A line fault intelligent diagnosis method and system

Abstract

The present invention relates to the field of power grid technology, and in particular to a line fault intelligent diagnosis method and system, comprising: S00: a line library stores line information of each line section; a judgment factor library is established to store fault related information; S10: when a line fault occurs, fault traveling waves and recorded wave information are recorded; the line information is converted into line fault information and stored in the judgment factor library; S20: a fault waveform identification model is constructed to obtain fault information; S30: the line section where the fault point is located is calculated; S40: monitoring data of the line section is obtained, and information is updated irregularly and input into the judgment factor library, corresponding to the line fault information; S50: an intelligent diagnosis model is constructed to diagnose factors in the judgment factor library to obtain line fault diagnosis results; S60: monitoring personnel send the results to operation and maintenance personnel for fault handling and inspection; S70: the handling and inspection results are transmitted back and stored in a historical fault sample library together with the factors; the present invention effectively improves the accuracy of line fault diagnosis.

Description

Technical Field

[0001] The present invention relates to the field of power grid technology, and in particular to a line fault intelligent diagnosis method and system. Background Art

[0002] With the development of social economy, electricity has become one of the important energy sources that are indispensable to modern society; the power grid, as the main carrier of power transmission, undertakes the task of efficiently and reliably transmitting the electricity produced by power plants to thousands of households; the safe and stable operation of power grid lines directly affects the quality and reliability of power supply, and thus affects all aspects of the national economy and social life; in recent years, with the development of smart grid technology, more and more advanced technologies and equipment have been applied to the monitoring and maintenance of power grid lines, greatly improving the level of intelligence of power grid operation.

[0003] Therefore, ensuring the normal operation of power grid lines and timely discovering and eliminating potential faults are crucial to ensuring the safety and stability of the power system; however, traditional line fault inspections mainly rely on on-site operators to use "drone + manual" inspection methods to check discharge points, combined with the operators' fault screening experience to conduct a closed-loop line fault elimination. However, this method is too dependent on the experience of on-site operators and is restricted by data collection conditions. It has limitations in terms of fault location accuracy and fault type identification accuracy, making it impossible for operation and maintenance personnel to deal with faults in a timely and effective manner, and is often affected by bad weather, making it difficult to guarantee timeliness.

[0004] Therefore, there is an urgent need for a method that can automatically and accurately detect and diagnose faults in order to improve the operation and maintenance efficiency and reliability of power grid lines. Summary of the invention

[0005] In view of at least one of the above technical problems, the present invention provides a line fault intelligent diagnosis method and system, which can effectively improve the accuracy of line fault diagnosis.

[0006] According to a first aspect of the present invention, there is provided a line fault intelligent diagnosis method, comprising the following steps:

[0007] S00: Establish a line library, record the line between two adjacent towers as a line section, and store basic information of each line section in the line library;

[0008] Real-time detection of current and voltage waveforms and various tower monitoring data of each line section, input into the line library, and store in correspondence with each corresponding line section;

[0009] The relevant account information, environmental information and operation and maintenance information of each line section are stored in the line library in correspondence with the corresponding line sections, forming irregularly updated information;

[0010] Establish a judgment factor library to store fault-related information;

[0011] S10: When a line fault occurs, the time when the traveling wave reaches different line sections and the characteristic information of the traveling wave are recorded as the fault traveling wave information; the fault recording wave and the line section where it occurs are recorded as the fault recording wave information; and the line information corresponding to the line fault stored in the line library is converted into line fault information and stored in the judgment factor library; the line information where the fault occurs and its corresponding basic information are stored as the line fault information of this time in the judgment factor library;

[0012] S20: construct a fault waveform identification model based on the fault recording information and the fault traveling wave information, obtain the fault time, fault distance, fault insulation characteristics and waveform characteristic parameter information, and store them in the judgment factor library corresponding to the line fault information;

[0013] S30: Calculate the line section where the fault point is located according to the fault recording and record it as the simulated line section, and determine whether the simulated line section is consistent with the line section where the fault occurs in the fault recording information; if they are consistent, proceed to the subsequent steps;

[0014] S40: Acquire various tower monitoring data of the line section where the line fault point is located, update the information irregularly and input it into the analysis factor library, and correspond it with the line fault information;

[0015] S50: setting the line fault information in the judgment factor library and the fault time, fault distance, fault insulation characteristics, waveform characteristic parameter information, various tower monitoring data, and irregular update information corresponding to the line fault information as factors; setting a same initial weight for each factor, multiplying the weight value of each factor in this line fault with the weight value of the factor stored in the judgment factor library to obtain each sub-factor, constructing an intelligent diagnosis model to diagnose the sub-factor, and obtaining the line fault diagnosis result;

[0016] S60: The monitoring personnel of the monitoring line sends the line fault diagnosis result to the mobile app of the line operation and maintenance personnel; the operation and maintenance personnel handle the fault and conduct inspection according to the line fault diagnosis result;

[0017] S70: The fault handling situation and inspection results of the operation and maintenance personnel are transmitted back to the monitoring personnel in real time; based on the fault handling situation and inspection results, the monitoring personnel verify and confirm the accuracy of the line fault diagnosis results, and store the factors and actual situation of this line fault in the historical fault sample library.

[0018] Furthermore, the method further includes step S80: according to the historical fault sample library, using a multimodal learning method, dynamically calculating the weights of the factors in the analysis and judgment factor library, and forming factor correlation rule criteria for various fault types.

[0019] Furthermore, in step S80, the weights of the factors in the dynamic judgment factor library are specifically calculated as follows:

[0020] After each new fault sample is stored in the historical fault sample library, the weights of the factors related to the new fault sample in the judgment factor library are dynamically updated;

[0021] Set an initial weight W(1) for each factor; the initial weight of the i-th factor is Wi(1);

[0022] Set Wi(t+1)=Wi(t)+ΔWi;

[0023] Where Wi(t+1) is the new weight of factor i after the t+1th update, Wi(t) is the old weight of factor i after the tth update, and ΔWi is the weight increment of factor i.

[0024] Furthermore, in step S80, the weight increment ΔWi is calculated as follows:

[0025] When the diagnosis is correct, ΔWi = α × Si;

[0026] When the diagnosis is wrong, ΔWi=(-β)×Si;

[0027] Among them, α and β are update rates, which are artificially set values ​​and are constants; Si is the importance score of factor i.

[0028] Furthermore, in step S20, the fault waveform identification model is constructed as follows:

[0029] Firstly, pre-process the fault recording information and fault traveling wave information;

[0030] Extract characteristic parameters of fault waveform from preprocessed data in three aspects: time domain, frequency domain and time-frequency domain;

[0031] The extracted feature parameters are then input into the shared hidden layer to capture the features among the fault time, fault distance, fault insulation characteristics and waveform feature parameter information, and the features are fused to form a universal feature vector output;

[0032] Finally, the universal feature vector is input into multiple output layers, and the fault time, fault distance, fault insulation characteristics and waveform characteristic parameter information are refined and outputted respectively according to the universal feature vector.

[0033] Furthermore, in step S20, establishing a shared hidden layer is specifically as follows:

[0034] Establish the first layer of feature fusion, from the time domain f t , frequency domain f f and the time-frequency domain f tfThe extracted features are initially integrated to form a feature vector f1=Concat(f t , f f );

[0035] Establish the second layer of feature fusion and introduce a weighting mechanism. Different features have different contributions when fused; form a feature vector f2=Concat(W1·f1,W2·f tf ), where W1 and W2 are artificially set weight coefficients and are constants;

[0036] The third layer of feature fusion is established, and adaptive feature fusion is introduced. The importance of features is dynamically adjusted according to the current features to adapt to the identification of fault waveforms, forming a feature vector f3=Concat(f2,g); where g is an adaptive function, which is a linear function set artificially;

[0037] The final feature vector f3 is the feature vector of the shared hidden layer.

[0038] Furthermore, in step S50, the smart diagnosis model is specifically:

[0039] Type identification level: Input the line fault information and the factors consisting of the fault time, fault distance, fault insulation characteristics, waveform characteristic parameter information, various tower monitoring data, and irregularly updated information corresponding to the line fault information, and extract multiple sub-sample sets from the factors by replacement sampling method; compare the information in each sub-sample set with the manually set fault type to obtain the most similar fault type for each sub-sample set; select the fault type with the largest proportion from all sub-sample sets in proportion, so as to determine the line fault type and cause;

[0040] Position estimation level: input fault waveform information; filter the current and voltage waveforms between the line sections where the fault point is located according to the passage of time, and record the time of the current and voltage waveform mutation; calculate the mutation time from the time of the current and voltage waveform mutation to the initial time of the line entering the line section where the fault point is located, and then calculate the total time of the line entering and flowing out of the line section where the fault point is located, and finally obtain the proportion of the mutation time to the total time, so as to obtain the proportion of the length of the line section where the fault point is located, and determine the specific location of the line fault;

[0041] Comprehensive diagnosis level: receives the line fault type and cause determined by the type identification level, and the specific location of the line fault determined by the location estimation level, and comprehensively explains the final fault diagnosis result.

[0042] Furthermore, in step S50, the intelligent diagnosis model will periodically reference new fault samples in the historical fault sample library to update the model.

[0043] Furthermore, in step S30, the line section where the fault point is located is calculated as follows:

[0044] Set the nth pair of towers T n With T n+1 The distance between the fault point and the tower T1 in the fault recording wave is dn; the fault distance between the fault point and the tower T1 in the fault recording wave is Df;

[0045] Calculate the cumulative distance Dc=d1+d2+…+dn;

[0046] When Dc accumulates to the i-th pair of towers T i With T i+1 , and Dc ≥ Df, stop calculating the cumulative distance Dc, and record the fault point at the i-th pair of towers T at this time i With T i+1 Thus, the line section T where the fault point is located can be obtained. i With T i+1 between.

[0047] According to a second aspect of the present invention, there is also provided a line fault intelligent diagnosis system, comprising:

[0048] Data management module: including line library, which is used to store line information, current and voltage waveforms, various tower monitoring data, and irregularly updated information of each line section; including analysis factor library, which is used to store information related to faults; including historical fault sample library, which is used to store factors of historical line faults and their corresponding real situations;

[0049] Fault collection module: record the time when the traveling wave arrives at different line sections to obtain fault traveling wave information; obtain fault recording information based on line information; and convert the line information corresponding to the line fault stored in the line library into line fault information and store it in the judgment factor library; obtain various tower monitoring data of the line section where the line fault point is located, update the information irregularly and input it into the judgment factor library, and correspond it with the line fault information;

[0050] Fault detection module: obtain fault time, fault distance, fault insulation characteristics and waveform characteristic parameter information based on the fault waveform identification model; calculate the line section where the fault point is located based on the tower span information;

[0051] Fault diagnosis module: builds an intelligent diagnosis model to diagnose and judge the factors in the factor library to obtain line fault diagnosis results;

[0052] Communication feedback module: used for communication between monitoring personnel and operation and maintenance personnel.

[0053] The beneficial effects of the present invention are as follows: by collecting historical fault information, extracting fault characteristic parameters from three aspects, namely, time domain, frequency domain and time-frequency domain, and establishing a fault waveform identification model, the characteristic waveform data of each type of fault is obtained, and the cause of the line fault is accurately analyzed and located by combining the factor correlation rule judgment criteria of the fault type; by effectively linking the line fault diagnosis results with fault handling and inspection, the line fault is reviewed in real time and incorporated into the historical fault sample library, and the multimodal learning method is used to dynamically calculate the factor weights in the analysis and judgment factor library, so as to further improve the accuracy and reliability of line fault diagnosis. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0055] Figure 1 Schematic diagram of the steps of the line fault intelligent diagnosis method of the present invention. DETAILED DESCRIPTION

[0056] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments.

[0057] In the description of the present invention, it should be noted that the orientations or positional relationships indicated by “center”, “up”, “down”, “left”, “right”, “vertical”, “horizontal”, “inside” and “outside”, etc., are based on the orientations or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the referred device or element must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation on the present invention.

[0058] In the description of the present invention, it should be noted that, unless otherwise clearly specified and limited, the terms "installed", "connected", and "connected" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection or an indirect connection through an intermediate medium, or it can be the internal communication of two components. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0059] A line fault intelligent diagnosis method, such as Figure 1 As shown, the following steps are included:

[0060] S00: Establish a line library, record the line between two adjacent towers as a line section, and store basic information of each line section in the line library; the basic information will include the line name, phase mark, tower span information and corresponding voltage level of each line section. These basic information are fixed and not updated;

[0061] Real-time detection of current and voltage waveforms of each line section and various tower monitoring data, input into the line library, and store them in correspondence with the corresponding line sections, which are updated in real time;

[0062] The relevant account information, environmental information and operation and maintenance information of each line section are stored in the line library in correspondence with the corresponding line sections, forming irregularly updated information, which will be updated once after each line failure and operation and maintenance;

[0063] Among them, the ledger information includes but is not limited to the tower location, tower height, tower type, line design parameters, conductor arrangement, phase sequence, line direction, insulator string suspension method, insulator material type, insulator operation time, whether the insulator anti-pollution flashover material is sprayed, line commissioning time, line rated current, and line parameter information; environmental information includes but is not limited to terrain, landform, vegetation information, and disaster distribution map information of the faulty line section (the disaster distribution map is divided into bird damage, ice cover, wind damage, filth, mountain fire, lightning damage, and dancing); operation and maintenance information includes but is not limited to line defect library, channel hidden danger library, tower grounding resistance information, historical fault information, and work maintenance plan information;

[0064] Establishing an analysis and judgment factor library to store fault-related information is equivalent to extracting fault-related information from a large amount of data in the line library and storing it in the analysis and judgment factor library for subsequent intelligent diagnosis of line faults.

[0065] S10: When a line fault occurs, a special band that is significantly different from normal transmission will be generated at the fault location, and the band will travel along the line to form a traveling wave. The sensor installed on the tower will detect the arrival time of the traveling wave, thereby recording the time when the traveling wave arrives at different line sections, and adding the characteristic information of the traveling wave, such as the amplitude, waveform, propagation speed and polarity of the traveling wave, as the fault traveling wave information. There are usually tools such as fault recorders on the tower, which can automatically and accurately record the changes in various electrical quantities before and after the fault when a system fault occurs, and convert these data into fault recordings to transmit them. The fault recording and the line section where it occurs are recorded as fault recording information. At the same time, the faulty line information and its corresponding basic information are stored in the judgment factor library as the line fault information of this time; for example, if a fault occurs in the line between towers 4 and 5, the line section between towers 4 and 5 is the corresponding faulty line section. At this time, the basic information of the faulty line section will be stored in the judgment factor library, such as the line name, phase mark, tower spacing information and corresponding voltage level between towers 4 and 5, which will be recorded as the line fault information of this fault.

[0066] S20: According to the fault recording information and the fault traveling wave information, a fault waveform identification model is constructed to obtain the fault time, fault distance, fault insulation characteristics and waveform characteristic parameter information, and store them in the judgment factor library corresponding to the line fault information. For example, continuing with the previous example, at this time, the line fault information corresponding to the No. 4~5 towers is stored in the judgment factor library, and the fault time, fault distance, fault insulation characteristics and waveform characteristic parameter information obtained in this step are stored;

[0067] Among them, the fault insulation characteristics include metallicity and high resistance; the waveform characteristic parameter information includes but is not limited to steepness, wave head time, half-peak time, and wave tail time.

[0068] S30: Calculate the line section where the fault point is located based on the fault recording and record it as the simulated line section, and then determine whether the simulated line section is consistent with the line section where the fault occurs in the fault recording information; if they are consistent, it means that the calculated fault point is consistent with the actual detection, which indirectly proves that the fault recording is not distorted during the transmission process, and this data can be used to proceed to the subsequent steps; if they are inconsistent, it means that the fault recording may have been interfered with during transmission, resulting in the calculation result being inconsistent with the actual detection result. At this time, a corresponding processing method needs to be adopted, such as: issuing a data distortion alarm and stopping subsequent steps; or the interference of the fault recording can be corrected according to the calculation result and the actual detection error, and then proceed to the subsequent steps.

[0069] S40: Obtain various tower monitoring data and irregularly updated information of the line section where the line fault point is located, input them into the judgment factor library, and correspond them with the line fault information. For example, continuing with the previous example, at this time, the line fault information corresponding to towers 4 to 5 is stored in the judgment factor library, and then the monitoring data and irregularly updated information corresponding to towers 4 to 5 are stored;

[0070] Among them, various types of pole tower monitoring data include: image data, conductor and hardware temperature measurement data, leakage current monitoring data, icing monitoring data, dancing monitoring data, conductor sag data, meteorological data, conductor vibration data, pole tower tilt and deformation monitoring data and wind deviation monitoring data; image data include but are not limited to drone images and channel monitoring device images, meteorological data include but are not limited to lightning monitoring data, meteorological disaster warning information, micrometeorological monitoring data, meteorological station monitoring data, weather forecast data.

[0071] S50: setting the line fault information in the judgment factor library and the fault time, fault distance, fault insulation characteristics, waveform characteristic parameter information, various tower monitoring data, and irregular update information corresponding to the line fault information as factors; setting a same initial weight for each factor, multiplying the weight value of each factor in this line fault by the weight value of the factor stored in the judgment factor library to obtain each sub-factor, constructing an intelligent diagnosis model to diagnose the sub-factor, and the sub-factor with a large weight value will be diagnosed first to obtain the line fault diagnosis result;

[0072] S60: The monitoring personnel of the monitoring line will couple the line fault diagnosis results with the application planning and scheduling capabilities, combine the fault diagnosis results with the task scheduling of the operation and maintenance personnel, and intelligently dispatch auxiliary work orders (a work order refers to a document or electronic record used to record and track work tasks or requests) to the mobile app of the line operation and maintenance personnel to improve response speed and work efficiency; the operation and maintenance personnel will handle the faults and conduct inspections according to the line fault diagnosis results; form a timely and effective linkage between fault diagnosis and handling work orders, and improve the intelligence level of fault emergency response and emergency repair.

[0073] S70: The fault handling situation and inspection results of the operation and maintenance personnel are transmitted back to the monitoring personnel in real time; based on the fault handling situation and inspection results, the monitoring personnel verify and confirm the accuracy of the line fault diagnosis results, and store the factors and actual situation of this line fault in the historical fault sample library.

[0074] The method further includes step S80: dynamically calculating the weights of factors in the analysis and judgment factor library based on the historical fault sample library using a multimodal learning method to form factor correlation rule criteria for various fault types;

[0075] Among them, the fault types include: lightning strike, external damage during construction, external damage by foreign objects, bird damage, wind deflection, dancing, pollution flashover, ice flashover, wildfire, tree line, and main body;

[0076] By dynamically calculating the weights of factors in the factor library, the actual proportion of different factors in fault diagnosis can be more accurately reflected; this means that factors that have a greater impact on the diagnosis results will receive higher weights, thereby improving the accuracy of the diagnosis results; and multimodal learning methods can extract features from different types of data and dynamically adjust factor weights based on information in the historical fault sample library, which enables the diagnosis system to better adapt to different fault types and environmental changes; by continuously updating and optimizing factor weights, the diagnosis system can reduce misdiagnosis caused by certain factors being too heavy or too light, thereby improving the reliability of line fault diagnosis; more accurate line fault diagnosis results can help operation and maintenance personnel locate fault points more efficiently, reduce unnecessary inspection scopes, and optimize the allocation of human and material resources.

[0077] In step S80, the weights of the factors in the dynamic judgment factor library are specifically calculated as follows:

[0078] After each new fault sample is stored in the historical fault sample library, the weights of the factors related to the new fault sample in the judgment factor library are dynamically updated, so that the diagnosis system can continuously optimize its own diagnosis capabilities according to the new fault samples;

[0079] Set an initial weight W(1) for each factor; the initial weight of the i-th factor is Wi(1);

[0080] Set Wi(t+1)=Wi(t)+ΔWi;

[0081] Where Wi(t+1) is the new weight of factor i after the t+1th update, Wi(t) is the old weight of factor i after the tth update, and ΔWi is the weight increment of factor i.

[0082] Preferably, in step S80, the weight increment ΔWi is calculated as follows:

[0083] When the diagnosis is correct, ΔWi = α × Si;

[0084] When the diagnosis is wrong, ΔWi=(-β)×Si;

[0085] By increasing the weight of important factors in the correct diagnosis when the diagnosis is correct, and reducing the weight of important factors in the wrong diagnosis when the diagnosis is wrong, and dynamically calculating the weight of factors in the judgment factor library after each new fault sample is obtained, the diagnosis system can better identify faults and provide more accurate line fault diagnosis results;

[0086] Among them, α and β are update rates, which are artificially set values ​​and are constants. They are used to control the speed of weight update to prevent the weight update from being too fast or too slow, so as to find a suitable balance point so that the diagnosis system can enter a faster and better state for the line fault type.

[0087] Si is the importance score of factor i; it can be obtained by measuring the correlation between the factor and the line fault diagnosis result through statistical methods, or by setting it after human experience and knowledge judgment, or by feature selection algorithm and other methods.

[0088] In step S20, the fault waveform identification model is constructed as follows:

[0089] Firstly, the fault recording information and fault traveling wave information are preprocessed to remove noise, smooth the signal, and improve the accuracy of subsequent feature extraction;

[0090] The pre-processed data can be used to extract 38 characteristic parameters of the fault waveform from the time domain, frequency domain and time-frequency domain, and stored in the established line fault transient waveform database to capture fault information from different aspects and improve the comprehensiveness and accuracy of diagnosis;

[0091] The extracted feature parameters are then input into a shared hidden layer, which consists of a series of neural network layers, such as fully connected layers, convolutional layers, etc., to capture the features between fault time, fault distance, fault insulation characteristics and waveform feature parameter information, and perform feature fusion to form a common feature vector output from data features from different sources or different processing stages;

[0092] Finally, the universal feature vector is input into multiple output layers, and the information specific to each output layer is refined and extracted according to the universal feature vector, and the fault time, fault distance, fault insulation characteristics and waveform characteristic parameter information are output in a targeted manner.

[0093] Preferably, in step S20, establishing a shared hidden layer is specifically as follows:

[0094] Establish the first layer of feature fusion, from the time domain f t , frequency domain f f and the time-frequency domain f tf The extracted features are first extracted in the time domain f t and frequency domain f f In terms of information features, the information features are initially integrated to form a feature vector f1=Concat(f t , f f );

[0095] Establish the second layer of feature fusion and introduce the weighting mechanism. Different features have different contributions when fused. tfIt is a comprehensive perspective to describe the characteristics of the signal. It takes into account the changes of the signal in both time and frequency dimensions. tf The first layer has been integrated with the time domain f t and frequency domain f f The feature vector f1 is weighted to form a feature vector f2 = Concat(W1·f1,W2·f tf ), so that the information obtained from the three aspects is more comprehensive, and the importance of different features can be adjusted by changing the weights; where W1 and W2 are artificially set weight coefficients and are constants;

[0096] The third layer of feature fusion is established, and adaptive feature fusion is introduced. The importance of features is dynamically adjusted according to the current features to adapt to the identification of fault waveforms, forming a feature vector f3=Concat(f2,g). The adaptive function added in this layer can dynamically adjust features according to the needs of current fault waveform identification, so that the fault waveform identification model can better capture key features; where g is an adaptive function, which is a linear function set manually, and the linear function g can be obtained through manual training;

[0097] The final feature vector f3 is the feature vector of the shared hidden layer.

[0098] Preferably, in step S50, the smart diagnosis model is specifically:

[0099] Type identification level: Input the factors consisting of the line fault information and the fault time, fault distance, fault insulation characteristics, waveform characteristic parameter information, various tower monitoring data, and irregularly updated information corresponding to the line fault information, and extract multiple sub-sample sets from the factors by replacement sampling. Factors with large weights are easier to be extracted than factors with small weights, forming a situation where factors with large weights are preferentially involved in diagnosis; compare the information in each sub-sample set with the manually set fault type to obtain the most similar fault type for each sub-sample set; select the fault type with the largest proportion from all sub-sample sets in proportion, so as to determine the line fault type and cause;

[0100] For example, there are five sub-sample sets, A, B, C, D, and E, and each sub-sample set contains two factors extracted by replacement sampling (i.e., line fault information and the fault time, fault distance, fault insulation characteristics, waveform characteristic parameter information, various tower monitoring data, and two extracted from irregularly updated information corresponding to the line fault information); sub-sample set A is compared with the set fault type, and the most similar to sub-sample set A is type one; sub-sample set B is compared with the set fault type, and the most similar to sub-sample set B is type two; repeating this operation, it is obtained that the most similar to sub-sample set C is type three, the most similar to sub-sample set D is type one, and the most similar to sub-sample set E is type one; then among the five sub-sample sets, the proportion of fault type one obtained by comparison is three-fifths, which is the largest, so it is judged that the line fault type is type one.

[0101] Position estimation level: input fault waveform information; filter the current and voltage waveforms between the line sections where the fault point is located according to the passage of time, and record the time of the current and voltage waveform mutation; calculate the mutation time from the time of the current and voltage waveform mutation to the initial time of the line entering the line section where the fault point is located, and then calculate the total time for the line to enter and flow out of the line section where the fault point is located, and finally obtain the proportion of the mutation time to the total time, so as to obtain the proportion of the length of the line section where the fault point is located, and judge the specific position of the line fault; the specific position here refers to the specific position of the fault point in the line between a pair of towers, rather than calculating which pair of towers the line fault is located between in the entire line as in step S30.

[0102] Comprehensive diagnosis level: Receives the line fault type and cause determined by the type identification level, and the specific location of the line fault determined by the location estimation level, and comprehensively explains the final fault diagnosis results, making it convenient for operation and maintenance personnel to go to the site to handle the fault and conduct inspections based on the fault diagnosis results.

[0103] Preferably, in step S50, the intelligent diagnosis model will periodically reference new fault samples in the historical fault sample library to update the model; with the aging of power grid equipment, changes in environmental conditions and other factors, new fault types may continue to appear. By periodically referencing the latest fault samples, the intelligent diagnosis model can learn the types and characteristics of the most recent faults. The latest data can usually better reflect the current fault situation, thereby improving the accuracy of diagnosis.

[0104] In step S30, the line section where the fault point is located is calculated as follows:

[0105] Set the nth pair of towers T n With T n+1By reading the information in the fault recording, it can be obtained that the fault distance between the fault point and the tower T1 recorded in the fault recording is Df, that is, the distance between the fault point and the first tower;

[0106] Calculate the cumulative distance Dc=d1+d2+…+dn;

[0107] When calculating, n starts from 1 and gradually increases. When Dc accumulates to the i-th pair of towers T i With T i+1 , and Dc ≥ Df, stop calculating the cumulative distance Dc, and record the fault point at the i-th pair of towers T at this time i With T i+1 Thus, the line section T where the fault point is located can be obtained. i With T i+1 between;

[0108] The calculated fault point is located at T i ~T i+1 between line sections, which can then be compared with the T at the actual fault point. x ~T x+1 Perform a comparison to determine whether the fault recording is distorted.

[0109] The present invention also provides a line fault intelligent diagnosis system, comprising:

[0110] Data management module: including line library, which is used to store line information, current and voltage waveforms, various tower monitoring data, and irregularly updated information of each line section; including analysis factor library, which is used to store information related to faults; including historical fault sample library, which is used to store factors of historical line faults and their corresponding real situations;

[0111] Fault collection module: record the time when the traveling wave arrives at different line sections to obtain fault traveling wave information; obtain fault recording information based on line information; and convert the line information corresponding to the line fault stored in the line library into line fault information and store it in the judgment factor library; obtain various tower monitoring data of the line section where the line fault point is located, update the information irregularly and input it into the judgment factor library, and correspond it with the line fault information;

[0112] Fault detection module: obtain fault time, fault distance, fault insulation characteristics and waveform characteristic parameter information based on the fault waveform identification model; calculate the line section where the fault point is located based on the tower span information;

[0113] Fault diagnosis module: builds an intelligent diagnosis model to diagnose and judge the factors in the factor library to obtain line fault diagnosis results;

[0114] Communication feedback module: used for communication between monitoring personnel and operation and maintenance personnel.

[0115] The above shows and describes the basic principles, main features and advantages of the present invention. It should be understood by those skilled in the art that the present invention is not limited to the above embodiments. The above embodiments and descriptions are only for explaining the principles of the present invention. Without departing from the spirit and scope of the present invention, the present invention may have various changes and improvements, which fall within the scope of the present invention to be protected. The scope of protection of the present invention is defined by the attached claims and their equivalents.

Claims

1. A line fault intelligent diagnosis method, characterized in that: The steps include: S00: Establish a line library, record the line between two adjacent towers as a line section, and store basic information of each line section in the line library; Real-time detection of current and voltage waveforms and various tower monitoring data of each line section, input into the line library, and store in correspondence with each corresponding line section; The relevant account information, environmental information and operation and maintenance information of each line section are stored in the line library in correspondence with the corresponding line sections, forming irregularly updated information; Establish a judgment factor library to store fault-related information; S10: When a line fault occurs, the time when the traveling wave arrives at different line sections and characteristic information of the traveling wave are recorded as fault traveling wave information; The fault recording and the line section where it occurs are recorded as fault recording information; the line information where the fault occurs and its corresponding basic information are stored in the analysis factor library as the line fault information of this time; S20: construct a fault waveform identification model based on the fault recording information and the fault traveling wave information, obtain the fault time, fault distance, fault insulation characteristics and waveform characteristic parameter information, and store them in the judgment factor library corresponding to the line fault information; S30: Calculate the line section where the fault point is located according to the fault recording and record it as the simulated line section, and determine whether the simulated line section is consistent with the line section where the fault occurs in the fault recording information; if they are consistent, proceed to the subsequent steps; S40: Acquire various tower monitoring data of the line section where the line fault point is located, update the information irregularly and input it into the analysis factor library, and correspond it with the line fault information; S50: setting the line fault information in the analysis factor library and the fault time, fault distance, fault insulation characteristics, waveform characteristic parameter information, various tower monitoring data, and irregular update information corresponding to the line fault information as factors; Each factor is given the same initial weight, and the weight values ​​of each factor in this line fault are multiplied by the weight values ​​of the factors stored in the judgment factor library to obtain each sub-factor, and a smart diagnosis model is constructed to diagnose the sub-factors and obtain the line fault diagnosis results; S60: The monitoring personnel of the monitoring line sends the line fault diagnosis result to the mobile app of the line operation and maintenance personnel; The operation and maintenance personnel will handle the faults and conduct inspections based on the line fault diagnosis results; S70: The fault handling situation and inspection results of the operation and maintenance personnel are transmitted back to the monitoring personnel in real time; based on the fault handling situation and inspection results, the monitoring personnel verify and confirm the accuracy of the line fault diagnosis results, and store the factors and actual situation of this line fault in the historical fault sample library.

2. The line fault intelligent diagnosis method according to claim 1, characterized in that: The method further includes step S80: according to the historical fault sample library, using a multimodal learning method, dynamically calculating the weights of the factors in the analysis and judgment factor library, and forming factor correlation rule criteria for various fault types.

3. The line fault intelligent diagnosis method according to claim 2, characterized in that: In step S80, the weights of the factors in the dynamic judgment factor library are specifically calculated as follows: After each new fault sample is stored in the historical fault sample library, the weights of the factors related to the new fault sample in the judgment factor library are dynamically updated; Set an initial weight W(1) for each factor; the initial weight of the i-th factor is Wi(1); Set Wi(t+1)=Wi(t)+ΔWi; Where Wi(t+1) is the new weight of factor i after the t+1th update, Wi(t) is the old weight of factor i after the tth update, and ΔWi is the weight increment of factor i.

4. The line fault intelligent diagnosis method according to claim 3, characterized in that: In step S80, the weight increment ΔWi is calculated as follows: When the diagnosis is correct, ΔWi = α × Si; When the diagnosis is wrong, ΔWi=(-β)×Si; Among them, α and β are update rates, which are artificially set values ​​and are constants; Si is the importance score of factor i.

5. The line fault intelligent diagnosis method according to claim 1, characterized in that: In step S20, the fault waveform identification model is constructed as follows: Firstly, pre-process the fault recording information and fault traveling wave information; Extract characteristic parameters of fault waveform from preprocessed data in three aspects: time domain, frequency domain and time-frequency domain; The extracted feature parameters are then input into the shared hidden layer to capture the features among the fault time, fault distance, fault insulation characteristics and waveform feature parameter information, and the features are fused to form a universal feature vector output; Finally, the universal feature vector is input into multiple output layers, and the fault time, fault distance, fault insulation characteristics and waveform characteristic parameter information are refined and outputted respectively according to the universal feature vector.

6. The line fault intelligent diagnosis method according to claim 5, characterized in that: In step S20, establishing a shared hidden layer is specifically as follows: Establish the first layer of feature fusion, from the time domain f t , frequency domain f f and the time-frequency domain f tf The extracted features are initially integrated to form a feature vector f1=Concat(f t , f f ); Establish the second layer of feature fusion and introduce a weighting mechanism. Different features have different contributions when fused; form a feature vector f2=Concat(W1·f1,W2·f tf ), where W1 and W2 are artificially set weight coefficients and are constants; The third layer of feature fusion is established, and adaptive feature fusion is introduced. The importance of features is dynamically adjusted according to the existing features to adapt to the identification of fault waveforms, forming a feature vector f3=Concat(f2,g); where g is an adaptive function, which is a linear function set artificially; The final feature vector f3 is the feature vector of the shared hidden layer.

7. The line fault intelligent diagnosis method according to claim 1, characterized in that: In step S50, the smart diagnosis model is specifically: Type identification level: Input the line fault information and the factors consisting of the fault time, fault distance, fault insulation characteristics, waveform characteristic parameter information, various tower monitoring data, and irregularly updated information corresponding to the line fault information, and extract multiple sub-sample sets from the factors by replacement sampling method; compare the information in each sub-sample set with the manually set fault type to obtain the most similar fault type for each sub-sample set; select the fault type with the largest proportion from all sub-sample sets in proportion, so as to determine the line fault type and cause; Position estimation level: input fault waveform information; Screen the current and voltage waveforms between the line sections where the fault point is located according to the passage of time, and record the time of sudden changes in the current and voltage waveforms; Calculate the mutation time from the time when the current and voltage waveforms suddenly change to the initial time when the line enters the line section where the fault point is located, and then calculate the total time for the line to enter and flow out of the line section where the fault point is located, and finally obtain the proportion of the mutation time to the total time, so as to obtain the proportion of the length of the line section where the fault point is located, and determine the specific location of the line fault; Comprehensive diagnosis level: receives the line fault type and cause determined by the type identification level, and the specific location of the line fault determined by the location estimation level, and comprehensively explains the final fault diagnosis result.

8. The line fault intelligent diagnosis method according to claim 7, characterized in that: In step S50, the intelligent diagnosis model periodically references new fault samples in the historical fault sample library to update the model.

9. The line fault intelligent diagnosis method according to claim 1, characterized in that: In step S30, the line section where the fault point is located is calculated as follows: Set the nth pair of towers T n With T n+1 The gear distance between them is dn; The fault distance between the fault point and the tower T1 in the fault recording wave is recorded as Df; Calculate the cumulative distance Dc=d1+d2+…+dn; When Dc accumulates to the i-th pair of towers T i With T i+1 , and Dc ≥ Df, stop calculating the cumulative distance Dc, and record the fault point at the i-th pair of towers T at this time i With T i+1 Thus, the line section T where the fault point is located can be obtained. i With T i+1 between.

10. A line fault intelligent diagnosis system, characterized in that: The line fault intelligent diagnosis method according to any one of claims 1 to 9 comprises: Data management module: including line library, which is used to store line information, current and voltage waveforms, various tower monitoring data, and irregularly updated information of each line section; including analysis factor library, which is used to store information related to faults; including historical fault sample library, which is used to store factors of historical line faults and their corresponding real situations; Fault collection module: record the time when the traveling wave arrives at different line sections to obtain fault traveling wave information; obtain fault recording information based on line information; and convert the line information corresponding to the line fault stored in the line library into line fault information and store it in the judgment factor library; obtain various tower monitoring data of the line section where the line fault point is located, update the information irregularly and input it into the judgment factor library, and correspond it with the line fault information; Fault detection module: obtain fault time, fault distance, fault insulation characteristics and waveform characteristic parameter information based on the fault waveform identification model; calculate the line section where the fault point is located based on the tower span information; Fault diagnosis module: builds an intelligent diagnosis model to diagnose and judge the factors in the factor library to obtain line fault diagnosis results; Communication feedback module: used for communication between monitoring personnel and operation and maintenance personnel.

Images