A fault location detection system for a photovoltaic power plant cable joint

Abstract

The application relates to the technical field of photovoltaic power station cable detection, and discloses a fault positioning and detection system for a photovoltaic power station cable joint. The fault positioning and detection system for the photovoltaic power station cable joint obtains operation data and bidirectional reflection waveforms of multiple working condition sampling windows through a bidirectional observation module, forms a multi-working condition bidirectional observation data set, screens high-confidence structural feature positions and calculates a common propagation parameter set based on the data set and physical topology information through a parameter estimation module, establishes a frequency domain forward propagation model and extracts state-related local reflection parameters through a joint inversion module, determines a contact impedance abnormal candidate joint through a model matching module, re-estimates the common propagation parameter set through a closed-loop review module, reviews the contact impedance abnormal candidate joint based on the complete multi-working condition bidirectional observation data set, and outputs a detection result. The application can be used for abnormal positioning and detection of the photovoltaic power station cable joint.

Description

Technical Field

[0001] This invention relates to the field of photovoltaic power plant cable testing technology, specifically a fault location and detection system for photovoltaic power plant cable joints. Background Technology

[0002] The DC side of a photovoltaic power station typically includes a large number of module connection lines, string cables, and cable joints. Cable joints are constantly exposed to the combined effects of outdoor temperature variations, current-carrying heat generation, humid and hot environments, and fluctuating operating currents, making them prone to problems such as contact surface oxidation, loose crimping, and increased contact resistance. While joint abnormalities may not immediately lead to open-circuit faults in the early stages, they can alter the local impedance state and cause localized heating during continuous operation, affecting the operational reliability of the photovoltaic strings.

[0003] Current cable fault detection methods mostly employ single-end time-domain reflectometry (TDRS), manual inspection, or thermal imaging. TDRS primarily relies on changes in the position and amplitude of the reflection peak to determine the location of anomalies. However, in photovoltaic power plant operation scenarios, the reflected waveform is easily affected by variations in string operating current, cable temperature, inverter state switching, and drift in line propagation parameters. These factors can cause changes in the amplitude, phase, and arrival time of the reflection peak, making it difficult to distinguish between local contact impedance anomalies at joints and changes in line propagation status.

[0004] Furthermore, photovoltaic string cables often contain multiple component connection points and joint locations. Reflected signals from different structural locations may approach each other or undergo multiple reflections and superpositions in the time domain. Judging solely based on a single reflection waveform or the arrival time in a single direction is easily affected by channel time offset, propagation speed variations, and background reflection components, making it difficult to reliably obtain the location of joint anomalies and their corresponding abnormal characteristics.

[0005] Therefore, this invention proposes a fault location and detection system for cable joints in photovoltaic power plants to address the shortcomings of existing technologies. Summary of the Invention

[0006] To address the shortcomings of existing technologies, this invention provides a fault location and detection system for cable joints in photovoltaic power plants. This system solves the problem that existing photovoltaic power plant cable joint detection is susceptible to changes in operating conditions, drift in line propagation parameters, and background reflection interference, resulting in insufficient stability in joint abnormal location and contact impedance abnormality determination.

[0007] To achieve the above objectives, the present invention provides the following technical solution: a fault location and detection system for cable joints in photovoltaic power stations, comprising:

[0008] The bidirectional observation module is used to acquire operating data and bidirectional reflection waveforms from multiple operating condition sampling windows. The bidirectional reflection waveforms include reflection waveforms acquired in the first detection direction and reflection waveforms acquired in the second detection direction. The operating data includes current data, voltage data, and inverter operating status information. The bidirectional observation module is also used to calculate thermal state excitation and electrical transient excitation to form a multi-operating condition bidirectional observation dataset.

[0009] The parameter estimation module is used to determine the location of structural features based on the multi-condition bidirectional observation dataset and physical topology information, filter high-confidence structural feature locations, and calculate the common propagation parameter set.

[0010] The joint inversion module is used to establish a frequency domain forward propagation model based on the multi-condition bidirectional observation dataset and the common propagation parameter set, and to extract state-related local reflection parameters.

[0011] The model matching module is used to determine candidate joints based on the state-related local reflection parameters, and to calculate the reconstruction residual under the normal assumption of the target joint using the candidate joints as the target joints, thereby determining the contact impedance anomaly candidate joints that meet the local contact impedance anomaly judgment rules but do not meet the line propagation parameter anomaly judgment rules.

[0012] The closed-loop verification module is used to shield the abnormal data window corresponding to the candidate joint with abnormal contact impedance to re-estimate the common propagation parameter set, and to verify the candidate joint with abnormal contact impedance based on the re-estimated common propagation parameter set and the complete multi-condition bidirectional observation dataset without abnormal data window shielding. The candidate joint with abnormal contact impedance that passes the verification is confirmed as a faulty joint, and the detection result including the faulty joint location and the abnormality type determination result is output.

[0013] Preferably, the bidirectional observation module is used to acquire the current data of the photovoltaic string within the operating condition sampling window when calculating the thermal state excitation, introduce candidate thermal response parameters, and calculate the thermal state excitation using a discrete accumulation method with continuous integration; the thermal state excitation is calculated according to the following formula:

[0014] ;

[0015] In the formula, Indicates the first Each operating condition sampling window in candidate thermal response parameters Thermal excitation quantity under the following conditions; Indicates the first The start time of each working condition sampling window; Indicates the first The end time of each working condition sampling window; Indicates the operating time of the photovoltaic string Current data; Indicates the first One candidate thermal response parameter; This represents the integral variable at runtime.

[0016] Preferably, the parameter estimation module is used to extract the reflection arrival time of each structural feature position in the first detection direction and the second detection direction when establishing a bidirectional distance closure constraint;

[0017] The first detection direction is the direction in which a reflection test signal is injected from the first detection end of the cable under test and the reflected waveform is acquired at the first detection end;

[0018] The second detection direction is the direction in which a reflection test signal is injected from the second detection end of the cable under test and the reflected waveform is acquired at the second detection end;

[0019] The parameter estimation module is also used to combine the nominal propagation speed of the cable and the initial channel time offset to calculate the first initial unidirectional propagation distance from the structural feature location to the first detection end and the second initial unidirectional propagation distance to the second detection end. When the difference between the sum of the first initial unidirectional propagation distance and the second initial unidirectional propagation distance and the physical length of the cable under test is not greater than the bidirectional distance closure tolerance, the structural feature location is marked as a high-confidence structural feature location.

[0020] Preferably, the parameter estimation module is used to establish and solve the arrival time residual objective function corresponding to the working condition sampling window when calculating the common propagation parameter set; the arrival time residual objective function is constructed according to the following formula:

[0021] ;

[0022] In the formula, Indicates the first The arrival time residual objective function value corresponding to each sampling window under each operating condition; Represents the set of indexes for high-confidence structural features; An index representing the location of a high-confidence structural feature; Indicates the first Within the sampling window of the first working condition Robust weights for high-confidence structural feature locations; and They represent the first The reflection arrival time of each structural feature location in the first and second detection directions; Indicates the first The physical coordinates of the locations of each structural feature; Indicates the physical length of the cable path being measured; Indicates the first The equivalent propagation speed of the line corresponding to each operating condition sampling window; and They represent the first Channel time offset in the first and second detection directions within each working condition sampling window.

[0023] Preferably, the joint inversion module is used to convert the bidirectional reflection waveforms in the multi-condition bidirectional observation dataset into complex frequency domain reflection responses when constructing the frequency domain forward propagation model, determine the round-trip propagation response of each structural feature location in each detection direction based on the common propagation parameter set, and define the frequency domain physical mapping relationship between the complex frequency domain reflection response and the round-trip propagation response, reference reflection component, state-dependent reflection component, electrical transient interference component, and noise component;

[0024] The electrical transient interference component is determined by the frequency domain response coefficient corresponding to the electrical transient excitation and the detection direction.

[0025] Preferably, the joint inversion module is used to decompose the state-related reflection components into thermal state-related reflection components, string common background variation components, and local anomaly components, and to establish a hierarchical joint inversion objective function with the reconstruction error of the frequency domain forward propagation model, the spatial sparsity of the fault-related components, and the consistency of the string common background variation components as constraints.

[0026] The joint inversion module is also used to iteratively solve the hierarchical joint inversion objective function by using an augmented Lagrangian approach that introduces a dual variable matrix, and to extract the state-related local reflection parameters corresponding to each structural feature position. The state-related local reflection parameters include local anomaly components and fault-related components.

[0027] Preferably, the model matching module is used to set the fault-related component of the candidate joint to zero and retain the reference reflection component, normal reference thermal state response component and string common background change component of the candidate joint when using the candidate joint as the target joint and calculating the reconstruction residual under the normal assumption of the target joint.

[0028] The model matching module is also used to retain the state-related reflection components of other structural feature locations besides the candidate joint, and to calculate the predicted reflection waveform under the normal assumption of the target joint by combining the common propagation parameter set and electrical transient interference components under the current operating condition sampling window.

[0029] The model matching module is also used to calculate the waveform difference between the actual observed complex frequency domain reflection response and the predicted reflection waveform, and to use the waveform difference as the reconstruction residual corresponding to the candidate connector.

[0030] Preferably, the model matching module is used to calculate the local residual energy based on the reconstructed residual, and to convert the residual energy peak coordinates based on the local residual energy peak.

[0031] When the residual energy peak coordinate is located within the physical coordinate range of the candidate connector obtained by expanding the installation coordinate range of the candidate connector by a preset positioning tolerance, or within the range of multiple reflection paths that can be derived by the round-trip propagation distance between the candidate connector, adjacent impedance discontinuity positions, and line ends, and the difference between the residual energy peak coordinate calculated by the first detection direction and the second detection direction is not greater than the bidirectional positioning consistency threshold, the model matching module confirms that the candidate connector satisfies the residual position concentration condition and the bidirectional position consistency condition.

[0032] The model matching module is also used to obtain the thermal state excitation variable sequence by removing the mean and normalizing the standard deviation of the thermal state excitation quantity sequence corresponding to the candidate joint, and to calculate the thermal state correlation coefficient between the local residual energy characteristic quantity sequence corresponding to the candidate joint and the thermal state excitation variable sequence. When the absolute value of the thermal state correlation coefficient is not less than the correlation threshold, the candidate joint is confirmed to meet the thermal state correlation condition and the local contact impedance anomaly judgment rule.

[0033] Preferably, the model matching module is used to determine the set of affected structural feature positions after passing the candidate joint according to the line propagation parameter anomaly judgment rules, along the first detection direction and the second detection direction respectively, and calculate the phase shift, time delay shift and attenuation change obtained from the amplitude change based on the phase change, reflection arrival time change and amplitude change between the actual observed reflection response and the reconstructed reflection response, respectively, and determine whether at least two adjacent affected structural feature positions have a phase shift, time delay shift or attenuation change that exceeds the sum of the mean and three times the standard deviation of the corresponding type of phase shift, time delay shift or attenuation change under the initial working condition in the same detection direction, and the corresponding changes have the same change direction;

[0034] When the judgment result is yes, the model matching module determines that the candidate joint meets the abnormal judgment rule of line propagation parameters, and does not output the candidate joint as a candidate joint with abnormal contact impedance.

[0035] Preferably, the closed-loop verification module is used to merge the direct reflection window and the deducible multiple reflection path window of the candidate joint with abnormal contact impedance into an abnormal data window, and construct an abnormal data mask based on the abnormal data window;

[0036] The closed-loop verification module is also used to use anomaly data mask to prevent sampling points within the anomaly data window from participating in the secondary estimation of common propagation parameters, and to generate a remaining observation dataset for re-estimating the common propagation parameter set.

[0037] The detection results output by the closed-loop verification module include a contact impedance anomaly characterization value, which is calculated according to the following formula:

[0038] ;

[0039] In the formula, This indicates that the first faulty connector has been confirmed upon review. A characteristic quantity of contact impedance anomaly for a candidate joint with abnormal contact impedance. Indicates the first Fault-related component matrix at each structural feature location; Represents the Frobenius norm of a matrix; This indicates the operation of finding the maximum value. Indicates the first Reference reflection component matrix at each structural feature location; Indicates the number of sampling windows for operating conditions; Indicates the number of discrete frequency points; This represents the noise scale estimate of the reconstructed residual.

[0040] This invention provides a fault location and detection system for cable joints in photovoltaic power plants. It has the following beneficial effects:

[0041] 1. This invention integrates multi-condition operating data, bidirectional reflection waveforms, common propagation parameter estimation, abnormal joint identification, and closed-loop verification into a single detection process through data coordination between a bidirectional observation module, a parameter estimation module, a joint inversion module, a model matching module, and a closed-loop verification module. The system can locate and verify cable joint anomalies under changing photovoltaic string operating conditions, reducing the susceptibility to changes in operating conditions when relying solely on single reflection waveform judgments.

[0042] 2. This invention utilizes the bidirectional arrival time of structural feature locations to calculate a common propagation parameter set, and incorporates high-confidence structural feature locations into parameter estimation, enabling propagation velocity, attenuation characteristics, phase characteristics, and channel time offset to be updated with the operating condition sampling window. This processing method helps reduce the impact of changes in line propagation state under different operating conditions on subsequent inversion results, and improves the consistency of basic parameters for anomaly location judgment.

[0043] 3. This invention extracts state-related local reflection parameters from the frequency domain forward propagation model and combines them with the reconstruction residuals under the assumption of a normal target joint for judgment. This makes the determination of candidate joints with abnormal contact impedance not only dependent on the reflection peak amplitude, but also on the spatial distribution, bidirectional positional consistency, and thermal state correlation of the reconstruction residuals. This method helps to distinguish between abnormal local contact impedance of the joint and abnormal line propagation parameters, reducing the possibility of misjudging changes in line propagation as joint abnormalities.

[0044] 4. After obtaining candidate joints with abnormal contact impedance, this invention re-estimates the common propagation parameters using an anomaly data window, and then verifies the results based on the re-estimated common propagation parameter set and the complete multi-condition bidirectional observation dataset without anomaly data window shielding. This closed-loop processing method can further constrain the initial judgment results, making the output fault joint location and contact impedance anomaly characterization quantities more stable. Attached Figure Description

[0045] Figure 1 This is a schematic diagram of the system architecture according to an embodiment of the present invention;

[0046] Figure 2 This is a schematic diagram of the method flow according to an embodiment of the present invention;

[0047] Figure 3 This is a schematic diagram of the bidirectional observation module according to an embodiment of the present invention;

[0048] Figure 4 This is a schematic diagram of the parameter estimation module in an embodiment of the present invention;

[0049] Figure 5 This is a schematic diagram of the joint inversion module according to an embodiment of the present invention;

[0050] Figure 6 This is a schematic diagram of the model matching module and closed-loop verification module in an embodiment of the present invention;

[0051] Figure 7 This is a schematic diagram of the bidirectional observation reflection waveform and residual energy spatial distribution in an embodiment of the present invention, wherein (a) is the initial time-domain reflection waveform in the first detection direction and the second detection direction, and (b) is the reconstructed residual energy distribution under the normal assumption of the fault joint;

[0052] Figure 8 This is a schematic diagram of the correlation curve between the thermal state excitation quantity and the local residual energy characteristic quantity in an embodiment of the present invention;

[0053] Figure 9 This is a schematic diagram showing the correspondence between the contact impedance anomaly characterization quantity and the measured contact resistance value in an embodiment of the present invention. Detailed Implementation

[0054] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0055] Reference Figure 1The present invention provides a fault location and detection system for cable joints in photovoltaic power plants, including a bidirectional observation module, a parameter estimation module, a joint inversion module, a model matching module, and a closed-loop verification module.

[0056] The bidirectional observation module includes an operational data acquisition interface, a test control interface, a signal generation unit, a direction switching unit, a reflected waveform acquisition interface, and a reflection acquisition unit. The test control interface is connected to both the signal generation unit and the direction switching unit, and the reflected waveform acquisition interface is connected to the reflection acquisition unit. The bidirectional observation module acquires current and voltage data from the photovoltaic string and inverter operating status information through the operational data acquisition interface. It controls the signal generation unit and the direction switching unit to perform reflection tests through the test control interface, and acquires the bidirectional reflected waveforms of the cable under test through the reflection acquisition unit.

[0057] The first detection direction is the direction in which a reflection test signal is injected into the first detection end of the cable under test and the reflected waveform is acquired at the first detection end; the second detection direction is the direction in which a reflection test signal is injected into the second detection end of the cable under test and the reflected waveform is acquired at the second detection end.

[0058] The bidirectional observation module calculates the thermal state excitation based on the current data, and calculates the electrical transient excitation based on the rate of change of voltage and current data and the inverter operating status information.

[0059] The bidirectional observation module associates bidirectional reflected waveforms, thermal state excitation quantities, electrical transient excitation quantities, and corresponding operating condition timestamps to form a multi-operating condition bidirectional observation dataset.

[0060] The parameter estimation module receives a multi-condition bidirectional observation dataset and determines the locations of structural features based on the physical topology information of the photovoltaic string. These structural feature locations include at least one of the following: cable joint locations, module connection locations, and line end reflection locations.

[0061] The parameter estimation module establishes bidirectional distance closure constraints for the location of structural features and filters high-confidence structural feature locations based on the arrival time relationship in the bidirectional reflection waveforms.

[0062] The parameter estimation module uses the bidirectional arrival times corresponding to the high-confidence structural feature locations to calculate the common propagation parameter set under the current operating conditions. The common propagation parameter set is used to characterize the propagation speed, attenuation characteristics, phase characteristics, and detection channel time offset of the cable under test under the current operating conditions.

[0063] The joint inversion module receives a multi-condition bidirectional observation dataset and a common propagation parameter set, and constructs a frequency domain forward propagation model.

[0064] The joint inversion module is based on the frequency domain forward propagation model and decomposes the reflected waveform into operating condition invariant reflection components, state-dependent reflection components, electrical transient interference components, and noise components.

[0065] The joint inversion module further decomposes the state-related reflection components into thermal state-related reflection components, string common background change components, and local anomaly components.

[0066] The joint inversion module calculates the state-related local reflection parameters of each cable joint based on the thermal state excitation, electrical transient excitation, bidirectional position consistency condition, and spatial sparsity condition.

[0067] The model matching module receives state-related local reflection parameters and performs waveform reconstruction under the normal assumption of the target joint for candidate joints with local anomalous components.

[0068] The model matching module generates reconstruction residuals based on the difference between the reconstructed predicted waveform and the actual observed waveform.

[0069] The model matching module determines whether candidate joints meet the local contact impedance anomaly judgment rules based on the spatial distribution, bidirectional positional consistency, and thermal state correlation of the reconstructed residuals.

[0070] Meanwhile, the model matching module determines whether candidate anomalies meet the line propagation parameter anomaly judgment rules based on the phase shift, time delay shift, or attenuation changes at multiple structural feature locations.

[0071] When a candidate joint meets the local contact impedance anomaly determination rule but does not meet the line propagation parameter anomaly determination rule, the model matching module determines the candidate joint as a contact impedance anomaly candidate joint.

[0072] The closed-loop verification module determines the anomalous data window based on the reflection locations corresponding to the candidate anomalous connectors. During the secondary estimation of common propagation parameters, the closed-loop verification module masks the anomalous data window and generates the remaining observation dataset for re-estimating the common propagation parameters.

[0073] The parameter estimation module recalculates the common propagation parameter set based on the remaining observation dataset.

[0074] Subsequently, the closed-loop verification module inputs the recalculated common propagation parameter set and the original multi-condition bidirectional observation dataset into the joint inversion module and the model matching module, and performs joint inversion and anomaly detection again.

[0075] When the secondary judgment result is consistent with the initial judgment result, the closed-loop verification module outputs the detection results including the location of the faulty joint, bidirectional positioning consistency parameters, and contact impedance anomaly characterization.

[0076] Reference Figure 2 This invention provides a fault location and detection method for cable joints in photovoltaic power plants, comprising the following steps:

[0077] S100: Acquire current data, voltage data, and inverter operating status information under multiple operating conditions; send test trigger commands from the bidirectional observation module to the signal generation unit and the direction switching unit, causing the signal generation unit to inject reflection test signals sequentially from the first detection terminal and the second detection terminal, respectively, and the reflection acquisition unit to acquire the corresponding reflection waveforms; calculate the thermal state excitation quantity based on the current data, and calculate the electrical transient excitation quantity based on the rate of change of voltage data, current data, and inverter operating status information; establish a multi-condition bidirectional observation dataset;

[0078] S200: Determine the location of structural features based on the physical topology information of the photovoltaic string; establish bidirectional distance closure constraints for the structural feature locations and screen high-confidence structural feature locations; use the bidirectional arrival time corresponding to the high-confidence structural feature locations to calculate the common propagation parameter set under the current operating conditions;

[0079] S300: Construct a frequency domain forward propagation model based on a multi-condition bidirectional observation dataset and a common propagation parameter set; decompose the reflected waveform based on the frequency domain forward propagation model and calculate the state-related local reflection parameters corresponding to each cable joint;

[0080] S400: For candidate joints with local abnormal components, perform waveform reconstruction under the normal assumption of the target joint and generate a predicted waveform; generate reconstruction residuals based on the difference between the predicted waveform and the actual observed waveform; determine whether the candidate joint is a candidate abnormal joint by using the local contact impedance anomaly judgment rule and the line propagation parameter anomaly judgment rule.

[0081] In the secondary estimation of common propagation parameters, the S500 shields the abnormal data window corresponding to the candidate joint with abnormal contact impedance; it recalculates the common propagation parameter set using the remaining observation dataset after shielding; it uses the recalculated common propagation parameter set together with the complete multi-condition bidirectional observation dataset without abnormal data window shielding for secondary verification; when the secondary judgment result is consistent with the initial judgment result, it outputs the detection result including the location of the faulty joint, the bidirectional positioning consistency parameter, and the contact impedance abnormality characterization quantity.

[0082] The technical solution of the present invention will be described in detail below with reference to each processing step.

[0083] Reference Figure 3S100 is executed by the bidirectional observation module. The bidirectional observation module collects photovoltaic string operation data and bidirectional reflection waveforms of the tested cable in multiple operating condition sampling windows, and establishes a correlation between the reflection waveforms and the thermal state excitation and electrical transient excitation under the corresponding operating conditions to form a multi-operating condition bidirectional observation dataset, which is then transmitted to the parameter estimation module and the joint inversion module respectively.

[0084] S110: Acquire multi-condition operating data of photovoltaic strings and collect bidirectional reflection waveforms.

[0085] The bidirectional monitoring module acquires current and voltage data of the photovoltaic strings, as well as inverter operating status information, during multiple operating sampling periods. These sampling periods cover the photovoltaic system's operation under different irradiance and ambient temperature conditions.

[0086] To reduce the impact of sudden operational changes on the reflected waveform, the bidirectional observation module first determines whether the current sampling period belongs to a stable operating phase. As one implementation method, if the fluctuation amplitude of the photovoltaic string output current within a 5-minute time window is no greater than 5% of the average current within that time window, the sampling period is considered to be in a stable operating phase. The 5-minute time window and the 5% fluctuation amplitude are used for screening stable operating phases, and can also be initialized based on the inverter data sampling period and on-site operational fluctuations.

[0087] Within the sampling window that meets the reflection detection conditions, the bidirectional observation module sends test trigger commands to the signal generation unit and the direction switching unit, causing the signal generation unit to inject reflection test signals sequentially according to the first detection direction and the second detection direction; the bidirectional observation module acquires the corresponding reflection waveforms through the reflection acquisition unit. The first detection direction refers to the direction in which the reflection test signal is injected from the first detection end of the cable under test and the reflection waveform is acquired at the first detection end; the second detection direction refers to the direction in which the reflection test signal is injected from the second detection end of the cable under test and the reflection waveform is acquired at the second detection end.

[0088] The reflection test signal uses a broadband pulse reflection test signal. The highest effective frequency of the reflection test signal is denoted as . The sampling frequency of the reflected wave is denoted as and satisfy .in, The determination is based on the propagation attenuation characteristics and predetermined spatial resolution of the cable under test. Configure the parameters for the reflection acquisition device and record them in the device configuration data for the current detection task.

[0089] The operating condition sampling window that meets the reflection detection conditions includes: the inverter not performing a maximum power point tracking step scan, and the DC-side contactor not opening or closing. If the current operating condition sampling window does not meet the above conditions, the bidirectional observation module pauses the injection of the reflection test signal and waits for the next operating condition sampling window to re-evaluate.

[0090] Each of the two detection ends uses the injection trigger time of its own reflected test signal as the fast time zero point of the corresponding reflected waveform. The bidirectional observation module uses the operating condition timestamp to correlate the reflected waveforms acquired by the two detection ends to the same operating condition sampling window. The channel time offset of the two detection directions is determined by the channel calibration reflection response and recorded as the input parameters for the estimation of the common propagation parameters.

[0091] The bidirectional observation module records the operating condition timestamp, detection direction, reflection test signal configuration parameters, and sampling configuration parameters corresponding to each reflection waveform. After recording, the bidirectional observation module outputs multi-condition bidirectional reflection waveform data and provides it to the parameter estimation module and the joint inversion module.

[0092] S120: Calculate the thermal excitation.

[0093] The bidirectional observation module calculates the thermal state excitation quantity based on the squared time series of the photovoltaic string operating current. The thermal state excitation quantity is used to characterize the current-carrying heating state of the cable joint within the corresponding operating condition sampling window.

[0094] For the The sampling window for the operating conditions, at the first Candidate thermal response parameters The thermal excitation quantity is calculated according to the following formula:

[0095] ;

[0096] In the formula, Indicates the first Each operating condition sampling window in candidate thermal response parameters Thermal excitation quantity under the following conditions, in units of A 2 ·s; Indicates the first The start time of each working condition sampling window; Indicates the first The end time of each working condition sampling window; Indicates the operating time of the photovoltaic string Current data, in amperes (A); Indicates the first There are 1 candidate thermal response parameters, in seconds; This represents the integral variable at runtime.

[0097] In digital systems, the continuous integration described above is implemented using a discrete accumulation method, with the accumulation step size taken as the sampling period of the running data. The candidate thermal response parameter set consists of positive numbers. The lower limit of the candidate thermal response parameters is not less than the sampling period of the operating data, and the upper limit of the candidate thermal response parameters is not greater than the historical observation duration used to calculate the thermal state excitation. The bidirectional observation module calculates the corresponding thermal state excitation based on each candidate thermal response parameter, forming a thermal state excitation sequence corresponding to the operating condition sampling window. After the thermal state excitation sequence is associated with the corresponding operating condition sampling window, it becomes part of the multi-operating condition bidirectional observation dataset.

[0098] S130: Calculate the electrical transient excitation.

[0099] The bidirectional observation module calculates electrical transient excitation quantities based on the rate of change of current, the rate of change of voltage, and the inverter's operating status information. These electrical transient excitation quantities characterize the transient impact of inverter control actions, operating state switching, or rapid changes in current and voltage on the acquisition of reflected waveforms.

[0100] The bidirectional observation module performs discrete difference calculations on the current and voltage data within the operating condition sampling window. When the time difference between adjacent data points is less than the operating data sampling period, the difference is calculated according to the operating data sampling period; when a duplicate timestamp is detected, the data point corresponding to the duplicate timestamp is skipped. If the time difference between adjacent data points is greater than the disconnection judgment time threshold... If the difference result corresponding to the adjacent data point is invalid, the bidirectional observation module will not include it in the maximum value statistics of the current change rate and voltage change rate. As one implementation method, Take running data sampling period 3 to 10 times. If the first If the number of valid difference results within a sampling window for a given operating condition is less than 80% of the theoretical number of differences for that sampling window, then that sampling window is marked as an invalid sampling window. The theoretical number of differences is the number of sampling points that should be obtained within that sampling window according to the operating data sampling cycle minus 1.

[0101] No. The electrical transient excitation quantity corresponding to each operating condition sampling window is denoted as: Calculate according to the following formula:

[0102] ;

[0103] In the formula, Indicates the first The electrical transient excitation quantity corresponding to each operating condition sampling window is a dimensionless quantity; Indicates the first The result of the current change rate within the sampling window under each operating condition being normalized to the reference change rate; Indicates the first The result of the voltage change rate within each operating condition sampling window after normalization to the reference change rate; This indicates a change in the inverter's operating status. When a preset operating status switch occurs within the sampling window for this operating condition,... ,otherwise ; , and These represent the weights for the rate of change of current, the rate of change of voltage, and the change in operating status, respectively.

[0104] Weights satisfy , , ,and The weights are determined by the operating data collected during initialization and the noise data before the injection of the reflection test signal, and are solved using constrained least squares to minimize the sum of squared residuals between the electrical transient excitation and the noise power within the corresponding operating condition sampling window.

[0105] The reference rate of change of current and the reference rate of change of voltage are determined by the effective operating condition sampling window during initialization. The preset quantiles of the absolute value sequences of current and voltage changes during initialization are taken respectively, and compared with the minimum resolvable rate of change of the corresponding acquisition device, and the larger value is taken. As one implementation method, the preset quantile is the 95th quantile.

[0106] The calculated electrical transient excitation is not directly equivalent to the electrical transient interference component in the frequency domain reflection waveform. The electrical transient excitation, as an input variable of the operating condition, is written together with the thermal state excitation into the multi-operating condition bidirectional observation dataset, and serves as an input field for the joint inversion module to distinguish between thermal state-related reflection changes and electrical transient reflection changes.

[0107] S140: Construct a multi-condition bidirectional observation dataset.

[0108] The bidirectional observation module uses operating condition sampling windows as units, and timestamps the reflected waveforms with the operating condition data. Each reflected waveform uses a fast-time sampling index. Each operating condition sampling window uses an operating condition index. The operating condition data is not copied to each fast time sampling point of the reflected waveform, but is instead used as the status data of the operating condition sampling window to which the reflected waveform belongs.

[0109] No. The first working condition sampling window, the first The time-domain reflectance waveforms corresponding to each detection direction are represented as follows: .in, , Indicates the first detection direction. Indicates the second detection direction; The fast time sampling index represents the time-domain reflected waveform.

[0110] The two-way observation module uses the operating condition sampling window index. and detection direction index As a combined index, the time-domain reflection waveform Thermal state excitation quantity Electrical transient excitation quantity The system binds fields such as working condition timestamp, detection direction information, and reflection test signal configuration parameters, and outputs a multi-condition bidirectional observation dataset.

[0111] If the complete reflection waveforms of the first and second detection directions are not successfully acquired within a certain working condition sampling window, the bidirectional observation module marks the working condition sampling window as an invalid working condition sampling window and does not input the data of the working condition sampling window into the parameter estimation module and the joint inversion module.

[0112] Reference Figure 4 S200 is executed by the parameter estimation module. The parameter estimation module takes the multi-condition bidirectional observation dataset formed in step S100 and the physical topology information of the photovoltaic string as input, determines the location of structural features, filters high-confidence structural feature locations, and calculates the common propagation parameter set corresponding to each condition sampling window.

[0113] S210: Extract and integrate the location coordinates of structural features.

[0114] When the reflected test signal propagates along the cable under test, it will be reflected at locations of impedance discontinuity, such as joints, component connections, and line ends. The parameter estimation module reads the engineering design documents of the photovoltaic string, extracts the string wiring relationship, cable laying length, component connection information, joint installation location information, and line end location information, and determines the location of structural features on the cable under test accordingly.

[0115] Structural feature locations include at least one of the following: cable joint locations, component connection locations, and line end reflection locations. Each structural feature location corresponds to a location index. ,in , This represents the total number of structural feature locations.

[0116] The parameter estimation module establishes a one-dimensional physical space coordinate system with the first detection end of the cable under test as the origin of the spatial coordinate system, along the cable laying path pointing towards the second detection end. The physical length of the cable path under test is denoted as... The unit is m; the first The coordinates of each structural feature location in the one-dimensional physical space coordinate system are: The unit is m, and it satisfies .

[0117] The parameter estimation module determines the positions of each structural feature according to... Sort by size from smallest to largest to form a set of structural feature locations. ,in, Represents the set of structural feature location indices. This represents the total number of structural feature locations. The set of structural feature locations includes the location index, physical coordinates, structural type, and corresponding topology source information.

[0118] S220: Perform bidirectional distance closure check and filter high-confidence structural feature locations.

[0119] In actual reflection waveforms, non-target impedance variations, multiple reflections, and noise can all create additional peaks. The parameter estimation module targets a set of structural feature locations. For each structural feature position, its arrival time in the corresponding reflected waveform in the first and second detection directions is extracted.

[0120] Specifically, the parameter estimation module calculates the theoretical arrival time based on the physical coordinates of the structural feature location, the nominal propagation speed of the cable, and the initial channel time offset, and sets an arrival time search window centered on the theoretical arrival time. The reflected waveform within the search window is subjected to DC removal, filtering, and envelope detection, and the local peak with the amplitude exceeding the noise threshold and the smallest deviation from the theoretical arrival time is selected as the reflected arrival peak. When there are multiple adjacent peaks within the search window, the peaks are assigned according to the principle of minimizing the bidirectional distance closure residual. When the peak amplitude does not exceed the noise threshold or the bidirectional assignment does not meet the distance closure condition, the corresponding structural feature location is marked as a low-confidence structural feature location.

[0121] No. The first detection direction, the first The reflection arrival time corresponding to each structural feature location is denoted as . ,in .

[0122] Since the reflected test signal in single-ended reflection measurement needs to propagate from the detection end to the structural feature location and then return to the detection end, the parameter estimation module calculates the initial unidirectional propagation distance based on the cable's nominal propagation speed and the initial channel time offset:

[0123] ;

[0124] In the formula, Indicates the first The structural feature location to the first The initial unidirectional propagation distance of each detection end, in meters; This indicates the nominal propagation speed of the cable, expressed in m / s. Indicates the first The initial channel time offset for each detection direction, in seconds, is determined by the channel calibration reflection response.

[0125] The parameter estimation module performs bidirectional distance closure verification on the same structural feature location:

[0126] ;

[0127] In the formula, Indicates the first The initial unidirectional propagation distance from each structural feature location to the first detection end, in meters; Indicates the first The initial unidirectional propagation distance from each structural feature location to the second detection end, in meters; Indicates the physical length of the cable path being measured, in meters (m). This indicates the two-way distance closure tolerance, in meters (m).

[0128] Two-way distance closure tolerance The distance closure residual is determined based on the distance closure residual obtained during the initial calibration process. As one implementation method, the parameter estimation module selects the line end reflection position and the connector at a known installation position as the calibration object, calculates the bidirectional distance closure residual for each calibration object, and determines the maximum absolute value of the bidirectional distance closure residual for each calibration object as the bidirectional distance closure tolerance. .

[0129] ;

[0130] ;

[0131] ;

[0132] In the formula, Indicates the first The two-way distance closure residuals at each structural feature location, in meters; This represents the set of known structural feature location indices used for initial calibration.

[0133] When the When the first structural feature location satisfies the above two-way distance closure condition, the parameter estimation module marks it as a high-confidence structural feature location; when the second structural feature location satisfies the above two-way distance closure condition, the parameter estimation module marks it as a high-confidence structural feature location; when the third structural feature location satisfies the above two-way distance closure condition, the parameter estimation module marks it as a high-confidence structural feature location. When a structural feature location does not satisfy the aforementioned two-way distance closure condition, the parameter estimation module marks it as a low-confidence structural feature location. The parameter estimation module then forms a high-confidence structural feature location index set from the structural feature locations that satisfy the two-way distance closure condition. .

[0134] S230: Calculate the public communication parameter set.

[0135] The parameter estimation module utilizes a set of high-confidence structural feature location indices. , establish the first The objective function for the arrival time residuals corresponding to each sampling window under each operating condition is:

[0136] ;

[0137] In the formula, Indicates the first The objective function value of the arrival time residual corresponding to each sampling window under each operating condition, in seconds. 2 ; Represents the set of indexes for high-confidence structural features; Indicates the first Within the sampling window of the first working condition Robust weights for high-confidence structural feature locations; and They represent the first The reflection arrival time of each structural feature location in the first and second detection directions, in seconds; Indicates the first The physical coordinates of each structural feature location, in meters; Indicates the physical length of the cable path being measured, in meters (m). Indicates the first The equivalent propagation velocity of the line corresponding to each operating condition sampling window is expressed in m / s. and They represent the first The channel time offset in the first and second detection directions within each working condition sampling window, in seconds; An index representing the location of a high-confidence structural feature.

[0138] Stable weighting Determined based on the standardized arrival time residuals of each structural feature location. As one implementation method, the Huber weighting rule is used:

[0139] ;

[0140] In the formula, This represents the standardized result of dividing the arrival time residuals by the robust scale of the set of residuals; when At that time, the corresponding structural feature location does not participate in the estimation of common propagation parameters in the current operating condition sampling window.

[0141] The robustness metric is calculated using the median absolute deviation. For the set of arrival time residuals in the current sampling window, the median of the residuals is calculated first, then the median of the absolute values ​​of the differences between each residual and the median is calculated, and 1.4826 times this median is determined as the robustness metric. The robustness metric used to standardize the arrival time residuals is the larger of the initial robustness metric and the time resolution of the detection device. The time resolution of the detection device is a positive number; therefore, when all arrival time residuals are the same and the median absolute deviation is 0, the denominator of the standardization calculation is still a positive number.

[0142] The parameter estimation module has By minimizing the solution, we obtain the th The equivalent propagation speed of the line corresponding to each operating condition sampling window Time offset of the first detection direction channel and the time offset of the second detection direction channel .in, The solution range is determined based on the speed factor range provided by the cable manufacturer, the reflection position at the line end, and the initial calibration results.

[0143] The parameter estimation module performs frequency domain transformation on the reflection waveform segments corresponding to the high-confidence structural feature locations to obtain each discrete frequency point. The complex frequency domain reflection response. For the same high-confidence structural feature location. The parameter estimation module extracts the complex frequency domain reflection responses of the first and second detection directions respectively, and calculates the amplitude ratio term and the phase difference term.

[0144] When calculating the amplitude ratio term, the parameter estimation module adds a lower amplitude limit to the amplitude terms in both the numerator and denominator. To reduce logarithmic overflow caused by weak response frequencies. Lower amplitude limit. Determined based on the background noise amplitude of the reflection acquisition channel; assuming the complex frequency domain reflection response has been normalized. 10 can be taken -6 Up to 10 -4 :

[0145] ;

[0146] ;

[0147] In the formula, Indicates the first The first working condition sampling window, the first The location of a high-confidence structural feature at a frequency point The amplitude ratio term below; This represents the corresponding phase difference term; and These represent the complex frequency domain reflection responses after correction of the injected signal spectrum and the detection channel response in the first and second detection directions, respectively. Indicates the phase expansion operation; This indicates taking the complex phase.

[0148] Under the condition that the local reflection coefficient at the same structural feature location remains consistent in both detection directions, the parameter estimation module uses multiple high-confidence structural feature locations to perform robust regression on the following relationship:

[0149] ;

[0150] ;

[0151] In the formula, Indicates the first Each operating condition sampling window at frequency point The frequency-dependent attenuation parameters are given below, in Np / m. Indicates the first Each operating condition sampling window at frequency point The frequency-dependent phase parameters are given below, in rad / m. This represents the amplitude channel difference between the two detection directions; This represents the phase channel difference between the two detection directions.

[0152] When there are at least three structural feature locations with different physical coordinates among the high-confidence structural feature locations, the parameter estimation module performs robust regression based on the amplitude ratio term and the phase difference term; otherwise, the attenuation parameter and phase parameter obtained from the initial calibration are used.

[0153] The parameter estimation module outputs the first... The common propagation parameter set corresponding to each operating condition sampling window:

[0154] ;

[0155] In the formula, Indicates the first The common propagation parameter set corresponding to each operating condition sampling window is transmitted to the joint inversion module along with the corresponding operating condition sampling window number.

[0156] Reference Figure 5 S300 is executed by the joint inversion module, which takes the multi-condition bidirectional observation dataset and the common propagation parameter set as input, converts the bidirectional reflection waveform under the sampling window of each condition into a complex frequency domain reflection response, and separates the condition-invariant reflection component, state-dependent reflection component, electrical transient interference component and noise component in the frequency domain forward propagation model.

[0157] The joint inversion module further extracts thermal state-related reflection components, string common background change components, and local anomaly components from the state-related reflection components to obtain the state-related local reflection parameters corresponding to each structural feature location.

[0158] S310: Calculate the complex frequency domain reflection response and establish a frequency domain forward propagation model.

[0159] The joint inversion module reads the time-domain reflection waveforms from the multi-condition bidirectional observation dataset. Then, perform a frequency domain transformation on it to obtain the corresponding complex frequency domain reflection response. .in, Indicates the direction of detection. Indicates the index of the operating condition sampling window. The fast time sampling index represents the time-domain reflected waveform. Indicates the first A discrete frequency point.

[0160] Complex frequency domain reflection response It is the frequency domain response after correction of the injected signal spectrum and the detection channel response, which simultaneously contains amplitude and phase information.

[0161] The joint inversion module uses the common propagation parameter set output by the parameter estimation module. Define the complex frequency domain reflection response. Frequency domain forward physical mapping relationship between round-trip propagation response, reference reflection component, state-dependent reflection component, electrical transient interference component, and noise component:

[0162] ;

[0163] In the formula, This represents the total number of structural feature locations; Indicates the first Within the sampling window of the first working condition The position of the structural feature relative to the first Round-trip propagation response in each detection direction; Indicates the first Reference reflection components at each structural feature location; Indicates the first The structural feature location is at the [number]th [location]. State-related reflection components under each operating condition sampling window; Indicates the first The first detection direction, the first Electrical transient interference components under each operating condition sampling window; This represents the noise component. All the frequency domain response quantities mentioned above are complex dimensionless quantities.

[0164] The round-trip propagation response is calculated according to the following formula:

[0165] ;

[0166] In the formula, Indicates the first The structural feature location to the first The unidirectional propagation distance of each detection end, in meters; when hour, ;when hour, ; Indicates the first The coordinates of each structural feature position relative to the first detection end are in meters. This represents the frequency-dependent attenuation parameter, in Np / m. This represents frequency-dependent phase parameters, in rad / m. It represents the imaginary unit.

[0167] The electrical transient interference component is determined by the frequency domain response coefficients corresponding to the electrical transient excitation and the detection direction.

[0168] ;

[0169] In the formula, Indicates the first Each detection direction at the frequency point The frequency domain response coefficients of electrical transient interference are determined. These frequency domain response coefficients, as parameters to be determined by the joint inversion module, are solved jointly by the complex frequency domain reflection responses and electrical transient excitation quantities from multiple operating condition sampling windows.

[0170] S320: Decompose state-related reflection components.

[0171] The joint inversion module decomposes the state-related reflection components based on the thermal state excitation sequence and electrical transient excitation obtained in step S100.

[0172] For the At each structural feature location, the joint inversion module iterates through each candidate thermal response parameter in the candidate thermal response parameter set one by one. Substituting the corresponding thermal state excitation sequences into the frequency domain forward propagation model, the normalized reconstruction residuals of the structural feature location in two detection directions and multiple operating condition sampling windows are calculated. The candidate thermal response parameter with the smallest normalized reconstruction residual is then determined as... .

[0173] The joint inversion module is based on Achieve the first The thermal state excitation sequence corresponding to each structural feature location is then processed by removing the mean and normalizing the standard deviation to obtain the dimensionless thermal state excitation variables:

[0174] ;

[0175] In the formula, Indicates the first The structural feature location is at the [number]th [location]. Dimensionless thermal state excitation variables under a single operating condition sampling window; Indicates adoption The calculated thermal excitation quantity; express The mean value in each sampling window under each operating condition; express The standard deviation of each operating condition sampling window.

[0176] like Less than the positive threshold Then Take as Positive threshold Determined based on the minimum resolvable current change of the operating current acquisition device; as one implementation method The standard deviation of the minimum non-zero thermal state excitation quantity is obtained by substituting the minimum resolvable current change into the calculation formula of the thermal state excitation quantity.

[0177] State-dependent reflection components are decomposed according to the following formula:

[0178] ;

[0179] In the formula, Indicates the first Normal reference thermal state response coefficients at each structural feature location; Indicates the first The deviation of the response coefficient of each structural feature location relative to the normal reference thermal state; Indicates the common background variation components of the string; This indicates a local abnormal component.

[0180] Normal reference thermal state response coefficient The normal reference thermal state response coefficient is determined based on the normal historical operating data of the structural feature location. When no normal historical operating data exists, for each discrete frequency point, the median of the thermal state response coefficient of the same type of high confidence structural feature location within the same string group is taken as the normal reference thermal state response coefficient of the structural feature location. When there are fewer than 3 high confidence structural feature locations of the same type, the normal reference thermal state response coefficient obtained from the initial calibration is used.

[0181] The joint inversion module defines the following formula as the first... Fault-related components at each structural feature location:

[0182] ;

[0183] In the formula, Indicates the first The structural feature location is at the [number]th [location]. The first working condition sampling window and the first Fault-related components at discrete frequency points.

[0184] S330: Solve for state-dependent local reflection parameters.

[0185] The joint inversion module establishes a hierarchical joint inversion objective function, constrained by the reconstruction error of the frequency domain forward propagation model, the spatial sparsity of fault-related components, and the consistency of string common background variation components:

[0186] ;

[0187] In the formula, This represents the objective function value obtained through hierarchical joint inversion. This represents the reconstructed frequency domain reflection response calculated using the frequency domain forward propagation model; Indicates by The fault-related component matrix formed at all operating condition sampling windows and all discrete frequency points; This represents the Frobenius norm, used to constrain local anomalies to be concentrated in a few structural feature locations; Indicates the common background variation component of the string. The background change matrix is ​​composed of; The nuclear norm is used to constrain the common background variation components of a string to have common variation characteristics across multiple structural feature locations. Indicates spatial sparse weights; This indicates the weight of changes in the public context.

[0188] and The weights are non-negative. The joint inversion module uses the verification condition sampling window from the initialization data to determine the weights. and The value of is determined by the joint inversion module, which iterates through a preset candidate set. and The parameters are calculated by taking the possible combinations of values, calculating the normalized reconstruction error for each combination, and selecting the combination that minimizes the normalized reconstruction error as the current parameter combination. The preset candidate set can be set with equal intervals, for example, 10... -3 Up to 10 1 Select multiple candidate values ​​within the range.

[0189] The joint inversion module solves the objective function using an augmented Lagrangian approach. During the solution process, the joint inversion module introduces a dual variable matrix. And a quadratic penalty term, and perform the following updates in each iteration:

[0190] First, fixed-string common background variation components The fault-related components are updated using a soft-threshold shrinkage operator and the thermal response coefficient. .

[0191] Second, fix the fault-related components. The singular value threshold shrinkage operator is used to update the common background change components of the string. .

[0192] Third, fixed and Updated using regularized least squares method , And the frequency domain coefficients corresponding to the electrical transient interference components.

[0193] Fourth, update the dual variable matrix. And calculate the original residual and dual residual of the current iteration.

[0194] In regularized least squares, the joint inversion module adds a diagonal regularization term to the normal equation matrix. The coefficient of this diagonal regularization term is positive and determined based on the scale of the current normal equation matrix. As one implementation, the joint inversion module sets the coefficient of the diagonal regularization term to 10 times the largest diagonal element of the normal equation matrix. -4 Up to 10 -2 Alternatively, it can be set to 10 of the largest eigenvalue of the normal equation matrix. -4 Up to 10 -2 .

[0195] When the relative change in the objective function between two consecutive iterations is no greater than 1.0 × 10 -4 Furthermore, both the normalized original residual and the normalized dual residual are no greater than 1.0 × 10⁻⁶. -3 The joint inversion module stops iterating when the objective function is satisfied for two consecutive iterations. When the number of iterations reaches 500, the joint inversion module stops iterating and outputs the parameter result with the minimum current objective function value.

[0196] The joint inversion module outputs state-related local reflection parameters, which include... , , and the corresponding thermal response parameters The joint inversion module transmits state-related local reflection parameters to the model matching module.

[0197] Reference Figure 6 S400 is executed by the model matching module. The model matching module receives the state-related local reflection parameters output by the joint inversion module, determines the candidate abnormal joints, and compares and judges the candidate abnormal joints according to the local contact impedance anomaly judgment rules and the line propagation parameter anomaly judgment rules respectively.

[0198] The model matching module identifies structural feature locations that meet the following conditions as candidate abnormal joints: candidate abnormal joints are those initially screened by fault correlation components; candidate abnormal joints confirmed by the local contact impedance anomaly determination rule but not meeting the line propagation parameter anomaly determination rule are recorded as contact impedance anomaly candidate joints; and contact impedance anomaly candidate joints that pass the closed-loop verification module are recorded as faulty joints.

[0199] ;

[0200] In the formula, Represents the set of candidate abnormal connector indices; Indicates the first Fault-related component matrix at each structural feature location; Denotes the Frobenius norm; This represents the candidate anomaly threshold. Candidate anomaly threshold The norm distribution of the fault-related components at high-confidence structural feature locations under initial operating conditions is determined. As one implementation method, the model matching module calculates the mean of the norms of the fault-related components at each high-confidence structural feature location under initial operating conditions. and standard deviation And set the candidate anomaly threshold as follows:

[0201] ;

[0202] In the formula, This represents the mean norm of the fault-related components at the high-confidence structural feature location under initial operating conditions. This represents the corresponding standard deviation.

[0203] S410: Waveform reconstruction and reconstruction residual calculation under the normal assumption of the target joint.

[0204] The model matching module targets the candidate abnormal connector index set. The first in From the candidate abnormal connectors, we establish the assumption that the target connector is normal.

[0205] Under the normal assumption of the target connector, the model matching module will... Fault-related components of candidate abnormal connectors Set to 0, and retain the reference reflection component, normal reference thermal state response component, and string common background variation component of the candidate anomalous connector.

[0206] The model matching module also retains all models except the first one. State-related reflection components, electrical transient interference components, and common propagation parameter sets at other structural feature locations besides the candidate anomalous joints.

[0207] The predicted reflection waveform under the normal assumptions of the target connector is calculated using the following formula:

[0208] ;

[0209] In the formula, Indicates the first When the candidate abnormal joint is under the normal assumption of the target joint, the first... The first working condition sampling window, the first The first detection direction, the first The predicted reflection waveforms corresponding to each discrete frequency point; Indicates except the first Index of other structural features besides the candidate anomalous connectors; and These represent the round-trip propagation responses at the corresponding structural feature locations; and These represent the reference reflection components; Indicates the first State-related reflection components at each structural feature location; Indicates the first Normal reference thermal state response components of each candidate abnormal joint; Indicates the common background variation components of the string; This represents the electrical transient interference component.

[0210] The model matching module calculates the first value based on the difference between the actual complex frequency domain reflection response and the predicted reflection waveform. The reconstruction residuals corresponding to each candidate anomalous connector:

[0211] ;

[0212] In the formula, Indicates that for the first Complex frequency domain reconstruction residuals of candidate anomalous connectors; This represents the actual complex frequency domain reflection response.

[0213] S420: Judgment of local contact impedance anomaly.

[0214] The model matching module reconstructs the residuals in the complex frequency domain. Perform an inverse frequency domain transform to obtain the time-domain reconstructed residual. .

[0215] The model matching module sets a local energy statistics window on the temporal reconstruction residuals and calculates the local residual energy:

[0216] ;

[0217] In the formula, Indicates the first Local residual energy corresponding to each candidate anomalous connector; Indicates the first A local energy statistics window centered on a fast-time sampling point; This represents the index of the sampling point within the local energy statistics window. The length of the local energy statistics window is determined based on the main lobe width of the reflection test signal obtained from the initial calibration.

[0218] The model matching module calculates the coordinates of the residual energy peak in the first and second detection directions based on the fast-time sampling point corresponding to the local residual energy peak, the equivalent propagation speed of the line, and the channel time offset.

[0219] When the residual energy peak coordinate is located at the first When the physical coordinates of a candidate abnormal connector are within a range, or within a derivable multiple reflection path formed by the candidate abnormal connector and adjacent impedance discontinuities, the model matching module determines that the candidate abnormal connector satisfies the residual location concentration condition. The range of multiple reflection paths is calculated and determined based on the round-trip propagation distance between the candidate abnormal connector, adjacent impedance discontinuities, and the line ends.

[0220] Specifically, for a deducible multiple reflection path that includes the detection end, candidate abnormal connectors, adjacent impedance discontinuity locations, and line ends, the path nodes are arranged according to the reflection propagation order, and the absolute values ​​of the physical coordinate differences between adjacent path nodes are summed to obtain the total propagation distance of the multiple reflection path. Then, based on the line equivalent propagation speed of the current operating condition sampling window and the channel time offset of the corresponding detection direction, the total propagation distance is converted into the arrival time of multiple reflections and the equivalent residual peak coordinates. With the equivalent residual peak coordinates as the center, the range determined by the main lobe width and positioning tolerance of the reflection test signal space is expanded to both sides to obtain the range of the deducible multiple reflection path.

[0221] The model matching module further compares the residual energy peak coordinates of the first and second detection directions. When the difference between the residual energy peak coordinates obtained from the two detection directions is not greater than the bidirectional positioning consistency threshold... When the candidate abnormal connector is determined to meet the bidirectional positional consistency condition, it is determined that the abnormal connector meets the condition.

[0222] Bidirectional positioning consistency threshold The coordinate deviation distribution of the same known reflection position in two detection directions during the initial calibration process is determined. The model matching module calculates the mean of the absolute values ​​of the coordinate deviations. and standard deviation And set the bidirectional positioning consistency threshold to .

[0223] Bidirectional positioning consistency threshold This is used to constrain the deviation between the residual energy peak coordinates obtained from the first and second detection directions for the same candidate anomalous connector. The bidirectional distance closure tolerance in step S220... The two methods are used to constrain the deviation between the sum of the unidirectional propagation distances from the same structural feature location to the two detection ends and the physical length of the cable under test. The calculation objects and judgment purposes of the two methods are different.

[0224] Among them, bidirectional distance closure tolerance The maximum value of the distance closure residual at the known structural feature locations used in the initial calibration is taken to cover the distance closure error that has already occurred during the initial calibration process; bidirectional positioning consistency threshold. The distribution of repeated positioning deviations at the same known reflection position in two detection directions is used to evaluate the consistency of bidirectional positioning results for candidate abnormal joints.

[0225] The model matching module is for the first For each candidate abnormal connector, the local residual energy characteristic is calculated within the sampling window of each operating condition. The local residual energy characteristic is obtained by summing the temporal reconstruction residual energy within the direct reflection window corresponding to the candidate anomalous connector and its derivable multiple reflection path window.

[0226] The model matching module calculates the sequence of local residual energy features. With thermal state excitation variable sequence The correlation between them was determined, and the thermal state correlation coefficient was obtained. :

[0227] ;

[0228] In the formula, Indicates the first Thermal state correlation coefficient of each candidate abnormal joint; This indicates the calculation of the correlation coefficient.

[0229] When a candidate anomalous joint simultaneously satisfies the residual location concentration condition, the bidirectional location consistency condition, and the thermal state correlation condition, the model matching module determines that the candidate anomalous joint meets the local contact impedance anomaly determination rule. The thermal state correlation condition is determined by the thermal state correlation coefficient. Determined by comparison with a correlation threshold; correlation threshold The correlation coefficients of the thermal states at high-confidence structural feature locations under initial operating conditions are determined. The model matching module uses high-confidence structural feature locations not included in the candidate abnormal joint index set as reference locations, and takes the 95th percentile of the absolute value of the thermal state correlation coefficient at each reference location as the correlation threshold. .when When the candidate abnormal joint is determined to meet the thermal state correlation condition, it is determined that the abnormal joint meets the thermal state correlation condition.

[0230] S430: Determination of abnormal line propagation parameters.

[0231] The model matching module determines whether candidate anomalies meet the line propagation parameter anomaly judgment rules based on the time-domain reconstruction residual and the complex frequency-domain reconstruction residual.

[0232] When at least two adjacent structural feature locations following a candidate location exhibit phase shift, time delay shift, or attenuation changes in the same detection direction, and the corresponding changes all exceed the sum of the mean and three standard deviations of the corresponding type of phase shift, time delay shift, or attenuation change under the initialization condition, and have the same direction of change, the model matching module records this change as a propagation influence feature. Phase shift is determined based on the phase change of the complex frequency domain reconstruction residual, time delay shift is determined based on the reflection arrival time change, and attenuation change is determined based on the amplitude change of the complex frequency domain reflection response; the aforementioned phase change, reflection arrival time change, and amplitude change are data obtained based on the difference between the actual observed reflection response and the reconstructed reflection response.

[0233] Among them, the effective frequency point is the frequency point where the amplitude of both the actual observed complex frequency domain reflection response and the reconstructed complex frequency domain reflection response is not less than the lower limit of amplitude; the phase offset is obtained by phase expansion and summing of the difference between the actual observed phase and the reconstructed phase at the effective frequency point; the time delay offset is obtained by the difference between the arrival time of the actual observed reflection peak and the arrival time of the reconstructed reflection peak; the amplitude change is obtained by the ratio or logarithmic difference between the actual observed amplitude and the reconstructed amplitude at the effective frequency point, and is converted into the attenuation change; the phase offset, time delay offset and attenuation change under the initial operating condition are respectively used as the benchmark statistical samples of the same type of change to calculate the mean and standard deviation of the corresponding type.

[0234] For the For each candidate anomaly connector, the model matching module determines the set of affected structural feature locations in two detection directions. The set of affected structural feature locations in the first detection direction includes the structural feature locations after passing through the candidate locations when propagating along the first detection direction; the set of affected structural feature locations in the second detection direction includes the structural feature locations after passing through the candidate locations when propagating along the second detection direction.

[0235] When the set of affected structural feature locations under two detection directions corresponds to the propagation path passing through the candidate location, and at least two adjacent affected structural feature locations satisfy the propagation influence feature determination condition, the model matching module determines that the candidate anomaly satisfies the line propagation parameter anomaly determination rule.

[0236] If a candidate abnormal joint meets the local contact impedance abnormality judgment rule but does not meet the line propagation parameter abnormality judgment rule, the model matching module marks the candidate abnormal joint as a contact impedance abnormality candidate joint.

[0237] If a candidate anomaly meets the rule for determining line propagation parameter anomalies, the model matching module will mark the anomaly as a line propagation parameter anomaly and will not output it as a candidate connector for contact impedance anomalies.

[0238] The model matching module outputs the index, physical coordinates, reconstruction residual, bidirectional positioning consistency parameters, thermal state correlation coefficient, and anomaly type determination results for candidate joints with abnormal contact impedance.

[0239] Reference Figure 6 The S500 process is executed by the closed-loop verification module. The closed-loop verification module receives the contact impedance anomaly candidate joints, reconstruction residuals, residual energy peak coordinates, bidirectional positioning consistency parameters, and thermal state correlation coefficients output by the model matching module, and constructs an anomaly data mask. The parameter estimation module recalculates the common propagation parameter set based on the anomaly data mask. The closed-loop verification module then uses the recalculated common propagation parameter set to perform consistency verification on the contact impedance anomaly candidate joints and outputs the detection results.

[0240] S510: Construct anomaly data mask and generate the remaining observation dataset.

[0241] For the output of the model matching module, the first For each candidate connector with abnormal contact impedance, the closed-loop verification module determines its corresponding fast-time sampling position based on the residual energy peak coordinates of the candidate connector in the first and second detection directions.

[0242] The closed-loop verification module determines the direct reflection window centered on the fast-time sampling position, based on the main lobe width of the reflected test signal. The main lobe width is determined by the system impulse response obtained during the initialization calibration process. For windows with derivable multiple reflection paths, the closed-loop verification module calculates the corresponding arrival time of multiple reflections and the fast-time sampling position based on the total propagation distance of the path formed by candidate joints, adjacent impedance discontinuities, and the end of the line. Using this fast-time sampling position as the center, the module determines the half-width of the window by combining the main lobe width of the system impulse response and the uncertainty of the propagation parameters. The direct reflection window is then combined with each derivable multiple reflection path window to form the abnormal data window.

[0243] The closed-loop review module also follows the first Based on the candidate joint with abnormal contact impedance, the location of adjacent impedance discontinuity, and the round-trip propagation distance between the ends of the line, the derivable multiple reflection path window corresponding to the candidate joint is determined.

[0244] The closed-loop review module will be the first The direct reflection window and the deducible multiple reflection path window corresponding to each candidate joint with abnormal contact impedance are merged to obtain the abnormal data window. When multiple candidate joints with abnormal contact impedance exist, the closed-loop verification module takes the union of the abnormal data windows corresponding to each candidate joint:

[0245] ;

[0246] In the formula, This represents the set of candidate joint indices for contact impedance anomalies. Indicates the first The first working condition sampling window, the first An abnormal data window consists of a direct reflection window and a window with derivable multiple reflection paths in each detection direction.

[0247] The closed-loop verification module uses the union of abnormal data windows. Construct a unified mask for anomalous data:

[0248] ;

[0249] In the formula, Indicates the first The first working condition sampling window, the first A unified anomaly data mask corresponding to each detection direction; This represents the union of all candidate connectors with abnormal contact impedance corresponding to abnormal data windows; The fast time sampling index represents the time-domain reflected waveform.

[0250] The closed-loop verification module generates the remaining observation dataset based on the outlier data mask. The remaining observation dataset is only used as input for the secondary estimation of the common propagation parameters; sampling points with an outlier data mask value of 0 do not participate in the common propagation parameter estimation, while sampling points with an outlier data mask value of 1 participate in the common propagation parameter estimation.

[0251] S520: Secondary correction of public communication parameters and execution of consistency verification.

[0252] The parameter estimation module receives the remaining observation dataset and recalculates the parameters according to the common propagation parameter estimation method in step S200. The common propagation parameter set corresponding to each sampling window under each operating condition is used to obtain the re-estimated common propagation parameter set, i.e., the second-corrected common propagation parameter set:

[0253] ;

[0254] In the formula, Indicates the first The set of secondary corrected common propagation parameters corresponding to each working condition sampling window; This represents the equivalent propagation speed of the line after the second correction; This represents the frequency-dependent attenuation parameter after secondary correction; This represents the frequency-dependent phase parameters after the second correction. and These represent the channel time offsets in the first and second detection directions after the second correction.

[0255] After the second estimation of the common propagation parameters is completed, the closed-loop verification module reads the complete multi-condition bidirectional observation dataset without abnormal data masking and inputs the complete multi-condition bidirectional observation dataset and the second-corrected common propagation parameter set into the joint inversion module.

[0256] The joint inversion module re-executes the hierarchical joint inversion based on the second-corrected common propagation parameter set to obtain the local reflection parameters related to the second-inversion state. The model matching module recalculates the predicted reflection waveform and reconstructed residual under the normal assumption of the target joint based on the local reflection parameters related to the second-inversion state, and re-executes the local contact impedance anomaly judgment rules and the line propagation parameter anomaly judgment rules.

[0257] The closed-loop review module compares the initial judgment result with the secondary judgment result. When the... If a candidate joint with abnormal contact impedance still meets the local contact impedance abnormality judgment rule in the secondary judgment, but does not meet the line propagation parameter abnormality judgment rule, the closed-loop verification module will... One candidate connector with abnormal contact impedance was confirmed as a faulty connector. This faulty connector is a candidate connector with abnormal contact impedance that has been verified as a faulty connector.

[0258] If the first If a candidate joint with abnormal contact impedance no longer meets the local contact impedance abnormality judgment rule or meets the line propagation parameter abnormality judgment rule in the secondary judgment, the closed-loop verification module will not identify the candidate joint as a faulty joint and will output a verification failure mark.

[0259] S530: Outputs the detection results.

[0260] For target objects identified as faulty connectors, the closed-loop verification module outputs the detection results. The detection results include the physical coordinates of the faulty connector, connector number, bidirectional positioning consistency parameters, thermal state correlation coefficient, contact impedance anomaly characterization quantity, and anomaly type determination result.

[0261] The bidirectional positioning consistency parameter is the deviation between the residual energy peak coordinates obtained in the first and second detection directions.

[0262] The contact impedance anomaly characterization quantity is denoted as Calculate according to the following formula:

[0263] ;

[0264] In the formula, This indicates that the first faulty connector has been confirmed upon review. The contact impedance anomaly characterization quantity of a candidate joint with abnormal contact impedance. Indicates the first Fault-related component matrix at each structural feature location; Indicates the first Reference reflection component matrix at each structural feature location; Indicates the number of sampling windows for operating conditions; Indicates the number of discrete frequency points; This represents the noise scaling estimate of the reconstructed residuals. , This represents the set of reconstructed residual sample indices outside the anomalous data window; This indicates the number of samples in the index set; Represents the Frobenius norm of a matrix; This indicates the operation of finding the maximum value.

[0265] The contact impedance anomaly characterization quantity is a relative state parameter calculated based on the reflection response, and does not represent the absolute value of the contact resistance obtained through independent resistance measurement.

[0266] In one implementation, the closed-loop verification module encapsulates the detection results into a data packet and sends it to the host computer monitoring system via a communication interface. The data packet includes the physical coordinates of the faulty connector, connector number, bidirectional positioning consistency parameters, thermal state correlation coefficient, contact impedance anomaly characterization quantity, and anomaly type determination result.

[0267] The host computer monitoring system parses the data packets and displays the location of the faulty connector and the corresponding detection results in the preset cable topology diagram.

[0268] The following application examples, based on a set of on-site testing data from a photovoltaic power station, illustrate the working process of the technical solution of this invention.

[0269] The target of the test is set as a 1500V DC photovoltaic string in a photovoltaic power station. This string uses 4mm... 2The photovoltaic-specific cable under test has a physical laying length of 120m from the first testing end to the second testing end. According to the engineering topology information, there are 5 structural feature locations (all of which are component cable joints) on this line, with physical coordinates of 15m, 45m, 65m, 95m, and 115m respectively. The system performed testing over 3 consecutive typical sunny days, acquiring a total of 45 valid operating condition sampling windows.

[0270] During the test, the bidirectional observation module collected the string operating current and voltage in real time and calculated the thermal excitation. The operating current reached a peak of 13.5A at midday, corresponding to a peak thermal excitation of approximately 3.8 × 10⁻⁶. 4 A 2 After acquiring the multi-condition bidirectional observation dataset, the parameter estimation module, using the known distances to the end reflection location and other high-confidence structural feature locations, calculated the equivalent propagation velocity of the line under different sampling windows, which is between 1.95 × 10⁻⁶. 8 m / s to 1.98×10 8 Between m / s.

[0271] After decomposing the bidirectional reflection waveform, the joint inversion module extracts the local anomalous components of each joint at various frequencies and under various operating conditions. The model matching module performs waveform reconstruction under the normal assumptions of the target joint for the candidate joint at the 65m location and calculates the reconstruction residual.

[0272] Reference Figure 7 , Figure 7 The study presents the bidirectional observed reflection waveforms and spatial distribution of residual energy for the aforementioned 65m joint. Figure 7 In the diagram, (a) represents the initial time-domain reflection waveforms in the first and second detection directions, with the horizontal axis representing physical coordinates (unit: m) and the vertical axis representing the normalized reflection amplitude (dimensionless). From Figure 7 As can be seen in (a), there is a reflection peak at the 65m position, but it is superimposed with cable attenuation drift caused by temperature changes along the entire line and background noise.

[0273] Figure 7 Figure (b) shows the reconstructed residual energy distribution under the normal assumption of the faulty joint. The horizontal axis represents the physical coordinates (unit: m), and the vertical axis represents the local residual energy (expressed as normalized energy, dimensionless). Curve 1 in the figure corresponds to the residual energy distribution calculated in the first detection direction, with its peak coordinate appearing at 65.2 m; curve 2 corresponds to the residual energy distribution calculated in the second detection direction, with its peak coordinate appearing at 64.9 m. By comparison, the difference in the peak coordinates of the residual energy obtained from the two detection directions is 0.3 m, which is less than the bidirectional positioning consistency threshold of 0.6 m set by the system based on the initial calibration distribution. Therefore, the joint is determined to meet the bidirectional position consistency condition.

[0274] To verify the thermal state correlation conditions, the model matching module extracted the local residual energy feature sequence of the 65m joint. (Refer to...) Figure 8 , Figure 8 The correlation curve between the thermal excitation quantity and the local residual energy characteristic quantity is shown. Figure 8 The horizontal axis represents the sampling window number under the operating condition (dimensionless), and the vertical axis on the left represents the thermal excitation quantity (unit: A). 2 The right-hand ordinate represents the local residual energy characteristic quantity (dimensionless). The solid line marked with a circle represents the curve of thermal excitation quantity variation, and the dashed line marked with a square represents the curve of local residual energy characteristic quantity. Data comparison shows that the local residual energy characteristic quantity increases with the increase of thermal excitation quantity within the operating condition sampling window. The system calculates the thermal state correlation coefficient between the two to be 0.92, which is higher than the set correlation threshold of 0.45. Therefore, the candidate joint is determined to meet the thermal state correlation condition and the local contact impedance anomaly judgment rule.

[0275] The closed-loop verification module, centered at the 65m location, shielded the abnormal data window and regenerated the common propagation parameter set using the remaining observation dataset for secondary verification. The verification results showed that the candidate joint still met the local contact impedance anomaly judgment rule in the secondary judgment, but did not meet the line propagation parameter anomaly judgment rule. The closed-loop verification module output the detection results: the faulty joint coordinates were 65.05m (bidirectional average), the anomaly type was contact impedance anomaly, and the contact impedance anomaly characterization value was 16.4. Subsequent on-site investigation revealed oxidation traces and heat discoloration on the internal metal contacts of the joint near this coordinate, consistent with the detection results.

[0276] To verify the effectiveness of the present invention in extracting local anomaly components, a physical test platform with adjustable contact resistance connectors was constructed. The test platform utilized a 120m cable topology and connected test connectors in different degradation states at designated locations. The actual contact resistance of the test connectors was measured using a high-precision micro-resistance meter, with the resistance value distribution ranged from 1mΩ to 50mΩ. During the test, irradiance variations in a photovoltaic power station were simulated, and electrical transient interference components generated by inverter state switching were superimposed.

[0277] The same dataset was processed using both the traditional unidirectional time-domain reflectometry detection method and the method of this invention. (Refer to...) Figure 9 , Figure 9 The correlation between the anomaly characterization of contact impedance and the measured value of contact resistance in the experiment is shown. Figure 9 The horizontal axis represents the measured value of contact resistance (unit: mΩ), and the vertical axis represents the contact impedance anomaly characterization quantity (dimensionless). In the figure, the solid line marked with a diamond represents the bidirectional joint inversion detection result of the present invention, and the dashed line marked with a triangle represents the traditional unidirectional time-domain reflectometry detection result.

[0278] Depend on Figure 9 Observations show that when the measured contact resistance is between 1 mΩ and 10 mΩ, the traditional unidirectional time-domain reflectometry method does not separate the common background variation component and the thermal state-related reflection component of the string. Its characterization value fluctuates significantly with operating conditions within this low contact resistance range, leading to reduced differentiation between normal and slightly degraded joints. The bidirectional joint inversion detection result of this invention remains between 1.0 and 2.0 within this range, reducing common-mode signal interference caused by ambient temperature fluctuations.

[0279] When the measured contact resistance is greater than 15mΩ and increases, the contact impedance anomaly characterization quantity of this invention increases with the increase of the measured contact resistance, showing a monotonically increasing trend in the test data. The joint inversion module reduces the influence of electrical transient interference components, string common background variation components, and operating condition invariant reflection components on the extraction of local anomaly components through spatial sparsity conditions and bidirectional position consistency conditions, making the reconstruction residual more concentrated in reflecting the reflection changes near the candidate joint.

[0280] In multiple sets of experimental test data, when the measured contact resistance was greater than 15mΩ, the present invention was able to output contact impedance anomaly judgment results corresponding to the test joint position. Under simulated irradiation changes and inverter state switching conditions, no cases of misjudging contact impedance anomalies due to abnormal line propagation parameters were observed. In each set of tests, the deviation between the residual energy peak coordinates and the test joint installation coordinates was no greater than 0.4m. The above results demonstrate that the present invention can maintain the correspondence between the contact impedance anomaly judgment results and the joint position under multiple operating conditions.

[0281] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention.

Claims

1. A fault location and detection system for cable joints in photovoltaic power plants, characterized in that, include: The bidirectional observation module is used to acquire operating data and bidirectional reflection waveforms from multiple operating condition sampling windows. The bidirectional reflection waveforms include reflection waveforms acquired in the first detection direction and reflection waveforms acquired in the second detection direction. The operating data includes current data, voltage data, and inverter operating status information. The bidirectional observation module is also used to calculate thermal state excitation and electrical transient excitation to form a multi-operating condition bidirectional observation dataset. The parameter estimation module is used to determine the location of structural features based on the multi-condition bidirectional observation dataset and physical topology information, filter high-confidence structural feature locations, and calculate the common propagation parameter set. The joint inversion module is used to establish a frequency domain forward propagation model based on the multi-condition bidirectional observation dataset and the common propagation parameter set, and to extract state-related local reflection parameters. The model matching module is used to determine candidate joints based on the state-related local reflection parameters, and to calculate the reconstruction residual under the normal assumption of the target joint using the candidate joints as the target joints, thereby determining the contact impedance anomaly candidate joints that meet the local contact impedance anomaly judgment rules but do not meet the line propagation parameter anomaly judgment rules. The closed-loop verification module is used to shield the abnormal data window corresponding to the candidate joint with abnormal contact impedance to re-estimate the common propagation parameter set, and to verify the candidate joint with abnormal contact impedance based on the re-estimated common propagation parameter set and the complete multi-condition bidirectional observation dataset without abnormal data window shielding. The candidate joint with abnormal contact impedance that passes the verification is confirmed as a faulty joint, and the detection result including the faulty joint location and the abnormality type determination result is output.

2. The fault location and detection system for photovoltaic power station cable joints according to claim 1, characterized in that, The bidirectional observation module is used to acquire the current data of the photovoltaic string within the operating condition sampling window when calculating the thermal state excitation, introduce candidate thermal response parameters, and calculate the thermal state excitation using a discrete accumulation method with continuous integration; the thermal state excitation is calculated according to the following formula: ； In the formula, Indicates the first Each operating condition sampling window in candidate thermal response parameters Thermal excitation quantity under the following conditions; Indicates the first The start time of each working condition sampling window; Indicates the first The end time of each working condition sampling window; Indicates the operating time of the photovoltaic string Current data; Indicates the first One candidate thermal response parameter; This represents the integral variable at runtime.

3. The fault location and detection system for photovoltaic power station cable joints according to claim 1, characterized in that, The parameter estimation module is used to extract the reflection arrival time of each structural feature position in the first and second detection directions when establishing bidirectional distance closure constraints. The first detection direction is the direction in which a reflection test signal is injected from the first detection end of the cable under test and the reflected waveform is acquired at the first detection end; The second detection direction is the direction in which a reflection test signal is injected from the second detection end of the cable under test and the reflected waveform is acquired at the second detection end; The parameter estimation module is also used to combine the nominal propagation speed of the cable and the initial channel time offset to calculate the first initial unidirectional propagation distance from the structural feature location to the first detection end and the second initial unidirectional propagation distance to the second detection end. When the difference between the sum of the first initial unidirectional propagation distance and the second initial unidirectional propagation distance and the physical length of the cable under test is not greater than the bidirectional distance closure tolerance, the structural feature location is marked as a high-confidence structural feature location.

4. The fault location and detection system for photovoltaic power station cable joints according to claim 3, characterized in that, The parameter estimation module is used to establish and solve the arrival time residual objective function corresponding to the working condition sampling window when calculating the common propagation parameter set; the arrival time residual objective function is constructed according to the following formula: ； In the formula, Indicates the first The arrival time residual objective function value corresponding to each sampling window under each operating condition; Represents the set of indexes for high-confidence structural features; An index representing the location of a high-confidence structural feature; Indicates the first Within the sampling window of the first working condition Robust weights for high-confidence structural feature locations; and They represent the first The reflection arrival time of each structural feature location in the first and second detection directions; Indicates the first The physical coordinates of the locations of each structural feature; Indicates the physical length of the cable path being measured; Indicates the first The equivalent propagation speed of the line corresponding to each operating condition sampling window; and They represent the first Channel time offset in the first and second detection directions within each working condition sampling window.

5. The fault location and detection system for photovoltaic power station cable joints according to claim 1, characterized in that, The joint inversion module is used to convert the bidirectional reflection waveforms in the multi-condition bidirectional observation dataset into complex frequency domain reflection responses when constructing the frequency domain forward propagation model. Based on the common propagation parameter set, it determines the round-trip propagation response of each structural feature location in each detection direction and defines the frequency domain physical mapping relationship between the complex frequency domain reflection response and the round-trip propagation response, reference reflection component, state-related reflection component, electrical transient interference component, and noise component. The electrical transient interference component is determined by the frequency domain response coefficient corresponding to the electrical transient excitation and the detection direction.

6. The fault location and detection system for photovoltaic power station cable joints according to claim 5, characterized in that, The joint inversion module is used to decompose the state-related reflection component into thermal state-related reflection component, string common background variation component and local anomaly component, and establish a hierarchical joint inversion objective function with the reconstruction error of the frequency domain forward propagation model, the spatial sparsity of the fault-related component and the consistency of the string common background variation component as constraints. The joint inversion module is also used to iteratively solve the hierarchical joint inversion objective function by using an augmented Lagrangian approach that introduces a dual variable matrix, and to extract the state-related local reflection parameters corresponding to each structural feature position. The state-related local reflection parameters include local anomaly components and fault-related components.

7. The fault location and detection system for photovoltaic power station cable joints according to claim 1, characterized in that, The model matching module is used to set the fault-related component of the candidate joint to zero and retain the reference reflection component, normal reference thermal state response component and string common background change component of the candidate joint when using the candidate joint as the target joint and calculating the reconstruction residual under the normal assumption of the target joint. The model matching module is also used to retain the state-related reflection components of other structural feature locations besides the candidate joint, and to calculate the predicted reflection waveform under the normal assumption of the target joint by combining the common propagation parameter set and electrical transient interference components under the current operating condition sampling window. The model matching module is also used to calculate the waveform difference between the actual observed complex frequency domain reflection response and the predicted reflection waveform, and to use the waveform difference as the reconstruction residual corresponding to the candidate connector.

8. The fault location and detection system for photovoltaic power station cable joints according to claim 7, characterized in that, The model matching module is used to calculate the local residual energy based on the reconstructed residual, and to convert the residual energy peak coordinates based on the local residual energy peak. When the residual energy peak coordinate is located within the physical coordinate range of the candidate connector obtained by expanding the installation coordinate range of the candidate connector by a preset positioning tolerance, or within the range of multiple reflection paths that can be derived by the round-trip propagation distance between the candidate connector, adjacent impedance discontinuity positions, and line ends, and the difference between the residual energy peak coordinate calculated by the first detection direction and the second detection direction is not greater than the bidirectional positioning consistency threshold, the model matching module confirms that the candidate connector satisfies the residual position concentration condition and the bidirectional position consistency condition. The model matching module is also used to obtain the thermal state excitation variable sequence by removing the mean and normalizing the standard deviation of the thermal state excitation quantity sequence corresponding to the candidate joint, and to calculate the thermal state correlation coefficient between the local residual energy characteristic quantity sequence corresponding to the candidate joint and the thermal state excitation variable sequence. When the absolute value of the thermal state correlation coefficient is not less than the correlation threshold, the candidate joint is confirmed to meet the thermal state correlation condition and the local contact impedance anomaly judgment rule.

9. The fault location and detection system for photovoltaic power station cable joints according to claim 7, characterized in that, The model matching module is used to determine the set of affected structural feature positions after passing the candidate joint according to the line propagation parameter anomaly judgment rules, along the first detection direction and the second detection direction respectively. Based on the phase change, reflection arrival time change and amplitude change between the actual observed reflection response and the reconstructed reflection response, the module calculates the phase shift, time delay shift and attenuation change obtained from the amplitude change, and determines whether at least two adjacent affected structural feature positions have a phase shift, time delay shift or attenuation change that exceeds the sum of the mean and three times the standard deviation of the corresponding type of phase shift, time delay shift or attenuation change under the initial working condition in the same detection direction, and whether the corresponding changes have the same direction of change. When the judgment result is yes, the model matching module determines that the candidate joint meets the abnormal judgment rule of line propagation parameters, and does not output the candidate joint as a candidate joint with abnormal contact impedance.

10. The fault location and detection system for photovoltaic power station cable joints according to claim 1, characterized in that, The closed-loop verification module is used to merge the direct reflection window and the deducible multiple reflection path window of the candidate joint with abnormal contact impedance into an abnormal data window, and construct an abnormal data mask based on the abnormal data window. The closed-loop verification module is also used to use anomaly data mask to prevent sampling points within the anomaly data window from participating in the secondary estimation of common propagation parameters, and to generate a remaining observation dataset for re-estimating the common propagation parameter set. The detection results output by the closed-loop verification module include a contact impedance anomaly characterization value, which is calculated according to the following formula: ； In the formula, This indicates that the first faulty connector has been confirmed upon review. A characteristic quantity of contact impedance anomaly for a candidate joint with abnormal contact impedance. Indicates the first Fault-related component matrix at each structural feature location; Represents the Frobenius norm of a matrix; This indicates the operation of finding the maximum value. Indicates the first Reference reflection component matrix at each structural feature location; Indicates the number of sampling windows for operating conditions; Indicates the number of discrete frequency points; This represents the noise scale estimate of the reconstructed residual.

Images