阅读说明：本技术 一种电缆局部放电在线监测缺陷辨识方法 (Cable partial discharge online monitoring defect identification method ) 是由 潘文霞 李昕芮 熊蕙 卢为 刘东超 于 2020-06-22 设计创作，主要内容包括：本发明涉及一种电缆局部放电在线监测缺陷辨识方法,基于φ-OTDR原理,针对局部放电的电缆实现在线监测,获得对电缆缺陷类型的辨识,其中利用在线监测数据,建立局部放电监测数据的自回归滑动平均模型,获取局部放电的特征矢量,最后建立并训练随机森林分类模型,从而获得应用电缆缺陷识别模型,用于实际当中电缆绝缘缺陷类型的识别；如此设计方案突破了现有技术的限制,弥补了现有技术的不足,能够在基于φ-OTDR原理电缆局部放电在线监测中,实现电缆绝缘缺陷类型辨识,且方法辨识准确度较高,无需额外添加设备,实际应用中对电缆运行状态监测和故障检修具有重要意义。(The invention relates to a cable partial discharge on-line monitoring defect identification method, which is based on a phi-OTDR principle, realizes on-line monitoring aiming at a cable with partial discharge and obtains the identification of the cable defect type, wherein an autoregressive moving average model of the partial discharge monitoring data is established by utilizing on-line monitoring data, a characteristic vector of the partial discharge is obtained, and finally a random forest classification model is established and trained, so that an application cable defect identification model is obtained and is used for identifying the cable insulation defect type in practice; the design scheme breaks through the limitation of the prior art, overcomes the defects of the prior art, can realize the type identification of the insulation defect of the cable in the online monitoring of the partial discharge of the cable based on the phi-OTDR principle, has higher identification accuracy, does not need additional equipment, and has important significance for the monitoring of the running state of the cable and the fault maintenance in practical application.)

1. A cable partial discharge online monitoring defect identification method is used for identifying defects of partial discharge cables and is characterized by comprising a cable defect identification model construction method and defect identification aiming at the partial discharge cables by applying a cable defect identification model; the cable defect identification model construction method comprises the following steps A to E;

step A, collecting cables respectively corresponding to different defect types to form a sample set, wherein one cable corresponds to one defect type, and then entering step B;

b, respectively aiming at each cable in the sample set, positioning each partial discharge cable section in the cable so as to obtain each partial discharge cable section in the sample set, and then entering the step C;

c, extracting a backward Rayleigh scattering light intensity variation data sequence of the partial discharge cable segments respectively aiming at each partial discharge cable segment in the sample set, and performing validity verification to eliminate the interference of white noise; obtaining a data sequence of light intensity variation of backward Rayleigh scattering light of non-white noise corresponding to each partial discharge cable segment in the sample set, and then entering the step D;

d, respectively aiming at each partial discharge cable segment in the sample set, establishing an autoregressive moving average model corresponding to the partial discharge cable segment according to a non-white noise backward Rayleigh scattering light intensity variation data sequence corresponding to the partial discharge cable segment, and extracting a model coefficient in the autoregressive moving average model as a characteristic vector corresponding to the partial discharge cable segment; obtaining the characteristic vectors corresponding to the partial discharge cable sections respectively, and then entering the step E;

and E, taking the characteristic vector corresponding to each partial discharge cable section as input, taking the defect type corresponding to the cable to which each partial discharge cable section belongs as output, and training aiming at a preset classification model to obtain a cable defect identification model.

2. The method for identifying the defect of the cable partial discharge online monitoring according to claim 1, wherein the defect identification is implemented for the partial discharge cable by applying a cable defect identification model, and the method comprises the following steps:

step I, aiming at a target cable with partial discharge, positioning each partial discharge target cable section in the target cable, and then entering step II;

step II, respectively aiming at each partial discharge target cable section in the target cable, executing the methods from the step C to the step D to obtain a characteristic vector corresponding to each partial discharge target cable section, and then entering the step III;

and III, respectively aiming at the characteristic vectors corresponding to the partial discharge target cable sections, applying a cable defect identification model to obtain the defect types corresponding to the partial discharge target cable sections, namely obtaining the defect types corresponding to the target cables.

3. The method for identifying the partial discharge online monitoring defect of the cable according to claim 1 or 2, wherein the step B comprises the following steps:

step B1, respectively aiming at each cable in the sample set, averagely dividing the cable into N sections to obtain N monitoring unit sections corresponding to the cable, and entering step B2 after the division of all the cables is completed;

step B2, respectively aiming at each cable in the sample set, in a preset monitoring period, executing the measurement of the light intensity variation of the backward scattering light for preset M times aiming at the cable, and respectively aiming at each monitoring unit section on the cable, according to the following formula:

obtaining the vibration coefficient K of each monitoring unit section on the cable_nWherein N is more than or equal to 1 and less than or equal to N, z_nIndicating the nth section of the monitoring unit, K, on the cable_nRepresenting the vibration coefficient, z, of the nth section of the monitoring unit on the cable_stM is more than or equal to 1 and less than or equal to M, and delta I represents the section of the monitoring unit without confirmed partial discharge_m(z_n) Indicating the variation of the intensity of the backward scattered light, delta I, obtained by the m-th measurement in the preset monitoring period corresponding to the nth monitoring unit section on the cable_m(z_n) Indicating the variation of the intensity of the backward scattered light obtained by the mth measurement in the preset monitoring period corresponding to the monitoring unit section without partial discharge; further obtaining the vibration coefficient of each monitoring unit section on each cable, and then entering step B3;

b3, respectively aiming at each monitoring unit section on each cable, judging whether the vibration coefficient of the monitoring unit section is not less than a preset vibration coefficient threshold value, if so, judging that the monitoring unit section has partial discharge, and the monitoring unit section is a partial discharge cable section; otherwise, judging that the monitoring unit section has no partial discharge; and then positioning to obtain each partial discharge cable section on each cable in the sample set, and then entering the step C.

4. The method for identifying the cable partial discharge online monitoring defect according to claim 1 or 2, wherein in the step C, the following steps C1 to C5 are performed by using an Ljung-Box inspection method for each partial discharge cable segment in the sample set, so as to verify the validity of the data sequence of the light intensity variation of the backward rayleigh scattered light of the extracted partial discharge cable segment and eliminate the interference of white noise;

c1, extracting a backward Rayleigh scattered light intensity variable quantity data sequence of the partial discharge cable segment, and then entering a step C2;

step C2. is based on the preset maximum delay order T, T is smaller than the length L of the backward Rayleigh scattering light intensity variation data sequence, and T is not less than 1 and not more than T, so as to obtain the autocorrelation coefficients of the backward Rayleigh scattering light intensity variation data sequence corresponding to each T-order lag respectivelyThen proceed to step C3;

step C3. determines the respective autocorrelation coefficients

step C4, according to the following formula:

obtaining a statistic Q (T) of a data sequence of the light intensity variation of the backward Rayleigh scattering light, and then entering a step C5;

step C5., according to the predetermined significance level α, determine whether Q (T) is greater than χ with T degree of freedom corresponding to 1- α²If the distribution value is the white noise data, judging that the backward Rayleigh scattering light intensity variation data sequence is non-white noise data, namely obtaining the non-white noise backward Rayleigh scattering light intensity variation data sequence corresponding to the partial discharge cable section; otherwise, the procedure returns to step C1.

5. The method for identifying the cable partial discharge online monitoring defect according to claim 1 or 2, wherein in the step D, for each partial discharge cable segment in the sample set, steps D1 to D8 are performed, an autoregressive moving average model corresponding to the partial discharge cable segment is established, and model coefficients therein are extracted as the feature vector corresponding to the partial discharge cable segment;

d1, obtaining the stationarity of a non-white noise backward Rayleigh scattering light intensity variation data sequence corresponding to the partial discharge cable segment by applying a unit root inspection method, and judging that the data sequence is a stable sequence or a non-stable sequence, wherein if the data sequence is judged to be the stable sequence, the step D2 is carried out; if the sequence is determined to be a non-stationary sequence, step D8 is entered;

d2. according to the length L of the data sequence of the variation of the light intensity of the non-white noise backward Rayleigh scattering light corresponding to the partial discharge cable segment, randomly extracting each value in the specified Gauss standard sequence in turn, and forming a sequence for making a final product according to the extraction sequence₁、…、_LThen step D3 is entered;

step D3, according to the following formula:

establishing an autoregressive moving average model corresponding to the non-white noise backward Rayleigh scattering light intensity variation data sequence, and then entering the step D4; wherein phi is₀、φ₁、…、φ_pAnd theta₁、θ₂、…、θ_qThe coefficient of the autoregressive moving average model and the coefficient form a characteristic vector [ phi ] of the data sequence of the light intensity variation of the non-white noise backward Rayleigh scattering light₀、φ₁、…、φ_p,θ₁、θ₂、…、θ_q]；x_lA fitting value, X, representing the data of the No. l in the data series { X } of the variation of the intensity of the non-white noise backward Rayleigh scattered light_l-1、x_l-pRespectively representing the l-1 th data and the l-p th data in the data sequence { X } of the light intensity variation of the non-white noise backward Rayleigh scattering light, wherein l-p is more than or equal to 1,_l、_l-1、_l-qrespectively representing sequences₁、…、_LThe l th data, the l-1 st data and the l-q th data in the data are obtained, wherein l-q is more than or equal to 1; e () represents the sequence₁、…、_LMean value of (f), Var () representing the sequence (f)₁、…、_LThe variance of } satisfies

step D4., determining whether p and q are 0 by using the autocorrelation coefficient and partial autocorrelation coefficient curve, and then entering step D5;

step D5., based on the determination result of step D4 regarding p and q, selecting the values of p and q of each group, and applying a preset method to the p and q values of each group respectively to obtain the p and q values of each group through calculationCorresponding feature vector [ phi ]₀、φ₁、…、φ_p,θ₁、θ₂、…、θ_q]Then, go to step D6;

step D6. targets each set of feature vectors [ phi ] separately₀、φ₁、…、φ_p,θ₁、θ₂、…、θ_q]The feature vector [ phi ]₀、φ₁、…、φ_p,θ₁、θ₂、…、θ_q]Substituting the characteristic vector into the autoregressive moving average model established in the step D3, and fitting to obtain a backward Rayleigh scattered light intensity variable quantity data sequence corresponding to the characteristic vector; further obtaining the data sequence of the light intensity variation of the backward rayleigh scattered light corresponding to each group of feature vectors, and then entering step D7;

step D7., checking the data sequence of the variation of the light intensity of the backward Rayleigh scattering light corresponding to each group of the feature vectors, and selecting the feature vector [ phi ] corresponding to the optimal checking result₀、φ₁、…、φ_p,θ₁、θ₂、…、θ_q]Uniquely representing a data sequence of the actually measured light intensity variation of the backward Rayleigh scattering light of the partial discharge cable segment, namely forming a characteristic vector corresponding to the partial discharge cable segment;

step D8. is to execute difference operation for the data sequence of the variation of the light intensity of the non-white noise backward rayleigh scattering light corresponding to the partial discharge cable segment, update the data sequence of the variation of the light intensity of the non-white noise backward rayleigh scattering light to a stable sequence, and then return to step D2.

6. The method for identifying the partial discharge on-line monitoring defect of the cable according to claim 5, wherein: in the step D5, any one of the available moment estimation method, the maximum likelihood estimation method, and the least square method is applied to each group of p and q values, respectively, to calculate and obtain the feature vector [ phi ] corresponding to each group of p and q values, respectively₀、φ₁、…、φ_p,θ₁、θ₂、…、θ_q]。

7. The method for identifying the partial discharge on-line monitoring defect of the cable according to claim 5, wherein: in the step D7, aiming at the data sequence of the light intensity variation of the backward rayleigh scattered light corresponding to each group of feature vectors, an AIC criterion or a BIC criterion is adopted for verification;

wherein, the AIC criterion expression is as follows:

in the formula (I), the compound is shown in the specification,representing residual variance between a backward Rayleigh scattered light intensity variation data sequence obtained by fitting an autoregressive moving average model and an actually measured backward Rayleigh scattered light intensity variation data sequence, wherein r represents the number r of parameters of the autoregressive moving average model to be p +1+ q;

and in the AIC criterion, the optimal parameter r'₀Satisfies the following conditions:

wherein M (L) is

The BIC criteria expression is as follows:

in the formula (I), the compound is shown in the specification,representing results obtained by fitting an autoregressive moving average modelThe residual variance between the backward Rayleigh scattered light intensity variation data sequence and the actually measured backward Rayleigh scattered light intensity variation data sequence is shown as r, and the number r of the parameters of the autoregressive moving average model is p +1+ q;

and in BIC criterion, the optimal parameter r'₀Satisfies the following conditions:

wherein M (L) isOr

8. The method for identifying the partial discharge on-line monitoring defect of the cable according to claim 1, wherein: the preset classification model is a random forest classification model.

Technical Field

The invention relates to a method for identifying a cable partial discharge online monitoring defect, and belongs to the technical field of cable online monitoring.

Background

The cable is widely applied to urban power distribution networks, cross-sea power transmission and other special occasions. The large-scale cable application is accompanied by the frequent occurrence of cable accidents, and statistics show that 43.7 percent of cable accidents are caused by cable insulation problems under the condition of not counting external force damage. The early manifestation of the problem of cable insulation is closely related to the partial discharge of the cable, which is also different due to different insulation defects. Therefore, the defect type of the cable can be identified by utilizing the information of the partial discharge of the cable, the defect forming reason is analyzed, and measures are taken to reduce the occurrence of the insulation defect of the cable. The method has the advantages that partial discharge of the cable is analyzed, the type of the insulation defect of the cable is identified, and the method has important significance for reducing the insulation defect of the cable and ensuring safe and stable operation of a power system. In the traditional partial discharge monitoring method, in the cable partial discharge monitoring, the positioning method is complex, the precision is poor, and crosstalk is easy to occur in monitoring signals.

Disclosure of Invention

The invention aims to solve the technical problem of providing a cable partial discharge online monitoring defect identification method, which is based on the phi-OTDR principle, realizes online monitoring aiming at cable partial discharge, can efficiently realize the identification of cable defects and ensures the stability of the actual work of cables.

The invention adopts the following technical scheme for solving the technical problems: the invention designs a cable partial discharge online monitoring defect identification method, which is used for identifying defects aiming at partial discharge cables and comprises a cable defect identification model construction method and an application cable defect identification model, wherein the defect identification aiming at the partial discharge cables is realized; the cable defect identification model construction method comprises the following steps A to E;

step A, collecting cables respectively corresponding to different defect types to form a sample set, wherein one cable corresponds to one defect type, and then entering step B;

b, respectively aiming at each cable in the sample set, positioning each partial discharge cable section in the cable so as to obtain each partial discharge cable section in the sample set, and then entering the step C;

c, extracting a backward Rayleigh scattering light intensity variation data sequence of the partial discharge cable segments respectively aiming at each partial discharge cable segment in the sample set, and performing validity verification to eliminate the interference of white noise; obtaining a data sequence of light intensity variation of backward Rayleigh scattering light of non-white noise corresponding to each partial discharge cable segment in the sample set, and then entering the step D;

d, respectively aiming at each partial discharge cable segment in the sample set, establishing an autoregressive moving average model corresponding to the partial discharge cable segment according to a non-white noise backward Rayleigh scattering light intensity variation data sequence corresponding to the partial discharge cable segment, and extracting a model coefficient in the autoregressive moving average model as a characteristic vector corresponding to the partial discharge cable segment; obtaining the characteristic vectors corresponding to the partial discharge cable sections respectively, and then entering the step E;

and E, taking the characteristic vector corresponding to each partial discharge cable section as input, taking the defect type corresponding to the cable to which each partial discharge cable section belongs as output, and training aiming at a preset classification model to obtain a cable defect identification model.

As a preferred technical solution of the present invention, the method for identifying a defect of a partial discharge cable by using a cable defect identification model includes the following steps:

step I, aiming at a target cable with partial discharge, positioning each partial discharge target cable section in the target cable, and then entering step II;

step II, respectively aiming at each partial discharge target cable section in the target cable, executing the methods from the step C to the step D to obtain a characteristic vector corresponding to each partial discharge target cable section, and then entering the step III;

and III, respectively aiming at the characteristic vectors corresponding to the partial discharge target cable sections, applying a cable defect identification model to obtain the defect types corresponding to the partial discharge target cable sections, namely obtaining the defect types corresponding to the target cables.

As a preferred technical solution of the present invention, the step B includes the steps of:

step B1, respectively aiming at each cable in the sample set, averagely dividing the cable into N sections to obtain N monitoring unit sections corresponding to the cable, and entering step B2 after the division of all the cables is completed;

step B2, respectively aiming at each cable in the sample set, in a preset monitoring period, executing the measurement of the light intensity variation of the backward scattering light for preset M times aiming at the cable, and respectively aiming at each monitoring unit section on the cable, according to the following formula:

obtaining the vibration coefficient K of each monitoring unit section on the cable_nWherein N is more than or equal to 1 and less than or equal to N, z_nIndicating the nth section of the monitoring unit, K, on the cable_nRepresenting the vibration coefficient, z, of the nth section of the monitoring unit on the cable_stM is more than or equal to 1 and less than or equal to M, and delta I represents the section of the monitoring unit without confirmed partial discharge_m(z_n) Indicating the variation of the intensity of the backward scattered light, delta I, obtained by the m-th measurement in the preset monitoring period corresponding to the nth monitoring unit section on the cable_m(z_n) Indicating the variation of the intensity of the backward scattered light obtained by the mth measurement in the preset monitoring period corresponding to the monitoring unit section without partial discharge; further obtaining the vibration coefficient of each monitoring unit section on each cable, and then entering step B3;

b3, respectively aiming at each monitoring unit section on each cable, judging whether the vibration coefficient of the monitoring unit section is not less than a preset vibration coefficient threshold value, if so, judging that the monitoring unit section has partial discharge, and the monitoring unit section is a partial discharge cable section; otherwise, judging that the monitoring unit section has no partial discharge; and then positioning to obtain each partial discharge cable section on each cable in the sample set, and then entering the step C.

As a preferred technical solution of the present invention, in the step C, the following steps C1 to C5 are performed by using an Ljung-Box inspection method for each partial discharge cable segment in the sample set, so as to implement validity verification on the data sequence of the light intensity variation of the backward rayleigh scattered light of the extracted partial discharge cable segment and eliminate the interference of white noise;

c1, extracting a backward Rayleigh scattered light intensity variable quantity data sequence of the partial discharge cable segment, and then entering a step C2;

step C2. is based on the preset maximum delay order T, T is smaller than the length L of the backward Rayleigh scattering light intensity variation data sequence, and T is not less than 1 and not more than T, so as to obtain the autocorrelation coefficients of the backward Rayleigh scattering light intensity variation data sequence corresponding to each T-order lag respectively

Then proceed to step C3;

step C3. determines the respective autocorrelation coefficientsIf the light intensity variation data sequence is not equal to 0, judging that the light intensity variation data sequence of the backward rayleigh scattered light is white noise data, deleting the light intensity variation data sequence of the backward rayleigh scattered light, and returning to the step C1; otherwise go to step C4;

step C4, according to the following formula:

obtaining a statistic Q (T) of a data sequence of the light intensity variation of the backward Rayleigh scattering light, and then entering a step C5;

step C5. is performed according to a preset significance level αJudging whether Q (T) is larger than X with T degree of freedom corresponding to 1- α²If the distribution value is the white noise data, judging that the backward Rayleigh scattering light intensity variation data sequence is non-white noise data, namely obtaining the non-white noise backward Rayleigh scattering light intensity variation data sequence corresponding to the partial discharge cable section; otherwise, the procedure returns to step C1.

As a preferred technical solution of the present invention, in the step D, for each partial discharge cable segment in the sample set, the following steps D1 to D8 are performed, an autoregressive moving average model corresponding to the partial discharge cable segment is established, and a model coefficient therein is extracted as a feature vector corresponding to the partial discharge cable segment;

d1, obtaining the stationarity of a non-white noise backward Rayleigh scattering light intensity variation data sequence corresponding to the partial discharge cable segment by applying a unit root inspection method, and judging that the data sequence is a stable sequence or a non-stable sequence, wherein if the data sequence is judged to be the stable sequence, the step D2 is carried out; if the sequence is determined to be a non-stationary sequence, step D8 is entered;

d2. according to the length L of the data sequence of the variation of the light intensity of the non-white noise backward Rayleigh scattering light corresponding to the partial discharge cable segment, randomly extracting each value in the specified Gauss standard sequence in turn, and forming a sequence for making a final product according to the extraction sequence₁、…、_LThen step D3 is entered;

step D3, according to the following formula:

establishing an autoregressive moving average model corresponding to the non-white noise backward Rayleigh scattering light intensity variation data sequence, and then entering the step D4; wherein phi is₀、φ₁、…、φ_pAnd theta₁、θ₂、…、θ_qThe coefficient of the autoregressive moving average model and the coefficient form a characteristic vector [ phi ] of the data sequence of the light intensity variation of the non-white noise backward Rayleigh scattering light₀、φ₁、…、φ_p,θ₁、θ₂、…、θ_q]；x_lA fitting value, X, representing the data of the No. l in the data series { X } of the variation of the intensity of the non-white noise backward Rayleigh scattered light_l-1、x_l-pRespectively representing the l-1 th data and the l-p th data in the data sequence { X } of the light intensity variation of the non-white noise backward Rayleigh scattering light, wherein l-p is more than or equal to 1,_l、_l-1、_l-qrespectively representing sequences₁、…、_LThe l th data, the l-1 st data and the l-q th data in the data are obtained, wherein l-q is more than or equal to 1; e () represents the sequence₁、…、_LMean value of (f), Var () representing the sequence (f)₁、…、_LThe variance of } satisfies sigma²Positive distribution of (c ∈ {1, …, L }, d ∈ {1, …, L }, and c ≠ d, E: (L) (C ≠ D))_{c d}) Representation sequence₁、…、_LAverage of all pairwise data products in a ∈ {1, …, L }, b ∈ {1, …, L }, E (x)_{a b}) Representing all respective images from X, a₁、…、_LMean of the product of two by two data;

step D4., determining whether p and q are 0 by using the autocorrelation coefficient and partial autocorrelation coefficient curve, and then entering step D5;

step D5., based on the determination result of step D4 regarding p and q, selecting the values of p and q of each group, and applying a preset method to the p and q values of each group respectively to calculate the feature vectors [ phi ] corresponding to the p and q values of each group respectively₀、φ₁、…、φ_p,θ₁、θ₂、…、θ_q]Then, go to step D6;

step D6. targets each set of feature vectors [ phi ] separately₀、φ₁、…、φ_p,θ₁、θ₂、…、θ_q]The feature vector [ phi ]₀、φ₁、…、φ_p,θ₁、θ₂、…、θ_q]Substituting the characteristic vector into the autoregressive moving average model established in the step D3, and fitting to obtain a backward Rayleigh scattered light intensity variable quantity data sequence corresponding to the characteristic vector; further obtaining the data sequence of the light intensity variation of the backward Rayleigh scattering light corresponding to each group of characteristic vectors respectively, and thenEntering step D7;

step D7., checking the data sequence of the variation of the light intensity of the backward Rayleigh scattering light corresponding to each group of the feature vectors, and selecting the feature vector [ phi ] corresponding to the optimal checking result₀、φ₁、…、φ_p,θ₁、θ₂、…、θ_q]Uniquely representing a data sequence of the actually measured light intensity variation of the backward Rayleigh scattering light of the partial discharge cable segment, namely forming a characteristic vector corresponding to the partial discharge cable segment;

step D8. is to execute difference operation for the data sequence of the variation of the light intensity of the non-white noise backward rayleigh scattering light corresponding to the partial discharge cable segment, update the data sequence of the variation of the light intensity of the non-white noise backward rayleigh scattering light to a stable sequence, and then return to step D2.

As a preferred technical scheme of the invention: in the step D5, any one of the available moment estimation method, the maximum likelihood estimation method, and the least square method is applied to each group of p and q values, respectively, to calculate and obtain the feature vector [ phi ] corresponding to each group of p and q values, respectively₀、φ₁、…、φ_p,θ₁、θ₂、…、θ_q]。

As a preferred technical scheme of the invention: in the step D7, aiming at the data sequence of the light intensity variation of the backward rayleigh scattered light corresponding to each group of feature vectors, an AIC criterion or a BIC criterion is adopted for verification;

wherein, the AIC criterion expression is as follows:

in the formula (I), the compound is shown in the specification,representing residual variance between a backward Rayleigh scattered light intensity variation data sequence obtained by fitting an autoregressive moving average model and an actually measured backward Rayleigh scattered light intensity variation data sequence, wherein r represents the number r of parameters of the autoregressive moving average model to be p +1+ q;

and in the AIC criterion, the optimal parameter r'₀Satisfies the following conditions:

wherein M (L) isOr

The BIC criteria expression is as follows:

in the formula (I), the compound is shown in the specification,representing residual variance between a backward Rayleigh scattered light intensity variation data sequence obtained by fitting an autoregressive moving average model and an actually measured backward Rayleigh scattered light intensity variation data sequence, wherein r represents the number r of parameters of the autoregressive moving average model to be p +1+ q;

and in BIC criterion, the optimal parameter r'₀Satisfies the following conditions:

wherein M (L) is

Or

As a preferred technical scheme of the invention: the preset classification model is a random forest classification model.

Compared with the prior art, the method for identifying the cable partial discharge on-line monitoring defects has the following technical effects:

the invention relates to a method for identifying the defect of the partial discharge on-line monitoring of a designed cable, which is based on the phi-OTDR principle, realizes on-line monitoring aiming at the partial discharge cable and obtains the identification of the defect type of the cable, wherein, the on-line monitoring data is utilized to establish an autoregressive moving average model of the partial discharge monitoring data, obtain the characteristic vector of the partial discharge, and finally establish and train a random forest classification model, thereby obtaining an applied cable defect identification model which is used for identifying the insulation defect type of the cable in practice; the design scheme breaks through the limitation of the prior art, overcomes the defects of the prior art, can realize the type identification of the insulation defect of the cable in the online monitoring of the partial discharge of the cable based on the phi-OTDR principle, has higher identification accuracy, does not need additional equipment, and has important significance for the monitoring of the running state of the cable and the fault maintenance in practical application.

Drawings

FIG. 1 is a schematic diagram of a sample cable installation in the method for identifying defects by online monitoring of partial discharge of a cable according to the present invention;

fig. 2 is a schematic flow chart of the method for identifying the defect of the cable partial discharge online monitoring according to the present invention.

Detailed Description

The following description will explain embodiments of the present invention in further detail with reference to the accompanying drawings.

The invention designs a cable partial discharge on-line monitoring defect identification method, which is used for identifying defects aiming at partial discharge cables and comprises a cable defect identification model construction method and an application cable defect identification model, wherein the defect identification is realized aiming at the partial discharge cables; in practical application, as shown in fig. 2, the following steps a to E are performed to implement the construction of the cable defect identification model.

And step A, collecting cables respectively corresponding to different defect types to form a sample set, wherein one cable corresponds to one defect type, and then entering step B.

In step a, regarding the collection of cables respectively corresponding to different defect types, a manufacturing method may be adopted, that is, cables with different types of partial discharge are manufactured, and one cable corresponds to one defect type, and then, by using an optical fiber sensing technology, a backward rayleigh scattered light intensity variation data sequence in an optical fiber along the cable is monitored, so as to obtain a backward rayleigh scattered light intensity variation data sequence at each position along the cable, and record the corresponding partial discharge type.

And B, respectively aiming at each cable in the sample set, positioning each partial discharge cable section in the cable so as to obtain each partial discharge cable section in the sample set, and then entering the step C.

In practical applications, the step B specifically executes the following steps B1 to B3 to complete the execution design of the step B.

And B1, respectively aiming at each cable in the sample set, averagely dividing the cable into N sections to obtain N monitoring unit sections corresponding to the cable, and entering the step B2 after the division of all the cables is finished.

Step B2, respectively aiming at each cable in the sample set, in a preset monitoring period, executing the measurement of the light intensity variation of the backward scattering light for preset M times aiming at the cable, and respectively aiming at each monitoring unit section on the cable, according to the following formula:

obtaining the vibration coefficient K of each monitoring unit section on the cable_nWherein N is more than or equal to 1 and less than or equal to N, z_nIndicating the nth section of the monitoring unit, K, on the cable_nRepresenting the vibration coefficient, z, of the nth section of the monitoring unit on the cable_stM is more than or equal to 1 and less than or equal to M, and delta I represents the section of the monitoring unit without confirmed partial discharge_m(z_n) Indicating the variation of the intensity of the backward scattered light, delta I, obtained by the m-th measurement in the preset monitoring period corresponding to the nth monitoring unit section on the cable_m(z_n) Indicating the variation of the intensity of the backward scattered light obtained by the mth measurement in the preset monitoring period corresponding to the monitoring unit section without partial dischargeAn amount; and obtaining the vibration coefficient of each monitoring unit section on each cable, and then entering step B3.

B3, respectively aiming at each monitoring unit section on each cable, judging whether the vibration coefficient of the monitoring unit section is not less than a preset vibration coefficient threshold value K_thIf yes, judging that the monitoring unit section has partial discharge, wherein the monitoring unit section is a partial discharge cable section; otherwise, judging that the monitoring unit section has no partial discharge; and then positioning to obtain each partial discharge cable section on each cable in the sample set, and then entering the step C. In practical application, K_thCan be determined experimentally, K_thToo large a setting results in low monitoring sensitivity, K_thIf the setting is too small, the system is easy to generate false alarm.

C, extracting a backward Rayleigh scattering light intensity variation data sequence of the partial discharge cable segments respectively aiming at each partial discharge cable segment in the sample set, and performing validity verification to eliminate the interference of white noise; and further obtaining a data sequence of the light intensity variation of the backward Rayleigh scattering light of the non-white noise corresponding to each partial discharge cable segment in the sample set, and then entering the step D.

In practical application, the step C is to perform the following steps C1 to C5 by using Ljung-Box inspection method for each partial discharge cable segment in the sample set, so as to implement validity verification on the data sequence of the light intensity variation of the backward rayleigh scattered light of the extracted partial discharge cable segment and eliminate the interference of white noise.

Original hypothesis for Ljung-Box test: the data sequences of the light intensity variation of the backward Rayleigh scattering light are independent, and the overall correlation coefficient is 0, namelySome correlation coefficients that are not 0 are computed, simply due to errors resulting from random sampling. The standby assumption is that: the monitoring data of the light intensity variation of the backward Rayleigh scattering light is not independent, and at least one of the monitoring data exists

Wherein T is more than or equal to 1 and less than or equal to T.

And C1, extracting a backward Rayleigh scattered light intensity variation data sequence of the partial discharge cable segment, and then entering the step C2.

Step C2. is based on the preset maximum delay order T, T is smaller than the length L of the backward Rayleigh scattering light intensity variation data sequence, and T is not less than 1 and not more than T, so as to obtain the autocorrelation coefficients of the backward Rayleigh scattering light intensity variation data sequence corresponding to each T-order lag respectivelyThen proceed to step C3.

Step C3. determines the respective autocorrelation coefficientsIf the light intensity variations are all equal to 0, determining that the backward rayleigh scattered light intensity variation data sequence is white noise data, deleting the backward rayleigh scattered light intensity variation data sequence, namely the original hypothesis is established, and returning to the step C1; otherwise, go to step C4.

Step C4, according to the following formula:

and obtaining the statistic Q (T) of the data sequence of the light intensity variation of the backward Rayleigh scattering light, and then entering the step C5.

Step C5., according to the predetermined significance level α, determine whether Q (T) is greater than χ with T degree of freedom corresponding to 1- α²If the distribution value is positive, judging that the backward Rayleigh scattering light intensity variation data sequence is non-white noise data, namely obtaining the non-white noise backward Rayleigh scattering light intensity variation data sequence corresponding to the partial discharge cable segment, namely Q (T) falls in the rejection region of the original hypothesis; otherwise, the procedure returns to step C1.

Step D, respectively aiming at each partial discharge cable segment in the sample set, establishing an autoregressive moving average (ARMA) model corresponding to the partial discharge cable segment according to a non-white noise backward Rayleigh scattering light intensity variation data sequence corresponding to the partial discharge cable segment, and extracting a model coefficient in the ARMA model as a characteristic vector corresponding to the partial discharge cable segment; and further acquiring the characteristic vectors corresponding to the partial discharge cable sections respectively, and then entering the step E.

Step D, in the actual implementation, the following steps D1 to D8 are performed for each partial discharge cable segment in the sample set, respectively, to establish an autoregressive moving average model (ARMA) corresponding to the partial discharge cable segment, and extract the model coefficients therein as the feature vectors corresponding to the partial discharge cable segment.

D1, applying a unit root inspection method (ADF) to obtain the stationarity of a non-white noise backward Rayleigh scattering light intensity variation data sequence corresponding to the partial discharge cable segment, and judging that the stationarity is a stable sequence or a non-stable sequence, wherein if the stationarity is judged to be the stable sequence, the step D2 is carried out; if it is determined to be a non-stationary sequence, the process proceeds to step D8.

D2. according to the length L of the data sequence of the variation of the light intensity of the non-white noise backward Rayleigh scattering light corresponding to the partial discharge cable segment, randomly extracting each value in the specified Gauss standard sequence in turn, and forming a sequence for making a final product according to the extraction sequence₁、…、_LAnd then proceeds to step D3.

Step D3, according to the following formula:

establishing an autoregressive moving average (ARMA) corresponding to the data sequence of the light intensity variation of the non-white noise backward Rayleigh scattering light, and then entering the step D4; wherein phi is₀、φ₁、…、φ_pAnd theta₁、θ₂、…、θ_qThe coefficient of the autoregressive moving average model and the coefficient form a characteristic vector [ phi ] of the data sequence of the light intensity variation of the non-white noise backward Rayleigh scattering light₀、φ₁、…、φ_p,θ₁、θ₂、…、θ_q]；x_lRepresenting the intensity variation of the non-white noise backward Rayleigh scattering lightFitting value, X, of the ith data in the sequence of quantitative data { X }_l-1、x_l-pRespectively representing the l-1 th data and the l-p th data in the data sequence { X } of the light intensity variation of the non-white noise backward Rayleigh scattering light, wherein l-p is more than or equal to 1,_l、_l-1、_l-qrespectively representing sequences₁、…、_LThe l th data, the l-1 st data and the l-q th data in the data are obtained, wherein l-q is more than or equal to 1; e () represents the sequence₁、…、_LMean value of (f), Var () representing the sequence (f)₁、…、_LThe variance of } satisfies sigma²Positive distribution of (c ∈ {1, …, L }, d ∈ {1, …, L }, and c ≠ d, E: (L) (C ≠ D))_{c d}) Representation sequence₁、…、_LAverage of all pairwise data products in a ∈ {1, …, L }, b ∈ {1, …, L }, E (x)_{a b}) Representing all respective images from X, a₁、…、_LMean of the product of two by two data of.

Step D4. determines whether p and q are 0 using the autocorrelation coefficient and partial autocorrelation coefficient curve, and then proceeds to step D5.

In the step D4, the determination of whether p and q are 0 is actually performed by using an autocorrelation coefficient and partial autocorrelation coefficient curve as follows, and the principle in table 1 is that ar (p) in table 1 indicates that p ≠ 0, q ═ 0, ma (q) indicates that p ≠ 0, q ≠ 0, and the calculation result is that p ≠ 0, and the data monitoring sequence of the light intensity variation of the backward rayleigh scattered light with q ═ 0 is a white noise sequence and is rejected in step S2.

ACF	PACF	ARMA model types
			Tailing	P orderTruncation tail	AR (p) model
q-order truncation	Tailing	MA (q) model
			Tailing	Tailing	ARMA (q) model

TABLE 1

Step D5., based on the determination result of step D4 regarding p and q, selecting the values of p and q of each group, and applying a preset method to the p and q values of each group respectively to calculate the feature vectors [ phi ] corresponding to the p and q values of each group respectively₀、φ₁、…、φ_p,θ₁、θ₂、…、θ_q]Then, the process proceeds to step D6. In the actual implementation, any one of the available moment estimation method, maximum likelihood estimation method and least square method is applied to each group of p and q values, and the feature vector [ phi ] corresponding to each group of p and q values is obtained through calculation₀、φ₁、…、φ_p,θ₁、θ₂、…、θ_q]。

Step D6. targets each set of feature vectors [ phi ] separately₀、φ₁、…、φ_p,θ₁、θ₂、…、θ_q]The feature vector [ phi ]₀、φ₁、…、φ_p,θ₁、θ₂、…、θ_q]Substituting the characteristic vector into the autoregressive moving average model established in the step D3, and fitting to obtain a backward Rayleigh scattered light intensity variable quantity data sequence corresponding to the characteristic vector; and then obtaining the data sequence of the light intensity variation of the backward rayleigh scattered light corresponding to each group of feature vectors, and then entering step D7.

Step D7. is to determine the backward rayls corresponding to each set of feature vectorsUtilizing scattered light intensity variable data sequence to execute verification and selecting characteristic vector [ phi ] corresponding to optimum verification result₀、φ₁、…、φ_p,θ₁、θ₂、…、θ_q]And uniquely representing a data sequence of the actually measured backward Rayleigh scattering light intensity variation of the partial discharge cable segment, namely forming a characteristic vector corresponding to the partial discharge cable segment.

In step D7, in practical applications, the data sequence of the light intensity variations of the backward rayleigh scattered light corresponding to each set of feature vectors may specifically be verified by using an AIC criterion or a BIC criterion.

Wherein, the AIC criterion expression is as follows:

in the formula (I), the compound is shown in the specification,

and a residual variance between the backward Rayleigh scattered light intensity variation data sequence obtained by the autoregressive moving average model fitting and the actually measured backward Rayleigh scattered light intensity variation data sequence is shown, wherein r represents the number of parameters of the autoregressive moving average model, and is p +1+ q.

And in the AIC criterion, the optimal parameter r'₀Satisfies the following conditions:

wherein M (L) is

Or

The BIC criteria expression is as follows:

in the formula (I), the compound is shown in the specification,

and a residual variance between the backward Rayleigh scattered light intensity variation data sequence obtained by the autoregressive moving average model fitting and the actually measured backward Rayleigh scattered light intensity variation data sequence is shown, wherein r represents the number of parameters of the autoregressive moving average model, and is p +1+ q.

And in BIC criterion, the optimal parameter r'₀Satisfies the following conditions:

wherein M (L) is

Or

Step D8. is to execute difference operation for the data sequence of the variation of the light intensity of the non-white noise backward rayleigh scattering light corresponding to the partial discharge cable segment, update the data sequence of the variation of the light intensity of the non-white noise backward rayleigh scattering light to a stable sequence, and then return to step D2.

And E, taking the characteristic vector corresponding to each partial discharge cable section as input, taking the defect type corresponding to the cable to which each partial discharge cable section belongs as output, training a preset classification model, specifically training a random forest classification model, and obtaining a cable defect identification model.

In the practical application process, after the steps a to E are executed to obtain the cable defect identification model, the cable defect identification model can be applied in the subsequent practical application, and the following steps I to III are specifically executed for the partial discharge cable to realize the discrimination of the cable defect.

And step I, aiming at the target cable with partial discharge, positioning each partial discharge target cable section in the target cable, and then entering the step II.

And II, respectively aiming at each partial discharge target cable section in the target cable, executing the methods from the step C to the step D to obtain the characteristic vector corresponding to each partial discharge target cable section, and then entering the step III.

And III, respectively aiming at the characteristic vectors corresponding to the partial discharge target cable sections, applying a cable defect identification model to obtain the defect types corresponding to the partial discharge target cable sections, namely obtaining the defect types corresponding to the target cables.

The method for identifying the partial discharge on-line monitoring defect of the cable designed by the invention is applied to the cable shown in the embodiment shown in figure 1, wherein the length of the sample cable is 5 meters, the length of the sensing optical fiber is 5 meters, the sensing optical fiber is a common single-mode optical fiber, the sensing optical fiber is tightly coated on the surface of the cable, and the average apparent discharge amount of the cable defect is 100pC under the action of 7kV voltage.

In this embodiment, the designed method for identifying the defect of the cable partial discharge online monitoring is utilized to identify the type of the defect of the cable partial discharge, three types of defects are set in an experiment, 125 groups of samples are collected for each type of defect, 375 groups of samples are collected in total, and the method has high identification accuracy by cross inspection, wherein the identification accuracy is 98.36%.

According to the method for identifying the cable partial discharge online monitoring defect, which is disclosed by the invention, different cable insulation defect types can be identified, the identification accuracy is higher, additional equipment is not required, the cost is low, the cable defect types can be reported online, and the method has important significance for monitoring the running state of the cable and troubleshooting.

The embodiments of the present invention have been described in detail with reference to the drawings, but the present invention is not limited to the above embodiments, and various changes can be made within the knowledge of those skilled in the art without departing from the gist of the present invention.