A power equipment life prediction method and device based on a hierarchical asynchronous mechanism

Abstract

The embodiments of this disclosure provide a method and apparatus for predicting the lifespan of power equipment based on a hierarchical asynchronous mechanism. The method includes: acquiring operating data of a target power equipment; preprocessing the operating data and extracting features; constructing a similarity matrix based on the extracted features; clustering the similarity matrix using a spectral clustering algorithm to obtain multiple clustering results; inputting each clustering result into a pre-constructed target model to output the remaining lifespan of the target power equipment.

Description

Technical Field

[0001] This disclosure relates to the fields of intelligent operation and maintenance of power equipment and artificial intelligence technology, specifically to providing a method and device for predicting the lifespan of power equipment based on a hierarchical asynchronous mechanism. Background Technology

[0002] In the field of power equipment lifespan prediction and condition assessment, a centralized data analysis approach is typically used. This involves aggregating operational data from various equipment onto a central server and then using machine learning or deep learning models for unified modeling and prediction. However, this method has shortcomings in practical applications. Power equipment is widely distributed, and data is managed separately by different units. Centralized transmission can easily lead to privacy leaks and cybersecurity risks, while also increasing communication burdens. Differences in operating environments, load conditions, and usage patterns among different devices result in inconsistent data distribution, making it difficult for a unified model to account for various characteristics and leading to poor stability in prediction results. Some terminal devices have limited computing power and need to wait for all nodes to complete updates during synchronous training, making them susceptible to the influence of weaker nodes and reducing overall efficiency. While existing distributed or federated learning methods alleviate the problem of data centralization, they often employ fully synchronous or fully asynchronous mechanisms. The former suffers from waiting issues, while the latter is prone to affecting model convergence stability due to update discrepancies. With the increasing scale of equipment and the growing complexity of data, these problems become more pronounced, making it difficult to meet the needs of practical applications. Summary of the Invention

[0003] This disclosure provides a method and apparatus for predicting the lifespan of power equipment based on a hierarchical asynchronous mechanism.

[0004] According to a first aspect of this disclosure, a method for predicting the lifetime of power equipment based on a hierarchical asynchronous mechanism is provided. The method includes: The operating data of the target power equipment is acquired, and the operating data is preprocessed and features are extracted. Based on the extracted features, a similarity matrix is ​​constructed; The similarity matrix was clustered using a spectral clustering algorithm to obtain multiple clustering results. Each clustering result is input into a pre-built target model, which outputs the remaining lifespan of the target power equipment.

[0005] In addition to the aspects and any possible implementations described above, a further implementation is provided, wherein the runtime data includes: Voltage, current, temperature, load conditions, and operating status.

[0006] In addition to the aspects and any possible implementations described above, a further implementation is provided in which the preprocessing of the runtime data includes: The running data is processed by outlier removal, missing value completion, and data normalization.

[0007] As described above and in any possible implementation, a further implementation is provided, wherein the feature extraction includes: The preprocessed running data is segmented using a sliding time window method, and statistical feature parameters are extracted from the segmented running data. The statistical feature parameters include mean, variance, maximum value and minimum value. Features are generated based on the statistical feature parameters.

[0008] In addition to the aspects and any possible implementations described above, a further implementation is provided in which the construction of a similarity matrix based on the extracted features includes: Based on the extracted features, the cosine similarity between the features of any two target power devices is calculated, and a similarity matrix is ​​constructed based on the cosine similarity.

[0009] As described above and in any possible implementation, a further implementation is provided, wherein the similarity matrix is ​​clustered using a spectral clustering algorithm to obtain multiple clustering results, including: The similarity matrix is ​​decomposed using a spectral clustering algorithm. The eigenvectors corresponding to the top K largest eigenvalues ​​are selected to form a new feature space, where K is the preset number of clusters. The target power equipment is clustered using the K-means clustering algorithm in the new feature space to obtain multiple clustering results.

[0010] In addition to the aspects and any possible implementations described above, a further implementation is provided in which the training process of the pre-built target model includes the following steps: A randomly initialized regression prediction model is generated on the central server, and the central server sends the initialized regression prediction model to the client. N clients train simultaneously. On each client participating in the training, the client's local model is initialized with the regression prediction model transmitted from the central server. The extracted features are used to train the client's regression prediction model. The training is iterated until the preset local training rounds M or the error threshold condition is met. The trained regression prediction model is then uploaded to the group's server, and the regression prediction model parameters are updated using the stochastic gradient descent algorithm. After training is completed on N clients, the updated regression prediction model parameters are uploaded to the group's server. After receiving the updated regression prediction model parameters, the group server performs weighted aggregation on the trained regression prediction model based on the amount of client data to obtain a group-level model. The group-level server uploads the group-level model to the central server. After receiving the group-level model, the central server performs synchronous aggregation on the group-level model using a weighted average method to generate a target model. The central server then sends the target model to the client.

[0011] According to a second aspect of this disclosure, a power equipment lifetime prediction device based on a hierarchical asynchronous mechanism is provided. The device includes: The acquisition module is used to acquire the operating data of the target power equipment, and to preprocess and extract features from the operating data; The building block is used to construct a similarity matrix based on the extracted features; The clustering module is used to cluster the similarity matrix using a spectral clustering algorithm to obtain multiple clustering results. The output module is used to input each clustering result into the pre-built target model and output the remaining service life of the target power equipment.

[0012] According to a third aspect of this disclosure, an electronic device is provided. The electronic device includes a memory and a processor, wherein the memory stores a computer program, and the processor executes the program to implement the method described above.

[0013] According to a fourth aspect of this disclosure, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the methods as described in the first and / or second aspects of this disclosure.

[0014] It should be understood that the description in the Summary of the Invention is not intended to limit the key or essential features of the embodiments of this disclosure, nor is it intended to restrict the scope of this disclosure. Other features of this disclosure will become readily apparent from the following description. Attached Figure Description

[0015] The above and other features, advantages, and aspects of the embodiments of this disclosure will become more apparent from the accompanying drawings and the following detailed description. The drawings are provided for a better understanding of the invention and are not intended to limit the scope of this disclosure. In the drawings, the same or similar reference numerals denote the same or similar elements, wherein: Figure 1 A flowchart of a power equipment lifetime prediction method 100 based on a hierarchical asynchronous mechanism according to an embodiment of the present disclosure is shown; Figure 2 A block diagram of a power equipment lifetime prediction device 200 based on a hierarchical asynchronous mechanism according to an embodiment of the present disclosure is shown. Figure 3 A block diagram of an exemplary electronic device 300 capable of implementing embodiments of the present disclosure is shown. Detailed Implementation

[0016] To make the objectives, technical solutions, and advantages of the embodiments of this disclosure clearer, the technical solutions of the embodiments of this disclosure will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this disclosure, and not all embodiments. Based on the embodiments of this disclosure, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this disclosure.

[0017] Furthermore, the term "and / or" in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. Additionally, the character " / " in this article generally indicates that the preceding and following related objects have an "or" relationship.

[0018] In this disclosure, without leaving the local data, a fusion of graph clustering and hierarchical asynchronous federated learning mechanisms is used to achieve collaborative utilization of power equipment operation data in multiple regions, improve the model's adaptability to heterogeneous data, and construct a hierarchical federated learning framework with a two-level structure of intra-group and inter-group. An asynchronous update mechanism is used within groups, while a synchronous aggregation method is used between groups to reduce node waiting time and ensure the consistency of the global model. In the model training process, an intra-group consistency regularization term is introduced to constrain the differences in model parameters within the same group, enhance the intra-group collaborative learning capability, improve the lifetime prediction accuracy, reduce the impact of outdated updates on model training, and improve the stability of the model convergence process.

[0019] Figure 1 A flowchart of a power equipment lifetime prediction method 100 based on a hierarchical asynchronous mechanism according to an embodiment of the present disclosure is shown.

[0020] like Figure 1 As shown, the power equipment lifetime prediction method based on a hierarchical asynchronous mechanism includes: S101: Acquire the operating data of the target power equipment, preprocess the operating data, and extract features; S102: Construct a similarity matrix based on the extracted features; S103: The similarity matrix is ​​clustered using a spectral clustering algorithm to obtain multiple clustering results; S104: Input each clustering result into the pre-built target model and output the remaining service life of the target power equipment.

[0021] In some embodiments, the runtime data includes: Voltage, current, temperature, load conditions, and operating status.

[0022] In some embodiments, each power device acts as an independent client, deployed in different regions or operating environments, and continuously collects local operating data.

[0023] In some embodiments, the preprocessing of the runtime data includes: The running data is processed by outlier removal, missing value completion, and data normalization.

[0024] In some embodiments, the normalization process employs a minimum-maximum normalization method to map data of different dimensions to a uniform numerical range, thereby ensuring numerical stability during model training and eliminating the impact of dimensional differences on model training.

[0025] In some embodiments, the feature extraction includes: segmenting the preprocessed running data using a sliding time window method, and extracting statistical feature parameters from the segmented running data, wherein the statistical feature parameters include mean, variance, maximum value, and minimum value; Features are generated based on the statistical feature parameters.

[0026] In some embodiments, the feature extraction employs a sliding time window method to segment the power equipment operation data, extracting statistical feature parameters, including mean, variance, maximum and minimum values, within each time window, and constructing a feature vector by combining the time series change trend.

[0027] In some embodiments, constructing a similarity matrix based on the extracted features includes: Based on the extracted features, the cosine similarity between the features of any two target power devices is calculated, and a similarity matrix is ​​constructed based on the cosine similarity.

[0028] In some embodiments, based on the local data distribution characteristics of each client, feature descriptors that can characterize the statistical properties of the data are extracted. By measuring the differences in data distribution between different clients, the similarity between clients is calculated, and a similarity matrix is ​​constructed. This matrix is ​​used to characterize the data association relationships between various power equipment nodes. The similarity calculation uses a Gaussian kernel function to measure the differences in data distribution between clients. Its calculation form is an exponential decay function based on Euclidean distance. By setting a scale parameter to control the degree of similarity decay, a continuously differentiable similarity matrix is ​​constructed to characterize the nonlinear association relationships between different power equipment.

[0029] In some embodiments, the similarity matrix is ​​clustered using a spectral clustering algorithm to obtain multiple clustering results, including: The similarity matrix is ​​decomposed using a spectral clustering algorithm. The eigenvectors corresponding to the top K largest eigenvalues ​​are selected to form a new feature space, where K is the preset number of clusters. The target power equipment is clustered using the K-means clustering algorithm in the new feature space to obtain multiple clustering results.

[0030] In some embodiments, an undirected weighted graph is constructed based on a similarity matrix, where nodes represent clients and edge weights reflect the degree of similarity between clients. Clients are then partitioned using graph clustering methods to obtain... The goal is to divide clients into client groups, ensuring high data distribution consistency within each group. This is achieved by minimizing the following objective function, resulting in a partitioning effect where clients within groups are closely connected and clients between groups show significant differences: .

[0031] In some embodiments, an intra-group consistency regularization term is introduced to constrain the differences between client model parameters within the same group, thereby enhancing the consistency of the intra-group model: .

[0032] In some embodiments, a joint optimization objective is obtained, and clients are assigned to corresponding groups. This ensures that the grouping results satisfy both the requirements for graph structure partitioning and the collaborative optimization capability of model parameters.

[0033] Among them, the number of groups Alternatively, a similarity threshold can be set as a preset parameter. This step is used to reduce the impact of differences in data distribution, and the output grouping results are used to construct the subsequent hierarchical training structure.

[0034] In some embodiments, the graph clustering method employs a spectral clustering algorithm, which constructs the Laplacian matrix of the graph, decomposes its eigenvalues, selects the top k eigenvectors to form a low-dimensional embedding space, and performs clustering partitioning in this space to achieve grouping of clients.

[0035] In some embodiments, the joint optimization objective is solved by an iterative optimization method. In each iteration, the grouping variables are fixed and the model parameters are updated. Then, the grouping results are updated by fixing the model parameters. The optimal solution is gradually approached through an alternating optimization method until the objective function changes tend to stabilize.

[0036] In some embodiments, the training process of the pre-built target model includes the following steps: A randomly initialized regression prediction model is generated on the central server, and the central server sends the initialized regression prediction model to the client. N clients train simultaneously. On each client participating in the training, the client's local model is initialized with the regression prediction model transmitted from the central server. The extracted features are used to train the client's regression prediction model. The training is iterated until the preset local training rounds M or the error threshold condition is met. The trained regression prediction model is then uploaded to the group's server, and the regression prediction model parameters are updated using the stochastic gradient descent algorithm. After training is completed on N clients, the updated regression prediction model parameters are uploaded to the group's server. After receiving the updated regression prediction model parameters, the group server performs weighted aggregation on the trained regression prediction model based on the amount of client data to obtain a group-level model. The group-level server uploads the group-level model to the central server. After receiving the group-level model, the central server performs synchronous aggregation on the group-level model using a weighted average method to generate a target model. The central server then sends the target model to the client.

[0037] In some embodiments, asynchronous training and model updates within a group are implemented: Within each client group, each client independently trains the model based on local data and updates the model parameters according to a preset number of training epochs or local iterations. After completing local training, each client uploads the model parameters to the group's server without waiting for other clients to complete their updates. .

[0038] In some embodiments, the number of local training rounds, learning rate, and batch size are preset parameters; each client needs to have model training and communication capabilities; it is used to improve training efficiency, and its output serves as the input for model aggregation within the group.

[0039] In some embodiments, the model employs a neural network-based regression prediction model. Model parameters are updated using a stochastic gradient descent algorithm. During each round of local training, gradients are calculated based on mini-batch samples, and parameter updates are performed to progressively minimize the local loss function. Time decay weights are calculated based on the number of delayed iterations of the model update; the larger the delay, the smaller the corresponding weight, thus ensuring that the most recently uploaded model update carries a higher weight in the aggregation process, reducing the interference of outdated historical information on the current model update.

[0040] In some embodiments, intra-group model aggregation: the intra-group server performs weighted aggregation of each client model based on the amount of client data and weight information to obtain the shared model parameters for the group. This process enables information fusion and model optimization within the group, improving the model's adaptability to similar data distributions. .

[0041] In some embodiments, the group server has received client model updates and uses them to generate a group-level model, the output of which serves as input for global aggregation. The weighting coefficients are calculated based on the number of samples contained in the client; clients with a larger number of samples have higher weights during aggregation, ensuring that nodes with larger data volumes contribute more fully to model updates.

[0042] In some embodiments, global model synchronization and aggregation: after completing aggregation within their respective groups, each server uploads its group model to the global server, which then performs unified synchronization and aggregation to generate global model parameters. This process ensures effective fusion of information between different client groups while maintaining the consistency and stability of model updates. .

[0043] In some embodiments, each group completes a model update to generate a globally shared model, the results of which are then distributed to each client for the next round of training. The global aggregation uses a weighted average method, weighting and fusing the model parameters of each group according to the data size within each group, thereby achieving information integration on a global scale.

[0044] In some embodiments, convergence determination involves detecting changes in model parameters after each training round to determine if a preset convergence condition is met. If the convergence requirement is not met, the process returns to the asynchronous training step within the group to continue iterative updates until the model converges or reaches the maximum number of training rounds. The convergence threshold or maximum number of iterations is a preset parameter. This step controls the training termination condition and forms a cyclic iterative relationship. The convergence condition is determined by calculating the difference between model parameters in two consecutive rounds. When the change in model parameters is lower than the preset threshold, the model is considered to have reached a convergent state.

[0045] In some embodiments, the lifespan prediction output is as follows: After model training is completed, the global model is distributed to each power equipment node. Each node uses the updated model to analyze the local equipment's operating status and outputs the predicted remaining lifespan of the equipment, thus achieving dynamic evaluation and prediction of the power equipment. This step, using the converged model as input, is used to output the final prediction result and is the target implementation stage of the entire method. The lifespan prediction result is represented by continuous numerical values ​​output by the model and, combined with a preset equipment lifespan mapping relationship, transforms the model output into the remaining lifespan of the equipment.

[0046] According to the embodiments of this disclosure, the following technical effects are achieved: collaborative utilization of power equipment operation data from multiple regions is realized without the data leaving the local area, improving the model's adaptability to heterogeneous data; a hierarchical federated learning framework with a two-level structure within and between groups is constructed; an asynchronous update mechanism is adopted within groups, and a synchronous aggregation method is adopted between groups, reducing node waiting time while ensuring the consistency of the global model; enhancing the collaborative learning capability within groups; improving lifetime prediction accuracy; and improving the stability of the model convergence process.

[0047] It should be noted that, for the sake of simplicity, the foregoing method embodiments are all described as a series of actions. However, those skilled in the art should understand that this disclosure is not limited to the described order of actions, because according to this disclosure, some steps can be performed in other orders or simultaneously. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are all optional embodiments, and the actions and modules involved are not necessarily essential to this disclosure.

[0048] The above is an introduction to the method embodiments. The following describes the solution described in this disclosure further through device embodiments.

[0049] Figure 2 A block diagram of a power equipment lifetime prediction device 200 based on a hierarchical asynchronous mechanism according to an embodiment of the present disclosure is shown.

[0050] like Figure 2 As shown, the device 200 includes: The acquisition module 201 is used to acquire the operating data of the target power equipment, and to preprocess and extract features from the operating data; Module 202 is used to construct a similarity matrix based on the extracted features; Clustering module 203 is used to cluster the similarity matrix using a spectral clustering algorithm to obtain multiple clustering results; The output module 204 is used to input each clustering result into the pre-built target model and output the remaining service life of the target power equipment.

[0051] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working process of the described module can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0052] The acquisition, storage, and application of user personal information involved in the technical solution disclosed herein comply with the provisions of relevant laws and regulations and do not violate public order and good morals.

[0053] According to embodiments of this disclosure, this disclosure also provides an electronic device, a readable storage medium, and a computer program product.

[0054] Figure 3 A schematic block diagram of an electronic device 300 that can be used to implement embodiments of the present disclosure is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device may also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the present disclosure described and / or claimed herein.

[0055] Electronic device 300 includes a computing unit 301, which can perform various appropriate actions and processes based on a computer program stored in read-only memory (ROM) 302 or a computer program loaded from storage unit 308 into random access memory (RAM) 303. The RAM 303 may also store various programs and data required for the operation of device 300. The computing unit 301, ROM 302, and RAM 303 are interconnected via bus 304. An input / output (I / O) interface 305 is also connected to bus 304.

[0056] Multiple components in electronic device 300 are connected to I / O interface 305, including: input unit 306, such as keyboard, mouse, etc.; output unit 307, such as various types of displays, speakers, etc.; storage unit 308, such as disk, optical disk, etc.; and communication unit 309, such as network card, modem, wireless transceiver, etc. Communication unit 309 allows device 300 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0057] The computing unit 301 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of the computing unit 301 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. The computing unit 301 performs the various methods and processes described above, such as method 100. For example, in some embodiments, method 100 may be implemented as a computer software program tangibly contained in a machine-readable medium, such as storage unit 308. In some embodiments, part or all of the computer program may be loaded and / or installed on the electronic device 300 via ROM 302 and / or communication unit 309. When the computer program is loaded into RAM 303 and executed by the computing unit 301, one or more steps of method 300 described above may be performed. Alternatively, in other embodiments, the computing unit 301 may be configured to perform method 300 by any other suitable means (e.g., by means of firmware).

[0058] Various embodiments of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), systems-on-a-chip (SoCs), payload-programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transmitting data and instructions to the storage system, the at least one input device, and the at least one output device.

[0059] The program code used to implement the methods of this disclosure may be written in any combination of one or more programming languages. This program code may be provided to a processor or controller of a general-purpose computer, special-purpose computer, or other programmable data processing apparatus, such that when executed by the processor or controller, the program code causes the functions / operations specified in the flowcharts and / or block diagrams to be implemented. The program code may be executed entirely on a machine, partially on a machine, as a standalone software package partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0060] In the context of this disclosure, a machine-readable medium can be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, apparatus, or device. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium can be, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0061] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device for displaying information to the user (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor); and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the computer. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).

[0062] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as a data server), or computing systems that include middleware components (e.g., an application server), or computing systems that include frontend components (e.g., a user computer with a graphical user interface or web browser through which a user can interact with implementations of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., a communication network). Examples of communication networks include local area networks (LANs), wide area networks (WANs), and the Internet.

[0063] Computer systems can include clients and servers. Clients and servers are generally located far apart and typically interact via communication networks. Client-server relationships are created by computer programs running on the respective computers and having a client-server relationship with each other. Servers can be cloud servers, servers in distributed systems, or servers incorporating blockchain technology.

[0064] It should be understood that the various forms of processes shown above can be used to rearrange, add, or delete steps. For example, the steps described in this disclosure can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution disclosed in this disclosure can be achieved, and this is not limited herein.

[0065] The specific embodiments described above do not constitute a limitation on the scope of protection of this disclosure. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this disclosure should be included within the scope of protection of this disclosure.

Claims

1. A method for predicting the lifetime of power equipment based on a hierarchical asynchronous mechanism, comprising: The operating data of the target power equipment is acquired, and the operating data is preprocessed and features are extracted. Based on the extracted features, a similarity matrix is ​​constructed; The similarity matrix was clustered using a spectral clustering algorithm to obtain multiple clustering results. Each clustering result is input into a pre-built target model, which outputs the remaining lifespan of the target power equipment.

2. The method according to claim 1, characterized in that, The operational data includes: Voltage, current, temperature, load conditions, and operating status.

3. The method according to claim 1, characterized in that, The preprocessing of the operational data includes: The running data is processed by outlier removal, missing value completion, and data normalization.

4. The method according to claim 1, characterized in that, The feature extraction includes: The preprocessed running data is segmented using a sliding time window method, and statistical feature parameters are extracted from the segmented running data. The statistical feature parameters include mean, variance, maximum value and minimum value. Features are generated based on the statistical feature parameters.

5. The method according to claim 1, characterized in that, The construction of a similarity matrix based on the extracted features includes: Based on the extracted features, the cosine similarity between the features of any two target power devices is calculated, and a similarity matrix is ​​constructed based on the cosine similarity.

6. The method according to claim 1, characterized in that, The similarity matrix is ​​clustered using a spectral clustering algorithm to obtain multiple clustering results, including: The similarity matrix is ​​decomposed using a spectral clustering algorithm. The eigenvectors corresponding to the top K largest eigenvalues ​​are selected to form a new feature space, where K is the preset number of clusters. The target power equipment is clustered using the K-means clustering algorithm in the new feature space to obtain multiple clustering results.

7. The method according to claim 1, characterized in that, The training process of the pre-built target model includes the following steps: A randomly initialized regression prediction model is generated on the central server, and the central server sends the initialized regression prediction model to the client. N clients train simultaneously. On each client participating in the training, the client's local model is initialized with the regression prediction model transmitted from the central server. The extracted features are used to train the client's regression prediction model. The training is iterated until the preset local training rounds M or the error threshold condition is met. The trained regression prediction model is then uploaded to the group's server, and the regression prediction model parameters are updated using the stochastic gradient descent algorithm. After training is completed on N clients, the updated regression prediction model parameters are uploaded to the group's server. After receiving the updated regression prediction model parameters, the group server performs weighted aggregation on the trained regression prediction model based on the amount of client data to obtain a group-level model. The group-level server uploads the group-level model to the central server. After receiving the group-level model, the central server performs synchronous aggregation on the group-level model using a weighted average method to generate a target model. The central server then sends the target model to the client.

8. A power equipment lifetime prediction device based on a hierarchical asynchronous mechanism, comprising: The acquisition module is used to acquire the operating data of the target power equipment, and to preprocess and extract features from the operating data; The building block is used to construct a similarity matrix based on the extracted features; The clustering module is used to cluster the similarity matrix using a spectral clustering algorithm to obtain multiple clustering results. The output module is used to input each clustering result into the pre-built target model and output the remaining service life of the target power equipment.

9. An electronic device, comprising: At least one processor; as well as A memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform the method of any one of claims 1-7.

10. A non-transitory computer-readable storage medium storing computer instructions, wherein, The computer instructions are used to cause the computer to perform the method according to any one of claims 1-7.

Images