Cable defect identification method based on robust random segmentation forest algorithm

Abstract

The invention discloses a cable defect identification method based on a robust random segmentation forest algorithm, and belongs to the field of cable defect measurement, and the method comprises the following steps: 1, connecting all structures of a measurement device; 2, setting measurement parameters according to the actual condition of the cable, wherein the measurement parameters comprise a sweep frequency range, a point number, power and a measurement target; step 3, performing measurement to obtain a reflection coefficient spectrum of a cable test end, adjusting a frequency range, and processing reflection coefficient spectrum data through inverse fast Fourier transform (IFFT) to obtain a positioning oscillogram of a spatial domain; and step 4, extracting characteristic peaks in the positioning oscillogram through a robust random segmentation forest anomaly detection algorithm to obtain an anomaly score of each data point. On the basis that the FDR method is adopted to test the cable, the positioning data are processed in combination with the robust random segmentation forest algorithm, and accurate identification and positioning of defects are achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of cable defect measurement, and in particular to a cable defect recognition method based on a robust random segmentation forest algorithm. Background Art

[0002] Cross-linked polyethylene cables in power transmission and distribution networks are prone to local defects such as electrical and water dendrites due to their complex operating environment, temperature, humidity, mechanical force, and chemical corrosion. In addition, flaws in the cable manufacturing and installation process may also cause local defects, which in turn cause partial discharge and lead to insulation performance degradation. Therefore, cable operation status detection is crucial, especially the accurate identification and location of cable faults and defects.

[0003] At present, the mainstream cable defect location technology is based on frequency domain reflectometry (FDR) technology. By sending a sweep test signal, a reflected signal with the same frequency as the transmitted signal but delayed in time will be generated at the location where the cable characteristic impedance does not match. These signals are analyzed by Fourier transform, and the distance to the line obstacle point is calculated by measuring the frequency of the peak of the reflected signal. However, the existing frequency domain reflectometry data analysis method lacks a unified standard for fault or defect judgment. It only relies on the protrusion of the peak of the positioning waveform to judge whether the cable has a fault or defect at that location, which is very easy to cause misjudgment due to noise and signal attenuation. Therefore, based on the FDR method, a cable defect recognition method based on the robust random segmentation forest (RRCF) algorithm is proposed. Summary of the invention

[0004] The purpose of the present invention is to provide a cable defect identification method based on a robust random segmentation forest algorithm. On the basis of testing the cable using the FDR method, the positioning data is processed in combination with the robust random segmentation forest (RRCF) algorithm to achieve accurate identification and positioning of defects.

[0005] To achieve the above object, the present invention provides a cable defect identification method based on a robust random segmentation forest algorithm, comprising the following steps:

[0006] Step 1, connecting various structures of the measuring device;

[0007] Step 2: Set the measurement parameters according to the actual situation of the cable, including sweep frequency range, number of points, power, and measurement target;

[0008] Step 3, measuring and obtaining the reflection coefficient spectrum of the cable test end, adjusting the frequency range, and processing the reflection coefficient spectrum data by inverse fast Fourier transform IFFT to obtain a positioning waveform diagram in the spatial domain;

[0009] Step 4: The robust random segmentation forest (RRCF) anomaly detection algorithm is used to extract the characteristic peaks in the positioning waveform and obtain the anomaly score of each data point.

[0010] Preferably, the measuring device comprises a PC, a vector network analyzer VNA, a test fixture, and a cable to be tested. The vector network analyzer VNA is used to generate a low voltage signal, and the test fixture is used to connect the VNA and the cable.

[0011] Preferably, step 1 comprises the following steps:

[0012] Step 11, connect the PC to the VNA and open the corresponding test software;

[0013] Step 12. Connect one end of the test fixture to the VNA.

[0014] Preferably, step 2 comprises the following steps:

[0015] Step 22, set the measurement parameters on the VNA according to the actual situation of the cable;

[0016] Step 23, performing open circuit automatic calibration on the VNA;

[0017] Step 24, connect the core alligator clip and shield alligator clip at the other end of the test fixture to the core and shield of the cable under test respectively.

[0018] Preferably, in step 3, the frequency range of the VNA is adjusted according to the reflection coefficient display effect on the VNA or the PC;

[0019] Combining the calculation formula of distributed capacitance and characteristic impedance, the cable with defective section is analyzed, and the reflection coefficient spectrum RCS at the right end of the defect, the left end of the defect and the head end of the cable is deduced as follows:

[0020]

[0021] Where L is the cable length, ω is the angular frequency, and Z L is the load impedance, is the input impedance spectrum at the right end of the defect, is the input impedance spectrum at the left end of the defect, x1 is the distance from the left end of the defect to the head end, d1 is the length of the defect segment, γ0 and γ s are the propagation coefficients of the intact and defective cable segments, Z0 and Z s are the characteristic impedances of the intact segment and the defective segment respectively;

[0022] From the complex expression of the propagation constant, we can expand equation (3) according to Euler's formula to obtain

[0023]

[0024] In the formula, f is the frequency, v is the electromagnetic wave velocity, α is the attenuation coefficient, and j is the imaginary unit;

[0025] By performing Fourier transform on the collected head-end RCS and combining it with the wave velocity, the RCS in the frequency domain is converted into a positioning waveform in the space domain, and the equivalent frequency corresponds to time.

[0026] Preferably, in step 4, the RRCF anomaly detection algorithm is used to extract characteristic peaks in the positioning graph. For abnormal points, they are defined from the model complexity. As the complexity of the model increases with a certain data point, the point is an abnormal point; the amplitude A of the positioning data is selected as input, and a data is randomly selected as a cutting point p within the data range of the amplitude. According to the cutting point p, the data is divided into two groups: a left subtree and a right subtree; the above process is repeated for each child node until the stop condition is met and the tree is constructed; the abnormal score of each data point is finally obtained by averaging the path lengths of all trees in the forest, and points with small abnormal scores represent normal values, and points with large abnormal scores are abnormal values.

[0027] Preferably, the calculation process of the anomaly score is as follows:

[0028] Locate the leaf node corresponding to the target point p in the tree T, and traverse upward from the leaf node layer by layer to the root node; at each layer, calculate the impact of removing point p on the balance of the left and right subtrees of the current node: if p belongs to the left subtree, the impact factor is the size of the right subtree divided by the size of the left subtree; if p belongs to the right subtree, it is the size of the left subtree divided by the size of the right subtree; record the impact factors of all layers during the traversal process, and finally use the largest impact factor as the abnormality score of point p; the higher the score, the greater the damage to the tree structure caused by point p, and the stronger its abnormality.

[0029] Therefore, the present invention adopts the above-mentioned cable defect identification method based on robust random segmentation forest algorithm, which has the following beneficial effects:

[0030] (1) The present invention adopts a self-made test fixture, firstly leads out the shield and core of the coaxial cable through double alligator clips, and then allows the alligator clips to clamp the copper tape to extend the test distance, which increases the signal injection rate to a certain extent and weakens the influence of impedance mismatch at the connection on the signal;

[0031] (2) Based on the FDR method for testing the cable, the present invention combines the RRCF algorithm to process the positioning data, which can eliminate the influence of interference peaks and realize accurate identification and positioning of defects.

[0032] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1The test principle diagram of the FDR method of an embodiment of the present invention;

[0034] Figure 2 This is a result diagram of numerical simulation of a cable of model YJV-6 / 10kV according to an embodiment of the present invention;

[0035] Figure 3 This is a result diagram of actual measurement of a 823m long cross-linked polyethylene cable (cable model: YJV-6 / 10kV) according to an embodiment of the present invention;

[0036] Figure 4 This is a schematic diagram of the structure of a self-made test fixture according to an embodiment of the present invention;

[0037] Figure 5 A pseudo code diagram for constructing a single robust random partitioning tree (RRCT) according to an embodiment of the present invention. DETAILED DESCRIPTION

[0038] In order to make the purpose, technical scheme and advantages disclosed in the embodiments of the present invention clearer, the embodiments of the present invention are further described in detail in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the embodiments of the present invention and are not intended to limit the embodiments of the present invention. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application. Examples of the embodiments are shown in the accompanying drawings, where the same or similar reference numerals throughout represent the same or similar elements or elements with the same or similar functions.

[0039] It should be noted that the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or server that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units that are not explicitly listed or inherent to these processes, methods, products or devices.

[0040] Like reference numerals and letters denote similar items in the following drawings, and thus, once an item is defined in one drawing, further definition and explanation thereof is not required in subsequent drawings.

[0041] In the description of the present invention, it should be noted that the terms "upper", "lower", "inside", "outside", etc. indicate directions or positional relationships based on the directions or positional relationships shown in the accompanying drawings, or are directions or positional relationships in which the product of the invention is usually placed when in use. They are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific direction, be constructed and operated in a specific direction, and therefore should not be understood as a limitation on the present invention.

[0042] In the description of the present invention, it is also necessary to explain that, unless otherwise clearly specified and limited, the terms "setting", "installation" and "connection" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be an indirect connection through an intermediate medium, or it can be the internal communication of two components. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0043] Example

[0044] The present invention provides a cable defect recognition method based on a robust random segmentation forest algorithm, comprising the following steps:

[0045] Step 1: Connect various structures of the measuring device. The measuring device includes a PC, a vector network analyzer VNA, a test fixture, and a cable to be tested. The vector network analyzer VNA is used to generate a low voltage signal, and the test fixture is used to connect the VNA and the cable.

[0046] Step 1 includes the following steps:

[0047] Step 11, connect the PC to the VNA and open the corresponding test software;

[0048] Step 12, connect one end of the homemade test fixture to the VNA. Figure 4 .

[0049] Step 2: Set the measurement parameters according to the actual cable conditions. The measurement parameters include sweep frequency range, number of points, power, and measurement target.

[0050] Step 2 includes the following steps:

[0051] Step 22, set the measurement parameters on the VNA according to the actual situation of the cable;

[0052] Step 23, performing open circuit automatic calibration on the VNA;

[0053] Step 24, connect the core alligator clip and the shield alligator clip at the other end of the test fixture to the core and the shield (or grounding terminal) of the cable under test respectively.

[0054] Step 3, measure and obtain the reflection coefficient spectrum of the cable test end, adjust the frequency range, and process the reflection coefficient spectrum data through inverse fast Fourier transform IFFT to obtain a positioning waveform diagram in the spatial domain.

[0055] In step 3, when the cable has local defects such as insulation aging, its insulation performance will also change. The more serious the aging state, the greater the change in the complex dielectric constant of the insulation material. From the calculation formula of distributed capacitance and characteristic impedance, it can be seen that the change in the relative dielectric constant of the insulation causes the change in the cable capacitance, which in turn causes the change in characteristic impedance. When the signal propagates in the cable, refraction and reflection of the signal will occur at the discontinuity of the characteristic impedance.

[0056] Adjust the frequency range of the VNA according to the reflection coefficient display effect on the VNA or PC;

[0057] Combining the calculation formula of distributed capacitance and characteristic impedance, the cable with defective section is analyzed, and the reflection coefficient spectrum RCS at the right end of the defect, the left end of the defect and the head end of the cable is deduced as follows:

[0058]

[0059] Where L is the cable length, ω is the angular frequency, and Z L is the load impedance, is the input impedance spectrum at the right end of the defect, is the input impedance spectrum at the left end of the defect, x1 is the distance from the left end of the defect to the head end, d1 is the length of the defect segment, γ0 and γ s are the propagation coefficients of the intact and defective cable segments, Z0 and Z s are the characteristic impedances of the intact segment and the defective segment respectively;

[0060] From the complex expression of the propagation coefficient, formula (3) is expanded according to the Euler formula to obtain

[0061]

[0062] In the formula, f is the frequency, v is the electromagnetic wave velocity, α is the attenuation coefficient, and j is the imaginary unit;

[0063] It can be seen from formula (5) that the RCS at the head end of the cable is obtained by superimposing exponentially decaying complex sinusoidal signals of multiple frequencies. In the intact state, the RCS at the head end of the cable shows a pseudo-periodic distribution with frequency changes, and the equivalent frequency f'=2l / v. When the cable contains a defective section at a distance of x1 from the head end, other periodic components will be introduced, including RCS of f'=2x1 / v. The position of the left end of the defective section can be located by x1=f'v / 2.

[0064] By performing Fourier transform on the collected head-end RCS and combining it with the wave velocity, the RCS in the frequency domain can be converted into a positioning diagram in the space domain, and the equivalent frequency corresponds to time. However, when the test object is a medium or high voltage cable, the defect peak in the positioning diagram is not prominent due to the influence of signal attenuation, so the manual judgment of the defect is often subjective, resulting in misjudgment.

[0065] Step 4: The robust random segmentation forest (RRCF) anomaly detection algorithm is used to extract the characteristic peaks in the positioning waveform and obtain the anomaly score of each data point.

[0066] In step 4, the RRCF anomaly detection algorithm is used to extract characteristic peaks in the positioning map to accurately identify defects. For abnormal points, the algorithm is defined from the model complexity. If the complexity of the model increases with a certain data point, then the point is an abnormal point. First, the characteristics of the positioning data are taken as input. This patent selects the amplitude A of the positioning data as input, and then randomly selects a data within the data range of the amplitude as the cutting point p. According to the cutting point p, the data is divided into two groups: the left subtree and the right subtree. Repeat the above process for each child node until the stopping condition is met. The above is the construction of the tree. A complete RRCF contains multiple randomly cut trees. Each tree reflects the local characteristics of the data distribution, and the collection of forests can capture the global distribution characteristics. The abnormal score of each data point is finally obtained by averaging the path length of all trees in the forest. Points with smaller abnormal scores represent normal values, and larger points are abnormal values.

[0067] The calculation process of the anomaly score is as follows: first locate the leaf node corresponding to the target point p in the tree T, and then traverse upward from the leaf node layer by layer to the root node. At each layer, calculate the impact of removing point p on the balance of the left and right subtrees of the current node: if p belongs to the left subtree, the impact factor is the size of the right subtree divided by the size of the left subtree; if p belongs to the right subtree, it is the size of the left subtree divided by the size of the right subtree. During the traversal process, the impact factors of all layers are recorded, and finally the largest impact factor is used as the anomaly score of point p. The higher the score, the greater the damage to the tree structure caused by point p, and thus the stronger its anomaly.

[0068] Figure 1 This is the test principle diagram of the FDR method. Its working principle is to inject a low-voltage swept frequency signal at one end of the cable and obtain the reflection coefficient spectrum (RCS) or broadband impedance spectrum (BIS) at the test end. Then, RCS and BIS can be used to locate discontinuities through a series of data processing methods such as inverse fast Fourier transform (IFFT) and windowing. Finally, combined with the propagation speed of the signal in the cable, the peak indicating the fault location is obtained.

[0069] Figure 2 This is the result of numerical simulation of a cable model YJV-6 / 10kV. The length is set to 823m, two defects are set at 420m and 570m, the sweep range is set to 30kHz~30MHz, the number of points is 3000, and the end is open-circuited. The real part of the reflection coefficient is obtained as follows Figure 2 As shown in (a), the simulated wave velocity is calculated according to the actual structural parameters of the cable, and the positioning diagram is obtained as follows Figure 2 (b) as shown.

[0070] Figure 3 This is the result of an actual test on a 823m long cross-linked polyethylene cable (cable model: YJV-6 / 10kV). Two defects were set before the test. The copper foil was stripped from the semi-conductive part at 420m from the cable head end and then part of the copper foil was reconnected. The installation of the intermediate joint was simulated at 620m. The original positioning diagram is as follows: Figure 3 As shown in (a), the location map obtained by extracting abnormal peaks using the RRCF algorithm is as follows: Figure 3 As shown in (b), the peaks of the two defects can be clearly seen, eliminating the influence of the interference peak.

[0071] Figure 5 The pseudo code for constructing a single robust random partitioning tree (RRCT) is given, and RRCF is the collection of RRCTs.

[0072] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that they can still modify or replace the technical solution of the present invention with equivalents, and these modifications or equivalent replacements cannot cause the modified technical solution to deviate from the spirit and scope of the technical solution of the present invention.

Claims

1. A cable defect recognition method based on a robust random segmentation forest algorithm, characterized in that: The following steps are involved: Step 1, connecting various structures of the measuring device; Step 2: Set the measurement parameters according to the actual situation of the cable, including sweep frequency range, number of points, power, and measurement target; Step 3, measuring and obtaining the reflection coefficient spectrum of the cable test end, adjusting the frequency range, and processing the reflection coefficient spectrum data by inverse fast Fourier transform IFFT to obtain a positioning waveform diagram in the spatial domain; Step 4: The robust random segmentation forest (RRCF) anomaly detection algorithm is used to extract the characteristic peaks in the positioning waveform and obtain the anomaly score of each data point.

2. The cable defect identification method based on the robust random segmentation forest algorithm according to claim 1 is characterized in that: The measuring device includes a PC, a vector network analyzer VNA, a test fixture, and a cable to be tested. The vector network analyzer VNA is used to generate a low voltage signal, and the test fixture is used to connect the VNA and the cable.

3. The cable defect identification method based on the robust random segmentation forest algorithm according to claim 2 is characterized in that: Step 1 includes the following steps: Step 11, connect the PC to the VNA and open the corresponding test software; Step 12. Connect one end of the test fixture to the VNA.

4. The cable defect identification method based on the robust random segmentation forest algorithm according to claim 3 is characterized in that: Step 2 includes the following steps: Step 22, set the measurement parameters on the VNA according to the actual situation of the cable; Step 23, performing open circuit automatic calibration on the VNA; Step 24, connect the core alligator clip and shield alligator clip at the other end of the test fixture to the core and shield of the cable under test respectively.

5. The cable defect identification method based on the robust random segmentation forest algorithm according to claim 4 is characterized in that: In step 3, adjust the frequency range of the VNA according to the reflection coefficient display effect on the VNA or PC; Combining the calculation formula of distributed capacitance and characteristic impedance, the cable with defective section is analyzed, and the reflection coefficient spectrum RCS at the right end of the defect, the left end of the defect and the head end of the cable is deduced as follows: Where L is the cable length, ω is the angular frequency, and Z L is the load impedance, is the input impedance spectrum at the right end of the defect, is the input impedance spectrum at the left end of the defect, x1 is the distance from the left end of the defect to the head end, d1 is the length of the defect segment, γ0 and γ s are the propagation coefficients of the intact and defective cable segments, Z0 and Z s are the characteristic impedances of the intact segment and the defective segment respectively; From the complex expression of the propagation coefficient, formula (3) is expanded according to the Euler formula to obtain In the formula, f is the frequency, v is the electromagnetic wave velocity, α is the attenuation coefficient, and j is the imaginary unit; By performing Fourier transform on the collected head-end RCS and combining it with the wave velocity, the RCS in the frequency domain is converted into a positioning waveform in the space domain, and the equivalent frequency corresponds to time.

6. The cable defect identification method based on robust random segmentation forest algorithm according to claim 5 is characterized in that: In step 4, the RRCF anomaly detection algorithm is used to extract characteristic peaks in the positioning graph. For abnormal points, they are defined from the model complexity. When the complexity of the model increases with a certain data point, the point is an abnormal point. The amplitude A of the positioning data is selected as input, and a data is randomly selected as the cutting point p within the data range of the amplitude. According to the cutting point p, the data is divided into two groups: the left subtree and the right subtree. The above process is repeated for each child node until the stopping condition is met and the tree is constructed. The abnormal score of each data point is finally obtained by averaging the path lengths of all trees in the forest. Points with small abnormal scores represent normal values, and points with large abnormal scores are abnormal values.

7. The cable defect identification method based on robust random segmentation forest algorithm according to claim 6 is characterized in that: The calculation process of anomaly score is as follows: Locate the leaf node corresponding to the target point p in the tree T, and traverse upward from the leaf node layer by layer to the root node; at each layer, calculate the impact of removing point p on the balance of the left and right subtrees of the current node: if p belongs to the left subtree, the impact factor is the size of the right subtree divided by the size of the left subtree; if p belongs to the right subtree, it is the size of the left subtree divided by the size of the right subtree; record the impact factors of all layers during the traversal process, and finally use the largest impact factor as the abnormality score of point p; the higher the score, the greater the damage to the tree structure caused by point p, and the stronger its abnormality.

Images