Transformer fault diagnosis method based on whale optimization algorithm improved fuzzy clustering
By introducing whale optimization algorithm in fuzzy clustering analysis, dynamically updating the cluster center and introducing adaptive weights, the problems of initial value sensitivity and local extreme value points in fuzzy clustering analysis are solved, and the accuracy of transformer fault diagnosis is improved and Clustering effect.
Patent Information
- Application Number
- CN202410952634.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-16
- Publication Date
- 2024-11-22
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
The existing fuzzy clustering analysis methods have initial value sensitivity and local extreme value points problems in transformer fault diagnosis, which limits the analysis effect of DGA data.
Using the fuzzy clustering method based on the improved whale optimization algorithm, the fuzzy clustering model is improved by constructing whale optimization algorithm, dynamically update the clustering center, introduce adaptive weights and learning factors, balance global and local searches, avoid premature convergence, and improve clustering Class effects.
It effectively improves the accuracy of transformer fault diagnosis, can more accurately identify transformer fault types, avoid local extreme points, and improves the comprehensiveness and accuracy of clustering results.
Smart Images

Figure CN119004156A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of transformer fault diagnosis, and in particular to a transformer fault diagnosis method based on improved fuzzy clustering using a whale optimization algorithm. Background Art
[0002] With the development of power systems, transformers, as an important part of power transmission and transformation systems, have an important impact on the stability and safety of power systems. However, due to the influence of various factors, transformers are prone to various faults. Common fault types include low temperature faults, medium temperature faults, high temperature faults, partial discharges, low energy discharges and arc discharges. Therefore, timely and accurate diagnosis of transformer fault types is of great significance to ensure the stable operation of power systems.
[0003] Dissolved Gas Analysis (DGA) in transformer oil can timely detect the potential fault risk of transformers and avoid serious accidents caused by transformer faults. There are five main gases commonly used to judge the internal fault type of transformers: hydrogen (H2), methane (CH4), ethane (C2H6), ethylene (C2H4), and acetylene (C2H2). In the research field, the three ratio method, characteristic gas method, David triangle method and other methods are commonly used to judge the fault type of transformers, but these methods have problems such as too absolute boundaries and low fault identification accuracy. In machine learning models that use DGA data as feature parameter mining, optimization models such as artificial neural networks and support vector machines are used for transformer fault diagnosis, but these methods require a large number of samples to be trained, are usually more complicated, and are limited in practical applications. Fuzzy clustering analysis methods have good application prospects in DGA data analysis, but fuzzy clustering analysis methods have problems such as initial value sensitivity and local extreme points, which limit the analysis effect of DGA data. Summary of the invention
[0004] In order to solve the problems of initial value sensitivity and local extreme points in DGA data analysis of fuzzy clustering analysis method, the present invention provides a method for transformer fault diagnosis based on whale optimization algorithm to improve fuzzy clustering, which can effectively improve the accuracy of transformer fault diagnosis.
[0005] The technical solution adopted by the present invention is:
[0006] The transformer fault diagnosis method based on improved fuzzy clustering based on whale optimization algorithm includes the following steps:
[0007] Step S1: Collect data of dissolved gas in transformer oil as an initial data set;
[0008] Step S2: Initialize the initial data set;
[0009] Step S3: construct a whale optimization algorithm to improve the fuzzy clustering model and obtain the cluster center;
[0010] Step S4: diagnose the transformer fault type and determine the fault result.
[0011] In step S1, the initial DGA data set of the transformer includes: hydrogen dissolved in the oil when the transformer fails (H 2 ), methane (CH 4 ), ethane (C 2 H 6 ), ethylene (C 2 H 4 ), acetylene (C 2 H 2 ) gas content, the data includes six types: low temperature fault, medium temperature fault, high temperature fault, partial discharge, spark discharge and arc discharge.
[0012] In step S2, the similarity of DGA data of the same fault type is mainly reflected in the similarity of gas composition, that is, the proportion of various gases is similar. It is necessary to initialize the DGA data to obtain the proportion of gases. The initialization process includes:
[0013] As shown in formula (1), calculate the percentage of hydrogen in hydrogen hydrocarbons;
[0014]
[0015] in, is the proportion of hydrogen; Hydrogen (H 2 ) content; Methane (CH 4 ) content; For ethane (C 2 H 6 ) content; Ethylene (C 2 H 4 ) content; For acetylene (C 2 H 2 ) content.
[0016] As shown in formula (2), the percentage of various hydrocarbon gases in the total hydrocarbons is calculated;
[0017]
[0018] Among them, P i Indicates the percentage of the four hydrocarbon gases in the total hydrocarbons; the subscript i indicates CH 4 , C 2 H 6 , C2 H 4 , C 2 H 2 Any one of .
[0019] Fault types with similar hydrogen contents are classified into one category, and fault types with large differences in hydrogen contents are classified into different categories.
[0020] The step S3 comprises the following steps:
[0021] S3.1: Given the number of cluster centers c, fuzzy index m, population size N, maximum number of iterations T, and maximum speed v max and error value ε;
[0022] S3.2: Calculate the fitness of each whale, the expression is:
[0023]
[0024] Where fitness(i) is the fitness of the whale; J m (U, V) represents the clustering objective function after the introduction of fuzzy parameters; i is the number of iterations, J m is the objective function, U represents the membership matrix, and V represents the cluster center vector;
[0025] S3.3: Update the position of each whale to generate the next generation of whales. The position update expression is:
[0026] D 1 =|CX * (t)-X(t)| (4);
[0027] X(t+1)=X * (t)-AD 1 (5);
[0028] A=2a r -a (6);
[0029] C=2r 2 (7);
[0030] a=2-2t / t max (8);
[0031] Among them, D 1 is the distance between the whale and the current optimal individual; A and C are coefficient vectors used to adjust the position update of the individual; t is the number of iterations; X * (t) is the position vector of the current optimal solution; X(t) is the position of the individual whale, r 1 、r 2is a randomly generated number between [0,1], a is the convergence factor, which decreases from 2 to 0 as the number of iterations increases; t max is the maximum number of iterations.
[0032] X(t+1) represents the whale position after iteration t+1.
[0033] The design of adaptive weight ω adjustment is shown in formula (9):
[0034]
[0035] Among them, ω initial ,ω final are the initial and final values of the inertia weight respectively; m is the influencing factor of the change speed.
[0036] After introducing the adaptive weight ω, the whale's position update formula is as follows:
[0037]
[0038] Among them, D 2 is the distance between the whale and the current optimal individual; b is the constant that constrains the spiral shape; l is a randomly generated number between [-1,1], p is a randomly generated number between [0,1], D rand is the distance from other whales to the target whale; X rand is a vector representing the position of a random individual whale.
[0039] Learning factor c 1 and c 2 The design of is shown in equations (11) and (12):
[0040] c 1 =rand()×e(11);
[0041] c 2 =e-rand() (12);
[0042] In formula (11) and formula (12), rand() is a random function that varies in the range [0,1], and e is a natural constant;
[0043] S3.4: Continuously update and iterate the whales to find the optimal fitness value for individuals and the optimal fitness value for the group;
[0044] For each whale, its updated individual fitness value fitness is calculated by formula (3). id (i+1) and its previous generation individual fitness value fitness id (i) Compare, if the individual fitness value of the whale is updated id(i+1) is greater than the individual fitness value of the previous generation id (i), then take it as the optimal fitness value of the individual at this time id (i+1), otherwise it remains the same, and the optimal fitness value of the individual is shown in formula (13);
[0045] fitness id (i+1)=max{fitness id (i), fitness id (i+1)} (13);
[0046] In formula (13), i is the number of iterations, fitness id (i+1) is the individual fitness value of the whale after update, fitness id (i) is the individual fitness value of the whale before updating;
[0047] For each whale, its updated individual fitness value fitness is calculated by formula (3). id (i+1) and the optimal fitness value of the group gd (i) Compare, if the updated fitness value is id (i+1) is greater than the optimal fitness value of the group gd (i) is taken as the optimal fitness value of the group at this time. gd (i+1), otherwise it remains the same. The optimal fitness value of the group is shown in formula (14):
[0048] fitness gd (i+1)=max{fitness gd (i), fitness id (i+1)} (14);
[0049] Among them, i is the number of iterations, fitness gd (i+1) is the group fitness value after particle update, fitness gd (i) is the fitness value of the group before the particle update, fitness id (i+1) is the individual fitness value of the particle after update.
[0050] S3.5: Increase the number of iterations. When the algorithm reaches the maximum number of iterations, the algorithm stops, otherwise jump to S3.2;
[0051] S3.6: Output initial cluster centers;
[0052] S3.7: The generated cluster centers are clustered using the FCM algorithm. The model of the FCM algorithm is shown in the following formula:
[0053]
[0054] Among them, J m (U, V) represents the clustering objective function after the introduction of fuzzy parameters; x i is the i-th data to be clustered; v j represents the jth cluster center; u ij The i-th data x to be clustered i The membership degree belonging to the jth cluster center, c is the number of cluster centers, n is the data x i The number of, m is the fuzzy parameter, m = 2, U is the membership matrix, V is the cluster center vector, d ij Represents the i-th sample element x i With the jth cluster center v j The difference between
[0055] According to the constraints in formula (15), the Lagrangian function is constructed using the Lagrangian multiplier method:
[0056]
[0057] In formula (16), λ is the Lagrange multiplier, for the cluster center v in formula (16) j and membership u ij Calculate partial derivatives respectively and get the cluster center v j and membership u ij The update method is as follows:
[0058]
[0059] In formula (17), Represents the i-th data x to be clustered i The membership degree belonging to the jth cluster center; m is the fuzzy parameter;
[0060]
[0061] In formula (18), k>0 and k≠j;
[0062] Using sample x i The distance d from the cluster center ij Construct the sub-similarity p of sample i to the jth class ijk , as shown below:
[0063] d ij =x i -v j (19);
[0064]
[0065] Where: d ij For sample x i Distance from cluster center v j The distance; p ijk Sample x i With cluster center v j The similarity between them; e is a natural constant;
[0066] The sample x i The sub-similarity p ijk Multiply them together to form the sample x i The comprehensive similarity p ij ; As shown below:
[0067]
[0068] Where c is the number of cluster centers, and k is a constant greater than 0;
[0069] Using similarity p ij The sample x i The proportion of the sum of similarities of all cluster centers is taken as the membership degree u ij , rewrite formula (18) as:
[0070]
[0071] Among them, p ik Represents sample x i For the kth cluster center v k The similarity of, c is the number of cluster centers;
[0072] S3.8: Update the cluster center v by iteration j and membership u ij , when the stop condition is met, the algorithm ends. The algorithm ends expression
[0073] The formula is:
[0074]
[0075] Where k is an integer greater than 0, ε is the error value, is the membership degree of the k+1th iteration, is the membership degree of the kth iteration.
[0076] S3.9: When the condition after the algorithm runs satisfies formula (23), the algorithm stops and the cluster center v of the transformer DGA data is obtained. j .
[0077] In step S4, the cluster center v of the DGA data set obtained in step S3 is j Performing fault diagnosis on the transformer and determining the fault type of the transformer to be diagnosed includes the following distributed steps:
[0078] S4.1: Calculate the dissolved gas data x in the transformer oil to be diagnosed by formula (19) i With cluster center v j The difference between the gases in
[0079] S4.2: Calculate the dissolved gas data x in the transformer oil to be diagnosed by formula (20) in step S3 i With cluster center v j The sub-similarity p between each gas in ij ;
[0080] S4.3: Calculate the dissolved gas data x in the transformer oil to be diagnosed by using equations (21) and (22) in step S3 i For cluster center v j The membership degree u ij ;
[0081] S4.4: After obtaining the membership degree u ij In the above equation, the membership degree u ij Corresponding to various fault types of transformers, the membership degree u ij The largest proportion of fault types is the fault type of the transformer to be diagnosed.
[0082] The present invention provides a method for transformer fault diagnosis based on whale optimization algorithm to improve fuzzy clustering, and the technical effects are as follows:
[0083] 1) In step 1 of the present invention, data on dissolved gases in various transformer oils that have failed are collected so that the algorithm can provide comprehensive clustering results for various transformer fault types.
[0084] 2) In step 2 of the present invention, the components of dissolved gas in transformer oil of the same fault type are similar. The data are normalized by calculating the percentage of hydrogen in hydrogen hydrocarbons and the percentage of various hydrocarbon gases in total hydrocarbons to improve the clustering effect of the algorithm.
[0085] 3) In step 3 of the present invention, by dynamically and randomly changing the learning factor c 1 and c 2 , which can enable the algorithm to jump out of the local optimal solution and find the global optimal solution; the design of the adaptive weight ω can balance the global and local search, avoid premature convergence, and improve the accuracy of the transformer fault diagnosis results.
[0086] 4) In step 4 of the present invention, the degree of membership is improved by using the similarity between the dissolved gas in the transformer oil to be diagnosed and each gas in the cluster center, and the transformer fault type is diagnosed according to the degree of membership, thereby improving the accuracy of the transformer fault diagnosis result. BRIEF DESCRIPTION OF THE DRAWINGS
[0087] The present invention is further described in detail below with reference to the accompanying drawings and specific embodiments:
[0088] Figure 1 is a flow chart of the present invention;
[0089] Figure 2 The flowchart of the whale optimization algorithm of the present invention is used to improve the fuzzy clustering algorithm. DETAILED DESCRIPTION
[0090] The present invention is further described below in conjunction with the accompanying drawings and specific embodiments.
[0091] A certain number of transformer oil dissolved gas data with known fault types are selected, covering six fault types: low-temperature overheating, medium-temperature overheating, high-temperature overheating, partial discharge, arc discharge and spark discharge, to form a transformer oil dissolved gas data set.
[0092] According to the above steps S2 and S3, the cluster center of the fault is obtained, as shown in Table 1.
[0093] Table 1 Cluster centers of faults
[0094]
[0095] Specific example: The oil chromatogram data of a faulty transformer is H 2 =54,CH 4 =23.4,C 2 H 6 =4.8,C 2 H 4 =24,C 2 H 2 =78, use this method to diagnose the fault, the steps are as follows:
[0096] a. According to formula (1) and formula (2), the data is normalized to calculate the percentage of hydrogen in hydrogen hydrocarbons and the percentage of various hydrocarbon gases in total hydrocarbons, and the data obtained are:
[0097]
[0098] b. According to formula (19), the difference between the data of the above-mentioned fault to be diagnosed and each fault gas in Table 1 is calculated. The results are shown in Table 2.
[0099] Table 2 The difference between the data of the fault to be diagnosed and the various fault gases in Table 1
[0100]
[0101] From Table 2, we can see the difference between the data to be diagnosed and the cluster center. The higher the content of different gases to be diagnosed, the greater the difference.
[0102] c. According to formula (20), the sub-similarity between the data to be diagnosed and each gas in the cluster center is calculated, as shown in Table 3.
[0103] Table 3 Sub-similarity between the data to be diagnosed and each gas in the cluster center
[0104]
[0105] From Table 3, we can see the sub-similarity between the data to be diagnosed and the gases in the cluster center. The sub-similarity decreases as the difference increases.
[0106] d. According to formula (21), the comprehensive similarity between the data to be diagnosed and each gas in the cluster center is calculated, as shown in Table 4
[0107] Table 4 Comprehensive similarity between the data to be diagnosed and each gas in the cluster center
[0108]
[0109] From Table 4, we can see the comprehensive similarity between the data to be diagnosed and each gas in the cluster center.
[0110] e. According to formula (22), the comprehensive similarity between the data to be diagnosed and each gas in the cluster center is normalized to obtain the membership degree, as shown in Table 5.
[0111] Table 5 Membership degree of the data to be diagnosed to various faults
[0112]
[0113] From Table 5, we can see the degree of membership of the data to be diagnosed to different types of faults, and thus the fault judgment of the transformer can be obtained.
[0114] f. According to the membership, the spark discharge has the highest membership of 0.6599, followed by the arc discharge with a membership of 0.3214. The transformer has a discharge fault and tends to be spark discharge.
[0115] The above specific embodiments of the present invention are only for clarifying the present invention, rather than limiting the implementation methods of the present invention. Those skilled in the art may modify the techniques and solutions described in the embodiments. Any modification, equivalent substitution, improvement, etc. made within the spirit and principle of the present invention shall be included in the scope of the present invention.
Claims
1. A transformer fault diagnosis method based on improved fuzzy clustering based on whale optimization algorithm, characterized in that The following steps are involved: Step S1: Collect data of dissolved gas in transformer oil as an initial data set; Step S2: Initialize the initial data set; Step S3: construct a whale optimization algorithm to improve the fuzzy clustering model and obtain the cluster center; Step S4: diagnose the transformer fault type and determine the fault result; In step S1, the DGA initial data set of the transformer Including: the gas content of hydrogen (H2), methane (CH4), ethane (C2H6), ethylene (C2H4), and acetylene (C2H2) dissolved in the oil when the transformer fails. The data includes six types: low-temperature fault, medium-temperature fault, high-temperature fault, partial discharge, spark discharge and arc discharge.
2. The transformer fault diagnosis method based on improved fuzzy clustering based on whale optimization algorithm according to claim 1 is characterized in that: In step S2, the similarity of DGA data of the same fault type is reflected in the similarity of gas composition, that is, the proportion of various gases is similar, and the DGA data is initialized to obtain the proportion of gases; the initialization process includes: As shown in formula (1), calculate the percentage of hydrogen in hydrogen hydrocarbons; in, is the proportion of hydrogen; is the content of hydrogen (H2); is the content of methane (CH4); is the content of ethane (C2H6); is the content of ethylene (C2H4); is the content of acetylene (C2H2); As shown in formula (2), the percentage of various hydrocarbon gases in the total hydrocarbons is calculated; Among them, P i Indicates the percentage of four hydrocarbon gases in the total hydrocarbons; the subscript i represents any one of CH4, C2H6, C2H4, and C2H2.
3. The method for transformer fault diagnosis based on improved fuzzy clustering based on whale optimization algorithm according to claim 1 is characterized in that: The step S3 comprises the following steps: S3.1: Given the number of cluster centers c, fuzzy index m, population size N, maximum number of iterations T, and maximum speed v max and error value ε; S3.2: Calculate the fitness of each whale, the expression is: Where fitness(i) is the fitness of the whale; J m (U, V) represents the clustering objective function after the introduction of fuzzy parameters; i is the number of iterations, J m is the objective function, U represents the membership matrix, and V represents the cluster center vector; S3.3: Update the position of each whale to generate the next generation of whales. The position update expression is: D1=|CX * (t)-X(t)| (4); X(t+1)=X * (t)-AD1 (5); A = 2ar1-a (6); C=2r2 (7); a=2-2t / t max (8); Where D1 is the distance between the whale and the current optimal individual; A and C are coefficient vectors used to adjust the position update of the individual; t is the number of iterations; X * (t) is the position vector of the current optimal solution; X(t) is the position of the individual whale, r1 and r2 are randomly generated numbers between [0,1], and a is the convergence factor; t max is the maximum number of iterations; X(t+1) represents the whale position after iteration t+1; The design of adaptive weight adjustment is shown in formula (9): in, t , are the initial and final values of the inertia weight respectively; m is the influencing factor of the change speed; The whale's position update formula after introducing adaptive weights is as follows: Among them, is the distance between the whale and the current optimal individual; 8 is the constant constraining the spiral shape; 9 is a randomly generated number between -1 and 1, (is a randomly generated number between 0 and 1, 5 6 is the distance from other whales to the target whale; 5 6 is a vector representing the position of a random individual whale; The design of learning factors c1 and c2 is shown in equations (11) and (12): c1 = rand() × e (11); c2 = e-rand() (12); In formula (11) and formula (12), rand() is a random function that varies in the range [0,1], and e is a natural constant; S3.4: Continuously update and iterate the whales to find the optimal fitness value for individuals and the optimal fitness value for the group; For each whale, its updated individual fitness value fitness is calculated by formula (3). id (i+1) and its previous generation individual fitness value fitness id (i) Compare, if the individual fitness value of the whale is updated id (i+1) is greater than the individual fitness value of the previous generation id (i), then take it as the optimal fitness value of the individual at this time id (i+1), otherwise it remains the same, and the optimal fitness value of the individual is shown in formula (13); fitness id (i+1)=max{fitness id (i),fitness id (i+1)}(13); In formula (13), i is the number of iterations, fitness id (i+1) is the individual fitness value of the whale after update, fitness id (i) is the individual fitness value of the whale before updating; For each whale, its updated individual fitness value fitness is calculated by formula (3). id (i+1) and the optimal fitness value of the group gd (i) Compare, if the updated fitness value is id (i+1) is greater than the optimal fitness value of the group gd (i) is taken as the optimal fitness value of the group at this time. gd (i+1), otherwise it remains the same. The optimal fitness value of the group is shown in formula (14): fitness gd (i+1)=max{fitness gd (i),fitness id (i+1)}(14); Among them, i is the number of iterations, fitness gd (i+1) is the group fitness value after particle update, fitness gd (i) is the fitness value of the group before the particle update, fitness id (i+1) is the individual fitness value of the particle after update; S3.5: Increase the number of iterations. When the algorithm reaches the maximum number of iterations, the algorithm stops, otherwise jump to S3.2; S3.6: Output initial cluster centers; S3.7: The generated cluster centers are clustered using the FCM algorithm. The model of the FCM algorithm is shown in the following formula: Among them, J m (U, V) represents the clustering objective function after the introduction of fuzzy parameters; x i is the i-th data to be clustered; v j represents the jth cluster center; u ij The i-th data x to be clustered i The membership degree belonging to the jth cluster center, c is the number of cluster centers, n is the data x i The number of, m is the fuzzy parameter, m = 2, U is the membership matrix, V is the cluster center vector, d ij Represents the i-th sample element x i With the jth cluster center v j The difference between According to the constraints in formula (15), the Lagrangian function is constructed using the Lagrangian multiplier method: In formula (16), λ is the Lagrange multiplier, for the cluster center v in formula (16) j and membership u ij Calculate partial derivatives respectively and get the cluster center v j and membership u ij The update method is as follows: In formula (17), Represents the i-th data x to be clustered i The membership degree belonging to the jth cluster center; m is the fuzzy parameter; In formula (18), k>0 and k≠j; Using sample x i The distance d from the cluster center ij Construct the sub-similarity p of sample i to the jth class ijk , as shown below: d ij =x i -v j (19); Where: d ij For sample x i Distance from cluster center v j distance; :; Sample x i With cluster center v j The similarity between them; e is a natural constant; The sample x i The sub-similarity p ijk Multiply them together to form the sample x i The comprehensive similarity p ij ; As shown below: Where c is the number of cluster centers, and k is a constant greater than 0; Using similarity p ij The sample x i The proportion of the sum of similarities of all cluster centers is taken as the membership degree u ij , rewrite formula (18) as: Among them, p ik Represents sample x i For the kth cluster center v k The similarity of, c is the number of cluster centers; S3.8: Update the cluster center v by iteration j and membership u ij , when the stop condition is met, the algorithm ends; the algorithm ends expression The formula is: Where k is an integer greater than 0, ε is the error value, is the membership degree of the k+1th iteration, is the membership degree of the kth iteration; S3.9: When the condition after the algorithm runs satisfies formula (23), the algorithm stops and the cluster center v of the transformer DGA data is obtained. j .
4. The method for transformer fault diagnosis based on improved fuzzy clustering based on whale optimization algorithm according to claim 3 is characterized in that: In step S4, the cluster center v of the DGA data set obtained in step S3 is j Performing fault diagnosis on the transformer and determining the fault type of the transformer to be diagnosed includes the following distributed steps: S4.1: Calculate the dissolved gas data x in the transformer oil to be diagnosed by formula (19) i With cluster center v j The difference between the gases in S4.2: Calculate the dissolved gas data x in the transformer oil to be diagnosed by formula (20) in step S3 i With cluster center v j The sub-similarity p between each gas in ij ; S4.3: Calculate the dissolved gas data x in the transformer oil to be diagnosed by using equations (21) and (22) in step S3 i For cluster center v j The membership degree u ij ; S4.4: After obtaining the membership degree u ij In the above equation, the membership degree u ij Corresponding to various fault types of transformers, the membership degree u ij The largest proportion of fault types is the fault type of the transformer to be diagnosed.
Citation Information
Patent Citations
Method for optimizing turn-off angle of switched reluctance motor based on improved whale optimization algorithm
CN116743027A
Generalized nonlinear equivalent modeling method and system based on data driving
CN117195686A
Transformer fault diagnosis method based on particle swarm improved fuzzy clustering
CN118013330A
Belt conveyor bearing fault diagnosis method, system, equipment and medium
CN118094360A
Bearing fault diagnosis method based on fuzzy broad learning model
US11928613B1
Cited By
Intelligent 5G network equipment anomaly detection method and system
CN120011907A
An intelligent 5G network device anomaly detection method and system
CN120011907B
A transformer fault diagnosis method and device based on interpretable XGBoost and whale optimization and a storage medium
CN122796710A
