An overhead transmission line fault diagnosis method and computer readable medium

Abstract

The application provides an overhead transmission line fault diagnosis method and a computer readable medium. The application collects distributed fault recording samples of historical records; determines the reliability of recording data of different recording devices through a TOPSIS method; assigns different weights according to the reliability of the recording data of the recording devices; obtains fused waveform data of the historical records through data fusion processing; extracts fault characteristic quantities of the fused waveform data obtained from the historical record sample data; constructs a diagnosis model between the fault characteristic quantities and a transmission line fault type by using a random forest method; determines the distance between different distributed fault recorders and a fault point according to a traveling wave reflection network diagram; and determines the fault type in combination with the diagnosis model between the fault characteristic quantities and the transmission line fault type. The application solves the problem that a large amount of human resources and time cost are consumed in traditional fault diagnosis and the accuracy is low, and improves the efficiency and accuracy of fault diagnosis.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of overhead transmission line fault diagnosis, and particularly relates to an overhead transmission line fault diagnosis method and a computer readable medium. BACKGROUND

[0002] With the rapid development of global economy and the continuous increase of population, the structure of modern power grid is becoming more and more complex, the scale is expanding, and the capacity is increasing. The reliability of power supply has become a common problem faced by countries all over the world, and the overhead transmission line fault mainly caused by lightning has caused great impact and damage to the stable power supply of the power grid.

[0003] With the continuous development of society, the demand for power supply is expanding, and the total length of transmission lines is expanding. This trend makes the stable operation of transmission lines have more and more obvious influence on the power grid. In order to ensure power supply, it is necessary to accurately, quickly and effectively diagnose and identify the type of line fault when the overhead transmission line fault occurs. The human patrol has low accuracy and low efficiency. In order to solve this problem, an automatic diagnosis method for overhead transmission line fault needs to be developed. The type of transmission line fault itself has diversity and complexity, which increases the difficulty of developing fault diagnosis methods.

[0004] A large amount of research work has been done on the diagnosis method of transmission line fault at home and abroad. According to the different data sources, the steps are divided into centralized fault diagnosis and distributed fault diagnosis. The centralized fault diagnosis refers to taking the data of a single unit monitoring point near the fault line point in the power system, processing and analyzing the data, and diagnosing the type of line fault, that is, all the data and metrics of the unit can be used as the standard for judging the fault. Distributed fault diagnosis refers to installing a monitoring terminal every 20-30km based on the double-end traveling wave positioning principle, which divides the line into several sections. The monitoring terminal has the function of collecting power frequency current and high frequency traveling wave current. When the line fault occurs, first, the power frequency fault current is used to judge the fault phase and fault interval; then, the high frequency traveling wave current collected by the monitoring terminal close to the fault interval is selected for synchronous analysis, and the waveform characteristics of the high frequency traveling wave current are used for fault type identification.

[0005] Compared with each other, the centralized fault diagnosis has the disadvantage that the power system is a large-scale spatially distributed system, and a unit thereof cannot obtain the measurement values of all units. The fault recorder responsible for collecting key data is far from the fault point of the transmission line in terms of physical distance, so that the fault waveform data is prone to distortion when transmitted on the line, the accuracy of the fault recording data transmitted to the fault recorder is greatly reduced, and the accuracy of fault diagnosis is low. At the same time, the centralized fault diagnosis mode is to call the steady-state data, monitoring alarm information, fault recording information and the like of each dispersed system by experienced operation personnel for reasoning and analysis, so that the fault diagnosis efficiency is low, and a large amount of manpower investment is not in line with the maximization of economic benefits pursued by the power grid. The distributed diagnosis method has the advantages of high fault tolerance and high diagnosis dimension in data analysis, and well compensates for the shortcomings of the former. The distributed fault diagnosis system is installed with multiple sets of distributed diagnosis devices on the line conductor to monitor the power frequency fault current and the traveling wave current, and overcomes the influence of traveling wave attenuation; the fault type is rapidly and accurately judged by analyzing and comparing the fault traveling waveforms. This makes the intelligent diagnosis of the distributed fault traveling wave of the transmission line have superior performance and strong practicability.

[0006] At present, there are few and immature studies on fault diagnosis using a distributed model. There are defects such as inaccurate recording of wave data, improper extraction and fusion of data characteristics, and low accuracy of algorithm model. Therefore, it is of great significance to research a distributed fault diagnosis method that can reasonably fuse the data recorded by multiple fault recorders, has high prediction accuracy, and has good fault tolerance in the case of uncertain information, and realizes rapid identification of fault types, to ensure the safe and stable operation of the power grid and improve the power supply reliability of the system. SUMMARY

[0007] The problem to be solved by the present application is to overcome the technical defects of the original technology, and the present application provides an overhead transmission line fault diagnosis method and a computer readable medium.

[0008] To solve the above technical problems, the technical scheme of the method of the present application is an overhead transmission line fault diagnosis method, characterized in that it comprises the following steps:

[0009] Step 1: Collecting historical recording distributed fault recording samples;

[0010] Step 2: Determining the reliability of the data recorded by different fault recorders by the TOPSIS method;

[0011] Step 3: Assigning different weights according to the reliability of the data recorded by the fault recorders;

[0012] Step 4: Performing data fusion processing on the historical recording distributed fault recording data based on the assigned weights to obtain historical recording fusion waveform data;

[0013] Step 5: Extract the fault feature quantity of the fusion waveform data obtained from the historical record sample data;

[0014] Step 6: Process a large number of historical samples according to steps 2 to 5, establish a database, and use the random forest method to construct a diagnostic model between the fault feature quantity and the fault type of the transmission line;

[0015] Step 7: Determine the distance between the different distributed fault recorders and the fault point according to the traveling wave reflection network diagram of the real-time fault;

[0016] Step 8: Real-time fault record data is processed by the same method as steps 2 to 5, and combined with the diagnostic model between the fault feature quantity and the fault type of the transmission line, the fault type;

[0017] As preferred, the historical record distributed fault recorder sample in step 1 is:

[0018] Sample * =Sample * ((s1,x1),(s2,x2),...,(s i ,x i ),...,(s n ,x n ),label

[0019] Wherein, s i is the sample recorded by the i-th recorder within a period of time after the fault occurs, x i is the distance between the installation point and the fault point of the i-th recorder on the overhead line, n is the total number of recorders installed on the overhead transmission line, label represents the fault type, and the value range is {label1, label2, label3, label4}, wherein label1 is lightning strike fault, label2 is bird damage fault, label3 is mountain fire fault, and label4 is external force damage fault;

[0020] In step 1, the sample recorded by each recorder is specifically represented as:

[0021] s i =[V i I i ]

[0022] Wherein, V i is the voltage in the sample recorded by the i-th recorder, and I i is the current in the sample recorded by the i-th recorder;

[0023] In step 1, the matrix of the fault waveform data recorded by all recorders is constructed as:

[0024]

[0025] wherein V i is the voltage in the sample recorded by the i-th recording wave instrument, I i is the current in the sample recorded by the i-th recording wave instrument, x i is the distance between the installation point and the fault point on the overhead line of the i-th recording wave instrument;

[0026] The data recorded at the moment of the fault occurrence are:

[0027]

[0028]

[0029] wherein, is the value of the matrix E at t = 0, is the voltage instantaneous value recorded by the i-th recording wave instrument at t = 0, is the current instantaneous value recorded by the i-th recording wave instrument at t = 0;

[0030] In step 2, the elements of the matrix are respectively subjected to forward processing to obtain the forward-processed matrix E'

[0031] The voltage is an intervalization index, and the method for forward processing thereof is:

[0032]

[0033]

[0034] wherein M(V) is an intermediate transition amount, and are the maximum and minimum values of the voltage possibly recorded by the fault recording wave instrument, is the voltage instantaneous value recorded by the i-th recording wave instrument at t = 0, V i is the result of the forward processing of the voltage instantaneous value of the i-th recording wave instrument at t = 0;

[0035] The current is an intervalization index, and the method for forward processing thereof is:

[0036]

[0037]

[0038] wherein M(I) is an intermediate transition amount, and are respectively the maximum and minimum values of the current possibly recorded by the fault recording wave instrument, is the current instantaneous value recorded by the i-th current recorder at t=0 i is the result of the forward processing of the current instantaneous value recorded by the i-th current recorder at t=0

[0039] The distance between the installation point and the fault point of the current recorder on the overhead line is a minimum index, and the forward processing method thereof is as follows:

[0040]

[0041] wherein, x i is the result of the forward processing of the distance between the installation point and the fault point of the current recorder on the overhead line, x i is the distance between the installation point and the fault point of the i-th current recorder on the overhead line;

[0042] The forward-processed matrix E' is as follows:

[0043]

[0044] In step 2, the maximum vector E' + and the minimum vector E' - are obtained by using the forward-processed matrix E':

[0045] E' + =(max{V1', V2',..., V i ',..., V n '}, max{I1', I2',..., I i ',..., I n '}, max{x1', x2',..., x i ',..., x n '})

[0046] E' - =(min{V1', V2',..., V i ',..., V n '}, min{I1', I2',..., I i ',..., I n '}, min{x1', x2',..., x i ',..., x n})

[0047] wherein, E' + is the maximum vector, E' - is the minimum vector, V i ' is the result of the forward processing of the voltage instantaneous value recorded by the i-th current recorder at t=0, I iE' is the result of the positive treatment of the current instantaneous value of the i-th recording wave at t=0, x i E' is the result of the positive treatment of the distance between the installation point and the fault point of the i-th recording wave on the overhead line, max indicates the maximum treatment, and min indicates the minimum treatment;

[0048] In step 2, the positive treatment data of the i-th recording wave are respectively subtracted from the maximum vector E' + and the minimum vector E' -

[0049] A i =(V i ' I i ' x i ' )-E' + Bi=(V i ' I i ' x i ' )-E' -

[0050] A i =(a i1 a i2 a i3 )

[0051] B i =(b i1 b i2 b i3 )

[0052] Wherein, E' is the maximum vector, E' is the minimum vector, V + ' is the result of the positive treatment of the voltage instantaneous value of the i-th recording wave at t=0, I - ' is the result of the positive treatment of the current instantaneous value of the i-th recording wave at t=0, x i ' is the result of the positive treatment of the distance between the installation point and the fault point of the recording wave on the overhead line, A i and B i are the vectors obtained after subtraction, a i is the difference between the result of the positive treatment of the voltage instantaneous value of the i-th recording wave at t=0 and the maximum value of the results of the positive treatment of the voltage instantaneous values of n recording waves at t=0, a i is the difference between the result of the positive treatment of the current instantaneous value of the i-th recording wave at t=0 and the maximum value of the results of the positive treatment of the current instantaneous values of n recording waves at t=0, and a i1 is the difference between the result of the positive treatment of the voltage instantaneous value of the i-th recording wave at t=0 and the maximum value of the results of the positive treatment of the voltage instantaneous values of n recording waves at t=0, a i2 is the difference between the result of the positive treatment of the current instantaneous value of the i-th recording wave at t=0 and the maximum value of the results of the positive treatment of the current instantaneous values of n recording waves at t=0, and a i3 ​The difference between the result of the forward processing of the distance between the installation point and the fault point on the overhead line of the i-th recorder and the maximum value in the result of the forward processing of the distance between the installation point and the fault point on the overhead line of the n recorders, b i1 The difference between the result of the forward processing of the voltage instantaneous value at t=0 of the i-th recorder and the minimum value in the result of the forward processing of the voltage instantaneous value at t=0 of the n recorders, b i2 The difference between the result of the forward processing of the current instantaneous value at t=0 of the i-th recorder and the minimum value in the result of the forward processing of the current instantaneous value at t=0 of the n recorders, b i3 The difference between the result of the forward processing of the distance between the installation point and the fault point on the overhead line of the i-th recorder and the minimum value in the result of the forward processing of the distance between the installation point and the fault point on the overhead line of the n recorders.

[0053]

[0054] Wherein, m i The reliability of the data recorded by the i-th recorder, a ij The element of the vector A i after the subtraction, a i1 The difference between the result of the forward processing of the voltage instantaneous value at t=0 of the i-th recorder and the maximum value in the result of the forward processing of the voltage instantaneous value at t=0 of the n recorders, a i2 The difference between the result of the forward processing of the current instantaneous value at t=0 of the i-th recorder and the maximum value in the result of the forward processing of the current instantaneous value at t=0 of the n recorders, a i3 The difference between the result of the forward processing of the distance between the installation point and the fault point on the overhead line of the i-th recorder and the maximum value in the result of the forward processing of the distance between the installation point and the fault point on the overhead line of the n recorders, b ij The element of the vector B i after the subtraction, b i1 The difference between the result of the forward processing of the voltage instantaneous value at t=0 of the i-th recorder and the minimum value in the result of the forward processing of the voltage instantaneous value at t=0 of the n recorders, b i2 The difference between the result of the forward processing of the current instantaneous value at t=0 of the i-th recorder and the minimum value in the result of the forward processing of the current instantaneous value at t=0 of the n recorders, b i3 The difference between the result of the forward processing of the distance between the installation point and the fault point on the overhead line of the i-th recorder and the minimum value in the result of the forward processing of the distance between the installation point and the fault point on the overhead line of the n recorders.

[0055] In step 3, the weight value assigned to the recorded data according to the credibility criterion is represented as:

[0056] D = (D1, D2,..., D i ,..., D n )

[0057] wherein D i represents the weight value of the recorded data of the i-th recorder;

[0058] In step 3, the weight value set of the recorded data is assigned according to the credibility criterion:

[0059]

[0060] wherein D i represents the weight value of the recorded data of the i-th recorder, m j represents the credibility of the recorded data of the j-th recorder, and n represents the total number of recorders;

[0061] In step 4, the distributed fault recorded data of the historical record is weighted to obtain the fusion waveform data of the historical record, which is represented as:

[0062]

[0063] wherein S represents the fusion waveform data obtained by weighting the data, D i represents the weight value of the recorded data of the i-th recorder, and s i represents the sample recorded by the i-th recorder;

[0064] In step 5, the specific steps of extracting the fault feature quantity are as follows: according to the obtained fusion waveform data, the head time and tail time can be directly obtained from the fusion waveform data, and the ground resistance corresponding to the fusion waveform data is calculated by the equivalent circuit method, the Fourier decomposition of the fusion waveform data is performed to obtain the harmonic content and other electrical characteristics, and the extracted fault feature quantity is represented as:

[0065]

[0066] wherein T1 and T2 respectively represent the head time and tail time of the fusion waveform data, R represents the ground resistance corresponding to the fusion waveform data, represents the frequency spectrum obtained by Fourier decomposition of the fusion waveform data, and Γ j (S) represents the j-th feature quantity extracted from the fusion waveform data, and p-1 represents the number of feature quantities extracted from the fusion waveform data.

[0067] In step 6, a diagnostic model between the fault feature quantity and the fault type of the power transmission line is constructed by using a random forest algorithm, and the mapping relationship between the fault feature quantity set Γ of different types of faults and the fault type label is embodied in the diagnostic model, and the fault feature quantity set and the fault type of k different types of faults constitute an active data matrix:

[0068]

[0069] Wherein, SS is an active data matrix composed of fault feature quantity sets of different types of faults and fault types, T 1(i) , T 2(i) , R (i) , The wave head time, the wave tail time, the ground resistance corresponding to the fused waveform data, and the frequency spectrum obtained by Fourier decomposition of the fused waveform data of the fault feature quantity of the i-th data are respectively denoted as Γ j (S) (i) The j-th feature quantity extracted from the i-th data is denoted as p-1, and the number of feature quantities extracted from the fused waveform data is denoted as label (i) The fault type of the i-th data is denoted as label.

[0070] Wherein, the definitions are as follows:

[0071]

[0072]

[0073] Wherein, ΓΓ is a matrix composed of fault feature quantities of k faults, T 1(i) , T 2(i) , R (i) , The wave head time, the wave tail time, the ground resistance corresponding to the fused waveform data, and the frequency spectrum obtained by Fourier decomposition of the fused waveform data of the fault feature quantity of the i-th data are respectively denoted as Γ j (S) (i) The j-th feature quantity extracted from the i-th data is denoted as p-1, and the number of feature quantities extracted from the fused waveform data is denoted as llabel, and label (i) The fault type of the i-th data is denoted as label, and p-1 represents the number of types of fault feature quantities.

[0074] In step 6, P sub-matrices ss1, ss2, …, ss i , …, ss P

[0075]

[0076] Wherein, ssi represents the i-th sub-matrix generated randomly, ΓΓis a matrix composed of fault feature quantities of k faults, and llabelis a matrix composed of k fault types, r(ΓΓ) m×(q-1) is a sub-matrix of the fault feature quantity set, γ(llabel) m×1 is a sub-matrix of the fault type matrix llabel, represents a mathematical symbol of a block matrix, and the matrix is divided into left and right two matrices;

[0077] In the above formula, k, p, m and q satisfy the constraint condition:

[0078]

[0079] wherein k represents the number of fault feature quantity sets of different types of historical records and fault types, p-1 is the number of types of fault feature quantities, m represents the number of randomly sampled fault feature quantity sets of different types of k historical records and fault types, and q-1 represents the number of types of fault feature quantities randomly sampled from p-1 fault feature quantities;

[0080] In step 6, P decision trees are obtained by random forest training through the P subsets.

[0081] In step 6, the historical fault recording data is processed according to steps 2-5, and the processed data is input into the P trees of the random forest to obtain P fault type results, and the number of fault types is counted to determine the most, and the fault type is taken as the final prediction result;

[0082] In step 7, the obtained real-time distributed fault recording sample is represented as:

[0083]

[0084] wherein s i is the sample recorded by the i-th recorder in a period of time after the fault occurs, is a distributed fault recording sample recorded by n fault recorders in real time;

[0085] According to the traveling wave reflection network diagram, the distance of different distributed fault recorders from the fault point is obtained as:

[0086] Data = ((s1, x1), (s2, x2),..., (s i , x i ),..., (s n , x n ))

[0087] wherein s ix is the sample recorded by the i-th wave recorder in real time within a period of time after the fault occurs, i L is the distance between the installation point and the fault point of the i-th wave recorder on the overhead line, x is the distributed fault wave recording sample recorded by the n fault wave recorders in real time;

[0088] In step 8, the distributed fault wave recording sample recorded by the n fault wave recorders in real time is processed by the method in step 2 to obtain the credibility of the recording data of different wave recorders;

[0089] In step 8, the credibility of the recording data of different wave recorders is processed by the method in step 3 to obtain the weight of different wave recorders;

[0090] In step 8, further, based on the different weights assigned to different wave recorders, the method in step 4 is used to process the fusion waveform data recorded in real time;

[0091] In step 8, the step of distinguishing the fault type is, according to the mapping relationship between the fault feature quantity set Γ and the fault type label of different types of faults in the constructed random forest model, the type of the fault is obtained from the extracted fault feature quantity set Γ.

[0092] The application also provides a computer readable medium storing a computer program executed by an electronic device, which executes the steps of the overhead transmission line fault diagnosis method when the computer program runs on the electronic device.

[0093] The application has the beneficial effects of overcoming the technical defects of the original technology, providing a method based on machine learning to solve the problem of distributed fault wave recording data overhead transmission line fault diagnosis. When a fault occurs, the data of multiple distributed fault wave recorders is judged for credibility, and after the unreliable data is removed, the data fusion processing is performed to obtain the fusion waveform, the fault feature quantity is extracted from the fusion waveform, and the machine learning algorithm is used to diagnose the fault type from the database and mathematical model established by historical samples, and the adaptive function of the machine learning algorithm is further realized, solving the problem of low accuracy and high cost of manpower and time in traditional fault diagnosis. The fault type is determined by the machine learning model instead of manual experience, improving the efficiency and accuracy of fault diagnosis. BRIEF DESCRIPTION OF DRAWINGS

[0094] Figure 1 : Method flowchart of the embodiment of the application;

[0095] Figure 2 : Schematic diagram of the random forest principle of the embodiment of the application. DETAILED DESCRIPTION

[0096] With reference to the drawings of the embodiments of the present application, the technical solutions in the embodiments of the present application will be described clearly and completely. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments of the present application, all the other embodiments obtained by a person of ordinary skill in the art without creative work fall within the protection scope of the present application.

[0097] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings of the embodiments of the present application. Figures 1-2 The specific implementation of the present application is a kind of crowd source satellite remote sensing image data set optimization method, specifically as follows:

[0098] As shown in Figure 1 , the present application discloses a kind of overhead transmission line fault diagnosis method using distributed fault recording data, comprising the following steps:

[0099] Step 1, collect the historical record of distributed fault recording sample.

[0100] First, the historical record of distributed fault recording sample is represented:

[0101] Sample * =Sample * ((s1,x1),(s2,x2),...,(s i ,x i ),...,(s n ,x n ),label)

[0102] Wherein, s i is the sample recorded by the i-th recorder in a period of time after the fault occurs, x i is the distance between the installation point and the fault point on the overhead line of the i-th recorder, n=100 is the total number of recorders installed on the overhead transmission line, label represents the fault type, and the value range is {label1, label2, label3, label4}, wherein label1 is lightning stroke fault, label2 is bird damage fault, label3 is mountain fire fault, and label4 is external force damage fault.

[0103] The sample recorded by each recorder is specifically represented as step 9:

[0104] s i =[V i I i ]

[0105] Wherein, V i is the voltage in the sample recorded by the i-th recorder, and I i is the current in the sample recorded by the i-th recorder.

[0106] Then, the matrix of all fault recorder recorded fault waveform data is constructed:

[0107]

[0108] Where, V i is the voltage in the sample recorded by the i-th fault recorder, I i is the current in the sample recorded by the i-th fault recorder, x i is the distance between the installation point and the fault point of the i-th fault recorder on the overhead line.

[0109] The data recorded at the moment of fault occurrence is:

[0110]

[0111]

[0112] Where, is the value of matrix E at t = 0, is the voltage instantaneous value recorded by the i-th fault recorder at t = 0, is the current instantaneous value recorded by the i-th fault recorder at t = 0.

[0113] Step 2, determine the credibility of the data recorded by different fault recorders by the TOPSIS method. The specific way is as follows:

[0114] The elements of matrix are processed respectively to obtain the normalized matrix E'

[0115] The voltage is an interval index, and the method of normalization processing is:

[0116]

[0117]

[0118] Where: M(V) is the intermediate transition amount, and are the maximum and minimum values of the voltage that the fault recorder may record, is the voltage instantaneous value recorded by the i-th fault recorder at t = 0, V i ' is the result of the normalized processing of the voltage instantaneous value of the i-th fault recorder at t = 0.

[0119] Similarly, the current is an interval index, and the method of normalization processing is:

[0120]

[0121]

[0122] M(I) is the intermediate transition amount, and is the maximum and minimum current values that the fault recorder can record, is the current instantaneous value recorded by the i-th recorder at t=0, I i ′ is the result of the forward processing of the current instantaneous value of the i-th recorder at t=0.

[0123] The distance between the installation point of the recorder on the overhead line and the fault point is a minimum type index, and the forward processing method is:

[0124]

[0125] wherein: x i ′ is the result of the forward processing of the distance between the installation point of the recorder on the overhead line and the fault point, x i is the distance between the installation point of the i-th recorder on the overhead line and the fault point.

[0126] Therefore, the forward processed matrix E′ is

[0127]

[0128] The maximum vector E′ + and the minimum vector E′ - are obtained by using the forward processed matrix E′:

[0129] E′ + =(max{V1′,V2′,...,V i ′,...,V n ′},max{I1′,I2′,...,I i ′,...,I n ′},max{x1′,x2′,...,x i ′,...,x n ′})

[0130] E′ - =(min{V1′,V2′,...,V i ′,...,V n ′},min{I1′,I2′,...,I i ′,...,I n ′},min{x1′,x2′,...,x i ′,...,x n ′})

[0131] wherein E′ + is the maximum vector, and E′- is the maximum vector, V i is the result of the forward processing of the voltage instantaneous value of the i-th wave recorder at t=0, I i is the result of the forward processing of the current instantaneous value of the i-th wave recorder at t=0, x i is the result of the forward processing of the distance between the installation point and the fault point of the wave recorder on the overhead line, max represents the maximum processing, and min represents the minimum processing;

[0132] The forward-processed data of the i-th wave recorder is respectively subtracted from the maximum vector E' + and the minimum vector E' - :

[0133] A i = (V i ′ I i ′ x i ′)-B ′+

[0134] B i = (V i ′ I i ′ x i ′)-E′-

[0135] A i = (a i1 a i2 a i3 )

[0136] B i = (b i1 b i2 b i3 )

[0137] wherein E' + is the maximum vector, E' - is the minimum vector, V i ′ is the result of the forward processing of the voltage instantaneous value of the i-th wave recorder at t=0, I i ′ is the result of the forward processing of the current instantaneous value of the i-th wave recorder at t=0, x i ′ is the result of the forward processing of the distance between the installation point and the fault point of the wave recorder on the overhead line, A i and B i are the vectors obtained after the subtraction, a i1 is the difference between the result of the forward processing of the voltage instantaneous value of the i-th wave recorder at t=0 and the maximum value of the results of the forward processing of the voltage instantaneous values of n wave recorders at t=0, and a i2The difference between the result of the forward processing of the current instantaneous value of the i-th recorder at t=0 and the maximum value in the results of the forward processing of the current instantaneous values of the n recorders at t=0, a i3 The difference between the result of the forward processing of the distance between the installation point and the fault point on the overhead line of the i-th recorder and the maximum value in the results of the forward processing of the distances between the installation point and the fault point on the overhead line of the n recorders, b i1 The difference between the result of the forward processing of the voltage instantaneous value of the i-th recorder at t=0 and the minimum value in the results of the forward processing of the voltage instantaneous values of the n recorders at t=0, b i2 The difference between the result of the forward processing of the current instantaneous value of the i-th recorder at t=0 and the minimum value in the results of the forward processing of the current instantaneous values of the n recorders at t=0, b i3 The difference between the result of the forward processing of the distance between the installation point and the fault point on the overhead line of the i-th recorder and the minimum value in the results of the forward processing of the distances between the installation point and the fault point on the overhead line of the n recorders.

[0138]

[0139] Wherein, m i represents the reliability of the data recorded by the i-th recorder, a ij is an element of the vector A i obtained after the difference, a i1 The difference between the result of the forward processing of the voltage instantaneous value of the i-th recorder at t=0 and the maximum value in the results of the forward processing of the voltage instantaneous values of the n recorders at t=0, a i2 The difference between the result of the forward processing of the current instantaneous value of the i-th recorder at t=0 and the maximum value in the results of the forward processing of the current instantaneous values of the n recorders at t=0, a i3 The difference between the result of the forward processing of the distance between the installation point and the fault point on the overhead line of the i-th recorder and the maximum value in the results of the forward processing of the distances between the installation point and the fault point on the overhead line of the n recorders, b ij is an element of the vector B i obtained after the difference, b i1 The difference between the result of the forward processing of the voltage instantaneous value of the i-th recorder at t=0 and the minimum value in the results of the forward processing of the voltage instantaneous values of the n recorders at t=0, b i2 The difference between the result of the forward processing of the current instantaneous value of the i-th recorder at t=0 and the minimum value in the results of the forward processing of the current instantaneous values of the n recorders at t=0, b i3The difference between the result of the distance between the installation point and the fault point on the overhead line of the i-th wave recorder after the forward processing and the minimum value in the results of the distance between the installation point and the fault point on the overhead line of the n wave recorders after the forward processing.

[0140] In step 3, different weights are assigned according to the reliability of the wave recorder data. Specifically, the different weights can be represented as:

[0141] D = (D1, D2,..., D i ,..., D n )

[0142] wherein D i represents the weight of the wave data recorded by the i-th wave recorder.

[0143]

[0144] wherein D i represents the weight of the wave data recorded by the i-th wave recorder, m j represents the reliability of the data recorded by the j-th wave recorder, and n represents the total number of wave recorders.

[0145] In step 4, the historical record of the distributed fault wave data is weighted to obtain the fusion waveform data of the historical record based on the assigned weights. The specific implementation is as follows:

[0146]

[0147] wherein S represents the fusion waveform data obtained by weighting the data, D i represents the weight of the wave data recorded by the i-th wave recorder, and s i represents the sample recorded by the i-th wave recorder.

[0148] In step 5, the fault characteristic quantity of the fusion waveform data obtained by extracting the historical record sample data is extracted. The specific steps of extracting the fault characteristic quantity are as follows: according to the obtained fusion waveform data, the head time and the tail time can be directly obtained from the fusion waveform data, and the ground resistance corresponding to the fusion waveform data is calculated by the equivalent circuit method, the Fourier decomposition of the fusion waveform data is performed to obtain the harmonic content and other electrical characteristic quantities, and the extracted fault characteristic quantity is represented as:

[0149]

[0150] wherein T1 and T2 respectively represent the head time and the tail time of the fusion waveform data, R represents the ground resistance corresponding to the fusion waveform data, represents the frequency spectrum obtained by performing Fourier decomposition on the fusion waveform data, and Γ j(S) represents the jth feature quantity extracted from the fused waveform data, and p-1 = 100 represents the number of feature quantities extracted from the fused waveform data.

[0151] Step 6: A large number of historical samples are processed according to steps 2 to 5 to establish a database, and a diagnostic model between fault feature quantities and transmission line fault types is constructed by using the random forest method.

[0152] As shown in Figure 2 , a diagnostic model between fault feature quantities and transmission line fault types is constructed by using the random forest algorithm. The diagnostic model reflects the mapping relationship between the fault feature quantity set Γ of different types of faults and the fault type label. k = 10000 fault feature quantity sets of different types of faults and fault types constitute an active data matrix:

[0153]

[0154] where SS is an active data matrix composed of fault feature quantity sets of different types of faults and fault types, T 1(i) , T 2(i) , R (i) , are the wave head time, wave tail time, ground resistance corresponding to the fused waveform data, and frequency spectrum obtained by Fourier decomposition of the fused waveform data of the ith data, respectively, Γ j (S) (i) represents the jth feature quantity extracted from the ith data, and p-1 represents the number of feature quantities extracted from the fused waveform data, and label (i) is the fault type of the ith data.

[0155] wherein:

[0156]

[0157]

[0158] where ΓΓ is a matrix composed of fault feature quantities of k faults, R 1(i) , T 2(i) , R (i) , are the wave head time, wave tail time, ground resistance corresponding to the fused waveform data, and frequency spectrum obtained by Fourier decomposition of the fused waveform data of the ith data, respectively, Γ j (S) (i) represents the jth feature quantity extracted from the ith data, and p-1 represents the number of feature quantities extracted from the fused waveform data, and ΓΓ is a matrix composed of k fault types, and label (i)p-1 represents the number of types of fault feature quantities.

[0159] Subsequently, P = 1000 sub-matrices ss1, ss2, …, ssP are randomly generated by the SS. i , …, ss P

[0160]

[0161] wherein ss i represents the i-th sub-matrix randomly generated, ΓΓ is a matrix composed of fault feature quantities of k faults, llabel is a matrix composed of k fault types, r(ΓΓ) m×(q-1) is a sub-matrix of the fault feature quantity set, γ(llabel) m×1 is a sub-matrix of the fault type matrix llabel, represents the mathematical symbol of the block matrix, and the matrix is divided into left and right two matrices.

[0162] In the above formula, k, p, m and q satisfy the constraint condition:

[0163]

[0164] wherein k represents the number of fault feature quantity sets of different types of historical records and fault types, p-1 is the number of types of fault feature quantities, m = 1000 represents the number of random samples from the fault feature quantity sets of different types of k historical records and fault types, and q-1 = 50 represents the number of types of fault feature quantities randomly sampled from p-1 types of fault feature quantities.

[0165] Finally, P decision trees are obtained by random forest training through the P subsets.

[0166] The historical fault recording data is processed according to steps 2-5, the processed data is put into the P trees of the random forest, P fault type results are obtained, and the number of the fault type is counted. The most is taken as the final prediction result.

[0167] Step 7, according to the traveling wave reflection network diagram, the distance of different distributed fault recorders from the fault point is determined.

[0168] Specifically, the obtained real-time distributed fault recording sample step 9 is represented as:

[0169]

[0170] wherein s i is the sample recorded by the i-th recorder in a period of time after the fault occurs, The distributed fault recording samples recorded by the n fault recorders in real time.

[0171] The distance of different distributed fault recorders from the fault point is determined according to the traveling wave reflection network diagram, and the distance is obtained as follows:

[0172] Data = ((s1, x1), (s2, x2),..., (s i , x i ),..., (s n , x n ))

[0173] wherein s i is a sample recorded by the i-th recorder in real time within a period of time after the fault occurs, x i is the distance between the installation point of the i-th recorder on the overhead line and the fault point, The distributed fault recording samples recorded by the n fault recorders in real time.

[0174] Step 8, the real-time fault recording data is processed by the same method as steps 2 to 5, and the fault type is determined by the established diagnosis model.

[0175] First, the distributed fault recording samples recorded by the n fault recorders in real time are processed by the method of step 2 to obtain the reliability of the data recorded by different recorders. The reliability of the data recorded by different recorders is processed by the method of step 3 to obtain the weight of different recorders. Further, based on the different weights allocated to different recorders, the real-time recording fusion waveform data is processed by the method of step 4.

[0176] The step of distinguishing the fault type is, according to the mapping relationship between the fault feature quantity set Γ and the fault type label of different types of faults in the constructed random forest model, to obtain the type of the fault from the extracted fault feature quantity set Γ.

[0177] Step 9, the fault feature quantity of the real-time fault fusion waveform data and the fault type thereof are archived and integrated, the database is updated, and the machine learning algorithm with adaptive function continuously improves the accuracy of the diagnosis model.

[0178] The specific method is to archive and incorporate the fault feature quantity of the real-time fault fusion waveform data and the fault type diagnosed by machine learning in the database in step 8, to provide machine learning, and to improve the accuracy of the model fault diagnosis.

[0179] Step 9, based on the allocated weight, the real-time recording distributed fault recording data is processed to obtain fusion waveform data. The calculation method of this step is the same as that of step 3.

[0180] Step 10, extract the fault feature quantity of the fusion waveform data obtained by the real-time recording and wave data.

[0181] The calculation method of this step is the same as that of step 4.

[0182] Step 11, diagnose the fault type of the real-time fault by using the constructed machine learning model.

[0183] Specifically, according to the mapping relationship between the fault feature quantity set Γ of different types of faults in the constructed machine learning model and the fault type label, the type of the fault is obtained.

[0184] Step 12, archive the fault feature quantity of the fusion waveform data of the real-time fault and its fault type.

[0185] The fault feature quantity of the fusion waveform data and the fault type diagnosed by machine learning are archived and incorporated into the database, the database is updated, and the machine learning is provided, so that the machine learning algorithm has the self-adaptive function, and the accuracy of the diagnosis model is continuously improved with the increase of the real-time fault recording data.

[0186] The specific embodiments of the present application also provide a computer readable medium.

[0187] The computer readable medium is a server workstation;

[0188] The server workstation stores a computer program executed by an electronic device, and when the computer program runs on the electronic device, the electronic device executes the steps of the crowd source satellite remote sensing image data set optimization method of the embodiments of the present application.

[0189] It should be understood that parts not described in detail in the specification are all prior art.

[0190] It should be understood that the above description of the preferred embodiments is more detailed, and therefore should not be considered as a limitation on the scope of patent protection of the present application. Those skilled in the art can make substitutions or modifications under the inspiration of the present application without departing from the scope of protection claimed by the present application, and all fall within the scope of protection of the present application. The scope of protection of the present application should be subject to the appended claims.

Claims

1. An overhead transmission line fault diagnosis method, characterized in that: based on the assigned weight, the historical record of distributed fault recording data is processed to obtain the fusion waveform data of the historical record; extracting the fault feature quantity of the fusion waveform data obtained from the historical record sample data; establishing a database and using the random forest method to construct a diagnosis model between the fault feature quantity and the transmission line fault type; determining the distance of different distributed fault recorders from the fault point according to the traveling wave reflection network diagram of the real-time fault; combining the diagnosis model between the fault feature quantity and the transmission line fault type to determine the fault type; comprising the following steps: Step 1: Collecting historical record of distributed fault recording samples; Step 2: Determine the reliability of the recording data of different recorders by TOPSIS method; Step 3: Assign different weights according to the reliability of the recording data of the recorders; Step 4: Based on the assigned weight, the historical record of distributed fault recording data is processed to obtain the fusion waveform data of the historical record; Step 5: Extracting the fault feature quantity of the fusion waveform data obtained from the historical record sample data; Step 6: Process a large number of historical samples according to steps 2 to 5, establish a database, and use the random forest method to construct a diagnosis model between the fault feature quantity and the transmission line fault type; Step 7: Determine the distance of different distributed fault recorders from the fault point according to the traveling wave reflection network diagram of the real-time fault; Step 8: Real-time fault recording data is processed by the same method in steps 2 to 5, and the fault type is determined by combining the diagnosis model between the fault feature quantity and the transmission line fault type; The real-time distributed fault recording sample obtained in step 7 is represented as: wherein, is the real-time recorded sample of the ith fault recorder within a period of time after the fault occurs, is the distributed fault recording sample recorded by the n fault recorders in real time; According to the traveling wave reflection network diagram, the distance of different distributed fault recorders from the fault point is obtained: wherein, is the real-time sample recorded by the i-th wave recorder after the fault occurs, is the distance between the installation point of the i-th wave recorder on the overhead line and the fault point, is the distance between the installation point of the i-th wave recorder on the overhead line and the fault point, is the distributed fault wave recording sample recorded by the n fault wave recorders in real time; In step 8, the reliability of the recording data of different recorders is obtained by processing the real-time recording of the distributed fault recording sample of n fault recorders by the method of step 2; In step 8, the weight of different recorders is obtained by processing the reliability of the recording data of different recorders by the method of step 3; In step 8, further, based on the different weights assigned to different recorders, the real-time recording fusion waveform data is obtained by the method of step 4; In Step 8, the step of distinguishing the fault type is performed by mapping the extracted fault feature set to the fault feature set of different types of faults in the constructed random forest model and the mapping relationship between the fault type , to obtain the type of the fault.

2. The overhead power transmission line fault diagnostic method of claim 1, characterized in that: The historical record of distributed fault recording sample in step 1 is: wherein, is the sample recorded by the i-th wave recorder in a period of time after the fault occurs, is the distance between the installation point and the fault point on the overhead line for the i-th wave recorder, and n is the total number of wave recorders installed on the overhead transmission line, represents the fault type, and the value range is } wherein is a lightning strike fault, is a bird damage fault, is a forest fire fault, is an external force damage fault; In step 1, the sample recorded by each recorder is represented as: wherein, V is the voltage in the sample recorded by the i-th recorder, I is the current in the sample recorded by the i-th recorder; In step 1, the matrix of fault waveform data recorded by all recorders is constructed as: wherein, V is the voltage in the sample recorded by the i-th wave recorder, I is the current in the sample recorded by the i-th wave recorder, Di is the distance between the installation point and the fault point on the overhead line for the i-th wave recorder. The data recorded at the moment of fault occurrence is: wherein is a matrix is the value at t = 0, is the voltage instantaneous value recorded by the i-th recorder at t = 0, is the current instantaneous value recorded by the i-th recorder at t = 0.

3. The overhead transmission line fault diagnosis method according to claim 2, characterized in that: Step 2 is as follows: The elements of the matrix are respectively processed by forward processing to obtain the forward-processed matrix The voltage is an interval index, and the positive processing method is: wherein, is the intermediate transition amount, and is the maximum and minimum voltage that the fault recorder can record, is the voltage instantaneous value recorded by the i-th recorder at t = 0, is the result of the forward processing of the voltage instantaneous value of the i-th recorder at t = 0. The current is an interval index, and the positive processing method is: wherein, is the intermediate transition amount, and are the maximum and minimum current values, respectively, that the fault recorder can record, is the current instantaneous value recorded by the i-th recorder at t = 0, is the result of the forward transformation of the current instantaneous value recorded by the i-th recorder at t = 0. The distance between the installation point of the recorder on the overhead line and the fault point is a minimum index, and the positive processing method is: wherein, is the result of the forward processing of the distance between the installation point and the fault point on the overhead line for the recorder, is the distance between the installation point and the fault point on the overhead line for the i-th recorder; matrix after forward is: In step 2, the matrix after normalization is used The maximum vector is found And the minimum vector : wherein, is the maximum vector, is the minimum vector, is the result of the forward transformation of the voltage instantaneous value of the i-th wave recorder at time t = 0, is the result of the forward transformation of the current instantaneous value of the i-th wave recorder at time t = 0, is the result of the forward transformation of the distance between the installation point and the fault point of the i-th wave recorder on the overhead line, max indicates the maximum processing, and min indicates the minimum processing; In step 2, the data of the i-th wave recorder after normalization is respectively subtracted from the maximum vector and the minimum vector ​ wherein, is the maximum vector, is the minimum vector, is the result of the forward transformation of the voltage instantaneous value of the i-th wave recorder at t=0, is the result of the forward transformation of the current instantaneous value of the i-th wave recorder at t=0, is the result of the forward transformation of the distance between the installation point and the fault point of the wave recorder on the overhead line, and is the vector obtained after the difference, is the difference between the result of the forward transformation of the voltage instantaneous value of the i-th wave recorder at t=0 and the maximum of the results of the forward transformation of the voltage instantaneous values of the n wave recorders at t=0, is the difference between the result of the forward transformation of the current instantaneous value of the i-th wave recorder at t=0 and the maximum of the results of the forward transformation of the current instantaneous values of the n wave recorders at t=0, is the difference between the result of the forward transformation of the distance between the installation point and the fault point of the i-th wave recorder on the overhead line and the maximum of the results of the forward transformation of the distances between the installation point and the fault point of the n wave recorders on the overhead line, is the difference between the result of the forward transformation of the voltage instantaneous value of the i-th wave recorder at t=0 and the minimum of the results of the forward transformation of the voltage instantaneous values of the n wave recorders at t=0, is the difference between the result of the forward transformation of the current instantaneous value of the i-th wave recorder at t=0 and the minimum of the results of the forward transformation of the current instantaneous values of the n wave recorders at t=0, is the difference between the result of the forward transformation of the distance between the installation point and the fault point of the i-th wave recorder on the overhead line and the minimum of the results of the forward transformation of the distances between the installation point and the fault point of the n wave recorders on the overhead line. in, This represents the reliability of the data recorded by the i-th waveform recorder. The vector obtained after subtraction elements, The difference between the forward-processed instantaneous voltage value of the i-th recorder at t=0 and the maximum value among the forward-processed instantaneous voltage values ​​of the n-th recorders at t=0 is given. The difference between the forward-processed instantaneous current value of the i-th recorder at t=0 and the maximum value among the forward-processed instantaneous current values ​​of the n-th recorders at t=0 is given. Let be the difference between the forward-processed result of the distance between the installation point of the i-th recorder and the fault point on the overhead line and the maximum value among the forward-processed results of the distances between the installation points of the n recorders and the fault points on the overhead line. The vector obtained after subtraction elements, The difference between the minimum value of the instantaneous voltage at time t=0 of the i-th recorder after positive transformation and the minimum value of the instantaneous voltage at time t=0 of the n-th recorder. The difference between the minimum value of the instantaneous current at time t=0 of the i-th recorder after positive transformation and the minimum value of the instantaneous current at time t=0 of the n-th recorder after positive transformation. The difference between the minimum value of the forward-processed distance between the installation point and the fault point of the i-th recorder on the overhead line and the forward-processed distance between the installation points and the fault points of the n recorders on the overhead line.

4. The overhead transmission line fault diagnosis method according to claim 3, characterized in that: The weight assigned to the recording data of the recorder according to the reliability criterion in step 3 is represented as: wherein, represents the weight of the recording data recorded by the i-th recorder; In step 3, the set of weight values of the recorded wave data is assigned according to the credibility criterion: wherein, represents the weight of the recording data recorded by the i-th recorder, represents the reliability of the data recorded by the j-th recorder, and n represents the total number of recorders.

5. The overhead transmission line fault diagnosis method according to claim 4, characterized in that: The weighted processing of the distributed fault recorded wave data of the historical record in step 4 obtains the fusion waveform data of the historical record: wherein, the fused waveform data obtained by the representative data weighting processing, representative data of the i-th recording waveforms, a weight of the recording waveform data recorded by the i-th recording wave device, a sample recorded by the i-th recording wave device.

6. The overhead transmission line fault diagnosis method according to claim 5, characterized in that: The specific steps of extracting the fault characteristic quantity in step 5 are: According to the obtained fusion waveform data, the wave head time and the wave tail time can be directly obtained from the fusion waveform data, and the ground resistance corresponding to the fusion waveform data is calculated through the equivalent circuit method, the Fourier decomposition of the fusion waveform data is performed to obtain the harmonic content and other electrical characteristic quantities, and the extracted fault characteristic quantity is represented as: wherein, and respectively represent a wave head time and a wave tail time of the fused waveform data, represents a ground resistance corresponding to the fused waveform data, represents a frequency spectrum obtained by performing Fourier decomposition on the fused waveform data, represents the jth feature quantity extracted from the fused waveform data, and p-1 represents the number of feature quantities extracted from the fused waveform data.

7. The overhead transmission line fault diagnosis method according to claim 6, characterized in that: The diagnostic model between the fault feature quantity and the fault type of the power transmission line is constructed by using the random forest algorithm in step 6, and the diagnostic model reflects the mapping relationship between the fault feature quantity set of different types of faults and the fault type and the fault type The fault feature quantity set of k different types of faults and the fault type constitute an active data matrix: wherein, a source data matrix constituted by a set of fault feature quantities of different types of faults and fault types, , , , a wave head time, a wave tail time, a ground resistance corresponding to the fusion waveform data, and a frequency spectrum obtained by Fourier decomposition of the fusion waveform data of the fault feature quantity of the ithdata, respectively, represents the jthfeature quantity extracted from the ithdata, and p-1 represents the number of feature quantities extracted from the fusion waveform data, a fault type of the ithdata; wherein the definition is: wherein, is a matrix of fault feature quantities of k faults, , , , is a wave head time, a wave tail time, a ground resistance corresponding to the fusion waveform data, and a frequency spectrum obtained by Fourier decomposition of the fusion waveform data, respectively, of the fault feature quantity of the i-th data, represents the j-th feature quantity extracted from the i-th data, and p-1 represents the number of feature quantities extracted from the fusion waveform data, is a matrix of k fault types, is a fault type of the i-th data, and p-1 represents the number of fault feature quantities; In step 6, from randomly generated sub-matrices , , , wherein, denotes the i-th sub-matrix generated randomly, is a matrix of fault feature quantities of k faults, is a matrix of k fault types, is a sub-matrix of the fault feature quantity set, is a sub-matrix of the fault type matrix denotes the mathematical symbol of the block matrix, which divides the matrix into left and right two matrices;​ In the above formulae, , , and satisfy the constraint: wherein k represents the set of fault characteristic quantities of different types of faults of the historical record and the number of fault types, p-1 is the number of types of fault characteristic quantities, m represents the number of randomly sampled sets of fault characteristic quantities of different types of faults of k historical records and fault types, and q-1 represents the number of types of fault characteristic quantities randomly sampled from p-1 types of fault characteristic quantities; In step 6, through this A subset is trained using a random forest to obtain A decision tree; In step 6, the historical fault recording data is processed according to steps 2-5, and the processed data is input into the random forest of trees to obtain fault type results. The fault type with the largest number of counts is taken as the final prediction result.

8. A computer readable medium characterized by The storage electronic device executes a computer program, when the computer program runs on the electronic device, so that the electronic device executes the steps of the method according to any one of claims 1-7.

Images