CN112595913B - Cable local aging detection method and detection device - Google Patents
Cable local aging detection method and detection device Download PDFInfo
- Publication number
- CN112595913B CN112595913B CN202011431893.7A CN202011431893A CN112595913B CN 112595913 B CN112595913 B CN 112595913B CN 202011431893 A CN202011431893 A CN 202011431893A CN 112595913 B CN112595913 B CN 112595913B
- Authority
- CN
- China
- Prior art keywords
- aging
- cable
- cnn
- neural network
- transfer function
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
- G01R31/00—Arrangements for testing electric properties; Arrangements for locating electric faults; Arrangements for electrical testing characterised by what is being tested not provided for elsewhere
- G01R31/003—Environmental or reliability tests
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
- G01R31/00—Arrangements for testing electric properties; Arrangements for locating electric faults; Arrangements for electrical testing characterised by what is being tested not provided for elsewhere
- G01R31/12—Testing dielectric strength or breakdown voltage ; Testing or monitoring effectiveness or level of insulation, e.g. of a cable or of an apparatus, for example using partial discharge measurements; Electrostatic testing
- G01R31/1227—Testing dielectric strength or breakdown voltage ; Testing or monitoring effectiveness or level of insulation, e.g. of a cable or of an apparatus, for example using partial discharge measurements; Electrostatic testing of components, parts or materials
- G01R31/1263—Testing dielectric strength or breakdown voltage ; Testing or monitoring effectiveness or level of insulation, e.g. of a cable or of an apparatus, for example using partial discharge measurements; Electrostatic testing of components, parts or materials of solid or fluid materials, e.g. insulation films, bulk material; of semiconductors or LV electronic components or parts; of cable, line or wire insulation
- G01R31/1272—Testing dielectric strength or breakdown voltage ; Testing or monitoring effectiveness or level of insulation, e.g. of a cable or of an apparatus, for example using partial discharge measurements; Electrostatic testing of components, parts or materials of solid or fluid materials, e.g. insulation films, bulk material; of semiconductors or LV electronic components or parts; of cable, line or wire insulation of cable, line or wire insulation, e.g. using partial discharge measurements
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
- G06F17/10—Complex mathematical operations
- G06F17/15—Correlation function computation including computation of convolution operations
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
- G06F17/10—Complex mathematical operations
- G06F17/16—Matrix or vector computation, e.g. matrix-matrix or matrix-vector multiplication, matrix factorization
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Abstract
The invention relates to a method and a device for detecting local aging of a cable, which are used for detecting the local aging of the cable and estimating an aging position and an aging degree by measuring a transfer function of the cable in a frequency range, according to the amplitude of the transfer function, based on a deep learning algorithm and by using a detection deep neural network D-CNN and a positioning deep neural network L-CNN. The transfer function is not influenced by phase angle errors caused by real-time communication in the system, so that online detection is easy to realize. The detection method provided by the invention is suitable for detection of various local aging modes, and can realize estimation of the aging degree, thereby providing a basis for subsequent preventive maintenance; meanwhile, the problem of local aging positioning under the variable load condition is solved based on transfer learning, the precision of the detection method under the variable load condition can be improved based on parameter fine adjustment, and the calculation cost and the calculation time are saved.
Description
Technical Field
The invention relates to the technical field of detection of power system equipment, in particular to a method and a device for detecting local aging of a cable.
Background
Underground cables in a power distribution network are often aged due to wetting or overheating, insulation breakdown and short-circuit faults are finally caused, safe and reliable operation of a power system is affected, and huge economic loss is brought. Especially for the urban distribution network with gradually enlarged scale at present, the running environment of underground cables is complex, and a branch cable can be overheated due to long-time heavy-load current running to generate sectional local aging or generate local hot-point aging due to certain insulation defects in the production and processing processes. However, most studies are focused on locating short-circuit faults of cables, and the short-circuit faults belong to a passive protection mode, and it is still difficult to avoid power failure in partial areas. The local aging of the underground cable is positioned, so that active preventive maintenance is realized, system faults can be fundamentally avoided, the reliability of the power distribution network is improved, and the method has great industrial application value.
At present, a method based on broadband impedance spectroscopy (hereinafter referred to as "BIS") is mainly adopted for positioning the local aging of the cable, and a schematic diagram of a method for detecting the local aging of the cable based on BIS is shown in fig. 1. Injecting a sweep frequency signal at one end of the cable, and measuring the broadband impedance spectrum at the cable input end. The impedance spectrum in the frequency domain is transformed into an impedance distribution spectrum in the time domain (spatially) using an inverse fourier transform (iFFT) on the measured impedance spectrum. Peaks in the spectrum are identified to enable localization of the aging section. The method has the following defects:
(1) only off-line detection can be realized, and the application of the method in an actual system is limited; (2) the method depends on the obvious change of the impedance of the end part of the aging section of the locally aged cable, so that the monitoring sensitivity is low, and the method is difficult to be applied to the aging condition with smooth impedance change; (3) the method can only determine the impedance change point, and the aging degree is difficult to estimate. And is also difficult to apply to end-aged conditions.
Disclosure of Invention
Based on the above situation of the prior art, the invention aims to estimate the aging position and the aging degree of a cable based on a deep learning algorithm so as to solve the problem of positioning the local aging of the underground cable.
In order to achieve the above object, according to one aspect of the present invention, there is provided a method for detecting local degradation of a cable, comprising the steps of:
measuring a transfer function of a cable to be tested according to a preset time interval, extracting characteristic information in a transfer function vector, inputting a trained detection depth neural network D-CNN, judging whether the cable to be tested has local aging or not by using the detection depth neural network D-CNN, and if so, carrying out the next step; if not, returning;
extracting characteristic information in the transfer function vector, inputting the trained positioning depth neural network L-CNN, and performing aging positioning and aging degree estimation by using the positioning depth neural network L-CNN.
Further, the cable local aging detection is realized by using the relationship between the change of the transfer function of the cable and the cable local aging, wherein the transfer function of the cable comprises the following formula:
wherein L is b Indicating the starting position of the cable aging section, L e Denotes the end position of the cable aging section, L denotes the total cable length, H 0 Representing the transfer function of the intact section of cable,
indicates for a bit located at x i The length of (d) is the transfer function of a small segment of Δ x.
Further, the utilized detection deep neural network D-CNN and the positioning deep neural network L-CNN both comprise a sparse self-encoder and a convolutional neural network.
Further, the determining whether the cable to be tested has local aging by using the detection deep neural network D-CNN includes:
the D-CNN outputs a 1 xN vector, wherein N represents that the cable to be tested is divided into N sections, and each element in the vector represents the possibility of aging of each section of the cable to be tested.
Furthermore, for each element in the vector output by the D-CNN, if the value of each element is greater than 0.5, the segment corresponding to the element is considered to be aged, and if the value of each element is less than or equal to 0.5, the segment corresponding to the element is considered to be intact.
Further, the aging localization and aging degree estimation by using the localization depth neural network L-CNN includes:
and the positioning depth neural network L-CNN outputs a 1 x 3 vector which respectively represents the initial position of the aging section of the cable to be detected, the termination position of the aging section of the cable to be detected and the aging degree of the cable to be detected.
Further, when the load changes, the weight coefficient matrix between the full connection layer and the output in the convolutional neural network is adjusted based on a back propagation algorithm.
According to another aspect of the present invention, there is provided a cable local aging detection apparatus, including a transfer function measurement module, a local aging detection module, and a local aging localization module; wherein the content of the first and second substances,
the transfer function measuring module is used for measuring the transfer function of the cable to be measured;
the local aging detection module extracts the characteristic information in the transfer function vector and inputs the characteristic information into a trained detection depth neural network D-CNN, and judges whether the cable to be detected has local aging or not by using the detection depth neural network D-CNN;
and the local aging positioning module extracts the characteristic information in the transfer function vector and inputs the characteristic information into the trained positioning depth neural network L-CNN, and when judging that the cable to be tested has local aging, the positioning depth neural network L-CNN is used for performing aging positioning and aging degree estimation.
Further, the local aging detection module judges whether the cable to be detected has local aging according to a 1 × N vector output by the detection deep neural network D-CNN, where N represents that the cable to be detected is divided into N sections, and each element in the vector represents the possibility of aging of each section of the cable to be detected.
Further, the local aging positioning module performs aging positioning and aging degree estimation according to a 1 × 3 vector output by the positioning depth neural network L-CNN, where the 1 × 3 vector represents an initial position of an aging section of the cable to be measured, an end position of the aging section of the cable to be measured, and an aging degree of the cable to be measured, respectively.
In summary, the invention provides a method and a device for detecting local aging of a cable, which detect local aging of the cable by measuring a transfer function of the cable in a frequency range, according to an amplitude of the transfer function, based on a deep learning algorithm, and by using a detection deep neural network D-CNN and a positioning deep neural network L-CNN, and estimate an aging position and an aging degree. Compared with the detection method in the prior art, the method provided by the invention has the following beneficial technical effects:
(1) the transfer function amplitude is used, so that the influence of phase angle errors caused by real-time communication in a system can be avoided, and the method is easy to realize in an actual system, thereby being easy to realize on-line detection.
(2) The method can be suitable for detecting various local aging modes, including uniform aging, non-uniform aging, end aging, smooth aging and the like.
(3) The aging degree can be estimated, so that a basis is provided for subsequent preventive maintenance.
(4) The method solves the problem of local aging positioning under the condition of variable load based on transfer learning, can improve the precision of the detection method under the condition of variable load based on parameter fine adjustment, and saves the calculation cost and the calculation time.
Drawings
FIG. 1 is a schematic diagram of a BIS-based cable local aging detection method;
FIG. 2 is a flow chart of an embodiment of the method for detecting local aging of a cable according to the present invention;
FIG. 3 is a schematic diagram of local aging of a cable;
FIG. 4 is a graph comparing the transfer function of an intact cable with a locally aged cable;
FIG. 5 is a schematic structural diagram of a deep probing neural network D-CNN and a deep positioning neural network L-CNN according to the present invention;
FIG. 6 is a diagram illustrating the detection results of detecting the deep neural network D-CNN, and FIGS. 6(a) -6(D) are the detection results of four typical samples in the test set, respectively;
FIG. 7 is a comparison graph of the detection results of the cable local aging detection method and the BIS method according to the present invention for different aging situations, wherein FIG. 7(a) shows uniform aging, FIG. 7(b) shows end aging, and FIG. 7(c) shows non-uniform aging;
fig. 8 is a schematic diagram of the detection result of the cable local aging detection method under the condition of smooth aging, fig. 8(a) is a typical example with a relatively severe aging degree, and fig. 8(b) is a typical example with a relatively light aging degree;
FIG. 9 is a diagram illustrating the result of the BIS method under the condition of smooth aging, wherein FIG. 9(a) is a typical example with a more severe aging degree, and FIG. 9(b) is a typical example with a less aging degree;
fig. 10 is a block diagram showing the structure of the cable partial deterioration detecting apparatus of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention will be described in further detail with reference to the accompanying drawings in conjunction with the following detailed description. It should be understood that the description is intended to be exemplary only, and is not intended to limit the scope of the present invention. Moreover, in the following description, descriptions of well-known structures and techniques are omitted so as to not unnecessarily obscure the concepts of the present invention.
The technical scheme of the invention is explained in detail in the following with reference to the attached drawings. According to an embodiment of the present invention, a method for detecting local degradation of a cable is provided, and a flowchart for implementing the method is shown in fig. 2. The method for detecting the local aging of the cable provided by the embodiment has a flow divided into two parts, namely local aging detection and local aging positioning. Since cable aging is a slow process, local aging detection may be performed at intervals (e.g., 3 months). In the local aging detection process, firstly, the transfer function of the cable to be monitored on a frequency band of 3-500 kHz is measured, and characteristic information in a transfer function vector is input into a trained detection depth neural network D-CNN and a positioning depth neural network L-CNN. And detecting and preliminarily positioning the local aging according to the trained detection deep neural network D-CNN. If the D-CNN does not find that the aging section exists in the system, recording the state of the cable system as 'healthy', returning the flow to the beginning, and waiting for the next detection; and if the D-CNN finds that the aging section exists in the system, further using the trained positioning deep neural network L-CNN to position aging and estimate the aging degree. And finally, comprehensively analyzing the detection and positioning results of the D-CNN and the L-CNN, and giving an alarm to maintenance personnel when serious aging and hot spot aging occur in the judgment system.
Relationship between cable system transfer function and local aging
The most significant characteristics of cable aging are a reduction in insulation resistance and an increase in insulation capacitance. The change in insulation capacitance is more pronounced in comparison to the insulation resistance in the frequency band considered by the present invention, and therefore the effect of the insulation resistance is not considered in the following analysis. The partial aging of the cable is schematically shown in FIG. 3, wherein R 0 Representing a cable conductor resistance matrix, L 0 Representing the inductance matrix of the conductors of the cable, C 0 Representing an insulated capacitor matrix. Representing the capacitance of the aging section as a function of position C 1 (x) To enable the depiction of different aging patterns, Z L Representing the impedance matrix of the system.
Due to the capacitance change in the aging section of the cable, the transfer function of the cable system will also change. The transfer function of the system can be measured using a carrier communication module (PLM) or a high precision current transformer (HFCT). A carrier communication module (PLM) or a high-precision current transformer (HFCT) is used for injecting a high-frequency signal at one end of the cable, and the other end of the cable is measured by the same device to obtain the transfer function of the system, so that the method is an online measurement mode. In order to solve the transfer function under various aging modes, the method comprises the following stepsThe aging section of the cable is divided into a number of small sections, and the capacitance value on each small section is considered to be constant. For being located at x i The transfer function of a small segment of length Δ x can be expressed as follows according to transmission line theory:
its equivalent end admittance can be expressed as follows:
wherein, the matrix U represents a unit matrix, and the matrix T is a characteristic vector matrix of the matrix YZ. Y ═ j ω C 1 (x i ) And Z is R 0 +jωL 0 . ω represents the angular frequency of the signal used and can be expressed as 2 π f, f being the frequency of the signal used. τ is the transmission coefficient of the cable, as the square root of the eigenvalues of the matrix YZ. Z is a linear or branched member C Can be represented as Y -1 TτT -1 And Y is C Can be represented as Z C -1 。ρ LI The reflection coefficient of the system can be expressed as follows:
wherein the content of the first and second substances,
is shown at x i+1 The equivalent input end admittance of the small segment. Thus, the transfer function of the entire cable system can be expressed as:
wherein L is b Indicating the starting position of the cable aging section, L e Indicating the termination point of the aged section of the cable. L represents the total cable length. H 0 Representing the transfer function of the intact section of cable. Because of the intact cable insulation capacitance C 0 Consistent, it can therefore be solved directly using transmission line theory. It should also be noted that the transfer functions of the good section at the head end of the line and the good section at the tail end of the line are related to the equivalent input admittance and the system load admittance at the start point of the aging section, respectively, and thus are not completely identical.
The following examples illustrate the change in the transfer function of a cable system in the case of local aging: for a typical medium voltage cable, the parameters are shown in table 1 below.
TABLE 1 typical Medium Voltage Cable parameters
For a 10km long cable, assuming local aging of the insulation between 2km and 5km, the insulation capacitance (relative permittivity) becomes 1.2 times that of the intact case, and the comparison of the transfer function with that of the intact case is shown in fig. 4. As can be seen from fig. 4, the position and magnitude of the peak occurrence of the transfer function vary, and therefore localized aging can be located using the cable system transfer function variation.
Detecting deep neural network D-CNN and positioning deep neural network L-CNN
In the detection method provided in this embodiment, the deep neural network D-CNN for detecting local aging and the deep neural network L-CNN for locating local aging are further adopted for detection, and the schematic structural diagrams of the two networks are shown in fig. 5. In the detection method provided in this embodiment, the transfer function vector obtained by measurement, for example, [ H (f) ] 1 ),H(f 2 ),……H(f M-1 )，H(f M )]Each element of the vector being the magnitude of the transfer function at a frequency, a 1 x M vector is obtained, the M' sThe values are related to the measurement frequency range and the measurement interval. In this embodiment, the measurement interval is set to 1kHz, the local aging is detected and positioned by using the transfer function amplitude obtained by measuring the frequency band of 3 to 500kHz, and M is 498. And inputting the measured 1 multiplied by M vectors into the detection deep neural network D-CNN and the positioning deep neural network L-CNN. For the exploration deep neural network D-CNN, a 1 × N vector is output, where N represents the division of the cable to be tested into N segments, each element in the vector represents the possibility of aging of each segment of the cable to be tested, and N is, for example, 20. For the L-CNN, outputting a 1 x 3 vector which respectively represents the initial position L of the cable aging section b Termination position L of aged section of cable e And the degree of ageing γ of the cable. The detection deep neural network D-CNN is used for detecting and preliminarily positioning local aging in the cable system. Unlike the traditional detection and positioning mode, the cable system to be monitored is divided into N sections, and the aging possibility of each section is estimated. The use of the structure can improve the generalization capability of the method so as to be suitable for positioning under various aging modes. The designed detection deep neural network D-CNN is composed of two parts, wherein the first part is a sparse self-encoder (hereinafter referred to as SAE) for extracting characteristic information in an input transfer function vector. SAE can reduce the dimension of the input vector, extract the relation implied by each element in the input vector, reduce the calculation complexity of the subsequent supervised learning algorithm and improve the detection and positioning accuracy. And a Convolutional Neural Network (CNN) is connected behind the sparse self-encoder, and estimation of local aging parameters is obtained through two convolutional layers and a full connection layer according to a feature matrix output by SAE. Because the detection deep neural network D-CNN and the positioning deep neural network L-CNN estimate the aging parameters based on the feature matrix extracted by SAE, the detection deep neural network D-CNN and the positioning deep neural network L-CNN adopt the same structure, and the output layer structures are different according to different tasks.
And testing the detection and positioning capabilities of the detection deep neural network D-CNN and the positioning deep neural network L-CNN. First, for the medium voltage cables used in table 1, set L b And L e Is 0: 0.05L: l, aging degree gamma is 1:0.01:1.5, 10500 data are obtainedThe composed data set. The data set is expressed as [0.6:0.2 ]]The method comprises the steps of dividing the training set, the verification set and the test set. The training set is used for training the proposed detection depth neural network D-CNN and the proposed positioning depth neural network L-CNN, the validation set is used for adjusting parameters in the neural network, and the test set is used for testing the detection and positioning performance of the obtained model.
The local aging detection capability of the detection deep neural network D-CNN is firstly tested. And for each element in the output vector of the D-CNN, if the element is greater than 0.5, the section is considered to be aged, and if the element is less than 0.5, the section is considered to be intact. If the number of aging sections in the whole cable is more than or equal to 1, the whole cable is recorded as a local aging phenomenon. Simulation results show that on a test set, the accuracy of detecting the local aging of the whole cable by the deep neural network D-CNN reaches 98.0%, and the accuracy of detecting the aging of each section reaches 90.0%. The results of the positioning of the four representative samples in the test set are shown in fig. 6. Researches show that the trained detection deep neural network D-CNN can carry out accurate diagnosis and preliminary positioning on the local aging of the cable.
And then researching the positioning capability and the aging degree estimation accuracy of the trained positioning deep neural network L-CNN. The results are shown in Table 2. As can be seen from Table 2, the L-CNN realizes more accurate positioning and estimation of the local aging position and aging degree of the cable. The error of positioning the aging start point is 5.24%, the error of positioning the aging end point is 5.63%, and the error of estimating the aging degree is 4.18%. Therefore, accurate positioning of local aging of the cable can be achieved by using the trained detection deep neural network D-CNN and the positioning deep neural network L-CNN.
TABLE 2 positioning accuracy of deep positioning neural network L-CNN
Accurate detection and positioning of local aging under load change condition based on transfer learning
As can be seen from the solution of the relationship between the transfer function and the local aging parameter of the cable, the transfer function of the cable system is closely related to the load size. The measured transfer function amplitude differs for different loads under the same aging. In order to solve this problem, the present embodiment further proposes a method for detecting and locating local aging under variable load conditions based on transfer learning. In the research, the transfer function amplitude changes with different loads, and the law of the transfer function peak value movement under the aging condition is quite similar under the condition of variable loads. Therefore, a transfer learning method based on parameter fine tuning is introduced. The structures of the detection deep neural network D-CNN and the positioning deep neural network L-CNN and most parameters of training are not changed, and only the weight coefficient matrix between the full connection layer connected with the output and the output is finely adjusted based on a Back Propagation (BP) algorithm, so that the calculation complexity is reduced, and the training time is saved.
Tests show that when the load is 100 omega, the accuracy of the D-CNN detection directly using the detection depth neural network is reduced to 83.05%. The detection accuracy can be improved to 85.96% by using the proposed algorithm based on the transfer learning. For the L-CNN, the initial position L of the cable aging section is determined without using transfer learning b Is reduced to 11.54%, and the terminal position L of the cable aging section is determined e The accuracy of the estimation was reduced to 15.93%, and the accuracy of the estimation of the degree of aging γ of the cable was reduced to 9.5%. The accuracy of the above estimates can be improved to 5.81%, 6.21% and 5.51%, respectively, close to the accuracy on the training set 1, using the proposed algorithm based on the transfer learning.
Compared with the BIS method in the prior art, the method provided by the embodiment only needs to measure the voltages at two ends of the cable system, is easy to implement, and can implement online measurement based on the PLM technology or the HFCT technology. And because the method estimates the aging possibility of each section of the cable to be monitored, the method is suitable for different aging modes and has better robustness. The comparison between the method and BIS method under the conditions of uniform aging, end aging, and non-uniform aging is shown in FIGS. 7(a), (b), and (c), respectively. The first action in the figure is an aging mode, the second action is used for detecting the estimation result of the deep neural network D-CNN, and the third action is used for obtaining the result by the BIS method. As can be seen from fig. 7, the method provided by the present implementation has higher local aging positioning accuracy compared to the BIS method. When the degree of aging is low, it is difficult for the BIS method to detect the presence of local aging, and the method can achieve high-precision positioning of local aging of low degree of aging. Meanwhile, the BIS method is difficult to be applied to the end portion aging, and the proposed method can obtain good effects in the above three aging modes.
In addition, the method provided by this example also performed better than BIS method in the case of smooth aging, as shown in fig. 8 and 9. Smooth aging is one of the aging modes common in the industry and may represent hot spot aging in cable systems. Since the aging parameters change smoothly in this aging mode, there is no place where the impedance changes significantly, so the BIS method is difficult to detect and locate the smooth aging. The method is based on transfer function measurement to position local aging, so that the overall information of the cable is utilized to the maximum extent, and the D-CNN structure of the deep sounding neural network has good generalization capability on various aging modes, so that the positioning of the local aging in the smooth aging mode can be realized, and the method has great advantages compared with a BIS method.
According to another embodiment of the present invention, there is provided a cable partial-aging detection apparatus, which is configured as shown in fig. 10 and includes a transfer function measurement module, a partial-aging detection module, and a partial-aging localization module.
And the transfer function measuring module is used for measuring the transfer function of the cable to be measured.
And the local aging detection module extracts the characteristic information in the transfer function vector and inputs the characteristic information into the trained detection depth neural network D-CNN, and judges whether the cable to be detected has local aging or not by using the detection depth neural network D-CNN. The module judges whether the cable to be tested has local aging or not according to a 1 multiplied by N vector output by a detection deep neural network D-CNN, wherein N represents that the cable to be tested is divided into N sections, and each element in the vector represents the possibility of aging of each section of the cable to be tested.
And the local aging positioning module extracts the characteristic information in the transfer function vector, inputs the characteristic information into the trained positioning depth neural network L-CNN, and utilizes the positioning depth neural network L-CNN to perform aging positioning and aging degree estimation when judging that the cable to be tested has local aging. The module carries out aging positioning and aging degree estimation according to a 1 x 3 vector output by a positioning depth neural network L-CNN, wherein the 1 x 3 vector respectively represents the initial position of an aging section of the cable to be detected, the termination position of the aging section of the cable to be detected and the aging degree of the cable to be detected.
In summary, the invention relates to a method and a device for detecting local aging of a cable, which detect the local aging of the cable by measuring a transfer function of the cable in a frequency range, according to an amplitude of the transfer function, based on a deep learning algorithm, and by using a detection deep neural network D-CNN and a positioning deep neural network L-CNN, and estimate an aging position and an aging degree. The transfer function amplitude is used, so that the influence of phase angle errors caused by real-time communication in a system can be avoided, and the method is easy to realize in an actual system, thereby being easy to realize on-line detection. The detection method provided by the invention is suitable for detecting various local aging modes, including uniform aging, non-uniform aging, end aging, smooth aging and the like; the aging degree can be estimated, so that a basis is provided for subsequent preventive maintenance; meanwhile, the problem of local aging positioning under the variable load condition is solved based on transfer learning, the precision of the detection method under the variable load condition can be improved based on parameter fine adjustment, and the calculation cost and the calculation time are saved.
It should be understood that the above-described embodiments of the present invention are merely illustrative of or explaining the principles of the invention and are not to be construed as limiting the invention. Therefore, any modifications, equivalents, improvements and the like which are made without departing from the spirit and scope of the present invention shall be included in the protection scope of the present invention. Further, it is intended that the appended claims cover all such variations and modifications as fall within the scope and boundaries of the appended claims or the equivalents of such scope and boundaries.
Claims (7)
1. A cable local aging detection method is characterized by comprising the following steps:
measuring a transfer function of a cable to be measured according to a preset time interval, inputting a transfer function vector obtained by measurement into a trained detection deep neural network D-CNN, wherein the detection deep neural network D-CNN outputs a 1 xN vector, N represents the possibility of dividing the cable to be measured into N sections, each element in the vector represents the aging of each section of the cable to be measured, judging whether the cable to be measured has local aging by using the detection deep neural network D-CNN, and if so, performing the next step; if not, returning;
the transfer function of the cable comprises the following equation:
wherein L is b Indicating the starting position of the aged section of the cable, L e Denotes the end position of the cable aging section, L denotes the total cable length, H 0 Represents the transfer function of the cable in the intact section,
indicates for a bit located at x i A transfer function for a small segment of length Δ x;
inputting a transfer function vector obtained by measurement into a trained positioning depth neural network L-CNN, outputting a 1 x 3 vector by the positioning depth neural network L-CNN, respectively representing the initial position of the aging section of the cable to be measured, the termination position of the aging section of the cable to be measured and the aging degree of the cable to be measured, and performing aging positioning and aging degree estimation by using the positioning depth neural network L-CNN;
wherein the vector of transfer functions is an l x M vector, each element in the vector being the magnitude of the transfer function at a frequency, the value of M being related to the measurement frequency range and the measurement interval.
2. The method of claim 1, wherein the utilized exploration deep neural network D-CNN and localization deep neural network L-CNN each comprise a sparse self-encoder and a convolutional neural network.
3. The method of claim 2, wherein for each element in the vector output by the D-CNN, if the value is greater than 0.5, the segment corresponding to the element is determined to be aged, and if the value is less than or equal to 0.5, the segment corresponding to the element is determined to be intact.
4. The method of claim 3, wherein the weight coefficient matrix between the fully-connected layer and the output in the convolutional neural network is adjusted based on a back propagation algorithm when the load changes.
5. The cable local aging detection device is characterized by comprising a transfer function measurement module, a local aging detection module and a local aging positioning module; wherein, the first and the second end of the pipe are connected with each other,
the transfer function measuring module is used for measuring the transfer function of the cable to be measured; the transfer function of the cable comprises the following equation:
wherein L is b Indicating the starting position of the aged section of the cable, L e Indicating the end position of the aged section of the cable, L the total length of the cable, H 0 Representing the transfer function of the intact section of cable,
for a position at x i A transfer function for a small segment of length Δ x;
the local aging detection module inputs the measured transfer function vector into a trained detection deep neural network D-CNN, the detection deep neural network D-CNN outputs a 1 xN vector, wherein N represents the possibility of aging of each section of the cable to be detected, and each element in the vector represents the possibility of aging of each section of the cable to be detected, and the detection deep neural network D-CNN is used for judging whether the cable to be detected has local aging;
the local aging positioning module inputs the measured transfer function vector into a trained positioning depth neural network L-CNN, the positioning depth neural network L-CNN outputs a 1 x 3 vector which respectively represents the initial position of the aging section of the cable to be measured, the termination position of the aging section of the cable to be measured and the aging degree of the cable to be measured, and when the cable to be measured is judged to have local aging, the positioning depth neural network L-CNN is used for aging positioning and aging degree estimation;
wherein the transfer function vector is a l x M vector, each element in the vector being the magnitude of the transfer function at a frequency, the value of M being related to the measurement frequency range and the measurement interval.
6. The apparatus of claim 5, wherein the local aging detection module determines whether the cable under test has local aging according to a 1 × N vector output by the deep neural network D-CNN, where N represents a division of the cable under test into N segments, and each element in the vector represents a possibility of aging of each segment of the cable under test.
7. The apparatus of claim 5, wherein the local aging positioning module performs aging positioning and aging degree estimation according to a 1 x 3 vector output by the positioning depth neural network L-CNN, wherein the 1 x 3 vector represents a start position of an aging section of the cable to be tested, an end position of the aging section of the cable to be tested, and an aging degree of the cable to be tested.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011431893.7A CN112595913B (en) | 2020-12-07 | 2020-12-07 | Cable local aging detection method and detection device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011431893.7A CN112595913B (en) | 2020-12-07 | 2020-12-07 | Cable local aging detection method and detection device |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112595913A CN112595913A (en) | 2021-04-02 |
| CN112595913B true CN112595913B (en) | 2022-07-26 |
Family
ID=75191382
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011431893.7A Active CN112595913B (en) | 2020-12-07 | 2020-12-07 | Cable local aging detection method and detection device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112595913B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117520983A (en) * | 2023-11-07 | 2024-02-06 | 浙江大学 | Structural damage identification robust method based on electromechanical impedance and CNN |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103901330A (en) * | 2014-04-15 | 2014-07-02 | 上海君世电气科技有限公司 | Partial discharge on-line monitoring method for XLPE cable |
| CN105629136A (en) * | 2015-12-28 | 2016-06-01 | 国网甘肃省电力公司金昌供电公司 | Cable insulation state online automatic monitoring and diagnosis system |
| CN108520357A (en) * | 2018-04-10 | 2018-09-11 | 北京中电普华信息技术有限公司 | A kind of method of discrimination, device and the server of line loss abnormal cause |
| CN110286303A (en) * | 2019-07-10 | 2019-09-27 | 国家电网有限公司 | A kind of coaxial cable insulation cable ageing state appraisal procedure based on BP neural network |
| CN110687408A (en) * | 2019-10-15 | 2020-01-14 | 清华大学 | Cable insulation monitoring method and device |
| US10698019B2 (en) * | 2018-11-16 | 2020-06-30 | Chung Yuan Christian University | Method of online estimating remaining life of moving power cable |
| CN111968077A (en) * | 2020-07-24 | 2020-11-20 | 苏州浪潮智能科技有限公司 | Wire detection method, system, equipment and medium |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110068748B (en) * | 2019-04-26 | 2020-09-25 | 中国矿业大学 | 10kV single-core cable early state identification method based on deep convolutional neural network |
| CN110335260B (en) * | 2019-06-27 | 2023-07-14 | 华东送变电工程有限公司 | Power cable damage detection method based on light convolutional neural network |
-
2020
- 2020-12-07 CN CN202011431893.7A patent/CN112595913B/en active Active
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103901330A (en) * | 2014-04-15 | 2014-07-02 | 上海君世电气科技有限公司 | Partial discharge on-line monitoring method for XLPE cable |
| CN105629136A (en) * | 2015-12-28 | 2016-06-01 | 国网甘肃省电力公司金昌供电公司 | Cable insulation state online automatic monitoring and diagnosis system |
| CN108520357A (en) * | 2018-04-10 | 2018-09-11 | 北京中电普华信息技术有限公司 | A kind of method of discrimination, device and the server of line loss abnormal cause |
| US10698019B2 (en) * | 2018-11-16 | 2020-06-30 | Chung Yuan Christian University | Method of online estimating remaining life of moving power cable |
| CN110286303A (en) * | 2019-07-10 | 2019-09-27 | 国家电网有限公司 | A kind of coaxial cable insulation cable ageing state appraisal procedure based on BP neural network |
| CN110687408A (en) * | 2019-10-15 | 2020-01-14 | 清华大学 | Cable insulation monitoring method and device |
| CN111968077A (en) * | 2020-07-24 | 2020-11-20 | 苏州浪潮智能科技有限公司 | Wire detection method, system, equipment and medium |
Non-Patent Citations (5)
| Title |
|---|
| Application of PRPD Spectrum Characteristic in the Recognition of the Aging State of EPDM Cable Insulation;Enxin Xiang 等;《2019 4th International Conference on Power and Renewable Energy》;20200316;度335-341页 * |
| 基于行波特征量与堆叠自动编码器的电缆早期故障定位方法;李成升 等;《四川电力技术》;20190831;第42卷(第4期);第72-79页 * |
| 抑制局部放电老化影响的XLPE电缆绝缘缺陷识别;符方达 等;《电工电气》;20200415(第4期);第16-24页 * |
| 电力电缆及隧道在线监测与移动巡检协同策略探讨;徐伟强 等;《电测与仪表》;20191010;第56卷(第19期);第121-125页 * |
| 高压电缆局部绝缘老化故障智能定位系统设计;李万超 等;《自动化与仪器仪表》;20190825(第8期);第41-45页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112595913A (en) | 2021-04-02 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| EP1932006B1 (en) | System and method for monitoring of electrical cables | |
| CN104937427B (en) | Monitor the method and system of cable status | |
| KR101570506B1 (en) | Apparatus and method for detecting cable fault location and impedance using linear chirp reflectometry | |
| KR101926995B1 (en) | Apparatus and method for detecting fault location of underground cable | |
| CN106841913B (en) | Distribution line fault location method | |
| US11061060B2 (en) | Method and system for locating defects on an electric cable | |
| CN108896872A (en) | Handheld cable fault detection system and method based on SSTDR | |
| CN113702754A (en) | Distribution cable defect positioning algorithm adopting windowed Fourier transform | |
| Xu et al. | Special requirements of high frequency current transformers in the on-line detection of partial discharges in power cables | |
| Sheng et al. | A novel on-line cable pd localisation method based on cable transfer function and detected pd pulse rise-time | |
| Hua et al. | Analysis and diagnosis of shielded cable faults based on finite-element method and time-reversal time-frequency domain reflectometry | |
| CN112595913B (en) | Cable local aging detection method and detection device | |
| Minullin et al. | Model and experimental detection of single phase-to-earth faults of overhead conductors in 6–10 kV distribution circuits by a location method | |
| CN113075501B (en) | Cable fault positioning method and system based on impedance spectrum periodic characteristics | |
| Zhang et al. | A method for locating and diagnosing cable abrasion based on broadband impedance spectroscopy | |
| Pinomaa et al. | Diagnostics of low-voltage power cables by using broadband impedance spectroscopy | |
| Li et al. | Location of cable joints with moisture in medium voltage distribution grid based on frequency domain reflection method | |
| CN116540016A (en) | Cable defect diagnosis method based on reflection coefficient discrete sequence | |
| CN114062852B (en) | Cable intermediate connector fault diagnosis method, device, equipment and readable storage medium | |
| Rao et al. | Cable defect location by using frequency domain reflectometry with synchrosqueezing generalized S-transform | |
| Hashmi et al. | Covered-conductor overhead distribution line modeling and experimental verification for determining its line characteristics | |
| CN114675128A (en) | Submarine cable insulation fault on-line positioning method based on sheath current and voltage | |
| Li et al. | A new method for on-line cable tanδ monitoring | |
| Ragusa et al. | Practical Evaluation of Electromagnetic Time Reversal to Locate Partial Discharges on Power Networks in the Presence of Noise | |
| Norouzi et al. | Localization of Deviations in Cable Geometry Using FDA Fault Localization Method |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |