Federal learning-based power grid line fault inspection method, equipment and medium

Abstract

The invention discloses a federated learning-based power grid line fault inspection method, equipment and a medium. The method comprises the following steps: each client obtains operation data of a local power grid line as a local data set; the central server generates and distributes global model parameters to each client; each client establishes and uses a local data set to train a local model according to the global model parameters, calculates generated gradient loss and uploads the gradient loss to the central server; the central server carries out aggregation processing on the gradient loss uploaded by each client to obtain a new round of global model parameters to continue a federated learning cycle until the global model parameters converge and / or the number of cycles reaches an upper limit; and each client uses the latest local model to carry out power grid line fault inspection work. According to the method, local computing resources of all participants are utilized, dependence on centralized computing power is reduced, model training efficiency is improved, it is ensured that data do not leave the local environment, and sensitive information is effectively protected.

Description

Technical Field

[0001] The present invention relates to the technical field of power grid line fault diagnosis, and in particular to a power grid line fault inspection method, equipment and medium based on federated learning. Background Art

[0002] As the construction of smart grids continues to advance, the grid system is becoming more and more complex, which requires not only efficient data processing capabilities, but also the ability to respond to various operating conditions in real time. In order to ensure the reliability and stability of the grid, accurate and timely fault diagnosis of grid lines has become an indispensable part. Traditional fault diagnosis methods such as manual inspections or conventional monitoring systems can no longer meet the needs of modern grids. They often have problems such as poor scalability and slow response speed, and are unable to cope with the processing of massive amounts of data.

[0003] However, there are many challenges in applying machine learning (ML) technology to power grid fault diagnosis. First, due to the huge amount of power grid data, traditional centralized learning frameworks may become inefficient and consume a lot of computing resources. In addition, power grid data usually contains sensitive information, so how to efficiently analyze data while ensuring data privacy and security has become an urgent problem to be solved. Furthermore, power grid failure modes are complex and changeable, and a single learning model is difficult to cope with all situations.

[0004] Although existing centralized and decentralized learning frameworks can improve the accuracy of fault diagnosis models to a certain extent, these frameworks require data to be collected and processed centrally, which not only increases the risk of data leakage, but also makes it difficult for the system to expand effectively when facing large-scale and high-dimensional data. Summary of the invention

[0005] In view of this, an embodiment of the present invention provides a power grid line fault inspection method, device and medium based on federated learning, which can ensure that the model is trained without leaving the local environment, thereby effectively protecting sensitive information.

[0006] A first aspect of the present invention provides a method for inspecting power line faults based on federated learning, comprising the following steps: Each client obtains the operating data of the local power grid line as a local data set; The central server generates initial model parameters and uses the initial model parameters as the global model parameters for the first round; The central server distributes global model parameters to each client, and each client establishes a local model according to the global model parameters; Each client uses a local data set to train the local model and calculates the gradient loss generated by the initial model parameters on the client local data set; Each client uploads the calculated gradient loss to the central server, which aggregates the gradient losses uploaded by each client to obtain a new round of global model parameters. Return to the step of distributing global model parameters from the central server to each client; each client re-executes the steps of establishing a local model according to the global model parameters, training the local model using a local data set, calculating the gradient loss generated by the initial model parameters on the client local data set, and uploading the calculated gradient loss to the central server; until the global model parameters converge and / or the number of gradient loss aggregation processing times reaches a preset upper limit; Each client uses the last established local model to perform power grid line fault inspection.

[0007] Furthermore, the operation data of the local power grid line includes at least sensor data, power grid status data and historical fault records.

[0008] Furthermore, the central server generates initial model parameters, specifically including the following steps: Get the preset initialization strategy ; According to the preset initialization strategy Preset model framework i Assign values ​​to get the initial model parameters i 0 .

[0009] Furthermore, the central server distributes the global model parameters to each client, specifically including the following steps: Create a client dataset ; According to the client data set Establish a separate data transmission channel for each client, and copy multiple sets of initial model parameters to each client. ; Distribute initial model parameters to each client through the established data transmission channel .

[0010] Furthermore, each client calculates the gradient loss of the initial model parameters on the client local data set, specifically including the following steps: For local datasets Annotate the local dataset Each sample in x All have corresponding tags y ; Use the local dataset For local models i t Perform parameter estimation and calculate the local model using the following formula i t Gradient loss: ; in,( x , y ) represents a local dataset A single sample in l represents the loss function, Indicates that a single sample is in the local model i t About the loss function l The gradient presented, Represents the local model i t The gradient loss.

[0011] Furthermore, the central server aggregates the gradient losses uploaded by each client, specifically including the following steps: Collect the gradient loss uploaded by each client , establish the gradient loss set ; Use the following formula to perform weighted average of the gradient loss uploaded by each client to obtain the global model parameters for the new round: i t+1 : ; in, i Represents each client, t Indicates the current round, m i Indicates the number of local dataset samples of each client, represents the gradient loss of each client.

[0012] Furthermore, the convergence of the global model parameters is specifically determined by the following formula: ; in, Represents the learning rate. When the learning rate When it is lower than the preset range, it is judged that the global model parameters have converged.

[0013] Furthermore, each client uses the last established local model to perform power grid line fault inspection, which specifically includes the following steps: Obtaining real-time operating parameters of power grid lines and performing data preprocessing on the real-time operating parameters of power grid lines; Inputting the real-time operation parameters of the power grid line into the local model established last time, and obtaining the power grid line fault parameters, specific fault type and fault location output by the local model established last time; Perform fault troubleshooting based on the output power grid line fault parameters, specific fault type and fault location.

[0014] A second aspect of the present invention discloses an electronic device, comprising a processor and a memory; The memory is used to store programs; The processor executes the program to implement the power grid line fault inspection method based on federated learning.

[0015] A third aspect of the present invention is a computer-readable storage medium, wherein the storage medium stores a program, and the program is executed by a processor to implement the method for inspecting power grid line faults based on federated learning.

[0016] The embodiment of the present invention also discloses a computer program product or a computer program, which includes a computer instruction stored in a computer-readable storage medium. A processor of a computer device can read the computer instruction from the computer-readable storage medium, and the processor executes the computer instruction, so that the computer device executes the above method.

[0017] The embodiments of the present invention have the following beneficial effects: The embodiments of the present invention propose a method, device and medium for power line fault inspection based on federated learning, which can ensure that the data is trained in the model without leaving the local environment, thereby effectively protecting sensitive information. In addition, in view of the low efficiency and high resource consumption of centralized learning when processing large-scale data sets, the present invention reduces the dependence on centralized computing power by utilizing the local computing resources of each participant, and improves the efficiency of model training. At the same time, for the deficiencies in scalability of centralized and decentralized learning frameworks, the federated learning method adopted by the present invention has better scalability, can adapt to the growing amount and complexity of data, and ensure the stable operation of the system. In addition, the present invention is also committed to improving the delay problem of model training, especially when using random forest and linear regression models, by optimizing algorithm design, reducing training time and improving the real-time response capability of the system. In summary, the present invention aims to solve the deficiencies of existing power line fault diagnosis methods in data privacy protection, resource efficiency, scalability and model training delay by introducing federated learning technology.

[0018] Additional aspects and advantages of the present invention will be given in part in the following description, and in part will be obvious from the following description, or will be learned through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0020] Figure 1 The present invention is a flowchart of the steps of a power grid line fault inspection method based on federated learning.

[0021] Figure 2 Schematic diagram of the federated learning framework of the present invention.

[0022] Figure 3 It is a structural schematic diagram of an electronic device of the present invention.

[0023] Figure 4 It is a schematic diagram of the structure of a computer-readable storage medium of the present invention. DETAILED DESCRIPTION

[0024] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0025] In the field of power grid line fault diagnosis, a centralized learning framework is a commonly used method that improves the accuracy of the model by building and training machine learning models through integrated data sets. However, this method faces the problems of low efficiency and high resource consumption when processing large-scale and complex data sets, and the need to store and process data in a centralized manner leads to hidden dangers in data privacy and security, especially in the power industry involving sensitive operational data. In addition, with the increase in the scale and complexity of power grid data, the effective scalability of centralized learning frameworks has also encountered difficulties. Specifically, in the centralized data processing method of the prior art, all data needs to be uploaded to a central server for processing. This method is prone to the leakage of sensitive data, especially in the power industry, where power grid data contains a large amount of information on critical infrastructure, which may bring serious security risks once leaked.

[0026] As for the distributed framework, although the distributed learning framework distributes the data set to multiple computing nodes for parallel processing and uses technologies such as Apache Spark to improve processing efficiency and solve the problems caused by centralized data processing, there are still some significant disadvantages. First, since the data is distributed among multiple nodes, a complex mechanism is required to ensure the balanced distribution of data. Otherwise, some nodes may become bottlenecks due to processing more data, resulting in an unbalanced load in the overall training process and the overall performance failing to meet expectations. In addition, since the intermediate results need to be transmitted across nodes, the network communication overhead is large, which further affects the speed and efficiency of model training.

[0027] To solve the above problems, Figure 1 As shown, the first embodiment of the present invention provides a method for inspecting power line faults based on federated learning, comprising the following steps: S1. Each client obtains the operation data of the local power grid line as a local data set; S2. The central server generates initial model parameters and uses the initial model parameters as the global model parameters of the first round; S3. The central server distributes global model parameters to each client, and each client establishes a local model based on the global model parameters; S4. Each client uses a local data set to train the local model and calculates the gradient loss of the initial model parameters on the client local data set; S5. Each client uploads the calculated gradient loss to the central server, and the central server aggregates the gradient losses uploaded by each client to obtain a new round of global model parameters; S6. Return to the step of distributing global model parameters from the central server to each client; each client re-executes the steps of establishing a local model according to the global model parameters, training the local model using a local data set, calculating the gradient loss generated by the initial model parameters on the client local data set, and uploading the calculated gradient loss to the central server; until the global model parameters converge and / or the number of gradient loss aggregation processing times reaches a preset upper limit; S7. Each client uses the last established local model to perform power grid line fault inspection.

[0028] The embodiment of the present invention reduces the reliance on centralized computing power and improves the efficiency of model training by utilizing the local computing resources of each participant. At the same time, for the deficiencies of centralized and decentralized learning frameworks in scalability, the federated learning method adopted by the present invention has better scalability, can adapt to the growing amount and complexity of data, and ensure the stable operation of the system.

[0029] Federated Learning is an advanced distributed machine learning method that has significant advantages in data privacy protection and data utilization efficiency. Figure 2 In federated learning, multiple clients can share the results of model training while keeping data local.

[0030] Specifically, federated learning changes the object of data interaction from a data set to a global model. The central server sends the global model to the client, and the client returns the gradient loss generated by the global model update to the central server. Therefore, in the work of power grid line fault inspection, federated learning can train a relatively comprehensive power grid fault identification model using data collected from different locations without sharing the original power grid data and protecting the security of the power grid data.

[0031] On the other hand, the number of clients in federated learning is not specific, and the access or exit of each power grid client terminal to the federated learning framework will not affect the operation of the federated learning framework. Therefore, the power grid line fault inspection method of the embodiment of the present invention has strong scalability and can accommodate an unlimited number of power grid clients connected to the central server. Since the client trains the global model locally on the client, the central server only needs to complete the aggregation processing of the data uploaded by the client, which reduces the load requirements of the central server and improves the training efficiency of the global model.

[0032] The following is a detailed description of the implementation process of each step of the embodiment of the present invention: S1. Each client obtains the operating data of the local power grid line as a local data set.

[0033] In the embodiment of the present invention, the operation data of the local power grid line includes at least sensor data, power grid status data and historical fault records. Since the fault type and fault cause of the power grid line are related to various factors, such as geographical location, weather conditions, equipment operation time, etc., multiple sensors will be deployed in the power grid line to monitor the operation status of the power grid line in real time. At the same time, the power grid status data and historical fault records will be recorded in each power grid terminal for maintenance personnel to review. The embodiment of the present invention collects the operation data of the local power grid line including at least sensor data, power grid status data and historical fault records, and establishes a local data set. Train the local model.

[0034] S2. The central server generates initial model parameters and uses the initial model parameters as the global model parameters of the first round.

[0035] In the embodiment of the present invention, the central server generates initial model parameters, which specifically includes the following steps: Get the preset initialization strategy ; According to the preset initialization strategy Assign values ​​to the preset model framework θ to obtain the initial model parameters i 0 .

[0036] In the field of machine learning, using a suitable initialization strategy to initialize model parameters can enable the model to generate a suitable gradient during the training process, so that it is not too large or too small, so as to further optimize the model. The initialization strategies that can be used in the embodiments of the present invention include random initialization strategy (Random Initialization), He initialization strategy (HeInitialization), orthogonal initialization strategy, etc. Taking the random initialization strategy as an example, the embodiments of the present invention can establish a distribution (such as a normal distribution or a uniform distribution) for each model parameter, and randomly assign values ​​to the model parameters according to the preset parameter distribution, so that the generated global model has good randomness.

[0037] S3. The central server distributes the global model parameters to each client, and each client establishes a local model according to the global model parameters.

[0038] In the embodiment of the present invention, the central server distributes the global model parameters to each client, specifically including the following steps: S3-1. Create a client dataset ; S3-2. Based on the client data set Establish a separate data transmission channel for each client, and copy multiple sets of initial model parameters to each client. ; S3-3. Distribute initial model parameters to each client through the established data transmission channel .

[0039] In the embodiment of the present invention, since the power grid client connected to the central server may change, a client data set needs to be established in the central server for storage, and the central server determines the power grid client participating in the federated learning based on the established client data set. When the client status changes later, the client data set is synchronously updated to avoid suspension or confusion of federated learning.

[0040] In some embodiments, if the resources of the central server are available, encryption operations can be performed on the data transmission channel established between the central server and the power grid client. By using encryption technologies such as homomorphic encryption or secure multi-party computing, data privacy can be protected during data transmission and processing.

[0041] After establishing the data transmission channel, copy multiple groups of initial model parameters and distribute them to each client. , to ensure that each client can collaborate under unified starting conditions and jointly build a more powerful and accurate global model.

[0042] S4. Each client uses a local data set to train the local model and calculates the gradient loss generated by the initial model parameters on the client local data set.

[0043] S4-1. In the embodiment of the present invention, each client calculates the gradient loss generated by the initial model parameters on the client local data set, specifically including the following steps: S4-2. For local datasets Annotate the local dataset Each sample in x All have corresponding tags y ; S4-3. Using local datasets For local models i t Perform parameter estimation and calculate the local model using the following formula i t Gradient loss: ; in,( x , y ) represents a local dataset A single sample in l represents the loss function, Indicates that a single sample is in the local model i t About the loss function l The gradient presented, Represents the local model i t The gradient loss.

[0044] In the embodiment of the present invention, the sample x The corresponding label y It is used to indicate the power grid fault information recorded by the data, and is manually labeled by the actual inspection personnel of the power grid client. In some embodiments, the historical fault records established in the form of logs may have recorded the fault information corresponding to the power grid data sample, so the historical fault records can also be directly imported to complete the automatic labeling of the sample.

[0045] The calculation process of the gradient loss involves the calculation of the minimum value of the loss function. The minimum value is used to represent the gap between the sample label and the model output, and the gap is used as the gradient loss of the model. The loss function used in the embodiment of the present invention is the Mean Squared Error function; in some embodiments, functions including Cross-Entropy Loss and Mean Absolute Error can also be used for regression calculation to obtain the gradient loss generated by the model. In the embodiment of the present invention, the loss function used by different clients is the same loss function to ensure the commonality of the gradient loss obtained by training.

[0046] S5. Each client uploads the calculated gradient loss to the central server. The central server aggregates the gradient losses uploaded by each client to obtain a new round of global model parameters.

[0047] In the embodiment of the present invention, the central server aggregates the gradient losses uploaded by each client, specifically including the following steps: S5-1. Collect the gradient loss uploaded by each client , establish the gradient loss set ; S5-2. Use the following formula to perform weighted average on the gradient loss uploaded by each client to obtain the global model parameters for the new round: i t+1 : ; in, i Represents each client, t Indicates the current round, m i Indicates the number of local dataset samples of each client, represents the gradient loss of each client.

[0048] In an embodiment of the present invention, the client uploads the model gradient loss obtained by local training to the central server, and the central server aggregates these gradient losses using weighted average to generate new global model parameters.

[0049] In some embodiments, those skilled in the art can add noise to the gradient loss uploaded by each client through differential privacy and other methods. Adding noise to the data uploaded by the client can not only improve the robustness of the model, but also protect the local data privacy of the client to a certain extent. However, when the noise level is high, it may affect the performance of the model. Therefore, for scenarios with high requirements for calculation accuracy, it is not advisable to mix noise into the gradient loss.

[0050] In some embodiments, only the gradient losses of some clients can be selected for aggregation in each gradient loss aggregation process, and the gradient losses of other clients in this round do not need to be obtained. This operation can reduce the communication volume of the federated learning framework interaction and reduce the data transmission burden.

[0051] S6. Return to the step of distributing the global model parameters from the central server to each client; each client re-executes the steps of establishing a local model according to the global model parameters, training the local model using a local data set, calculating the gradient loss generated by the initial model parameters on the client's local data set, and uploading the calculated gradient loss to the central server; until the global model parameters converge and / or the number of gradient loss aggregation processing times reaches a preset upper limit.

[0052] After the central server calculates the new global model parameters, it distributes them to each client. The client receives the new global model parameters and uses them to update the local model and enter the next training round. This process will continue until the constructed global model converges or reaches the preset maximum number of iterations T.

[0053] In the embodiment of the present invention, the convergence of the global model parameters is specifically determined by the following formula: ; in, Represents the learning rate. When the learning rate When it is lower than the preset range, it is judged that the global model parameters have converged.

[0054] S7. Each client uses the last established local model to perform power grid line fault inspection.

[0055] In step S7, after the local model established last time is applied to the power grid line fault inspection work, the specific workflow is as follows: Data collection: Various sensors and monitoring equipment installed along the power grid lines are used to collect real-time status data of the power grid lines, including but not limited to environmental parameters such as voltage, current, temperature, and humidity.

[0056] Data preprocessing: On each client (representing different geographical areas), the received data is preprocessed, including steps such as data cleaning (removing invalid or erroneous data) and standardization (ensuring that the data is at the same scale) so that it can be used as model input.

[0057] Fault prediction: The client uses the optimized model trained by the federated learning framework to predict the preprocessed data and determine whether there are potential faults or abnormal conditions. The model output may include information such as the type of fault and the location where it may occur.

[0058] Result upload: The client uploads the prediction results to the server or cloud management system for reference by operation and maintenance personnel.

[0059] Decision support: Based on the uploaded prediction results, operation and maintenance personnel can make corresponding decisions, such as dispatching staff to the site for inspection, maintenance or preventive measures.

[0060] Feedback loop: Operation and maintenance personnel feed back actual inspection results to the system, which is used as supplementary data for model training to further optimize model performance. Through continuous feedback and adjustment, the model can gradually improve the accuracy and reliability of fault prediction.

[0061] The embodiment of the present invention utilizes federated learning technology to enable each client to participate in model training without sharing the original data, thereby effectively protecting the privacy of the power system data. This is one of the core technical points of the present invention and is also the key to achieving data privacy protection. The model gradient losses uploaded by each client are aggregated through a federal averaging strategy to generate new global model parameters. This technical point ensures that the diagnostic model parameters can be effectively updated in a distributed environment of power grid data, thereby improving the overall efficiency of model training. In summary, the present invention aims to solve the deficiencies of existing power line fault diagnosis methods in terms of data privacy protection, resource efficiency, scalability, and model training delay by introducing federated learning technology.

[0062] Figure 3 : is a schematic diagram of the structure of the electronic device proposed in the second embodiment of the present invention. In the memory of this embodiment, program instructions for implementing the method for inspecting power line faults based on federated learning of any of the above embodiments are stored. The processor is used to execute the program instructions stored in the memory to perform inspection of power line faults based on federated learning. Among them, the processor can also be called a CPU (Central Processing Unit). The processor may be an integrated circuit chip with signal processing capabilities. The processor may also be a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0063] The contents of the method in the first embodiment of the present invention are all applicable to the electronic device embodiment. The functions specifically implemented by the electronic device embodiment are the same as those of the above method embodiment, and the beneficial effects achieved are also the same as those achieved by the above method.

[0064] Figure 4It is a schematic diagram of the structure of the computer-readable storage medium of the third embodiment of the present invention. The computer-readable storage medium of the fourth embodiment of the present invention stores program instructions that can implement the above-mentioned power grid line fault inspection method based on federated learning, wherein the program instructions can be stored in the above-mentioned storage medium in the form of a software product, including a number of instructions to enable a computer device (which can be a personal computer, server, or network device, etc.) or a processor (processor) to perform all or part of the steps of the methods of each embodiment of the present invention. The aforementioned computer-readable storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM, Read-Only Memory), a random access memory (RAM, Random Access Memory), a disk or an optical disk, or a terminal device such as a computer, a server, a mobile phone, and a tablet.

[0065] The contents of the method in the first embodiment of the present invention are all applicable to the computer-readable storage medium embodiment. The functions specifically implemented by the computer-readable storage medium embodiment are the same as those of the above method embodiment, and the beneficial effects achieved are also the same as those achieved by the above method.

[0066] This embodiment also provides a computer program product. When the computer program product runs on a computer, it enables the computer to execute the above-mentioned related steps to implement the power grid line fault inspection method based on federated learning provided in the above embodiment.

[0067] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in the embodiments of the present invention are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data need to comply with the relevant laws, regulations and standards of relevant countries and regions, and provide corresponding operation entrances for users to choose to authorize or refuse.

[0068] Those skilled in the art will appreciate that the modules in the devices in the embodiments of the present invention may be adaptively changed and arranged in one or more devices different from the embodiments. The modules or units or components in the embodiments of the present invention may be combined into one module or unit or component, and in addition they may be divided into a plurality of submodules or subunits or subcomponents. Except that at least some of such features and / or processes or units are mutually exclusive, all features disclosed in this specification (including the corresponding claims, abstracts and drawings) and all processes or units of any method or device disclosed in this manner may be combined in any combination. Unless otherwise expressly stated, each feature disclosed in this specification (including the corresponding claims, abstracts and drawings) may be replaced by an alternative feature providing the same, equivalent or similar purpose.

[0069] Through the description of the above implementation methods, technical personnel in the relevant field can understand that for the convenience and simplicity of description, only the division of the above-mentioned functional modules is used as an example. In actual applications, the above-mentioned functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above.

[0070] It should be noted that the logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be specifically implemented in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute instructions), or in combination with these instruction execution systems, devices or apparatuses. For the purposes of this specification, "computer-readable medium" can be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in combination with these instruction execution systems, devices or apparatuses. More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection portion with one or more wirings (electronic device), a portable computer disk box (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium and then editing, interpreting or processing in other suitable ways if necessary, and then stored in a computer memory.

[0071] In addition, each embodiment in this specification is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other. In particular, for embodiments such as devices and equipment, since they are basically similar to method embodiments, the relevant parts can refer to the partial description of the method embodiments. The embodiments of the devices and equipment described above are merely schematic, wherein the modules, units, etc. described as separate components may or may not be physically separated, that is, they may be located in one place, or they may be distributed in multiple places, such as nodes of a system network. Specifically, some or all of the modules and units may be selected according to actual needs to achieve the purpose of the above-mentioned embodiment scheme. Those skilled in the art can understand and implement it without paying creative labor.

[0072] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, a plurality of steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0073] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "examples", "specific examples", or "some examples" means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner.

[0074] In addition, the terms "first", "second", etc. used in the embodiments of the present invention are only used for descriptive purposes and should not be understood as indicating or implying relative importance, or implicitly indicating the number of technical features indicated in the present embodiment. Therefore, the features defined by the terms "first", "second", etc. in the embodiments of the present invention can explicitly or implicitly indicate that the embodiment includes at least one of the features. In the description of the present invention, the word "multiple" means at least two or two or more, such as two, three, four, etc., unless otherwise clearly and specifically defined in the embodiments.

[0075] In the embodiments of the present invention, the terms "include", "comprises" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, an element defined by the sentence "includes a ..." does not exclude the existence of other identical elements in the process, method, article or device including the element. In addition, components, features, and elements with the same name in different embodiments of the present invention may have the same meaning or different meanings, and their specific meanings need to be determined by their explanation in the specific embodiment or further combined with the context in the specific embodiment.

[0076] Although embodiments of the present invention have been shown and described above, it is to be understood that the above embodiments are exemplary and cannot be construed as limitations of the present invention, and those of ordinary skill in the art may change, modify, replace and modify the above embodiments within the scope of the present invention. Those skilled in the art will readily come to think of other embodiments of the present invention after considering the specification and practicing the present invention. This application is intended to cover any variations, uses or adaptations of the present invention, which follow the general principles of the present invention and include common knowledge or customary technical means in the art that are not disclosed in the present invention. The specification and embodiments are intended to be exemplary only, and the true scope and spirit of the present invention are indicated by the following claims.

Claims

1. A method for fault inspection of power grid lines based on federated learning, characterized in that: The following steps are involved: Each client obtains the operating data of the local power grid line as a local data set; The central server generates initial model parameters and uses the initial model parameters as the global model parameters for the first round; The central server distributes global model parameters to each client, and each client establishes a local model according to the global model parameters; Each client uses a local data set to train the local model and calculates the gradient loss generated by the initial model parameters on the client local data set; Each client uploads the calculated gradient loss to the central server, which aggregates the gradient losses uploaded by each client to obtain a new round of global model parameters. Return to the step of distributing global model parameters from the central server to each client; each client re-executes the steps of establishing a local model according to the global model parameters, training the local model using a local data set, calculating the gradient loss generated by the initial model parameters on the client local data set, and uploading the calculated gradient loss to the central server; Until the global model parameters converge and / or the number of gradient loss aggregation processes reaches a preset upper limit; Each client uses the last established local model to perform power grid line fault inspection.

2. A method for inspecting power line faults based on federated learning according to claim 1, characterized in that: The operation data of the local power grid line at least includes sensor data, power grid status data and historical fault records.

3. A method for power grid line fault inspection based on federated learning according to claim 1, characterized in that: The central server generates initial model parameters, specifically including the following steps: Get the preset initialization strategy ; According to the preset initialization strategy Preset model framework θ Assign values ​​to get the initial model parameters θ 0.

4. A method for power grid line fault inspection based on federated learning according to claim 1, characterized in that: The central server distributes the global model parameters to each client, specifically including the following steps: Create a client dataset ; According to the client data set Establish a separate data transmission channel for each client, and copy multiple sets of initial model parameters to each client. ; Distribute initial model parameters to each client through the established data transmission channel .

5. A method for power grid line fault inspection based on federated learning according to claim 1, characterized in that: Each client calculates the gradient loss generated by the initial model parameters on the client local data set, specifically including the following steps: For local datasets Annotate the local dataset Each sample in x All have corresponding tags y ; Use the local dataset For local models θ t Perform parameter estimation and calculate the local model using the following formula θ t Gradient loss: ； in,( x , y ) represents a local dataset A single sample in l represents the loss function, Indicates that a single sample is in the local model θ t About the loss function l The gradient presented, Represents the local model θ t The gradient loss.

6. A method for power grid line fault inspection based on federated learning according to claim 1, characterized in that: The central server aggregates the gradient losses uploaded by each client, specifically including the following steps: Collect the gradient loss uploaded by each client , establish the gradient loss set ; Use the following formula to perform weighted average of the gradient loss uploaded by each client to obtain the global model parameters for the new round: θ t+1 : ； in, i For each client, t Indicates the current round. m i Indicates the number of local dataset samples of each client, represents the gradient loss of each client.

7. A method for power grid line fault inspection based on federated learning according to claim 6, characterized in that: The global model parameter convergence is specifically determined by the following formula: ； in, Represents the learning rate. When the learning rate When it is lower than the preset range, it is judged that the global model parameters have converged.

8. A method for power grid line fault inspection based on federated learning according to claim 1, characterized in that: Each client uses the last established local model to perform power grid line fault inspection work, which specifically includes the following steps: Obtaining real-time operating parameters of power grid lines and performing data preprocessing on the real-time operating parameters of power grid lines; Inputting the real-time operation parameters of the power grid line into the local model established last time, and obtaining the power grid line fault parameters, specific fault type and fault location output by the local model established last time; Perform fault troubleshooting based on the output power grid line fault parameters, specific fault type and fault location.

9. An electronic device, characterized in that: including a processor and a memory; The memory is used to store programs; The processor executes the program to implement a power grid line fault inspection method based on federated learning as described in any one of claims 1-8.

10. A computer-readable storage medium, characterized in that: The storage medium stores a program, and the program is executed by a processor to implement a power grid line fault inspection method based on federated learning as described in any one of claims 1 to 8.

Images