Diagnosis method suitable for transformer operation state monitoring and early warning

Abstract

The invention discloses a diagnosis method suitable for transformer operation state monitoring and early warning, and relates to the technical field of monitoring and early warning, and the diagnosis method comprises the following steps: oil chromatography data acquisition and standard exceeding determination are carried out, and a determination result is obtained; performing primary fault judgment according to the judgment result to obtain a primary judgment result; according to the judgment result, performing neural network initial classification fault judgment to obtain a neural network initial classification judgment result; performing comprehensive judgment according to the initial classification judgment result and the neural network initial classification judgment result to obtain an initial classification fault type; performing subdivision fault judgment according to the primary division fault type to obtain a subdivision judgment result; performing neural network subdivision fault judgment according to the primary division fault type to obtain a neural network subdivision judgment result; and performing comprehensive judgment according to a subdivision judgment result and a neural network subdivision judgment result to obtain a final fault type. The method improves the classification accuracy, improves the recognition capability of complex faults, and reduces the misjudgment and missed judgment risks.

Description

Technical Field

[0001] The present invention relates to the technical field of monitoring and early warning, and in particular to a diagnostic method suitable for transformer operating state monitoring and early warning. Background Art

[0002] As a key component of power systems, accurate monitoring of transformer operating conditions and fault diagnosis are crucial to ensuring grid stability. Currently, transformer fault detection relies primarily on oil chromatography, which detects dissolved gas components in the oil to determine whether potential faults exist within the transformer and further identify the fault type.

[0003] Existing technologies have shortcomings in fault type determination: While the three-ratio method, a common fault type determination method, is computationally simple and convenient for engineering applications, it cannot effectively identify complex fault patterns, especially for complex faults, resulting in low accuracy. Furthermore, traditional neural network models are susceptible to data distribution during training, resulting in unstable classification results for fault types with relatively small amounts of data. Existing fault type determination often relies on the results of only one method, resulting in biased judgments and potential misjudgments or omissions. Summary of the Invention

[0004] In view of the deficiencies in the prior art, the present invention provides a diagnostic method suitable for transformer operating status monitoring and early warning to solve the problems raised in the above background technology.

[0005] To achieve the above object, the present invention provides the following technical solutions:

[0006] In a first aspect, an embodiment of the present invention provides a diagnostic method for transformer operating status monitoring and early warning, comprising the following steps:

[0007] S1. Oil chromatography data collection and over-standard judgment to obtain judgment results;

[0008] S2. Perform preliminary fault judgment based on the judgment result to obtain the preliminary judgment result;

[0009] S3. Perform a neural network preliminary fault judgment based on the judgment result to obtain a neural network preliminary judgment result;

[0010] S4. Perform comprehensive judgment based on the initial judgment result and the neural network initial judgment result to obtain the initial fault type;

[0011] S5. Perform detailed fault judgment based on the initial fault type and obtain detailed judgment results;

[0012] S6. Perform neural network segmentation fault judgment based on the initial fault type to obtain a neural network segmentation judgment result;

[0013] S7. Perform comprehensive judgment based on the segmentation judgment result and the neural network segmentation judgment result to obtain the final fault type.

[0014] To further optimize this technical solution, the oil chromatographic data collection and exceeding-standard judgment in S1 include:

[0015] Based on the oil chromatography data, the IEC60599 dissolved gas analysis standard is used to compare it with the preset fault judgment threshold to determine whether it exceeds the threshold. If it exceeds the threshold, the diagnostic logic is activated.

[0016] To further optimize this technical solution, the neural network initial fault judgment in S3 includes:

[0017] After the diagnostic logic is started according to the judgment result, the initial fault judgment model is used to make an initial fault type judgment.

[0018] To further optimize this technical solution, the initial fault judgment model includes:

[0019] Hidden layer neural network calculation:

[0020] ;

[0021] in:

[0022] : The initial output of the hidden layer neurons;

[0023] : activation function;

[0024] : The connection weight from the input layer to the hidden layer;

[0025] : original gas concentration;

[0026] : Adaptive fault feature weight;

[0027] : The initial bias term of the hidden layer neurons;

[0028] Output layer fault category probability calculation:

[0029] ;

[0030] in:

[0031] : The probability of initial classification of fault categories;

[0032] , : The connection weight from the hidden layer to the output layer.

[0033] To further optimize this technical solution, the adaptive fault feature weights include:

[0034] ;

[0035] in:

[0036] , : The rate of change of gas concentration.

[0037] To further optimize this technical solution, the neural network segmentation fault judgment in S6 includes:

[0038] According to the initial fault type classification results, the fault type is further subdivided using the subdivision fault judgment model to obtain the neural network subdivision judgment results.

[0039] To further optimize this technical solution, the segmented fault judgment model includes:

[0040] Hidden layer calculation formula:

[0041] ;

[0042] in:

[0043] : The segmented output of the hidden layer neurons;

[0044] : gas concentration after adaptive normalization;

[0045] : The subdivision bias term of the hidden layer neurons;

[0046] Output layer calculation formula:

[0047] ;

[0048] in:

[0049] : The probability of breaking down the fault categories.

[0050] To further optimize this technical solution, the adaptively normalized gas concentration includes:

[0051] ;

[0052] in:

[0053] : the mean value of gas i;

[0054] : standard deviation of gas i;

[0055] : Weight coefficient.

[0056] To further optimize this technical solution, the comprehensive judgment in S7 includes:

[0057] According to the segmented judgment results obtained in steps S5 and S6, a weighted random forest model is used to make a final fault type judgment and determine the final fault.

[0058] To further optimize this technical solution, the weighted random forest model includes:

[0059] Judgment result fusion:

[0060] ;

[0061] in:

[0062] : Judgment result after fusion;

[0063] : weight coefficient;

[0064] : Judgment results of three judgment methods;

[0065] Decision tree construction:

[0066] ;

[0067] in:

[0068] : kth decision tree;

[0069] : training sample of the kth tree;

[0070] Calculate the classification probability of each decision tree:

[0071] ;

[0072] in:

[0073] : The predicted probability of the kth decision tree for the fault category;

[0074] : All fault categories;

[0075] Voting mechanism to obtain the final classification results:

[0076] ;

[0077] in:

[0078] : Final fault category prediction result;

[0079] : The total number of decision trees.

[0080] In a second aspect, an embodiment of the present invention provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: when the computer program instructions are executed by the processor, the steps of a diagnostic method suitable for transformer operating status monitoring and early warning as described in the first aspect of the present invention are implemented.

[0081] In a third aspect, an embodiment of the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program instructions are executed by a processor, the steps of a diagnostic method suitable for transformer operating status monitoring and early warning as described in the first aspect of the present invention are implemented.

[0082] Compared with the prior art, the present invention provides a diagnostic method suitable for transformer operating status monitoring and early warning, which has the following beneficial effects:

[0083] This diagnostic method, suitable for transformer operating status monitoring and early warning, uses an initial fault judgment model and a detailed fault judgment model, and adopts a dynamic weight adjustment mechanism to improve the classification accuracy of minority category data and enhance the ability to identify complex faults. Compared with traditional methods, it improves classification accuracy and reduces the occurrence of misjudgments and missed judgments.

[0084] Through the weighted random forest model, the results of multiple judgment methods are integrated, which improves the stability of the classification effect and reduces the risk of misjudgment. BRIEF DESCRIPTION OF THE DRAWINGS

[0085] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0086] Figure 1 This is a flow chart of a diagnostic method for transformer operating status monitoring and early warning proposed by the present invention;

[0087] Figure 2 This is a flow chart of a preliminary fault judgment model of a diagnostic method for transformer operating status monitoring and early warning proposed by the present invention;

[0088] Figure 3 This is a flow chart of a subdivided fault judgment model of a diagnostic method for transformer operation status monitoring and early warning proposed by the present invention;

[0089] Figure 4 This is a flow chart of a weighted random forest model for a diagnostic method suitable for transformer operating status monitoring and early warning proposed by the present invention. DETAILED DESCRIPTION

[0090] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0091] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0092] Secondly, the term "one embodiment" or "embodiment" herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in various places throughout this specification does not necessarily refer to the same embodiment, nor does it constitute a separate or selective embodiment that is mutually exclusive with other embodiments.

[0093] Example 1:

[0094] Reference Figures 1 to 4 , which is the first embodiment of the present invention, provides a diagnostic method suitable for transformer operation status monitoring and early warning, comprising the following steps:

[0095] S1. Oil chromatography data collection and over-standard judgment to obtain judgment results

[0096] In this embodiment, the oil chromatogram data collection and exceeding-standard judgment include:

[0097] Based on the oil chromatography data, including the current measured concentration values ​​of H2, CO, CO2, CH4, C2H4, C2H6, C2H2, and total hydrocarbons, the IEC60599 dissolved gas analysis standard is used to compare with the preset fault judgment threshold to determine whether it exceeds the threshold. If it exceeds the threshold, the diagnostic logic is activated.

[0098] S2. Perform preliminary fault judgment based on the judgment result to obtain a preliminary fault judgment result.

[0099] In this embodiment, the initial fault determination includes:

[0100] After the diagnostic logic is activated based on the judgment results, the current measured data of H2, C2H2, C2H4, CH4, C2H6 and total hydrocarbons are extracted for three ratio judgments, and the ratios of H2 / C2H2, C2H2 / C2H4, and CH4 / C2H6 are calculated. The fault type is judged according to the Rogers ratio rule. For example, when C2H2 / C2H4>3.0, it usually indicates the presence of high-energy discharge. The case-based reasoning judgment method is used to match the closest fault type based on similar cases in the historical case library, and the similarity between the current gas data and the historical fault case is calculated. If the similarity is higher than the set threshold, the conclusion of the case is referenced.

[0101] S3. Perform a neural network preliminary fault judgment based on the judgment result to obtain a neural network preliminary judgment result.

[0102] In this embodiment, the neural network initial fault determination includes:

[0103] After the diagnostic logic is started according to the judgment result, the initial fault judgment model is used to make an initial fault type judgment.

[0104] Furthermore, the initial fault judgment model includes:

[0105] Hidden layer neural network calculation:

[0106] ;

[0107] in:

[0108] : The initial output of the hidden layer neurons;

[0109] : Activation function, use Leaky ReLU activation function to avoid the gradient disappearance problem;

[0110] : The connection weight from the input layer to the hidden layer;

[0111] : original gas concentration;

[0112] : Adaptive fault feature weight, indicating the importance of the gas. The greater the change in gas concentration, the higher the weight.

[0113] : The initial bias term of the hidden layer neurons;

[0114] Output layer fault category probability calculation:

[0115] ;

[0116] in:

[0117] : The probability of initial classification of fault categories, including overheating or discharge;

[0118] , : The connection weight from the hidden layer to the output layer.

[0119] Furthermore, the adaptive fault feature weights include:

[0120] ;

[0121] in:

[0122] , : The rate of change of gas concentration. The larger the value, the more drastic the change of the gas during the fault development process.

[0123] This model makes an initial classification of fault types by calculating the changes in the concentration of each gas.

[0124] This neural network initial fault judgment model uses adaptive weights and performs real-time adjustments, which improves the dynamic adaptability of fault judgment. It also improves the credibility of judgment based on dynamic gas concentration data. The use of the Leaky ReLU activation function avoids the gradient vanishing problem and reduces the risk of overfitting.

[0125] Uses of the above models include:

[0126] Data input: Get gas concentration data from step S1 , as model input;

[0127] Hidden layer calculation: based on the rate of change of gas concentration , , calculate the weights of different gases, use them to calculate the hidden layer of the neural network, and get the hidden layer calculation results ;

[0128] Fault category calculation: Calculation results based on hidden layer Calculate and get the probability of the initial fault category , perform preliminary fault classification based on the probability. If the probability of a certain fault category is greater than 0.8, then output this fault category.

[0129] S4. Perform a comprehensive judgment based on the initial classification result and the neural network initial classification result to obtain the initial fault type.

[0130] In this embodiment, the comprehensive judgment includes:

[0131] Based on the preliminary judgment results obtained using the three-ratio judgment method, the case-based reasoning judgment method, and the neural network preliminary judgment results obtained using the neural network, the results of the three methods are combined for judgment, and the final diagnosis result is determined by two or more results. For example, if the judgment results of the three-ratio method and the neural network are the same, but different from the results of the case-based reasoning judgment method, the two identical judgment results shall prevail.

[0132] S5. Perform detailed fault judgment based on the initial fault type to obtain a detailed judgment result.

[0133] In this embodiment, the detailed fault determination includes:

[0134] Based on step S2, further calculate the ratios of CH4 / H2, C2H4 / C2H6, etc., and combine them with the extended ratio rules of IEC60599 to perform more detailed fault classification and determine the fault type;

[0135] Based on the case-based reasoning in step S2, additional parameters such as environmental factors and equipment operating load are added to perform secondary matching to determine the fault type.

[0136] S6. Perform neural network segmentation fault judgment based on the initial fault type to obtain the neural network segmentation judgment result.

[0137] In this embodiment, the neural network segmentation fault judgment includes:

[0138] According to the initial fault type classification results, the fault type is further subdivided using the segmentation fault judgment model to obtain the neural network segmentation judgment results to improve the accuracy of fault classification.

[0139] Furthermore, the subdivided fault judgment model includes:

[0140] Hidden layer calculation formula:

[0141] ;

[0142] in:

[0143] : The segmented output of the hidden layer neurons;

[0144] : gas concentration after adaptive normalization;

[0145] : The subdivision bias term of the hidden layer neurons;

[0146] Output layer calculation formula:

[0147] ;

[0148] in:

[0149] : The probability of breaking down the failure categories, including mild overheating, severe overheating, low discharge, high discharge, etc.

[0150] Furthermore, the adaptively normalized gas concentration includes:

[0151] ;

[0152] in:

[0153] : the mean value of gas i;

[0154] : standard deviation of gas i;

[0155] : Weight coefficient, which indicates the relative importance of the gas under the current fault type and is set according to the gas type.

[0156] The relative importance of gas under the current fault type is adjusted through adaptive normalization and dynamic weighting, and a neural network is used to implement subdivided fault classification and obtain the probabilities of different fault categories.

[0157] Adaptive normalization and dynamic weighting mechanisms can flexibly adjust the influence of gas in the classification process, improving the accuracy of fault classification.

[0158] The steps for using the model include:

[0159] Data normalization: According to the adaptive fault feature weights obtained in step S3 Gas concentration data obtained by S1 Calculate the gas concentration after adaptive normalization , used for calculating the breakdown of fault types;

[0160] Probability calculation: Based on the adaptively normalized gas concentration obtained , calculate the segmented output of the neurons in the hidden layer of the neural network , then calculate the probability of the subdivided fault category ;

[0161] Fault segmentation: According to the probability of the obtained segmented fault category , perform detailed fault type judgment, including mild overheating, severe overheating, low discharge, high discharge, etc. If the fault probability is greater than 0.8, this fault category is output.

[0162] S7. Perform comprehensive judgment based on the segmentation judgment result and the neural network segmentation judgment result to obtain the final fault type.

[0163] In this embodiment, the comprehensive judgment includes:

[0164] Based on the segmented judgment results of the three-ratio method, case-based reasoning method, and neural network obtained in steps S5 and S6, a weighted random forest model is used to make a final fault type judgment and determine the final fault.

[0165] Furthermore, the weighted random forest model includes:

[0166] Judgment result fusion:

[0167] ;

[0168] in:

[0169] : Judgment result after fusion;

[0170] : Weight coefficient, indicating the proportion of different judgment methods, the sum is 1;

[0171] : The judgment results of three judgment methods, including three-ratio judgment, case-based reasoning judgment, and neural network judgment;

[0172] Decision tree construction:

[0173] ;

[0174] in:

[0175] : kth decision tree;

[0176] : The training sample of the kth tree is randomly sampled from the judgment results;

[0177] Calculate the classification probability of each decision tree:

[0178] ;

[0179] in:

[0180] : The predicted probability of the kth decision tree for the fault category;

[0181] : All fault categories;

[0182] Voting mechanism to obtain the final classification results:

[0183] ;

[0184] in:

[0185] : Final fault category prediction result;

[0186] : The total number of decision trees.

[0187] By integrating the results of multiple judgment methods, a decision tree is constructed to determine the final fault.

[0188] This method achieves more stable and accurate transformer fault classification by integrating the results of multiple judgment methods and combining the ensemble learning ability of random forest, enhances the fault identification ability, and improves the classification stability and accuracy.

[0189] Uses of this model include:

[0190] Result fusion: According to steps S5 and S6, the judgment results of the three methods are obtained and fused to obtain the fused judgment result. ;

[0191] Decision tree construction: Decision tree is constructed based on the fusion judgment results to obtain the decision tree ;

[0192] Final fault judgment: Calculate the classification probability of each decision tree based on the constructed decision tree , voting mechanism to obtain the final classification results , determine the final fault.

[0193] Example 2:

[0194] This embodiment also provides a computer device, which is applicable to a diagnostic method for transformer operating status monitoring and early warning, including a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute computer-executable instructions to implement a diagnostic method for transformer operating status monitoring and early warning as proposed in the above embodiment.

[0195] This embodiment further provides a storage medium on which a computer program is stored. When the program is executed by a processor, a diagnostic method suitable for transformer operating status monitoring and early warning proposed in the above embodiment is implemented.

[0196] The computer device may be a terminal, comprising a processor, memory, a communication interface, a display, and an input device connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores an operating system and computer programs. The internal memory provides an environment for the operating system and computer programs stored in the non-volatile storage media. The communication interface of the computer device is used to communicate with external terminals via wired or wireless communication. Wireless communication may be achieved via Wi-Fi, a carrier network, NFC (near-field communication), or other technologies. The display of the computer device may be a liquid crystal display or an electronic ink display. The input device may be a touchscreen overlay on the display, buttons, a trackball, or a touchpad on the computer device housing, or an external keyboard, touchpad, or mouse.

[0197] If a function is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the portion that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the various embodiments of the method of the present invention. The aforementioned storage medium includes various media that can store program code, such as USB flash drives, mobile hard drives, read-only memories (ROMs), random access memories (RAMs), magnetic disks, or optical disks.

[0198] The logic and / or steps represented in the flowcharts or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing the logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (e.g., a computer-based system, a system including a processor, or other system that can fetch and execute instructions from an instruction execution system, apparatus, or device). For purposes of this specification, a "computer-readable medium" can be any device that can contain, store, communicate, propagate, or transport a program for use by, or in conjunction with, an instruction execution system, apparatus, or device.

[0199] More specific examples (a non-exhaustive list) of computer-readable media include the following: an electrical connection with one or more wires (electronic devices), a portable computer disk cartridge (magnetic devices), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), a fiber optic device, and a portable compact disc read-only memory (CDROM). In addition, the computer-readable medium may even be paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium and then editing, interpreting, or processing it in another suitable manner as necessary, and then storing it in a computer memory.

[0200] It should be understood that various parts of the present invention can be implemented using hardware, software, firmware, or a combination thereof. In the above-described embodiments, multiple steps or methods can be implemented using software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented using hardware, as in another embodiment, any one of the following technologies known in the art or a combination thereof can be used: a discrete logic circuit having logic gate circuits for implementing logic functions on data signals, an application-specific integrated circuit having suitable combinational logic gate circuits, a programmable gate array (PGA), a field-programmable gate array (FPGA), etc.

[0201] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A diagnostic method suitable for transformer operating status monitoring and early warning, characterized in that: The following steps are involved: S1. Oil chromatography data collection and over-standard judgment to obtain judgment results; S2. Perform preliminary fault judgment based on the judgment result to obtain the preliminary judgment result; S3. Perform a neural network preliminary fault judgment based on the judgment result to obtain a neural network preliminary judgment result; S4. Perform comprehensive judgment based on the initial judgment result and the neural network initial judgment result to obtain the initial fault type; S5. Perform detailed fault judgment based on the initial fault type and obtain detailed judgment results; S6. Perform neural network segmentation fault judgment based on the initial fault type to obtain a neural network segmentation judgment result; S7. Perform comprehensive judgment based on the segmentation judgment result and the neural network segmentation judgment result to obtain the final fault type.

2. A diagnostic method for transformer operating status monitoring and early warning according to claim 1, characterized in that: The oil chromatography data collection and exceeding standard judgment in S1 include: Based on the oil chromatography data, the IEC60599 dissolved gas analysis standard is used to compare it with the preset fault judgment threshold to determine whether it exceeds the threshold. If it exceeds the threshold, the diagnostic logic is activated.

3. A diagnostic method for transformer operating status monitoring and early warning according to claim 1, characterized in that: The neural network initial fault judgment in S3 includes: After the diagnostic logic is started according to the judgment result, the initial fault judgment model is used to make an initial fault type judgment.

4. A diagnostic method for transformer operating status monitoring and early warning according to claim 3, characterized in that: The initial fault judgment model includes: Hidden layer neural network calculation: ； in: : The initial output of the hidden layer neurons; : activation function; : The connection weight from the input layer to the hidden layer; : original gas concentration; : Adaptive fault feature weight; : The initial bias term of the hidden layer neurons; Output layer fault category probability calculation: ； in: : The probability of initial classification of fault categories; , : The connection weight from the hidden layer to the output layer.

5. A diagnostic method for transformer operating status monitoring and early warning according to claim 4, characterized in that: The adaptive fault feature weights include: ； in: , : The rate of change of gas concentration.

6. A diagnostic method for transformer operating status monitoring and early warning according to claim 1, characterized in that: The neural network segmentation fault judgment in S6 includes: According to the initial fault type classification results, the fault type is further subdivided using the subdivision fault judgment model to obtain the neural network subdivision judgment results.

7. A diagnostic method for transformer operating status monitoring and early warning according to claim 6, characterized in that: The subdivided fault judgment model includes: Hidden layer calculation formula: ； in: : The segmented output of the hidden layer neurons; : gas concentration after adaptive normalization; : The subdivision bias term of the hidden layer neurons; Output layer calculation formula: ； in: : The probability of breaking down the fault categories.

8. A diagnostic method for transformer operating status monitoring and early warning according to claim 7, characterized in that: The adaptively normalized gas concentration includes: ； in: : the mean value of gas i; : standard deviation of gas i; : Weight coefficient.

9. A diagnostic method for transformer operating status monitoring and early warning according to claim 1, characterized in that: The comprehensive judgment in S7 includes: According to the segmented judgment results obtained in steps S5 and S6, a weighted random forest model is used to make a final fault type judgment and determine the final fault.

10. A diagnostic method for transformer operating status monitoring and early warning according to claim 9, characterized in that: The weighted random forest model includes: Judgment result fusion: ； in: : Judgment result after fusion; : weight coefficient; : Judgment results of three judgment methods; Decision tree construction: ； in: : kth decision tree; : training sample of the kth tree; Calculate the classification probability of each decision tree: ； in: : The predicted probability of the kth decision tree for the fault category; : All fault categories; Voting mechanism to obtain the final classification results: ； in: : Final fault category prediction result; : The total number of decision trees.

Images