Breaker switch-based fault type determination model training method and device

Abstract

The invention provides a training method and device for a fault type determination model based on a circuit breaker switch. The method comprises the steps of obtaining a to-be-trained data set; wherein the to-be-trained data set comprises a historical data matrix when the circuit breaker switch breaks down and a tag vector corresponding to the historical data matrix, the historical data matrix comprises a plurality of sub-matrixes, and each row in the sub-matrixes represents the state of the circuit breaker switch during preset operation; the preset operation comprises at least one of switching-on operation, switching-off operation by using a main switching-off coil and switching-off operation by using an auxiliary switching-off coil, sub-matrixes in the historical data matrix are vertically arranged, and the tag vector represents an actual fault type of the circuit breaker switch; determining a target data matrix according to the historical data matrix; and according to the target data matrix and the corresponding label vector, training the initial model to obtain a trained fault type determination model. According to the method, the accuracy of determining the fault type of the circuit breaker switch is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of power systems, and in particular to a training method and device for a fault type determination model based on a circuit breaker switch. BACKGROUND

[0002] In a power system, a circuit breaker switch can refer to a mechanical switching device that can be closed, carried, and opened under normal circuit current conditions. It is an important device in the energy storage field and other power fields. For example, it can quickly cut off the circuit to protect the energy storage device from damage caused by overcharging, overdischarging, or short circuit. However, the operation or running of the circuit breaker switch can also face failures, and the fault type of the circuit breaker switch needs to be determined or judged to take targeted maintenance measures for the fault of the circuit breaker switch.

[0003] In the prior art, the fault type of the circuit breaker switch is determined by manual judgment. However, manual judgment relies on personal experience and professional knowledge, and there is a large difference between the judgment results of different personnel, resulting in poor stability of determining the fault type of the circuit breaker switch and high labor cost. At the same time, the power system is complex and variable, and the fault type is various and complex, so it is difficult to accurately identify the fault type by relying on manual experience.

[0004] Therefore, the accuracy of determining the fault type of the circuit breaker switch is low. SUMMARY

[0005] The present application provides a training method and device for a fault type determination model based on a circuit breaker switch to solve the technical problem of low accuracy in determining the fault type of the circuit breaker switch.

[0006] In a first aspect, the present application provides a training method for a fault type determination model based on a circuit breaker switch, comprising:

[0007] obtaining a training data set; wherein the training data set includes a historical data matrix of the circuit breaker switch when a fault occurs and a label vector corresponding to the historical data matrix, the historical data matrix includes a plurality of sub-matrices, each row in the sub-matrix represents the state of the circuit breaker switch when a preset operation is performed, the preset operation includes at least one of a closing operation, a main trip coil tripping operation, and a secondary trip coil tripping operation, the sub-matrices in the historical data matrix are arranged vertically, and the label vector represents the actual fault type of the circuit breaker switch;

[0008] determining a target data matrix according to the historical data matrix; wherein the target data matrix represents the elements in the historical data matrix and the positions of the elements in the historical data matrix;

[0009] According to the target data matrix and the corresponding label vector, the initial model is trained to obtain a trained fault type determination model; wherein the fault type determination model is used to predict the fault type of the circuit breaker switch.

[0010] Optionally, according to the above method, the initial model includes a plurality of network layers, each network layer corresponds to at least one perspective matrix, and the perspective matrix represents the prediction angle of the network layer to the fault type; according to the target data matrix and the corresponding label vector, the initial model is trained to obtain a trained fault type determination model, including:

[0011] According to the target data matrix and the perspective matrix corresponding to the network layer, a target feature matrix corresponding to the network layer is determined; wherein the target feature matrix represents the target data matrix under the prediction angle of the network layer;

[0012] Each target feature matrix is input into the corresponding network layer to obtain a probability sequence output by the network layer; wherein the probability sequence includes probability values of different fault types of the circuit breaker switch corresponding to the network layer, and the number of probability values in the probability sequence is consistent with the number of fault types;

[0013] According to the probability sequence output by each network layer and the label vector, the initial model is trained to obtain a trained fault type determination model.

[0014] Optionally, according to the above method, the network layer corresponds to three perspective matrices, and the three perspective matrices have the same dimension; according to the target data matrix and the perspective matrix corresponding to the network layer, the target feature matrix corresponding to the network layer is determined, including:

[0015] The target data matrix and the three perspective matrices corresponding to the network layer are multiplied respectively to obtain three output matrices corresponding to the three perspective matrices respectively; wherein the output matrix represents the form of the target data matrix under the perspective matrix;

[0016] According to the output matrix corresponding to each perspective matrix, the target feature matrix corresponding to the network layer is determined.

[0017] Optionally, according to the above method, the three perspective matrices corresponding to the network layer are a first perspective matrix, a second perspective matrix, and a third perspective matrix respectively; the output matrix is represented as:

[0018] ; ; ;

[0019] wherein, the target data matrix is represented as, the first perspective matrix corresponding to the i-th network layer is represented as, the second perspective matrix corresponding to the i-th network layer is represented as, a third view matrix corresponding to the i-th network layer, an output matrix corresponding to the first view matrix corresponding to the i-th network layer, an output matrix corresponding to the second view matrix corresponding to the i-th network layer, an output matrix corresponding to the third view matrix corresponding to the i-th network layer.

[0020] The target feature matrix is represented as:

[0021] ;

[0022] wherein, the target feature matrix corresponding to the i-th network layer, and h represents the column number of the view matrix corresponding to the network layer.

[0023] Optionally, according to the probability sequence output by each network layer and the label vector, the initial model is trained to obtain the trained fault type determination model, including:

[0024] According to the probability sequence output by each network layer, the prediction result of the fault type is determined.

[0025] According to the prediction result and the label vector, the initial model is trained to obtain the trained fault type determination model.

[0026] Optionally, according to the probability sequence output by each network layer, the prediction result of the fault type is determined, including:

[0027] For all probability sequences, the probability values in the same position in each probability sequence are determined, and the probability average value corresponding to the position is determined; wherein, the position corresponds to one of the fault types, and the same position in each probability sequence represents the same fault type;

[0028] According to the probability average value corresponding to each position, the prediction result of the fault type is determined.

[0029] Optionally, according to the historical data matrix, the target data matrix is determined, including:

[0030] According to the dimension of the historical data matrix, a position vector matrix is determined; wherein, the dimension of the position vector matrix is the same as that of the historical data matrix, and the elements in the position vector matrix represent the order of the positions of the elements in the historical data matrix;

[0031] According to the historical data matrix and the position vector matrix, the target data matrix corresponding to the historical data matrix is determined.

[0032] Optionally, the elements in the position vector matrix are represented as:

[0033] ;

[0034] wherein, n represents the total number of columns of elements in the position vector matrix.

[0035] In a second aspect, the present application provides a method for determining the fault type of a circuit breaker switch, comprising:

[0036] obtaining a current data matrix of the circuit breaker switch; wherein the current data matrix comprises a plurality of sub-matrices, each row in the sub-matrices representing the state of the circuit breaker switch when performing a preset operation, the preset operation comprising at least one of closing operation, opening operation using a main opening coil, and opening operation using a secondary opening coil, and the sub-matrices in the current data matrix are arranged vertically;

[0037] inputting the current data matrix into a fault type determination model to obtain the output fault type of the circuit breaker switch; wherein the fault type determination model is any one of the fault type determination models in the first aspect.

[0038] In a third aspect, the present application provides a training device for a fault type determination model of a circuit breaker switch, comprising:

[0039] a first obtaining unit configured to obtain a training data set; wherein the training data set comprises a historical data matrix of the circuit breaker switch when a fault occurs and a label vector corresponding to the historical data matrix, the historical data matrix comprises a plurality of sub-matrices, each row in the sub-matrices representing the state of the circuit breaker switch when performing a preset operation, the preset operation comprising at least one of closing operation, opening operation using a main opening coil, and opening operation using a secondary opening coil, and the sub-matrices in the historical data matrix are arranged vertically, and the label vector represents the actual fault type of the circuit breaker switch;

[0040] a determining unit configured to determine a target data matrix according to the historical data matrix; wherein the target data matrix represents the elements in the historical data matrix and the positions of the elements in the historical data matrix;

[0041] a training unit configured to train an initial model according to the target data matrix and the corresponding label vector to obtain a trained fault type determination model; wherein the fault type determination model is used to predict the fault type of the circuit breaker switch.

[0042] In a fourth aspect, the present application provides a device for determining the fault type of a circuit breaker switch, comprising:

[0043] The second acquisition unit is configured to acquire a current data matrix of the circuit breaker switch, wherein the current data matrix comprises a plurality of sub-matrices, each row in the sub-matrix represents a state of the circuit breaker switch when performing a preset operation, the preset operation comprises at least one of a closing operation, a tripping operation using a main tripping coil, and a tripping operation using a secondary tripping coil, and the sub-matrices in the current data matrix are arranged vertically.

[0044] The output unit is configured to input the current data matrix into the fault type determination model to obtain an output fault type of the circuit breaker switch, wherein the fault type determination model is the fault type determination model of any one of the first aspect and / or the various possible implementation manners of the first aspect.

[0045] In a fifth aspect, the present application provides an electronic device, comprising: a memory, a processor;

[0046] The memory stores computer execution instructions;

[0047] The processor executes the computer execution instructions stored in the memory, so that the processor executes the first aspect and / or various possible implementation manners of the first aspect, the second aspect and / or various possible implementation manners of the second aspect.

[0048] In a sixth aspect, the present application provides a computer readable storage medium, the computer readable storage medium stores computer execution instructions, and the computer execution instructions are executed by the processor to realize the first aspect and / or various possible implementation manners of the first aspect, the second aspect and / or various possible implementation manners of the second aspect.

[0049] In a seventh aspect, the present application provides a computer program product, comprising a computer program, the computer program is executed by the processor to realize the first aspect and / or various possible implementation manners of the first aspect, the second aspect and / or various possible implementation manners of the second aspect.

[0050] The application provides a training method and device of a fault type determination model based on a circuit breaker switch, historical data matrix and a label vector corresponding to the historical data matrix of the circuit breaker switch when a fault occurs are acquired, wherein the historical data matrix includes a plurality of sub-matrices, each row in the sub-matrices represents a state of the circuit breaker switch when a preset operation is performed, the preset operation includes one of closing operation, opening operation using a main opening coil and opening operation using a secondary opening coil, further, in order to distinguish the elements in the historical data matrix in a certain sequence, a target data matrix is determined according to the historical data matrix, further, an initial model is trained according to the target data matrix and the corresponding label vector, and a trained fault type determination model is obtained, and the fault type of the circuit breaker switch output by the fault type determination model can be obtained by inputting a current data matrix into the fault type determination model. The method of the application considers the time sequence of the elements in the historical data matrix in the acquisition process, and considers the case that the preset operation includes closing operation, opening operation using a main opening coil and opening operation using a secondary opening coil, thereby ensuring the richness of the elements in the historical data matrix of the circuit breaker switch when a fault occurs, improving the accuracy of determining the fault type of the circuit breaker switch, and further, the fault type of the current circuit breaker switch can be determined by acquiring a current data matrix in an actual scene, without manually determining the fault type of the current circuit breaker switch, thereby ensuring the efficiency of determining the fault type of the circuit breaker switch, and also ensuring the accuracy of determining the fault type of the circuit breaker switch. The method of the application improves the accuracy of determining the fault type of the circuit breaker switch. BRIEF DESCRIPTION OF DRAWINGS

[0051] The accompanying drawings, which are incorporated herein and form part of the specification, illustrate embodiments consistent with the present application and, together with the description, further serve to explain the principles of the application.

[0052] Figure 1 A flowchart of a training method of a fault type determination model based on a circuit breaker switch provided by the application Figure One ;

[0053] Figure 2 A flowchart of a training method of a fault type determination model based on a circuit breaker switch provided by the application Figure Two ;

[0054] Figure 3 A flowchart of a training method of a fault type determination model based on a circuit breaker switch provided by the application Figure Three ;

[0055] Figure 4 A flowchart of a fault type determination method based on a circuit breaker switch provided by the application

[0056] Figure 5 A structure diagram of a training device of a fault type determination model based on a circuit breaker switch provided by the present application Figure One ;

[0057] Figure 6 A structure diagram of a training device of a fault type determination model based on a circuit breaker switch provided by the present application Figure Two ;

[0058] Figure 7 A structure diagram of a fault type determination device based on a circuit breaker switch provided by the present application

[0059] Figure 8 A structure diagram of an electronic device provided by the present application.

[0060] The specific embodiments of the present application have been shown in the above-described drawings, and will be described in more detail hereinafter. These drawings and the written description are not intended to restrict the scope of the present application concept in any way, but to illustrate the present application concept to those skilled in the art by referring to specific embodiments. DETAILED DESCRIPTION

[0061] The exemplary embodiments will be described in detail herein with reference to the attached drawings. The same or similar components are denoted by the same reference numerals throughout the drawings and the following description, unless otherwise specified. The embodiments described in the following exemplary embodiments are not intended to represent all embodiments consistent with the present application. Rather, they are merely examples of apparatuses and methods consistent with some aspects of the present application as detailed in the appended claims.

[0062] It should be noted that the user information (including but not limited to user equipment information, user personal information, etc.) and data (including but not limited to data for analysis, stored data, displayed data, etc.) involved in the present application are all information and data authorized by the user or authorized by all parties, and the collection, use and processing of related data need to comply with relevant laws, regulations and standards, and provide corresponding operation portal for user to choose authorization or refusal.

[0063] In the power system, the circuit breaker switch can refer to a mechanical switching device that can be closed, carried and opened under normal loop current, and is an important device in the energy storage field and other power fields.

[0064] Specifically, the important role of the circuit breaker switch in the field of energy storage mainly lies in: when the energy storage system fails due to short circuit, overload, etc., the circuit breaker can quickly respond and cut off the fault circuit in time to prevent the fault current from causing irreversible damage to the energy storage equipment and ensure the safety of the energy storage system; when the energy storage system is maintained, repaired or equipment is replaced, the circuit breaker can safely isolate the energy storage equipment from the power grid or other electrical equipment to ensure the safety of the maintenance personnel and prevent electrical accidents caused by misoperation; the circuit breaker switch can flexibly realize the on-off control of the circuit according to the operating state and control strategy of the energy storage system, thereby accurately regulating the charging and discharging process of the energy storage equipment, for example, during the low load period of the power grid, the circuit breaker controls the energy storage system to charge and store excess power, while during the peak load period of the power grid, the circuit breaker controls the energy storage system to discharge and deliver power to the power grid, achieving the effect of peak load shifting to optimize the operating efficiency of the power grid.

[0065] However, the operation or running of the circuit breaker switch may also face failures, and the type of failure of the circuit breaker switch needs to be judged or determined so that targeted maintenance measures can be taken for the failure of the circuit breaker switch.

[0066] In the prior art, the type of failure of the circuit breaker switch is judged manually, however, manual judgment relies on personal experience and professional knowledge, and there is a large difference between the judgment results of different personnel, resulting in poor stability of judging the type of failure of the circuit breaker switch and high labor cost; at the same time, the power system is complex and variable, and the types of failure are various and complex, and it is difficult to accurately identify the type of failure only by manual experience.

[0067] Therefore, the accuracy of determining the type of failure of the circuit breaker switch is low.

[0068] The training method and device for the failure type determination model of the circuit breaker switch provided in the present application, by obtaining a historical data matrix of the circuit breaker switch when a failure occurs and a label vector corresponding to the historical data matrix, wherein the historical data matrix includes a plurality of sub-matrices, each row in the sub-matrix represents the state of the circuit breaker switch when a preset operation is performed, the preset operation includes one of closing operation, opening operation using a main opening coil, and opening operation using a secondary opening coil, further, in order to make the elements in the historical data matrix have a certain order, a target data matrix is determined according to the historical data matrix, further, the initial model is trained according to the target data matrix and the corresponding label vector to obtain a trained failure type determination model, which can realize that the current data matrix is input into the failure type determination model to obtain the type of failure of the circuit breaker switch output by the failure type determination model.

[0069] The method of the present application considers the time sequence of elements in the historical data matrix in the acquisition process, and considers the case that the preset operation includes closing operation, opening operation using the main opening coil, and opening operation using the auxiliary opening coil, ensures the richness of the elements in the historical data matrix of the circuit breaker switch when a fault occurs, improves the accuracy of determining the fault type of the circuit breaker switch, and further, in actual scenarios, the fault type of the current circuit breaker switch can be determined by acquiring the current data matrix, without the need for manual determination of the fault type of the current circuit breaker switch, while ensuring the efficiency of determining the fault type of the circuit breaker switch, also ensures the accuracy of determining the fault type of the circuit breaker switch.

[0070] The method of the present application improves the accuracy of determining the fault type of the circuit breaker switch.

[0071] The technical solutions of the present application and how the technical solutions of the present application solve the above technical problems will be described in detail below with specific examples. The following specific examples can be combined with each other, and the same or similar concepts or processes may not be described again in some examples. The embodiments of the present application will be described below with reference to the accompanying drawings.

[0072] Figure 1 A flowchart of a training method of a circuit breaker switch fault type determination model provided by the present application Figure One The execution subject of the method can be a server, a host or other equipment, as shown in Figure 1 The method can include:

[0073] S101, acquiring a to-be-trained data set; wherein the to-be-trained data set includes a historical data matrix of a circuit breaker switch when a fault occurs and a label vector corresponding to the historical data matrix, the historical data matrix includes a plurality of sub-matrices, each row in the sub-matrix represents the state of the circuit breaker switch when a preset operation is performed, the preset operation includes at least one of closing operation, opening operation using a main opening coil, and opening operation using an auxiliary opening coil, the sub-matrices in the historical data matrix are arranged vertically, and the label vector represents the actual fault type of the circuit breaker switch.

[0074] The source of the to-be-trained data set can be data in the historical fault record of the circuit breaker switch actually operated in the power system, and the source of the to-be-trained data set can also be other conditions. The to-be-trained data set includes a historical data matrix of a circuit breaker switch when a fault occurs and a label vector corresponding to the historical data matrix. Illustratively, the to-be-trained data set includes a plurality of historical data matrices and label vectors corresponding to the historical data matrices, for example, the to-be-trained data set includes 10,000 historical data matrices and label vectors corresponding to the historical data matrices.

[0075] The historical data matrix can refer to the states of the circuit breaker switch when performing a plurality of preset operations. The historical data matrix includes a plurality of sub-matrices, and each sub-matrix can refer to the state of the circuit breaker switch when performing a preset operation. For example, the historical data matrix can refer to the states of the circuit breaker switch when performing three preset operations, and the historical data matrix includes three sub-matrices.

[0076] The preset operation can refer to a preset operation performed on the circuit breaker switch, and the preset operation can include at least one of a closing operation, a main trip coil tripping operation, and a secondary trip coil tripping operation.

[0077] The closing operation can refer to connecting the moving contact and the stationary contact of the circuit breaker switch. The main trip coil tripping operation can refer to energizing the main trip coil, and after the main trip coil generates an electromagnetic force, the electromagnetic force separates the moving contact and the stationary contact of the circuit breaker switch. The secondary trip coil tripping operation can refer to energizing the secondary trip coil, and after the secondary trip coil generates an electromagnetic force, the electromagnetic force separates the moving contact and the stationary contact of the circuit breaker switch.

[0078] The state of the preset operation can refer to state parameter information of the circuit breaker switch when performing the preset operation. The state of the preset operation can include state parameter information of the circuit breaker switch when performing the closing operation, state parameter information of the circuit breaker switch when performing the main trip coil tripping operation, and state parameter information of the circuit breaker switch when performing the secondary trip coil tripping operation.

[0079] The state parameter information of the circuit breaker switch when performing the closing operation can include, but is not limited to, a closing speed, a closing time, a closing stroke, a closing bounce number, and a closing loop resistance of the switch. The closing speed can refer to the movement speed of the moving contact during the closing process. The closing time can refer to the movement time interval of the moving contact during the closing process. The closing stroke can refer to the movement stroke of the moving contact during the closing process. The closing bounce number can refer to the number of bounces of the moving contact after contacting the stationary contact during the closing process. The closing loop resistance of the switch can refer to the resistance value between the moving contact and the stationary contact of the circuit breaker switch in the closing state. The closing loop resistance of the switch can be measured by using a preset loop resistance tester and the like. The loop resistance is usually within a preset loop resistance range, for example, the loop resistance of a 10kV (kilovolt) vacuum circuit breaker switch is generally not greater than 100μΩ (micro-ohm).

[0080] The state parameter information of the circuit breaker switch during the opening operation using the main opening coil can include, but is not limited to, opening speed, opening time, opening stroke, contact opening distance, switch insulation resistance during opening, and overstroke. The opening speed can refer to the speed of the moving contact from the closed state to the completely disconnected state during the opening process. The opening time can refer to the time interval of the moving contact from the closed state to the completely disconnected state during the opening process. The opening stroke can refer to the movement stroke of the moving contact from the closed position to the completely disconnected position during the opening process. The contact opening distance can refer to the minimum distance between the moving contact and the static contact of the circuit breaker switch in the opening state. The switch insulation resistance during opening can refer to the resistance value between the moving contact and the static contact of the circuit breaker switch in the opening state. The insulation resistance can be measured by a pre-set insulation resistance tester, and the insulation resistance is usually within a pre-set insulation resistance range, for example, the insulation resistance of a 10kV circuit breaker switch is usually not less than 1000MΩ. The overstroke can refer to the distance of the moving contact continuing to move when the moving contact is in the opening position, and the overstroke is usually within a pre-set overstroke range, for example, the overstroke of a 10kV vacuum circuit breaker switch is 3-5mm.

[0081] The state parameter information of the circuit breaker switch during the opening operation using the auxiliary opening coil can include, but is not limited to, opening speed, opening time, opening stroke, contact opening distance, switch insulation resistance during opening, and overstroke.

[0082] The label vector can refer to a representation form for characterizing the actual fault type of the circuit breaker switch.

[0083] The fault type of the circuit breaker switch can include, but is not limited to, overload fault, short circuit fault, ground fault, insulation aging fault, and operating mechanism fault, and the fault type can be artificially pre-determined.

[0084] For example, the label vector can be in the form of a one-hot code, the length of the one-hot code is the number of fault types, one position in the one-hot code corresponds to one fault type, and only one position in the one-hot code has a value of 1, and the values of the other positions are all 0. The position in the one-hot code with a value of 1 corresponds to the fault type characterized by the label vector, which is the actual fault type of the circuit breaker switch. For example, the length of the one-hot code is 3, the first position of the one-hot code corresponds to fault type A, the second position of the one-hot code corresponds to fault type B, and the third position of the one-hot code corresponds to fault type C. If the one-hot code is (1, 0, 0), the one-hot code represents that the actual fault type of the circuit breaker switch is fault type A.

[0085] For example, one historical data matrix in the training data set can be represented as:

[0086] ;

[0087] wherein the historical data matrix includes three sub-matrices, respectively , , , the sub-matrices are arranged vertically, for one of the sub-matrices, the sub-matrix can be expressed as:

[0088] ;

[0089] wherein each row in the sub-matrix represents the state of the circuit breaker switch when performing a preset operation, represents state parameter information of the circuit breaker switch when performing a closing operation, represents state parameter information of the circuit breaker switch when performing a primary trip coil tripping operation, represents state parameter information of the circuit breaker switch when performing a secondary trip coil tripping operation.

[0090] Specifically, the state parameter information of the circuit breaker switch when performing a closing operation may include n elements, the state parameter information of the circuit breaker switch when performing a primary trip coil tripping operation may also include n elements, the state parameter information of the circuit breaker switch when performing a secondary trip coil tripping operation may also include n elements, that is , , The number of elements in each of

[0091] It should be noted that the state of the circuit breaker switch represented by the same row in each sub-matrix when performing a preset operation is consistent, for example, the order of each row in each sub-matrix is , or the order of each row in each sub-matrix is . The beneficial effect of such setting is that the order of each row in the sub-matrix conforms to the normal use scenario of the circuit breaker switch, so as to improve the accuracy and reliability of the subsequent fault type determination model training. For example, when the initial state of the circuit breaker switch is a tripping state, the circuit breaker switch is sequentially operated for the first time closing, primary trip coil tripping, second time closing, and secondary trip coil tripping, represents state parameter information of the circuit breaker switch corresponding to the first time closing, represents state parameter information of the circuit breaker switch corresponding to the primary trip coil tripping, represents state parameter information of the circuit breaker switch corresponding to the secondary trip coil tripping.

[0092] Exemplarily, one historical data matrix includes three sub-matrices, respectively , , , the historical data matrix can be expressed as:

[0093] = = ;

[0094] wherein, represents the state parameter information of the breaker switch in the first sub-matrix in the historical data matrix when performing closing operation, the i-th element in ; represents the state parameter information of the breaker switch in the first sub-matrix in the historical data matrix when performing opening operation using the main opening coil, the i-th element in ; represents the state parameter information of the breaker switch in the first sub-matrix in the historical data matrix when performing opening operation using the auxiliary opening coil, the i-th element in .

[0095] It can be understood that each historical data matrix corresponds to a label vector, and the one-hot code form of the label vector can be:

[0096] ;

[0097] wherein, there are w positions in the label vector , the number of different fault types of the breaker switch is w, each position corresponds to one fault type of the breaker switch, and the position with value 1 in the one-hot code corresponds to the fault type of the actual breaker switch represented by the label vector .

[0098] The historical data matrix and the label vector are in one-to-one correspondence.

[0099] By obtaining the to-be-trained data set, data support is provided for subsequent training of the initial model.

[0100] S102, determining a target data matrix according to the historical data matrix; wherein, the target data matrix represents the elements in the historical data matrix and the positions of the elements in the historical data matrix.

[0101] The target data matrix can refer to a data matrix obtained by performing data processing on the historical data matrix, and the target data matrix represents elements in the historical data matrix and positions of the elements in the historical data matrix.

[0102] It can be understood that, in the process of obtaining the historical data matrix, the elements in the state parameter information of the circuit breaker switch during the preset operation exist in time sequence, and the operation events of the circuit breaker switch during the preset operation also exist in sequence. By introducing the position information into the historical data matrix, the target data matrix corresponding to the historical data matrix can be obtained.

[0103] By determining the target data matrix, the model can better learn the patterns and relationships of the data in the target data matrix in the training process of the fault type determination model, and thus the accuracy of determining the fault type of the circuit breaker switch using the trained fault type determination model can be improved.

[0104] In an optional implementation, the step S102 can include:

[0105] According to the dimension of the historical data matrix, a position vector matrix is determined, wherein the dimension of the position vector matrix is the same as that of the historical data matrix, and each element in the position vector matrix represents the sequence of positions of each element in the historical data matrix; according to the historical data matrix and the position vector matrix, a target data matrix corresponding to the historical data matrix is determined.

[0106] The position vector matrix can refer to a matrix with the same dimension as the historical data matrix, i.e., the dimension of the position vector matrix is the same as that of the historical data matrix. For example, the dimension of the historical data matrix can be , and the dimension of the position vector matrix is also Each element in the position vector matrix has a one-to-one correspondence with each element in the historical data matrix.

[0107] Each element in the position vector matrix represents the position information of the corresponding element in the historical data matrix, i.e., the sequence of rows and columns of the corresponding element in the historical data matrix.

[0108] According to the historical data matrix and the position vector matrix, the target data matrix corresponding to the historical data matrix can be determined, which can include but is not limited to directly splicing each element in the historical data matrix with the element of the corresponding position vector matrix to obtain the target data matrix. It can be understood that the dimension of the position vector matrix is the same as that of the historical data matrix and the target data matrix, for example, all are .

[0109] In an optional implementation, the elements in the position vector matrix are represented as:

[0110] ;

[0111] wherein, represents the value of the element in the i-th row and j-th column of the position vector matrix, and n represents the total number of columns of the elements in the position vector matrix.

[0112] The beneficial effect of such arrangement is that, by using the sine function and the cosine function to determine the values of the elements in the position vector matrix, the target data matrix determined according to the historical data matrix and the position vector matrix can make the fault type determination model better capture the periodic patterns and relative position relationships in the data during the training process of the fault type determination model, thereby improving the accuracy of determining the fault type of the circuit breaker switch. At the same time, the use of the sine function and the cosine function also makes the calculation of the position vector matrix more simple and efficient, which is conducive to improving the training efficiency of the fault type determination model.

[0113] S103, training the initial model according to the target data matrix and the corresponding label vector to obtain a trained fault type determination model; wherein the fault type determination model is used to predict the fault type of the circuit breaker switch.

[0114] wherein, the initial model can refer to an untrained machine learning model, and the initial model can include but is not limited to neural network, support vector machine (SVM), random forest, etc.

[0115] By training the initial model, in the training process, by determining the difference between the label vector output by the initial model and the label vector representing the actual fault type of the circuit breaker switch, the parameters of the initial model are adjusted, and when the trained initial model meets the preset training completion condition, the trained fault determination model can be obtained. The fault type determination model is used to predict the fault type of the circuit breaker switch.

[0116] Exemplarily, the preset training completion condition can include but is not limited to model convergence, reaching the maximum number of training rounds.

[0117] The training method of the fault type determination model based on the circuit breaker switch provided by the present application considers the case that the elements in the historical data matrix have time sequence in the acquisition process, and considers the case that the preset operation includes closing operation, using main trip coil tripping operation, and using auxiliary trip coil tripping operation, which ensures the richness of the elements in the historical data matrix when the circuit breaker switch fails, and improves the accuracy of determining the fault type of the circuit breaker switch.

[0118] Figure 2 Flowchart of the training method of the fault type determination model based on the circuit breaker switch provided by the present application Figure TwoThe execution subject of the method can be a server, a host or other equipment, such as Figure 2 As shown in the figure, the method can include:

[0119] S201, obtaining a to-be-trained data set; wherein the to-be-trained data set includes a historical data matrix of a circuit breaker switch when a fault occurs and a label vector corresponding to the historical data matrix, the historical data matrix includes a plurality of sub-matrices, each row in the sub-matrix represents the state of the circuit breaker switch when a preset operation is performed, the preset operation includes at least one of a closing operation, a primary trip coil tripping operation, and a secondary trip coil tripping operation, the sub-matrices in the historical data matrix are arranged vertically, and the label vector represents the actual fault type of the circuit breaker switch.

[0120] Illustratively, this step can refer to step S101 described above, and will not be described again.

[0121] S202, determining a target data matrix according to the historical data matrix; wherein the target data matrix represents the elements in the historical data matrix and the positions of the elements in the historical data matrix.

[0122] Illustratively, this step can refer to step S102 described above, and will not be described again.

[0123] S203, determining a target feature matrix corresponding to a network layer according to the target data matrix and a perspective matrix corresponding to the network layer of an initial model; wherein the initial model includes a plurality of network layers, each network layer corresponds to at least one perspective matrix, the perspective matrix represents the prediction angle of the network layer to the fault type, and the target feature matrix represents the target data matrix under the prediction angle of the network layer.

[0124] Among them, the initial model can include a plurality of network layers, and one network layer can also refer to one network module, which can be understood as one calculation module in the initial model.

[0125] For each network layer, there is at least one perspective matrix, and in one possible implementation, each network layer corresponds to the same number of perspective matrices, for example, each network layer corresponds to three perspective matrices.

[0126] In one possible implementation, the dimension of the perspective matrix is opposite to the dimension of the target data matrix, for example, the dimension of the target data matrix is , and the dimension of the perspective matrix is .

[0127] In one possible implementation, the dimensions of the perspective matrices are consistent, for example, all are .

[0128] The view matrix is used to perform feature extraction on the target data matrix from different angles, so that the model can learn the data in the target data matrix from multiple angles to improve the accuracy of determining the fault type of the circuit breaker switch. It can be understood that each view matrix is different, and at least one view matrix corresponding to each network layer is different.

[0129] In a possible implementation, the value of each element in the view matrix is in the range of [0, 1].

[0130] The target feature matrix can represent the target data matrix at the prediction angle of the network layer. In a possible implementation, the dimension of the target feature matrix is consistent with the dimension of the target data matrix.

[0131] In a possible implementation, the target data matrix can be subjected to first operation processing with the view matrix corresponding to the network layer to obtain the target feature matrix. The first operation processing can include but is not limited to directly multiplying the target data matrix by the view matrix corresponding to the network layer, or multiplying the target data matrix by the view matrix corresponding to the network layer and then processing using an activation function.

[0132] It can be understood that by determining the target feature matrix corresponding to the network layer, the output result of the network layer can be determined by inputting the target feature matrix into the corresponding network layer.

[0133] In an optional implementation, the network layer corresponds to three view matrices, and the three view matrices have the same dimension. Step S203 can include:

[0134] The target data matrix and the three view matrices corresponding to the network layer are respectively multiplied to obtain three output matrices respectively corresponding to the three view matrices. The output matrix represents the representation of the target data matrix under the view matrix. The target feature matrix corresponding to the network layer is determined according to the output matrix corresponding to each view matrix.

[0135] The output matrix represents the representation of the target data matrix under the corresponding view matrix.

[0136] It can be understood that for the three view matrices corresponding to the network layer, multiplying the target data matrix by each of the three view matrices corresponding to the network layer can obtain a total of three output matrices corresponding to the three view matrices.

[0137] Further, the second operation processing can be performed according to the output matrix corresponding to the three view matrices to obtain the target feature matrix corresponding to the network layer. The second operation processing can include, but is not limited to, directly splicing the output matrix corresponding to the three view matrices, or performing weighted summation on the output matrix corresponding to the three view matrices according to a preset weight coefficient corresponding to the three view matrices, or fusing the three output matrices through a multi-layer perception machine, and taking the fused matrix as the target feature matrix corresponding to the network layer.

[0138] In an optional implementation, the three view matrices corresponding to the network layer are respectively a first view matrix, a second view matrix, and a third view matrix; and the output matrix is represented as:

[0139] ; ; ;

[0140] wherein, the target data matrix is represented as: the first view matrix corresponding to the i-th network layer is represented as: the second view matrix corresponding to the i-th network layer is represented as: the third view matrix corresponding to the i-th network layer is represented as: the output matrix corresponding to the first view matrix of the i-th network layer is represented as: the output matrix corresponding to the second view matrix of the i-th network layer is represented as: the output matrix corresponding to the third view matrix of the i-th network layer is represented as.

[0141] The target feature matrix is represented as:

[0142] ;

[0143] wherein, the target feature matrix corresponding to the i-th network layer is represented as: h represents the column number of the view matrix corresponding to the network layer.

[0144] The three view matrices have the same dimension, for example, h can be 9.

[0145] It can be understood that by combining the output matrix corresponding to the three view matrices and using the softmax function for processing, the target feature matrix corresponding to the network layer can be obtained.

[0146] The beneficial effect of such an arrangement is that for each network layer, there are three different perspective matrices corresponding to the network layer, and the target data matrix can be extracted from multiple angles, so that the output matrix corresponding to each perspective matrix captures different patterns and information in the target data matrix, thereby improving the accuracy of the model in determining the fault type of the circuit breaker switch.

[0147] Meanwhile, using the softmax function for processing can make the elements in the target feature matrix within the range of [0, 1], and the sum of all elements in the target feature matrix is 1, which can improve the stability and convergence speed of the model.

[0148] S204, input each target feature matrix into the corresponding network layer to obtain a probability sequence output by the network layer; wherein the probability sequence includes probability values of different fault types of the circuit breaker switch corresponding to the network layer, and the number of probability values in the probability sequence is consistent with the number of fault types.

[0149] It can be understood that for each target feature matrix, inputting the target feature matrix into the network layer corresponding to the target feature matrix can obtain a probability sequence output by the network layer, and in the probability sequence, probability values of different fault types of the circuit breaker switch corresponding to the network layer can be included.

[0150] Exemplarily, the number of different fault types of the circuit breaker switch is w, and the length of the probability sequence is w, and the probability sequence can be represented as:

[0151] ;

[0152] wherein each position in the probability sequence corresponds to a fault type of the circuit breaker switch, and the probability value corresponding to each position represents the possibility of the fault type corresponding to the position output by the network layer. The sum of the probability values corresponding to all positions in the probability sequence is 1.

[0153] Exemplarily, taking w as 3, the probability sequence can be represented as (0.1, 0.4, 0.5), that is, the probability value of the first fault type corresponding to the probability sequence output by the network layer is 0.1, the probability value of the second fault type is 0.4, and the probability value of the third fault type is 0.5. It can be understood that the probability sequence output by the network layer can show that the target feature matrix is more inclined to be mapped to the third fault type under the perspective of the network layer.

[0154] It can be understood that for each network layer, a probability sequence output by the network layer can be obtained.

[0155] S205, according to the probability sequence output by each network layer and the label vector, training the initial model to obtain a trained fault type determination model.

[0156] It can be understood that the probability sequence output by each network layer and the probability sequence output by each network layer are in one-to-one correspondence, and the probability sequence output by each network layer and the label vector are in many-to-one correspondence.

[0157] By training the initial model according to the probability sequence output by each network layer and the label vector, a trained fault type determination model can be obtained.

[0158] In an optional embodiment, step S205 can include:

[0159] According to the probability sequence output by each network layer, a prediction result of the fault type is determined; and according to the prediction result and the label vector, the initial model is trained to obtain a trained fault type determination model.

[0160] It can be understood that by combining the probability sequence output by each network layer, a prediction result of the fault type can be determined, and the prediction result can represent the fault type corresponding to the target data matrix output by the initial model.

[0161] For example, there are E network layers, and according to the probability sequence output by the E network layers, a prediction result of the fault type can be determined by conversion processing.

[0162] In an optional embodiment, according to the probability sequence output by each network layer, a prediction result of the fault type can include:

[0163] For all probability sequences, the probability values at the same position in each probability sequence are determined, and the probability average value corresponding to the position is determined; wherein the position corresponds to one of the fault types, and the same position in each probability sequence represents the same fault type; and according to the probability average value corresponding to each position, a prediction result of the fault type is determined.

[0164] For example, there are E network layers, the number of different fault types is w, E is taken as 2, w is taken as 3, and all probability sequences can be represented as , The average value of the probability value of the first position in each probability sequence is 0.2, the average value of the probability value of the second position in each probability sequence is 0.25, and the average value of the probability value of the third position in each probability sequence is 0.55.

[0165] According to the probability average value corresponding to each position, a prediction result of the fault type can be determined. For example, the fault type corresponding to the maximum value in each probability average value is taken as the prediction result of the fault type.

[0166] Exemplarily, the average of the probability values of the first position in each probability sequence is 0.2, the average of the probability values of the second position in each probability sequence is 0.25, and the average of the probability values of the third position in each probability sequence is 0.55, where the average of the probability values of the third position is the largest, and the fault type corresponding to the third position is taken as the prediction result of the fault type.

[0167] The beneficial effect of such arrangement is that by taking the average of the probability sequences output by the multiple network layers, the prediction results of different network layers can be integrated, the uncertainty of the prediction of a single network layer can be reduced, and the stability and accuracy of the prediction of the model can be improved.

[0168] In an optional implementation, the initial model is trained according to the prediction result and the label vector to obtain the trained fault type determination model, which can include:

[0169] The loss value of the preset loss function is determined according to the prediction result and the label vector, the parameters of the initial model are updated according to the loss value of the loss function to obtain an updated initial model, and if the updated initial model meets a preset training completion condition, the trained fault type determination model is obtained.

[0170] Wherein, meeting the training completion condition can refer to reaching a preset number of updates, for example, 200 times.

[0171] In a possible implementation, the above steps can further include:

[0172] The test data set is obtained, wherein the test data set includes a historical data matrix of the circuit breaker switch when the circuit breaker switch fails and a label vector corresponding to the historical data matrix, the historical data matrix includes a plurality of sub-matrices, each row in the sub-matrix represents the state of the circuit breaker switch when a preset operation is performed, the preset operation includes at least one of closing operation, using a main trip coil to trip operation, and using a secondary trip coil to trip operation, the sub-matrices in the historical data matrix are arranged vertically, and the label vector represents the actual fault type of the circuit breaker switch; the test data set and the training data set do not have an intersection; and the ratio between the data in the test data set and the data in the training data set can be 7:3.

[0173] The historical data matrix in the test data set is input into the trained fault type determination model to obtain the output fault type of the circuit breaker switch.

[0174] If the difference between the output fault type of the circuit breaker switch and the label vector corresponding to the historical data matrix is less than a preset threshold, it is determined that the trained fault type determination model can be used to determine the fault type of the circuit breaker switch.

[0175] The training method of the fault type determination model based on the circuit breaker switch provided by the application considers that the initial model includes multiple network layers, each network layer corresponds to at least a view matrix, feature extraction can be performed on the target data matrix from multiple angles, the prediction results of different network layers can be comprehensively considered, the uncertainty of single network layer prediction is reduced, and the stability and accuracy of model prediction are improved, and the accuracy of determining the fault type of the circuit breaker switch is improved.

[0176] Figure 3 The flowchart of the training method of the fault type determination model based on the circuit breaker switch provided by the application Figure Three The execution subject of the method can be a server, a host or other equipment, as shown in Figure 3 The method can include:

[0177] S301, a plurality of historical data matrices and a label vector corresponding to the historical data matrix are acquired.

[0178] The historical data matrix can be represented as:

[0179] = = ;

[0180] The label vector can be represented as:

[0181] ;

[0182] S302, for each historical data matrix , a position vector matrix is superimposed on the historical data matrix to obtain a target data matrix .

[0183] Wherein, ;

[0184] The position vector matrix can be represented as:

[0185] ;

[0186] The elements in the position vector matrix can be represented as:

[0187] ;

[0188] Wherein, n represents the total number of columns of elements in the position vector matrix.

[0189] S303, according to the target data matrix , determine E target feature matrices.

[0190] Wherein, the E target feature matrices can be expressed in turn as .

[0191] For the i-th target feature matrix in the E target feature matrices, the i-th target feature matrix Can be expressed as:

[0192] ;

[0193] ; ; ;

[0194] Wherein, , , All are the perspective matrix corresponding to the i-th target feature matrix.

[0195] S304, input the E target feature matrices into the network layer corresponding to the target feature matrix respectively, to obtain the probability sequence output by the E network layers; wherein, the E network layers are the components of the initial model.

[0196] S305, for all probability sequences, determine the probability values in the same position in each probability sequence, and determine the probability average value corresponding to the position; wherein, the position corresponds to one of the fault types, and the same position in each probability sequence represents the same fault type; according to the probability average value corresponding to each position, determine the maximum value in each probability average value, and the fault type corresponding to the maximum value is the prediction result of the fault type.

[0197] S306, according to the prediction result and the label vector, train the initial model to obtain the trained fault type determination model.

[0198] The method of the present application improves the accuracy of determining the fault type of the circuit breaker switch.

[0199] Figure 4 A flowchart of a fault type determination method based on a circuit breaker switch is provided, and the execution subject of the method can be a server, a host or other equipment, as shown in Figure 4 The method can include:

[0200] S401, acquire a current data matrix of the circuit breaker switch; wherein the current data matrix comprises a plurality of sub-matrices, each row in the sub-matrix represents a state of the circuit breaker switch when performing a preset operation, the preset operation comprises at least one of a closing operation, a primary trip coil tripping operation, and a secondary trip coil tripping operation, and the sub-matrices in the current data matrix are arranged vertically.

[0201] It can be understood that the current data matrix of the circuit breaker switch has the same form as the historical data matrix of the circuit breaker switch. For example, the current data matrix of the circuit breaker switch can be obtained by a real-time detection device preset at the circuit breaker switch.

[0202] S402, input the current data matrix into a fault type determination model to obtain an output fault type of the circuit breaker switch; wherein the fault type determination model is any fault type determination model of any embodiment of the present application.

[0203] The determination method of the fault type of the circuit breaker switch provided by the present application can realize inputting the current data matrix into the fault type determination model, so as to obtain the fault type of the circuit breaker switch output by the fault type determination model, without manually determining the fault type of the current circuit breaker switch. The efficiency of determining the fault type of the circuit breaker switch is ensured, and the accuracy of determining the fault type of the circuit breaker switch is also ensured.

[0204] Figure 5 The structure of the training device of the fault type determination model of the circuit breaker switch provided by the present application is shown in the figure Figure One As shown in the figure Figure 5 The training device 50 of the fault type determination model of the circuit breaker switch comprises a first acquisition unit 501, a determination unit 502, and a training unit 503.

[0205] The first acquisition unit 501 is configured to acquire a training data set; wherein the training data set comprises a historical data matrix of the circuit breaker switch when the circuit breaker switch fails and a label vector corresponding to the historical data matrix, the historical data matrix comprises a plurality of sub-matrices, each row in the sub-matrix represents a state of the circuit breaker switch when performing a preset operation, the preset operation comprises at least one of a closing operation, a primary trip coil tripping operation, and a secondary trip coil tripping operation, the sub-matrices in the historical data matrix are arranged vertically, and the label vector represents an actual fault type of the circuit breaker switch.

[0206] The determination unit 502 is configured to determine a target data matrix according to the historical data matrix; wherein the target data matrix represents elements in the historical data matrix and positions of the elements in the historical data matrix.

[0207] The training unit 503 is configured to train the initial model according to the target data matrix and the corresponding label vector, to obtain the trained fault type determination model; and the fault type determination model is configured to predict the fault type of the circuit breaker switch.

[0208] Figure 6 A structure of a training device for a circuit breaker switch-based fault type determination model is provided in the present application Figure Two As shown in Figure 6 The training device 60 for the circuit breaker switch-based fault type determination model includes a first acquisition unit 601, a determination unit 602, and a training unit 603, wherein the training unit 603 further includes a first processing module 6031, a second processing module 6032, and a third processing module 6033.

[0209] In an optional example, the initial model includes a plurality of network layers, and each network layer corresponds to at least one perspective matrix, and the perspective matrix represents a prediction angle of the network layer to the fault type.

[0210] The first processing module 6031 is configured to determine a target feature matrix corresponding to the network layer according to the target data matrix and the perspective matrix corresponding to the network layer; and the target feature matrix represents the target data matrix at the prediction angle of the network layer.

[0211] The second processing module 6032 is configured to input each target feature matrix into the corresponding network layer to obtain a probability sequence output by the network layer; and the probability sequence includes probability values of different fault types of the circuit breaker switch corresponding to the network layer, and the number of probability values in the probability sequence is consistent with the number of fault types.

[0212] The third processing module 6033 is configured to train the initial model according to the probability sequence output by each network layer and the label vector, to obtain the trained fault type determination model.

[0213] In an optional example, the network layer corresponds to three perspective matrices, and the three perspective matrices have the same dimension; and the first processing module 6031 is further configured to multiply the target data matrix and the three perspective matrices corresponding to the network layer respectively to obtain three output matrices corresponding to the three perspective matrices respectively; and the output matrix represents a form of the target data matrix under the perspective matrix; and the target feature matrix corresponding to the network layer is determined according to the output matrix corresponding to each perspective matrix.

[0214] In an optional example, the three perspective matrices corresponding to the network layer are a first perspective matrix, a second perspective matrix, and a third perspective matrix respectively; and the output matrix is represented as:

[0215] ; ; ;

[0216] wherein, characterizing the target data matrix, characterizing the first view matrix corresponding to the i-th network layer, characterizing the second view matrix corresponding to the i-th network layer, characterizing the third view matrix corresponding to the i-th network layer, characterizing the output matrix corresponding to the first view matrix corresponding to the i-th network layer, characterizing the output matrix corresponding to the second view matrix corresponding to the i-th network layer, characterizing the output matrix corresponding to the third view matrix corresponding to the i-th network layer;

[0217] The target feature matrix is characterized as:

[0218] ;

[0219] wherein, characterizing the target feature matrix corresponding to the i-th network layer, and h characterizing the column number of the view matrix corresponding to the network layer.

[0220] In an optional example, the third processing module 6033 is further configured to determine a prediction result of the fault type according to the probability sequences output by the network layers; and train the initial model according to the prediction result and the label vector to obtain a trained fault type determination model.

[0221] In an optional example, the third processing module 6033 is further configured to, for all the probability sequences, determine the probability values at the same position in each probability sequence, and determine the probability average value corresponding to the position; wherein the position corresponds to one of the fault types, and the same position in each probability sequence characterizes the same fault type; and determine the prediction result of the fault type according to the probability average value corresponding to each position.

[0222] Figure 7 A structure diagram of a fault type determination device based on a circuit breaker switch provided in the present application is shown in FIG. 7. Figure 7 As shown in FIG. 7, the fault type determination device 70 based on the circuit breaker switch includes a second acquisition unit 701 and an output unit 702.

[0223] The second acquisition unit 701 is configured to acquire a current data matrix of the circuit breaker switch; wherein the current data matrix includes a plurality of sub-matrices, each row in the sub-matrix characterizes a state of the circuit breaker switch when performing a preset operation, the preset operation includes at least one of a closing operation, a primary coil tripping operation, and a secondary coil tripping operation, and the sub-matrices in the current data matrix are arranged vertically;

[0224] The output unit 702 is configured to input the current data matrix into the fault type determination model to obtain an output fault type of the circuit breaker switch; and the fault type determination model is the fault type determination model of any of the embodiments of the present application.

[0225] Figure 8 A structural schematic diagram of an electronic device provided by the present application is shown in FIG. 8. As shown in FIG. 8, the electronic device 80 provided by the present embodiment includes at least one processor 801 and a memory 802. Optionally, the device 80 further includes a communication component 803. The processor 801, the memory 802 and the communication component 803 are connected through a bus 804. Figure 8

[0226] In the implementation process, the at least one processor 801 executes the computer-executed instructions stored in the memory 802, so that the at least one processor 801 executes the method described above.

[0227] The specific implementation process of the processor 801 can refer to the method embodiments described above, which have similar implementation principles and technical effects, and will not be described here again in the present embodiment.

[0228] In the above embodiments, it should be understood that the processor can be a central processing unit (CPU), and can also be other general-purpose processors, digital signal processors (DSP), application specific integrated circuits (ASIC) and the like. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor and the like. The steps of the method disclosed in the present application can be directly embodied as the execution of the hardware processor, or the execution of the combination of the hardware and software modules in the processor.

[0229] The memory can include a random access memory (RAM), and can also include a non-volatile memory (NVM), for example, at least one disk memory.

[0230] ​The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus, etc. The bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, the bus in the drawings of the present application does not limit to only one bus or one type of bus.

[0231] The present application also provides a computer program product, comprising a computer program, which, when executed by a processor, implements the method described above.

[0232] The present application also provides a computer-readable storage medium, which stores computer-executable instructions, and when a processor executes the computer-executable instructions, the method described above is implemented.

[0233] It should be noted that, for the foregoing method embodiments, in order to simply describe, they are all expressed as a series of action combinations, but those skilled in the art should know that the present application is not limited to the order of the actions described, because according to the present application, certain steps can be performed in other order or simultaneously. Secondly, those skilled in the art should know that the embodiments described in the specification all belong to optional embodiments, and the actions and modules involved are not necessarily required by the present application.

[0234] Further, it should be noted that, although each step in the flowchart is displayed in sequence according to the arrow indication, these steps are not necessarily executed in sequence according to the arrow indication. Unless explicitly stated in this article, the execution of these steps has no strict order limitation, and these steps can be executed in other order. Moreover, at least part of the steps in the flowchart can include multiple sub-steps or multiple stages, which are not necessarily executed at the same time, but can be executed at different times, and the execution order of these sub-steps or stages is not necessarily sequential, but can be executed with at least part of other steps or sub-steps or stages of other steps in rotation or alternation.

[0235] It should be understood that the above-mentioned device embodiments are only schematic, and the device of the present application can also be realized by other ways. For example, the division of units / modules in the above-mentioned embodiments is only a logical functional division, and actual implementation can have another division way. For example, multiple units, modules or components can be combined, or can be integrated into another system, or some features can be ignored or not executed.

[0236] In addition, each functional unit / module in the embodiments of the present application can be integrated in one unit / module, or each unit / module can exist physically, or two or more units / modules can be integrated together. The integrated unit / module can be realized in the form of hardware or in the form of a software program module.

[0237] If the integrated unit / module is realized in the form of hardware, the hardware can be a digital circuit, an analog circuit, etc. The physical implementation of the hardware structure includes but is not limited to transistors, memristors, etc. Unless otherwise specified, the processor can be any appropriate hardware processor, such as a CPU, a GPU, an FPGA, a DSP, an ASIC, etc. Unless otherwise specified, the storage unit can be any appropriate magnetic storage medium or magneto-optical storage medium, such as a resistive random access memory (RRAM), a dynamic random access memory (DRAM), a static random access memory (SRAM), an enhanced dynamic random access memory (EDRAM), a high-bandwidth memory (HBM), a hybrid memory cube (HMC), etc.

[0238] If the integrated unit / module is realized in the form of a software program module and sold or used as an independent product, it can be stored in a computer readable storage medium. Based on this understanding, the technical solutions of the present application essentially or the part that contributes to the prior art or the whole or part of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a storage medium, including a number of instructions to make a computer device (which can be a personal computer, a server, or a network device, etc.) execute all or part of the steps of the embodiments of the method of the present application. The aforementioned storage medium includes: a U disk, a read-only memory (ROM), a random access memory (RAM), a mobile hard disk, a magnetic disk or an optical disk, and various program codes that can be stored in the medium.

[0239] In the above embodiments, the description of each of the embodiments focuses on different aspects of the embodiments. The parts not described in detail in a certain embodiment can be seen in the relevant description of the other embodiments. The technical features of the above embodiments can be combined in any manner. In order to make the description simple, not all possible combinations of the technical features in the above embodiments are described, however, as long as the combinations of the technical features do not exist contradictory, they should be considered as falling within the scope of the disclosure.

[0240] Other embodiments of the application will be apparent to those skilled in the art from consideration of the specification and practice of the application disclosed herein. It is intended that the specification and examples be considered as exemplary only, with the true scope and spirit of the application being indicated by the following claims.

[0241] It should be understood that the application is not limited to the precise construction and combinations of the components and steps described above and shown in the accompanying drawings. Various modifications and changes can be made in the arrangement of the parts and steps without departing from the scope of the application. The scope of the application is indicated by the appended claims, rather than the description and figures.

Claims

1. A training method for a fault type determination model based on circuit breaker switches, characterized in that, include: Obtain the training dataset; wherein, the training dataset includes a historical data matrix of circuit breaker switches when a fault occurs and a label vector corresponding to the historical data matrix. The historical data matrix includes multiple sub-matrices, each row of which represents the state of the circuit breaker switch when a preset operation is performed. The preset operation includes at least one of closing operation, opening operation using the main opening coil, and opening operation using the auxiliary opening coil. The sub-matrices in the historical data matrix are arranged vertically, and the label vector represents the actual fault type of the circuit breaker switch. Based on the historical data matrix, a target data matrix is ​​determined; wherein, the target data matrix represents the elements in the historical data matrix and the positions of the elements in the historical data matrix; Based on the target data matrix and the corresponding label vector, the initial model is trained to obtain a trained fault type determination model; wherein, the fault type determination model is used to predict the fault type of the circuit breaker switch.

2. The method according to claim 1, characterized in that, The initial model includes multiple network layers, each network layer corresponding to at least one view matrix, the view matrix representing the network layer’s prediction angle for fault types; Based on the target data matrix and the corresponding label vector, the initial model is trained to obtain a trained fault type determination model, including: Based on the target data matrix and the viewpoint matrix corresponding to the network layer, the target feature matrix corresponding to the network layer is determined; wherein, the target feature matrix represents the target data matrix under the prediction angle of the network layer; Each of the target feature matrices is input into the corresponding network layer to obtain the probability sequence output by the network layer; wherein, the probability sequence includes the probability values ​​of different fault types of the circuit breaker switch corresponding to the network layer, and the number of probability values ​​in the probability sequence is consistent with the number of fault types; The initial model is trained based on the probability sequence output by each network layer and the label vector to obtain a trained fault type determination model.

3. The method according to claim 2, characterized in that, The network layer corresponds to three viewpoint matrices, and the three viewpoint matrices have the same dimension; based on the target data matrix and the viewpoint matrices corresponding to the network layer, the target feature matrix corresponding to the network layer is determined, including: The target data matrix and the three view matrices corresponding to the network layer are multiplied respectively to obtain the output matrices corresponding to the three view matrices; wherein, the output matrix represents the representation of the target data matrix under the view matrix; The target feature matrix corresponding to the network layer is determined based on the output matrix corresponding to each viewpoint matrix.

4. The method according to claim 3, characterized in that, The three viewpoint matrices corresponding to the network layer are the first viewpoint matrix, the second viewpoint matrix, and the third viewpoint matrix, respectively; the output matrix is ​​represented as: ； ； ； Among them, the Characterizing the target data matrix, the The first viewpoint matrix representing the i-th network layer, wherein The second-view matrix representing the i-th network layer, the The third-view matrix representing the i-th network layer, the The output matrix representing the first-view matrix of the i-th network layer, wherein The output matrix representing the second-view matrix corresponding to the i-th network layer, wherein The output matrix representing the third-view matrix corresponding to the i-th network layer; The target feature matrix is ​​characterized as follows: ； Among them, the The target feature matrix corresponding to the i-th network layer is represented by h, where h represents the number of columns in the view matrix corresponding to the network layer.

5. The method according to claim 2, characterized in that, Based on the probability sequences output by each network layer and the label vector, the initial model is trained to obtain a trained fault type determination model, including: Based on the probability sequence output by each network layer, the predicted fault type is determined. Based on the prediction results and the label vector, the initial model is trained to obtain a trained fault type determination model.

6. The method according to claim 5, characterized in that, Based on the probability sequences output by each network layer, the predicted fault type is determined, including: For all probability sequences, determine the probability value at the same position in each probability sequence, and determine the average probability value corresponding to that position; wherein, the position corresponds to one type of fault, and the same position in each probability sequence represents the same type of fault; The predicted fault type is determined based on the average probability of each location.

7. The method according to claim 1, characterized in that, Based on the historical data matrix, the target data matrix is ​​determined, including: Based on the dimension of the historical data matrix, a position vector matrix is ​​determined; wherein the dimension of the position vector matrix is ​​the same as the dimension of the historical data matrix, and the elements in the position vector matrix represent the chronological order of the positions of each element in the historical data matrix. Based on the historical data matrix and the location vector matrix, the target data matrix corresponding to the historical data matrix is ​​determined.

8. The method according to claim 7, characterized in that, The elements in the position vector matrix are represented as follows: ； Among them, the The value of the element in the i-th row and j-th column of the position vector matrix is ​​represented by n, where n represents the total number of columns in the position vector matrix.

9. A method for determining fault types based on circuit breaker switches, characterized in that, include: Obtain the current data matrix of the circuit breaker switch; wherein, the current data matrix includes multiple sub-matrices, each row of the sub-matrices represents the state of the circuit breaker switch when performing a preset operation, the preset operation includes at least one of closing operation, opening operation using the main opening coil, and opening operation using the auxiliary opening coil, and the sub-matrices in the current data matrix are arranged vertically; The current data matrix is ​​input into the fault type determination model to obtain the fault type of the circuit breaker switch; wherein the fault type determination model is the fault type determination model according to any one of claims 1-8.

10. A training device for a fault type determination model based on circuit breaker switches, characterized in that, include: The first acquisition unit is used to acquire a training dataset; wherein, the training dataset includes a historical data matrix of circuit breaker switches when a fault occurs and a label vector corresponding to the historical data matrix. The historical data matrix includes multiple sub-matrices, each row of which represents the state of the circuit breaker switch when a preset operation is performed. The preset operation includes at least one of closing operation, opening operation using the main opening coil, and opening operation using the auxiliary opening coil. The sub-matrices in the historical data matrix are arranged vertically, and the label vector represents the actual fault type of the circuit breaker switch. A determining unit is configured to determine a target data matrix based on the historical data matrix; wherein the target data matrix represents the elements in the historical data matrix and the positions of the elements in the historical data matrix; The training unit is used to train the initial model based on the target data matrix and the corresponding label vector to obtain a trained fault type determination model; wherein, the fault type determination model is used to predict the fault type of the circuit breaker switch.

11. A device for determining fault types based on circuit breaker switches, characterized in that, include: The second acquisition unit is used to acquire the current data matrix of the circuit breaker switch; wherein, the current data matrix includes multiple sub-matrices, each row of the sub-matrices represents the state of the circuit breaker switch when performing a preset operation, the preset operation includes at least one of closing operation, opening operation using the main opening coil, and opening operation using the auxiliary opening coil, and the sub-matrices in the current data matrix are arranged vertically; An output unit is used to input the current data matrix into a fault type determination model to obtain the output fault type of the circuit breaker switch; wherein the fault type determination model is the fault type determination model according to any one of claims 1-8.

12. An electronic device, characterized in that, include: Memory, processor; The memory stores computer-executed instructions; The processor executes computer execution instructions stored in the memory, causing the processor to perform the method as described in any one of claims 1-8 and / or claim 9.

13. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer-executable instructions, which, when executed by a processor, are used to implement the method as described in any one of claims 1-8 and / or claim 9.

14. A computer program product, characterized in that, Includes a computer program that, when executed by a processor, implements the method described in any one of claims 1-8 and / or claim 9.

Images