Power equipment fault diagnosis system based on deep learning and related equipment

Abstract

The invention discloses a power equipment fault diagnosis system based on deep learning and related equipment, and relates to the technical field of power equipment operation and maintenance and intelligent diagnosis. Comprising the steps of data acquisition, data preprocessing, deep learning model training and feature extraction, health scoring and fault diagnosis, fault data generation and enhancement, federal learning model optimization, online diagnosis and early warning and the like. Equipment operation data is collected through a multi-modal sensor, data features are extracted by using a deep learning model and fault analysis is carried out, a fault sample set is expanded in combination with a generative adversarial network, and the accuracy and robustness of diagnosis are improved; meanwhile, a global model is optimized through a federated learning framework, data privacy is protected, and distributed monitoring and diagnosis of equipment are achieved; the method can accurately identify the fault type, dynamically evaluate the health condition of the equipment and quickly generate an alarm signal and a maintenance suggestion, and is widely applied to intelligent operation and maintenance scenes of the power equipment.

Description

Technical Field

[0001] The present invention belongs to the technical field of operation and maintenance and intelligent diagnosis of power equipment, and particularly relates to a power equipment fault diagnosis system and related equipment based on deep learning. Background Art

[0002] With the continuous expansion of the scale of modern power systems, the stable operation and reliability of power equipment are crucial for the safe operation of the power grid. Power equipment usually includes transformers, switch cabinets, transmission lines, etc. These devices are prone to being affected by various factors such as mechanical wear, insulation aging, and changes in environmental conditions during long-term operation, thus resulting in failures of varying degrees. To ensure the reliability and safety of the power grid, it is necessary to monitor these devices in real time and diagnose faults in order to take maintenance measures in a timely manner.

[0003] Currently, power equipment fault diagnosis mainly relies on traditional rule-driven models or analysis methods based on expert experience. These methods usually judge the device status by setting fixed thresholds or rules and combine historical fault cases for diagnosis. However, traditional methods have the following deficiencies: on the one hand, the ability to comprehensively utilize multi-modal data by rules and thresholds is weak, and the correlation relationships between complex data cannot be effectively processed; on the other hand, the results of fault diagnosis largely rely on expert experience, lacking standardization and generality. In addition, the problem of scarce fault samples is particularly obvious in actual scenarios, affecting the generalization ability of diagnostic methods.

[0004] A major technical problem that urgently needs to be solved in the prior art is how to make full use of the multi-modal data characteristics during the operation of power equipment to improve the accuracy and robustness of fault diagnosis. Especially in the case where equipment is widely distributed and the data acquisition environment is complex, a more intelligent method is needed that can extract key features from multi-dimensional data, comprehensively analyze the device status, accurately identify the fault type, and evaluate the health status. Summary of the Invention

[0005] The purpose of the present invention is to provide a power equipment fault diagnosis system and related equipment based on deep learning to solve the technical problems of insufficient utilization of multi-modal data, low diagnostic accuracy, and scarce fault samples in existing power equipment fault diagnosis methods.

[0006] To achieve the above object, the present invention adopts the following technical solutions: A power equipment fault diagnosis method based on deep learning, comprising the following steps: Obtain multi-modal operation data of power equipment and perform preprocessing; Input the preprocessed multi-modal operation data of power equipment into a fault diagnosis model to generate a diagnosis result; Calculate the health score of the power equipment, identify the fault type, and evaluate the health status of the power equipment according to the fault diagnosis results; The fault diagnosis model is a combined model based on a convolutional neural network, a recurrent neural network, and a graph neural network; Based on a distributed architecture, optimize the fault diagnosis model through federated learning, and use a generative adversarial network to generate realistic fault data, supplement scarce fault samples, and expand the training dataset of the fault diagnosis model.

[0007] Furthermore, the multimodal operation data includes vibration signals, temperature signals, current signals, and partial discharge signals.

[0008] Furthermore, the process of inputting the preprocessed multimodal operation data of the power equipment into the fault diagnosis model to generate a diagnosis result is as follows: Use a convolutional neural network to extract spatio-temporal features from the preprocessed multimodal operation data of the power equipment, a recurrent neural network to capture time series dynamic features, and a graph neural network to model the topological relationship of the power equipment; The fault diagnosis model conducts diagnostic analysis based on spatio-temporal features, time series dynamic features, and the topological relationship of the power equipment, and generates a diagnosis result.

[0009] Furthermore, the graph neural network uses a graph convolutional network, and the topological relationship includes the connection relationship between power equipment and the fault propagation path; The formula for the feature propagation mechanism of the graph convolutional network is:

[0010] where is the node feature of the l-th layer, A is the adjacency matrix, D is the node degree matrix, is the weight matrix of the l-th layer, is the activation function.

[0011] Furthermore, the steps for optimizing the fault diagnosis model through federated learning based on the distributed architecture are as follows: Based on the distributed architecture, upload the parameters of the local training model of each device site to the central server; The central server aggregates the local models through a weighted average algorithm to generate a global optimization model, and uses the global optimization model to optimize the fault diagnosis model.

[0012] Furthermore, the formula for the health score is:

[0013] where is the health score at time t, is the initial health score, is the equipment deterioration rate, is the environmental correction factor.

[0014] Furthermore, when the health score is lower than the set threshold, an alarm signal is triggered, and a diagnostic report including the fault type, cause analysis, and maintenance suggestions is generated.

[0015] In a second aspect, the present invention provides a power equipment fault diagnosis system based on deep learning, including an acquisition module, a diagnosis module, and an evaluation module, where: The acquisition module: is used to acquire multi-modal operation data of power equipment and perform preprocessing; The diagnosis module: is used to input the preprocessed multi-modal operation data of power equipment into the fault diagnosis model to generate a diagnosis result; The evaluation module: is used to calculate the health score of power equipment, identify the fault type, and evaluate the health status of power equipment according to the fault diagnosis result; The fault diagnosis model is a combined model based on a convolutional neural network, a recurrent neural network, and a graph neural network; Based on a distributed architecture, the fault diagnosis model is optimized through federated learning, and a generative adversarial network is used to generate realistic fault data to supplement scarce fault samples and expand the training data set of the fault diagnosis model.

[0016] In a third aspect, a terminal device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the steps of the above method are implemented.

[0017] In a fourth aspect, a computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the above method are implemented.

[0018] Compared with the prior art, the present invention has the following beneficial technical effects: A power equipment fault diagnosis method based on deep learning according to the present invention performs feature extraction and diagnostic analysis on preprocessed multi-modal data through a combined model of a convolutional neural network, a recurrent neural network, and a graph neural network, fully excavating the spatio-temporal features, dynamic change characteristics, and association information between devices of the multi-modal data, and being able to efficiently process large-scale multi-modal data, thereby improving the accuracy and robustness of fault diagnosis; using a generative adversarial network to expand fault sample data, alleviating the problem of scarce data of rare fault types, and enhancing the generalization ability of the model; based on a distributed model training mechanism of federated learning, realizing the dynamic optimization of the global model; in summary, the present invention can effectively solve the problems of insufficient utilization of multi-modal data, low diagnostic accuracy, and scarce fault samples existing in traditional methods.

[0019] Preferably, the connection relationship and fault propagation path between power equipment are modeled by a graph convolutional network, clarifying the propagation mode of associated faults between power equipment.

[0020] Preferably, during the federated learning process, differential privacy technology is used to encrypt the parameters to ensure that the sensitive data of the device sites will not be leaked.

[0021] Preferably, through the real-time health scoring and diagnosis of fault types, the present invention can quickly trigger an alarm signal and generate a diagnostic report, providing efficient and reliable technical support for the operation and maintenance of power equipment. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 It is a flowchart of a power equipment fault diagnosis method based on deep learning in an embodiment of the present invention; Figure 2 It is a flowchart of a power equipment fault diagnosis method based on deep learning in another embodiment of the present invention; Figure 3 It is a schematic structural diagram of a power equipment fault diagnosis system based on deep learning in an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0023] In order to enable those skilled in the art to better understand the solution of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0024] It should be noted that: similar reference numerals and letters denote similar items in the following drawings. Therefore, once an item is defined in one drawing, it does not need to be further defined and explained in subsequent drawings.

[0025] It should be noted that the terms "first", "second", etc. in the description, claims and above-mentioned drawings of the present invention are used to distinguish similar objects, and do not necessarily describe a specific order or sequence. It should be understood that such data used can be interchanged under appropriate circumstances, so that the embodiments of the present invention described here can be implemented in an order other than those illustrated or described here. In addition, the terms "comprising" and "having" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device comprising a series of steps or units does not necessarily limit to those steps or units clearly listed, but may include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0026] The present invention will be further described in detail below with reference to the accompanying drawings: As Figure 1 shown, a power equipment fault diagnosis method based on deep learning includes the following steps: Step 1, obtain multi-modal operation data of power equipment and perform preprocessing; The above steps are specifically divided into data acquisition and preprocessing. Among them, data acquisition is to deploy a variety of sensors at key parts of power equipment, including vibration sensors, temperature sensors, current transformers and partial discharge signal collectors. The sampling frequency range is 10 Hz to 10 kHz, which is used to collect multi-modal data of the equipment operation state. These data include vibration signals for analyzing the mechanical state of the equipment, temperature signals for monitoring the thermal state of the equipment, current signals reflecting the electrical load condition and partial discharge signals for evaluating the insulation performance. The above data are transmitted to the data processing module through the industrial field bus or wireless network to ensure the real-time and integrity of the transmission.

[0027] Data preprocessing is to clean, time-synchronize, noise-remove and normalize the multi-modal operation data to generate standardized high-quality data, where: Time synchronization: The timestamps of different sensors may have deviations. Through master-slave clock synchronization or GPS signals, multi-source data are aligned to a unified time axis to ensure the consistency between different types of data; Noise removal: Environmental noise or interference signals may be mixed in the data acquisition process. For vibration signals, filters and empirical mode decomposition techniques are used to separate the noise components and extract effective signal features. The processed signals are used for subsequent feature extraction; Normalization processing is to eliminate the amplitude differences of different types of data, and the data are standardized by using the normalization method. Specifically, by calculating the mean and standard deviation of each type of data, the data amplitude is converted into a zero-mean and unit-variance distribution to meet the input requirements of the deep learning model.

[0028] Step 2: Input the preprocessed multimodal operation data of the power equipment into the fault diagnosis model to generate a diagnosis result; The fault diagnosis model consists of a convolutional neural network (CNN), a recurrent neural network (RNN), and a graph neural network (GNN). Specifically: CNN: Extract the spatio-temporal features of the signal. The CNN uses a multi-layer convolutional structure to extract local features in the vibration signal and partial discharge signal through convolutional kernels. The sliding operation of the convolutional kernel transforms the input signal into a high-dimensional feature representation. This feature extraction process can effectively extract the edge and frequency characteristics of the signal without specific formula description; RNN: Capture the dynamic features of the time series. The time series data is input into a bidirectional long short-term memory network (Bi-LSTM) to capture the time dependence of the equipment operation data. The RNN forms a comprehensive feature representation by processing the front and back correlation characteristics; GNN: Model the topological relationship between power equipment. Specifically, the graph neural network is used to model based on the topological relationship of the equipment. The topological relationship includes the connection relationship between equipment and the fault propagation path. Among them, the formula for the feature propagation mechanism of the graph convolutional network is:

[0029] Among them, is the node feature of the l-th layer, A is the adjacency matrix, D is the node degree matrix, is the weight matrix of the l-th layer, is the activation function.

[0030] Step 3: According to the fault diagnosis result, calculate the health score of the power equipment, identify the fault type, and evaluate the health status of the power equipment; The health score and fault classification tasks are jointly optimized through a multi-task learning framework. Each task uses an independent loss function and optimizes the overall objective through a weighted method. During the training process, the overall loss is reduced by using the accuracy of fault classification and the deviation of the health score; The formula for calculating the health score is:

[0031] Among them, is the health score at time t, is the initial health score, is the equipment deterioration rate, is the environmental correction factor; This scoring mechanism comprehensively considers the operation characteristics of the equipment and the influence of the external environment, and is an important basis for evaluating the health status of the equipment.

[0032] In addition, in actual working conditions, fault data is scarce, especially for rare and severe faults. The unbalanced data distribution may cause the model training effect to be biased towards normal data. The present invention uses a generative adversarial network (GAN) to generate realistic fault signals to expand the dataset. The GAN includes a generator and a discriminator. The generator deceives the discriminator through an optimization objective to make the generated signal distribution close to the real fault signal. Its objective function is:

[0033] where z represents a random noise vector, G represents the generator, and D represents the discriminator; Use the generative adversarial network to generate realistic fault data, supplement scarce fault samples, and expand the training dataset of the fault diagnosis model.

[0034] The present invention uses a federated learning optimization model to optimize the fault diagnosis model. Specifically, a deep learning model is locally trained at each device site, and the model parameters are uploaded to the central server. The central server aggregates the local models through a weighted average algorithm (such as FedAvg) to generate a globally optimized model, and uses the globally optimized model to optimize the fault diagnosis model; During the upload process, differential privacy technology is used to encrypt the parameters to ensure that the sensitive data of the device site will not be leaked.

[0035] In an optional implementation, as Figure 2 shown, it further includes an online diagnosis and real-time alarm step. Specifically, when the health score is lower than the set threshold, an alarm signal is triggered, and a diagnostic report including the fault type, cause analysis, and maintenance suggestions is generated.

[0036] In another embodiment of the present invention, a power equipment fault diagnosis system based on deep learning is provided, including an acquisition module, a diagnosis module, and an evaluation module, where: The acquisition module: is used to acquire multi-modal operation data of power equipment and perform preprocessing; The diagnosis module: is used to input the preprocessed multi-modal operation data of power equipment into the fault diagnosis model to generate a diagnosis result; The evaluation module: is used to calculate the health score of power equipment, identify the fault type, and evaluate the health status of power equipment according to the fault diagnosis result; The fault diagnosis model is a combined model based on a convolutional neural network, a recurrent neural network, and a graph neural network; Based on a distributed architecture, the fault diagnosis model is optimized through federated learning, and a generative adversarial network is used to generate realistic fault data to supplement scarce fault samples and expand the training dataset of the fault diagnosis model.

[0037] Optionally, as Figure 3As shown in the figure, the system includes a data acquisition module, a data preprocessing module, a deep learning diagnosis module, a fault data generation module, a federated learning module, and an online diagnosis and warning module. These modules are interconnected by data streams and control signals to form a closed-loop system architecture.

[0038] Specifically, the data acquisition module collects multimodal operation data of power equipment through sensors and transmits it to the data preprocessing module; After cleaning and standardizing the data, the data preprocessing module outputs high-quality feature data to the deep learning diagnosis module; The deep learning diagnosis module uses deep learning algorithms to complete fault type identification and health score calculation, and its diagnosis results are transmitted to the online diagnosis and warning module. At the same time, the feature learning results of the diagnosis module are also used by the fault data generation module to generate new fault samples and optimize the global model through the federated learning module; The federated learning module aggregates local models of multiple device sites through a distributed architecture to form an optimized global model, which is fed back to the deep learning diagnosis module for updating. The online diagnosis and warning module combines the diagnosis results and real-time health scores to provide alarm signals and maintenance suggestions for the equipment; Specifically, the functions of each module and the system implementation are as follows: 1. Data acquisition module The data acquisition module is the front end of the entire system and is used to obtain multimodal operation data of power equipment, including vibration signals, temperature signals, current signals, and partial discharge signals. This module comprehensively monitors the mechanical, thermal, electrical, and insulation states of the equipment through various types of sensors deployed at key parts of the equipment, such as vibration sensors, temperature sensors, current transformers, and partial discharge signal collectors. The collected data is sent to the data preprocessing module through an industrial field bus (such as CAN bus or Modbus protocol) or a wireless transmission network (such as LoRa or 5G). The module design supports high-frequency sampling from 10Hz to 10kHz to ensure the real-time and accuracy of the collected data.

[0039] 2. Data preprocessing module The data preprocessing module receives multimodal data from the data acquisition module, cleans, aligns, and normalizes the data, and outputs standardized data for subsequent analysis. This module first ensures the consistency of multi-source data on the time axis through a time synchronization mechanism (such as master-slave clock protocol or GPS time synchronization). Then, the module uses filtering techniques and empirical mode decomposition methods to remove high-frequency noise in vibration signals and partial discharge signals. Finally, the data is normalized by calculating the mean and standard deviation to eliminate amplitude differences and enhance the stability of model processing. The preprocessed data is transmitted to the deep learning diagnosis module in real time to ensure the quality and consistency of the model input data.

[0040] 3. Deep Learning Diagnosis Module The deep learning diagnosis module is the core of the system, responsible for feature extraction and fault analysis of the preprocessed standardized data. The module uses a convolutional neural network (CNN) to extract spatio-temporal features of vibration signals and partial discharge signals, captures the dynamic changes of time series data through a recurrent neural network (RNN), and simultaneously uses a graph neural network (GNN) to model the topological relationships between power equipment and analyze the associated fault propagation patterns between devices. The module outputs the health score and fault type diagnosis results of the device for use by the online diagnosis and warning module, and at the same time provides training parameters and feature representations for the fault data generation module and the federated learning module. The high-performance operation of the module depends on GPU or TPU hardware support and can efficiently process large-scale multimodal data.

[0041] 4. Fault Data Generation Module The fault data generation module expands the training dataset of the system through a generative adversarial network (GAN). The module uses the feature distribution learned by the deep learning diagnosis module to generate realistic fault signals to supplement scarce fault sample data. The generator optimizes the objective to deceive the discriminator so that the generated samples are close to real fault data in distribution, thereby enhancing the system's ability to identify rare faults. After the generated fault data is verified by experts, it can be directly fed back to the deep learning diagnosis module for model training to improve the generalization ability of the model.

[0042] 5. Federated Learning Module The federated learning module realizes the optimization of the global model through a distributed architecture. Each device site trains a deep learning model locally and uploads the trained model parameters to the central server. The central server aggregates the local model parameters through a weighted average method (such as the FedAvg algorithm) to generate a globally optimized model and feeds it back to the deep learning diagnosis module to update the model weights. The module uses differential privacy technology to encrypt the uploaded parameters to ensure the privacy and security of device data.

[0043] 6. Online Diagnosis and Warning Module The online diagnosis and warning module receives the output results of the deep learning diagnosis module in real time, dynamically calculates the device health score, and triggers an alarm signal in combination with the score threshold. The health score formula is:

[0044] where, is the health score at time t, is the initial health score, is the device deterioration rate, is the environmental correction factor.

[0045] The module outputs fault types and health status alerts through the alarm generation sub-module, and at the same time generates a detailed diagnostic report, including fault cause analysis, operation risk prediction, and recommended maintenance measures. The diagnostic results can be presented in a visual form through a remote monitoring platform (such as a Web terminal or a mobile terminal), facilitating users to track the device status in real time.

[0046] In summary, through the functional description of each module and the analysis of the interaction relationship, the modular structure of the system realizes the full-process intelligence of data collection, processing, analysis, optimization, and early warning, providing efficient and accurate technical support for the fault diagnosis of complex power equipment.

[0047] Those skilled in the art should understand that the embodiments of the present invention can be provided as a method, a system, or a computer program product. Therefore, the present invention can take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present invention can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0048] The present invention is described with reference to the flowcharts and / or block diagrams of methods, apparatuses (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowcharts and / or block diagrams, and the combination of processes and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing devices generate means for implementing the functions specified in Figure 1 one process or multiple processes and / or blocks Figure 1 one block or multiple blocks.

[0049] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer-readable memory generate a manufactured article including instruction means, and the instruction means implements the functions specified in Figure 1 one process or multiple processes and / or blocks Figure 1 one block or multiple blocks.

[0050] These computer program instructions can also be loaded onto a computer or other programmable data processing device, so that a series of operation steps are executed on the computer or other programmable device to generate a computer-implemented process. Thus, the instructions executed on the computer or other programmable device provide means for implementing the functions specified in Figure 1steps of one or more processes and / or boxes Figure 1 steps of functions specified in one or more boxes.

[0051] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than limit the scope of its protection. Although the present invention has been described in detail with reference to the above embodiments, those of ordinary skill in the art should understand that: after reading the present invention, those skilled in the art can still make various changes, modifications or equivalent replacements to the specific implementation manners of the invention, but these changes, modifications or equivalent replacements are all within the scope of protection of the pending claims of the invention.

Claims

1. A method for fault diagnosis of electric power equipment based on deep learning, characterized in that: The following steps are involved: Acquire multi-modal operation data of power equipment and perform pre-processing; Inputting the preprocessed multi-modal operation data of the power equipment into the fault diagnosis model to generate a diagnosis result; Based on the fault diagnosis results, calculate the power equipment health score, identify the fault type, and evaluate the health status of the power equipment; The fault diagnosis model is a combined model based on convolutional neural network, recurrent neural network and graph neural network; Based on a distributed architecture, the fault diagnosis model is optimized through federated learning, and a generative adversarial network is used to generate realistic fault data to supplement scarce fault samples and expand the training data set of the fault diagnosis model.

2. A method for diagnosing faults of electric power equipment based on deep learning according to claim 1, characterized in that: The multimodal operation data includes a vibration signal, a temperature signal, a current signal and a partial discharge signal.

3. The method for diagnosing faults of electric power equipment based on deep learning according to claim 1, characterized in that: The process of inputting the pre-processed multi-modal operation data of the power equipment into the fault diagnosis model to generate the diagnosis result is as follows: The pre-processed multi-modal operation data of power equipment is used to extract spatiotemporal features using convolutional neural networks, capture dynamic features of time series using recurrent neural networks, and model the topological relationship of power equipment using graph neural networks. The fault diagnosis model performs diagnostic analysis based on spatiotemporal characteristics, time series dynamic characteristics and topological relationships of power equipment to generate diagnostic results.

4. A method for diagnosing faults of electric power equipment based on deep learning according to claim 1 or 3, characterized in that: The graph neural network adopts a graph convolutional network, and the topological relationship includes the connection relationship between power equipment and the fault propagation path; The feature propagation mechanism formula of the graph convolutional network is: in, is the node feature of the first layer, A is the adjacency matrix, D is the node degree matrix, is the weight matrix of the lth layer, is the activation function.

5. The method for diagnosing faults of electric power equipment based on deep learning according to claim 1, characterized in that: The steps of optimizing the fault diagnosis model through federated learning based on the distributed architecture are as follows: Based on a distributed architecture, the parameters are uploaded to the central server through the local training model of each device site; The central server aggregates local models through a weighted average algorithm to generate a global optimization model, and uses the global optimization model to optimize the fault diagnosis model.

6. A method for diagnosing faults of electric power equipment based on deep learning according to claim 1, characterized in that: The health score calculation formula is: in, is the health score at time t, is the initial health score, is the equipment degradation rate, is the environmental correction factor.

7. A method for diagnosing faults of electric power equipment based on deep learning according to claim 1, characterized in that: When the health score is lower than a set threshold, an alarm signal is triggered and a diagnostic report including fault type, cause analysis and maintenance recommendations is generated.

8. A power equipment fault diagnosis system based on deep learning, characterized in that: It includes acquisition module, diagnosis module and evaluation module, among which: Acquisition module: used to acquire multi-modal operation data of power equipment and perform pre-processing; Diagnosis module: used to input the pre-processed multi-modal operation data of the power equipment into the fault diagnosis model to generate the diagnosis results; Evaluation module: used to calculate the health score of power equipment, identify the fault type, and evaluate the health status of power equipment based on the fault diagnosis results; The fault diagnosis model is a combined model based on convolutional neural network, recurrent neural network and graph neural network; Based on a distributed architecture, the fault diagnosis model is optimized through federated learning, and a generative adversarial network is used to generate realistic fault data to supplement scarce fault samples and expand the training data set of the fault diagnosis model.

9. A terminal device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images