Power equipment fault diagnosis method and system based on multi-source data fusion, and medium

Abstract

The invention belongs to the technical field of fault diagnosis, and discloses a multi-source data fusion power equipment fault diagnosis method and system and a medium, and the method comprises the steps: obtaining multi-source data of power equipment; preprocessing the multi-source data of the power equipment to obtain processed multi-source data; performing local diagnosis on the processed multi-source data based on a trained base learner of integrated learning to obtain a plurality of local diagnosis results; fusing the plurality of local diagnosis results based on a trained meta-learner of integrated learning to obtain a fault diagnosis result of the power equipment; and generating fault early warning information based on the fault diagnosis result of the power equipment, and visually displaying the fault early warning information. According to the method, multiple diagnosis results are combined through integrated learning, the over-fitting risk can be reduced, complementarity and relevance among multi-source data can be fully mined, and stability and diagnosis accuracy are improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of fault diagnosis, and in particular relates to a multi-source data fusion method, system and medium for diagnosing faults of electric power equipment. Background Art

[0002] With the rapid development of smart grids, the demand for accurate and real-time fault diagnosis of power equipment is increasing. Power equipment (such as transformers, GIS equipment, and transmission lines) generates multi-source heterogeneous data during operation, including vibration signals, partial discharge waveforms, infrared thermal imaging, and direct gas chromatography (DGA). This data reflects the equipment's status from various perspectives, but also exhibits significant differences: varying timescales (e.g., high-frequency vibration signals versus minute-level temperature data), modal heterogeneity (structured numerical data versus unstructured images / waveforms), and information redundancy and conflict (e.g., partial discharge and overheating can simultaneously indicate mechanical looseness).

[0003] Traditional fault diagnosis methods are mostly based on a single data source or shallow feature fusion (such as threshold judgment and simple weighted average), which makes it difficult to fully explore the complementarity and correlation between multi-source data. In particular, misjudgment is prone to occur when facing complex faults (such as mechanical-electrical coupling faults). Summary of the Invention

[0004] The purpose of the present invention is to provide a multi-source data fusion power equipment fault diagnosis method, system and medium to solve the problem that traditional fault diagnosis methods are mostly based on a single data source or shallow feature fusion, which makes it difficult to fully explore the complementarity and correlation between multi-source data, especially when facing complex faults, and are prone to misjudgment.

[0005] In order to achieve the above object, the present invention adopts the following technical solutions: In a first aspect, the present invention provides a method for diagnosing faults in electric power equipment using multi-source data fusion, the method comprising: Acquire multi-source data of power equipment; Preprocessing multi-source data of power equipment to obtain processed multi-source data; Based on the trained ensemble learning base learner, local diagnosis is performed on the processed multi-source data to obtain multiple local diagnosis results; The meta-learner based on the trained ensemble learning is used to fuse multiple local diagnosis results to obtain the fault diagnosis results of the power equipment; Based on the fault diagnosis results of the power equipment, fault warning information is generated and displayed visually.

[0006] Preferably, the trained base learner of the ensemble learning comprises: a plurality of diagnosis modules connected in series, each diagnosis module comprising: an input layer, a classification layer and an enhancement layer; The input layer, classification layer, and enhancement layer in the i-th diagnostic module are connected in sequence. The output of the enhancement layer of the i-th diagnostic module serves as the first input of the input layer of the i+1-th diagnostic module, and the output of the input layer of the first diagnostic module also serves as the second input of the input layers of the remaining diagnostic modules; where i is a positive integer; The classification layer of each diagnosis module includes multiple classifiers, each classifier is used to perform local diagnosis, and the enhancement layer of each diagnosis module is used to enhance the output of each classifier.

[0007] Preferably, the method further comprises: obtaining a sample training set, and training a base learner and a meta learner of ensemble learning based on the sample training set to obtain a trained base learner of ensemble learning and a trained meta learner of ensemble learning.

[0008] Preferably, the sample training set includes training samples and corresponding test samples; training the base learner and meta learner of ensemble learning based on the sample training set to obtain the trained base learner of ensemble learning and the trained meta learner of ensemble learning includes: The training samples are divided into five subsamples; The first subsample is used as training data, and the second subsample is used as test data to train the base learner to obtain a first prediction result; The first sub-sample and the second sub-sample are used as training data, and the third sub-sample is used as test data to train the base learner to obtain the second prediction result; The first to third subsamples are used as training data, and the fourth subsample is used as test data to train the base learner to obtain a third prediction result; The first to fourth subsamples are used as training data, and the fifth subsample is used as test data to train the base learner to obtain a fourth prediction result; The training sample is used as training data, and the corresponding test sample is used as test data to train the base learner to obtain the fifth prediction result; The first to fifth prediction results are used as training data for meta-learning, and the test sample is used as test data to train the meta-learner to obtain the prediction results of the meta-learner; Based on the prediction results of the meta-learner and the labels of the sample training set, the loss value is calculated, and the parameters of the base learner and the meta-learner are updated based on the loss value until convergence, thereby obtaining the trained ensemble learning base learner and the trained ensemble learning meta-learner.

[0009] Preferably, the classifier of the base learner is an LSTM network, a gradient boosting tree, a CNN network, a random forest, and a clustering algorithm; and the meta-learner is constructed using a Ridge regression model or a Lasso regression model.

[0010] Preferably, during the training process of the base learner using the training data and the test data, the method further comprises: Input the training data or test data as input data to the base learner of the ensemble learning; Input the input data to the input layer of the first diagnostic module of the base learner for feature construction to obtain input features; Input the input features into the classification layer of the first diagnostic module of the base learner for prediction to obtain the predicted value; The predicted value is input into the enhancement layer of the first diagnostic module of the base learner for enhancement to obtain an enhanced predicted value; The enhanced prediction value is input into the next diagnostic module of the base learner. The enhanced prediction value and the input feature are spliced ​​based on the input layer of the next diagnostic module of the base learner to obtain a new input feature. The new input feature is re-input into the classification layer of the first diagnostic module of the base learner for prediction until the enhancement layer of the last diagnostic module of the base learner outputs the final enhanced prediction value, and the final enhanced prediction value is used as the prediction output of the base learner.

[0011] Preferably, the enhancement layer of each diagnostic module of the base learner enhances the predicted value to obtain an enhanced predicted value, including: Calculate the difference between the predicted value output by the classification layer of the i-th diagnostic module and the label, and the difference between the predicted value output by the classification layer of the in-th diagnostic module and the label, and obtain a difference set; where n = 0, 1, ..., i-1; The minimum value in the difference set is extracted, and the prediction value corresponding to the minimum value in the difference set is used as the enhanced prediction value output by the enhancement layer of the i-th diagnosis module.

[0012] Preferably, the calculation expression for the difference between the predicted value output by the classification layer of the i-th diagnosis module and the label is: ; Where, is the difference between the predicted value and the label output by the classification layer of the i-th diagnostic module, J is the dimension of the input feature, is the jth input feature, is the predicted value output by the classification layer of the i-th diagnostic module, is the label corresponding to the j-th input feature.

[0013] In a second aspect, the present invention provides a multi-source data fusion power equipment fault diagnosis system for implementing the above-mentioned multi-source data fusion power equipment fault diagnosis method, the system comprising: A data acquisition module, used to acquire multi-source data of power equipment; A data processing module, used for preprocessing multi-source data of power equipment to obtain processed multi-source data; The local diagnosis module is used to perform local diagnosis on the processed multi-source data based on the trained ensemble learning base learner to obtain multiple local diagnosis results; The diagnostic fusion module is used to fuse multiple local diagnostic results based on the trained ensemble learning meta-learner to obtain the fault diagnosis results of the power equipment; The fault warning module is used to generate fault warning information based on the fault diagnosis results of the power equipment and to visualize the fault warning information.

[0014] In a third aspect, the present invention provides an electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the above-mentioned multi-source data fusion method for diagnosing faults of electric power equipment when executing the computer program.

[0015] In a fourth aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the above-mentioned multi-source data fusion method for fault diagnosis of electric power equipment.

[0016] Beneficial effects: The present invention utilizes a base learner of ensemble learning to analyze and diagnose multi-source data. The data from each source is diagnosed separately to obtain a diagnostic result for each source data, i.e., a local diagnostic result. The multiple local diagnostic results are then fused through a trained meta-learner of ensemble learning to obtain a final fault diagnosis result. Therefore, by combining multiple diagnostic results through ensemble learning, the risk of overfitting can be reduced, and at the same time, the complementarity and correlation between multi-source data can be fully explored, thereby improving stability and diagnostic accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] The accompanying drawings are used to provide a further understanding of the embodiments of the present invention and constitute a part of the specification. Together with the following detailed description, they are used to explain the embodiments of the present invention, but do not constitute a limitation of the embodiments of the present invention. In the accompanying drawings: Figure 1 This is a flow chart of a method for diagnosing faults of electric power equipment using multi-source data fusion provided by one embodiment of the present invention; Figure 2The present invention provides a block diagram of a multi-source data fusion power equipment fault diagnosis system according to an embodiment of the present invention. DETAILED DESCRIPTION

[0018] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the present invention will be briefly introduced below in conjunction with the drawings and the description of the embodiments or the prior art. Obviously, the following description of the structure of the drawings is only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative work. It should be noted that the description of these embodiments is used to help understand the present invention, but does not constitute a limitation of the present invention.

[0019] Example 1 Figure 1 This is a flow chart of a method for diagnosing faults in power equipment using multi-source data fusion provided by one embodiment of the present invention. Figure 1 As shown, this embodiment provides a method for diagnosing faults of electric power equipment by fusion of multi-source data. The method runs on a server and includes: Step S10: Acquire multi-source data of the power equipment; in this embodiment, the multi-source data includes but is not limited to multi-dimensional data such as physical sensor data, operating status data, environmental data, and unstructured data.

[0020] Among them, physical sensor data includes but is not limited to: temperature, vibration, partial discharge, current / voltage waveform, oil chromatography analysis (such as dissolved gas in transformer oil), etc. The operating status data includes but is not limited to: load rate, number of switching operations, insulation resistance, and historical equipment maintenance records; Environmental data includes but is not limited to: humidity, temperature, electromagnetic interference, and geographic location information; Among them, unstructured data includes but is not limited to: infrared thermal imaging, ultrasonic detection images, and visible light images (such as insulator cracks).

[0021] The multi-source data of the power equipment in this embodiment is collected by the DCS control system and related sensors of the power equipment. The DCS control system and related sensors upload the collected multi-source data to the server so that the server obtains the relevant multi-source data.

[0022] Step S20: pre-processing the multi-source data of the power equipment to obtain processed multi-source data, wherein the pre-processing includes but is not limited to data cleaning and data conversion.

[0023] Step S30: performing local diagnosis on the processed multi-source data based on the trained ensemble learning base learner to obtain multiple local diagnosis results.

[0024] In this embodiment, the base learner has multiple classifiers, which classify different multi-source data and identify abnormal data. The abnormal data is used to indicate that there is a fault phenomenon in the power equipment.

[0025] For example, the multi-source data includes vibration data, partial discharge data, and infrared thermal imaging, and the classifiers include LSTM (Long Short-Term Memory) network, gradient boosting tree (GBDT), and convolutional neural network (fine-tuned ResNet-18).

[0026] The vibration data is converted into frequency domain features, which are then input into the LSTM network. These frequency domain features are classified and abnormal frequency domain features are identified. These abnormal frequency domain features are used to characterize mechanical looseness faults in power equipment.

[0027] The partial discharge data is converted into UHF (partial discharge ultra-high frequency) pulse statistical features, which are then input into a gradient boosting tree to classify these UHF pulse statistical features and identify abnormal UHF pulse statistical features. These abnormal UHF pulse statistical features are used to characterize abnormal discharge faults in power equipment.

[0028] Infrared image features are extracted from infrared thermal imaging, and these infrared image features are input into the convolutional neural network for final classification to identify abnormal infrared image features. These abnormal infrared image features are used to characterize overheating faults in power equipment.

[0029] Step S40: A meta-learner based on the trained ensemble learning is used to fuse multiple local diagnosis results to obtain a fault diagnosis result of the power equipment.

[0030] In this embodiment, after the base learner of the integrated learning completes the classification task, it can output multiple local diagnostic results, and then the meta-learner of the integrated learning integrates the multiple diagnostic results to obtain the fault diagnosis result of the power equipment. The fault diagnosis result is a compound fault, such as: mechanical looseness + abnormal discharge, mechanical looseness + overheating, mechanical looseness + abnormal discharge + overheating and other compound faults.

[0031] Step S50: Based on the fault diagnosis result of the power equipment, fault warning information is generated and the fault warning information is visualized; in this embodiment, when the fault diagnosis result of the power equipment is the above-mentioned composite fault, fault warning information is generated, and the server sends the warning information to the user end, such as a mobile phone and other devices, to inform the management personnel that there is a fault in the power equipment, so that the management personnel can promptly repair the faulty equipment to avoid further expansion of the fault phenomenon.

[0032] The present invention utilizes a base learner of ensemble learning to analyze and diagnose multi-source data. The data from each source is diagnosed separately to obtain a diagnostic result for each source data, i.e., a local diagnostic result. The multiple local diagnostic results are then fused through a trained meta-learner of ensemble learning to obtain a final fault diagnosis result. Therefore, by combining multiple diagnostic results through ensemble learning, the risk of overfitting can be reduced, and at the same time, the complementarity and correlation between multi-source data can be fully explored, thereby improving stability and diagnostic accuracy.

[0033] As a further optimization of this embodiment, the trained ensemble learning base learner includes: multiple diagnostic modules connected in series, each diagnostic module includes: an input layer, a classification layer and an enhancement layer; in this embodiment, 4 to 10 diagnostic modules are used, and the diagnostic modules are connected in series.

[0034] The input layer, classification layer, and enhancement layer in the i-th diagnostic module are connected in sequence. The output of the enhancement layer of the i-th diagnostic module serves as the first input of the input layer of the i+1-th diagnostic module, and the output of the input layer of the first diagnostic module also serves as the second input of the input layers of the remaining diagnostic modules; where i is a positive integer; The classification layer of each diagnostic module includes multiple classifiers, each classifier is used to perform local diagnosis, and the enhancement layer of each diagnostic module is used to enhance the output of each classifier; in this embodiment, the classifiers of the base learner are LSTM network, gradient boosting tree, CNN network, random forest and clustering algorithm; the meta-learner is constructed using Ridge regression model or Lasso regression model.

[0035] As a further optimization of this embodiment, the method also includes: obtaining a sample training set, training the base learner and meta-learner of the ensemble learning based on the sample training set, and obtaining a trained base learner of the ensemble learning and a trained meta-learner of the ensemble learning; wherein the sample training set is historical data of the power equipment, and the historical data contains labels of various fault types, multi-dimensional data such as physical sensor data, operating status data, environmental data, and unstructured data of the historical data, as well as timestamps corresponding to these multi-dimensional data.

[0036] In this embodiment, the sample training set includes training samples and corresponding test samples. In this embodiment, a portion of the historical data is divided into training samples, and the rest is divided into test samples; for example, 80% of the historical data is used as training samples, and the remaining 20% ​​of the data is used as test samples. Specifically, the base learner and meta learner of the ensemble learning are trained based on the sample training set to obtain the trained base learner of the ensemble learning and the trained meta learner of the ensemble learning, including: Step a10: Divide the training sample into five sub-samples. In this embodiment, the training samples are divided into five sub-samples according to the time sequence of the training samples.

[0037] Step a20: Use the first sub-sample as training data and the second sub-sample as test data to train the base learner and obtain a first prediction result.

[0038] Step a30: Use the first sub-sample and the second sub-sample as training data, and the third sub-sample as test data to train the base learner to obtain a second prediction result.

[0039] Step a40: Use the first to third subsamples as training data and the fourth subsample as test data to train the base learner and obtain a third prediction result.

[0040] Step a50: Use the first to fourth subsamples as training data and the fifth subsample as test data to train the base learner and obtain a fourth prediction result.

[0041] Step a60: Use the training sample as training data and the corresponding test sample as test data to train the base learner to obtain a fifth prediction result.

[0042] Step a70: Use the first to fifth prediction results as training data for meta-learning and the test sample as test data to train the meta-learner and obtain the prediction results of the meta-learner.

[0043] Step a80: Based on the prediction results of the meta-learner and the labels of the sample training set, the loss value is calculated, and the parameters of the base learner and the meta-learner are updated based on the loss value until convergence, thereby obtaining a trained ensemble learning base learner and a trained ensemble learning meta-learner.

[0044] In this embodiment, through the training steps of step a10 to step a80, each classifier of the base learner can be trained and evaluated based on different training samples and test samples, and the training method can also maximize the utilization of data while avoiding future information leakage, thereby ensuring the robustness of model training and the accuracy of prediction.

[0045] As a further optimization of this embodiment, during the training process of the base learner using the training data and the test data, the method further includes: Step b10: Input the training data or test data as input data to the base learner of the ensemble learning.

[0046] Step b20: Input the input data to the input layer of the first diagnostic module of the base learner for feature construction to obtain input features. For example, infrared thermal imaging can be converted into infrared image features, vibration data can be converted into frequency domain features, and partial discharge data can be converted into UHF pulse statistical features. The input feature is a characteristic distribution vector.

[0047] Step b30: Input the input features to the classification layer of the first diagnostic module of the base learner for prediction to obtain a predicted value; the predicted value of this embodiment is a predicted probability distribution vector; Step b40: Input the predicted value into the enhancement layer of the first diagnostic module of the base learner for enhancement to obtain an enhanced predicted value.

[0048] Step b50: Input the enhanced prediction value into the next diagnostic module of the base learner, concatenate the enhanced prediction value and the input features based on the input layer of the next diagnostic module of the base learner to obtain new input features, and re-input the new input features into the classification layer of the first diagnostic module of the base learner for prediction, until the enhancement layer of the last diagnostic module of the base learner outputs the final enhanced prediction value, and the final enhanced prediction value is used as the prediction output of the base learner.

[0049] In this embodiment, in each diagnostic module, the output of the previous diagnostic module serves as the input of the next diagnostic module, and the output prediction value is enhanced in the diagnostic module, making the prediction value output by the diagnostic module more accurate, which can effectively improve the prediction ability of the base learner.

[0050] Specifically, the enhancement layer of each diagnostic module of the base learner enhances the predicted value to obtain an enhanced predicted value, including: Step c10: Calculate the difference between the predicted value output by the classification layer of the i-th diagnostic module and the label and the difference between the predicted value output by the classification layer of the in-th diagnostic module and the label to obtain a difference set; where n=0,1,…,i-1.

[0051] Among them, the calculation expression of the difference between the predicted value output by the classification layer of the i-th diagnosis module and the label is: ; Where, is the difference between the predicted value and the label output by the classification layer of the i-th diagnostic module, J is the dimension of the input feature, is the jth input feature, is the predicted value output by the classification layer of the i-th diagnostic module, is the label corresponding to the j-th input feature.

[0052] Step c20: extracting the minimum value in the difference set, and using the prediction value corresponding to the minimum value in the difference set as the enhanced prediction value output by the enhancement layer of the i-th diagnosis module.

[0053] In this embodiment, the prediction value with the smallest difference is selected from the trained diagnostic modules as the enhanced prediction value output by the currently trained diagnostic module, and the enhanced prediction value is then spliced ​​with the input feature; for example, it is directly connected in series and used as the input of the next diagnostic module; therefore, the deployment of the enhancement layer in the base learner of this embodiment can significantly improve the prediction performance of the algorithm.

[0054] As a further optimization of this embodiment, the method further includes: dynamically adjusting the number of diagnostic modules, including: Step d10: Initializing the initial number of diagnostic modules, for example, the initial number is 4 to 10; Step d20: When the difference between the predicted value and the label output by the classification layer of three consecutive diagnostic modules reaches the preset condition, the training of the base learner is terminated, and the number of diagnostic modules is obtained at this time; wherein, the preset condition is that the three differences remain unchanged or the rate of change is lower than the preset threshold.

[0055] Therefore, the present invention uses the base learner of ensemble learning to analyze and diagnose multi-source data. The data from each source is diagnosed separately to obtain the diagnostic results of each source data, that is, the local diagnostic results. The multiple local diagnostic results are then fused through the trained meta-learner of ensemble learning to obtain the final fault diagnosis result. Therefore, by combining multiple diagnostic results through ensemble learning, the risk of overfitting can be reduced, and at the same time, the complementarity and correlation between multi-source data can be fully explored, thereby improving stability and diagnostic accuracy.

[0056] Example 2 Figure 2 This is a block diagram of a multi-source data fusion power equipment fault diagnosis system provided by an embodiment of the present invention. Figure 2 As shown, this embodiment provides a multi-source data fusion power equipment fault diagnosis system, which is used to implement the multi-source data fusion power equipment fault diagnosis method in embodiment 1. The system includes: A data acquisition module, used to acquire multi-source data of power equipment; A data processing module, used for preprocessing multi-source data of power equipment to obtain processed multi-source data; The local diagnosis module is used to perform local diagnosis on the processed multi-source data based on the trained ensemble learning base learner to obtain multiple local diagnosis results; The diagnostic fusion module is used to fuse multiple local diagnostic results based on the trained ensemble learning meta-learner to obtain the fault diagnosis results of the power equipment; The fault warning module is used to generate fault warning information based on the fault diagnosis results of the power equipment and to visualize the fault warning information.

[0057] This embodiment also provides an electronic device, 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 method for diagnosing faults of electric power equipment using multi-source data fusion in the first embodiment is implemented.

[0058] This embodiment further provides a computer-readable storage medium having a computer program stored thereon. When the program is executed by a processor, the method for diagnosing faults of electric power equipment using multi-source data fusion in the first embodiment is implemented.

[0059] The present invention utilizes a base learner of ensemble learning to analyze and diagnose multi-source data. The data from each source is diagnosed separately to obtain a diagnostic result for each source data, i.e., a local diagnostic result. The multiple local diagnostic results are then fused through a trained meta-learner of ensemble learning to obtain a final fault diagnosis result. Therefore, by combining multiple diagnostic results through ensemble learning, the risk of overfitting can be reduced, and at the same time, the complementarity and correlation between multi-source data can be fully explored, thereby improving stability and diagnostic accuracy.

[0060] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware. Furthermore, the present application may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0061] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 a process or multiple processes and / or boxes Figure 1 A system that specifies the functions of a box or boxes.

[0062] The above are merely embodiments of the present application and are not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present application should all be included within the scope of the claims of the present application.

Claims

1. A multi-source data fusion method for power equipment fault diagnosis, characterized in that: The method comprises: Acquire multi-source data of power equipment; Preprocessing multi-source data of power equipment to obtain processed multi-source data; Based on the trained ensemble learning base learner, local diagnosis is performed on the processed multi-source data to obtain multiple local diagnosis results; The meta-learner based on the trained ensemble learning is used to fuse multiple local diagnosis results to obtain the fault diagnosis results of the power equipment; Based on the fault diagnosis results of the power equipment, fault warning information is generated and displayed visually.

2. The method for diagnosing faults of electric power equipment based on multi-source data fusion according to claim 1, characterized in that: The trained ensemble learning base learner includes: a plurality of diagnosis modules connected in series, each diagnosis module includes: an input layer, a classification layer and an enhancement layer; The input layer, classification layer, and enhancement layer in the i-th diagnostic module are connected in sequence. The output of the enhancement layer of the i-th diagnostic module serves as the first input of the input layer of the i+1-th diagnostic module, and the output of the input layer of the first diagnostic module also serves as the second input of the input layers of the remaining diagnostic modules; where i is a positive integer; The classification layer of each diagnosis module includes multiple classifiers, each classifier is used to perform local diagnosis, and the enhancement layer of each diagnosis module is used to enhance the output of each classifier.

3. The method for diagnosing faults of electric power equipment based on multi-source data fusion according to claim 2, characterized in that: The method further includes: obtaining a sample training set, and training a base learner and a meta learner of ensemble learning based on the sample training set to obtain a trained base learner of ensemble learning and a trained meta learner of ensemble learning.

4. The method for diagnosing faults of electric power equipment based on multi-source data fusion according to claim 3, characterized in that: The sample training set includes training samples and corresponding test samples; training the base learner and meta learner of the ensemble learning based on the sample training set to obtain the trained base learner of the ensemble learning and the trained meta learner of the ensemble learning, including: The training samples are divided into five subsamples; The first subsample is used as training data, and the second subsample is used as test data to train the base learner to obtain a first prediction result; The first sub-sample and the second sub-sample are used as training data, and the third sub-sample is used as test data to train the base learner to obtain the second prediction result; The first to third subsamples are used as training data, and the fourth subsample is used as test data to train the base learner to obtain a third prediction result; The first to fourth subsamples are used as training data, and the fifth subsample is used as test data to train the base learner to obtain a fourth prediction result; The training sample is used as training data, and the corresponding test sample is used as test data to train the base learner to obtain the fifth prediction result; The first to fifth prediction results are used as training data for meta-learning, and the test sample is used as test data to train the meta-learner to obtain the prediction results of the meta-learner; Based on the prediction results of the meta-learner and the labels of the sample training set, the loss value is calculated, and the parameters of the base learner and the meta-learner are updated based on the loss value until convergence, thereby obtaining the trained ensemble learning base learner and the trained ensemble learning meta-learner.

5. The method for diagnosing faults of electric power equipment based on multi-source data fusion according to claim 4, characterized in that: The classifiers of the base learner are LSTM network, gradient boosting tree, CNN network, random forest and clustering algorithm; the meta-learner is constructed using Ridge regression model or Lasso regression model.

6. The method for diagnosing faults of electric power equipment based on multi-source data fusion according to claim 4, characterized in that: During the training process of the base learner using the training data and the test data, the method further comprises: Input the training data or test data as input data to the base learner of the ensemble learning; Input the input data to the input layer of the first diagnostic module of the base learner for feature construction to obtain input features; Input the input features into the classification layer of the first diagnostic module of the base learner for prediction to obtain the predicted value; The predicted value is input into the enhancement layer of the first diagnostic module of the base learner for enhancement to obtain an enhanced predicted value; The enhanced prediction value is input into the next diagnostic module of the base learner. The enhanced prediction value and the input feature are spliced ​​based on the input layer of the next diagnostic module of the base learner to obtain a new input feature. The new input feature is re-input into the classification layer of the first diagnostic module of the base learner for prediction until the enhancement layer of the last diagnostic module of the base learner outputs the final enhanced prediction value, and the final enhanced prediction value is used as the prediction output of the base learner.

7. The method for diagnosing faults of electric power equipment based on multi-source data fusion according to claim 6, characterized in that: The enhancement layer of each diagnostic module of the base learner enhances the predicted value to obtain an enhanced predicted value, including: Calculate the difference between the predicted value output by the classification layer of the i-th diagnostic module and the label, and the difference between the predicted value output by the classification layer of the in-th diagnostic module and the label, and obtain a difference set; where n = 0, 1, ..., i-1; The minimum value in the difference set is extracted, and the prediction value corresponding to the minimum value in the difference set is used as the enhanced prediction value output by the enhancement layer of the i-th diagnosis module.

8. The method for diagnosing faults of electric power equipment based on multi-source data fusion according to claim 7, characterized in that: The calculation expression of the difference between the predicted value output by the classification layer of the i-th diagnostic module and the label is: ； Where, is the difference between the predicted value and the label output by the classification layer of the i-th diagnostic module, J is the dimension of the input feature, is the jth input feature, is the predicted value output by the classification layer of the i-th diagnostic module, is the label corresponding to the j-th input feature.

9. A multi-source data fusion power equipment fault diagnosis system, used to implement the multi-source data fusion power equipment fault diagnosis method according to any one of claims 1 to 8, characterized in that: The system comprises: A data acquisition module, used to acquire multi-source data of power equipment; A data processing module, used for preprocessing multi-source data of power equipment to obtain processed multi-source data; The local diagnosis module is used to perform local diagnosis on the processed multi-source data based on the trained ensemble learning base learner to obtain multiple local diagnosis results; The diagnostic fusion module is used to fuse multiple local diagnostic results based on the trained ensemble learning meta-learner to obtain the fault diagnosis results of the power equipment; The fault warning module is used to generate fault warning information based on the fault diagnosis results of the power equipment and to visualize the fault warning information.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the power equipment fault diagnosis method based on multi-source data fusion according to any one of claims 1 to 8 is implemented.

Images