A method for calculating the accuracy of wind turbine bearing fault early warning based on multi-dimensional parameter combination

Abstract

This invention relates to a method for calculating the accuracy of wind turbine bearing fault early warning based on multi-dimensional parameter combinations, belonging to the field of bearing fault early warning technology. The method includes: preprocessing historical monitoring data from the wind turbine SCADA system and performing quality screening on the preprocessed data; constructing multi-dimensional parameter combinations; establishing a wind turbine bearing temperature prediction model using a spatiotemporal graph convolutional neural network and setting a fault early warning threshold; constructing a spatiotemporal graph and feature matrix for each monitoring parameter combination in the multi-dimensional parameter combination as training data for the wind turbine bearing temperature prediction model; collecting wind turbine SCADA data in real time; performing fault judgment using the wind turbine bearing temperature prediction model for each monitoring parameter combination; and calculating the overall early warning accuracy by combining the early warning results of all combinations. This invention can accurately obtain the wind turbine bearing fault early warning accuracy even when monitoring parameters are abnormal.

Description

Technical Field

[0001] This invention belongs to the field of bearing fault early warning technology, and relates to a method for calculating the accuracy of wind power bearing fault early warning based on multi-dimensional parameter combination. Background Technology

[0002] Bearings are key components in wind turbines for energy conversion, including main bearings, gearbox bearings, and generator bearings. Under long-term harsh environments and variable load conditions, the failure rate of generator bearings reaches 60%-70%, while the failure rate of gearbox bearings is approximately 80%. Once a wind turbine bearing fails, high-altitude operations are difficult, disassembly and assembly are challenging, and repairs require considerable time and incurred high costs. Therefore, it is necessary to conduct research on wind turbine bearing failure early warning systems, which is of significant value in reducing wind turbine failure rates, operation and maintenance costs, and increasing the economic benefits of wind farms.

[0003] Currently, vibration or temperature data is mainly used for wind turbine bearing fault early warning. Vibration-based wind turbine condition monitoring systems are widely used in wind farms; however, this incurs additional purchase and installation costs. Wind turbine bearing temperature data can be directly obtained from existing SCADA systems. Temperature data has a long accumulation period and exhibits thermal inertia characteristics and strong anti-interference capabilities. Domestic and international scholars mostly use bearing temperature data and the residual method to achieve wind turbine bearing fault early warning. The residual method generally uses multi-dimensional monitoring parameters to establish a normal behavior prediction model, calculates the residual between the model's predicted values ​​and the measured values, and achieves bearing fault early warning by setting a residual threshold. Besides the prediction model itself affecting the fault early warning accuracy, the accuracy is also closely related to the model's input parameters. Currently, SCADA monitoring parameters can be divided into two categories: the first category is natural environmental parameters, mainly including wind speed, wind direction, and ambient temperature; the second category is operational parameters, such as generator electrical parameters, nacelle vibration parameters, and bearing temperature parameters. Wind turbines are complex power generation systems with highly coupled aerodynamic, mechanical, and electrical systems. Various operating parameters are correlated to varying degrees, and current research often uses both Type 1 and Type 2 monitoring parameters for fault early warning. While this method can achieve wind turbine bearing fault early warning, accurate warnings rely on the absence of anomalies in the input monitoring parameters. Anomalies in monitoring parameters can occur for various reasons, such as gearbox rear bearing temperature, generator stator coil temperature, gearbox oil inlet pressure, and converter power device temperature. Faults in these bearings, windings, oil circuits, and power devices will cause anomalies in their respective temperature data. Additionally, sensor malfunctions or interference can also lead to abnormal monitoring parameter data. The complex operating conditions of wind turbines due to wind speed make early detection of anomalies in monitoring parameters difficult. Ignoring anomalies in input monitoring parameters will significantly reduce the accuracy of bearing early warnings. Therefore, understanding the accuracy and trends of wind turbine bearing fault early warnings, and identifying abnormal bearings in advance, even when anomalies exist in the input monitoring parameters of the prediction model, will become a challenging problem to solve.

[0004] Existing early fault prediction studies treat all input monitoring parameters as a single combination for the prediction model. If all input monitoring parameters are divided and recombined into multiple combinations, each combination will contain a subset of the monitoring parameters. All combinations will fall into two categories: the first category includes anomalous parameters, and the second category does not. Anomalous parameter data will only affect the bearing fault prediction accuracy under the first category of combinations, and will not affect the bearing fault prediction accuracy under the second category of combinations. Summary of the Invention

[0005] In view of this, the purpose of this invention is to provide a method for calculating the accuracy of wind turbine bearing fault early warning based on multi-dimensional parameter combination.

[0006] To achieve the above objectives, the present invention provides the following technical solution:

[0007] A method for calculating the accuracy of wind turbine bearing fault early warning based on multi-dimensional parameter combination includes the following steps: S1: Preprocess the historical monitoring data of the wind turbine SCADA system and perform quality screening on the preprocessed data; S2: Constructing multi-dimensional parameter combinations, specifically including: S21: For the monitoring parameter set C Perform multi-dimensional parameter combinations from the monitoring parameter dataset C of j Select each parameter without repetition. l The parameters form a combination, where j > l ; S22: Generate multiple combinations of monitoring parameters through permutations and combinations, with the total number of combinations... H for:

[0008] S23: Constructing a system containing H A set of monitoring parameter combinations D ={ F 1, F 2,…, F H}={{ X 1, X 2, … , X l},{ X 1, X 2, …X l-1 , X l+1},…,{ X j-l+1 , X j-l+2 , … , X j Each combination contains l One monitoring parameter; S3: Establish a wind turbine bearing temperature prediction model using a spatiotemporal graph convolutional neural network (STGCN) and set a fault warning threshold; construct a spatiotemporal graph and feature matrix for each monitoring parameter combination in the multidimensional parameter combination as training data for the wind turbine bearing temperature prediction model; and train the wind turbine bearing temperature prediction model. S4: Real-time acquisition of SCADA data from wind turbine units; for each combination of monitoring parameters, fault diagnosis is performed using the wind turbine bearing temperature prediction model. S5: Calculate the overall early warning accuracy by combining the early warning results of all combinations:

[0009] In the formula H This represents the total number of combinations; U z For the first z The fault warning results of each combination.

[0010] Furthermore, the preprocessing of historical monitoring data from the wind turbine SCADA system described in step S1 includes: S11: Collect multi-source monitoring data from the wind turbine SCADA system, including temperature, speed, electrical parameters, vibration parameters, pressure parameters, and other environmental parameters; S12: Use the DBSCAN clustering algorithm to identify and remove outlier data; S13: Use the Xgboost algorithm to fill in the missing data; S14: Use the Z-Score standardization method to eliminate dimensional differences between different monitoring parameters.

[0011] Furthermore, the quality screening of the preprocessed data described in step S1 includes: Calculate the Spearman correlation coefficients between each monitoring parameter and the target wind turbine bearing temperature:

[0012] In the formula 、 Monitoring parameters X a , X b Data rank arranged in descending order; , These represent the average rank of the two monitoring parameter data; n Number of data points for each monitoring parameter; Monitoring parameters with correlation coefficients greater than a threshold were selected as the set of monitoring parameters strongly correlated with the temperature of the target wind turbine bearing. C ={ X 1, X 2,…, X j}

[0013] Furthermore, the Spatiotemporal Graph Convolutional Neural Network (STGCN) includes an input module for receiving a feature matrix and a spatial graph structure; a spatiotemporal convolution module containing temporal gate convolutional layers and spatial gate convolutional layers; and an output module for outputting prediction results through fully connected layers. The graph convolution operation employs a Chebyshev polynomial approximation. The model was trained using the Adam optimizer, with a loss function that was a combination of MAE and MSE.

[0014] Furthermore, the setting of the fault warning threshold specifically includes: Calculate combinations F k The residual between the predicted and measured bearing temperatures:

[0015] in, y t This is the measured temperature of the wind turbine bearing. This is the predicted temperature value for wind turbine bearings; The residual series is smoothed using an exponentially weighted moving average:

[0016] Where λ is the weight. λ (0,1]; q t-1 for t Residual statistics at time -1; t When =0, q 0 is set to the historical average residual value; Based on the smoothed residual data, the mean and standard deviation of the residuals are calculated, and then the following is obtained: F k Fault warning threshold T h and lower threshold T s :

[0017]

[0018] In the formula , These are the residual mean and standard deviation, respectively. w This is the threshold coefficient.

[0019] Furthermore, the spatiotemporal graph constructed in step S3 is as follows: For each combination of monitoring parameters, construct a spatial map. G =( O , V ),in O For a set of nodes, V Let it be the set of edges; Calculate the correlation coefficients between the monitoring parameters as elements of the adjacency matrix;

[0020] Binarize the correlation coefficient:

[0021] in, θ To control the adjacency matrix E (0-1) Sparsity threshold; 0 represents no connection between monitoring parameters; 1 represents a connection between monitoring parameters. Obtain the adjacency matrix of the monitoring parameters E (0-1) and the corresponding spatial diagram G , E (0-1) As shown in the following formula: .

[0022] Furthermore, the feature matrix described in step S3 is constructed as follows: The spatiotemporal graph uses monitored parameters as nodes and spatial dependencies between parameters as edges, combining them. F k The feature matrix is ​​shown below;

[0023] In the formula For monitoring parameters X k In the n The numerical value of the moment.

[0024] Furthermore, in step S4, the wind turbine bearing temperature prediction model is used to independently determine the faults of each combination, and the fault early warning results of a single monitoring parameter combination are obtained. U for:

[0025] In the formula U =0 indicates that the bearing is fault-free; U =1 indicates that the bearing is faulty.

[0026] The beneficial effects of this invention are as follows: This invention discloses a method for calculating the accuracy of wind turbine bearing fault early warning based on multi-dimensional parameter combinations. Addressing the problem that existing methods suffer a significant drop in fault early warning accuracy when input monitoring parameter data is abnormal, this invention proposes a wind turbine bearing fault early warning approach based on multi-dimensional parameter combinations. Through data preprocessing and parameter screening, multiple monitoring parameter combinations are constructed. A high-precision wind turbine bearing temperature prediction model is established by utilizing the STGCN network to capture the spatiotemporal correlation characteristics between monitoring parameters. An exponentially weighted moving average method is used to determine the fault early warning threshold for each combination, and the early warning results of all combinations are combined to obtain the wind turbine bearing fault early warning accuracy. The method proposed in this invention can still effectively obtain the accuracy of wind turbine bearing fault early warning and its changing trend even when some monitoring parameter data is abnormal, achieving early fault warning. This provides a more reliable scientific basis for condition-based maintenance and intelligent operation and maintenance of wind turbine units, and has significant engineering application value for improving the reliability and economy of wind farm operation.

[0027] Other advantages, objectives, and features of the invention will be set forth in part in the description which follows, and in part will be apparent to those skilled in the art from the following examination, or may be learned from practice of the invention. The objectives and other advantages of the invention can be realized and obtained through the following description. Attached Figure Description

[0028] To make the objectives, technical solutions, and advantages of the present invention clearer, the preferred embodiments of the present invention will be described in detail below with reference to the accompanying drawings, wherein: Figure 1 This is a schematic diagram of the overall process of the present invention; Figure 2 This is a schematic diagram of the data preprocessing workflow; Figure 3 Construct a schematic diagram for the spacetime graph; Figure 4 This is a schematic diagram of the STGCN network structure; Figure 5 Combination of Unit 13 F 1. Schematic diagram of predicted temperature of generator front bearing; Figure 6 This is a trend chart showing the change in the accuracy of early warnings during the case study. Detailed Implementation

[0029] The following specific examples illustrate the implementation of the present invention. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and various details in this specification can be modified or changed based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that the illustrations provided in the following embodiments are only schematic representations of the basic concept of the present invention. Unless otherwise specified, the following embodiments and features can be combined with each other.

[0030] It should be noted that the illustrations provided in the following embodiments are only schematic representations of the basic concept of the present invention. Therefore, the drawings only show the components related to the present invention and are not drawn according to the actual number, shape and size of the components in the actual implementation. In the actual implementation, the form, quantity and proportion of each component can be arbitrarily changed, and the layout of the components may also be more complex.

[0031] In the following description, numerous details are explored to provide a more thorough explanation of embodiments of the invention. However, it will be apparent to those skilled in the art that embodiments of the invention may be practiced without these specific details. In other embodiments, well-known structures and devices are shown in block diagram form rather than in detail to avoid obscuring embodiments of the invention.

[0032] Example 1: This invention provides a method for calculating the accuracy of wind turbine bearing fault early warning based on multi-dimensional parameter combinations. First, data preprocessing and multi-dimensional parameter combinations are performed to construct a monitoring parameter dataset. Then, an adjacency matrix and a spatiotemporal graph are constructed, and a wind turbine bearing temperature prediction model is established using spatiotemporal graph convolution (STGCN). Next, exponentially weighted moving average (EWMA) is used to obtain the fault early warning threshold for each combination. Based on the thresholds for each combination, the fault early warning results for each combination are obtained. Finally, the fault early warning results for all combinations are statistically analyzed to obtain the accuracy of wind turbine bearing fault early warning. Finally, the effectiveness of the proposed fault early warning method is verified using SCADA monitoring data from a wind farm.

[0033] This method specifically includes the following steps: S1. Preprocessing and quality screening of historical monitoring data from the wind turbine SCADA system, specifically including: S11. Collect multi-source monitoring data from the wind turbine SCADA system, including temperature, speed, electrical parameters, vibration parameters, pressure parameters, and other environmental parameters; S12. Use the DBSCAN clustering algorithm to identify and remove outlier data; S13. Use the Xgboost algorithm to supplement the missing data; S14. Use the Z-Score standardization method to eliminate dimensional differences between different monitoring parameters.

[0034] S2. Monitoring parameter screening and feature extraction, specifically including: S21. Calculate the Spearman correlation coefficient between each monitoring parameter and the bearing temperature, and select the set of monitoring parameters that are strongly correlated with the target wind turbine bearing temperature. C ={ X 1, X 2,…, X j};

[0035] In the formula: 、 Monitoring parameters X a , X b Data rank arranged in descending order; , These represent the average rank of the two monitoring parameter data; n Number of data points for each monitoring parameter; S22. Select monitoring parameters with an absolute value of correlation coefficient greater than 0.3 as input features.

[0036] S3, multi-dimensional parameter combination construction, specifically includes: S31, regarding the monitoring parameter set C Perform multi-dimensional parameter combinations. From the monitoring parameter dataset. C of j Selecting parameters without repetition from each monitoring parameter l The parameters form a combination, where j > l ; S32. Calculate the total number of multidimensional parameter combinations through permutations and combinations. H :

[0037] S33, Constructing includes H A set of monitoring parameter combinations D ={ F 1, F 2,…, F H}={{ X 1, X 2, … , X l},{ X 1, X 2,…X l-1 , X l+1},…,{ X j-l+1 , X j-l+2 , … , X j}}. Total H Each monitoring parameter combination contains [number] parameters. l One monitoring parameter.

[0038] S4. Spatiotemporal graph construction and feature representation, specifically including: S41. Construct a spatial map for each combination of monitoring parameters. G =( O , V ),in O For a set of nodes, V Let it be the set of edges; S42. Calculate the correlation coefficients between monitoring parameters as elements of the adjacency matrix;

[0039] S43. Binarize the correlation coefficient.

[0040] in, θ To control the adjacency matrix E (0-1) Sparsity threshold; 0 represents no connection between monitoring parameters; 1 represents a connection between monitoring parameters.

[0041] S44. Obtain the adjacency matrix of the monitoring parameters. E (0-1) and the corresponding spatial diagram G , E (0-1) As shown in the following formula:

[0042] S45, Constructing Combinations F k The feature matrix, wherein the spatiotemporal graph uses monitoring parameters as nodes and the spatial dependencies between parameters as edges, is combined. F k The feature matrix is ​​shown below;

[0043] In the formula: For monitoring parameters X k In the nThe numerical value of the moment.

[0044] S5. A prediction model is established using a Spatiotemporal Graph Convolutional Neural Network (STGCN). The network structure includes the following three parts: an input module for receiving the feature matrix and the spatial graph structure; a spatiotemporal convolution module containing temporal gate convolutional layers and spatial gate convolutional layers; and an output module for outputting the prediction results through fully connected layers. Specifically, it includes: S51. The graph convolution operation uses Chebyshev polynomial approximation:

[0045] S52. The model training uses the Adam optimizer, and the loss function is a combination of MAE and MSE.

[0046] S6. Fault warning threshold setting, specifically including: S61, Calculation Combinations F k The residual between the predicted and measured bearing temperatures:

[0047] in, y t This is the measured temperature of the wind turbine bearing. This is the predicted temperature value for wind turbine bearings.

[0048] S62. Smooth the residual series using Exponentially Weighted Moving Average (EWMA):

[0049] Where λ is the weight. λ (0,1]; q t-1 for t Residual statistics at time -1; t When =0, q 0 is set to the historical average residual.

[0050] S63. Based on the smoothed residual data, calculate the residual mean and standard deviation, and then obtain the combined... F k Fault warning threshold T h and lower threshold T s :

[0051]

[0052] In the formula: , These are the residual mean and standard deviation, respectively. w This is the threshold coefficient.

[0053] S7, multi-combination fault warning, specifically including: S71. Acquire real-time SCADA data of the wind turbine online. For each combination of monitoring parameters, calculate the current bearing temperature residual and obtain the fault warning result based on the warning threshold. (Fault warning result for a single combination of monitoring parameters.) U for:

[0054] In the formula: U =0, the bearing is fault-free; U =1, the bearing is faulty.

[0055] S71, Statistics All H The fault early warning results of a combination of monitoring parameters are used to calculate the accuracy P of the wind turbine bearing fault early warning.

[0056] In the formula: H This represents the total number of combinations; U z For the first z The fault warning results of each combination.

[0057] Example 2: Please see Figures 1-6 This invention presents a method for calculating the accuracy of wind turbine bearing fault early warning based on multi-dimensional parameter combinations. Taking a 2.3MW doubly-fed induction generator (DFIG) wind turbine in a wind farm as an example, the method is used for bearing fault early warning, mainly including three steps: data preprocessing and multi-dimensional parameter combination, bearing temperature prediction model, and fault early warning. S1. Data preprocessing and multidimensional parameter combination construction Historical SCADA data for wind turbine unit 12 of this wind farm were collected from January 16, 2019 to June 16, 2019, with a sampling interval of 10 minutes. Initially, 20 parameters related to the generator front bearing temperature were selected to form the dataset. A After removing outlier data segments and filling in missing values ​​using the DBSCAN algorithm, Z-Score standardization is performed to obtain the monitoring parameter dataset. B The Spearman correlation coefficients between each parameter and the generator inlet bearing temperature were calculated. Parameters with correlation coefficients greater than 0.3 were selected to obtain the key monitoring parameter dataset C, which contains 10 parameters. The Spearman correlation coefficients of the generator inlet bearing temperature are shown in Table 1. The number of parameters for each combination was set. l =6. Selecting 6 parameters from 10 results in 210 different combinations of monitoring parameters, forming a set.D ={ F 1, F 2,…, F 210}

[0058] Table 1

[0059] S2. Establishment and Training of Bearing Temperature Prediction Model For each combination F k Constructing its spatiotemporal graph: Calculate the Spearman correlation coefficients between each pair of the six parameters within the combination, set a threshold θ=0.5 for binarization, and obtain a 6x6 adjacency matrix. E (0-1) This defines the spatial graph structure. The time-series data for each parameter constitutes the node features. The STGCN model structure is adopted (containing two spatiotemporal convolutional blocks, each block containing temporal convolution and spatial graph convolution). Historical data is divided into time windows (e.g., 12 hours, 72 points), with the training and test sets split in a 7:3 ratio. The Adam optimizer and ReLU activation function are used to train the prediction models for each combination until the loss function converges.

[0060] S3, Setting the Early Warning Threshold for Single Combined Faults The trained models are used to predict the training set data, and the residuals are calculated. The residuals are then smoothed using EWMA (λ=0.2). The mean of the smoothed residuals is calculated. and standard deviation Set threshold coefficient. w =3, calculate the upper threshold for each combination. T h and lower threshold T s .

[0061] S4. Online Fault Early Warning and Assessment The same preprocessing was performed on the real-time data of the target unit (Unit 13, which later experienced a generator front bearing failure) from June 17, 2019 to July 17, 2019. For each time window (e.g., data from 00:00 to 12:00 on July 1, 2019), 210 combinations of data were input into their corresponding trained STGCN model to obtain 210 predicted generator front bearing temperatures. The real-time residuals of each combination were calculated and smoothed using EWMA. It was determined whether the smoothed residuals exceeded the threshold for that combination. Warning signals were issued from the 210 combinations. U k The number of combinations where 1 = 1. Calculate the fault warning accuracy P for this time window. Repeat this process to calculate the warning accuracy for each time window. P.

[0062] The method of this invention effectively solves the problem of decreased early warning accuracy caused by abnormal monitoring parameters, and provides a reliable guarantee for the safe operation of wind turbine units.

[0063] Example 3: An electronic device, including a memory and a processor; The memory is used to store computer programs; The processor is configured to implement the method described in Embodiment 1 when executing the computer program.

[0064] Example 4: A computer-readable storage medium storing a computer program that, when executed by a processor, implements the method described in Embodiment 1.

[0065] Example 5: A computer program product includes a computer program that, when executed by a processor, implements the method described in Example 1.

[0066] In the above embodiments, the reference to "this embodiment" in the specification indicates that a specific feature, structure, or characteristic described in connection with the embodiment is included in at least some embodiments, but not necessarily all embodiments. Multiple appearances of "this embodiment" do not necessarily refer to the same embodiment.

[0067] In the above embodiments, although the invention has been described in conjunction with specific embodiments thereof, many substitutions, modifications, and variations of these embodiments will be apparent to those skilled in the art from the foregoing description. For example, other memory structures (e.g., dynamic RAM (DRAM)) may be used with the embodiments discussed. The embodiments of the invention are intended to cover all such substitutions, modifications, and variations falling within the broad scope of the appended claims.

[0068] As will be understood by those skilled in the art, the computer-readable storage medium described in this embodiment allows for the implementation of all or part of the steps in the above method embodiments by computer program-related hardware. The aforementioned computer program can be stored in a computer-readable storage medium. When executed, the program performs the steps of the above method embodiments; and the aforementioned storage medium includes various media capable of storing program code, such as ROM, RAM, magnetic disks, or optical disks.

[0069] The electronic terminal provided in this embodiment includes a processor, a memory, a transceiver, and a communication interface. The memory and the communication interface are connected to the processor and the transceiver and complete communication between them. The memory is used to store computer programs, the communication interface is used to perform communication, and the processor and the transceiver are used to run the computer programs, so that the electronic terminal performs the steps of the above method.

[0070] In this embodiment, the memory may include random access memory (RAM) and may also include non-volatile memory, such as at least one disk storage device.

[0071] The processors mentioned above can be general-purpose processors, including central processing units (CPUs), network processors (NPs), etc.; they can also be digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.

[0072] This invention can be used in a wide range of general-purpose or special-purpose computing system environments or configurations. Examples include: personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer electronics, network PCs, minicomputers, mainframe computers, and distributed computing environments including any of the above systems or devices, etc.

[0073] This invention can be described in the general context of computer-executable instructions, such as program modules, that are executed by a computer. Generally, program modules include routines, programs, objects, components, data structures, etc., that perform a specific task or implement a specific abstract data type. This invention can also be practiced in distributed computing environments where tasks are performed by remote processing devices connected via a communication network. In distributed computing environments, program modules can reside in local and remote computer storage media, including storage devices.

[0074] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for calculating the accuracy of wind turbine bearing fault early warning based on multi-dimensional parameter combination, characterized in that: Includes the following steps: S1: Preprocess the historical monitoring data of the wind turbine SCADA system and perform quality screening on the preprocessed data; S2: Constructing multi-dimensional parameter combinations, specifically including: S21: For the monitoring parameter set C Perform multi-dimensional parameter combinations from the monitoring parameter dataset C of j Select each parameter without repetition. l The parameters form a combination, where j > l ; S22: Generate multiple combinations of monitoring parameters through permutations and combinations, with the total number of combinations... H for: S23: Constructing a system containing H A set of monitoring parameter combinations D ={ F 1, F 2,…, F H }={{ X 1, X 2, … , X l },{ X 1, X 2, … X l-1 , X l+1 },…,{ X j-l+1 , X j-l+2 , … , X j Each combination contains l One monitoring parameter; S3: Establish a wind turbine bearing temperature prediction model using a spatiotemporal graph convolutional neural network (STGCN) and set a fault warning threshold; construct a spatiotemporal graph and feature matrix for each monitoring parameter combination in the multidimensional parameter combination as training data for the wind turbine bearing temperature prediction model; and train the wind turbine bearing temperature prediction model. S4: Real-time acquisition of SCADA data from wind turbine units; for each combination of monitoring parameters, fault diagnosis is performed using the wind turbine bearing temperature prediction model. S5: Calculate the overall early warning accuracy by combining the early warning results of all combinations: In the formula H This represents the total number of combinations; U z For the first z The fault warning results of each combination.

2. The method for calculating the accuracy of wind turbine bearing fault early warning based on multi-dimensional parameter combination according to claim 1, characterized in that: The preprocessing of historical monitoring data from the wind turbine SCADA system described in step S1 includes: S11: Collect multi-source monitoring data from the wind turbine SCADA system, including temperature, speed, electrical parameters, vibration parameters, pressure parameters, and other environmental parameters; S12: Use the DBSCAN clustering algorithm to identify and remove outlier data; S13: Use the Xgboost algorithm to fill in the missing data; S14: Use the Z-Score standardization method to eliminate dimensional differences between different monitoring parameters.

3. The method for calculating the accuracy of wind turbine bearing fault early warning based on multi-dimensional parameter combination according to claim 1, characterized in that: Step S1, which involves quality screening of the preprocessed data, includes: Calculate the Spearman correlation coefficients between each monitoring parameter and the target wind turbine bearing temperature: In the formula 、 Monitoring parameters X a , X b Data rank arranged in descending order; , These represent the average rank of the two monitoring parameter data; n Number of data points for each monitoring parameter; Monitoring parameters with correlation coefficients greater than a threshold were selected as the set of monitoring parameters strongly correlated with the temperature of the target wind turbine bearing. C ={ X 1, X 2,…, X j } 4. The method for calculating the accuracy of wind turbine bearing fault early warning based on multi-dimensional parameter combination according to claim 1, characterized in that: The Spatiotemporal Graph Convolutional Neural Network (STGCN) includes an input module for receiving a feature matrix and a spatial graph structure; a spatiotemporal convolution module containing temporal gate convolutional layers and spatial gate convolutional layers; and an output module for outputting prediction results through fully connected layers. The graph convolution operation uses a Chebyshev polynomial approximation. The model was trained using the Adam optimizer, with a loss function that was a combination of MAE and MSE.

5. The method for calculating the accuracy of wind turbine bearing fault early warning based on multi-dimensional parameter combination according to claim 1, characterized in that: The setting of fault warning thresholds specifically includes: Calculate combinations F k The residual between the predicted and measured bearing temperatures: in, y t This is the measured temperature of the wind turbine bearing. This is the predicted temperature value for wind turbine bearings; The residual series is smoothed using an exponentially weighted moving average: Where λ is the weight. λ (0,1]; q t-1 for t Residual statistics at time -1; t When =0, q 0 is set to the historical average residual value; Based on the smoothed residual data, the mean and standard deviation of the residuals are calculated, and then the following is obtained: F k Fault warning threshold T h and lower threshold T s : In the formula , These are the residual mean and standard deviation, respectively. w This is the threshold coefficient.

6. The method for calculating the accuracy of wind turbine bearing fault early warning based on multi-dimensional parameter combination according to claim 1, characterized in that: The spatiotemporal graph constructed in step S3 is as follows: For each combination of monitoring parameters, construct a spatial map. G =( O , V ),in O For a set of nodes, V Let it be the set of edges; Calculate the correlation coefficients between the monitoring parameters as elements of the adjacency matrix; Binarize the correlation coefficient: in, θ To control the adjacency matrix E (0-1) Sparsity threshold; 0 represents no connection between monitoring parameters; 1 represents a connection between monitoring parameters. Obtain the adjacency matrix of the monitoring parameters E (0-1) and the corresponding spatial diagram G , E (0-1) As shown in the following formula: 。 7. The method for calculating the accuracy of wind turbine bearing fault early warning based on multi-dimensional parameter combination according to claim 1, characterized in that: The feature matrix constructed in step S3 is as follows: The spatiotemporal graph uses monitored parameters as nodes and spatial dependencies between parameters as edges, combining them. F k The feature matrix is ​​shown below; In the formula For monitoring parameters X k In the n The numerical value of the moment.

8. The method for calculating the accuracy of wind turbine bearing fault early warning based on multi-dimensional parameter combination according to claim 1, characterized in that: In step S4, the wind turbine bearing temperature prediction model is used to independently determine the faults of each combination, and the fault early warning results of a single monitoring parameter combination are obtained. U for: In the formula U =0 indicates that the bearing is fault-free; U =1 indicates that the bearing is faulty.

9. An electronic device, characterized in that, Including memory and processor; The memory is used to store computer programs; The processor is configured to, when executing the computer program, implement the method for calculating the accuracy of wind turbine bearing fault early warning based on multi-dimensional parameter combination as described in any one of claims 1-8.

10. A computer-readable storage medium, characterized in that, The storage medium stores a computer program, which, when executed by a processor, implements the method for calculating the accuracy of wind turbine bearing fault early warning based on multi-dimensional parameter combinations as described in any one of claims 1-8.

Images