Power grid fault identification method, model training method, device and equipment

Abstract

The embodiment of the invention provides a power grid fault identification method, a model training method, devices and equipment. The method comprises the following steps: acquiring real-time fault data of a target power grid; performing hierarchical decomposition processing on the real-time fault data to obtain at least one piece of fault feature data corresponding to the real-time fault data; and based on a triple weighted extreme learning machine, processing the fault feature data to obtain a fault identification result of the target power grid. The method can improve the accuracy and efficiency of power grid fault identification.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence technology, and particularly to a power grid fault identification method, a model training method, a device, and a device. Background Art

[0002] There are a variety of possible fault types in the power grid, such as line short circuits, grounding faults, equipment failures, etc. Each fault has its unique characteristics and manifestations, making fault identification complex. In this context, accurately identifying the operating conditions of the power grid is particularly important.

[0003] In the prior art, according to the existing training data set, a classification model is constructed by using machine learning methods to achieve the identification of power grid faults.

[0004] However, in the above method, the classification model cannot extract effective feature data, thereby reducing the accuracy and efficiency of power grid fault identification. Summary of the Invention

[0005] The embodiments of this application provide a power grid fault identification method, a model training method, a device, and a device, which can improve the accuracy and efficiency of power grid fault identification.

[0006] In a first aspect, the embodiments of this application provide a power grid fault identification method, including:

[0007] Obtain the real-time fault data of the target power grid; wherein, the real-time fault data represents the data generated when the current target power grid has an operating fault;

[0008] Perform hierarchical decomposition processing on the real-time fault data to obtain at least one fault feature data corresponding to the real-time fault data; wherein, the fault feature data represents the operating fault characteristics of the target power grid;

[0009] Based on a triple weighted extreme learning machine, process each of the fault feature data to obtain a fault identification result of the target power grid; wherein, the fault identification result includes the operating fault type of the target power grid.

[0010] In a possible implementation manner, the triple weighted extreme learning machine includes an output layer and a hidden layer; based on the triple weighted extreme learning machine, processing each of the fault feature data to obtain a fault identification result of the target power grid includes:

[0011] Based on the hidden layer, perform triple weighted processing on the fault feature data to obtain an output result corresponding to the fault feature data;

[0012] Based on the output layer, process each of the output results to obtain a fault identification result of the target power grid.

[0013] In a possible implementation, based on the hidden layer, triple-weight processing is performed on the fault feature data to obtain an output result corresponding to the fault feature data, including:

[0014] According to the triple-weight information in the hidden layer, triple-weight processing is performed on the fault feature data to obtain the output result; wherein, the triple-weight information includes a regularization weight, a sample weight, and a variable weight; the regularization weight represents the degree of regularization; the sample weight represents the importance of the data; the variable weight represents the contribution degree of the variable.

[0015] In a possible implementation, according to the triple-weight information in the hidden layer, triple-weight processing is performed on the fault feature data to obtain the output result, including:

[0016] According to the triple-weight information and hidden layer parameters in the hidden layer, weighted processing is performed on the fault feature data to obtain the output result; wherein, the hidden layer parameters include input parameters and bias parameters.

[0017] In a possible implementation, based on the output layer, processing is performed on each of the output results to obtain a fault identification result of the target power grid, including:

[0018] Based on the output layer, according to the output layer weight, processing is performed on each of the output results to obtain the fault identification result.

[0019] In a possible implementation, the method further includes:

[0020] Based on an autoencoder network, dimensionality reduction processing is performed on the fault feature data to obtain processed fault feature data.

[0021] In a second aspect, an embodiment of the present application provides a model training method applied to power grid fault identification, including:

[0022] Obtain training fault data to be trained; wherein, the training fault data to be trained represents data generated when the power grid has an operation fault during a historical time period;

[0023] Perform hierarchical decomposition processing on the training fault data to be trained to obtain at least one fault feature data corresponding to the training fault data to be trained; wherein, the fault feature data represents the operation fault features of the power grid.

[0024] Call the initial extreme learning machine; and perform training processing on the initial extreme learning machine according to each piece of the fault feature data to obtain a triple-weighted extreme learning machine; wherein, the triple-weighted extreme learning machine is used to process the real-time fault data as described in the first aspect to obtain a fault identification result; the fault identification result includes the type of operation fault.

[0025] In a possible implementation manner, the initial extreme learning machine includes an initial output layer and an initial hidden layer; performing training processing on the initial extreme learning machine according to each piece of the fault feature data to obtain the triple-weighted extreme learning machine includes:

[0026] Based on the initial hidden layer, process the fault feature data to obtain an output result corresponding to the fault feature data;

[0027] Based on the initial output layer, process each of the output results to obtain a predicted fault type;

[0028] According to the predicted fault type, perform training processing on the initial hidden layer to obtain a hidden layer;

[0029] According to the predicted fault type, perform training processing on the initial output layer to obtain an output layer; wherein, the output layer and the hidden layer constitute the triple-weighted extreme learning machine.

[0030] In a possible implementation manner, the predicted fault type has an actual fault type; performing training processing on the initial hidden layer according to the predicted fault type to obtain a hidden layer includes:

[0031] According to the predicted fault type and the actual fault type, perform optimization processing on the triple-weighted information of the initial hidden layer to obtain the hidden layer; wherein, the triple-weighted information includes a regularization weight, a sample weight, and a variable weight; the regularization weight represents the degree of regularization; the sample weight represents the importance of data; the variable weight represents the contribution degree of variables.

[0032] In a possible implementation manner, the method further includes:

[0033] Based on an autoencoder network, perform dimensionality reduction processing on each piece of the fault feature data to obtain at least one processed fault feature data.

[0034] In a third aspect, an embodiment of the present application provides a power grid fault identification device, including:

[0035] An acquisition module, configured to acquire real-time fault data of a target power grid; wherein, the real-time fault data represents data generated when the current target power grid has an operation fault;

[0036] A processing module, configured to hierarchically decompose and process the real-time fault data to obtain at least one fault feature data corresponding to the real-time fault data; wherein, the fault feature data characterizes the operation fault features of the target power grid;

[0037] An identification module, configured to process each of the fault feature data based on a triple-weighted extreme learning machine to obtain a fault identification result of the target power grid; wherein, the fault identification result includes the operation fault types of the target power grid.

[0038] In a possible implementation manner, the triple-weighted extreme learning machine includes an output layer and a hidden layer; the identification module is specifically configured to: based on the hidden layer, perform triple-weighted processing on the fault feature data to obtain an output result corresponding to the fault feature data; based on the output layer, process each of the output results to obtain a fault identification result of the target power grid.

[0039] In a possible implementation manner, the identification module is specifically configured to: according to the triple-weighted information in the hidden layer, perform triple-weighted processing on the fault feature data to obtain the output result; wherein, the triple-weighted information includes a regularization weight, a sample weight, and a variable weight; the regularization weight characterizes the degree of regularization; the sample weight characterizes the importance of data; the variable weight characterizes the contribution degree of variables.

[0040] In a possible implementation manner, the identification module is specifically configured to: according to the triple-weighted information and hidden layer parameters in the hidden layer, perform weighted processing on the fault feature data to obtain the output result; wherein, the hidden layer parameters include input parameters and bias parameters.

[0041] In a possible implementation manner, the identification module is further specifically configured to: based on the output layer, process each of the output results according to the output layer weights to obtain the fault identification result.

[0042] In a possible implementation manner, the apparatus is further configured to: based on an autoencoder network, perform dimensionality reduction processing on the fault feature data to obtain processed fault feature data.

[0043] In a fourth aspect, an embodiment of the present application provides a model training apparatus for power grid fault identification, including:

[0044] An acquisition module, configured to acquire training fault data to be trained; wherein, the training fault data to be trained characterizes the data generated by the power grid when an operation fault occurs during a historical time period;

[0045] A processing module, configured to perform hierarchical decomposition processing on the to-be-trained fault data to obtain at least one fault feature data corresponding to the to-be-trained fault data; wherein, the fault feature data characterizes the operation fault features of the power grid.

[0046] A training module, configured to call an initial extreme learning machine; and perform training processing on the initial extreme learning machine according to each piece of the fault feature data to obtain a triple-weighted extreme learning machine; wherein, the triple-weighted extreme learning machine is configured to process the real-time fault data as described in the third aspect to obtain a fault recognition result; the fault recognition result includes an operation fault type.

[0047] In a possible implementation manner, the initial extreme learning machine includes an initial output layer and an initial hidden layer; the training module is specifically configured to: process the fault feature data based on the initial hidden layer to obtain an output result corresponding to the fault feature data; process each of the output results based on the initial output layer to obtain a predicted fault type; perform training processing on the initial hidden layer according to the predicted fault type to obtain a hidden layer; perform training processing on the initial output layer according to the predicted fault type to obtain an output layer; wherein, the output layer and the hidden layer constitute the triple-weighted extreme learning machine.

[0048] In a possible implementation manner, the predicted fault type has an actual fault type; the training module is specifically configured to: optimize the triple-weight information of the initial hidden layer according to the predicted fault type and the actual fault type to obtain the hidden layer; wherein, the triple-weight information includes a regularization weight, a sample weight, and a variable weight; the regularization weight characterizes the regularization degree; the sample weight characterizes the data importance; the variable weight characterizes the variable contribution degree.

[0049] In a possible implementation manner, the device is further configured to: perform dimensionality reduction processing on each piece of the fault feature data based on an autoencoder network to obtain at least one processed fault feature data.

[0050] In a fifth aspect, an embodiment of the present application provides an electronic device, including: a memory, a processor;

[0051] The memory stores computer execution instructions;

[0052] The processor executes the computer execution instructions stored in the memory, so that the processor executes the method as described in the first aspect and / or the second aspect above.

[0053] Sixth aspect, an embodiment of the present application provides a computer-readable storage medium, in which computer-executable instructions are stored, and when the computer-executable instructions are executed by a processor, they are used to implement the method of the first aspect and / or the second aspect above.

[0054] Seventh aspect, an embodiment of the present application provides a computer program product, including a computer program, and when the computer program is executed by a processor, it implements the method of the first aspect and / or the second aspect above.

[0055] The power grid fault identification method, model training method, device, and equipment provided by the embodiments of the present application perform hierarchical feature extraction and decomposition on the real-time fault data of the target power grid through the analytic hierarchy process, and input the obtained multi-level fault feature data into a triple weighted extreme learning machine for processing to identify the operating fault type of the target power grid; furthermore, the accuracy and efficiency of power grid fault identification can be improved. Description of the Drawings

[0056] The drawings here are incorporated into the description and constitute a part of this description, showing embodiments consistent with the present application, and are used together with the description to explain the principles of the present application.

[0057] Figure 1 It is a schematic diagram of an application scenario provided by the present application;

[0058] Figure 2 It is a schematic flow chart of a power grid fault identification method provided by an embodiment of the present application;

[0059] Figure 3 It is a schematic flow chart of another power grid fault identification method provided by an embodiment of the present application;

[0060] Figure 4 It is a schematic structural diagram of a triple weighted extreme learning machine provided by an embodiment of the present application;

[0061] Figure 5 It is a schematic flow chart of a model training method applied to power grid fault identification provided by an embodiment of the present application;

[0062] Figure 6 It is a schematic flow chart of another model training method applied to power grid fault identification provided by an embodiment of the present application;

[0063] Figure 7 It is a schematic diagram of a power grid fault identification process provided by an embodiment of the present application;

[0064] Figure 8 It is a schematic diagram of the classification test effect of a triple weighted extreme learning machine provided by an embodiment of the present application;

[0065] Figure 9 The structural schematic diagram of a power grid fault identification device provided by an embodiment of the present application;

[0066] Figure 10 The structural schematic diagram of a model training device for power grid fault identification provided by an embodiment of the present application;

[0067] Figure 11 The structural schematic diagram of an electronic device provided by an embodiment of the present application.

[0068] Through the above-mentioned drawings, the specific embodiments of the present application have been shown, and there will be more detailed descriptions hereinafter. These drawings and text descriptions are not intended to limit the scope of the concept of the present application in any way, but to illustrate the concept of the present application to those skilled in the art by referring to specific embodiments. Specific Embodiments

[0069] Here, the exemplary embodiments will be described in detail, and the examples are shown in the drawings. When the following description refers to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present application. On the contrary, they are merely examples of devices and methods consistent with some aspects of the present application as detailed in the appended claims.

[0070] First, the nouns involved in the present application are explained:

[0071] Analytic Hierarchy Process (AHP for short): A systematic and hierarchical analysis method that combines qualitative and quantitative methods;

[0072] Autoencoder (AE for short): A type of artificial neural network (ANNs for short) used in semi-supervised learning and unsupervised learning, and its function is to perform representation learning on the input information by taking the input information as the learning target;

[0073] Extreme Learning Machine (ELM for short): A type of machine learning system or method based on a feedforward neural network (FNN for short), applicable to supervised learning and unsupervised learning problems.

[0074] It should be noted that the user information involved in this application (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data for analysis, stored data, displayed data, etc.) are all information and data that have been authorized by the user or fully authorized by all parties. Moreover, for the processing of relevant data such as collection, use, processing, transmission, provision, disclosure, and application, all comply with relevant laws, regulations, and standards, adopt necessary confidentiality measures, do not violate public order and good customs, and provide corresponding operation entrances for users to choose to authorize or refuse.

[0075] Furthermore, for the technical solution of this application that involves big data analysis of user information (including but not limited to personal biometric characteristics, identity data, consumption data, asset data, electronic terminal operation data, etc.), and uses artificial intelligence technology for automated decision-making, and makes decisions that have a significant impact on personal rights and interests based on the results of automated decision-making, a corresponding operation entrance is provided for users to choose to agree or refuse the results of automated decision-making; if the user chooses to refuse, then it enters the expert decision-making process.

[0076] It should be noted that this application can be used in the field of artificial intelligence technology, and can also be used in any field other than artificial intelligence technology. The application field of this application is not limited.

[0077] Figure 1 As a schematic diagram of an application scenario provided for this application, as Figure 1 shown, the specific application scenario of this application includes: based on the electronic device 101, receiving the training data set input by the tester 102, and using the machine learning method to construct a classification model, so as to realize the identification of power grid faults.

[0078] Combined with the above scenario, it can be seen that the classification model cannot extract effective feature data, thereby reducing the accuracy and efficiency of power grid fault identification.

[0079] The power grid fault identification method provided by this application performs hierarchical feature extraction and decomposition on the real-time fault data of the target power grid through the analytic hierarchy process, and inputs the obtained multi-level fault feature data into a triple weighted extreme learning machine for processing to identify the operating fault type of the target power grid, solving the technical problems of low accuracy and efficiency in power grid fault identification.

[0080] The following uses specific embodiments to elaborate in detail on the technical solution of this application and how the technical solution of this application solves the above technical problems. These several specific embodiments can be combined with each other, and the same or similar concepts or processes may not be repeated in some embodiments. The following will describe the embodiments of this application in conjunction with the accompanying drawings.

[0081] Figure 2The flowchart of a power grid fault identification method provided by an embodiment of this application is shown as Figure 2 follows. The method includes:

[0082] 201. Obtain real-time fault data of the target power grid; wherein, the real-time fault data represents the data generated when the current target power grid has an operating fault.

[0083] Exemplarily, the execution subject of this embodiment can be an electronic device, hereinafter referred to as the device. The device can obtain real-time fault data from the currently monitored target power grid based on a start instruction input by the user or based on an automatically enabled instruction. The real-time fault data represents the data generated when the current target power grid has an operating fault, such as: the date and timestamp of the fault occurrence, the current and voltage measurement values at the time of the fault occurrence, environmental factors such as weather conditions, temperature, and humidity at the time of the fault occurrence, the load conditions at the time of the fault occurrence including load type and load level, communication records related to the fault, such as the logs of the remote monitoring and control system, and if there are monitoring devices, it may include image or video records at the time of the fault occurrence.

[0084] 202. Perform hierarchical decomposition processing on the real-time fault data to obtain at least one fault feature data corresponding to the real-time fault data; wherein, the fault feature data represents the operating fault characteristics of the target power grid.

[0085] Exemplarily, the device performs hierarchical decomposition processing on the obtained real-time fault data based on a preset analytic hierarchy process. For example, use AHP to hierarchically decompose the power grid fault types and their characteristics, transform the complex fault identification problem into multiple relatively simple sub-problems, and obtain the fault characteristics of each level; by constructing a hierarchical structure model, perform qualitative and quantitative analysis on the fault characteristics of each level, thereby determining the importance weights of each fault characteristic in the overall identification, and then perform sorting processing from large to small according to the importance weights, and extract a preset number of top-ranked fault characteristics as the multiple fault feature data corresponding to the real-time fault data, that is, the operating fault characteristics of multiple target power grids. AHP can help determine the importance of each fault characteristic, optimize feature selection, and thus further improve the identification performance.

[0086] 203. Based on a triple weighted extreme learning machine, process each fault feature data to obtain a fault identification result of the target power grid; wherein, the fault identification result includes the operating fault type of the target power grid.

[0087] Exemplarily, the device invokes the trained triple-weighted extreme learning machine and inputs all the obtained fault feature data into the triple-weighted extreme learning machine. Based on the triple-weighted extreme learning machine, triple-weighted processing is performed on all the fault feature data. For example, first, determine the first preset weighting coefficient, the second preset weighting coefficient, and the third preset weighting coefficient corresponding to each fault feature data. Based on the first preset weighting coefficient, perform first-weighted processing on all the fault feature data. Based on the second preset weighting coefficient, perform second-weighted processing on the first-weighted result. Based on the third preset weighting coefficient, perform third-weighted processing on the second-weighted result to obtain the fault identification result of the target power grid; wherein, the fault identification result includes the operating fault type of the target power grid.

[0088] In this embodiment, a power grid fault identification method is provided. Hierarchical feature extraction and decomposition are performed on the real-time fault data of the target power grid through the analytic hierarchy process, and the obtained multi-level fault feature data are input into a triple-weighted extreme learning machine for processing to identify the operating fault type of the target power grid; thereby, the accuracy and efficiency of power grid fault identification can be improved.

[0089] Figure 3 As shown in the flowchart of another power grid fault identification method provided by the embodiment of the present application, Figure 3 as shown, the method includes:

[0090] 301. Obtain the real-time fault data of the target power grid; wherein, the real-time fault data represents the data generated when the current target power grid has an operating fault.

[0091] Exemplarily, this step can refer to step 201, which will not be elaborated here.

[0092] 302. Perform hierarchical decomposition processing on the real-time fault data to obtain at least one fault feature data corresponding to the real-time fault data; wherein, the fault feature data represents the operating fault feature of the target power grid.

[0093] Exemplarily, this step can refer to step 202, which will not be elaborated here.

[0094] In one example, it further includes: based on the autoencoder network, perform dimensionality reduction processing on the fault feature data to obtain the processed fault feature data.

[0095] Exemplarily, the device calls a preset autoencoder network and, based on this autoencoder network, performs dimensionality reduction processing on all the fault feature data. That is, an autoencoder network is introduced to further learn and optimize each fault feature processed by AHP. The autoencoder network automatically extracts the effective information in the fault features through unsupervised learning and constructs a low-dimensional representation that can accurately reflect each fault characteristic, thereby obtaining each processed fault feature data. By introducing an autoencoder for feature extraction, the robustness of the ELM network can be further enhanced, and the accuracy and efficiency of power grid fault recognition can be improved.

[0096] 303. Perform triple-weighting processing on the fault feature data according to the triple-weighting information in the hidden layer to obtain an output result.

[0097] Among them, the triple-weighted extreme learning machine includes an output layer and a hidden layer.

[0098] Exemplarily, Figure 4 is a schematic structural diagram of a triple-weighted extreme learning machine provided by an embodiment of the present application. As Figure 4 shown, the structure of the triple-weighted extreme learning machine ELM includes an input layer, a hidden layer, and an output layer. Through the hidden layer, triple-weighting processing is performed on each fault feature data (X1,..., X t ,..., X n ) or the dimensionality-reduced fault feature data in the input layer, and the output result (h1,..., h t ,..., h n ) corresponding to each fault feature data can be obtained. For example, the operation fault type label corresponding to each fault feature data is obtained for further processing.

[0099] In one example, step 303 includes the following steps: Perform triple-weighting processing on the fault feature data according to the triple-weighting information in the hidden layer to obtain an output result; among them, the triple-weighting information includes a regularization weight, a sample weight, and a variable weight; the regularization weight represents the degree of regularization; the sample weight represents the importance of the data; the variable weight represents the contribution degree of the variable.

[0100] Exemplarily, for the hidden layer, a triple-weighting mechanism is deployed, that is, the hidden layer has triple-weighting information, which includes a regularization weight, a sample weight, and a variable weight; the regularization weight represents the degree of regularization; the sample weight represents the importance of the data; the variable weight represents the contribution degree of the variable. Based on the triple-weighting mechanism of the hidden layer, triple-weighting processing is performed on each fault feature data through the regularization weight, the sample weight, and the variable weight to obtain the output result corresponding to each fault feature data.

[0101] For example, based on the triple-weighting mechanism of the hidden layer, each piece of fault feature data is subjected to regularization weighting processing according to the regularization weight, and the regularization weighting result is subjected to sample weighting processing according to the sample weight, and the sample weighting result is subjected to variable weighting processing according to the variable weight, so as to obtain the operating fault type label corresponding to each piece of fault feature data.

[0102] In one example, step 303 specifically includes: performing weighting processing on the fault feature data according to the triple-weighting information and hidden layer parameters in the hidden layer to obtain an output result; wherein, the hidden layer parameters include input parameters and bias parameters.

[0103] Specifically, the hidden layer includes hidden layer parameters, specifically including input parameters and bias parameters. Then, based on the hidden layer, triple-weighting processing is performed on each piece of fault feature data according to the triple-weighting information and hidden layer parameters to obtain the output result corresponding to each piece of fault feature data.

[0104] For example, when an activation function is deployed in the hidden layer, the activation function includes input parameters and bias parameters. Based on the triple-weighting mechanism and activation function of the hidden layer, each piece of fault feature data is subjected to regularization weighting processing according to the regularization weight and hidden layer parameters, and the regularization weighting result is subjected to sample weighting processing according to the sample weight and hidden layer parameters, and the sample weighting result is subjected to variable weighting processing according to the variable weight and hidden layer parameters, so as to obtain the operating fault type label corresponding to each piece of fault feature data.

[0105] 304. Based on the output layer, process each output result to obtain the fault identification result of the target power grid.

[0106] Exemplarily, as Figure 4 shown, based on the output layer, according to a preset analysis algorithm, such as a voting algorithm or a mean-taking algorithm, process the operating fault types represented by each output result to obtain the fault identification result Y of the target power grid, that is, identify the final operating fault type of the target power grid.

[0107] In one example, step 304 includes: based on the output layer, process each output result according to the output layer weight to obtain the fault identification result.

[0108] Exemplarily, the output layer in the triple-weighting extreme learning machine has an output layer weight, that is, the output layer weight coefficient corresponding to each output result can be determined to perform weighting processing on each output result to obtain the final fault identification result.

[0109] In this embodiment, based on the above embodiment, the Analytic Hierarchy Process (AHP) is combined with an autoencoder network to simplify multi-dimensional power grid data to reduce costs, and an improved Extreme Learning Machine (ELM) is adopted as the model, aiming to improve the model's processing ability for dynamic data streams and the efficiency of the real-time update learning process; furthermore, it provides an effective and accurate method for evaluating and predicting the operating conditions of the smart grid, helps to detect and respond to potential power grid faults in advance, ensures the stable and reliable operation of the power grid, and thus supports sustainable energy utilization and the urban smart grid.

[0110] Figure 5 It is a schematic flowchart of a model training method for power grid fault identification provided by an embodiment of this application. As Figure 5 shown, the method includes:

[0111] 401. Obtain the fault data to be trained; wherein, the fault data to be trained represents the data generated by the power grid during operation faults in a historical time period.

[0112] Exemplarily, the execution subject of this embodiment can be an electronic device. The device can obtain the fault data to be trained from a local or remote database based on a training instruction input by the user. The fault data to be trained represents the data generated by the power grid during operation faults in a historical time period, such as the date and timestamp of the fault occurrence, the current and voltage measurement values at the time of the fault, environmental factors such as weather conditions, temperature, and humidity at the time of the fault, the load situation at the time of the fault including load type and load level, communication records related to the fault, such as the logs of remote monitoring and control systems, and if there are monitoring devices, it may include image or video records at the time of the fault.

[0113] 402. Perform hierarchical decomposition processing on the fault data to be trained to obtain at least one fault feature data corresponding to the fault data to be trained; wherein, the fault feature data represents the operation fault features of the power grid.

[0114] Exemplarily, the device performs hierarchical decomposition processing on the obtained fault data to be trained based on a preset analytic hierarchy process (AHP). For example, the AHP is used to hierarchically decompose the power grid fault types and their characteristics, transforming the complex fault identification problem into multiple relatively simple sub-problems to obtain the fault characteristics at each level. By constructing a hierarchical structure model, qualitative and quantitative analysis is performed on the fault characteristics at each level to determine the importance weights of each fault characteristic in the overall identification. Then, sorting processing is carried out from large to small according to the importance weights, and a preset number of top-ranked fault characteristics are extracted as the multiple fault characteristic data corresponding to the fault data to be trained, that is, the operating fault characteristics of multiple power grids. As a multi-criteria decision analysis method, AHP can decompose complex decision problems into multiple levels and sub-problems, and provide support for decision-making by comparing and judging the importance of various factors. In power grid fault identification, AHP can help determine the priorities and correlation relationships between different fault types and characteristics, providing a basis for subsequent fault handling and decision-making.

[0115] 403. Invoke the initial extreme learning machine; and based on each fault characteristic data, perform training processing on the initial extreme learning machine to obtain a triple-weighted extreme learning machine; wherein, the triple-weighted extreme learning machine is used to process real-time fault data in a power grid fault identification method to obtain a fault identification result; the fault identification result includes the operating fault type.

[0116] Exemplarily, the device invokes a preset initial extreme learning machine, and based on a preset model training method, can input all the fault characteristic data into the initial extreme learning machine to perform training processing on the initial extreme learning machine. For example, by introducing a triple-weighting mechanism, weighting processing is performed on the input features, output labels, and model parameters of the extreme learning machine to obtain a triple-weighted extreme learning machine. The triple-weighted extreme learning machine is used to process real-time fault data in a power grid fault identification method to obtain a fault identification result; the fault identification result includes the operating fault type.

[0117] In this embodiment, a model training method applied to power grid fault identification is provided. By introducing a triple-weighting mechanism, training processing is performed on the initial extreme learning machine to obtain a triple-weighted extreme learning machine, thereby further improving the prediction accuracy and stability of the model. This step realizes the accurate prediction of the probability of power grid faults, providing strong support for the safe and stable operation of the power grid.

[0118] Figure 6 FIG. is a schematic flowchart of another model training method applied to power grid fault identification provided by an embodiment of the present application, as Figure 6 shown, the method includes:

[0119] 501. Obtain the fault data to be trained; wherein, the fault data to be trained represents the data generated when the power grid experiences an operating fault during a historical time period.

[0120] Exemplarily, this step can refer to step 401, which will not be elaborated here.

[0121] 502. Perform hierarchical decomposition processing on the fault data to be trained to obtain at least one fault feature data corresponding to the fault data to be trained; wherein, the fault feature data represents the operating fault characteristics of the power grid.

[0122] Exemplarily, this step can refer to step 402, which will not be elaborated here.

[0123] In one example, it further includes: Based on the autoencoder network, perform dimensionality reduction processing on the fault feature data to obtain the processed fault feature data.

[0124] Exemplarily, Figure 7 is a schematic diagram of a power grid fault identification process provided by an embodiment of the present application. As Figure 7 shown, after performing feature decomposition on the fault data to be trained of the power grid, that is, the power grid operation data, based on hierarchical analysis, the device calls a preset autoencoder network, and based on this autoencoder network, performs dimensionality reduction processing on all the fault feature data. That is, an autoencoder network is introduced to further learn and optimize each fault feature data processed by AHP. The autoencoder network automatically extracts the effective information in the fault features through unsupervised learning, and constructs a low-dimensional representation that can accurately reflect each fault characteristic, thereby obtaining each processed fault feature data. Using the data preprocessing scheme of combining AHP and AE to extract features and reduce the dimension of the original multi-dimensional power grid data, this step effectively reduces the complexity of fault identification, improves the accuracy and efficiency of feature extraction, reduces the model reconstruction error, and also improves the generalization ability of the model to unknown fault types.

[0125] 503. Call the initial extreme learning machine; and based on the initial hidden layer, process the fault feature data to obtain the output result corresponding to the fault feature data.

[0126] Wherein, the initial extreme learning machine includes an initial output layer and an initial hidden layer.

[0127] Exemplarily, the initial extreme learning machine called by the device includes an initial output layer and an initial hidden layer. Through the initial hidden layer, triple weighted processing is performed on each fault feature data in the input layer or the dimensionality-reduced fault feature data, and the output result corresponding to each fault feature data can be obtained. For example, the operating fault type label corresponding to each fault feature data can be obtained for further processing.

[0128] For example, based on the triple weighting mechanism of the initial hidden layer, each fault feature data is weighted according to the initial first weight, and the single-weighted result is double-weighted according to the initial second weight, and the double-weighted result is triply weighted according to the initial third weight to obtain the operating fault type label corresponding to each fault feature data.

[0129] 504. Based on the initial output layer, each output result is processed to obtain a predicted fault type.

[0130] Exemplarily, based on the initial output layer, the operating fault type represented by each output result can be processed according to a preset analysis algorithm, such as a voting algorithm or an averaging algorithm, to obtain a predicted fault type.

[0131] 505. According to the predicted fault type, the initial hidden layer is trained to obtain a hidden layer.

[0132] Exemplarily, the device performs training processing on the initial hidden layer according to the obtained predicted fault type, such as performing parameter optimization, to obtain the hidden layer.

[0133] In one example, the predicted fault type has the actual fault type; step 505 includes: optimizing the triple weighted information of the initial hidden layer according to the predicted fault type and the actual fault type to obtain the hidden layer; wherein the triple weighted information includes regularization weight, sample weight and variable weight; regularization weight represents the degree of regularization; sample weight represents the importance of data; variable weight represents the contribution of variable.

[0134] Exemplarily, each predicted fault type has a corresponding actual fault type. According to the predicted fault type and the actual fault type, the triple weighted information of the initial hidden layer is optimized, and the triple weighted information includes regularization weight, sample weight and variable weight; the regularization weight represents the degree of regularization; the sample weight represents the importance of the data; the variable weight represents the contribution of the variable. For example, according to the predicted fault type and the actual fault type, the corresponding loss function is calculated to optimize the regularization weight, sample weight and variable weight of the initial hidden layer until the loss function reaches the minimum value, thereby obtaining the hidden layer.

[0135] For example, due to the special nature of the single hidden layer structure, ELM has a faster learning speed, minimal human intervention, and is easier to implement than traditional networks. Based on the principles of empirical risk minimization and structural risk minimization, a regularized extreme learning machine (Robust Extreme Learning Machine, referred to as RELM) is used, which has better generalization performance than ELM. Assume that the nth historical input variable vector and output variable are represented as xn = [x n1 , x n2 ,..., x nm and t n , where m is the number of input variables, (x n , t n ) is the nth historical sample composed of x n and t n . The output function of the RELM with L hidden layer neurons can be expressed as: N is the number of training samples; among them, β i is the output weight of the ith hidden layer unit, ω i = [ω j1 ,..., ω jm and b i are the input weight and the bias connecting the input layer and the ith hidden layer unit respectively; x j = [x j1 , x j2 ,..., x jm is the input variable vector, t j represents the output corresponding to x j . g() is the activation function. Usually, g() is set to the sigmoid function. Rewritten in matrix form:

[0136]

[0137] Among them, since ω i and b i are randomly given. In order to obtain the output weight vector β, the optimization equation can be expressed as:

[0138]

[0139] Among them, C represents the regularization coefficient, which can adjust the proportion of empirical risk and structural risk.

[0140] ξ = [ξ1,..., ξ N T is the training error vector. By constructing the Lagrangian function, the solution is:

[0141]

[0142] Among them, I L ∈ R L×L , I N ∈ R N×N ​; R is the set of real numbers. For the sample weighting process, not all samples contribute equally to the output during model training. However, the original ELM considers all samples equally important and does not take into account the differences between different samples. Therefore, to obtain more realistic results, a sample weight matrix W is added s = diag(W s1 ,..., W sV ), and is expressed in the following form:

[0143]

[0144] The Lagrangian function can be expressed as follows:

[0145]

[0146] where λ = [λ1,.., λ N , representing the Lagrange multiplier vector. According to the KKT conditions, taking the derivative and setting the derivative to zero, we get:

[0147]

[0148] Then the Lagrange multiplier vector λ can be expressed as: Similarly, the expression for the output weight vector of the regularized extreme learning machine with sample weighting is: When the number of modeling samples is greater than the number of hidden neurons, it has a faster calculation speed. For variable weighting, to reflect the differences in input variables and obtain better quality-related features, the Pearson correlation coefficient is used to reflect the contribution degrees of different variables. On this basis, a regularized extreme learning machine with variable weighting is proposed. The Pearson correlation coefficient is defined as: where E(x) and E(t) are the expectations of a single input variable and the output variable respectively. ρ represents the degree of correlation between two variables. Two highly correlated variables will also have a larger ρ. Therefore, the variable contribution degree can be defined by ρ. For training samples (x n , t n ), n = 1,..., k, where each input sample x n has m dimensions, and the contribution degree of each variable can be defined as:

[0149]

[0150] where, ρ i represents the Pearson correlation coefficient between the i-th input variable and the output variable. The variable contribution degree matrix can be written as: V = diag(v1,..., v m ), where, Represents the input samples with variable weighting. Each dimension of the input sample is assigned a different weight to reflect the differences between variables. Through variable weighting, we get:

[0151]

[0152] When L < N, the output weights are: Furthermore, based on the sample weighting, regularization, and variable weighting strategies, the input features, output labels, and model parameters of the extreme learning machine are weighted, thereby further improving the prediction accuracy and stability of the model.

[0153] 506. According to the predicted fault type, the initial output layer is trained to obtain the output layer; among them, the output layer and the hidden layer constitute a triple-weighted extreme learning machine.

[0154] Exemplarily, the device trains the initial output layer according to the obtained predicted fault type, for example, by optimizing the parameters, to obtain the output layer; furthermore, the obtained output layer and the hidden layer constitute a trained model, that is, a triple-weighted extreme learning machine, for power grid fault identification.

[0155] For example, Figure 8 This is the classification test effect diagram of a triple-weighted extreme learning machine provided by the embodiments of this application. As Figure 8 shown, the specific software and hardware configurations of the test platform are shown in Table 1 below.

[0156] Table 1

[0157] Name Version Operating System W10 CPU 5500U with 2.10GHz Algorithm Implementation Language MLab

[0158] The triple-weighted ELM algorithm is implemented through M Lab code, and the data set for experimental verification comes from an open-source website. Table 2 shows some information of the experimental data set.

[0159] Table 2

[0160] Parameter Value Dataset Size 357 Number of Labels 3 Training Set Ratio 80％ Test Set Ratio 20％

[0161] For the tested sequences, first, the Analytic Hierarchy Process (AHP) is used to hierarchically decompose the power grid fault types and their characteristics, transforming the complex fault identification problem into multiple relatively simple sub-problems. By constructing a hierarchical structure model, qualitative and quantitative analyses are carried out on the fault characteristics at each level to determine the importance weights of each fault characteristic in the overall identification. Subsequently, an autoencoder network is introduced to further learn and optimize the fault characteristics processed by AHP. This process not only reduces the model reconstruction error but also improves the generalization ability of the model for unknown fault types. Then, a triple-weighted ELM training model is used, and the trained model is used to judge the type of fault to obtain the classification test effect of this embodiment on the experimental data set.

[0162] In this embodiment, based on the above embodiment, the extreme learning machine model based on AHP and autoencoder has high efficiency in the training stage and can complete the construction and optimization of the model in a short time. This enables the model to quickly respond to power grid faults in practical applications and achieve real-time fault identification and location. At the same time, due to the good generalization ability of ELM, the model also has high reliability for the identification of unknown fault types.

[0163] Figure 9 It is a schematic structural diagram of a power grid fault identification device provided by an embodiment of the present application, as Figure 9 shown, the device includes:

[0164] An acquisition module 601, configured to acquire real-time fault data of the target power grid; wherein, the real-time fault data represents the data generated when the current target power grid has an operation fault;

[0165] A processing module 602, configured to perform hierarchical decomposition processing on the real-time fault data to obtain at least one fault feature data corresponding to the real-time fault data; wherein, the fault feature data represents the operation fault characteristics of the target power grid;

[0166] An identification module 603, configured to process each fault feature data based on a triple-weighted extreme learning machine to obtain a fault identification result of the target power grid; wherein, the fault identification result includes the operation fault type of the target power grid.

[0167] In a possible implementation manner, the triple-weighted extreme learning machine includes an output layer and a hidden layer; the identification module 603 is specifically configured to: perform triple-weighted processing on the fault feature data based on the hidden layer to obtain an output result corresponding to the fault feature data; and process each output result based on the output layer to obtain a fault identification result of the target power grid.

[0168] In a possible implementation manner, the recognition module 603 is specifically configured to: perform triple weighting processing on the fault feature data according to the triple weighting information in the hidden layer to obtain an output result; wherein, the triple weighting information includes a regularization weight, a sample weight, and a variable weight; the regularization weight represents the degree of regularization; the sample weight represents the importance of the data; and the variable weight represents the contribution degree of the variable.

[0169] In a possible implementation manner, the recognition module 603 is specifically configured to: perform weighting processing on the fault feature data according to the triple weighting information and the hidden layer parameters in the hidden layer to obtain an output result; wherein, the hidden layer parameters include input parameters and bias parameters.

[0170] In a possible implementation manner, the recognition module 603 is further specifically configured to: based on the output layer, process each output result according to the output layer weight to obtain a fault recognition result.

[0171] In a possible implementation manner, the device is further configured to: perform dimensionality reduction processing on the fault feature data based on an autoencoder network to obtain processed fault feature data.

[0172] The device of this embodiment can execute the technical solutions in the above grid fault recognition method, and the specific implementation process and technical principle are the same, which will not be elaborated here.

[0173] Figure 10 As shown in the structural schematic diagram of a model training device for grid fault recognition provided by an embodiment of the present application, Figure 10 shown, the device includes:

[0174] An acquisition module 701, configured to acquire fault data to be trained; wherein, the fault data to be trained represents data generated by the power grid during an operation fault in a historical time period;

[0175] A processing module 702, configured to perform hierarchical decomposition processing on the fault data to be trained to obtain at least one fault feature data corresponding to the fault data to be trained; wherein, the fault feature data represents the operation fault features of the power grid;

[0176] A training module 703, configured to call an initial extreme learning machine; and perform training processing on the initial extreme learning machine according to each fault feature data to obtain a triple-weighted extreme learning machine; wherein, the triple-weighted extreme learning machine is used to process real-time fault data in a grid fault recognition device to obtain a fault recognition result; the fault recognition result includes an operation fault type.

[0177] In a possible implementation manner, the initial extreme learning machine includes an initial output layer and an initial hidden layer; the training module 703 is specifically configured to: process the fault feature data based on the initial hidden layer to obtain an output result corresponding to the fault feature data; process each output result based on the initial output layer to obtain a predicted fault type; train and process the initial hidden layer according to the predicted fault type to obtain a hidden layer; train and process the initial output layer according to the predicted fault type to obtain an output layer; wherein, the output layer and the hidden layer constitute a triple-weighted extreme learning machine.

[0178] In a possible implementation manner, the predicted fault type has an actual fault type; the training module 703 is specifically configured to: optimize and process the triple-weighted information of the initial hidden layer according to the predicted fault type and the actual fault type to obtain a hidden layer; wherein, the triple-weighted information includes a regularization weight, a sample weight, and a variable weight; the regularization weight represents the degree of regularization; the sample weight represents the importance of data; the variable weight represents the contribution degree of variables.

[0179] In a possible implementation manner, the device is further configured to: perform dimensionality reduction processing on each fault feature data based on an autoencoder network to obtain at least one processed fault feature data.

[0180] The device in this embodiment can execute the technical solutions in the above-mentioned model training method applied to power grid fault identification. The specific implementation process and technical principle are the same, and will not be elaborated here.

[0181] Figure 11 This is a schematic structural diagram of an electronic device provided by an embodiment of the present application. As Figure 11 shown, the electronic device includes: a memory 801, a processor 802; the memory 801 is a memory for storing executable instructions of the processor 802.

[0182] Wherein, the processor 802 is configured to execute the method provided in the above-mentioned embodiment.

[0183] The electronic device further includes a receiver 803 and a transmitter 804. The receiver 803 is used to receive instructions and data sent by other devices, and the transmitter 804 is used to send instructions and data to external devices.

[0184] The specific implementation process of the processor can refer to the above method embodiment, and its implementation principle and technical effect are similar, and will not be elaborated here in this embodiment.

[0185] In the above embodiments, it should be understood that the processor may be a central processing unit (CPU), or may also be other general-purpose processors, digital signal processors (DSPs), application specific integrated circuits (ASICs), etc. The general-purpose processor may be a microprocessor or any conventional processor, etc. The steps of the method disclosed in combination with the invention can be directly implemented by a hardware processor, or implemented by a combination of hardware and software modules in the processor.

[0186] The embodiments of the present application also provide a chip for running instructions, and the chip is used to execute the technical solutions in the above embodiments.

[0187] The embodiments of the present application also provide a computer-readable storage medium, and computer-executable instructions are stored in the computer-readable storage medium. When the computer-executable instructions run on a computer, the computer is enabled to execute the technical solutions in the above embodiments.

[0188] The above-mentioned readable storage medium may be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory, electrically erasable programmable read-only memory, erasable programmable read-only memory, programmable read-only memory, read-only memory, magnetic memory, flash memory, magnetic disk or optical disk. The readable storage medium may be any available medium accessible by a general-purpose or special-purpose computer.

[0189] An exemplary readable storage medium is coupled to the processor, so that the processor can read information from the readable storage medium and write information to the readable storage medium. Of course, the readable storage medium may also be a component of the processor. The processor and the readable storage medium may be located in an application specific integrated circuit. Of course, the processor and the readable storage medium may also exist as discrete components in the device.

[0190] The embodiments of the present application also provide a computer program product, and the computer program product includes a computer program, which is stored in a computer-readable storage medium. At least one processor can read the computer program from the computer-readable storage medium, and when at least one processor executes the computer program, the technical solutions in the above embodiments can be implemented.

[0191] In addition, in each embodiment of the present invention, each functional unit may be integrated in one processing unit, or each unit may exist physically alone, or two or more units may be integrated in one unit.

[0192] If a function is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art or a part of this technical solution can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods of the various embodiments of the present invention. The aforementioned storage medium includes: USB flash drives, mobile hard disks, read-only memories (ROM), random access memories (RAM), magnetic disks, or optical discs and other various media that can store program codes.

[0193] Those of ordinary skill in the art can understand that all or part of the steps for implementing the above method embodiments can be completed by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When this program is executed, it executes the steps including the above method embodiments; and the aforementioned storage medium includes: ROM, RAM, magnetic disks, or optical discs and other various media that can store program codes.

[0194] Finally, it should be noted that those skilled in the art will easily think of other implementation manners of the present invention after considering the specification and practicing the invention disclosed herein. The present invention is intended to cover any variations, uses, or adaptations of the present invention, and these variations, uses, or adaptations follow the general principles of the present invention and include common general knowledge or conventional technical means in the technical field not disclosed in the present invention. It is not limited to the precise structures described above and shown in the drawings, and various modifications and changes can be made without departing from its scope. The scope of the present invention is only limited by the appended claims.

Claims

1. A power grid fault identification method, characterized in that, Including: Obtain real-time fault data of the target power grid; wherein, the real-time fault data represents data generated when the current target power grid has an operation fault. Perform hierarchical decomposition processing on the real-time fault data to obtain at least one fault feature data corresponding to the real-time fault data; wherein, the fault feature data represents the operation fault feature of the target power grid. Based on a triple weighted extreme learning machine, process each of the fault feature data to obtain a fault identification result of the target power grid; wherein, the fault identification result includes the operation fault type of the target power grid.

2. The method according to claim 1, wherein The triple weighted extreme learning machine includes an output layer and a hidden layer; based on the triple weighted extreme learning machine, processing each of the fault feature data to obtain the fault identification result of the target power grid includes: Based on the hidden layer, perform triple weighted processing on the fault feature data to obtain an output result corresponding to the fault feature data. Based on the output layer, process each of the output results to obtain the fault identification result of the target power grid.

3. The method according to claim 2, characterized in that Based on the hidden layer, performing triple weighted processing on the fault feature data to obtain an output result corresponding to the fault feature data includes: According to the triple weighted information in the hidden layer, perform triple weighted processing on the fault feature data to obtain the output result; wherein, the triple weighted information includes a regularization weight, a sample weight, and a variable weight; the regularization weight represents the degree of regularization; the sample weight represents the importance of the data; the variable weight represents the contribution degree of the variable.

4. The method according to claim 3, wherein According to the triple weighted information in the hidden layer, performing triple weighted processing on the fault feature data to obtain the output result includes: According to the triple weighted information and hidden layer parameters in the hidden layer, perform weighted processing on the fault feature data to obtain the output result; wherein, the hidden layer parameters include input parameters and bias parameters.

5. The method according to claim 2, wherein Based on the output layer, processing each of the output results to obtain the fault identification result of the target power grid includes: Based on the output layer, according to the output layer weights, process each of the output results to obtain the fault identification result.

6. The method according to any one of claims 1-5, characterized in that, The method further includes: Based on an autoencoder network, perform dimensionality reduction processing on the fault feature data to obtain processed fault feature data.

7. A model training method applied to power grid fault identification, characterized in that, Including: Obtain fault data to be trained; wherein, the fault data to be trained represents data generated when the power grid has an operation fault during a historical time period. Perform hierarchical decomposition processing on the fault data to be trained to obtain at least one fault feature data corresponding to the fault data to be trained; wherein, the fault feature data represents the operation fault feature of the power grid. Invoke an initial extreme learning machine; and according to each of the fault feature data, perform training processing on the initial extreme learning machine to obtain a triple weighted extreme learning machine; wherein, the triple weighted extreme learning machine is used to process the real-time fault data as described in any one of claims 1-6 to obtain a fault identification result; the fault identification result includes the operation fault type.

8. The method according to claim 7, wherein The initial extreme learning machine includes an initial output layer and an initial hidden layer; according to each piece of the fault feature data, the initial extreme learning machine is trained to obtain the triple weighted extreme learning machine, including: Based on the initial hidden layer, the fault feature data is processed to obtain an output result corresponding to the fault feature data; Based on the initial output layer, each of the output results is processed to obtain a predicted fault type; According to the predicted fault type, the initial hidden layer is trained to obtain a hidden layer; According to the predicted fault type, the initial output layer is trained to obtain an output layer; wherein, the output layer and the hidden layer constitute the triple weighted extreme learning machine.

9. The method according to claim 8, characterized in that, The predicted fault type has an actual fault type; according to the predicted fault type, the initial hidden layer is trained to obtain a hidden layer, including: According to the predicted fault type and the actual fault type, the triple weighted information of the initial hidden layer is optimized to obtain the hidden layer; wherein, the triple weighted information includes a regularization weight, a sample weight, and a variable weight; the regularization weight represents the degree of regularization; the sample weight represents the importance of data; the variable weight represents the contribution degree of variables.

10. The method according to any one of claims 7-9, characterized in that, The method further includes: Based on an autoencoder network, each piece of the fault feature data is dimensionally reduced to obtain at least one processed fault feature data.

11. A power grid fault identification device, characterized in that, Including: An acquisition module, configured to acquire real-time fault data of a target power grid; wherein, the real-time fault data represents data generated when the current target power grid has an operation fault; A processing module, configured to perform hierarchical decomposition processing on the real-time fault data to obtain at least one fault feature data corresponding to the real-time fault data; wherein, the fault feature data represents the operation fault characteristics of the target power grid; An identification module, configured to process each piece of the fault feature data based on a triple weighted extreme learning machine to obtain a fault identification result of the target power grid; wherein, the fault identification result includes the operation fault type of the target power grid.

12. A model training device applied to power grid fault identification, characterized in that Including: An acquisition module, configured to acquire fault data to be trained; wherein, the fault data to be trained represents data generated when the power grid has an operation fault during a historical time period; A processing module, configured to perform hierarchical decomposition processing on the fault data to be trained to obtain at least one fault feature data corresponding to the fault data to be trained; wherein, the fault feature data represents the operation fault characteristics of the power grid; A training module, configured to call an initial extreme learning machine; and train the initial extreme learning machine according to each piece of the fault feature data to obtain a triple weighted extreme learning machine; wherein, the triple weighted extreme learning machine is configured to process the real-time fault data as described in claim 11 to obtain a fault identification result; the fault identification result includes an operation fault type.

13. An electronic device, characterized in that, Including: A memory, a processor; The memory stores computer execution instructions; The processor executes the computer execution instructions stored in the memory, so that the processor executes the method according to any one of claims 1-10.

14. A computer-readable storage medium, characterized in that The computer-readable storage medium stores computer-executable instructions, which are used to implement the method according to any one of claims 1-10 when executed by a processor.

15. A computer program product, characterized in that, It includes a computer program, which implements the method according to any one of claims 1-10 when executed by a processor.

Images