value on the left, and the timestamp on the right with the energy value on the right, ensuring that the time-energy correspondence of the data is not confused. After pairing, the system combines these four parameters into a unified data structure in a predefined order. This data structure is defined as a quadruple or a record with four fields, and its standard format is (initial traveling wave arrival timestamp on the left, initial traveling wave arrival timestamp on the right, terminal energy value on the left, terminal energy value on the right). Substituting the aforementioned specific values, the generated instance is (1627883251.000020, 1627883251.000023, 83.18, 45.62). This structured data set is the time-energy joint feature vector.

[0031] The steps for obtaining the dual-domain fault location estimation set are as follows: Based on the time-energy joint feature vector, the arrival timestamps of the traveling wave at the left and right ends are extracted. The absolute value of the difference between the arrival timestamps of the traveling wave at the left and right ends is calculated and the sign of the difference is recorded. The total length of the photovoltaic cable and the traveling wave propagation speed parameters are called to convert the absolute value of the difference into length and determine the direction according to the sign of the difference to obtain the time-domain positioning coordinates. Based on the time-domain positioning coordinates, the energy values ​​of the left and right terminals in the time-energy joint feature vector are read. The logarithmic ratio of the energy values ​​of the left and right terminals is calculated, and the cable attenuation coefficient parameter is called to map the logarithmic ratio to the length offset. The length offset is corrected in the same direction as the time-domain positioning coordinates and the value is limited to the total length of the photovoltaic cable to obtain the energy-domain positioning coordinates. Based on the energy domain location coordinates, the time domain location coordinates and energy domain location coordinates are checked within the total length of the photovoltaic cable and out-of-bounds values ​​are eliminated. The source is recorded according to the port number to form a dual-domain fault location estimation set.

[0032] Specifically, based on the time-energy joint feature vector (1627883251.000020, 1627883251.000021, 83.18, 45.62), the arrival timestamps of the left-end traveling wave (1627883251.000020 seconds) and the right-end traveling wave (1627883251.000021 seconds) are first extracted. Then, the time difference between the two is calculated. seconds, the absolute value of the difference is obtained. The system records the arrival time of the traveling wave at the right end as later than that at the left end, and sets the difference to a positive sign. Then, it retrieves the total length of the photovoltaic cable and the traveling wave propagation speed from a preset parameter library. The total length of the photovoltaic cable is determined based on the photovoltaic power station design drawings or actual on-site surveys and is set to 500 meters. The traveling wave propagation speed is determined through an online calibration method: During the initial system deployment, a standard pulse signal is injected into one end of the cable (e.g., the inverter side), and the arrival time of this pulse is recorded at the other end (the combiner box side). The propagation speed is then calculated based on the total cable length. For example, for a 500-meter cable, if the measured one-way propagation time is 2.94 microseconds, then the traveling wave propagation speed is... The system stores this value (m / s) as the traveling wave propagation velocity parameter. Next, it calculates the fault location based on the dual-end time-based positioning principle. The calculation formula is as follows: ,in This represents the distance from the fault point to the left end (the junction box side). This represents the total length of the photovoltaic cable. For the traveling wave propagation speed parameter, This represents the difference in arrival times of the traveling wave between the right and left ends. Since the difference is positive (the right end arrives later), it indicates that the fault point is closer to the left end. Therefore, a minus sign is used in the formula to substitute the values ​​into the calculation: The time-domain positioning coordinates are obtained from the meters.

[0033] Based on the time-domain positioning coordinates obtained in the previous step, the left endband energy value of 83.18 and the right endband energy value of 45.62 are read from the time-energy joint feature vector (1627883251.000020, 1627883251.000021, 83.18, 45.62). Then, the natural logarithm ratio of these two energy values ​​is calculated. Next, the system calls the cable attenuation coefficient parameter, which characterizes the attenuation rate of traveling wave energy propagating in the cable. Its value is related to the traveling wave frequency and the cable material properties, and is obtained through experimental calibration: In a laboratory environment, a section of photovoltaic cable of the same model as the one used on-site is cut, and a frequency of 100kHz (the center frequency of the target high-frequency sub-band) with energy of [missing value] is input to one end. A sinusoidal pulse signal is received, and the signal energy is measured at the other end of the cable. According to the traveling wave energy attenuation formula The cable attenuation coefficient parameters were calculated. ,in Let the length of the experimental cable be used. For example, for a 100-meter-long cable, with an input energy of 100 and an output energy of 80, then... The system uses this value as the cable attenuation coefficient parameter, and uses this parameter to map the logarithmic ratio of energy to the energy domain location coordinates of the fault point. The calculation formula is as follows: ,in This represents the distance from the fault point to the left end (the junction box side). and Let the energy values ​​be the subband energy values ​​at the left and right ends, respectively. Substitute these values ​​into the calculation: Finally, the calculation results were limited to a range of 0 to 500 meters (the total length of the photovoltaic cable). Since 116.1 meters was within the valid range, the value was confirmed as a valid energy domain positioning coordinate.

[0034] Based on the energy domain positioning coordinates of 116.1 meters and the time domain positioning coordinates of 165 meters obtained in the previous steps, the system first performs boundary checks on these two positioning coordinates. The check is based on the total length of the photovoltaic cable, which is 500 meters, and the effective range is set as [0, 500]. First, the time domain positioning coordinates are checked by comparing their value of 165 meters with the effective range. The coordinates were determined to be valid and retained. Next, the energy domain positioning coordinates were checked, comparing their value of 116.1 meters with the valid interval [0, 500]. Since... The coordinate is also considered valid and retained. If, during the calculation process, any coordinate exceeds the boundary due to signal interference or model error, for example, if a time-domain positioning coordinate is calculated to be -15 meters, the system will determine that the value is invalid in the boundary verification step because -15 is less than the lower limit of the interval, 0. In this case, the out-of-bounds value will be discarded and will not be included in the subsequent fusion calculation. After the verification is completed, the system records all valid positioning coordinates that have passed the verification and their source domains in a structured manner. Specifically, a data set is created, where each element contains two fields: source domain identifier and location coordinate value. In this example, the generated records are "Source: Time Domain, Coordinates: 165.0 meters" and "Source: Energy Domain, Coordinates: 116.1 meters". These two records together constitute the dual-domain fault location estimation set.

[0035] The steps for obtaining the local symmetry breaking factor sequence are as follows: Based on the dual-end synchronous transient traveling wave signal and the dual-domain fault location estimation set, an equally spaced starting point is set on the time axis of the dual-end synchronous transient traveling wave signal, and continuous segments are cut according to a fixed sample length. Each segment is divided into a first half signal and a second half signal while keeping the index consistent, and a sliding time window is generated. According to the sliding time window, the amplitude of the first half of the signal in each sliding time window is recorded point by point and stored as a sequence. The amplitude of the second half of the signal in each sliding time window is reversed according to the time index and recorded point by point and stored as a sequence. The first half signal sequence and the reversed second half signal sequence are paired according to a unified index to generate a pairing sequence of the first half signal and the reversed second half signal. Based on the paired sequences of the first half signal and the reversed second half signal, the difference ratio of the amplitudes of the two sets of sequences within each sliding time window is calculated and normalized. The normalization results of all sliding time windows are arranged in order and kept in correspondence with the time axis to form a sequence of local symmetry breaking factors.

[0036] Specifically, based on the dual-end synchronous transient traveling wave signal and the dual-domain fault location estimation set, the dual-end signals are first fused. The absolute values ​​of the transient traveling wave signals from the combiner box side and the inverter side are added point by point on the same time axis to form a comprehensive analysis signal. This signal can highlight the waveform characteristics detected by both ends. Then, based on the time-domain positioning coordinates of 165 meters and the energy-domain positioning coordinates of 116.1 meters in the dual-domain fault location estimation set, combined with the total length of the photovoltaic cable of 500 meters and the traveling wave propagation speed parameter of 1.7 × 10⁻⁶, the signal is analyzed. 8 Meters per second (m / s) is used to calculate the theoretical arrival time of the reflected wave from the fault point at both ends, thus defining an analysis time interval. For example, for a fault point at 165 meters, the time offset for the reflected wave to arrive at the junction box side (left end) is... The time offset for reaching the inverter side (right end, distance 500-165=335 meters) in microseconds is... Similarly, the arrival time offset of the reflected wave corresponding to the 116.1-meter fault point is calculated in microseconds. The system integrates these theoretical time points to determine a key analysis window encompassing all possible arrival times of the reflected wave. For example, based on the earliest and latest theoretical reflection times, the window is extended by 20 microseconds before and after each time interval to form a concentrated analysis range. Then, the parameters of the sliding time window are set on the time axis of this analysis range. The fixed sample length is set to 100 sampling points (100 microseconds) to ensure that the window can completely cover the main morphology of the reflected wave. The sliding step size is set to 1. The system first selects sampling points to achieve the highest time resolution. Then, starting from the beginning of the analysis interval, it slides backward in steps of 1 sampling point, each time extracting a continuous segment of 100 sampling points. For each extracted segment, the system equally divides it into a first half and a second half, with the first half containing the first 50 sampling points and the second half containing the last 50 sampling points, while maintaining their original time indices. By repeating this sliding, extracting, and dividing process throughout the entire analysis interval, a series of overlapping sliding time windows are finally generated.

[0037] Based on a series of sliding time windows generated in the previous step, the system processes the signal within each window. Taking a sliding time window starting from sample index k as an example, the sample interval covered by this window is [k, k+99], where the first half is [k, k+49] and the second half is [k, k+99]. First, the system reads the amplitude of the first half of the signal in this window, i.e., the 50 samples from index k to k+49, point by point, and stores these amplitudes in a sequence of length 50 according to the original time order, denoted as the first half signal sequence. Next, the system processes the second half of the signal in this window, i.e., the 50 samples from index k+50 to k+99. The system first reads the amplitude of these 50 samples, and then performs a time index order reversal operation, that is, transforming the sequence [amplitude (k+50), amplitude (k+51), ..., amplitude (k+99)] into [amplitude (k+99), amplitude (k+98), ..., The system generates two sequences, each with a length of 50. The first sequence is then paired with the second sequence (k+50), and the reversed sequence is stored as another sequence of length 50, denoted as the reversed second half signal sequence. After generating both sequences, the system pairs them according to a unified index. Specifically, the i-th element of the first half signal sequence is combined with the i-th element of the reversed second half signal sequence (where i ranges from 0 to 49) to form a data pair. For example, the first data pair is (amplitude (k), amplitude (k+99)), the second data pair is (amplitude (k+1), amplitude (k+98)), and so on, with the last data pair being (amplitude (k+49), amplitude (k+50)). Through this process, each sliding time window generates a pairing sequence containing 50 data pairs, which is the pairing sequence of the first half signal and the reversed second half signal.

[0038] Based on the matched sequences of the first half of the signal and the inverted second half of the signal generated for each sliding time window, the system calculates a difference ratio to quantify the degree of signal symmetry breaking within that window. This calculation is performed using the following formula: ,in, It is the difference ratio of the k-th sliding time window. It is the amplitude of the i-th sample in the first half of the signal sequence within the window. It is the amplitude of the i-th sample in the second half of the signal sequence after the window is reversed. It is a very small positive number (e.g.) The system repeats this calculation for all sliding time windows within the analysis interval to obtain an original sequence of difference ratios, for example, the sequence [0.12, 0.15, 0.23, 0.88, 0.31,...]. Then, the entire original sequence of difference ratios is normalized. First, the sequence is traversed to find the minimum (e.g., 0.05) and maximum (e.g., 0.95) values. Then, the min-max normalization formula is applied to each original difference ratio in the sequence to calculate the result. ,in It is the normalized value. and These are the minimum and maximum values ​​of the sequence, respectively. For example, for the original value 0.88, its normalized value is... The system arranges the normalized processing results calculated by all sliding time windows in chronological order on the time axis, and ensures that the timestamp of each normalized value is consistent with the timestamp of the center point of its corresponding window, thus forming a sequence of local symmetry breaking factors.

[0039] The steps for obtaining the fault points of weighted fusion photovoltaic cables are as follows: Based on the local symmetry breaking factor sequence, the amplitudes of adjacent samples are compared point by point according to the sample index, local maxima are screened, and finally the local maximum with the largest amplitude is selected and the corresponding sample index is recorded as the peak position. At the same time, the corresponding amplitude is recorded as the peak value, thus obtaining the peak value and peak position. Based on the peak value and its location, a core energy window and a supporting energy window centered on the peak value are established in the local symmetry breaking factor sequence. The energy concentration coefficient is calculated using the following formula: ; in, Indexed by peak position The energy concentration coefficient centered on the center, For local symmetry breaking factor sequences in the index The amplitude at that point, For peak position index, For the core energy window half-width, To support the energy window half-width and greater than , It is a very small constant used to prevent the denominator from being zero; Based on the energy concentration coefficient, the time-domain and energy-domain positioning coordinates of the dual-domain fault location estimation set are called. The peak value and energy concentration coefficient are synthesized according to the ratio and weighted fusion is performed to obtain a single fault location while keeping the result within the total length of the photovoltaic cable, thus generating a weighted fusion photovoltaic cable fault point.

[0040] Specifically, based on the locally symmetry-breaking factor sequence, the system initiates a peak detection program. This program first traverses the sequence point by point, starting from the second sample point and proceeding to the penultimate sample point. In each iteration, the program compares the amplitude of the current sample point with the amplitudes of its immediate preceding and following sample points. If the amplitude of the current sample point is simultaneously greater than the amplitudes of its two immediate neighbors, then the point is determined to be a local maximum, and its amplitude and corresponding sample index are recorded in a temporary list of local maxima. For example, for a locally symmetry-breaking factor sequence segment...[0.5, 0.9, 0.7, 0.4, 0.6, [0.5]..., When the program processes a sample point with index i and amplitude 0.9, it compares 0.9 with the previous value 0.5 and the next value 0.7. Since 0.9 is greater than both, (index i, amplitude 0.9) is recorded. When processing a sample point with index i+3 and amplitude 0.6, since 0.6 is greater than 0.4 and 0.5, (index i+3, amplitude 0.6) is also recorded. After traversing the entire sequence of local symmetry breaking factors, the system will obtain a sequence containing all... The system generates a list of local maxima. Then, it compares all records in the list by amplitude and selects the entry with the largest amplitude. This largest amplitude is officially recorded as the peak value, and its corresponding sample index is recorded as the peak position. Continuing the previous example, if the list contains (index i, amplitude 0.9) and (index i+3, amplitude 0.6) as well as other items with smaller amplitudes, the system compares and determines 0.9 as the largest amplitude value. Finally, the peak value is 0.922, and the peak position is sample index 150.

[0041] formula: The advantage of the formula is that by calculating the ratio of core energy to support energy, it provides a quantitative confidence assessment of the effectiveness of the identified symmetry-breaking peak. A peak caused by a real fault reflection is usually sharp and energy-concentrated, while a spurious peak caused by noise or interference is relatively flat and energy-dispersed. This feature is captured by using two windows of different widths. The core energy window in the numerator focuses on the most prominent part of the peak, while the support energy window in the denominator includes the peak and the local background around it. The larger the ratio, the more concentrated the energy is in the core of the peak, and the higher the probability that the peak is a real fault signal. For local symmetry breaking factor sequences in the index The amplitude at a given point is the result of the previous step's calculation. It consists of a series of values ​​normalized to the [0, 1] interval. Each value in the sequence reflects the degree of local symmetry breaking of the two-terminal transient traveling wave signal at the corresponding time point. This sequence is the direct data basis for calculating the energy concentration coefficient; its values ​​are directly retrieved from memory without any additional processing. During the calculation, the peak position index is used to determine the amplitude. And the window half-width, extracting amplitude data within the corresponding range from the sequence.

[0042] The peak position index is obtained from the previous step and represents the sample index number where the largest peak in the local symmetry breaking factor sequence is located. For example, based on the calculation results of the previous step, the peak position index... The value is 150.

[0043] The full width at half maximum (FWHM) of the core energy window defines the number of samples that expand outwards from the peak position when calculating the numerator (core energy). Its value determines the measure of peak "sharpness." This parameter is determined through statistical analysis of numerous simulated and field-acquired fault reflection waveforms. Specifically, it analyzes the full width at half maximum (FWHM) of the peaks formed by these waveforms in the local symmetry breaking factor sequence and takes half of their average value as the threshold. For example, by analyzing 100 different types of fault data, the average half-width at half-maximum (FWHM) is calculated to be 20 sample points. Therefore, the core energy window FWHM is... Set to 10.

[0044] To support the energy window half-width, this parameter defines the number of samples that expand outwards from the peak position when calculating the denominator (support energy), and its value must be greater than [missing value]. This parameter is used to establish an energy benchmark that includes the peak and its local background. The setting of this parameter aims to ensure that the window is large enough to include the complete peak shape and a segment of the background signal, but not so large as to introduce other irrelevant signal characteristics. Empirically, Set as Three to five times that of typical background noise fluctuations, here, based on the analysis of typical background noise fluctuation ranges, in order to obtain a stable local energy sum, the half-width of the supporting energy window is... Set to 40.

[0045] It is a very small constant used to prevent the denominator from being zero. The value is set to .

[0046] Calculation process: Based on the previous steps, obtain the parameter: peak position index. Core energy window half width Support energy window half width Minimal constant , Simultaneously, from the local symmetry breaking factor sequence Extracting The relevant data fragments centered on Calculate the molecular (core energy): ; Assuming the sum of squares of the amplitudes within this window is calculated to be 5.82, Calculate the denominator (supporting energy): ; Assuming the sum of squares of the amplitudes within this window is calculated to be 6.85, Substitute the values ​​into the formula: ; ; The results show that the energy concentration coefficient centered on sample 150 is 0.850, which is very close to 1. This means that about 85% of the energy within the supporting energy window is concentrated in the core energy window. This indicates that the detected peak is a very sharp, energy-concentrated, and effective signal with high confidence, and can be considered a reliable indicator of real fault reflection.

[0047] Based on the energy concentration coefficient of 0.850 and peak value of 0.922 calculated in the previous steps, the system first retrieves the time-domain location coordinates (165.0 meters) and energy-domain location coordinates (116.1 meters) from the dual-domain fault location estimation set. Then, the system employs a dynamic weighted fusion strategy to calculate the final fault location. The weighting factor of this strategy is jointly determined by the peak value and the energy concentration coefficient. Specifically, a comprehensive confidence score is defined, which is calculated as the product of the peak value and the energy concentration coefficient. This score reflects the overall reliability of the positioning results based on reflected wave analysis. The system then uses this comprehensive confidence score to assign weights to the energy domain positioning coordinates, while the weight of the time domain positioning coordinates is 1 minus this score, which represents the weight of the energy domain positioning coordinates. Weights of time-domain positioning coordinates Then, a weighted average calculation is performed: The system calculates the single fault location to be approximately 126.7 meters. Finally, the system performs a final boundary check on this result and compares it with the total length of the photovoltaic cable, which is 500 meters. Since 126.7 meters is within the effective range of [0, 500] meters, this result is confirmed as the final valid fault point, and a weighted fusion photovoltaic cable fault point is generated.

[0048] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A photovoltaic cable fault location method based on multi-source data fusion, characterized in that, The method comprises the following steps: A fault recording unit is arranged at both ends of a current combiner box and an inverter of a photovoltaic string, voltage waveform and current waveform at the time of fault occurrence are collected, band-pass filtering is performed, and double-end synchronous transient traveling wave signals are obtained; Based on the double-end synchronous transient traveling wave signals, wavelet decomposition is performed on the double-end synchronous transient traveling wave signals, time points corresponding to modulus maximum points in a target high-frequency sub-band are extracted as traveling wave arrival time stamps, and the traveling wave arrival time stamps and sub-band energy values are calculated and obtained, the traveling wave arrival time stamps and sub-band energy values from both ends are combined, and a time-energy joint feature vector is established; Based on the time-energy joint feature vector, the total length of a photovoltaic cable and a traveling wave propagation speed parameter are called, time difference operation is performed in combination with the double-end traveling wave arrival time stamps in the time-energy joint feature vector, time domain positioning coordinates are obtained, a cable attenuation coefficient parameter is called, and operation is performed in combination with the double-end sub-band energy values in the time-energy joint feature vector, energy domain positioning coordinates are obtained, and a double-domain fault position estimation set is obtained; Based on the double-end synchronous transient traveling wave signals and the double-domain fault position estimation set, a sliding time window is set on the double-end synchronous transient traveling wave signals, a local symmetry breaking factor sequence is obtained, and a weighted fusion photovoltaic cable fault point is calculated and obtained based on the local symmetry breaking factor sequence.

2. The method for photovoltaic cable fault location based on multi-source data fusion according to claim 1, characterized in that, The acquisition step of the double-end synchronous transient traveling wave signals is as follows: A fault recording unit is arranged at both ends of a current combiner box and an inverter of a photovoltaic string, a unified sampling rate and a trigger threshold are set, GPS synchronous timing is enabled to write a time stamp for each sampling point, a fault trigger interval is located, voltage waveform values and current waveform values in a corresponding time period are intercepted according to the time stamp, and double-end synchronous collected voltage waveform values and current waveform values are obtained; Based on the double-end synchronous collected voltage waveform values and current waveform values, lower limit cutoff frequency parameters and upper limit cutoff frequency parameters are set, band-pass filtering is performed, and forward and backward merging is performed to remove first and last samples according to a filter order parameter to reserve a fault trigger interval segment, and double-end band-pass filtered voltage waveform values and current waveform values are obtained; Based on the double-end band-pass filtered voltage waveform values and current waveform values, absolute time stamps of GPS synchronous timing are checked to sequence sampling points at both ends and resample on a unified time axis, only mutation components in the fault trigger interval segment are reserved and stable segments are discarded, and double-end synchronous transient traveling wave signals are obtained.

3. The method for photovoltaic cable fault location based on multi-source data fusion according to claim 1, characterized in that, The acquisition step of the traveling wave arrival time stamps and sub-band energy values is as follows: Based on the double-end synchronous transient traveling wave signals, multi-resolution wavelet decomposition is performed and modulus maximum points are detected in a target high-frequency sub-band, an absolute time index corresponding to a first modulus maximum point is extracted according to a sample order and is defined as an initial traveling wave arrival time stamp, and a left-end initial traveling wave arrival time stamp and a right-end initial traveling wave arrival time stamp are formed; According to the left-end initial traveling wave arrival time stamp and the right-end initial traveling wave arrival time stamp, an adaptive time window is established on the target high-frequency sub-band with the traveling wave arrival time stamp as a starting point, amplitude samples in the window are intercepted, sub-band energy values are calculated, and the traveling wave arrival time stamps and the sub-band energy values are obtained.

4. The method for photovoltaic cable fault location based on multi-source data fusion according to claim 1, characterized in that, The acquisition step of the time-energy joint feature vector is as follows: Based on the traveling wave arrival time stamp and the sub-band energy value, an initial left end traveling wave arrival time stamp, an initial right end traveling wave arrival time stamp, a left end sub-band energy value and a right end sub-band energy value are obtained, corresponding pairing is performed according to port numbers, and time and energy parameters are combined in a unified structure to form a time-energy joint feature vector.

5. The method for photovoltaic cable fault location based on multi-source data fusion according to claim 1, characterized in that, The obtaining step of the double-domain fault location estimation set is: Based on the time-energy joint feature vector, a left end traveling wave arrival time stamp and a right end traveling wave arrival time stamp are extracted, an absolute value of a difference between the left end traveling wave arrival time stamp and the right end traveling wave arrival time stamp is calculated and a difference sign is recorded, a total length of the photovoltaic cable and a traveling wave propagation speed parameter are called, the absolute value of the difference is converted into a length and a direction is determined according to the difference sign, and a time domain positioning coordinate is obtained; According to the time domain positioning coordinate, a left end sub-band energy value and a right end sub-band energy value in the time-energy joint feature vector are read, a logarithmic ratio of the left end sub-band energy value and the right end sub-band energy value is calculated, a cable attenuation coefficient parameter is called to map the logarithmic ratio into a length offset, the length offset is corrected in the same direction as the time domain positioning coordinate and is limited to a range of the total length of the photovoltaic cable, and an energy domain positioning coordinate is obtained. Based on the energy domain positioning coordinate, boundary checking is performed on the time domain positioning coordinate and the energy domain positioning coordinate within the total length of the photovoltaic cable, and out-of-range values are removed, and a source is recorded according to the port number to form the double-domain fault location estimation set.

6. The method for photovoltaic cable fault location based on multi-source data fusion according to claim 1, characterized in that, The obtaining step of the local symmetry breaking factor sequence is: Based on the double-end synchronous transient traveling wave signal and the double-domain fault location estimation set, an equal-interval starting point is set on a time axis of the double-end synchronous transient traveling wave signal, a continuous segment is intercepted according to a fixed sample length, each segment is divided into a front half segment signal and a rear half segment signal and an index is kept consistent, and a sliding time window is generated; According to the sliding time window, a point-by-point amplitude of the front half segment signal of each sliding time window is recorded and stored as a sequence, a point-by-point amplitude of the rear half segment signal of each sliding time window is recorded in reverse order according to a time index sequence and stored as a sequence, the front half segment signal sequence and the reversed rear half segment signal sequence are paired according to a unified index, and a pairing sequence of the front half segment signal and the reversed rear half segment signal is generated; Based on the pairing sequence of the front half segment signal and the reversed rear half segment signal, a difference ratio of amplitudes of two groups of sequences in each sliding time window is calculated and normalized, normalized processing results of all sliding time windows are arranged in order and kept corresponding to the time axis, and a local symmetry breaking factor sequence is formed.

7. The method for photovoltaic cable fault location based on multi-source data fusion according to claim 1, characterized in that, The obtaining step of the weighted fusion photovoltaic cable fault point is: Based on the local symmetry breaking factor sequence, adjacent sample amplitudes are compared point by point according to sample indexes, local maxima are screened, and finally a local maximum with the largest amplitude is selected and a corresponding sample index is recorded as a peak position, and a corresponding amplitude is recorded as a peak value, and a peak value and a peak position are obtained; According to the peak value and the peak position, a core energy window and a support energy window are established in the local symmetry breaking factor sequence with the peak position as the center, and an energy concentration coefficient is calculated.

8. The method for photovoltaic cable fault location based on multi-source data fusion according to claim 7, characterized in that, The acquisition step of the weighted fusion photovoltaic cable fault point further comprises: based on the energy concentration coefficient, calling the time domain positioning coordinates and the energy domain positioning coordinates in the double domain fault position estimation set, proportionally synthesizing the peak value and the energy concentration coefficient, obtaining a single fault position through weighted fusion and keeping the result within the total length range of the photovoltaic cable, and generating a weighted fusion photovoltaic cable fault point.

Images