Wind turbine tower fault prediction method and system based on multivariate data fusion, equipment and medium

Abstract

The invention relates to the technical field of wind power generation equipment detection, in particular to a wind turbine tower fault prediction method and system based on multivariate data fusion, equipment and a medium, and the method mainly comprises the steps that a neural network prediction model is established, the neural network prediction model is trained through a training set, and the fault prediction result is obtained. Optimizing the neural network prediction model through the verification set, and outputting an optimal neural network prediction model; and outputting a current prediction result through a neural network prediction model based on the basic data, and giving different execution actions according to different prediction results. According to the method, the current data can be collected in real time, and the collected data are analyzed to obtain the corresponding result, so that frequent manual inspection can be avoided to a certain extent, the manual inspection frequency is reduced, and the detection efficiency is improved.

Description

Technical Field

[0001] This invention relates to the field of wind power equipment testing technology, and more specifically, to a method, system, equipment, and medium for predicting wind turbine tower faults based on multi-source data fusion. Background Technology

[0002] The wind turbine tower is a key load-bearing structure supporting the nacelle and rotor, its main function being to elevate the power generation unit to the required height to obtain stable wind energy. Early tower designs varied, including truss-type (similar to power towers) and cable-stayed types, but these gradually faded out due to aesthetic, safety, and maintenance issues. The modern mainstream is the steel conical tower, composed of multiple tower sections connected by high-strength flange bolts, comprehensively optimizing structural strength, manufacturing convenience, transportation economy, and wind farm adaptability.

[0003] However, current technology generally relies on manual inspection to inspect wind turbine towers. Since wind turbines are typically installed at mountain tops where wind speeds are high, the process of reaching these locations presents challenges such as high workload and low efficiency for workers. Summary of the Invention

[0004] The purpose of this invention is to provide a method, system, equipment, and medium for predicting wind turbine tower faults based on multi-source data fusion, in order to solve the problems in the prior art.

[0005] This invention is achieved through the following technical solution:

[0006] In a first aspect, the present invention provides a method for predicting wind turbine tower faults based on multi-source data fusion, comprising:

[0007] Acquire the current basic data of the wind turbine tower and the historical data of the wind turbine tower during the collection period when it is faulty, and preprocess the normal historical data and basic data;

[0008] The preprocessed historical data is divided into training and validation sets. A neural network prediction model is established. The neural network prediction model is trained using the training set and optimized using the validation set. The optimal neural network prediction model is then output.

[0009] Based on the underlying data, a neural network prediction model outputs the current prediction result;

[0010] If the prediction result indicates a fault, the fault result is output, and an alarm is issued based on the current basic data and the output value of the neural network prediction model.

[0011] If the prediction result indicates that there is no fault, the time for the next collection week is updated based on the output value of the current neural network prediction model.

[0012] Preferably, the preprocessing of normal historical data and basic data includes:

[0013] Data cleaning was performed based on K-means clustering, and the cleaned data was then normalized.

[0014] The data cleaning includes:

[0015] A historical dataset will be created and then input into K-means for clustering, resulting in two clusters.

[0016] Calculate the Euclidean distance between each sampling point and the cluster centroid, and assign the sampling points to the clusters corresponding to the cluster centroids;

[0017] For each cluster, recalculate the cluster centroid position. When the set number of clusters is reached or the cluster centroid position no longer changes, output the clustering result.

[0018] A judgment threshold is set. If the distance between the centroids of two clusters is greater than the judgment threshold, the historical data at the current moment is abnormal, and the abnormal historical data is cleaned.

[0019] Preferably, the method further includes establishing a feature extraction model to extract features from the normalized data:

[0020]

[0021] In the formula, Fault type This is the current hidden layer state. For activation function, The weight matrix is ​​input to the hidden layer. For the current input, This is the weight matrix from hidden layer to hidden layer. This is the bias vector for the hidden layer.

[0022] Preferably, the establishment of the neural network prediction model includes:

[0023]

[0024]

[0025]

[0026] In the formula, For the Gate of Oblivion Let be the weight matrix of the s-th layer of the forget gate. Let be the hidden state of the s-th layer at time t-1. The input at time t, Let be the bias vector of the forget gate. For input gate, Let be the weight matrix of the s-th layer of the input gate. Let be the bias vector of the input gate. This represents the state of the cell at time t. This represents the candidate cell state.

[0027] Preferably, the loss function also includes:

[0028]

[0029] In the formula, The output of the loss function, The total number of fault types. The weights for calculating fault types, Let the intersection loss function be the type k fault. For hyperparameters, The weights for the s-th stage are calculated as follows: For loss measurement at different stages.

[0030] Preferably, issuing an alarm based on the current basic data and the output value of the neural network prediction model includes:

[0031] Obtain the confidence level of the current neural network model output, and establish an alarm evaluation model based on the confidence level. The alarm evaluation model includes:

[0032]

[0033] In the formula, As an evaluation index, The confidence score output by the neural network model. The standard pressure value is collected by a pressure sensor installed at the bolt connection of the tower. This is the current pressure value. This is the initial height from the top of the wind turbine to the ground. The current height from the top of the wind turbine to the ground. This is the maximum rotational speed of the wind turbine blades. This represents the current speed of the wind turbine blades;

[0034] Set a first threshold and a second threshold;

[0035] when When the value is less than or equal to the first threshold, a level one alarm is issued. When the value is greater than the first threshold and less than or equal to the second threshold, a level two alarm is issued. When the value exceeds the second threshold, a level 3 alarm is issued;

[0036] The severity of the level 1, level 2, and level 3 alarms increases sequentially.

[0037] Preferably, updating the time for the next data collection week based on the output value of the current neural network prediction model includes establishing a data collection cycle update model:

[0038]

[0039] In the formula, The updated collection cycle. This represents the current data collection cycle.

[0040] Secondly, the present invention also provides a wind turbine tower fault prediction system based on multi-source data fusion, characterized in that it includes a method for performing the above-mentioned wind turbine tower fault prediction method based on multi-source data fusion, and further includes:

[0041] The model building module is configured to acquire the basic data of the current wind turbine tower and the historical data of the wind turbine tower during the acquisition period when it is faulty, and preprocess the normal historical data and basic data; divide the preprocessed historical data into training set and validation set, build a neural network prediction model, train the neural network prediction model with the training set, optimize the neural network prediction model with the validation set, and output the optimal neural network prediction model; output the current prediction result based on the basic data and the neural network prediction model.

[0042] The output module is configured to output a faulty result if the prediction result indicates a fault exists, and issue an alarm based on the current basic data and the output value of the neural network prediction model; if the prediction result indicates no fault exists, the time for the next collection week is updated based on the current output value of the neural network prediction model.

[0043] Thirdly, the present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the above-described method for predicting wind turbine tower faults based on multi-source data fusion.

[0044] Fourthly, the present invention also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the above-described method for predicting wind turbine tower faults based on multi-source data fusion.

[0045] The technical solution of the present invention has at least the following advantages and beneficial effects:

[0046] The method provided by this invention mainly includes establishing a neural network prediction model, training the neural network prediction model using a training set, optimizing the neural network prediction model using a validation set, and outputting the optimal neural network prediction model. Based on basic data, the neural network prediction model outputs the current prediction result, and different execution actions are given according to different prediction results. This method allows for real-time data collection and analysis to obtain corresponding results, which can, to some extent, avoid frequent manual inspections, reduce the number of manual inspections, and improve detection efficiency. Attached Figure Description

[0047] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0048] Figure 1 This is a schematic diagram of the process of the present invention;

[0049] Figure 2 This is a schematic diagram of the system structure of the present invention. Detailed Implementation

[0050] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.

[0051] The terms "first," "second," etc., in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. The naming or numbering of steps in this application does not imply that the steps in the method flow must be executed in the time / logical order indicated by the naming or numbering. The execution order of named or numbered process steps can be changed according to the desired technical purpose, as long as the same or similar technical effect is achieved. The division of modules in this application is a logical division; in practical applications, other division methods may be used. For example, multiple modules may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the connection, coupling, or communication in this application can be a direct connection, coupling, or communication between related objects, or an indirect connection, coupling, or communication through other devices. Moreover, the connection, coupling, or communication between objects can be electrical or other similar forms, which are not limited in this application. The independently described modules or sub-modules may or may not be physically separated; they may be implemented in software or hardware, and some modules or sub-modules may be implemented in software, with the processor calling the software to implement the function of these modules or sub-modules, while other modules or sub-modules may be implemented in hardware, such as through hardware circuits. Furthermore, some or all of the modules can be selected to achieve the purpose of this application's solution according to actual needs.

[0052] Please refer to Figures 1-2 This invention provides a method for predicting wind turbine tower faults based on multi-source data fusion, comprising:

[0053] S101: Obtain the basic data of the current wind turbine tower and the historical data of the wind turbine tower during the collection period when it is faulty, and preprocess the normal historical data and basic data;

[0054] In this embodiment, the historical data mainly includes data that reflects the current condition of the tower, such as the stress of the bolts and the overall height of the tower. Based on this data, the subsequent training of the neural network model is carried out.

[0055] S102: Divide the preprocessed historical data into training and validation sets, establish a neural network prediction model, train the neural network prediction model using the training set, optimize the neural network prediction model using the validation set, and output the optimal neural network prediction model.

[0056] During training, the model can be updated in each iteration by changing network connection weights, learning rate, etc., to obtain the optimal network model.

[0057] S103: Output the current prediction result based on the basic data using a neural network prediction model;

[0058] In this embodiment, the output is a confidence value, which reflects the numerical similarity between the current data and the historical fault. By setting a confidence threshold, the output of the network model is obtained.

[0059] S104: If the prediction result indicates a fault, output the fault result and issue an alarm based on the current basic data and the output value of the neural network prediction model.

[0060] S105: If the prediction result indicates that there is no fault, then update the time of the next collection week based on the output value of the current neural network prediction model.

[0061] The method provided by this invention mainly includes establishing a neural network prediction model, training the neural network prediction model using a training set, optimizing the neural network prediction model using a validation set, and outputting the optimal neural network prediction model. Based on basic data, the neural network prediction model outputs the current prediction result, and different execution actions are given according to different prediction results. This method allows for real-time data collection and analysis to obtain corresponding results, which can, to some extent, avoid frequent manual inspections, reduce the number of manual inspections, and improve detection efficiency.

[0062] In one exemplary embodiment of the present invention, the preprocessing of normal historical data and basic data includes:

[0063] Data cleaning was performed based on K-means clustering, and the cleaned data was then normalized.

[0064] In this embodiment, all data is normalized to the range of 0 to 1.

[0065] In this embodiment, if the confidence value is less than or equal to 0.5, the output will show that there is no fault; if the confidence value is greater than 0.5, the output will show that there is a fault.

[0066] The data cleaning includes:

[0067] A historical dataset will be created and then input into K-means for clustering, resulting in two clusters.

[0068] Calculate the Euclidean distance between each sampling point and the cluster centroid, and assign the sampling points to the clusters corresponding to the cluster centroids;

[0069] For each cluster, recalculate the cluster centroid position. When the set number of clusters is reached or the cluster centroid position no longer changes, output the clustering result.

[0070] A judgment threshold is set. If the distance between the centroids of two clusters is greater than the judgment threshold, the historical data at the current moment is abnormal, and the abnormal historical data is cleaned.

[0071] Data cleaning is a crucial step in data preprocessing, primarily used to improve data quality and ensure the accuracy of subsequent analysis. This includes handling missing values, deleting missing records, filling in default values ​​(mean / median / mode), using interpolation, and using predictive models to complete the data, thus avoiding analytical bias or model errors caused by missing values.

[0072] Process duplicate data by identifying and deleting completely duplicate records, or merging partially duplicated data (such as forms submitted multiple times by users) to prevent duplicate data from interfering with statistical results or model training.

[0073] An exemplary embodiment of the present invention further includes establishing a feature extraction model to extract features from the normalized data:

[0074]

[0075] In the formula, Fault type This is the current hidden layer state. For activation function, The weight matrix is ​​input to the hidden layer. For the current input, This is the weight matrix from hidden layer to hidden layer. This is the bias vector for the hidden layer.

[0076] In one exemplary embodiment of the present invention, establishing the neural network prediction model includes:

[0077]

[0078]

[0079]

[0080] In the formula, For the Gate of Oblivion Let be the weight matrix of the s-th layer of the forget gate. Let be the hidden state of the s-th layer at time t-1. The input at time t, Let be the bias vector of the forget gate. For input gate, Let be the weight matrix of the s-th layer of the input gate. Let be the bias vector of the input gate. This represents the state of the cell at time t. This represents the candidate cell state.

[0081] Specifically, this also includes loss functions such as:

[0082]

[0083] In the formula, The output of the loss function, The total number of fault types. The weights for calculating fault types, Let the intersection loss function be the type k fault. For hyperparameters, The weights for the s-th stage are calculated as follows: For loss measurement at different stages.

[0084] In one exemplary embodiment of the present invention, issuing an alarm based on current basic data and the output value of the neural network prediction model includes:

[0085] Obtain the confidence level of the current neural network model output, and establish an alarm evaluation model based on the confidence level. The alarm evaluation model includes:

[0086]

[0087] In the formula, As an evaluation index, The confidence score output by the neural network model. The standard pressure value is collected by a pressure sensor installed at the bolt connection of the tower. This is the current pressure value. This is the initial height from the top of the wind turbine to the ground. The current height from the top of the wind turbine to the ground. This is the maximum rotational speed of the wind turbine blades. This represents the current speed of the wind turbine blades;

[0088] Set a first threshold and a second threshold;

[0089] when When the value is less than or equal to the first threshold, a level one alarm is issued. When the value is greater than the first threshold and less than or equal to the second threshold, a level two alarm is issued. When the value exceeds the second threshold, a level 3 alarm is issued;

[0090] The severity of the level 1, level 2, and level 3 alarms increases sequentially.

[0091] In this embodiment, a combination analysis of multiple data is used. The confidence level is used as the calculation weight to correct the entire evaluation and obtain a relatively objective result. The final alarm result is output by threshold segmentation. The setting of the first threshold and the second threshold can be determined according to the specific data when the severity of the fault is different.

[0092] Furthermore, updating the time for the next data collection cycle based on the output value of the current neural network prediction model includes establishing a data collection cycle update model:

[0093]

[0094] In the formula, The updated collection cycle. This represents the current data collection cycle.

[0095] If no fault is detected, the data collection period can be appropriately increased while the collection frequency can be decreased. Alternatively, an upper limit can be set for the data collection period to prevent it from becoming too long; the collection period could be N times per month.

[0096] Secondly, the present invention also provides a wind turbine tower fault prediction system based on multi-source data fusion, characterized in that it includes a method for performing the above-mentioned wind turbine tower fault prediction method based on multi-source data fusion, and further includes:

[0097] The model building module is configured to acquire the basic data of the current wind turbine tower and the historical data of the wind turbine tower during the acquisition period when it is faulty, and preprocess the normal historical data and basic data; divide the preprocessed historical data into training set and validation set, build a neural network prediction model, train the neural network prediction model with the training set, optimize the neural network prediction model with the validation set, and output the optimal neural network prediction model; output the current prediction result based on the basic data and the neural network prediction model.

[0098] The output module is configured to output a faulty result if the prediction result indicates a fault exists, and issue an alarm based on the current basic data and the output value of the neural network prediction model; if the prediction result indicates no fault exists, the time for the next collection week is updated based on the current output value of the neural network prediction model.

[0099] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0100] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. This computer software product, stored in a storage medium, 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 the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0101] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for predicting wind turbine tower faults based on multi-source data fusion, characterized in that, include: Acquire the current basic data of the wind turbine tower and the historical data of the wind turbine tower during the collection period when it is faulty, and preprocess the normal historical data and basic data; The preprocessed historical data is divided into training and validation sets. A neural network prediction model is established. The neural network prediction model is trained using the training set and optimized using the validation set. The optimal neural network prediction model is then output. Based on the underlying data, a neural network prediction model outputs the current prediction result; If the prediction result indicates a fault, the fault result is output, and an alarm is issued based on the current basic data and the output value of the neural network prediction model. If the prediction result indicates that there is no fault, the time for the next collection week is updated based on the output value of the current neural network prediction model.

2. The wind turbine tower fault prediction method based on multi-source data fusion according to claim 1, characterized in that, The preprocessing of normal historical data and basic data includes: Data cleaning was performed based on K-means clustering, and the cleaned data was then normalized. The data cleaning includes: A historical dataset will be created and then input into K-means for clustering, resulting in two clusters. Calculate the Euclidean distance between each sampling point and the cluster centroid, and assign the sampling points to the clusters corresponding to the cluster centroids; For each cluster, recalculate the cluster centroid position. When the set number of clusters is reached or the cluster centroid position no longer changes, output the clustering result. A judgment threshold is set. If the distance between the centroids of two clusters is greater than the judgment threshold, the historical data at the current moment is abnormal, and the abnormal historical data is cleaned.

3. The wind turbine tower fault prediction method based on multi-source data fusion according to claim 2, characterized in that, It also includes establishing a feature extraction model to extract features from the normalized data: ； In the formula, Fault type This is the current hidden layer state. For activation function, The weight matrix is ​​input to the hidden layer. For the current input, This is the weight matrix from hidden layer to hidden layer. This is the bias vector for the hidden layer.

4. The wind turbine tower fault prediction method based on multi-source data fusion according to claim 3, characterized in that, The establishment of the neural network prediction model includes: ； ； ； In the formula, For the Gate of Oblivion Let be the weight matrix of the s-th layer of the forget gate. Let be the hidden state of the s-th layer at time t-1. The input at time t, Let be the bias vector of the forget gate. For input gate, Let be the weight matrix of the s-th layer of the input gate. Let be the bias vector of the input gate. This represents the state of the cell at time t. This represents the candidate cell state.

5. The wind turbine tower fault prediction method based on multi-source data fusion according to claim 4, characterized in that, It also includes loss functions, including: ； In the formula, The output of the loss function, The total number of fault types. The weights for calculating fault types, Let the intersection loss function be the type k fault. For hyperparameters, The weights for the s-th stage are calculated as follows: For loss measurement at different stages.

6. The wind turbine tower fault prediction method based on multi-source data fusion according to claim 5, characterized in that, The alarm issuance based on current basic data and the output value of the neural network prediction model includes: Obtain the confidence level of the current neural network model output, and establish an alarm evaluation model based on the confidence level. The alarm evaluation model includes: ； In the formula, As an evaluation index, The confidence score output by the neural network model. The standard pressure value is collected by a pressure sensor installed at the bolt connection of the tower. This is the current pressure value. This is the initial height from the top of the wind turbine to the ground. The current height from the top of the wind turbine to the ground. This is the maximum rotational speed of the wind turbine blades. This represents the current speed of the wind turbine blades; Set a first threshold and a second threshold; when When the value is less than or equal to the first threshold, a level one alarm is issued. When the value is greater than the first threshold and less than or equal to the second threshold, a level two alarm is issued. When the value exceeds the second threshold, a level 3 alarm is issued; The severity of the level 1, level 2, and level 3 alarms increases sequentially.

7. The wind turbine tower fault prediction method based on multi-source data fusion according to claim 6, characterized in that, The step of updating the time for the next data collection week based on the output value of the current neural network prediction model includes establishing a data collection cycle update model: ； In the formula, The updated collection cycle. This represents the current data collection cycle.

8. A wind turbine tower fault prediction system based on multi-source data fusion, characterized in that, The method includes a method for predicting wind turbine tower faults based on multivariate data fusion as described in any one of claims 1-7, and further includes: The model building module is configured to acquire the basic data of the current wind turbine tower and the historical data of the wind turbine tower during the acquisition period when it is faulty, and preprocess the normal historical data and basic data; divide the preprocessed historical data into training set and validation set, build a neural network prediction model, train the neural network prediction model with the training set, optimize the neural network prediction model with the validation set, and output the optimal neural network prediction model; output the current prediction result based on the basic data and the neural network prediction model. The output module is configured to output a faulty result if the prediction result indicates a fault exists, and issue an alarm based on the current basic data and the output value of the neural network prediction model; if the prediction result indicates no fault exists, the time for the next collection week is updated based on the current output value of the neural network prediction model.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the wind turbine tower fault prediction method based on multi-source data fusion as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the wind turbine tower fault prediction method based on multi-source data fusion as described in any one of claims 1-7.

Images