Fault diagnosis method and system for power transformer
By normalizing and fusing the operational monitoring data of power transformers, combining the Pelican optimization algorithm and the firefly perturbation mechanism, a generative adversarial network model was constructed, which solved the problem of low accuracy in power transformer fault diagnosis and achieved efficient fault diagnosis and real-time prediction.
Patent Information
- Application Number
- CN202510717425.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-05-30
- Publication Date
- 2025-09-05
AI Technical Summary
Existing power transformer fault diagnosis technology is difficult to quickly capture the state changes of power transformers, resulting in low accuracy of fault diagnosis.
A fusion neural network model is used to normalize and perform feature fusion analysis on the operation monitoring data of power transformers. The Pelican optimization algorithm and the firefly disturbance mechanism are combined to screen data, and a fault diagnosis model based on generative adversarial network technology is constructed. Environmental monitoring data is used to improve the prediction accuracy and generalization ability of the model.
Through normalized processing and feature fusion, the accuracy and completeness of the data are improved, the accuracy and real-time performance of fault diagnosis are enhanced, the prediction accuracy and generalization ability of the model are improved, and the reliability of fault diagnosis is ensured.
Smart Images
Figure CN120596840A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of power system fault diagnosis, and in particular to a power transformer fault diagnosis method and system. Background Art
[0002] Power transformers are key equipment in the power system. The operating status of power transformers directly affects the stability of the entire power system. By monitoring the operating status of power transformers and diagnosing faults, potential problems of transformers can be discovered in a timely manner, the operating mode of the power system can be optimized, the operating efficiency of the power system can be improved, the expansion of faults can be prevented, and the collapse of the power system or large-scale power outages can be avoided.
[0003] However, due to the complex and changeable power environment, the existing power transformer fault diagnosis technology is difficult to quickly capture the state changes of the power transformer, resulting in low accuracy of power transformer fault diagnosis.
[0004] Therefore, how to improve the accuracy of power transformer fault diagnosis has become a technical problem that needs to be solved urgently by those skilled in the art. Summary of the Invention
[0005] The present invention provides a power transformer fault diagnosis method and system to solve the technical problem that the existing power transformer fault diagnosis technology is difficult to quickly capture the state change of the power transformer, resulting in low accuracy in power transformer fault diagnosis.
[0006] In order to solve the above technical problems, an embodiment of the present invention provides a fault diagnosis method for a power transformer.
[0007] Acquire each first operation monitoring data obtained by normalizing each operation monitoring history data of the target power transformer;
[0008] A fusion neural network model is used to perform a fusion analysis on the spatial features of all the extracted first operation monitoring data and the temporal features of all the extracted first operation monitoring data; wherein the fusion analysis is designed to use the Pelican optimization algorithm to screen the first operation monitoring data based on the data correlation between all the first operation monitoring data, and use the firefly perturbation mechanism to introduce random perturbations into the screening process of the first operation monitoring data;
[0009] Based on the fusion analysis results, an operation monitoring sample data set is constructed, and a trained target performance prediction model is obtained using the operation monitoring sample data set and the acquired environmental monitoring data; the second operation monitoring data of the target power transformer collected in real time is input into the target performance prediction model to obtain operation status prediction data of the target power transformer;
[0010] A fault diagnosis model of the target power transformer is constructed based on generative adversarial network technology, and the operating status prediction data is input into the fault diagnosis model to obtain a fault diagnosis log of the target power transformer.
[0011] As one preferred solution, the method further comprises:
[0012] generating a corresponding fault repair strategy according to the fault diagnosis log;
[0013] A simulation model of the target power transformer is constructed using simulation technology, and the fault repair strategy is input into the simulation model to obtain a simulated repair result of the target power transformer;
[0014] The fault repair strategy is modified based on the simulated repair result.
[0015] As one of the preferred solutions, the fault repair strategy at least includes cleaning the faulty part, replacing the faulty component, adjusting the load of the transformer, changing the operating state of the cooling system and adjusting the voltage.
[0016] As one preferred solution, the step of obtaining each first operation monitoring data obtained by normalizing each operation monitoring history data of the target power transformer includes:
[0017] Performing normalization processing on all the operation monitoring historical data, and performing missing value processing on the normalized monitoring data obtained by the normalization processing;
[0018] An outlier detection is performed on the result obtained by the missing value processing to obtain various abnormal historical data, and each of the abnormal historical data is replaced by the average value of all the abnormal historical data in the corresponding neighborhood window to obtain various first operation monitoring data.
[0019] As one preferred solution, the operating state prediction data at least includes voltage state prediction data and temperature state prediction data;
[0020] The step of inputting the operating state prediction data into the fault diagnosis model to obtain a fault diagnosis log of the target power transformer includes:
[0021] Comparing and analyzing the voltage state prediction data and the threshold value obtained by the fault diagnosis model to obtain a voltage diagnosis result of the target power transformer;
[0022] Acquire a temperature diagnosis result of the target power transformer based on the temperature state prediction data and a difference feature of the temperature fluctuation range generated by the fault diagnosis model;
[0023] A fault diagnosis log of the target power transformer is generated according to the voltage diagnosis result and the temperature diagnosis result.
[0024] As one of the preferred solutions, the fused neural network model is obtained by fusing a convolutional neural network model and a long short-term memory network model;
[0025] Among them, the activation function of the convolutional neural network model is the ReLU activation function, and the fusion neural network model uses mean square error as the loss function.
[0026] As one of the preferred schemes, the method also includes performing missing value processing on all the first operation monitoring data before using the fusion neural network model to perform fusion analysis on the spatial features of all the extracted first operation monitoring data and the time features of all the extracted first operation monitoring data, so that the dimensions of all the first operation monitoring data are the same.
[0027] As one of the preferred solutions, the method further includes performing data set augmentation processing on the operation monitoring sample data set before obtaining the trained target performance prediction model using the operation monitoring sample data set and the acquired environmental monitoring data.
[0028] As one of the preferred solutions, the disturbance intensity of the firefly disturbance mechanism is determined by the convergence speed of the screening process.
[0029] Another embodiment of the present invention provides a fault diagnosis system for a power transformer, comprising:
[0030] A data processing module, configured to obtain first operation monitoring data obtained by normalizing the operation monitoring history data of the target power transformer;
[0031] a fusion analysis module, configured to perform fusion analysis on the spatial features and temporal features of all the extracted first operation monitoring data using a fusion neural network model; wherein the fusion analysis is designed to use a Pelican optimization algorithm to filter the first operation monitoring data based on the data correlation between all the first operation monitoring data, and to use a firefly perturbation mechanism to introduce random perturbations into the filtering process of the first operation monitoring data;
[0032] A state prediction module is configured to construct an operation monitoring sample data set based on the fusion analysis results, and obtain a trained target performance prediction model using the operation monitoring sample data set and the acquired environmental monitoring data; input the second operation monitoring data of the target power transformer collected in real time into the target performance prediction model to obtain operation state prediction data of the target power transformer;
[0033] A fault diagnosis module is used to construct a fault diagnosis model of the target power transformer based on generative adversarial network technology, input the operating status prediction data into the fault diagnosis model, and obtain a fault diagnosis log of the target power transformer.
[0034] Compared with the prior art, the embodiments of the present invention have the following advantages:
[0035] Obtain each first operation monitoring data obtained by normalizing each operation monitoring history data of the target power transformer; by normalizing the operation monitoring history data of the target power transformer, it is possible to remove noise in the operation monitoring history data, fill missing values, correct erroneous data, ensure the accuracy and completeness of the data, reduce instability in the subsequent model processing process, and improve the convergence speed and performance of the model; a fusion neural network model is used to perform fusion analysis on the spatial features of all the extracted first operation monitoring data and the temporal features of all the extracted first operation monitoring data; wherein, the fusion analysis is designed to use the Pelican optimization algorithm to screen the first operation monitoring data based on the data correlation between all the first operation monitoring data, and use the firefly perturbation mechanism to introduce random perturbations into the screening process of the first operation monitoring data; through fusion analysis, spatial and temporal information are used at the same time to improve the feature expression ability, the Pelican optimization algorithm is used to screen data based on data correlation, and samples with high correlation are retained, thereby improving the quality of the data set, and random perturbations are introduced through the firefly perturbation mechanism to enhance global search capability, avoid falling into local optimality, construct a more representative operation monitoring sample data set, and provide high-quality input for subsequent model training to improve the convergence speed, thereby improving the accuracy and real-time performance of fault diagnosis; based on the fusion analysis results, an operation monitoring sample data set is constructed, and a trained target performance prediction model is obtained by using the operation monitoring sample data set and the acquired environmental monitoring data. The second operation monitoring data of the target power transformer collected in real time is input into the target performance prediction model to obtain the operation status prediction data of the target power transformer. The introduction of environmental monitoring data reflects the impact of external conditions on the operation status of the transformer, improves the prediction accuracy of the model, and thereby improves the accuracy of fault diagnosis; a fault diagnosis model of the target power transformer is constructed based on the generative adversarial network technology, which improves the generalization ability of the fault diagnosis model. The operation status prediction data is input into the fault diagnosis model to obtain the fault diagnosis log of the target power transformer. Through normalization processing, feature fusion, optimization screening and other steps, a high-quality data set is constructed to provide a reliable foundation for model training and real-time prediction, and comprehensively improve the accuracy of fault diagnosis. BRIEF DESCRIPTION OF THE DRAWINGS
[0036] Figure 1This is a flow chart of a method for diagnosing a fault of a power transformer according to one embodiment of the present invention;
[0037] Figure 2 The present invention is a block diagram of a fault diagnosis system for a power transformer according to one embodiment of the present invention. DETAILED DESCRIPTION
[0038] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of them. The purpose of providing these embodiments is to make the disclosure of the present invention more thorough and comprehensive. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts shall fall within the scope of protection of the present invention.
[0039] In the description of this application, the terms "first," "second," "third," etc. are used for descriptive purposes only and should not be understood to indicate or imply relative importance or implicitly specify the number of the technical features indicated. Therefore, a feature specified as "first," "second," "third," etc. may explicitly or implicitly include one or more of the features. In the description of this application, unless otherwise specified, "plurality" means two or more.
[0040] In the description of this application, it should be noted that, unless otherwise expressly specified and limited, the terms "installed", "connected" and "connected" should be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or an indirect connection through an intermediate medium, or it can be a communication between the two components. The terms "vertical", "horizontal", "left", "right", "up", "down" and similar expressions used herein are for illustrative purposes only, and do not indicate or imply that the device or component referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation on the present invention. The term "and / or" used herein includes any and all combinations of one or more related listed items. For those of ordinary skill in the art, the specific meanings of the above terms in this application can be understood according to specific circumstances.
[0041] In the description of this application, it should be noted that, unless otherwise defined, all technical and scientific terms used in this application have the same meanings as those commonly understood by those skilled in the art. The terms used in this specification are only for the purpose of describing specific embodiments and are not intended to limit the present invention. Those skilled in the art will understand the specific meanings of the above terms in this application in specific circumstances.
[0042] In order to solve the technical problem that the existing power transformer fault diagnosis technology is difficult to quickly capture the state change of the power transformer, resulting in low accuracy of power transformer fault diagnosis, an embodiment of the present invention provides a power transformer fault diagnosis method. For details, please refer to Figure 1 , Figure 1 The flowchart of a method for diagnosing a fault of a power transformer in one embodiment of the present invention is shown. Figure 2 , Figure 2 Shown is a structural block diagram of a fault diagnosis system for a power transformer in one embodiment of the present invention.
[0043] A flowchart of a fault diagnosis method for a power transformer in one embodiment of the present invention includes the following steps S1 to S4, which are specifically as follows:
[0044] Step S1: Acquire various first operation monitoring data obtained by normalizing various operation monitoring history data of a target power transformer.
[0045] It should be noted that the operation monitoring historical data is collected through the integrated sensor module, and the operation monitoring historical data includes at least voltage data, current data, electromagnetic induction data and vibration data. The sensor module includes at least voltage sensor, current sensor, electromagnetic induction sensor and vibration sensor.
[0046] Step S2: Use a fusion neural network model to perform a fusion analysis on the spatial features of all the extracted first operation monitoring data and the temporal features of all the extracted first operation monitoring data; wherein, the fusion analysis is designed to use the Pelican optimization algorithm to screen the first operation monitoring data based on the data correlation between all the first operation monitoring data, and use the firefly perturbation mechanism to introduce random disturbances into the screening process of the first operation monitoring data.
[0047] It should be noted that the fusion analysis is used to fuse the spatial features and corresponding temporal features of the first operation monitoring data through the fusion part of the fusion neural network model to obtain a comprehensive feature representation of the first operation monitoring data, and iteratively train the fusion neural network model using the Pelican optimization algorithm and the firefly perturbation mechanism. The fusion neural network model is obtained by fusing the convolutional neural network model and the long short-term memory network model; wherein, the activation function of the convolutional neural network model is the ReLU activation function, and the fusion neural network model uses the mean square error as the loss function. Random values and chaotic sequences are used to generate the initialization population for the first iterative training, and the reverse differential evolution mechanism and the firefly perturbation mechanism are used to obtain the corresponding initialization population for the next iterative training based on the position and fitness of each global optimal solution obtained in each iterative training.
[0048] Step S3: construct an operation monitoring sample data set based on the fusion analysis results, and obtain a trained target performance prediction model using the operation monitoring sample data set and the acquired environmental monitoring data; input the second operation monitoring data of the target power transformer collected in real time into the target performance prediction model to obtain the operation status prediction data of the target power transformer.
[0049] It should be noted that the environmental monitoring data includes at least temperature data. The first operation monitoring data and the second operation monitoring data are both obtained from the operation monitoring data of the target power transformer. The first operation monitoring data is the operation monitoring data of the target power transformer collected historically, and the second operation monitoring data is the operation monitoring data of the target power transformer collected in real time. By dividing the operation monitoring data of the target power transformer into the first operation monitoring data and the second operation monitoring data, the historically collected operation monitoring data and the real-time collected operation monitoring data are distinguished. A faulty power transformer will generate heat during operation, causing changes in the surrounding environment, such as an increase in temperature. By inputting the environmental monitoring data and the operation monitoring sample data set into the model for training, the reliability of the operation status prediction data acquisition can be improved.
[0050] Step S4: construct a fault diagnosis model of the target power transformer based on the generative adversarial network technology, input the operating status prediction data into the fault diagnosis model, and obtain the fault diagnosis log of the target power transformer.
[0051] It should be noted that the operating status prediction data to be judged is input into the fault diagnosis model constructed by generative adversarial network technology. The generator in the generative adversarial network generates new data samples similar to the operating status prediction data of the target power transformer, and the discriminator conducts a comprehensive evaluation and judgment on the input data and the data generated by the generator to determine whether there is a significant difference in the pattern of the operating status prediction data and the normal operating status data.
[0052] Preferably, the power transformer fault diagnosis method provided in this embodiment further includes:
[0053] Generate a corresponding fault repair strategy based on the fault diagnosis log; use simulation technology to build a simulation model of the target power transformer, input the fault repair strategy into the simulation model, and obtain a simulated repair result of the target power transformer; and modify the fault repair strategy based on the simulated repair result.
[0054] It should be noted that the fault repair strategy at least includes cleaning the fault location, replacing the faulty component, adjusting the load of the transformer, changing the operating state of the cooling system and adjusting the voltage.
[0055] Before using the fusion neural network model to perform fusion analysis on the spatial features of all the extracted first operation monitoring data and the temporal features of all the extracted first operation monitoring data in step S2, missing value processing is performed on all the first operation monitoring data to make the dimensions of all the first operation monitoring data the same.
[0056] It should be noted that due to the different sources of the first operation monitoring data obtained, there may be missing values, resulting in inconsistent data dimensions. The fusion neural network model requires the dimensions of the input data to be consistent, otherwise batch processing and feature fusion cannot be performed. By processing the missing values of the first operation monitoring data, the integrity and consistency of the first operation monitoring data can be ensured, and the input requirements of the fusion neural network model can be met to accurately extract the corresponding spatial features and time features, thereby improving the performance of the target performance prediction model and the fault diagnosis model, and ensuring the accuracy of fault diagnosis.
[0057] Preferably, the power transformer fault diagnosis method provided in this embodiment further includes:
[0058] In step S3, before obtaining a trained target performance prediction model using the operation monitoring sample data set and the acquired environmental monitoring data, a data set amplification process is performed on the operation monitoring sample data set.
[0059] It should be noted that dataset augmentation is a technique that increases the size of a dataset by transforming the original data or generating new data. When data is limited, dataset augmentation can effectively improve the generalization ability of the model and prevent overfitting. Dataset augmentation is a well-known technique and will not be described in detail in this embodiment.
[0060] In one embodiment, the step S1 of obtaining each first operation monitoring data obtained by normalizing each operation monitoring history data of the target power transformer includes:
[0061] All operation monitoring historical data are normalized, and the normalized monitoring data obtained by the normalization processing are processed for missing values; the results obtained by the missing value processing are subjected to outlier detection to obtain various abnormal historical data, and each abnormal historical data is replaced with the average value of all abnormal historical data in the corresponding neighborhood window to obtain various first operation monitoring data.
[0062] It should be noted that normalization scales the data to a specific range or distribution to eliminate the impact of dimension and numerical range differences on model training. Normalization and outlier detection are both well-known technologies and will not be described in detail in this embodiment.
[0063] In this embodiment, the 3σ principle is used to perform outlier detection on the results obtained by missing value processing.
[0064] In one embodiment, the perturbation intensity of the firefly perturbation mechanism in step S2 is determined by the convergence speed of the screening process.
[0065] In one embodiment, the operating state prediction data in step S3 at least includes voltage state prediction data and temperature state prediction data.
[0066] In one embodiment, in step S4, the operating state prediction data is input into the fault diagnosis model to obtain a fault diagnosis log of the target power transformer, including:
[0067] The voltage state prediction data and the threshold value obtained by the fault diagnosis model are compared and analyzed to obtain the voltage diagnosis result of the target power transformer; the temperature diagnosis result of the target power transformer is obtained based on the difference characteristics of the temperature state prediction data and the temperature fluctuation range generated by the fault diagnosis model; and the fault diagnosis log of the target power transformer is generated according to the voltage diagnosis result and the temperature diagnosis result.
[0068] The embodiment of the present invention provides a fault diagnosis method for a power transformer, which has the following advantages compared to the prior art:
[0069] Obtain each first operation monitoring data obtained by normalizing each operation monitoring history data of the target power transformer; by normalizing the operation monitoring history data of the target power transformer, it is possible to remove noise in the operation monitoring history data, fill missing values, correct erroneous data, ensure the accuracy and completeness of the data, reduce instability in the subsequent model processing process, and improve the convergence speed and performance of the model; a fusion neural network model is used to perform fusion analysis on the spatial features of all the extracted first operation monitoring data and the temporal features of all the extracted first operation monitoring data; wherein, the fusion analysis is designed to use the Pelican optimization algorithm to screen the first operation monitoring data based on the data correlation between all the first operation monitoring data, and use the firefly perturbation mechanism to introduce random perturbations into the screening process of the first operation monitoring data; through fusion analysis, spatial and temporal information are used at the same time to improve the feature expression ability, the Pelican optimization algorithm is used to screen data based on data correlation, and samples with high correlation are retained, thereby improving the quality of the data set, and random perturbations are introduced through the firefly perturbation mechanism to enhance global search capability, avoid falling into local optimality, construct a more representative operation monitoring sample data set, provide high-quality input for subsequent model training, improve the convergence speed, and thus improve the accuracy and real-time performance of fault diagnosis; construct an operation monitoring sample data set based on the fusion analysis results, and use the operation monitoring sample data set and the acquired environmental monitoring data to obtain a trained target performance prediction model, and input the second operation monitoring data of the target power transformer collected in real time into the target performance prediction model to obtain the operation status prediction data of the target power transformer. The introduction of environmental monitoring data reflects the impact of external conditions on the operation status of the transformer, improves the prediction accuracy of the model, and thus improves the accuracy of fault diagnosis; construct a fault diagnosis model for the target power transformer based on generative adversarial network technology, improves the generalization ability of the fault diagnosis model, inputs the operation status prediction data into the fault diagnosis model, and obtains the fault diagnosis log of the target power transformer. Through normalization processing, feature fusion, optimization screening and other steps, a high-quality data set is constructed to provide a reliable basis for model training and real-time prediction, and comprehensively improves the accuracy of fault diagnosis.
[0070] Another embodiment of the present invention provides a fault diagnosis system for a power transformer, comprising:
[0071] The data processing module 11 is used to obtain various first operation monitoring data obtained by normalizing various operation monitoring historical data of the target power transformer;
[0072] a fusion analysis module 12, configured to perform fusion analysis on the spatial features and temporal features of all extracted first operation monitoring data using a fusion neural network model; wherein the fusion analysis is designed to use the Pelican optimization algorithm to screen the first operation monitoring data based on the data correlation between all first operation monitoring data, and to use the firefly perturbation mechanism to introduce random perturbations into the screening process of the first operation monitoring data;
[0073] The state prediction module 13 is used to construct an operation monitoring sample data set based on the fusion analysis results, and obtain a trained target performance prediction model using the operation monitoring sample data set and the acquired environmental monitoring data; input the second operation monitoring data of the target power transformer collected in real time into the target performance prediction model to obtain the operation state prediction data of the target power transformer;
[0074] The fault diagnosis module 14 is used to build a fault diagnosis model of the target power transformer based on the generative adversarial network technology, input the operating status prediction data into the fault diagnosis model, and obtain the fault diagnosis log of the target power transformer.
[0075] The above-described embodiments merely illustrate several implementations of the present invention, and while their descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the present invention. It should be noted that a person skilled in the art would be able to make numerous variations and improvements without departing from the spirit of the present invention, all of which fall within the scope of protection of the present invention. Therefore, the scope of protection of the present invention shall be determined by the appended claims.
Claims
1. A fault diagnosis method for a power transformer, characterized in that: The method comprises: Acquire each first operation monitoring data obtained by normalizing each operation monitoring history data of the target power transformer; A fusion neural network model is used to perform a fusion analysis on the spatial features of all the extracted first operation monitoring data and the temporal features of all the extracted first operation monitoring data; wherein the fusion analysis is designed to use the Pelican optimization algorithm to screen the first operation monitoring data based on the data correlation between all the first operation monitoring data, and use the firefly perturbation mechanism to introduce random perturbations into the screening process of the first operation monitoring data; Based on the fusion analysis results, an operation monitoring sample data set is constructed, and a trained target performance prediction model is obtained using the operation monitoring sample data set and the acquired environmental monitoring data; the second operation monitoring data of the target power transformer collected in real time is input into the target performance prediction model to obtain operation status prediction data of the target power transformer; A fault diagnosis model of the target power transformer is constructed based on generative adversarial network technology, and the operating status prediction data is input into the fault diagnosis model to obtain a fault diagnosis log of the target power transformer.
2. A fault diagnosis method for a power transformer according to claim 1, characterized in that: The method further comprises: generating a corresponding fault repair strategy according to the fault diagnosis log; A simulation model of the target power transformer is constructed using simulation technology, and the fault repair strategy is input into the simulation model to obtain a simulated repair result of the target power transformer; The fault repair strategy is modified based on the simulated repair result.
3. A fault diagnosis method for a power transformer according to claim 2, characterized in that: The fault repair strategy at least includes cleaning the fault location, replacing the faulty component, adjusting the load of the transformer, changing the operating state of the cooling system and adjusting the voltage.
4. A fault diagnosis method for a power transformer according to claim 1, characterized in that: The obtaining of each first operation monitoring data obtained by normalizing each operation monitoring history data of the target power transformer includes: Performing normalization processing on all the operation monitoring historical data, and performing missing value processing on the normalized monitoring data obtained by the normalization processing; An outlier detection is performed on the result obtained by the missing value processing to obtain various abnormal historical data, and each of the abnormal historical data is replaced by the average value of all the abnormal historical data in the corresponding neighborhood window to obtain various first operation monitoring data.
5. A fault diagnosis method for a power transformer according to claim 1, characterized in that: The operating state prediction data at least includes voltage state prediction data and temperature state prediction data; The step of inputting the operating state prediction data into the fault diagnosis model to obtain a fault diagnosis log of the target power transformer includes: Comparing and analyzing the voltage state prediction data and the threshold value obtained by the fault diagnosis model to obtain a voltage diagnosis result of the target power transformer; Acquire a temperature diagnosis result of the target power transformer based on the temperature state prediction data and a difference feature of the temperature fluctuation range generated by the fault diagnosis model; A fault diagnosis log of the target power transformer is generated according to the voltage diagnosis result and the temperature diagnosis result.
6. A method for diagnosing a fault of a power transformer according to claim 1, characterized in that: The fused neural network model is obtained by fusing a convolutional neural network model and a long short-term memory network model; Among them, the activation function of the convolutional neural network model is the ReLU activation function, and the fusion neural network model uses mean square error as the loss function.
7. A fault diagnosis method for a power transformer according to claim 1, characterized in that: The method also includes performing missing value processing on all the first operation monitoring data before using the fusion neural network model to perform fusion analysis on the spatial features of all the extracted first operation monitoring data and the time features of all the extracted first operation monitoring data, so that the dimensions of all the first operation monitoring data are the same.
8. A method for diagnosing a fault of a power transformer according to claim 1, characterized in that: The method further includes performing data set augmentation processing on the operation monitoring sample data set before obtaining a trained target performance prediction model using the operation monitoring sample data set and the acquired environmental monitoring data.
9. A method for diagnosing a fault of a power transformer according to claim 1, characterized in that: The perturbation intensity of the firefly perturbation mechanism is determined by the convergence speed of the screening process.
10. A fault diagnosis system for a power transformer, characterized in that: The system comprises: A data processing module, configured to obtain first operation monitoring data obtained by normalizing the operation monitoring history data of the target power transformer; a fusion analysis module, configured to perform fusion analysis on the spatial features and temporal features of all the extracted first operation monitoring data using a fusion neural network model; wherein the fusion analysis is designed to use a Pelican optimization algorithm to filter the first operation monitoring data based on the data correlation between all the first operation monitoring data, and to use a firefly perturbation mechanism to introduce random perturbations into the filtering process of the first operation monitoring data; A state prediction module is configured to construct an operation monitoring sample data set based on the fusion analysis results, and obtain a trained target performance prediction model using the operation monitoring sample data set and the acquired environmental monitoring data; input the second operation monitoring data of the target power transformer collected in real time into the target performance prediction model to obtain operation state prediction data of the target power transformer; A fault diagnosis module is used to construct a fault diagnosis model of the target power transformer based on generative adversarial network technology, input the operating status prediction data into the fault diagnosis model, and obtain a fault diagnosis log of the target power transformer.