Wind power plant single fault tracing method and device

Abstract

The application provides a single fault tracing method and device for a wind power equipment, and relates to the field of fault analysis of wind power equipment. The method comprises the following steps: in response to a fault tracing operation for the wind power equipment, acquiring a state monitoring variable corresponding to a target subsystem; calculating a quantity control limit corresponding to the target subsystem through a PCA model; judging whether the wind power equipment is in a fault state according to the quantity control limit; if it is confirmed that the wind power equipment is in the fault state, acquiring a training sample representation, training a SOM network structure and weights through the training sample representation, and judging whether the target subsystem is a fault subsystem by inputting the state monitoring variable; if it is confirmed that the target subsystem is the fault subsystem, acquiring a historical fault component corresponding to the target subsystem, and taking the state monitoring variable and the historical fault component as nodes of a Bayesian network to construct a Bayesian network model; calculating respective conditional probabilities of the target subsystem being multiple fault modes through the Bayesian network model; if it is confirmed that a first conditional probability is greater than a second conditional probability, taking a fault mode corresponding to the first conditional probability as a fault source corresponding to the wind power equipment, and the first conditional probability and the second conditional probability are any two different probabilities in multiple different conditional probabilities. The application solves the problem that simple sensor alarm or historical data analysis cannot provide a clear tracing path, so that maintenance personnel can only try to find the fault source through repeated maintenance, thereby increasing the maintenance cost and downtime.

Description

Technical Field

[0001] This application relates to the field of wind power equipment fault analysis, and in particular to a method and apparatus for tracing the source of a single fault in wind power equipment. Background Technology

[0002] As one of the earliest developed and most important renewable energy sources in the integrated energy system, wind energy has enormous potential for greenhouse gas emission reduction due to its abundant reserves, renewability, and clean and pollution-free characteristics. It has become the mainstream direction of global renewable energy development and an important force in promoting industrial upgrading and driving energy system adjustment.

[0003] The existing wind power equipment has poor operational reliability due to the complex and variable working environment. Downtime caused by faults accounts for 25.6% of the rated power generation time. For wind turbines with an operating life of 20 years, maintenance costs account for about 20% to 25% of revenue. High operation and maintenance costs have become an important factor restricting the development of wind power, and there is an urgent need to develop efficient and reliable fault tracing technology.

[0004] Currently, common fault tracing methods in wind power equipment typically rely on sensor fault alarms and simple backtracking of historical data. While these methods can identify the time of fault occurrence and roughly indicate the fault area, the complex operating environment of wind power equipment makes it difficult for these traditional methods to accurately locate the specific fault source. For example, when multiple subsystems exist in a wind power unit and a fault occurs, the operation of each subsystem is often highly coupled, making it difficult for simple sensor alarms or historical data analysis to determine which subsystem is the initial fault source. This forces maintenance personnel to repeatedly inspect and attempt to find the initial fault source, thereby increasing maintenance costs and downtime.

[0005] Therefore, there is an urgent need for methods and devices for tracing the source of single faults in wind power equipment. Summary of the Invention

[0006] This application provides a method and device for tracing the source of a single fault in wind power equipment, which solves the problem that simple sensor alarms or historical data analysis cannot provide a clear tracing path, causing maintenance personnel to have to repeatedly inspect and try to find the source of the fault, thereby increasing maintenance costs and downtime.

[0007] The first aspect of this application provides a method for tracing the source of a single fault in wind power equipment. The method includes: in response to a fault tracing operation for wind power equipment, obtaining state monitoring variables corresponding to a target subsystem, wherein the target subsystem is any subsystem in the wind power equipment; calculating the quantity control constraints corresponding to the target subsystem using a PCA model; determining whether the wind power equipment is in a fault state based on the quantity control constraints; if the wind power equipment is confirmed to be in a fault state, obtaining training sample representations, training the SOM network structure and weights using the training sample representations, and determining whether the target subsystem is a faulty subsystem by inputting the state monitoring variables; if the target subsystem is confirmed to be a faulty subsystem, obtaining the historical faulty components corresponding to the target subsystem, and using the state monitoring variables and historical faulty components as Bayesian network nodes to construct a Bayesian network model; calculating the conditional probabilities corresponding to multiple fault modes of the target subsystem using the Bayesian network model, wherein each fault mode corresponds to one conditional probability; if it is confirmed that the first conditional probability is greater than the second conditional probability, taking the fault mode corresponding to the first conditional probability as the fault source corresponding to the wind power equipment, wherein the first conditional probability and the second conditional probability are any two different probabilities among multiple different conditional probabilities.

[0008] Optionally, obtain the state monitoring variables corresponding to the target subsystem: obtain the output variables corresponding to the target subsystem, including impeller speed variables, generator power variables, and wind angle variables; use the output variables as the state monitoring variables corresponding to the target subsystem.

[0009] Optionally, the quantity control limits include the quantity control limits corresponding to the first statistic and the second statistic. The quantity control limits corresponding to the target subsystem are calculated using the PCA model, specifically by calculating the first statistic using the following formula:

[0010]

[0011] Where SPE is the first statistic, P is a matrix composed of k eigenvectors, I is the identity matrix, and x new For the new sample, T is the transpose of the sample; then the control limits corresponding to the first statistic are calculated according to the following formula:

[0012]

[0013] Among them, Q α Let θ be the control limit corresponding to the first statistic. r The eigenvalues ​​corresponding to the sample covariance matrix are calculated using the following formula: c α Let h0 be the confidence limit of the standard normal distribution, and h0 be a distribution that follows θ. r The control limits are used to calculate the characteristic statistic, and the formula is as follows:

[0014] The second statistic is calculated using the following formula:

[0015]

[0016] Among them, T 2 The second statistic, Λ -1 To take the exponent of -1 for each non-zero element in the matrix, the control limits corresponding to the second statistic are calculated according to the following formula:

[0017]

[0018] Among them, T α The control limits for the second statistic are given, 1-α is the confidence level, and F0 is the value of the second statistic. α (a, na) is an F-distribution with a first degree of freedom and na second degree of freedom.

[0019] Optionally, the determination of whether the wind power equipment is in a fault state based on the quantity control limit includes: if it is confirmed that the first statistic is less than the quantity control limit corresponding to the first statistic and the second statistic is less than the quantity control limit corresponding to the second statistic, then the wind power equipment is determined not to be in a fault state; if it is confirmed that the first statistic is greater than or equal to the quantity control limit corresponding to the first statistic, or the second statistic is greater than or equal to the quantity control limit corresponding to the second statistic, then the wind power equipment is determined to be in a fault state.

[0020] Optionally, the SOM network structure and weights are trained using training sample representations, specifically including: S1, obtaining the first sample representation from the training sample representations, where the first sample representation is any sample representation from the training sample representations; S2, obtaining the winning neuron, which is the neuron with the smallest distance from the first sample representation among multiple output neurons; S3, obtaining the connection weight coefficients corresponding to the output neurons within the winning neighborhood, and adjusting the connection weight coefficients according to the following formula, where the winning neighborhood is the neighborhood where the winning neuron is located:

[0021] W ij (t+1)=W ij (t)+η(t)h c,j (t)(X i -W ij (t));

[0022] Among them, W ij (t+1) represents the updated weights of the output neurons, η(t) is the learning rate function, and 0 < η(t) < 1. The learning rate function is the gain function, h c,j (t) is the neighborhood function; S4, the neighborhood function is updated using the following formula:

[0023]

[0024] Where, d cj Let r be the Euclidean distance between the winning neuron and the activated output neurons in its neighborhood, r be the neighborhood radius, and t be the iteration number; S5, update the neighborhood radius using the following formula:

[0025]

[0026] Where r(t+1) is the updated neighborhood radius, INT is the floor function, and T is the total number of training iterations; S6, the initial learning rate is updated using the following formula:

[0027]

[0028] Where η(t+1) is the updated learning rate function and η(0) is the initial learning rate function; S7, let t = t+1, repeat steps S1-S6 until t = T or the neighborhood function converges, to complete the training of the SOM network structure and weights.

[0029] Optionally, by inputting state monitoring variables, it is determined whether the target subsystem is a faulty subsystem. Specifically, this includes: performing sensitivity analysis on the target subsystem based on the state monitoring variables and network slicing, whereby the sensitivity analysis is used to calculate the target sensitivity value corresponding to the target subsystem; determining whether the target sensitivity value is the largest sensitivity value among multiple sensitivity values, wherein the multiple sensitivity values ​​correspond one-to-one with multiple subsystems; if it is confirmed that the target sensitivity value is the largest sensitivity value among multiple sensitivity values, then the target subsystem is determined to be a faulty subsystem.

[0030] Optionally, the conditional probabilities corresponding to multiple failure modes of the target subsystem are calculated, specifically including: calculating the conditional probabilities using the following formula:

[0031]

[0032] Wherein, P(X) i |Y) represents the conditional probability, X1,…,X n It is a complete event set consisting of state monitoring variables and failure modes, and X i ,X j ∈{X1,…,X n}

[0033] The second aspect of this application provides a single fault tracing device for wind power equipment. The device includes a first construction module, a second construction module, a model building module, and an analysis and calculation module, wherein...

[0034] The first construction module is used to respond to the fault tracing operation for wind power equipment, obtain the state monitoring variables corresponding to the target subsystem, where the target subsystem is any subsystem in the wind power equipment; and calculate the quantitative control limits corresponding to the target subsystem through the PCA model.

[0035] The second construction module is used to determine whether the wind power equipment is in a fault state based on the quantity control constraints; if the wind power equipment is confirmed to be in a fault state, the module obtains training sample representations, trains the SOM network structure and weights through the training sample representations, and determines whether the target subsystem is a faulty subsystem by inputting state monitoring variables.

[0036] The model building module is used to obtain the historical faulty components corresponding to the target subsystem if the target subsystem is confirmed to be a faulty subsystem, and to use the state monitoring variables and historical faulty components as Bayesian network nodes to build a Bayesian network model.

[0037] The analysis and calculation module is used to calculate the conditional probabilities corresponding to multiple fault modes of the target subsystem using a Bayesian network model. Each fault mode corresponds to one conditional probability. If it is confirmed that the first conditional probability is greater than the second conditional probability, the fault mode corresponding to the first conditional probability is taken as the fault source of the wind power equipment. The first conditional probability and the second conditional probability are any two different probabilities among multiple different conditional probabilities.

[0038] A third aspect of this application provides an electronic device including a processor, a memory, a user interface, and a network interface. The memory is used to store instructions, the user interface and the network interface are used to communicate with other devices, and the processor is used to execute the instructions stored in the memory to cause the electronic device to perform the method as described above.

[0039] A fourth aspect of this application provides a computer-readable storage medium storing a computer program, which is executed by a processor according to any of the methods described above.

[0040] One or more technical solutions provided in the embodiments of this application have at least the following technical effects or advantages:

[0041] 1. In response to fault tracing operations for wind power equipment, the system acquires the state monitoring variables corresponding to the target subsystem; calculates the quantity control constraints corresponding to the target subsystem using a PCA model; determines whether the wind power equipment is in a fault state based on the quantity control constraints; if the wind power equipment is confirmed to be in a fault state, it acquires training sample representations, trains the SOM network structure and weights using the training sample representations, and determines whether the target subsystem is a faulty subsystem by inputting the state monitoring variables; if the target subsystem is confirmed to be a faulty subsystem, it acquires the historical faulty components corresponding to the target subsystem, and uses the state monitoring variables and historical faulty components as Bayesian network nodes to construct a Bayesian network model; through the Bayesian network model, it calculates the conditional probabilities corresponding to multiple fault modes of the target subsystem; if it is confirmed that the first conditional probability is greater than the second conditional probability, it takes the fault mode corresponding to the first conditional probability as the fault source of the wind power equipment. The first conditional probability and the second conditional probability are any two different probabilities among multiple different conditional probabilities, thereby achieving accurate location of the fault source of the wind power equipment and improving the accuracy and efficiency of fault tracing and identification.

[0042] 2. By using multiple output variables as the status monitoring variables corresponding to the target subsystem, the operating status of different subsystems can be comprehensively reflected, thereby ensuring that the fault sources of wind power equipment can be more comprehensively identified in subsequent fault tracing operations.

[0043] 3. Based on the status monitoring variables and network slicing, sensitivity analysis is performed on the target subsystem. The sensitivity analysis is used to calculate the target sensitivity value corresponding to the target subsystem and determine whether the target sensitivity value is the largest among multiple sensitivity values. If it is confirmed that the target sensitivity value is the largest among multiple sensitivity values, the target subsystem is determined to be a faulty subsystem, thereby improving the accuracy of fault diagnosis and reducing the possibility of misjudgment. Attached Figure Description

[0044] Figure 1 This is a flowchart illustrating the method for tracing the source of a single fault in wind power equipment provided in an embodiment of this application;

[0045] Figure 2 This is a schematic diagram of a first statistic and a second statistic provided in an embodiment of this application;

[0046] Figure 3 This is a schematic diagram of a Bayesian network model construction provided in an embodiment of this application;

[0047] Figure 4 This is a schematic diagram of the module of the wind power equipment single fault tracing device provided in the embodiments of this application;

[0048] Figure 5This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application.

[0049] Explanation of reference numerals in the attached drawings: 41, First construction module; 42, Second construction module; 43, Model building module; 44, Analysis and calculation module; 501, Processor; 502, Communication bus; 503, User interface; 504, Network interface; 505, Memory. Detailed Implementation

[0050] To enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments.

[0051] The terminology used in the following embodiments of this application is for the purpose of describing particular embodiments only and is not intended to be limiting of this application. As used in the specification of this application, the singular expressions “a,” “an,” “the,” “the,” “the,” and “this” are intended to include the plural expressions as well, unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used in this application refers to and includes any or all possible combinations of one or more of the listed items.

[0052] Hereinafter, the terms "first" and "second" are used for descriptive purposes only and should not be construed as implying or suggesting relative importance or implicitly indicating the number of indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature, and in the description of the embodiments of this application, unless otherwise stated, "multiple" means two or more.

[0053] To enable those skilled in the art to better understand the technical solution of the present invention, the present invention will be further described in detail below with reference to the accompanying drawings.

[0054] As one of the earliest developed and most important renewable energy sources in the integrated energy system, wind energy has enormous potential for greenhouse gas emission reduction due to its abundant reserves, renewability, and clean and pollution-free characteristics. It has become the mainstream direction of global renewable energy development and an important force in promoting industrial upgrading and driving energy system adjustment.

[0055] Existing wind power equipment suffers from poor operational reliability due to its complex and variable working environment. Downtime caused by faults accounts for 25.6% of the rated power generation time. For wind turbines with a service life of 20 years, maintenance costs account for about 20% to 25% of revenue. High operation and maintenance costs have become an important factor restricting the development of wind power, and there is an urgent need to develop efficient and reliable fault tracing technology.

[0056] Currently, common fault tracing methods in wind power equipment typically rely on sensor fault alarms and simple backtracking of historical data. While these methods can identify the time of fault occurrence and roughly indicate the fault area, the complex operating environment of wind power equipment makes it difficult for these traditional methods to accurately locate the specific fault source. For example, when multiple subsystems exist in a wind power unit and a fault occurs, the operation of each subsystem is often highly coupled, making it difficult for simple sensor alarms or historical data analysis to determine which subsystem is the initial fault source. This forces maintenance personnel to repeatedly inspect and attempt to find the initial fault source, thereby increasing maintenance costs and downtime.

[0057] Therefore, there is an urgent need for methods and devices for tracing the source of single faults in wind power equipment.

[0058] Please refer to Figure 1 The flowchart illustrates a single fault tracing method for wind power equipment provided in this application embodiment. The method is applied to a server, and the flowchart mainly includes the following steps: S101 to S107.

[0059] Step S101: In response to the fault tracing operation for wind power equipment, obtain the status monitoring variables corresponding to the target subsystem, where the target subsystem is any subsystem in the wind power equipment.

[0060] Specifically, in the embodiments of this application, the server can perform automatic fault tracing operations on the wind power equipment in real time, or the user can manually perform fault tracing operations to investigate the fault sources in the wind power equipment. The wind power equipment consists of multiple subsystems. When it is necessary to perform fault tracing operations on the wind power equipment, the status monitoring variables corresponding to each subsystem are obtained. For ease of explanation, this application uses the target subsystem as an example. The target subsystem can be any subsystem in the wind power equipment, and the subsystems in the wind power equipment include, but are not limited to: pitch system, gearbox system, frequency converter system, generator system, yaw system, nacelle system, etc. For the acquisition of status monitoring variables, the corresponding output variables of the target subsystem can be used as the status monitoring variables of the target subsystem. The output variables include, but are not limited to: rotor speed variable, generator power variable, wind angle variable, X-axis vibration variable, high-speed shaft drive end bearing temperature variable, turbine-side power module temperature variable, etc. The output variables corresponding to different subsystems are shown in Table 1.

[0061] [Table 1]

[0062]

[0063] It should be noted that in wind power equipment, there may be subsystems without corresponding output variables. In such cases, by integrating them with associated subsystems, the output variable corresponding to the associated subsystem can be used as the output variable for this subsystem. For example, consider a hydraulic system and a gearbox system. The hydraulic system lacks a state detection variable, but the two systems share similar fault characteristics. Therefore, they can be associated, with the high-speed shaft drive end bearing temperature used as a shared state detection variable for fault tracing. Furthermore, the purpose of associating subsystems is to decouple and trace faults. For subsystems without corresponding output variables, only one subsystem will be identified as its associated subsystem; multiple associations will not be performed.

[0064] Step S102: Calculate the quantity control limits corresponding to the target subsystem using the PCA model.

[0065] Specifically, the PCA model can be trained using the output variables of each subsystem. After the PCA model is completed, the quantitative control limits corresponding to the target subsystem are calculated using the PCA model.

[0066] In one possible implementation, step S102 further includes: calculating a first statistic using the following formula:

[0067]

[0068] Where SPE is the first statistic, P is a matrix composed of k eigenvectors, I is the identity matrix, and x new For the new sample, T is the transpose of the sample; then the control limits corresponding to the first statistic are calculated according to the following formula: Among them, Q α Let θ be the control limit corresponding to the first statistic. r The eigenvalues ​​corresponding to the sample covariance matrix are calculated using the following formula: c α Let h0 be the confidence limit of the standard normal distribution, and h0 be a distribution that follows θ. r The control limits are used to calculate the characteristic statistic, and the formula is as follows:

[0069] The second statistic is calculated using the following formula:

[0070]

[0071] Among them, T 2 The second statistic, Λ -1 To take the exponent of -1 for each non-zero element in the matrix, the control limits corresponding to the second statistic are calculated according to the following formula:

[0072]

[0073] Among them, T α The control limits for the second statistic are given, 1-α is the confidence level, and F0 is the value of the second statistic. α (a, na) is an F-distribution with a first degree of freedom and na second degree of freedom.

[0074] Specifically, the PCA model is used to determine whether the wind turbine as a whole is in a fault state, but not to identify the specific cause of the fault. When determining whether the wind turbine is in a fault state, the PCA model is first used to calculate the quantity control limits corresponding to the target subsystem. These limits include the quantity control limits corresponding to the first statistic and the second statistic. The calculation steps for these quantity control limits are as follows: Multiple output variables are standardized using the following formula. This standardization process is used to calculate standardized values:

[0075]

[0076] Where, x s.i For standardized values, x i S is the i-th output variable. i Let the standard deviations of multiple output variables be the standard deviations. Covariance decomposition is performed on the standardized values ​​using the following formula, which is used to calculate the covariance:

[0077]

[0078] Wherein, cov(x) i ,x i The first statistic is calculated for multiple output variables using the following formula:

[0079]

[0080] Where SPE is the first statistic, and P is a matrix composed of k eigenvectors, which is [p1, p2, ..., p...]. k ], where I is the identity matrix, x new For the new sample, T is the transpose of the sample; then the control limits corresponding to the first statistic are calculated according to the following formula:

[0081]

[0082] Among them, Q α Let θ be the control limit corresponding to the first statistic. r The eigenvalues ​​corresponding to the sample covariance matrix are calculated using the following formula: c α Let h0 be the confidence limit of the standard normal distribution, and h0 be a distribution that follows θ. r The control limits are used to calculate the characteristic statistic, and the formula is as follows:

[0083] The second statistic is calculated using the following formula:

[0084]

[0085] Among them, T 2 The second statistic, Λ -1 To take the exponent of -1 for each non-zero element in the matrix, the control limits corresponding to the second statistic are calculated according to the following formula:

[0086]

[0087] Among them, T α The control limits for the second statistic are given, 1-α is the confidence level, and F0 is the value of the second statistic. α (a, na) represents an F-distribution with first degree of freedom 'a' and second degree of freedom 'na'. Please refer to [reference needed]. Figure 2 The document presents a schematic diagram of a first statistic and a second statistic provided in an embodiment of this application.

[0088] Step S103: Determine whether the wind power equipment is in a fault state based on the quantity control limit.

[0089] Specifically, based on the calculation of the quantity control limit in step S102, if it is confirmed that the first statistic is less than the quantity control limit corresponding to the first statistic and the second statistic is less than the quantity control limit corresponding to the second statistic, then the wind power equipment is determined to be in a fault state; if it is confirmed that the first statistic is greater than or equal to the quantity control limit corresponding to the first statistic, or the second statistic is greater than or equal to the quantity control limit corresponding to the second statistic, then the wind power equipment is determined to be in a fault state. At this time, the fault tracing operation continues, and the fault source of the wind power equipment needs to be found.

[0090] Step S104: If the wind power equipment is confirmed to be in a fault state, obtain training sample representations, train the SOM network structure and weights through the training sample representations, and determine whether the target subsystem is a fault subsystem by inputting state monitoring variables.

[0091] Specifically, once a fault is confirmed in the wind turbine equipment, the fault tracing process continues, requiring the identification of the fault source. A System of Origin (SOM) network structure and weights are then constructed. The SOM network structure and weights are built using training sample representations, which represent the state monitoring variables corresponding to each subsystem under normal operating conditions. Subsequently, by inputting the state monitoring variables into the trained SOM network structure and weights, sensitivity analysis is performed based on network slices to determine whether the target subsystem is a faulty subsystem.

[0092] In one possible implementation, step S104 further includes: S1, obtaining a first sample representation from the training sample representations, wherein the first sample representation is any sample representation from the training sample representations; S2, obtaining the winning neuron, wherein the winning neuron is the neuron with the smallest distance from the first sample representation among multiple output neurons; S3, obtaining the connection weight coefficients corresponding to the output neurons within the winning neighborhood, and adjusting the connection weight coefficients according to the following formula, wherein the winning neighborhood is the neighborhood where the winning neuron is located: W ij (t+1)=W ij (t)+η(t)h c,j (t)(X i -W ij (t)); where W ij (t+1) is the learning rate function, η(t) is the learning rate function, and 0 < η(t) < 1. The learning rate function is the gain function, h c,j (t) is the neighborhood function; S4, the neighborhood function is updated using the following formula:

[0093]

[0094] Where, d cj Let r be the Euclidean distance between the winning neuron and the activated output neurons in its neighborhood, r be the neighborhood radius, and t be the iteration number; S5, update the neighborhood radius using the following formula:

[0095]

[0096] Where r(t+1) is the updated neighborhood radius, INT is the floor function, and T is the total number of training iterations; S6, the initial learning rate is updated using the following formula:

[0097]

[0098] Where η(t+1) is the updated learning rate function and η(0) is the initial learning rate function; S7, let t = t+1, repeat steps S1-S6 until t = T or the neighborhood function converges, to complete the training of the SOM network structure and weights.

[0099] Specifically, a learning process is constructed, and the training of the SOM network structure and weights is completed when this learning process converges. The training steps of this learning process are as follows: S1, obtain the first sample representation from the training sample representations. The first sample representation is any sample representation from the training sample representations. In this step, the initial neighborhood is first set, at which time t=0 and W can be initialized. ij (0), and establish a radius of N around the output neuron j. j(0) has a larger initial neighborhood. S2, obtain the winning neuron, which is the neuron with the smallest distance to the first sample representation among multiple output neurons: select any sample X from the state monitoring variables for input, and calculate the distance d between the input sample X and each output neuron j. j The output neuron c with the smallest distance is selected as the winning neuron, and the calculation formula is as follows:

[0100]

[0101] ||XW j || = min(d j );

[0102] Where, d j Let X be the distance between the input sample X and each output neuron j, where X is the input sample, and ||XW|| is the distance between the input sample X and each output neuron j. j ‖ represents the Euclidean distance, min(d j W represents the minimum distance. ij (t) represents the weights of the output neuron, X i (t) represents the iii-th variable in the input sample; S3, obtain the connection weight coefficients corresponding to the output neurons in the winning region, and adjust the connection weight coefficients according to the following formula, where the winning region is the region where the winning neuron is located: W ij (t+1)=W ij (t)+η(t)h c,j (t)(X i -W ij (t)); where W ij (t+1) represents the updated output neuron weights, η(t) is the learning rate function, and 0 < η(t) < 1. The learning rate function is the gain function, h c,j (t) is the neighborhood function; S4, the neighborhood function is updated using the following formula:

[0103]

[0104] Where, d cj Let r be the Euclidean distance between the winning neuron and the activated output neurons in its neighborhood, r be the neighborhood radius, and t be the iteration number; S5, update the neighborhood radius using the following formula:

[0105]

[0106] Where r(t+1) is the updated neighborhood radius, INT is the floor function, and T is the number of iterations; S6, the initial learning rate is updated using the following formula:

[0107]

[0108] Where η(t+1) is the updated learning rate function and η(0) is the initial learning rate function; S7, repeat the above steps S1-S6 until the neighborhood function, neighborhood radius and convergent initial learning rate are all converged, so as to complete the training of SOM network structure and weights.

[0109] In one possible implementation, step S104 further includes: performing sensitivity analysis on the target subsystem based on the state monitoring variables and network slices, wherein the sensitivity analysis is used to calculate the target sensitivity calculation value corresponding to the target subsystem; determining whether the target sensitivity calculation value is the maximum sensitivity calculation value among multiple sensitivity calculation values, wherein the multiple sensitivity calculation values ​​correspond one-to-one with multiple subsystems; if it is confirmed that the target sensitivity calculation value is the maximum sensitivity calculation value among multiple sensitivity calculation values, then the target subsystem is determined to be a faulty subsystem.

[0110] Specifically, after inputting the state monitoring variables, sensitivity analysis is performed on each state monitoring variable based on network slicing to obtain the sensitivity calculation value corresponding to each state monitoring variable. The target subsystem corresponds to the target sensitivity calculation value. If the target subsystem is a faulty subsystem, the target sensitivity calculation value corresponding to the target subsystem should be the maximum calculated value of all sensitivity calculation values. If this condition is met, the state monitoring variable corresponding to the target subsystem is the state monitoring variable that has experienced a fault, and the target subsystem corresponding to the state monitoring variable is the subsystem that has experienced a fault. The sensitivity calculation formula based on network slicing can be calculated as follows:

[0111]

[0112] Where, ω fe W is the sensitivity calculation value. c After inputting the state monitoring variable, the weight changes at the winning neuron location in the network slice fe, W max The maximum weight in network slice fe.

[0113] Step S105: If the target subsystem is confirmed to be a faulty subsystem, obtain the historical faulty components corresponding to the target subsystem, and use the state monitoring variables and historical faulty components as Bayesian network nodes to construct a Bayesian network model.

[0114] Specifically, if the target subsystem is confirmed to be a faulty subsystem, it is also necessary to obtain the specific fault mode corresponding to the faulty subsystem. Fault modes include, but are not limited to, sensor failure, mechanical damage, abnormal temperature, overload operation, and component wear. At this point, the historical faulty components corresponding to the target subsystem are obtained, and the state monitoring variables and historical faulty components are used as Bayesian network nodes to construct a Bayesian network model. The dependencies between network nodes in the Bayesian network are obtained from a given dataset, which includes, but is not limited to, historical fault data, sensor record data, equipment operation logs, and maintenance records. Based on the assumptions of the BN fault path tracing algorithm, a network node is only affected by its parent node, and only the set of parent nodes of that network node participates in Bayesian inference, thus further simplifying the conditional probability space. The conditional probability of a node is calculated by using the mutual information between state monitoring variables as the conditional probability of the node. The calculation formula is as follows:

[0115]

[0116] P(X|Y)=I(x;y);

[0117] Where I(x;y) is the mutual information of x and y, x and y are different state monitoring variables, p(x,y) is the joint probability density function of x and y, p(x) is the marginal probability density function of x, and p(y) is the marginal probability density function of y; the conditional probability between nodes is calculated by using the frequency data of historical failure modes of the state monitoring variables, and the calculation formula is as follows:

[0118]

[0119] Where P(X|Y) is the conditional probability of the state monitoring variable and the fault mode, n ci-xy The frequency, n, of the occurrence of fault mode X while a certain state monitoring variable Y is abnormal. ci-y The total number of times the status monitoring variable Y is determined to be abnormal, independent of whether a specific fault mode X occurs.

[0120] Step S106: Using a Bayesian network model, calculate the conditional probabilities corresponding to multiple failure modes of the target subsystem, with each failure mode corresponding to one conditional probability.

[0121] Specifically, the conditional probabilities corresponding to all failure modes of the target subsystem are calculated using a Bayesian network model.

[0122] Step S107: If it is confirmed that the first conditional probability is greater than the second conditional probability, the fault mode corresponding to the first conditional probability is taken as the fault source corresponding to the wind power equipment. The first conditional probability and the second conditional probability are any two different probabilities among a plurality of different conditional probabilities.

[0123] Specifically, when obtaining the fault mode, each corresponding conditional probability is identified, and the largest conditional probability, i.e. the first conditional probability, is found through filtering. The fault mode corresponding to the largest conditional probability is then obtained. At this point, the fault mode is the fault source corresponding to the wind power equipment.

[0124] In one possible implementation, step S107 further includes: calculating a first conditional probability using the following formula:

[0125]

[0126] Wherein, P(X) i |Y) represents the conditional probability, X1,…,X n It is a complete event set consisting of state monitoring variables and failure modes, and X i ,X j ∈{X1,…,X n}

[0127] For details, please refer to Figure 3 This illustration shows a schematic diagram of a Bayesian network model construction provided in an embodiment of this application, where I1 and I2 are state monitoring variables of the component layer, FM1 and FM2 are fault modes of the component layer, the conditional probability between I1 and I2 is calculated based on mutual information, and the mutual information of random variables is used to measure the correlation between events.

[0128] Based on the location of the target subsystem and the state monitoring variables indicating a fault, the fault probability of the target subsystem is set to 1.0 as the prior probability. Starting from this, Bayesian inference is performed using the following formula:

[0129]

[0130] Wherein, P(X) i |Y) represents the conditional probability, X1,…,X n It is a complete event set consisting of state monitoring variables and failure modes, and X i ,X j ∈{X1,…,X n By using Bayesian inference, the conditional probability of each failure mode occurring in the target subsystem is calculated. The higher the probability, the greater the likelihood of that failure mode occurring. Therefore, the failure mode with the highest conditional probability is selected as the failure source.

[0131] This application employs the aforementioned method to respond to fault tracing operations for wind power equipment by obtaining the state monitoring variables corresponding to the target subsystem; calculating the quantity control constraints corresponding to the target subsystem using a PCA model; determining whether the wind power equipment is in a fault state based on the quantity control constraints; if the wind power equipment is confirmed to be in a fault state, obtaining training sample representations, training the SOM network structure and weights using the training sample representations, and determining whether the target subsystem is a faulty subsystem by inputting the state monitoring variables; if the target subsystem is confirmed to be a faulty subsystem, obtaining the historical faulty components corresponding to the target subsystem, and using the state monitoring variables and historical faulty components as Bayesian network nodes to construct a Bayesian network model; calculating the conditional probabilities corresponding to multiple fault modes of the target subsystem using the Bayesian network model; if it is confirmed that the first conditional probability is greater than the second conditional probability, taking the fault mode corresponding to the first conditional probability as the fault source of the wind power equipment, where the first conditional probability and the second conditional probability are any two different probabilities among multiple different conditional probabilities, thereby achieving precise location of the fault source of the wind power equipment and improving the accuracy and efficiency of fault tracing and identification.

[0132] Please refer to Figure 4 It shows a schematic diagram of the modules of the wind power equipment single fault tracing device provided in the embodiment of this application. The device is a server, which includes a first construction module 41, a second construction module 42, a model building module 43, and an analysis and calculation module 44.

[0133] The first construction module 41 is used to respond to the fault tracing operation for wind power equipment, obtain the state monitoring variables corresponding to the target subsystem, where the target subsystem is any subsystem in the wind power equipment; and calculate the quantity control limits corresponding to the target subsystem through the PCA model.

[0134] The second construction module 42 is used to determine whether the wind power equipment is in a fault state based on the quantity control limit; if the wind power equipment is confirmed to be in a fault state, the module obtains training sample representations, trains the SOM network structure and weights through the training sample representations, and determines whether the target subsystem is a fault subsystem by inputting state monitoring variables.

[0135] The model building module 43 is used to obtain the historical faulty components corresponding to the target subsystem if the target subsystem is confirmed to be a faulty subsystem, and to use the state monitoring variables and historical faulty components as Bayesian network nodes to build a Bayesian network model.

[0136] The analysis and calculation module 44 is used to calculate the conditional probabilities corresponding to multiple fault modes of the target subsystem through a Bayesian network model. Each fault mode corresponds to a conditional probability. If it is confirmed that the first conditional probability is greater than the second conditional probability, the fault mode corresponding to the first conditional probability is taken as the fault source of the wind power equipment. The first conditional probability and the second conditional probability are any two different probabilities among multiple different conditional probabilities.

[0137] In one possible implementation, the first construction module 41 is used to obtain the state monitoring variables corresponding to the target subsystem: obtain the output variables corresponding to the target subsystem, including impeller speed variables, generator power variables and wind angle variables; and use the output variables as the state monitoring variables corresponding to the target subsystem.

[0138] In one possible implementation, the quantity control limits include quantity control limits corresponding to a first statistic and quantity control limits corresponding to a second statistic. The first construction module 41 is used to calculate the quantity control limits corresponding to the target subsystem using a PCA model, specifically including: calculating the first statistic using the following formula:

[0139]

[0140] Where SPE is the first statistic, P is a matrix composed of k eigenvectors, I is the identity matrix, and x new For the new sample, T is the transpose of the sample; then the control limits corresponding to the first statistic are calculated according to the following formula:

[0141]

[0142] Among them, Q α Let θ be the control limit corresponding to the first statistic. r The eigenvalues ​​corresponding to the sample covariance matrix are calculated using the following formula: c α Let h0 be the confidence limit of the standard normal distribution, and h0 be a distribution that follows θ. r The control limits are used to calculate the characteristic statistic, and the formula is as follows:

[0143] The second statistic is calculated using the following formula:

[0144]

[0145] Among them, T 2 The second statistic, Λ -1 To take the exponent of -1 for each non-zero element in the matrix, the control limits corresponding to the second statistic are calculated according to the following formula:

[0146]

[0147] Among them, T α The control limits for the second statistic are given, 1-α is the confidence level, and F0 is the value of the second statistic. α (a, na) is an F-distribution with a first degree of freedom and na second degree of freedom.

[0148] In one possible implementation, the second construction module 42 is used to determine whether the wind power equipment is in a fault state based on the quantity control limit, specifically including: if it is confirmed that the first statistic is less than the quantity control limit corresponding to the first statistic and the second statistic is less than the quantity control limit corresponding to the second statistic, then it is determined that the wind power equipment is not in a fault state; if it is confirmed that the first statistic is greater than or equal to the quantity control limit corresponding to the first statistic, or the second statistic is greater than or equal to the quantity control limit corresponding to the second statistic, then it is determined that the wind power equipment is in a fault state.

[0149] In one possible implementation, the second construction module 42 is used to train the SOM network structure and weights using training sample representations, specifically including: S1, obtaining a first sample representation from the training sample representations, wherein the first sample representation is any sample representation from the training sample representations; S2, obtaining the winning neuron, wherein the winning neuron is the neuron with the smallest distance from the first sample representation among multiple output neurons; S3, obtaining the connection weight coefficients corresponding to the output neurons within the winning neighborhood, and adjusting the connection weight coefficients according to the following formula, wherein the winning neighborhood is the neighborhood where the winning neuron is located:

[0150] W ij (t+1)=W ij (t)+η(t)h c,j (t)(X i -W ij (t));

[0151] Among them, W ij (t+1) represents the updated weights of the output neurons, η(t) is the learning rate function, and 0 < η(t) < 1. The learning rate function is the gain function, h c,j (t) is the neighborhood function; S4, the neighborhood function is updated using the following formula:

[0152]

[0153] Where, d cj Let r be the Euclidean distance between the winning neuron and the activated output neurons in its neighborhood, r be the neighborhood radius, and t be the iteration number; S5, update the neighborhood radius using the following formula:

[0154]

[0155] Where r(t+1) is the updated neighborhood radius, INT is the floor function, and T is the total number of training iterations; S6, the initial learning rate is updated using the following formula:

[0156]

[0157] Where η(t+1) is the updated learning rate function and η(0) is the initial learning rate function; S7, let t = t+1, repeat steps S1-S6 until t = T or the neighborhood function converges, to complete the training of the SOM network structure and weights.

[0158] In one possible implementation, the second construction module 42 is used to determine whether the target subsystem is a faulty subsystem by inputting state monitoring variables. Specifically, this includes: performing sensitivity analysis on the target subsystem based on the state monitoring variables and network slices, wherein the sensitivity analysis is used to calculate the target sensitivity calculation value corresponding to the target subsystem; determining whether the target sensitivity calculation value is the largest sensitivity calculation value among multiple sensitivity calculation values, wherein the multiple sensitivity calculation values ​​correspond one-to-one with multiple subsystems; if it is confirmed that the target sensitivity calculation value is the largest sensitivity calculation value among multiple sensitivity calculation values, then the target subsystem is determined to be a faulty subsystem.

[0159] In one possible implementation, the analysis and calculation module 44 is used to calculate the conditional probabilities corresponding to multiple failure modes of the target subsystem, specifically including: calculating the conditional probabilities using the following formula:

[0160]

[0161] Wherein, P(X) i |Y) represents the conditional probability, X1,…,X n It is a complete event set consisting of state monitoring variables and failure modes, and X i ,X j ∈{X1,…,X n}

[0162] It should be noted that the above embodiments of the apparatus are only illustrated by the division of the above functional modules. In practical applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above. In addition, the apparatus and method embodiments provided in the above embodiments belong to the same concept, and the specific implementation process can be found in the method embodiments, which will not be repeated here.

[0163] This application also provides an electronic device. (See reference...) Figure 5 , Figure 5This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. The electronic device may include: at least one processor 501, at least one communication bus 502, a user interface 503, at least one network interface 504, and a memory 505.

[0164] The communication bus 502 is used to enable communication between these components.

[0165] The user interface 503 may include a display screen and a camera. Optionally, the user interface 503 may also include a standard wired interface and a wireless interface.

[0166] The network interface 504 may optionally include a standard wired interface or a wireless interface (such as a Wi-Fi interface).

[0167] The processor 501 may include one or more processing cores. The processor 501 connects to various parts of the server using various interfaces and lines, and performs various server functions and processes data by running or executing instructions, programs, code sets, or instruction sets stored in memory 505, and by calling data stored in memory 505. Optionally, the processor 501 may be implemented using at least one hardware form of Digital Signal Processing (DSP), Field-Programmable Gate Array (FPGA), or Programmable Logic Array (PLA). The processor 501 may integrate one or a combination of several of the following: Central Processing Unit (CPU), Graphics Processing Unit (GPU), and modem. The CPU primarily handles the operating system, user interface, and applications; the GPU is responsible for rendering and drawing the content required for display; and the modem handles wireless communication. It is understood that the modem may also not be integrated into the processor 501 and may be implemented as a separate chip.

[0168] The memory 505 may include random access memory (RAM) or read-only memory. Optionally, the memory 505 may include a non-transitory computer-readable storage medium. The memory 505 may be used to store instructions, programs, code, code sets, or instruction sets. The memory 505 may include a program storage area and a data storage area, wherein the program storage area may store instructions for implementing an operating system, instructions for at least one function (such as touch function, sound playback function, image playback function, etc.), instructions for implementing the above-described method embodiments, etc.; the data storage area may store data involved in the above-described method embodiments, etc. Optionally, the memory 505 may also be at least one storage device located remotely from the aforementioned processor 501. (Refer to...) Figure 3 The memory 505, which serves as a computer storage medium, may include an operating system, a network communication module, a user interface module, and a wind power equipment single fault tracing application.

[0169] exist Figure 3 In the illustrated electronic device, the user interface 503 is primarily used to provide an input interface for the user and acquire user input data; while the processor 501 can be used to call the wind power equipment single fault tracing application stored in the memory 505. When executed by one or more processors 501, the electronic device performs one or more of the methods described in the above embodiments. It should be noted that, for the foregoing method embodiments, for the sake of simplicity, they are all described as a series of actions. However, those skilled in the art should understand that this application is not limited to the described order of actions, because according to this application, some steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also understand that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily essential to this application.

[0170] This application also provides a computer-readable storage medium storing instructions. When executed by one or more processors, these instructions cause an electronic device to perform one or more of the methods described in the above embodiments.

[0171] In the above embodiments, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.

[0172] In the various embodiments provided in this application, it should be understood that the disclosed apparatus can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some service interface; the indirect coupling or communication connection between apparatuses or units may be electrical or other forms.

[0173] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0174] Furthermore, the functional units in the various embodiments of this application 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.

[0175] 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 device (CMD). Based on this understanding, the technical solution of this application, 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 memory 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 of the various embodiments of this application. The aforementioned memory includes various media capable of storing program code, such as USB flash drives, portable hard drives, magnetic disks, or optical disks.

[0176] The above description is merely an exemplary embodiment disclosed in this application and should not be construed as limiting the scope of this application. Any equivalent changes and modifications made in accordance with the teachings of this application shall still fall within the scope of this application. Those skilled in the art, upon considering the disclosure of the specification and practical truths, will readily conceive of other embodiments disclosed in this application.

[0177] This application is intended to cover any variations, uses, or adaptations disclosed herein that follow the general principles disclosed herein and include common knowledge or customary technical means in the art that are not described in this application.

Claims

1. A method for tracing the source of a single fault in wind power equipment, characterized in that, The method includes: In response to a fault tracing operation for wind power equipment, the status monitoring variables corresponding to the target subsystem are obtained, wherein the target subsystem is any subsystem in the wind power equipment; The quantity control constraints corresponding to the target subsystem are calculated using a PCA model. These constraints include quantity control limits corresponding to a first statistic and a second statistic. Specifically, the calculation of the quantity control constraints for the target subsystem using the PCA model includes: The first statistic is calculated using the following formula: ； in, This is the first statistic. for A matrix composed of eigenvectors It is the identity matrix. For new samples, To perform the transpose operation, the control limits corresponding to the first statistic are calculated according to the following formula: ； in, The control limit corresponding to the first statistic. The eigenvalues ​​corresponding to the sample covariance matrix are calculated using the following formula: , The confidence limit for the standard normal distribution. To obey The control limits are used to calculate the characteristic statistic, and the formula is as follows: ; The second statistic is calculated using the following formula: ； in, This is the second statistic. To take each non-zero element in the matrix The index; then the control limits corresponding to the second statistic are calculated according to the following formula: ； in, The control limits for the second statistic are... For confidence level, To obey the first degree of freedom The second degree of freedom is of distributed; Based on the aforementioned quantity control limits, determine whether the wind power equipment is in a fault state; If the wind power equipment is confirmed to be in the fault state, a training sample representation is obtained, and the SOM network structure and weights are trained using the training sample representation. The target subsystem is then determined to be a faulty subsystem by inputting the state monitoring variables. The specific steps of training the SOM network structure and weights using the training sample representation include: S1, obtain the first sample representation in the training sample representation, wherein the first sample representation is any sample representation in the training sample representation; S2, obtain the winning neuron, which is the neuron with the smallest distance from the first sample representation among multiple output neurons; S3, obtain the connection weight coefficients corresponding to the output neuron within the winning region, and adjust the connection weight coefficients according to the following formula, where the winning region is the region where the winning neuron is located: ； in, The updated weights of the output neurons are as follows: Let be the learning rate function, and The learning rate function is a gain function. For domain functions; S4, update the domain function using the following formula: ； in, The Euclidean distance is the distance between the winning neuron and the activated output neurons in its neighborhood. Let t be the neighborhood radius and t be the number of iterations. S5, update the neighborhood radius using the following formula: ； in, For the updated domain radius, For the floor function, This represents the total number of training iterations. S6, the learning rate is updated using the following formula: ； in, The updated learning rate function, This is the initial learning rate function; S7, making Repeat steps S1-S6 until... Or the domain function converges to complete the training of the SOM network structure and weights; If the target subsystem is confirmed to be the faulty subsystem, obtain the historical faulty components corresponding to the target subsystem, and use the state monitoring variables and the historical faulty components as Bayesian network nodes to construct a Bayesian network model. Using the Bayesian network model, the conditional probabilities corresponding to the target subsystem when it has multiple failure modes are calculated, with each failure mode corresponding to one conditional probability. If it is confirmed that the first conditional probability is greater than the second conditional probability, the fault mode corresponding to the first conditional probability is taken as the fault source corresponding to the wind power equipment. The first conditional probability and the second conditional probability are any two different probabilities among a plurality of different conditional probabilities.

2. The method according to claim 1, characterized in that, The state monitoring variables corresponding to the target subsystem are obtained as follows: Obtain the output variables corresponding to the target subsystem, including impeller speed variable, generator power variable, and wind angle variable; The output variable is used as the state monitoring variable corresponding to the target subsystem.

3. The method according to claim 1, characterized in that, The step of determining whether the wind power equipment is in a fault state based on the quantity control limit specifically includes: If it is confirmed that the first statistic is less than the quantity control limit corresponding to the first statistic, and the second statistic is less than the quantity control limit corresponding to the second statistic, then it is determined that the wind power equipment is not in the fault state. If it is confirmed that the first statistic is greater than or equal to the quantity control limit corresponding to the first statistic, or the second statistic is greater than or equal to the quantity control limit corresponding to the second statistic, then the wind power equipment is determined to be in the fault state.

4. The method according to claim 1, characterized in that, The step of determining whether the target subsystem is a faulty subsystem by inputting the status monitoring variables specifically includes: Based on the state monitoring variables, and based on network slicing, a sensitivity analysis is performed on the target subsystem, wherein the sensitivity analysis is used to calculate the target sensitivity value corresponding to the target subsystem; Determine whether the target sensitivity calculation value is the largest sensitivity calculation value among multiple sensitivity calculation values, wherein the multiple sensitivity calculation values ​​correspond one-to-one with the multiple subsystems; If the target sensitivity calculation value is confirmed to be the maximum sensitivity calculation value among multiple sensitivity calculation values, then the target subsystem is determined to be the faulty subsystem.

5. The method according to claim 1, characterized in that, The calculation of the conditional probabilities corresponding to the target subsystem for multiple failure modes specifically includes: The conditional probability is calculated using the following formula: ； in, Let be the conditional probability. It is a complete event set consisting of the state monitoring variables and the fault modes, and .

6. A single fault tracing device for wind power equipment, characterized in that, The device includes a first construction module, a second construction module, a model building module, and an analysis and calculation module, wherein... The first construction module is used to respond to a fault tracing operation for wind power equipment by acquiring the state monitoring variables corresponding to the target subsystem, wherein the target subsystem is any subsystem in the wind power equipment; acquiring a PCA model, and calculating the quantity control limits corresponding to the target subsystem using the PCA model; the quantity control limits include quantity control limits corresponding to a first statistic and quantity control limits corresponding to a second statistic, and calculating the quantity control limits corresponding to the target subsystem using the PCA model specifically includes: The first statistic is calculated using the following formula: ； in, This is the first statistic. for A matrix composed of eigenvectors It is the identity matrix. For new samples, To perform the transpose operation, the control limits corresponding to the first statistic are calculated according to the following formula: ； in, The control limit corresponding to the first statistic. The eigenvalues ​​corresponding to the sample covariance matrix are calculated using the following formula: , The confidence limit for the standard normal distribution. To obey The control limits are used to calculate the characteristic statistic, and the formula is as follows: ; The second statistic is calculated using the following formula: ； in, This is the second statistic. To take each non-zero element in the matrix The index; then the control limits corresponding to the second statistic are calculated according to the following formula: ； in, The control limits for the second statistic are... For confidence level, To obey the first degree of freedom The second degree of freedom is of distributed; The second construction module is used to determine whether the wind power equipment is in a fault state based on the quantity control limit; if the wind power equipment is confirmed to be in a fault state, a training sample representation is obtained, the SOM network structure and weights are trained using the training sample representation, and the target subsystem is determined to be a faulty subsystem by inputting the state monitoring variable; the step of training the SOM network structure and weights using the training sample representation specifically includes: S1, obtain the first sample representation in the training sample representation, wherein the first sample representation is any sample representation in the training sample representation; S2, obtain the winning neuron, which is the neuron with the smallest distance from the first sample representation among multiple output neurons; S3, obtain the connection weight coefficients corresponding to the output neuron within the winning region, and adjust the connection weight coefficients according to the following formula, where the winning region is the region where the winning neuron is located: ； in, The updated weights of the output neurons are as follows: Let be the learning rate function, and The learning rate function is a gain function. For domain functions; S4, update the domain function using the following formula: ； in, The Euclidean distance is the distance between the winning neuron and the activated output neurons in its neighborhood. Let t be the neighborhood radius and t be the number of iterations. S5, update the neighborhood radius using the following formula: ； in, For the updated domain radius, For the floor function, This represents the total number of training iterations. S6, the learning rate is updated using the following formula: ； in, The updated learning rate function, This is the initial learning rate function; S7, making Repeat steps S1-S6 until... Or the domain function converges to complete the training of the SOM network structure and weights; The model building module is used to, if the target subsystem is confirmed to be the faulty subsystem, obtain the historical faulty components corresponding to the target subsystem, and use the state monitoring variables and the historical faulty components as Bayesian network nodes to build a Bayesian network model. The analysis and calculation module is used to calculate the conditional probabilities corresponding to multiple fault modes of the target subsystem through the Bayesian network model, with each fault mode corresponding to one conditional probability; if it is confirmed that the first conditional probability is greater than the second conditional probability, the fault mode corresponding to the first conditional probability is taken as the fault source corresponding to the wind power equipment, and the first conditional probability and the second conditional probability are any two different probabilities among multiple different conditional probabilities.

7. An electronic device, characterized in that, The device includes a processor, a communication bus, a user interface, a network interface, and a memory. The memory is used to store instructions. The user interface and the network interface are used to communicate with other devices. The processor is used to execute the instructions stored in the memory to cause the electronic device to perform the method as described in any one of claims 1 to 5.

8. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores instructions that, when executed, perform the method as described in any one of claims 1 to 5.

Images