Cable fault detection method and device, electronic equipment and storage medium

Abstract

The invention discloses a cable fault detection method and device, electronic equipment and a storage medium, and relates to the technical field of intelligent diagnosis of a power system. The method comprises the following steps: acquiring cable operation data; inputting the cable operation data into a pre-trained cable fault detection model to obtain an alternative cable fault type predicted by each cable fault detection sub-model; wherein the cable fault detection model comprises at least one cable fault detection sub-model; and voting each alternative cable fault type, and determining a target cable fault type of the cable operation data. According to the technical scheme of the embodiment of the invention, the detection efficiency and the detection accuracy of cable fault detection are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent diagnosis of power systems, and in particular to a cable fault detection method, device, electronic equipment and storage medium. Background Art

[0002] The modern power system is the nerve center of modern society. It provides indispensable energy support for all walks of life through a complex and sophisticated network. The stable supply of electricity is a key factor in maintaining the rhythm of daily life, industrial production efficiency and economic development. In this system, cables, as the blood vessels of power transmission, play a vital role. The health of cables is directly related to the reliability and safety of power transmission. The occurrence of cable faults will not only cause local or even large-scale power supply interruptions, but also may cause damage to connected equipment, trigger a chain reaction, lead to production stagnation and economic losses. More seriously, cable faults may also bring safety hazards and threaten the safety of personnel and facilities. Therefore, timely detection, accurate positioning and rapid repair of cable faults are of great significance to maintaining the integrity and stability of the power system.

[0003] With the advancement of science and technology, intelligent management has become an important direction for improving the efficiency of cable operation and maintenance. The intelligent cable system uses advanced sensors, Internet of Things technology and big data analysis to monitor the operating status of the cable in real time, predict potential fault risks, and provide auxiliary decision support when a fault occurs. However, although intelligence has brought many conveniences to cable management, the existing intelligent cable system still has limitations in automated fault diagnosis. When a fault actually occurs, it is often difficult for the intelligent system to automatically identify the fault type independently. This means that once a fault occurs, it is still necessary to rely on professionals for on-site inspection and fault diagnosis, which not only increases the cost of operation and maintenance, but also slows down the speed of fault handling. Summary of the invention

[0004] The present invention provides a cable fault detection method, device, electronic equipment and storage medium, which improve the detection efficiency and detection accuracy of cable fault detection.

[0005] According to one aspect of the present invention, a cable fault detection method is provided, the method comprising:

[0006] Obtain cable operation data;

[0007] Inputting the cable operation data into a pre-trained cable fault detection model to obtain candidate cable fault types predicted by each cable fault detection sub-model; wherein the cable fault detection model includes at least one cable fault detection sub-model;

[0008] Voting is performed on each of the candidate cable fault types to determine a target cable fault type of the cable operation data.

[0009] According to another aspect of the present invention, there is provided a cable fault detection device, the device comprising:

[0010] A cable operation data acquisition module is used to acquire cable operation data;

[0011] A cable fault type prediction model, used for inputting the cable operation data into a pre-trained cable fault detection model to obtain candidate cable fault types predicted by each cable fault detection sub-model; wherein the cable fault detection model includes at least one cable fault detection sub-model;

[0012] The cable fault type voting module is used to vote on each of the candidate cable fault types to determine the target cable fault type of the cable operation data.

[0013] According to another aspect of the present invention, an electronic device is provided, the electronic device comprising:

[0014] at least one processor; and

[0015] a memory communicatively connected to the at least one processor; wherein,

[0016] The memory stores a computer program executable by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can execute the cable fault detection method described in any embodiment of the present invention.

[0017] According to another aspect of the present invention, a computer-readable storage medium is provided, wherein the computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a processor to implement the cable fault detection method described in any embodiment of the present invention when executed.

[0018] According to another aspect of the present invention, a computer program product is provided. The computer program product comprises a computer program. When the computer program is executed by a processor, the cable fault detection method according to any embodiment of the present invention is implemented.

[0019] The technical solution of the embodiment of the present invention performs real-time detection of cable operation data by pre-training each cable fault detection sub-model in the cable fault detection model, thereby realizing automatic detection of cable faults, reducing the operation and maintenance cost of cable fault detection, and improving the detection efficiency of cable fault detection. By voting on each alternative cable fault type and determining the target cable fault type, the detection accuracy of the cable fault can be further improved.

[0020] It should be understood that the contents described in this section are not intended to identify the key or important features of the embodiments of the present invention, nor are they intended to limit the scope of the present invention. Other features of the present invention will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0022] Figure 1 is a flow chart of a cable fault detection method provided according to Embodiment 1 of the present invention;

[0023] Figure 2 is a flow chart of a cable fault detection method provided according to Embodiment 2 of the present invention;

[0024] Figure 3 is a flow chart of a cable fault detection method provided according to Embodiment 2 of the present invention;

[0025] Figure 4 is a structural schematic diagram of a cable fault detection model provided according to Embodiment 2 of the present invention;

[0026] Figure 5 is a structural schematic diagram of a cable fault detection device provided according to Embodiment 3 of the present invention;

[0027] Figure 6 It is a structural schematic diagram of an electronic device for implementing the cable fault detection method according to an embodiment of the present invention. DETAILED DESCRIPTION

[0028] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described 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, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0029] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0030] Embodiment 1

[0031] Figure 1 A flowchart of a cable fault detection method provided in Embodiment 1 of the present invention. The embodiment of the present invention is applicable to the case of detecting cable faults in a distribution network. The method can be performed by a cable fault detection device, which can be implemented in the form of hardware and / or software. The cable fault detection device can be configured in an electronic device that carries a cable fault detection function, such as a fault analysis server.

[0032] See also Figure 1 The cable fault detection method shown comprises:

[0033] S110. Obtain cable operation data.

[0034] The cable operation data can be used to characterize the operation status of the cable in the distribution network. For example, the cable operation data can include information such as cable voltage, cable current, cable temperature, humidity of the environment where the cable is located, and / or cable load.

[0035] In an optional embodiment of the present invention, the cable operation data includes cable current, cable temperature and cable load. The cable current may be the carrying current of the cable. The cable temperature may be the surface temperature of the cable. The cable load may be the load connected to the cable. This solution further improves the detection efficiency of cable fault detection by concretizing the cable operation data into cable current, cable temperature and cable load, and by selecting typical data in the cable operation data to detect the cable fault type.

[0036] Specifically, the cable operation data may be detected using sensors provided in the distribution network.

[0037] S120: Input the cable operation data into a pre-trained cable fault detection model to obtain candidate cable fault types predicted by each cable fault detection sub-model.

[0038] The cable fault detection model can be used to detect the cable fault type in the distribution network. The cable fault detection model includes at least one cable fault detection submodel. A single cable fault detection submodel has a high detection accuracy for a corresponding cable fault type. Therefore, each cable fault detection submodel can detect the cable fault type, and the detection accuracy for different cable fault types is different. Exemplarily, the cable fault detection submodel can be a width learning model. The input data of the cable fault detection model can be cable operation data, and the output result can be each candidate cable fault type predicted by each cable fault detection submodel. The candidate cable fault type can be a cable fault type predicted by the cable fault detection submodel. Optionally, the number of candidate cable fault types can be single or multiple. The number of candidate cable fault types is single, which can be understood as the candidate cable fault type is the final prediction result of the cable fault detection submodel, for example, the candidate cable fault type is the reference cable fault type with the highest confidence among the reference cable fault types. The reference cable fault type can be all cable fault types in the distribution network or part of the cable fault types with a preset output number with a high confidence. The preset output quantity may be the number of output results of the pre-set cable fault detection sub-model. The number of candidate cable fault types is multiple, which can be understood as the candidate cable fault types are the reference cable fault types or further screening results of the reference cable fault types.

[0039] Specifically, the cable operation data may be input into a pre-trained cable fault detection model, and the cable operation data may be detected by each cable fault detection sub-model to obtain each candidate cable fault type output by each cable fault detection sub-model.

[0040] S130: Vote for each candidate cable fault type to determine a target cable fault type for the cable operation data.

[0041] The target cable fault type may be a cable fault type of the finally determined cable operation data.

[0042] Specifically, voting may be performed on each candidate cable fault type, and the candidate cable fault type that appears the most times may be determined as the target cable fault type of the cable operation data.

[0043] For example, it is assumed that the number of candidate cable fault types is M; the candidate cable fault types are [y1, y2, y3, ..., y M ]; where y a It is the prediction result of the a-th cable fault detection sub-model on the cable operation data (i.e., the alternative cable fault type of the a-th cable fault detection sub-model).

[0044] The following formula can be used to vote for each alternative cable fault type:

[0045]

[0046] In the formula, Count(Y l ) is the cable fault type Y l The number of votes indicates cable fault type Y l The number of times predicted by all cable fault detection sub-models; Y l is the lth cable fault type; l is the number of the cable fault type; y a is the candidate cable fault type of the a-th cable fault detection sub-model; a is the number of the cable fault detection sub-model or the candidate cable fault type; M is the number of the cable fault detection sub-models; I(y a =Y l ) is the indicator function, if (y a =Y l ) is equal to 1, otherwise it is 0.

[0047] The target cable fault type for cable operation data can be calculated using the following formula:

[0048]

[0049] In the formula, argmax is the target cable fault type for cable operation data; M To select the maximum value function; Count(Y l ) is the cable fault type Y l The number of votes indicates cable fault type Y l The number of times predicted by all cable fault detection sub-models. It can be understood that the cable fault type with the most votes is selected as the final prediction result, so the target cable fault type of the cable operation data can be obtained.

[0050] In an optional example, the alternative cable fault types and confidences of the alternative cable fault types of each cable fault detection sub-model may be statistically analyzed to determine the target cable fault type of the cable operation data. For example, the alternative cable fault type with the highest confidence may be determined as the target cable fault type. For another example, the confidences of the same alternative cable fault type output by different cable fault detection sub-models may be averaged to obtain the average confidences of the alternative cable fault types, and the alternative cable fault type with the highest average confidence may be determined as the target cable fault type of the cable operation data.

[0051] The technical solution of the embodiment of the present invention performs real-time detection of cable operation data by pre-training each cable fault detection sub-model in the cable fault detection model, thereby realizing automatic detection of cable faults, reducing the operation and maintenance cost of cable fault detection, and improving the detection efficiency of cable fault detection. By voting on each alternative cable fault type and determining the target cable fault type, the detection accuracy of the cable fault can be further improved.

[0052] Embodiment 2

[0053] Figure 2 A flow chart of a cable fault detection method provided for the second embodiment of the present invention. Based on the above embodiments, the embodiment of the present invention further adds "obtaining cable operation samples and actual cable fault types; determining the cable fault detection sub-models corresponding to each actual cable fault type and the classification loss weights of each cable fault detection sub-model; for a single cable fault detection sub-model, based on the cable operation samples, the classification loss weights of the cable fault detection sub-models and the corresponding actual cable fault types, calculating the target connection weight matrix of the cable fault detection sub-model; adjusting the parameters of each cable fault detection sub-model according to the target connection weight matrix of each cable fault detection sub-model", and pre-training cable fault detection sub-models with different classification loss weights to obtain cable fault detection sub-models with different detection sensitivities to different actual cable fault types, thereby improving the detection accuracy and efficiency of cable fault detection using cable fault detection models. It should be noted that for the parts not described in detail in the embodiment of the present invention, please refer to the description of other embodiments.

[0054] See also Figure 2 The cable fault detection method shown comprises:

[0055] S210. Obtain cable operation samples and actual cable fault types.

[0056] The cable operation sample may be the cable operation data collected before the current moment in the distribution network. The cable operation sample may be used to train the cable fault detection model. The actual cable fault type may be the actual cable fault type of a certain group of cable operation samples. Optionally, the actual cable fault type may be pre-calibrated by a technician. The number of cable operation samples may be multiple groups. A single group of cable operation samples may correspond to a certain actual cable fault type.

[0057] Specifically, the cable operation data collected by the sensors in the distribution network before the current moment can be obtained as the cable operation sample, and the actual cable fault type corresponding to the pre-calibrated cable operation sample can be obtained.

[0058] S220: Determine the cable fault detection sub-model corresponding to each actual cable fault type and the classification loss weight of each cable fault detection sub-model.

[0059] The classification loss weight can be used to adjust the detection accuracy of the cable fault detection sub-model for different cable fault types. It can be understood that a cable fault detection sub-model can correspond to an actual cable fault type, and the weight value of the actual cable fault type corresponding to the cable fault detection sub-model in the classification loss weight of the cable fault detection sub-model is higher. That is, by differentially improving the weight value of the actual cable fault type corresponding to the cable fault detection sub-model in the classification loss weight of the cable fault detection sub-model, the cable fault detection sub-model can have a higher detection accuracy for the actual cable fault type. Exemplarily, the weight value of the actual cable fault type corresponding to the cable fault detection sub-model in the classification loss weight of the cable fault detection sub-model can be greater than 1; the weight value of other actual cable fault types can be 1. Optionally, in the classification loss weights of different cable fault detection sub-models, the higher weight values ​​can be the same or different.

[0060] Specifically, the cable fault detection sub-model corresponding to each actual cable fault type can be determined according to a one-to-one correspondence. For a single cable fault detection sub-model, the weight value of the corresponding actual cable fault type can be determined to be higher than the weight values ​​of other actual cable fault types in the classification loss weight of the cable fault detection sub-model, thereby obtaining the classification loss weight of each cable fault detection sub-model.

[0061] S230. For a single cable fault detection sub-model, based on the cable operation samples, the classification loss weights of the cable fault detection sub-model and the corresponding actual cable fault types, calculate the target connection weight matrix of the cable fault detection sub-model.

[0062] The target connection weight matrix can be used to characterize the model parameters of the trained cable fault detection sub-model. The target connection weight matrix can be used to adjust the parameters of the cable fault detection sub-model so that the cable fault detection sub-model training is completed.

[0063] Specifically, for a single cable fault detection sub-model, the cable operation samples are obtained through the input layer of the cable fault detection sub-model; the cable operation samples are feature mapped through the feature layer of the cable fault detection sub-model to obtain the cable operation feature matrix; the cable operation feature matrix is ​​nonlinearly activated through the enhancement layer of the cable fault detection sub-model to obtain the cable operation enhancement matrix, and the cable operation feature matrix and the cable operation enhancement matrix are connected to obtain the cable operation matrix; through the output layer of the cable fault detection sub-model, a loss function is used to calculate the target connection weight matrix of the cable fault detection sub-model based on the cable operation matrix, the actual cable fault type and the classification loss weight.

[0064] For example, it is assumed that the cable fault detection submodel is a width learning model. The cable fault detection submodel has n feature windows, each feature window is set with C feature nodes, and the cable operation data can be mapped to n groups of random feature spaces, each group of feature spaces contains C features. Among them, the feature window can be used to extract features from the cable operation data.

[0065] The cable operation characteristic matrix can be calculated using the following formula:

[0066]

[0067] In the formula, Z i is the feature matrix in the i-th group of random feature space; i is the number of the feature window; is the characteristic mapping function; X is the cable running sample; and are the weights and biases of the randomly generated feature layers; after obtaining all Z i ∈R N×M Then, all Z i Connect them to generate the cable run characteristic matrix Z n , the feature space of cable operation data can be constructed.

[0068] In order to further enhance the mapping capability of the cable fault detection sub-model and further mine the deep feature information of the cable operation data, the cable operation feature matrix Z n Perform another nonlinear mapping to obtain the cable operation enhancement matrix. The cable operation enhancement matrix can be calculated using the following formula:

[0069]

[0070] In the formula, H g is the g-th group enhancement matrix; g is the group number of the enhancement layer feature node; α represents the nonlinear activation function; Z n Run the characteristic matrix for the cable; and represents the weight and bias of the randomly generated enhancement layer; m represents the total number of feature nodes in the enhancement layer; g Connect to generate cable run enhancement matrix H m .

[0071] The cable operation matrix can be calculated by connecting the cable operation characteristic matrix and the cable operation enhancement matrix using the following formula:

[0072]

[0073] In the formula, is the cable operation matrix; Z n is the cable operation characteristic matrix; H m Run an enhanced matrix for the cables.

[0074] Assuming that the actual cable fault type corresponding to the cable fault detection sub-model D is the first type, the classification loss weight of the cable fault detection sub-model is a D＝ [f(β1X1:X2:X3:..:X M )]; where β is an enhancement weight with a value greater than 1; and f is a reduction function set according to the greatest common divisor of different sample numbers.

[0075] The loss function of the cable fault detection sub-model D can be expressed by the following formula:

[0076]

[0077] In the formula, is the target connection weight matrix; λ is the penalty coefficient; f is the reduction function set according to the greatest common divisor of different sample numbers; β is the enhancement weight with a value greater than 1; is the cable operation matrix; (X1, X2, X3, .., X M ) is a cable running sample; (Y1, Y2, Y3, .., Y M ) is the actual cable fault type corresponding to the cable operation sample.

[0078] The loss function of the cable fault detection sub-model D can be derivated and the derivative can be set to 0 to obtain the target connection weight matrix. Specifically, the target connection weight matrix can be calculated using the following formula:

[0079]

[0080] In the formula, is the target connection weight matrix; λ is the penalty coefficient; f is the reduction function set according to the greatest common divisor of different sample numbers; β is the enhancement weight with a value greater than 1; is the cable operation matrix; (X1, X2, X3, .., X M ) is a cable running sample; (Y1, Y2, Y3, .., Y M ) is the actual cable fault type corresponding to the cable operation sample; l is the number of the actual cable fault type.

[0081] In an optional embodiment of the present invention, for a single cable fault detection sub-model, based on the cable operation samples, the classification loss weights of the cable fault detection sub-model and the corresponding actual cable fault types, a target connection weight matrix of the cable fault detection sub-model is calculated, including: dividing the cable operation samples according to the sample proportions corresponding to each actual cable fault type in the cable operation samples and the number of cable fault detection sub-models to obtain unit cable operation samples of each cable fault detection sub-model; for a single cable fault detection sub-model, based on the unit cable operation samples of the cable fault detection sub-model, the classification loss weights of the cable fault detection sub-model and the corresponding actual cable fault types, the target connection weight matrix of the cable fault detection sub-model is calculated.

[0082] The sample ratio corresponding to each actual cable fault type can be used to characterize the sample ratio of the cable operation samples corresponding to different actual cable fault types. The cable operation samples are divided using the sample ratio corresponding to each actual cable fault type to obtain the unit cable operation samples of each cable fault detection sub-model. It can be understood that the cable operation samples are divided in an equal proportion, which can avoid the problem of too many samples of a certain type of actual cable fault type in different cable fault detection sub-models, thereby overcoming the imbalance of samples corresponding to different actual cable fault types in the cable operation samples.

[0083] Specifically, the cable operation samples can be divided into the number of cable fault detection sub-models in equal proportion according to the sample ratio corresponding to each actual cable fault type in the cable operation samples, and each cable operation sample can be determined as a unit cable operation sample of each cable fault detection sub-model. For a single cable fault detection sub-model, the unit cable operation sample can be obtained through the input layer of the cable fault detection sub-model; the cable operation sample is feature mapped through the feature layer of the cable fault detection sub-model to obtain a cable operation feature matrix; the cable operation feature matrix is ​​nonlinearly activated through the enhancement layer of the cable fault detection sub-model to obtain a cable operation enhancement matrix, and the cable operation feature matrix and the cable operation enhancement matrix are connected to obtain a cable operation matrix; through the output layer of the cable fault detection sub-model, a loss function is used to calculate the target connection weight matrix of the cable fault detection sub-model based on the cable operation matrix, the actual cable fault type and the classification loss weight.

[0084] For example, assuming that the cable data sample is (X, Y), {(X, Y)|X∈R N×A , Y∈R N×M}. Where N represents the total number of cable operation samples; A represents the characteristic dimension of the cable operation samples; M represents the number of categories of actual cable fault types; and R represents a real number. Wherein, the number of M is determined by the number of categories of actual cable fault types. Accordingly, the cable fault detection model can be composed of M cable fault detection sub-models. First, the cable operation samples are divided into M parts in equal proportion according to the inter-class ratio (i.e., sample ratio) of each actual cable fault type. The M unit cable operation samples after division are (X1, Y1), (X2, Y2), ..., (X M , Y M ). Each unit cable operation sample can train a cable fault detection sub-model separately. The equal-proportional division method can avoid the problem of too many samples in a certain category. Since the number of cable fault detection sub-models is equal to the number of categories of actual cable fault types, each cable fault detection sub-model will perform enhanced training on the unit cable operation samples of the corresponding actual cable fault type. This enhanced training method can make the cable fault detection sub-model more sensitive to specific actual cable fault types, thereby overcoming the problem of category imbalance.

[0085] This scheme divides the cable operation samples in equal proportion according to the sample proportions corresponding to each actual cable fault type in the cable operation samples and the number of cable fault detection sub-models, and obtains unit cable operation samples of each cable fault detection sub-model. The cable fault detection sub-model is trained based on the unit cable operation samples, which can solve the problems of insufficient samples and imbalance of samples of different categories in the training process of the cable fault detection model, and can improve the sample balance, thereby improving the detection accuracy of the cable fault detection model.

[0086] S240. Adjust parameters of each cable fault detection sub-model according to the target connection weight matrix of each cable fault detection sub-model.

[0087] Specifically, the connection weight matrix of each cable fault detection sub-model can be adjusted according to the target connection weight matrix of each cable fault detection sub-model, thereby completing the training of the cable fault detection model.

[0088] S250, obtaining cable operation data.

[0089] S260: Input the cable operation data into a pre-trained cable fault detection model to obtain candidate cable fault types predicted by each cable fault detection sub-model.

[0090] The cable fault detection model includes at least one cable fault detection sub-model.

[0091] S270: Vote for each candidate cable fault type to determine a target cable fault type for the cable operation data.

[0092] The technical solution of the embodiment of the present invention determines the cable fault detection sub-model corresponding to each actual cable fault type and the classification loss weight of each cable fault detection sub-model by obtaining cable operation samples and actual cable fault types, wherein the weight value of the corresponding actual cable fault type in the classification loss weight of the cable fault detection sub-model is higher, and for a single cable fault detection sub-model, the target connection weight matrix of the cable fault detection sub-model is calculated based on the cable operation samples, the classification loss weight of the cable fault detection sub-model and the corresponding actual cable fault type, and the parameters of each cable fault detection sub-model are adjusted according to the target connection weight matrix of each cable fault detection sub-model, and cable fault detection sub-models with different detection sensitivities to different actual cable fault types are obtained by pre-training cable fault detection sub-models with different classification loss weights, thereby improving the detection accuracy and detection efficiency of cable fault detection using cable fault detection models.

[0093] In an optional embodiment of the present invention, after voting on each candidate cable fault type to determine the target cable fault type of the cable operation data, it also includes: periodically acquiring cable operation data and the corresponding actual cable fault type; for a single cable fault detection sub-model, feature mapping the cable operation data through the newly added feature node of the feature layer of the cable fault detection sub-model to obtain a newly added cable operation feature matrix; for a single cable fault detection sub-model, nonlinearly activate the newly added cable operation feature matrix through the first newly added enhancement node of the enhancement layer of the cable fault detection sub-model to obtain a first newly added cable operation enhancement matrix; for a single cable fault detection sub-model, obtain the original cable operation matrix through the first newly added enhancement node of the enhancement layer of the cable fault detection sub-model, and connect the original cable operation matrix, the newly added cable operation feature matrix and the first newly added cable operation enhancement matrix to obtain a first newly added cable operation matrix; for a single cable fault detection sub-model, update the target connection weight matrix through the output layer of the cable fault detection sub-model based on the first newly added cable operation matrix and the corresponding actual cable fault type to obtain a first target connection weight matrix; and adjust the parameters of each cable fault detection sub-model according to each first target connection weight matrix.

[0094] Real-time monitoring of cables will generate a large amount of cable operation data. Regularly using the newly added cable operation data to update the cable fault detection model can continuously improve the accuracy of cable fault monitoring. The update cycle of the cable fault detection model can be preset and adjusted. Exemplarily, the update cycle of the cable fault detection submodel can be daily, weekly, monthly or quarterly. The actual cable fault type corresponding to the cable operation data may be consistent with or inconsistent with the target cable fault type corresponding to the cable operation data. It can be understood that the actual cable fault type corresponding to the cable operation data can be calibrated and corrected by a technician. The newly added feature node can be used to perform feature mapping on the newly added cable operation data of the cable fault detection submodel. The newly added cable operation feature matrix can be used to characterize the feature extraction results of the newly added cable operation data. The first newly added enhancement node can be used to nonlinearly activate the newly added cable operation feature matrix of the cable fault detection submodel. The first newly added cable operation enhancement matrix can be used to characterize the feature enhancement results of the newly added cable operation data based on the first newly added enhancement node. The first newly added cable operation matrix can be used to characterize the fusion results of the feature extraction results and feature enhancement results of the newly added cable operation data using the incremental method of the feature node. The first target connection weight matrix may be a target connection weight matrix updated in an incremental manner using feature nodes.

[0095] Specifically, after voting on each candidate cable fault type and determining the target cable fault type of the cable operation data, the cable operation data and the actual cable fault type corresponding to the pre-calibrated cable operation data can be periodically obtained according to the pre-set update cycle of the cable fault detection sub-model. For a single cable fault detection sub-model, the cable operation data can be feature mapped through the newly added feature nodes of the feature layer of the cable fault detection sub-model to obtain a newly added cable operation feature matrix. For a single cable fault detection sub-model, the newly added cable operation feature matrix is ​​nonlinearly activated through the first newly added enhancement node of the enhancement layer of the cable fault detection sub-model to obtain the first newly added cable operation enhancement matrix. For a single cable fault detection sub-model, the original cable operation matrix is ​​obtained through the first newly added enhancement node of the enhancement layer of the cable fault detection sub-model, and the original cable operation matrix, the newly added cable operation feature matrix and the first newly added cable operation enhancement matrix are connected to obtain the first newly added cable operation matrix. For a single cable fault detection sub-model, the first target connection weight matrix is ​​calculated through the output layer of the cable fault detection sub-model and the loss function of the cable fault detection sub-model is used based on the first newly added cable operation matrix and the corresponding actual cable fault type; according to the updated first target connection weight matrices of each cable fault detection sub-model, the parameters of each cable fault detection sub-model are adjusted, thereby completing the update of the cable fault detection model.

[0096] Referring to the above example, the following formula can be used to calculate the new cable operation characteristic matrix:

[0097]

[0098] In the formula, Z n+1 is the feature matrix in the n+1th group of random feature space, that is, the feature matrix in the random feature space corresponding to the newly added feature window, that is, the newly added cable operation feature matrix; n+1 is the number of the newly added feature window; X is the cable operation data obtained periodically; and They are the weights and biases generated for the newly added feature nodes respectively.

[0099] The first additional cable operation enhancement matrix can be calculated using the following formula:

[0100]

[0101] In the formula, H j is the jth group enhancement matrix; j is the group number of the enhancement node; α represents the nonlinear activation function; Z n+1 is the characteristic matrix in the n+1th group of random characteristic space, that is, the newly added cable operation characteristic matrix; represents the weight and bias of the randomly generated enhancement layer; m represents the total number of groups of feature nodes of the enhancement layer; wherein j can be multiple values.

[0102] The following formula can be used to calculate the first newly added cable operation matrix:

[0103]

[0104] In the formula, Added a cable run matrix for the first; is the original cable operation matrix; Z n+1 is the characteristic matrix in the n+1th group of random characteristic space, that is, the newly added cable operation characteristic matrix; H j is the jth group of enhancement matrix; j is the group number of the enhancement layer feature node.

[0105] The updated first target connection weight matrix can be calculated using the following formula:

[0106]

[0107] In the formula, is the updated first target connection weight matrix; is the first newly added cable operation matrix; Y is the actual cable fault type corresponding to the first newly added cable operation matrix; is the target connection weight matrix; E1 is the first intermediate variable; K1 Tis the second intermediate variable.

[0108] in,

[0109]

[0110] A n+1 =[|Z n+1 |H j ];

[0111]

[0112] In the formula, E1 is the first intermediate variable; A is the first newly added cable operation matrix; n+1 It is an independent new cable operation matrix generated for the new cable operation data; Z n+1 is the feature matrix in the n+1th group of random feature space, that is, the feature matrix in the random feature space corresponding to the newly added feature window; H j is the jth group enhancement matrix; j is the group number of the enhancement layer feature node; F1 is the third intermediate variable; Matrix for original cable operation; K1 T is the second intermediate variable.

[0113] This solution adopts an incremental approach of feature nodes, which can perform incremental learning on each individual cable fault detection sub-model without training the model from scratch. Compared with traditional machine learning models and deep learning models, which need to be trained from scratch when remodeling, it will undoubtedly occupy huge hardware resources to a great extent and will also impose a heavy burden on edge devices. This solution can significantly reduce the resource consumption of updating the cable fault detection model.

[0114] In an optional embodiment of the present invention, after voting on the cable fault types predicted by each cable fault detection sub-model to determine the target cable fault type of the cable operation data, it also includes: periodically acquiring cable operation data and the corresponding actual cable fault type; for a single cable fault detection sub-model, feature mapping the cable operation data through the feature nodes of the feature layer of the cable fault detection sub-model to obtain a cable operation feature matrix; for a single cable fault detection sub-model, nonlinearly activating the cable operation feature matrix through the second newly added enhancement node of the enhancement layer of the cable fault detection sub-model to obtain a second newly added cable operation enhancement matrix; for a single cable fault detection sub-model, obtaining the original cable operation matrix through the second newly added enhancement node of the enhancement layer of the cable fault detection sub-model, and connecting the original cable operation matrix and the second newly added cable operation enhancement matrix to obtain a second newly added cable operation matrix; for a single cable fault detection sub-model, updating the target connection weight matrix through the output layer of the cable fault detection sub-model based on the second newly added cable operation matrix and the corresponding actual cable fault type to obtain a second target connection weight matrix; and adjusting the parameters of each cable fault detection sub-model according to each second target connection weight matrix.

[0115] The second newly added enhancement node can also be used to perform nonlinear activation on the newly added cable operation feature matrix of the cable fault detection sub-model. In comparison, the first newly added enhancement node and the second newly added enhancement node are both newly added enhancement nodes of the cable fault detection sub-model. The difference is that the first newly added enhancement node corresponds to the incremental mode of the feature node and is used in combination with the newly added feature node; the second newly added enhancement node corresponds to the enhanced mode of the enhancement node, and no additional feature node is required, and it can be used alone. The second newly added cable operation enhancement matrix can be used to characterize the feature enhancement results of the newly added cable operation data based on the second newly added enhancement node. The second newly added cable operation matrix can be used to characterize the fusion results of the feature extraction results and feature enhancement results of the newly added cable operation data using the incremental mode of the enhancement node. The second target connection weight matrix can be the target connection weight matrix updated using the incremental mode of the feature node.

[0116] Specifically, after voting on the cable fault types predicted by each cable fault detection sub-model and determining the target cable fault type of the cable operation data, the cable operation data and the actual cable fault type corresponding to the pre-calibrated cable operation data can be periodically obtained according to the pre-set update cycle of the cable fault detection sub-model. For a single cable fault detection sub-model, the cable operation data is feature mapped through the feature nodes of the feature layer of the cable fault detection sub-model to obtain a cable operation feature matrix. For a single cable fault detection sub-model, the cable operation feature matrix is ​​nonlinearly activated through the second newly added enhancement node of the enhancement layer of the cable fault detection sub-model to obtain a second newly added cable operation enhancement matrix. For a single cable fault detection sub-model, the original cable operation matrix is ​​obtained through the second newly added enhancement node of the enhancement layer of the cable fault detection sub-model, and the original cable operation matrix and the second newly added cable operation enhancement matrix are connected to obtain a second newly added cable operation matrix. For a single cable fault detection sub-model, the output layer of the cable fault detection sub-model uses the loss function of the cable fault detection sub-model to calculate the updated second target connection weight matrix based on the second newly added cable operation matrix and the corresponding actual cable fault type. According to the updated second target connection weight matrices, the parameters of each cable fault detection sub-model are adjusted, thereby completing the update of the cable fault detection model.

[0117] Referring to the above example, the second additional cable run enhancement matrix can be calculated using the following formula:

[0118]

[0119] In the formula, H m+1 is the m+1th group enhancement matrix, i.e., the second newly added cable operation enhancement matrix; m+1 is the group number of the newly added enhancement node; α represents the nonlinear activation function; Z n Run the characteristic matrix for the cable; and Represents the weights and biases of the randomly generated enhancement layers for the newly added enhancement nodes.

[0120] The second additional cable operation matrix can be calculated using the following formula:

[0121]

[0122] In the formula, Added cable runs matrix for the second addition; is the cable operation matrix; H n+1 It is the m+1th group of enhanced matrix, that is, the second newly added cable operation enhanced matrix; m+1 is the group number of the newly added enhanced node.

[0123] The updated second target connection weight matrix can be calculated using the following formula:

[0124]

[0125] In the formula, The updated second target connection weight matrix; is the second newly added cable operation matrix; Y is the actual cable fault type corresponding to the first newly added cable operation matrix; is the target connection weight matrix; E2 is the fourth intermediate variable; K2 T is the fifth intermediate variable.

[0126] in,

[0127]

[0128]

[0129] Wherein, E2 is the fourth intermediate variable; is the original cable operation matrix; H m+1 is the m+1th group of enhanced matrix, i.e., the second newly added cable operation enhanced matrix; m+1 is the group number of the newly added enhanced node; F2 is the sixth intermediate variable; K2 T is the fifth intermediate variable.

[0130] This solution adopts an incremental method of enhancing nodes, which can perform incremental learning on each individual cable fault detection sub-model without training the model from scratch. Compared with traditional machine learning models and deep learning models, which need to be trained from scratch when remodeling, it will undoubtedly occupy huge hardware resources to a great extent and impose a heavy burden on edge devices. This solution can significantly reduce the resource consumption of updating the cable fault detection model.

[0131] With the rapid development of artificial intelligence technology, cable fault identification has gradually developed towards automatic diagnosis. By constructing a cable fault detection model, the automatic identification of cable faults can be achieved, the cable faults can be effectively detected, and the labor cost can be reduced. The intelligent cable operation monitoring system can include a data acquisition cluster, a fault analysis server, and a data storage device, and the data acquisition cluster and the data storage device can both be connected to the fault analysis server. By installing the data acquisition cluster at different locations of the intelligent cable, various normal and fault data of the cable and the cable fault type are collected and stored as cable operation samples, and the faults are monitored and identified through the results stored in the fault analysis server.

[0132] Figure 3 FIG. 1 is a flow chart of a cable fault detection method. Based on the above embodiment, Figure 3This is a preferred embodiment of the present invention. Figure 3 The cable fault detection method shown comprises:

[0133] Assume that the cable data sample is (X, Y), {(X, Y)|X∈R N×A , Y∈R N×M}. Where N represents the total number of cable operation samples; A represents the characteristic dimension of the cable operation samples; M represents the number of categories of actual cable fault types; and R represents a real number. Wherein, the number of M is determined by the number of categories of actual cable fault types. Accordingly, the cable fault detection model can be composed of M cable fault detection sub-models. First, the cable operation samples are divided into M parts in equal proportion according to the inter-class ratio (i.e., sample ratio) of each actual cable fault type. The M unit cable operation samples after division are (X1, Y1), (X2, Y2), ..., (X M , Y M ). Each unit cable operation sample can train a cable fault detection sub-model separately. The equal-proportional division method can avoid the problem of too many samples in a certain category. Since the number of cable fault detection sub-models is equal to the number of categories of actual cable fault types, each cable fault detection sub-model will perform enhanced training on the unit cable operation samples of the corresponding actual cable fault type. This enhanced training method can make the cable fault detection sub-model more sensitive to specific actual cable fault types, thereby overcoming the problem of category imbalance. Figure 4 Figure 1 is a schematic diagram of the structure of a cable fault detection model. Figure 4 As shown, each cable fault detection submodel can include an input layer, a feature layer, an enhancement layer, and an output layer. The cable fault detection submodel has n feature windows, each of which has C feature nodes, and the cable operation data can be mapped to n groups of random feature spaces, each of which contains C features. Among them, the feature window can be used to extract features from the cable operation data.

[0134] For a single cable fault detection sub-model, the feature layer of the cable fault detection sub-model can be used to extract features from the cable operation samples to obtain the cable operation feature matrix:

[0135]

[0136] In the formula, Z i is the feature matrix in the i-th group of random feature space; i is the number of the feature window; is the characteristic mapping function; X is the cable running sample; and are the weights and biases of the randomly generated feature layers; after obtaining Z n ∈R N×M Then, all Zi Connect them to generate the cable run characteristic matrix Z n , the feature space of cable operation data can be constructed.

[0137] In order to further enhance the mapping capability of the cable fault detection sub-model and further mine the deep feature information of the cable operation data, the cable operation feature matrix Z n Perform another nonlinear mapping to obtain the cable operation enhancement matrix. The cable operation enhancement matrix can be calculated using the following formula:

[0138]

[0139] In the formula, H g is the g-th group enhancement matrix; g is the group number of the enhancement layer feature node; α represents the nonlinear activation function; Z n Run the characteristic matrix for the cable; and represents the weight and bias of the randomly generated enhancement layer; m represents the total number of feature nodes in the enhancement layer; g Connect to generate cable run enhancement matrix H m .

[0140] Finally, the cable operation characteristic matrix Z of the characteristic layer can be n and the cable running reinforcement matrix H of the reinforcement layer m Connect in sequence to get the cable run matrix

[0141]

[0142] In the formula, is the cable operation matrix; Z n is the cable operation characteristic matrix; H m Run an enhanced matrix for the cables.

[0143] For the single cable fault detection sub-model D, the cable running sample (X D , Y D ) is classified according to the actual cable fault type, and the classified data matrix is ​​X D =[X1, X2, ..., X M ] and Y D =[Y1, Y2, ..., Y M ], the sample ratio corresponding to different actual cable fault types is a D =[X1:X2:X3:..:X M], the weight ratio corresponding to the cable fault detection sub-model D can be enhanced so that the cable fault detection sub-model D can enhance the training ability of the samples of the actual cable fault type. Assuming that the actual cable fault type corresponding to the cable fault detection sub-model D is the first type, the classification loss weight of the cable fault detection sub-model is a D =[f(β1X1:X2:X3:..:X M )]; where β1 is an enhancement weight with a value greater than 1; f is a reduction function set according to the greatest common divisor of different sample numbers. Figure 3 As shown in the figure, the enhanced weights of the cable fault detection sub-models corresponding to different actual cable fault types can be [β1, β2, ..., β M ].

[0144] The loss function of the cable fault detection sub-model D can be expressed by the following formula:

[0145]

[0146] In the formula, is the target connection weight matrix; λ is the penalty coefficient; f is the reduction function set according to the greatest common divisor of different sample numbers; β1 is the enhancement weight with a value greater than 1; is the cable operation matrix; (X1, X2, X3, ..., X M ) is a cable running sample; (Y1, Y2, Y3, ..., Y M ) is the actual cable fault type corresponding to the cable operation sample.

[0147] The loss function of the cable fault detection sub-model D can be derivated and the derivative can be set to 0 to obtain the target connection weight matrix. The target connection weight matrix can be calculated using the following formula:

[0148]

[0149] In the formula, is the target connection weight matrix; λ is the penalty coefficient; f is the reduction function set according to the greatest common divisor of different sample numbers; β is the enhancement weight with a value greater than 1; is the cable operation matrix; (X1, X2, X3, ..., X M ) is a cable running sample; (Y1, Y2, Y3, ..., Y M ) is the actual cable fault type corresponding to the cable operation sample; l is the number of the actual cable fault type.

[0150] Real-time monitoring of cables will generate a large amount of cable operation data. Regularly updating the cable fault detection model with the newly added cable operation data can continuously improve the accuracy of cable fault monitoring. Traditional machine learning models and deep learning models need to be trained from scratch when remodeling, which will undoubtedly occupy huge hardware resources to a great extent and will also be a heavy burden on edge devices. This solution can perform incremental learning on each individual subsystem without training the model from scratch, greatly reducing the consumption of training resources. Optionally, two methods can be used to perform incremental training of the cable fault detection submodel, namely the incremental method of feature nodes and the incremental method of enhanced nodes.

[0151] First, for the incremental method of feature nodes, assume that a new n+1th group of feature nodes is added to the cable fault detection sub-model, and the additional feature node Z n+1 An additional newly added enhancement node H will be generated separately j , and then generate the first new cable operation matrix This does not affect the originally trained model.

[0152] The following formula can be used to calculate the new cable operation characteristic matrix:

[0153]

[0154] In the formula, Z n+1 is the feature matrix in the n+1th group of random feature space, that is, the feature matrix in the random feature space corresponding to the newly added feature window; n+1 is the number of the newly added feature window; X is the cable operation data obtained periodically; and They are the weights and biases generated for the newly added feature nodes respectively.

[0155] The first additional cable operation enhancement matrix can be calculated using the following formula:

[0156]

[0157] In the formula, H j is the jth group enhancement matrix; j is the group number of the enhancement node; α represents the nonlinear activation function; Z n+1 is the feature matrix in the n+1th group of random feature space, that is, the feature matrix in the random feature space corresponding to the newly added feature window; represents the weight and bias of the randomly generated enhancement layer; m represents the total number of groups of feature nodes of the enhancement layer; wherein j can be multiple values.

[0158] The following formula can be used to calculate the first newly added cable operation matrix:

[0159]

[0160] In the formula, Added a cable run matrix for the first; is the original cable operation matrix; Z n+1 is the feature matrix in the n+1th group of random feature space, that is, the feature matrix in the random feature space corresponding to the newly added feature window; H j is the jth group of enhancement matrix; j is the group number of the enhancement layer feature node.

[0161] The updated first target connection weight matrix can be calculated using the following formula:

[0162]

[0163] In the formula, is the updated first target connection weight matrix; is the first newly added cable operation matrix; Y is the actual cable fault type corresponding to the first newly added cable operation matrix; is the target connection weight matrix; E1 is the first intermediate variable; K1 T is the second intermediate variable.

[0164] in,

[0165]

[0166] A n+1 =[|Z n+1 |H j ];

[0167]

[0168] In the formula, E1 is the first intermediate variable; A is the first newly added cable operation matrix; n+1 It is an independent new cable operation matrix generated for the new cable operation data; Z n+1 is the feature matrix in the n+1th group of random feature space, that is, the feature matrix in the random feature space corresponding to the newly added feature window; H j is the jth group enhancement matrix; j is the group number of the enhancement layer feature node; F1 is the third intermediate variable; Matrix for original cable operation; K1 T is the second intermediate variable.

[0169] Secondly, for the incremental method of enhanced nodes, assuming that a new m+1 group of enhanced nodes is added to the cable fault detection sub-model, the additional enhanced nodes H m+1 The second additional cable operation matrix of the cable fault detection submodel will be generated This does not affect the originally trained model.

[0170] The second additional cable operation enhancement matrix can be calculated using the following formula:

[0171]

[0172] In the formula, H m+1 is the m+1th group enhancement matrix, i.e., the second newly added cable operation enhancement matrix; m+1 is the group number of the newly added enhancement node; α represents the nonlinear activation function; Z n Run the characteristic matrix for the cable; and Represents the weights and biases of the randomly generated enhancement layers for the newly added enhancement nodes.

[0173] The second additional cable operation matrix can be calculated using the following formula:

[0174]

[0175] In the formula, Added cable runs matrix for the second addition; is the cable operation matrix; H m+1 It is the m+1th group of enhanced matrix, that is, the second newly added cable operation enhanced matrix; m+1 is the group number of the newly added enhanced node.

[0176] The updated second target connection weight matrix can be calculated using the following formula:

[0177]

[0178] In the formula, The updated second target connection weight matrix; is the second newly added cable operation matrix; Y is the actual cable fault type corresponding to the first newly added cable operation matrix; is the target connection weight matrix; E2 is the fourth intermediate variable; K2 T is the fifth intermediate variable.

[0179] in,

[0180]

[0181] Wherein, E2 is the fourth intermediate variable; is the original cable operation matrix; H m+1 is the m+1th group of enhanced matrix, i.e., the second newly added cable operation enhanced matrix; m+1 is the group number of the newly added enhanced node; F2 is the sixth intermediate variable; K2 T is the fifth intermediate variable.

[0182] At present, intelligent diagnosis of cables is generally achieved by building automatic recognition models. Traditional machine learning methods such as support vector machines, BP neural networks or random forests have been widely used in the field of intelligent diagnosis. With the rapid development of deep learning technology, advanced models such as convolutional neural networks, recurrent neural networks, generative adversarial networks, autoencoders and long short-term memory networks have gradually been introduced into the field of intelligent diagnosis to further improve the accuracy and efficiency of diagnosis. However, traditional machine learning algorithms often require complex feature extraction and screening of the original data of the cable, and the fault recognition rate depends to a large extent on the selected features, which not only requires certain expert knowledge and increases labor costs, but also may affect the diagnosis effect. In contrast, deep learning methods usually adopt an end-to-end classification mode, which can achieve automatic diagnosis without feature extraction. However, deep learning algorithms have high requirements on hardware resources, long model training time, and if the model is deployed in cable edge devices, it needs to consume a lot of computing resources. In addition, in the construction process of the cable intelligent diagnosis model, the number of normal data samples is usually much larger than that of abnormal samples. How to overcome the class imbalance problem has become one of the key challenges to improve the diagnosis efficiency.

[0183] This solution can not only effectively overcome the problem of data imbalance, but also significantly reduce the dependence on hardware resources while ensuring the accuracy of diagnosis, and achieve more efficient and accurate cable fault detection; the cable fault detection model in this solution can be quickly updated without the need to retrain the model, reducing resource loss and achieving higher cost-effectiveness, making the power grid more adaptive and responsive, and able to adjust predictions and decisions in real time, thereby improving the reliability and efficiency of power grid operation. With the continuous expansion of the scale and complexity of the power grid, traditional model training methods have been unable to cope with the processing needs of massive data. The incremental algorithm of this solution will continue to play its advantages and promote the development of the power grid in a more intelligent and automated direction.

[0184] Embodiment 3

[0185] Figure 5 A schematic diagram of the structure of a cable fault detection device provided in Embodiment 3 of the present invention. The embodiment of the present invention can be applied to the case of detecting cable faults in a distribution network, the device can execute a cable fault detection method, the device can be implemented in the form of hardware and / or software, and the device can be configured in an electronic device that carries a cable fault detection function, such as a fault analysis server.

[0186] See also Figure 5The cable fault detection device shown includes: a cable operation data acquisition module 510, a cable fault type prediction model 520 and a cable fault type voting module 530, wherein the cable operation data acquisition module 510 is used to obtain cable operation data; the cable fault type prediction model 520 is used to input the cable operation data into a pre-trained cable fault detection model to obtain the alternative cable fault type predicted by each cable fault detection sub-model; wherein the cable fault detection model includes at least one cable fault detection sub-model; the cable fault type voting module 530 is used to vote for each of the alternative cable fault types to determine the target cable fault type of the cable operation data.

[0187] The technical solution of the embodiment of the present invention performs real-time detection of cable operation data by pre-training each cable fault detection sub-model in the cable fault detection model, thereby realizing automatic detection of cable faults, reducing the operation and maintenance cost of cable fault detection, and improving the detection efficiency of cable fault detection. By voting on each alternative cable fault type and determining the target cable fault type, the detection accuracy of the cable fault can be further improved.

[0188] In an optional embodiment of the present invention, the device also includes: a cable operation sample acquisition module, which is used to acquire cable operation samples and actual cable fault types before acquiring cable operation data; a classification loss weight determination module, which is used to determine the cable fault detection sub-model corresponding to each of the actual cable fault types and the classification loss weight of each of the cable fault detection sub-models; wherein the weight value of the corresponding actual cable fault type in the classification loss weight of the cable fault detection sub-model is higher; a target connection weight matrix calculation module, which is used to calculate the target connection weight matrix of the cable fault detection sub-model for a single cable fault detection sub-model based on the cable operation sample, the classification loss weight of the cable fault detection sub-model and the corresponding actual cable fault type; a cable fault detection sub-model training module, which is used to adjust the parameters of each of the cable fault detection sub-models according to the target connection weight matrix of each of the cable fault detection sub-models.

[0189] In an optional embodiment of the present invention, the device also includes: a first periodic data acquisition module, which is used to periodically acquire cable operation data and the corresponding actual cable fault type after voting on each of the alternative cable fault types to determine the target cable fault type of the cable operation data; for a single cable fault detection sub-model, feature mapping is performed on the cable operation data through the feature nodes of the feature layer of the cable fault detection sub-model to obtain a cable operation feature matrix; a newly added cable operation feature matrix generation module is used to feature map the cable operation data through the newly added feature nodes of the feature layer of the cable fault detection sub-model for a single cable fault detection sub-model to obtain a newly added cable operation feature matrix; a first newly added cable operation enhancement matrix generation module is used to map the cable operation data through the first newly added enhancement node of the enhancement layer of the cable fault detection sub-model for a single cable fault detection sub-model. The cable operation feature matrix is ​​nonlinearly activated to obtain a first newly added cable operation enhancement matrix; the first newly added cable operation matrix generation module is used to obtain the original cable operation matrix for a single cable fault detection sub-model through the first newly added enhancement node of the enhancement layer of the cable fault detection sub-model, and connect the original cable operation matrix, the newly added cable operation feature matrix and the first newly added cable operation enhancement matrix to obtain the first newly added cable operation matrix; the first target connection weight matrix update module is used to update the target connection weight matrix for a single cable fault detection sub-model through the output layer of the cable fault detection sub-model based on the first newly added cable operation matrix and the corresponding actual cable fault type to obtain the first target connection weight matrix; the first cable fault detection sub-model update module is used to adjust the parameters of each cable fault detection sub-model according to each of the first target connection weight matrices.

[0190] In an optional embodiment of the present invention, the device also includes: a second periodic data acquisition module, which is used to periodically acquire cable operation data and the corresponding actual cable fault type after voting on the cable fault types predicted by each of the cable fault detection sub-models to determine the target cable fault type of the cable operation data; a cable operation feature matrix generation module, which is used to perform feature mapping on the cable operation data for a single cable fault detection sub-model through the feature nodes of the feature layer of the cable fault detection sub-model to obtain a cable operation feature matrix; a second newly added cable operation enhancement matrix generation module, which is used to perform nonlinear activation on the cable operation data for a single cable fault detection sub-model through the second newly added enhancement node of the enhancement layer of the cable fault detection sub-model to obtain a second newly added cable operation Enhanced matrix; a second newly added cable operation matrix generating module, used for obtaining the original cable operation matrix through the second newly added enhanced node of the enhanced layer of the cable fault detection submodel for a single cable fault detection submodel, and connecting the original cable operation matrix and the second newly added cable operation enhanced matrix to obtain the second newly added cable operation matrix; a second target connection weight matrix updating module, used for updating the target connection weight matrix through the output layer of the cable fault detection submodel for a single cable fault detection submodel based on the second newly added cable operation matrix and the corresponding actual cable fault type to obtain the second target connection weight matrix; a second cable fault detection submodel updating module, used for adjusting the parameters of each cable fault detection submodel according to each second target connection weight matrix.

[0191] In an optional embodiment of the present invention, the target connection weight matrix calculation module includes: a cable operation sample division unit, used to divide the cable operation samples according to the sample proportion corresponding to each actual cable fault type in the cable operation samples and the number of the cable fault detection sub-models, so as to obtain unit cable operation samples of each cable fault detection sub-model; a target connection weight matrix calculation unit, used to calculate the target connection weight matrix of the cable fault detection sub-model for a single cable fault detection sub-model based on the unit cable operation samples of the cable fault detection sub-model, the classification loss weight of the cable fault detection sub-model and the corresponding actual cable fault type.

[0192] In an optional embodiment of the present invention, the cable operation data includes cable current, cable temperature and cable load.

[0193] The cable fault detection device provided in the embodiment of the present invention can execute the cable fault detection method provided in any embodiment of the present invention, and has the corresponding functional modules and beneficial effects of the execution method.

[0194] In the technical solution of the embodiment of the present invention, the acquisition, storage and application of cable operation data, cable operation samples and actual cable fault types involved all comply with the provisions of relevant laws and regulations and do not violate public order and good morals.

[0195] Embodiment 4

[0196] Figure 6 A schematic diagram of the structure of an electronic device 600 that can be used to implement an embodiment of the present invention is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices (such as helmets, glasses, watches, etc.) and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present invention described and / or required herein.

[0197] like Figure 6 As shown, the electronic device 600 includes at least one processor 601, and a memory connected to the at least one processor 601, such as a read-only memory (ROM) 602, a random access memory (RAM) 603, etc., wherein the memory stores a computer program that can be executed by at least one processor, and the processor 601 can perform various appropriate actions and processes according to the computer program stored in the read-only memory (ROM) 602 or the computer program loaded from the storage unit 608 to the random access memory (RAM) 603. In the RAM 603, various programs and data required for the operation of the electronic device 600 can also be stored. The processor 601, the ROM 602, and the RAM 603 are connected to each other via a bus 604. An input / output (I / O) interface 605 is also connected to the bus 604.

[0198] Multiple components in the electronic device 600 are connected to the I / O interface 605, including: an input unit 606, such as a keyboard, a mouse, etc.; an output unit 607, such as various types of displays, speakers, etc.; a storage unit 608, such as a disk, an optical disk, etc.; and a communication unit 609, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 609 allows the electronic device 600 to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks.

[0199] The processor 601 may be a variety of general and / or special processing components with processing and computing capabilities. Some examples of the processor 601 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, a digital signal processor (DSP), and any appropriate processor, controller, microcontroller, etc. The processor 601 performs the various methods and processes described above, such as a cable fault detection method.

[0200] In some embodiments, the cable fault detection method may be implemented as a computer program, which is tangibly contained in a computer-readable storage medium, such as a storage unit 608. In some embodiments, part or all of the computer program may be loaded and / or installed on the electronic device 600 via the ROM 602 and / or the communication unit 609. When the computer program is loaded into the RAM 603 and executed by the processor 601, one or more steps of the cable fault detection method described above may be performed. Alternatively, in other embodiments, the processor 601 may be configured to perform the cable fault detection method in any other appropriate manner (e.g., by means of firmware).

[0201] Various implementations of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), complex programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various implementations can include: being implemented in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0202] Computer programs for implementing the methods of the present invention may be written in any combination of one or more programming languages. These computer programs may be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, so that when the computer program is executed by the processor, the functions / operations specified in the flow chart and / or block diagram are implemented. The computer program may be executed entirely on the machine, partially on the machine, partially on the machine and partially on a remote machine as a stand-alone software package, or entirely on a remote machine or server.

[0203] In the context of the present invention, a computer-readable storage medium may be a tangible medium that may contain or store a computer program for use by or in combination with an instruction execution system, device or equipment. A computer-readable storage medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices or equipment, or any suitable combination of the foregoing. Alternatively, a computer-readable storage medium may be a machine-readable signal medium. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0204] To provide interaction with a user, the systems and techniques described herein may be implemented on an electronic device having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and a pointing device (e.g., a mouse or trackball) through which the user can provide input to the electronic device. Other types of devices may also be used to provide interaction with the user; for example, the feedback provided to the user may be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user may be received in any form (including acoustic input, voice input, or tactile input).

[0205] The systems and techniques described herein may be implemented in a computing system that includes backend components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes frontend components (e.g., a user computer with a graphical user interface or a web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such backend components, middleware components, or frontend components. The components of the system may be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: a local area network (LAN), a wide area network (WAN), a blockchain network, and the Internet.

[0206] A computing system may include a client and a server. The client and the server are generally remote from each other and usually interact through a communication network. The relationship between the client and the server is generated by computer programs running on the corresponding computers and having a client-server relationship with each other. The server may be a cloud server, also known as a cloud computing server or a cloud host, which is a host product in the cloud computing service system to solve the defects of difficult management and weak business scalability in traditional physical hosts and VPS (Virtual Private Server) services.

[0207] It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps described in the present invention can be executed in parallel, sequentially or in different orders, as long as the desired results of the technical solution of the present invention can be achieved, and this document does not limit this.

[0208] The above specific implementations do not constitute a limitation on the protection scope of the present invention. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modification, equivalent substitution and improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A cable fault detection method, characterized in that: The method comprises: Obtain cable operation data; Inputting the cable operation data into a pre-trained cable fault detection model to obtain candidate cable fault types predicted by each cable fault detection sub-model; wherein the cable fault detection model includes at least one cable fault detection sub-model; Voting is performed on each of the candidate cable fault types to determine a target cable fault type of the cable operation data.

2. The method according to claim 1, characterized in that Before obtaining the cable operation data, the method further includes: Obtain cable operation samples and actual cable fault types; Determine the cable fault detection sub-model corresponding to each of the actual cable fault types and the classification loss weight of each of the cable fault detection sub-models; wherein the weight value of the actual cable fault type corresponding to the classification loss weight of the cable fault detection sub-model is higher; For a single cable fault detection sub-model, based on the cable operation sample, the classification loss weight of the cable fault detection sub-model and the corresponding actual cable fault type, a target connection weight matrix of the cable fault detection sub-model is calculated; According to the target connection weight matrix of each cable fault detection sub-model, the parameters of each cable fault detection sub-model are adjusted.

3. The method according to claim 2, characterized in that After voting on each of the candidate cable fault types to determine the target cable fault type of the cable operation data, the method further includes: Periodically obtain cable operation data and corresponding actual cable fault types; For a single cable fault detection sub-model, feature mapping is performed on the cable operation data through a newly added feature node of the feature layer of the cable fault detection sub-model to obtain a newly added cable operation feature matrix; For a single cable fault detection sub-model, nonlinearly activate the newly added cable operation feature matrix through a first newly added enhancement node of the enhancement layer of the cable fault detection sub-model to obtain a first newly added cable operation enhancement matrix; For a single cable fault detection submodel, an original cable operation matrix is ​​obtained through a first newly added enhancement node of the enhancement layer of the cable fault detection submodel, and the original cable operation matrix, the newly added cable operation feature matrix and the first newly added cable operation enhancement matrix are connected to obtain a first newly added cable operation matrix; For a single cable fault detection sub-model, the target connection weight matrix is ​​updated through the output layer of the cable fault detection sub-model based on the first newly added cable operation matrix and the corresponding actual cable fault type to obtain a first target connection weight matrix; According to each of the first target connection weight matrices, parameters of each of the cable fault detection sub-models are adjusted.

4. The method according to claim 2, characterized in that: After voting on the cable fault types predicted by the cable fault detection sub-models to determine the target cable fault type of the cable operation data, the method further includes: Periodically obtain cable operation data and corresponding actual cable fault types; For a single cable fault detection sub-model, feature mapping is performed on the cable operation data through feature nodes of a feature layer of the cable fault detection sub-model to obtain a cable operation feature matrix; For a single cable fault detection sub-model, nonlinearly activate the cable operation data through a second newly added enhancement node of the enhancement layer of the cable fault detection sub-model to obtain a second newly added cable operation enhancement matrix; For a single cable fault detection sub-model, an original cable operation matrix is ​​obtained through a second newly added enhancement node of the enhancement layer of the cable fault detection sub-model, and the original cable operation matrix and the second newly added cable operation enhancement matrix are connected to obtain a second newly added cable operation matrix; For a single cable fault detection sub-model, the target connection weight matrix is ​​updated through the output layer of the cable fault detection sub-model based on the second newly added cable operation matrix and the corresponding actual cable fault type to obtain a second target connection weight matrix; According to each of the second target connection weight matrices, parameters of each of the cable fault detection sub-models are adjusted.

5. The method according to claim 2, characterized in that: The step of calculating the target connection weight matrix of the cable fault detection submodel for a single cable fault detection submodel based on the cable operation sample, the classification loss weight of the cable fault detection submodel and the corresponding actual cable fault type comprises: According to the sample proportion corresponding to each actual cable fault type in the cable operation sample and the number of the cable fault detection sub-models, the cable operation sample is divided to obtain unit cable operation samples of each cable fault detection sub-model; For a single cable fault detection sub-model, a target connection weight matrix of the cable fault detection sub-model is calculated based on the unit cable operation samples of the cable fault detection sub-model, the classification loss weight of the cable fault detection sub-model and the corresponding actual cable fault type.

6. The method according to claim 1, characterized in that The cable operation data includes cable current, cable temperature and cable load.

7. A cable fault detection device, characterized in that: The device comprises: A cable operation data acquisition module is used to acquire cable operation data; A cable fault type prediction model, used for inputting the cable operation data into a pre-trained cable fault detection model to obtain candidate cable fault types predicted by each cable fault detection sub-model; wherein the cable fault detection model includes at least one cable fault detection sub-model; The cable fault type voting module is used to vote on each of the candidate cable fault types to determine the target cable fault type of the cable operation data.

8. An electronic device, characterized in that: The electronic device comprises: at least one processor; and a memory communicatively connected to the at least one processor; wherein, The memory stores a computer program executable by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can perform the cable fault detection method according to any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a processor to implement the cable fault detection method according to any one of claims 1 to 6 when executed.

10. A computer program product, characterized in that The computer program product comprises a computer program, and when the computer program is executed by a processor, the computer program implements the cable fault detection method according to any one of claims 1 to 6.

Images