A transformer bushing state comprehensive evaluation and fault diagnosis method based on mechanism-data fusion

Abstract

The application discloses a kind of transformer bushing state comprehensive evaluation and fault diagnosis method based on mechanism-data fusion, to multidimensional state characteristic data as foundation, by constructing the whole-process data preprocessing system based on physical mechanism, using Arrhenius thermal aging equation, henry law and other core mechanisms are decoupled and checked for reasonableness, obtain the standard characteristic value reflecting the intrinsic state of bushing;Establish the weighted naive Bayes evaluation model with physical level constraint, according to the characteristic physical property module processing and introducing weight constraint, break through the limitation of traditional independence assumption, realize the balance of data fitting and physical mechanism;Introduce the particle swarm optimization algorithm with mechanism penalty mechanism to realize the global optimization of feature weight, and combine the fault characteristic fingerprint method based on feature contribution degree, on the basis of state grade evaluation accurately locate fault type and degradation mechanism, significantly improve the engineering practicability and explainability of evaluation result, provide direct decision support for field operation and maintenance.

Description

Technical Field

[0001] This invention relates to the technical field of transformer bushing diagnosis, and in particular to a method for comprehensive evaluation and fault diagnosis of transformer bushing status based on mechanism-data fusion. Background Technology

[0002] Power transformers are core hub equipment in the power transmission and distribution chain of a power system, and their safe and stable operation directly determines the reliability of the power grid. Transformer bushings are critical components connecting the internal windings of the transformer to the external transmission lines, undertaking the core functions of current carrying and ground insulation. They are among the core components most prone to transformer failure. Statistics show that faults caused by bushing insulation deterioration, poor contact, and other defects account for more than 30% of all transformer faults. If these defects are not identified and assessed in a timely manner, they can develop into serious accidents such as bushing breakdown, transformer fires, and explosions, causing large-scale power outages and significant economic losses. Therefore, accurate and efficient condition assessment of transformer bushings is one of the core requirements of power equipment operation and maintenance.

[0003] Existing methods for identifying transformer bushing faults typically rely on a single threshold for fault determination, combined with Naive Bayes for evaluation. However, these methods only focus on outliers at the data level when processing transformer bushing data, neglecting the physical constraints and operating condition coupling effects of transformer bushing characteristic quantities. They also fail to consider the influence of operating condition parameters such as ambient temperature, operating oil temperature, and load current on core insulation characteristics such as dielectric loss factor and dissolved gases in oil. Consequently, these characteristic quantities cannot reflect the intrinsic state of bushing insulation. Furthermore, traditional Naive Bayes evaluation methods rely solely on fitting feature weights to pure data without explicit physical constraints, which can easily lead to weight inversion and evaluation results that do not conform to the physical logic of power equipment fault diagnosis. Limited by the strict feature condition independence assumption of traditional Naive Bayes, probability calculations suffer from systematic biases, resulting in insufficient evaluation accuracy. Moreover, they cannot further pinpoint specific fault types and degradation mechanisms, making it difficult to meet the actual operation and maintenance needs of engineering sites. Summary of the Invention

[0004] To address the shortcomings of existing technologies in transformer bushing assessment, such as insufficient accuracy and inability to precisely pinpoint specific fault types and degradation mechanisms, which hinders the fulfillment of actual operation and maintenance needs in engineering sites, this application provides a comprehensive assessment and fault diagnosis method for transformer bushing status based on mechanism-data fusion. This method can accurately pinpoint fault types and degradation mechanisms based on transformer bushing status level assessment, significantly improving the engineering applicability and interpretability of the assessment results and providing direct decision support for on-site operation and maintenance.

[0005] Firstly, the above-mentioned inventive objective of this application is achieved through the following technical solution: A method for comprehensive assessment and fault diagnosis of transformer bushing condition based on mechanism-data fusion, the method comprising: A multi-dimensional state feature of the transformer bushing is obtained to construct an original dataset. The original dataset is then subjected to mechanism rationality verification and decoupling correction of working condition physical properties to generate a bushing feature vector to be evaluated. Calculate the posterior probability of the casing feature vector, assign physical constraint weights to the casing feature vector, and construct a weighted Naive Bayes model for casing status evaluation. The state evaluation accuracy of the weighted Naive Bayes model is calculated. Combined with the preset mechanism constraint penalty value, a fitness function is constructed to perform optimal feature weight iteration processing to obtain the global optimal feature weight. Based on the global optimal weight, bushing status is classified, and the feature contribution score of the bushing feature vector under the fault status classification is calculated. Based on the calculation results, fault type classification is performed to obtain transformer bushing fault diagnosis data.

[0006] In a preferred embodiment, this application can be further configured as follows: the process of acquiring multi-dimensional state features of transformer bushings to construct an original dataset, performing mechanism rationality verification and decoupling correction on the original dataset, and generating the mechanism rationality verification process in the bushing feature vector to be evaluated specifically includes: The casing feature vectors are compared with preset mechanism constraint criteria, and the casing feature vectors are determined to be invalid data based on the comparison results. The mechanism constraint criteria include gas feature logic constraints, thermal feature physical constraints, and operating condition coupling. When the casing feature vector satisfies any one of the mechanism constraint criteria, it is determined to be invalid data and discarded.

[0007] In a preferred embodiment, this application can be further configured as follows: the process of acquiring multi-dimensional state features of transformer bushings to construct an original dataset, performing mechanism rationality verification and operating condition decoupling correction on the original dataset, and generating the operating condition decoupling correction process in the bushing feature vector to be evaluated, specifically includes: The original dataset is subjected to temperature correction processing for the dielectric loss factor, and the temperature correction expression is as follows: (1) in, To convert the dielectric standard value to a reference temperature of 20℃, tanδ meas The dielectric loss factor was measured on-site. E a The activation energy of oil paper insulation, R The molar gas constant, T insThe measured thermodynamic temperature of the bushing insulation; The original dataset is subjected to oil temperature correction processing for dissolved gases in the oil. The oil temperature correction expression is as follows: (2) in, X i,meas gaseous components i The actual measured volume fraction on site, k i ( T ) is a gaseous component i oil temperature T Henry coefficient under, T oil,meas The measured thermodynamic temperature of the top layer oil.

[0008] In a preferred embodiment, this application can be further configured as follows: the process of calculating the posterior probability of the casing feature vector, assigning physical constraint weights to the casing feature vector, and constructing a weighted Naive Bayes model for casing state assessment includes: The posterior probability of the sleeve feature vector is calculated using formula (3), which is shown below: (3) in, P ( c j | x ) represents the posterior probability. P ( x | c j ) is the state c j Lower eigenvectors x The conditional probability, P ( x ) is the normalized evidence factor. P ( c j ) is the state c j The prior probability is calculated as follows: (4) in, N This represents the total number of eigenvectors of the casing. N j The state in the casing feature vector is c j The number of samples; For each casing feature vector, physical constraint weights are assigned. Conditional probabilities are calculated based on the posterior probability and the assigned physical constraint weights. A weighted Naive Bayes model is then constructed based on the calculation results. The expression for the conditional probability calculation is as follows: (5) in, w i For the first i The weights of the eigenvectors of each sleeve, P ( x i | c j ) is the state c j Next i The characteristic vector value of each sleeve is x i The conditional probability density, M The total number of physical modules. P ( x m | c j ) represents the physical module m medium state c j Below m The joint conditional probability of the eigenvector values ​​of each sleeve.

[0009] In a preferred embodiment, this application can be further configured as follows: the process of calculating the posterior probability of the casing feature vector, assigning physical constraint weights to the casing feature vector, and constructing a weighted Naive Bayes model for casing state assessment includes: The conditional probability density is estimated using a normal distribution, and the expression for the conditional probability density estimation is as follows: (6) in, μ ji , σ ji These are the training set states. c j In the sample i The mean and standard deviation of each feature value; Based on the conditional probability estimation results, the maximum posterior probability is calculated, and the maximum posterior probability is selected as the state level of the corresponding pipe sleeve feature vector for pipe sleeve state evaluation and classification. The quantification expression of the pipe sleeve state is as follows: (7) in, P ( xi | c j () represents the estimated value of the conditional probability density. n This indicates the number of eigenvectors of the casing. P ( c j ) represents the maximum probability of delay.

[0010] In a preferred embodiment, this application can be further configured as follows: The calculation of the state evaluation accuracy of the weighted Naive Bayes model, combined with a preset mechanistic constraint penalty value, constructs a fitness function for iterative processing of the optimal feature weights to obtain the globally optimal feature weights, specifically including: Calculate the state evaluation accuracy of the weighted Naive Bayes model, set accuracy weight coefficients and mechanistic constraint penalty values ​​based on the state evaluation accuracy, and construct a fitness function. The expression of the fitness function is as follows: (8) in, Fit α is the fitness value; α is the accuracy weighting coefficient. Indicates the accuracy of the status assessment. N correct To evaluate the correct number of samples for the model, N total This represents the total number of samples used in the model validation. Penalty Indicates the penalty value for mechanism constraints; Obtain the historical optimal position of each casing feature vector and the historical global optimal position of the population. Under the constraint of the fitness function, iteratively update each casing feature vector. When the preset iteration termination condition is met, the global optimal weight of each pipe feature vector under the optimal weight is obtained.

[0011] In a preferred embodiment, this application can be further configured as follows: classifying bushing states based on the globally optimal weights, calculating the feature contribution score of the bushing feature vector under the fault state classification, and classifying fault types based on the calculation results to obtain transformer bushing fault diagnosis data, specifically including: Obtain the casing status classification result corresponding to the globally optimal weight, calculate the feature contribution score of the casing feature vector classified as a fault state, and construct the ranking matrix related to the dominant features of each fault state. The expression for calculating the feature contribution score is as follows: (9) in, express The characteristic contribution score of the state, Represents the characteristic vector of the casing The globally optimal weight, , These represent the characteristic vectors of the tubing. In the current state Log-likelihood value and normal state The log-likelihood value is given below. Based on the sorting matrix and the physical fault logic of each dominant feature, fault type inference and classification are performed to obtain transformer bushing fault diagnosis data.

[0012] Secondly, the above-mentioned inventive objective of this application is achieved through the following technical solutions: A system for comprehensive assessment and fault diagnosis of transformer bushing condition based on mechanism-data fusion, the system comprising: The data acquisition and preprocessing module is used to acquire multi-dimensional state characteristics of transformer bushings to construct an original dataset, perform mechanism rationality verification and decoupling correction of working condition physical properties on the original dataset, and generate bushing feature vectors to be evaluated. The model building module is used to calculate the posterior probability of the casing feature vector, assign physical constraint weights to the casing feature vector, and build a weighted Naive Bayes model for casing status evaluation. The data calculation module is used to calculate the state evaluation accuracy of the weighted Naive Bayes model, and combine it with the preset mechanism constraint penalty value to construct a fitness function for optimal feature weight iteration processing to obtain the global optimal feature weight. The fault diagnosis module is used to classify the bushing status according to the global optimal weight, calculate the feature contribution score of the bushing feature vector under the fault status classification, classify the fault type based on the calculation results, and obtain transformer bushing fault diagnosis data.

[0013] Thirdly, the above-mentioned objectives of this application are achieved through the following technical solutions: A computer device includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the above-described method for comprehensive evaluation and fault diagnosis of transformer bushing status based on mechanism-data fusion.

[0014] Fourthly, the above-mentioned objectives of this application are achieved through the following technical solutions: A computer-readable storage medium storing a computer program that, when executed by a processor, implements the aforementioned method for comprehensive evaluation and fault diagnosis of transformer bushing status based on mechanism-data fusion.

[0015] In summary, this application includes at least one of the following beneficial technical effects: 1. A full-process preprocessing system for multi-dimensional characteristics of transformer bushings based on physical mechanisms was constructed. It integrates the 3σ statistical criterion and the invalid data dual elimination method based on core mechanism constraints. At the same time, based on the Arrhenius thermal aging equation and Henry's law, temperature correction of dielectric loss factor and oil temperature correction of dissolved gas in oil were completed respectively. The operating conditions of characteristic quantities were decoupled, the interference of environmental and operating condition fluctuations on insulation characteristic quantities was eliminated, and standard characteristic values ​​that can truly reflect the intrinsic state of bushing insulation were obtained, laying a high-quality data foundation for subsequent evaluation models. 2. A particle swarm optimization algorithm with a mechanism penalty is proposed. The mechanism constraint violation penalty term is embedded in the fitness function. The compliance of physical constraints is checked simultaneously during the weight optimization process. This ensures that the optimal weight obtained by optimization not only meets the high accuracy requirements of data classification, but also fully conforms to the physical logic of insulation degradation of power equipment. This avoids model overfitting and greatly improves the generalization and engineering robustness of the model. 3. A fault feature fingerprinting method based on feature contribution scores was constructed. By calculating the contribution ranking of each feature under abnormal conditions and matching it with a preset fault feature fingerprint database, the specific fault type and deterioration physical mechanism can be accurately located on the basis of completing the casing condition level assessment. This enables full-process diagnosis from "condition level assessment" to "precise fault location". 4. To address the shortcomings of existing Naive Bayes models, such as the limitation of conditional independence assumptions and the tendency for weight inversion due to the lack of physical constraints on weights, this invention proposes a weighted Naive Bayes model with physical hierarchical constraints. By processing data in modules according to physical attributes, it satisfies the conditional independence assumption of Bayes' theorem and adapts to the physical coupling relationship between features, thus solving the systematic bias problem in the probability calculation of traditional models. Furthermore, by setting weight constraints at three physical levels, the weight boundaries of core insulation features, operational features, and environmental features are clearly defined, fundamentally eliminating the problem of weight inversion. This ensures that the model's evaluation results fully conform to the physical logic of power equipment fault diagnosis. Compared to purely data-driven models, it exhibits stronger generalization capabilities and maintains high accuracy in bushing samples across different voltage levels and operating conditions, significantly improving its engineering practicality. 5. To address the shortcomings of existing technologies that can only assess the state level but cannot accurately pinpoint the fault type, this invention achieves integrated state assessment and fault diagnosis through the fault feature fingerprint method. After identifying abnormal / critical states, the dominant degradation features can be quickly located through feature contribution scores, matching the fault type with the physical mechanism, and providing corresponding maintenance strategies, which greatly improves the efficiency of on-site operation and maintenance and the accuracy of decision-making. Attached Figure Description

[0016] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the accompanying drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. In all the drawings, similar elements or parts are generally identified by similar reference numerals. In the drawings, the elements or parts are not necessarily drawn to scale.

[0017] Figure 1 This is a flowchart illustrating the overall data processing of the transformer bushing condition comprehensive assessment and fault diagnosis method in this embodiment.

[0018] Figure 2 This is a flowchart illustrating the implementation of the transformer bushing condition comprehensive evaluation and fault diagnosis method in this embodiment.

[0019] Figure 3 This is a curve of the PSO algorithm optimization iteration for the transformer bushing condition comprehensive evaluation and fault diagnosis method in this embodiment.

[0020] Figure 4 This is a confusion matrix comparison diagram of the test results of the comprehensive evaluation and fault diagnosis method for transformer bushing status in this embodiment.

[0021] Figure 5 This is a verification and analysis diagram of the fault feature fingerprint method for the comprehensive evaluation and fault diagnosis method of transformer bushing status in this embodiment.

[0022] Figure 6 This is a structural block diagram of the transformer bushing condition comprehensive assessment and fault diagnosis system in this embodiment.

[0023] Figure 7 This is a schematic diagram of the internal structure of a computer device used to implement a comprehensive assessment of transformer bushing conditions and a fault diagnosis method. Detailed Implementation

[0024] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0025] It should be understood that, when used in this specification and the appended claims, the terms "comprising" and "including" indicate the presence of the described features, integrals, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or collections thereof.

[0026] It should also be understood that the terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the invention. As used in this specification and the appended claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise.

[0027] It should also be further understood that the term "and / or" as used in this specification and the appended claims refers to any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.

[0028] In one embodiment, such as Figure 1 As shown, this invention proposes a comprehensive evaluation and fault diagnosis method for transformer bushing conditions based on mechanism-data fusion. Based on multi-dimensional state characteristic data, it first constructs a full-process data preprocessing system based on physical mechanisms, utilizing core mechanisms such as the Arrhenius thermal aging equation and Henry's law to decouple operating conditions and verify their rationality, obtaining standard characteristic values ​​reflecting the intrinsic state of the bushing. Then, a weighted Naive Bayes evaluation model with physical hierarchy constraints is established, processing features by physical attributes and introducing weight constraints to overcome the limitations of traditional independence assumptions and achieve a balance between data fitting and physical mechanisms. Finally, a particle swarm optimization algorithm with a mechanism penalty mechanism is introduced to achieve global optimization of feature weights, combined with a fault feature fingerprint method based on feature contribution, to accurately locate fault types and degradation mechanisms based on state level evaluation, significantly improving the engineering practicality and interpretability of the evaluation results, providing direct decision support for on-site operation and maintenance. Specifically, as shown... Figure 2 As shown in the figure, the method for comprehensive evaluation and fault diagnosis of transformer bushing condition based on mechanism-data fusion in this embodiment specifically includes the following steps: S1: Obtain multi-dimensional state features of transformer bushings to construct the original dataset, perform mechanism rationality verification and decoupling correction of working condition physical properties on the original dataset, and generate bushing feature vectors to be evaluated.

[0029] Specifically, environmental characteristic values ​​(ambient temperature) are obtained through real-time and historical data from the transformer substation's online monitoring system, annual preventative test data, laboratory bushing aging and defect simulation test data, and measured data from typical on-site fault cases. x 1. Ambient relative humidity x 2) Operating characteristic values ​​(bushing load current) x 3. Transformer top oil temperature x 4) Condition monitoring characteristic value (buffer medium loss factor) x 5. Total hydrocarbon production rate x 6. Volume fraction of C2H2 x 7. Volume fraction of H2 x8) To eliminate the impact of outliers, missing values, and dimensional differences on the model, the original dataset was preprocessed. This included using the 3σ criterion to remove outliers from the original data; that is, when the value of a feature exceeds the mean ± 3 times the standard deviation of that feature, it is identified as an outlier and removed. Missing values ​​were filled using linear interpolation to ensure the integrity of the dataset. After data preprocessing, the original dataset underwent mechanistic rationality verification and decoupling correction of working conditions.

[0030] The mechanism rationality verification process specifically includes: S11: Compare the casing feature vector with the preset mechanism constraint criteria respectively, and determine whether the casing feature vector is invalid data based on the comparison results. The mechanism constraint criteria include gas feature logic constraint, thermal feature physical constraint and operating condition coupling.

[0031] Specifically, the gas characteristic logic constraints in this embodiment include: samples with C2H2 volume fraction > 0 but total hydrocarbon production rate of 0 are determined to be invalid data; samples with H2 volume fraction exceeding the standard but no other hydrocarbon gas growth are determined to be invalid data.

[0032] The physical constraints on thermal characteristics in this embodiment include: samples with a load current of 0 but a transformer top oil temperature far exceeding the ambient temperature are considered invalid data; samples with a sudden increase or decrease in dielectric loss factor > 0.003 under the same operating condition within the 20~100℃ range are considered invalid data.

[0033] The operating condition coupling constraints in this embodiment include: samples where the load current is negatively correlated with the top oil temperature and samples where the ambient temperature is negatively correlated with the top oil temperature are considered invalid data.

[0034] S12: When the casing feature vector satisfies any one of the mechanism constraint criteria, it is determined to be invalid data and discarded.

[0035] If the feature vector of the casing satisfies any one of the mechanism constraint criteria, the corresponding data is determined to be invalid data and removed. After removing invalid data, the missing values ​​in the dataset are supplemented by linear interpolation to ensure the integrity of the dataset.

[0036] The physical decoupling correction process under operating conditions in this embodiment specifically includes: S13: Perform temperature correction processing on the original dataset for the dielectric loss factor. The temperature correction expression is as follows: (1) in, To convert the dielectric standard value to a reference temperature of 20℃, tanδ meas The dielectric loss factor was measured on-site.E a The activation energy of oil-paper insulation is generally taken as 88 kJ / mol. R The molar gas constant is 8.314 J·mol⁻¹. -1 ·K -1 , T ins The measured thermodynamic temperature of the bushing insulation.

[0037] In this embodiment, the measured tanδ at different temperatures is converted to the standard value at a reference temperature of 20℃ based on the Arrhenius thermal aging equation of dielectric physics, so as to perform temperature correction on the dielectric loss factor.

[0038] S14: Perform oil temperature correction processing on the original dataset for dissolved gases in the oil. The oil temperature correction expression is as follows: (10) in, X i,meas gaseous components i The actual measured volume fraction on site, k i ( T ) is a gaseous component i oil temperature T Henry coefficient under, T oil,meas The measured thermodynamic temperature of the top layer oil.

[0039] In this embodiment, based on Henry's Law, the measured gas volume fraction at different oil temperatures is converted to the standard value at a reference oil temperature of 40℃, thereby eliminating the influence of oil temperature on the gas dissolution-precipitation equilibrium.

[0040] After verifying the rationality of the mechanism and decoupling the physical conditions of the working condition from the original dataset, all feature quantities are normalized to generate the casing feature vector to be evaluated. The feature normalization expression is as follows: (2) in, x i For the first i The original values ​​of each feature, x imax , x imin The first i The maximum and minimum values ​​of each feature in the dataset; x′ i These are the normalized eigenvalues.

[0041] In this embodiment, the bushing insulation status is divided into four states: "normal," "caution," "abnormal," and "serious." This embodiment uses... c 1. c 2. c 3. c 4 represents and specifies the corresponding operation and maintenance strategy for each state. The fault state level and maintenance strategy in this embodiment are shown in Table 1: Table 1 S2: Calculate the posterior probability of the casing feature vector, assign physical constraint weights to the casing feature vector, and construct a weighted Naive Bayes model to evaluate the casing status.

[0042] Specifically, this embodiment adopts c j ( j =1,2,3,4) represent the four status levels of transformer bushings: normal, warning, abnormal, and severe, respectively. Let the characteristic vector of the bushing to be evaluated be... x =( x 1, x 2,..., x n ),in n The total number of input features. x i For the first i The normalized values ​​of each feature are used to calculate the posterior probability according to Bayes' theorem. The model construction process in step S2 includes: S21: Calculate the posterior probability of the sleeve feature vector using formula (3), which is shown below: (3) in, P ( c j | x ) represents the posterior probability. P ( x | c j ) is the state c j Lower eigenvectors x The conditional probability, P ( x () is the normalized evidence factor, which is the same for all state levels. P ( c j ) is the state c j The prior probability is calculated as follows: (4) in, N This represents the total number of eigenvectors of the casing. N j The state in the casing feature vector is c j The number of samples.

[0043] S22: Assign physical constraint weights to each sleeve feature vector, calculate conditional probabilities based on the posterior probability and the physical constraint weights, and construct a weighted Naive Bayes model based on the calculation results. The expression for calculating the conditional probability is shown below: (5) in, w i For the first i The weights of each casing feature vector are assigned; the larger the weight, the greater the influence of the casing features on the evaluation results. P ( x i | c j ) is the state c j Next i The characteristic vector value of each sleeve is x i The conditional probability density, M The total number of physical modules. P ( x m | c j ) represents the physical module m medium state c j Below m The joint conditional probability of the eigenvector values ​​of each sleeve.

[0044] In this embodiment, the system is divided into three modules based on characteristic physical attributes: environment module. m 1=( x 1, x 2) Running module m 2=( x 3, x 4) Insulation status module m 3=( x 5, x 6, x 7, x 8) The modules satisfy the conditional independence assumption, and the modules within the module satisfy the conditional independence assumption. P ( x m | c jThe joint conditional probability of the modules is calculated based on the physical coupling relationship within the modules, overcoming the limitations of the single feature conditional independence assumption. The physical constraints of the feature weights in this embodiment are shown in Table 2. Table 2 In this embodiment, to overcome the limitations of the feature conditional independence assumption in traditional Naive Bayes and to address the shortcomings of pure data weights lacking physical constraints and prone to weight inversion, feature weights with physical upper and lower bound constraints are introduced. w i ( w i ∈[0,1]), construct a weighted Naive Bayes classification model.

[0045] Specifically, the pipe condition assessment process in step S2 includes: S23: Estimate the conditional probability density using a normal distribution. The expression for the conditional probability density estimation is shown below: (6) in, μ ji , σ ji These are the training set states. c j In the sample i The mean and standard deviation of each feature value.

[0046] In this embodiment, since the characteristic quantities of the casing are continuous values ​​and the statistical data in the project basically conform to the normal distribution law, the normal distribution probability density function is used for estimation. P ( x i | c j ).

[0047] S24: Based on the conditional probability estimation results, the maximum posterior probability is calculated. The maximum posterior probability is selected as the state level of the corresponding pipe casing feature vector for pipe casing state evaluation and classification. The pipe casing state quantification expression is as follows: (7) in, P ( x i | c j () represents the estimated value of the conditional probability density. n This indicates the number of eigenvectors of the casing. P ( c j ) represents the maximum probability of delay.

[0048] S3: Calculate the state evaluation accuracy of the weighted Naive Bayes model, combine it with the preset mechanism constraint penalty value, construct the fitness function to perform optimal feature weight iteration processing, and obtain the global optimal feature weight.

[0049] Specifically, to obtain the optimal weights of each feature in the weighted Naive Bayes model, a particle swarm optimization algorithm is used for global optimization, with the model's state evaluation accuracy as the fitness function. Mechanistic constraints are embedded in the optimization process to ensure that the optimal weights simultaneously satisfy data fitting accuracy and physical mechanism consistency. Step S3 includes: In this embodiment, the particle swarm optimization algorithm is used for optimal weight optimization. Before optimization, parameter initialization is required, where each sleeve feature vector is set as a particle, and the position vector of each particle is set as a set of feature weights. w =( w 1, w 2,..., w n Initialize the particle swarm size, maximum number of iterations, and inertia factor. λ acceleration constant c 1 and c 2. Initialize the velocity and position of all particles.

[0050] S31: Calculate the state assessment accuracy of the weighted Naive Bayes model, set the accuracy weight coefficients and mechanism constraint penalty values ​​based on the state assessment accuracy, and construct the fitness function. The fitness function expression is as follows: (8) in, Fit α is the fitness value; α is the accuracy weighting coefficient, which is set to 0.85 in this embodiment. Indicates the accuracy of the status assessment. N correct To evaluate the correct number of samples for the model, N total This represents the total number of samples used in the model validation. Penalty The penalty value represents the mechanism constraint. For each violation, such as exceeding the upper or lower limit of the weight, weight inversion, or distribution parameters not conforming to the physical trend, the penalty value is increased by 1. The maximum penalty value is equal to the total number of features.

[0051] Specifically, in this embodiment, the accuracy of the weighted Naive Bayes model in evaluating the state of the training samples is taken as the core, and a penalty term for violation of mechanistic constraints is added to construct a fitness function that balances data fitting and physical rationality.

[0052] S32: Obtain the historical optimal position of each casing feature vector and the historical global optimal position of the population. Under the constraint of the fitness function, iteratively update each casing feature vector.

[0053] Specifically, in each iteration of the particle, each particle, i.e., the sleeve feature vector, is determined based on its own historical best position. p i with the global optimal position of the group p g Under the constraint of the fitness function, that is, the particle updates and iterates to satisfy the fitness function, updating its own velocity and position in an iterative update.

[0054] The speed update formula is as follows: (2) The position update formula is as follows: (3) in, v i ( t ), z i ( t ) are respectively the first t During the nth iteration i The velocity vector and position vector of each particle; r 1. r 2 is a random number in the range [0,1].

[0055] S33: When the preset iteration termination condition is reached, the global optimal weight of each tube feature vector under the optimal weight is obtained.

[0056] Specifically, the iteration is terminated when the number of iterations reaches the preset maximum number of iterations, or when the optimal fitness value of the population does not improve in consecutive iterations. The global optimal position vector at this time is the optimal weight of each feature.

[0057] S4: Classify the bushing status according to the global optimal weight, calculate the feature contribution score of the bushing feature vector under the fault status classification, classify the fault type based on the calculation results, and obtain transformer bushing fault diagnosis data.

[0058] Specifically, step S4 includes: S41: Obtain the casing status classification result corresponding to the globally optimal weight, calculate the feature contribution score of the casing feature vector classified as a fault state, and construct the ranking matrix related to the dominant features of each fault state. The expression for calculating the feature contribution score is as follows: (9) in, express The characteristic contribution score of the state, Represents the characteristic vector of the casing The globally optimal weight, , These represent the characteristic vectors of the tubing. In the current state Log-likelihood value and normal state The log-likelihood value.

[0059] Specifically, the state level corresponding to the maximum posterior probability calculated from the globally optimal weights is used as the current pipeline state classification result. The pipe casing feature vectors under "abnormal" and "severe" fault states are then used to determine the specific fault type using the fault feature fingerprint method. The feature contribution score is used to measure the feature's contribution. x 1. In the current state c j Log-likelihood value under normal state c 1 The difference between the log-likelihood values ​​is calculated. A larger difference indicates a greater contribution of that feature to the transition of the transformer bushing condition from "normal" to "severe". By calculating the feature contribution scores of all features, a ranking matrix can be constructed, as shown in Table 3. Table 3 S42: Based on the sorting matrix and the physical fault logic of each dominant feature, perform fault type inversion and classification to obtain transformer bushing fault diagnosis data.

[0060] Specifically, based on the sorting matrix and the fault physical process interpretation of each dominant feature or combination of dominant features, i.e., physical fault logic, specific fault types are reversed and classified to obtain transformer bushing fault diagnosis results.

[0061] In one embodiment, a total of 200 data samples (50 samples for each of the four states) are used to simulate real-world scenarios involving environmental interference and feature coupling in field monitoring data. This dataset poses a significant challenge to the traditional Naive Bayes model. During the data preprocessing stage, outlier removal and min-max normalization are performed according to the above scheme, ultimately dividing the dataset into a training set of 140 samples and a test set of 60 samples. Addressing the inherent limitation of the traditional Naive Bayes model's "feature equal weighting," the PSO algorithm with L2 regularization is used to optimize the weights of eight core features (ambient temperature, ambient relative humidity, bushing load current, transformer top oil temperature, bushing dielectric loss factor, total hydrocarbon production rate, C2H2 volume fraction, and H2 volume fraction). The PSO algorithm's optimization iteration curve is shown below. Figure 3 As shown, the regularization constraint effectively avoids model overfitting, and the test results are as follows. Figure 4 As shown.

[0062] The results show that the fitness value stabilizes (0.8168) after 80 iterations, indicating good convergence in the optimization process. Test set validation results show that traditional Naive Bayes, unable to distinguish feature importance, achieves only 70.00% accuracy on highly polluted data; while the weighted model of this invention, through adaptive optimization of feature weights, improves the test set accuracy to 85.00%, representing an accuracy improvement of 15.00%. Will Figure 4 The analysis focused on 13 abnormal and 13 severe transformer bushing fault samples, fully replicating the diagnostic logic for three common types of bushing faults encountered in the field. Figure 5 As shown.

[0063] The measured data change in Case 1 shows that C2H2 increased from 0 μL / L to 30 μL / L, leading to the conclusion that the fault type was high-energy arc discharge and the physical process was instantaneous high-temperature breakdown in the oil.

[0064] The measured data changes in Case 2 show that the humidity increased from 50% to 90% and the tanδ value increased from 0.3% to 2.0%. The conclusion is that the fault type is insulation dampness and external contamination, and the physical process is seal failure and moisture intrusion.

[0065] Case 3: Measured data changes: Load increased from 0.5 to 0.9 pu, total hydrocarbons increased from 0.3 to 20 μL / L·d. Conclusion: Fault type is poor contact overheating, physical process is contact surface heating, leading to thermal decomposition of hydrocarbons in the oil.

[0066] The successful derivation of the three cases verifies the effectiveness of this method.

[0067] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0068] In one embodiment, a transformer bushing condition comprehensive assessment and fault diagnosis system based on mechanism-data fusion is provided. This system corresponds one-to-one with the transformer bushing condition comprehensive assessment and fault diagnosis method based on mechanism-data fusion described in the above embodiments. For example... Figure 6 As shown, the transformer bushing condition comprehensive assessment and fault diagnosis system based on mechanism-data fusion includes a data acquisition and preprocessing module, a model building module, a data calculation module, and a fault diagnosis module. Detailed descriptions of each functional module are as follows: The data acquisition and preprocessing module is used to acquire multi-dimensional state characteristics of transformer bushings to construct the original dataset, perform mechanism rationality verification and decoupling correction of operating conditions on the original dataset, and generate bushing feature vectors to be evaluated.

[0069] The model building module is used to calculate the posterior probability of the casing feature vector, assign physical constraint weights to the casing feature vector, and build a weighted Naive Bayes model for casing status evaluation.

[0070] The data calculation module is used to calculate the state evaluation accuracy of the weighted Naive Bayes model. It combines the preset mechanism constraint penalty value to construct a fitness function and perform iterative processing of the optimal feature weights to obtain the globally optimal feature weights.

[0071] The fault diagnosis module is used to classify bushing status according to the global optimal weight, calculate the feature contribution score of the bushing feature vector under the fault status classification, classify the fault type based on the calculation results, and obtain transformer bushing fault diagnosis data.

[0072] Preferably, the mechanism rationality verification in the data acquisition and preprocessing module specifically includes: The physical logic constraint submodule is used to compare the bushing feature vector with the preset mechanism constraint criteria, and determine whether the bushing feature vector is invalid data based on the comparison results. The mechanism constraint criteria include gas feature logic constraints, thermal feature physical constraints, and operating condition coupling.

[0073] The invalid data removal submodule is used to determine and remove invalid data when the casing feature vector satisfies any one of the mechanism constraint criteria.

[0074] Preferably, the decoupling correction process for operating conditions in the data acquisition and preprocessing module specifically includes: The dielectric loss factor correction submodule is used to perform temperature correction processing on the dielectric loss factor of the original dataset. The temperature correction expression is shown below: (1) in, To convert the dielectric standard value to a reference temperature of 20℃, tanδ meas The dielectric loss factor was measured on-site. E a The activation energy of oil paper insulation, R The molar gas constant, T ins The measured thermodynamic temperature of the bushing insulation.

[0075] The oil temperature correction submodule is used to correct the oil temperature of dissolved gases in the original dataset. The oil temperature correction expression is shown below: (2) in, X i,meas gaseous components i The actual measured volume fraction on site, k i ( T ) is a gaseous component i oil temperature T Henry coefficient under, T oil,meas The measured thermodynamic temperature of the top layer oil.

[0076] Preferably, the model building process in the model building module includes: The posterior probability of the sleeve feature vector is calculated using formula (3), which is shown below: (3) in, P ( c j | x ) represents the posterior probability. P ( x | c j ) is the state c j Lower eigenvectors x The conditional probability, P ( x ) is the normalized evidence factor. P ( c j ) is the state c j The prior probability is calculated as follows: (4) in, N This represents the total number of eigenvectors of the casing. N j The state in the casing feature vector is c j The number of samples.

[0077] For each casing feature vector, physical constraint weights are assigned. Conditional probabilities are calculated based on the posterior probability and the assigned physical constraint weights. A weighted Naive Bayes model is then constructed based on the calculation results. The expression for the conditional probability calculation is shown below: (5) in, w i For the first i The weights of the eigenvectors of each sleeve,P ( x i | c j ) is the state c j Next i The characteristic vector value of each sleeve is x i The conditional probability density, M The total number of physical modules. P ( x m | c j ) represents the physical module m medium state c j Below m The joint conditional probability of the eigenvector values ​​of each sleeve.

[0078] Preferably, the pipe condition assessment process in the model building module includes: The conditional probability density is estimated using a normal distribution. The expression for the conditional probability density estimation is shown below: (6) in, μ ji , σ ji These are the training set states. c j In the sample i The mean and standard deviation of each feature value.

[0079] Based on the conditional probability estimation results, the maximum posterior probability is calculated. The maximum posterior probability is selected as the state level of the corresponding pipe sheath feature vector for pipe sheath state evaluation and classification. The pipe sheath state quantification expression is as follows: (7) in, P ( x i | c j () represents the estimated value of the conditional probability density. n This indicates the number of eigenvectors of the casing. P ( c j ) represents the maximum probability of delay.

[0080] Preferably, the data calculation module specifically includes: The fitness function construction submodule is used to calculate the state evaluation accuracy of the weighted Naive Bayes model. Based on the state evaluation accuracy, it sets the accuracy weight coefficients and the mechanistic constraint penalty value, and constructs the fitness function. The fitness function expression is shown below: (8) in, Fit α is the fitness value; α is the accuracy weighting coefficient. Indicates the accuracy of the status assessment. N correct To evaluate the correct number of samples for the model, N total This represents the total number of samples used in the model validation. Penalty This represents the penalty value for mechanism constraints.

[0081] The iterative update submodule is used to obtain the historical best position of each casing feature vector and the historical global best position of the population. Under the constraint of the fitness function, the feature vector of each casing is iteratively updated.

[0082] The iteration termination determination submodule is used to obtain the global optimal weight of each pipe feature vector under the optimal weight when the preset iteration termination condition is met.

[0083] Preferably, the fault diagnosis module specifically includes: The feature contribution calculation submodule is used to obtain the casing status classification result corresponding to the globally optimal weight, calculate the feature contribution score of the casing feature vector classified as a fault state, and construct the ranking matrix related to the dominant features of each fault state. The expression for calculating the feature contribution score is as follows: (9) in, express The characteristic contribution score of the state, Represents the characteristic vector of the casing The globally optimal weight, , These represent the characteristic vectors of the tubing. In the current state Log-likelihood value and normal state The log-likelihood value.

[0084] The fault type inversion submodule is used to perform fault type inversion and classification based on the sorting matrix and the physical fault logic of each dominant feature to obtain transformer bushing fault diagnosis data.

[0085] Specific limitations regarding the mechanism-data fusion-based transformer bushing condition comprehensive assessment and fault diagnosis system can be found in the limitations of the mechanism-data fusion-based transformer bushing condition comprehensive assessment and fault diagnosis method described above, and will not be repeated here. Each module in the aforementioned mechanism-data fusion-based transformer bushing condition comprehensive assessment and fault diagnosis system can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the memory of a computer device as software, so that the processor can call and execute the corresponding operations of each module.

[0086] In one embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 7 As shown, the computer device includes a processor, memory, network interface, and database connected via a system bus. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and database. The internal memory provides the environment for the operation of the operating system and computer programs in the non-volatile storage media. The database stores relevant data for transformer bushing condition assessment and fault diagnosis. The network interface communicates with external terminals via a network connection. When the computer program is executed by the processor, it implements a mechanism-data fusion-based method for comprehensive transformer bushing condition assessment and fault diagnosis.

[0087] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements a method for comprehensive evaluation and fault diagnosis of transformer bushing status based on mechanism-data fusion.

[0088] Those skilled in the art will recognize that the units of the various examples described in connection with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application of the technical solution and the constraints involved. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of the invention.

[0089] In the embodiments provided by the present invention, it should be understood that the division of units is only a logical functional division. In actual implementation, there may be other division methods, such as multiple units can be combined into one unit, one unit can be split into multiple units, or some features can be ignored.

[0090] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0091] If the integrated unit 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, in essence, or the part that contributes to the prior art, or all or part 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 to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, read-only memory (ROM), random access memory (RAM), portable hard drives, magnetic disks, or optical disks.

[0092] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention, and they should all be covered within the scope of the claims and specification of the present invention.

Claims

1. A method for comprehensive assessment and fault diagnosis of transformer bushing condition based on mechanism-data fusion, characterized in that, The method includes: A multi-dimensional state feature of the transformer bushing is obtained to construct an original dataset. The original dataset is then subjected to mechanism rationality verification and decoupling correction of working condition physical properties to generate a bushing feature vector to be evaluated. Calculate the posterior probability of the casing feature vector, assign physical constraint weights to the casing feature vector, and construct a weighted Naive Bayes model for casing status evaluation. The state evaluation accuracy of the weighted Naive Bayes model is calculated. Combined with the preset mechanism constraint penalty value, a fitness function is constructed to perform optimal feature weight iteration processing to obtain the global optimal feature weight. Based on the global optimal weight, bushing status is classified, and the feature contribution score of the bushing feature vector under the fault status classification is calculated. Based on the calculation results, fault type classification is performed to obtain transformer bushing fault diagnosis data.

2. The method for comprehensive evaluation and fault diagnosis of transformer bushing condition based on mechanism-data fusion according to claim 1, characterized in that, The process of acquiring multi-dimensional state features of transformer bushings to construct an original dataset, performing mechanism rationality verification and decoupling correction on the original dataset to generate the mechanism rationality verification process in the bushing feature vector to be evaluated, specifically includes: The casing feature vectors are compared with preset mechanism constraint criteria, and the casing feature vectors are determined to be invalid data based on the comparison results. The mechanism constraint criteria include gas feature logic constraints, thermal feature physical constraints, and operating condition coupling. When the casing feature vector satisfies any one of the mechanism constraint criteria, it is determined to be invalid data and discarded.

3. The method for comprehensive evaluation and fault diagnosis of transformer bushing condition based on mechanism-data fusion according to claim 1, characterized in that, The process of acquiring multi-dimensional state features of transformer bushings to construct an original dataset, performing mechanistic rationality verification and operational condition decoupling correction on the original dataset, and generating the operational condition decoupling correction process in the bushing feature vector to be evaluated, specifically includes: The original dataset is subjected to temperature correction processing for the dielectric loss factor, and the temperature correction expression is as follows: （1） in, To convert the dielectric standard value to a reference temperature of 20℃, tanδ meas The dielectric loss factor was measured on-site. E a The activation energy of oil paper insulation, R The molar gas constant, T ins The measured thermodynamic temperature of the bushing insulation; The original dataset is subjected to oil temperature correction processing for dissolved gases in the oil. The oil temperature correction expression is as follows: （2） in, X i,meas gaseous components i The actual measured volume fraction on site, k i ( T ) is a gaseous component i oil temperature T Henry coefficient under, T oil,meas The measured thermodynamic temperature of the top layer oil.

4. The method for comprehensive evaluation and fault diagnosis of transformer bushing condition based on mechanism-data fusion according to claim 1, characterized in that, The process of calculating the posterior probability of the casing feature vector, assigning physical constraint weights to the casing feature vector, and constructing a weighted Naive Bayes model for casing condition assessment includes: The posterior probability of the sleeve feature vector is calculated using formula (3), which is shown below: （3） in, P ( c j | x ) represents the posterior probability. P ( x | c j ) is the state c j Lower eigenvectors x The conditional probability, P ( x ) is the normalized evidence factor. P ( c j ) is the state c j The prior probability is calculated as follows: （4） in, N This represents the total number of eigenvectors of the casing. N j The state in the casing feature vector is c j The number of samples; For each casing feature vector, physical constraint weights are assigned. Conditional probabilities are calculated based on the posterior probability and the assigned physical constraint weights. A weighted Naive Bayes model is then constructed based on the calculation results. The expression for the conditional probability calculation is as follows: （5） in, w i For the first i The weights of the eigenvectors of each sleeve, P ( x i | c j ) is the state c j Next i The characteristic vector value of each sleeve is x i The conditional probability density, M The total number of physical modules. P ( x m | c j ) represents the physical module m medium state c j Below m The joint conditional probability of the eigenvector values ​​of each sleeve.

5. The method for comprehensive evaluation and fault diagnosis of transformer bushing condition based on mechanism-data fusion according to claim 4, characterized in that, The process of calculating the posterior probability of the casing feature vector, assigning physical constraint weights to the casing feature vector, and constructing a weighted Naive Bayes model for casing status assessment includes: The conditional probability density is estimated using a normal distribution, and the expression for the conditional probability density estimation is as follows: （6） in, μ ji , σ ji These are the training set states. c j In the sample i The mean and standard deviation of each feature value; Based on the conditional probability estimation results, the maximum posterior probability is calculated, and the maximum posterior probability is selected as the state level of the corresponding pipe sleeve feature vector for pipe sleeve state evaluation and classification. The quantification expression of the pipe sleeve state is as follows: （7） in, P ( x i | c j () represents the estimated value of the conditional probability density. n This indicates the number of eigenvectors of the casing. P ( c j ) represents the maximum probability of delay.

6. The method for comprehensive evaluation and fault diagnosis of transformer bushing condition based on mechanism-data fusion according to claim 1, characterized in that, The process of calculating the state evaluation accuracy of the weighted Naive Bayes model, combining a preset mechanistic constraint penalty value, constructing a fitness function, and performing iterative processing on the optimal feature weights to obtain the globally optimal feature weights specifically includes: Calculate the state evaluation accuracy of the weighted Naive Bayes model, set accuracy weight coefficients and mechanistic constraint penalty values ​​based on the state evaluation accuracy, and construct a fitness function. The expression of the fitness function is as follows: （8） in, Fit α is the fitness value; α is the accuracy weighting coefficient. Indicates the accuracy of the status assessment. N correct To evaluate the correct number of samples for the model, N total This represents the total number of samples used in the model validation. Penalty Indicates the penalty value for mechanism constraints; Obtain the historical optimal position of each casing feature vector and the historical global optimal position of the population. Under the constraint of the fitness function, iteratively update each casing feature vector. When the preset iteration termination condition is met, the global optimal weight of each pipe feature vector under the optimal weight is obtained.

7. The method for comprehensive evaluation and fault diagnosis of transformer bushing condition based on mechanism-data fusion according to claim 1, characterized in that, The process of classifying bushing states based on the globally optimal weights, calculating the feature contribution score of the bushing feature vector under the fault state classification, and classifying fault types based on the calculation results to obtain transformer bushing fault diagnosis data specifically includes: Obtain the casing status classification result corresponding to the globally optimal weight, calculate the feature contribution score of the casing feature vector classified as a fault state, and construct the ranking matrix related to the dominant features of each fault state. The expression for calculating the feature contribution score is as follows: （9） in, express The characteristic contribution score of the state, Represents the characteristic vector of the casing The globally optimal weight, , These represent the characteristic vectors of the tubing. In the current state Log-likelihood value and normal state The log-likelihood value is given below. Based on the sorting matrix and the physical fault logic of each dominant feature, fault type inference and classification are performed to obtain transformer bushing fault diagnosis data.

8. A comprehensive evaluation and fault diagnosis system for transformer bushing condition based on mechanism-data fusion, characterized in that, The system includes: The data acquisition and preprocessing module is used to acquire multi-dimensional state characteristics of transformer bushings to construct an original dataset, perform mechanism rationality verification and decoupling correction of working condition physical properties on the original dataset, and generate bushing feature vectors to be evaluated. The model building module is used to calculate the posterior probability of the casing feature vector, assign physical constraint weights to the casing feature vector, and build a weighted Naive Bayes model for casing status evaluation. The data calculation module is used to calculate the state evaluation accuracy of the weighted Naive Bayes model, and combine it with the preset mechanism constraint penalty value to construct a fitness function for optimal feature weight iteration processing to obtain the global optimal feature weight. The fault diagnosis module is used to classify the bushing status according to the global optimal weight, calculate the feature contribution score of the bushing feature vector under the fault status classification, classify the fault type based on the calculation results, and obtain transformer bushing fault diagnosis data.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the method for comprehensive evaluation and fault diagnosis of transformer bushing status based on mechanism-data fusion as described in any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the mechanism-data fusion-based comprehensive evaluation and fault diagnosis method for transformer bushing status as described in any one of claims 1 to 7.

Images