Machine learning-based land wind power fault prediction method and system
Through the machine learning-based fault prediction method, the operating data and environmental parameters of the wind turbine unit are used to build a fault diagnosis model, which realizes the advance identification of potential faults of the wind turbine unit, solves the problem of fault diagnosis delay in the existing technology, reduces maintenance costs and improves operating reliability.
Patent Information
- Application Number
- CN202510234771.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-02-28
- Publication Date
- 2025-06-27
AI Technical Summary
In the prior art, fault diagnosis of wind turbines is usually only done after a fault occurs, resulting in high maintenance costs and long downtime.
Using machine learning-based fault prediction method, preprocessing and feature extraction is performed by collecting vibration data, temperature data and speed data of wind turbine units, as well as environmental parameters such as wind speed, humidity and air pressure, the training set is constructed and a fault diagnosis model is constructed using convolutional neural networks and long and short-term memory networks to perform real-time diagnosis and fault prediction.
It realizes the advance identification of potential faults of wind turbines, reduces the probability of failures, reduces maintenance costs, and improves the operating reliability and efficiency of wind turbines.
Smart Images
Figure CN120216906A_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the technical field of fault prediction, and particularly relates to a method and system for onshore wind power fault prediction based on machine learning. Background Art
[0002] With the acceleration of the world's new energy technology transformation, as one of the directions of the global new energy transformation, wind energy has the characteristics of being clean, pollution-free and renewable. China has rich wind energy resources. In the existing technology, wind turbines are often used to convert wind energy into electrical energy. The working principle of a wind turbine is that the wind drives the windmill blades to rotate, and then the rotation speed is increased by a speed increaser, and then the generator is driven to rotate to convert mechanical energy into electrical energy.
[0003] The development of the wind power industry is becoming increasingly rapid, and the operation safety and reliability of wind turbines are attracting more and more attention. The traditional fault diagnosis methods of wind turbines can usually only diagnose after a fault occurs, resulting in high maintenance costs and long downtime. Therefore, how to provide an effective technical solution to achieve early identification of potential faults and reduce the probability of faults has become an urgent problem to be solved in the existing technology. Summary of the Invention
[0004] The purpose of the present invention is to provide a method and system for onshore wind power fault prediction based on machine learning, so as to solve the problem that the existing fault diagnosis methods can usually only diagnose after a fault occurs, resulting in high maintenance costs and long downtime.
[0005] To achieve the above purpose, the present invention adopts the following technical solutions:
[0006] In the first aspect, the present invention provides a method for onshore wind power fault prediction based on machine learning, including:
[0007] Collect the first wind turbine data and the first environmental parameters, preprocess the first wind turbine data and the first environmental parameters to obtain the second wind turbine data and the second environmental parameters. The first environmental parameters include wind speed parameters, humidity parameters and air pressure parameters at each time period, and the first wind turbine data includes vibration data, temperature data and rotation speed data at each time period;
[0008] Extract features from the second wind turbine data to obtain wind turbine feature data, perform correlation screening on the wind turbine feature data and the second environmental parameters, and construct a training set based on the screened data;
[0009] Construct a fault diagnosis model based on a machine learning model, input the training set into the fault diagnosis model for training, and obtain a trained fault diagnosis model;
[0010] Obtain the real-time operation data of the wind turbine generator set, diagnose the real-time operation data of the wind turbine generator set using the trained fault diagnosis model to obtain a diagnosis result, and predict whether there is a future fault in the wind turbine generator set according to the diagnosis result. Among them, the real-time operation data of the wind turbine generator set includes real-time vibration data, real-time temperature data, and real-time speed data.
[0011] In a possible design, preprocess the first wind turbine generator set data and the first environmental parameters, including:
[0012] Denoise, data clean, and normalize the wind speed parameter, humidity parameter, air pressure parameter, vibration data, temperature data, and speed data, and use the sliding window method to align the processed wind speed parameter, humidity parameter, air pressure parameter, vibration data, temperature data, and speed data for each time period.
[0013] In a possible design, extract features from the second wind turbine generator set data to obtain wind turbine generator set feature data, including:
[0014] Perform centering processing on the second wind turbine generator set data to obtain the centered second wind turbine generator set data;
[0015] Based on the centered second wind turbine generator set data, construct a data matrix, calculate the data matrix to obtain a covariance matrix;
[0016] Perform eigen-decomposition on the covariance matrix to obtain the eigenvalues of the covariance matrix and the eigenvectors corresponding to the eigenvalues of the covariance matrix;
[0017] According to the magnitudes of the eigenvalues, extract the corresponding eigenvectors to obtain wind turbine generator set feature data.
[0018] In a possible design, perform correlation screening on the wind turbine generator set feature data and the second environmental parameters, including:
[0019] Use kernel density estimation to calculate the joint probability distribution and marginal probability distribution;
[0020] Calculate the mutual information between the wind turbine generator set feature data and the second environmental parameters based on the joint probability distribution and marginal probability distribution. The mutual information is used to characterize the degree of association between the wind turbine generator set feature data and the second environmental parameters;
[0021] Use the recursive feature elimination method to iterate on the wind turbine generator set feature data, and remove the wind turbine generator set feature data with the smallest mutual information until the number of remaining wind turbine generator set feature data reaches a preset value.
[0022] In a possible design, the fault diagnosis model is constructed based on a convolutional neural network and a long short-term memory network. Among them, the fault diagnosis model includes a CNN layer and an LSTM layer. The CNN layer of the fault diagnosis model is used to extract spatial features from the input data according to the time series; the LSTM layer is used to input the features extracted in the CNN layer and perform analysis and diagnosis based on the time series information prediction ability.
[0023] In a possible design, predicting whether there will be a future fault in the wind turbine according to the diagnosis result includes:
[0024] Obtain the historical fault data of the wind turbine, and determine the fault threshold based on the historical fault data of the wind turbine;
[0025] Judge the diagnosis result based on the fault threshold. If the diagnosis result is greater than the fault threshold, it is determined that there is a future fault.
[0026] In a possible design, after it is determined that there is a future fault if the diagnosis result is greater than the fault threshold, the method further includes:
[0027] Classify the determined future fault, where the levels include red warning and yellow warning;
[0028] Generate a fault warning message according to the level of the determined future fault, and visually display the fault warning message.
[0029] In a second aspect, the present invention provides an onshore wind power fault prediction system based on machine learning, including:
[0030] A data acquisition module, configured to collect first wind turbine data and first environmental parameters, preprocess the first wind turbine data and the first environmental parameters to obtain second wind turbine data and second environmental parameters. The first environmental parameters include wind speed parameters, humidity parameters, and air pressure parameters at each time period, and the first wind turbine data includes vibration data, temperature data, and rotation speed data at each time period;
[0031] A feature extraction module, configured to extract features from the second wind turbine data to obtain wind turbine feature data, perform correlation screening on the wind turbine feature data and the second environmental parameters, and construct a training set based on the screened data;
[0032] A model training module, configured to construct a fault diagnosis model based on a machine learning model, input the training set into the fault diagnosis model for training, and obtain a trained fault diagnosis model;
[0033] A fault prediction module, which is used to obtain the real-time operation data of a wind turbine, diagnose the real-time operation data of the wind turbine by using a trained fault diagnosis model to obtain a diagnosis result, and predict whether there will be a fault in the future of the wind turbine according to the diagnosis result. The real-time operation data of the wind turbine includes real-time vibration data, real-time temperature data, and real-time speed data.
[0034] In a third aspect, the present invention provides a computer-readable storage medium, on which instructions are stored. When the instructions run on a computer, they execute the machine learning-based onshore wind power fault prediction method according to any one of the above first aspects.
[0035] In a fourth aspect, the present invention provides a computer device, including a memory, a processor, and a transceiver that are communicatively connected in sequence. The memory is used to store a computer program, the transceiver is used to send and receive messages, and the processor is used to read the computer program and execute the machine learning-based onshore wind power fault prediction method according to any one of the above first aspects.
[0036] The beneficial effects of the present invention are as follows:
[0037] The present invention discloses a machine learning-based onshore wind power fault prediction method and system. The method includes the following steps: collecting first wind turbine data and first environmental parameters, preprocessing the first wind turbine data and the first environmental parameters to obtain second wind turbine data and second environmental parameters. The first environmental parameters include wind speed parameters, humidity parameters, and air pressure parameters at each time period. The first wind turbine data includes vibration data, temperature data, and speed data at each time period; extracting features from the second wind turbine data to obtain wind turbine feature data, performing correlation screening on the wind turbine feature data and the second environmental parameters, and constructing a training set based on the screened data; constructing a fault diagnosis model based on a machine learning model, inputting the training set into the fault diagnosis model for training to obtain a trained fault diagnosis model; obtaining the real-time operation data of the wind turbine, diagnosing the real-time operation data of the wind turbine by using the trained fault diagnosis model to obtain a diagnosis result, and predicting whether there will be a fault in the future of the wind turbine according to the diagnosis result. The real-time operation data of the wind turbine includes real-time vibration data, real-time temperature data, and real-time speed data. The present invention uses a fault diagnosis model to perform predictive analysis on the operation data of the wind turbine to obtain a diagnosis result. According to the diagnosis result, it can predict the future faults of the wind turbine in advance, give an early warning of abnormal trends in the wind turbine, assist maintenance personnel in timely troubleshooting potential hazards, improve maintenance efficiency, and ensure the normal operation of the wind turbine. Description of the Drawings
[0038] Figure 1It is a flowchart of a method for onshore wind power fault prediction based on machine learning shown in the first aspect of this embodiment;
[0039] Figure 2 It is a block diagram of a module of a system for onshore wind power fault prediction based on machine learning shown in the second aspect of this embodiment. Detailed implementation manners
[0040] To more clearly illustrate the technical solutions in the embodiments of the present invention or in the prior art, the following will briefly introduce the present invention in combination with the accompanying drawings and the descriptions of the embodiments or the prior art. Obviously, the following descriptions of the structures of the accompanying drawings are only some embodiments of the present invention. For those of ordinary skill in the art, other accompanying drawings can be obtained based on these drawings without creative efforts. It should be noted here that the descriptions of these embodiment modes are used to help understand the present invention, but do not constitute a limitation to the present invention.
[0041] It should be understood that although terms such as first and second may be used herein to describe various objects, these objects should not be limited by these terms. These terms are only used to distinguish one object from another. For example, the first object can be called the second object, and similarly, the second object can be called the first object, without departing from the scope of the exemplary embodiments of the present invention.
[0042] Embodiment:
[0043] As Figure 1 shown, the first aspect of this embodiment provides a method for onshore wind power fault prediction based on machine learning, which can be, but is not limited to, executed by a computer device or virtual machine with certain computing resources, such as an electronic device such as a personal computer or a smart phone, or a virtual machine; as Figure 1 shown, the method for onshore wind power fault prediction based on machine learning can, but is not limited to, include the following steps:
[0044] S1. Collect the first wind turbine data and the first environmental parameters, preprocess the first wind turbine data and the first environmental parameters to obtain the second wind turbine data and the second environmental parameters. The first environmental parameters include wind speed parameters, humidity parameters, and air pressure parameters at each time period, and the first wind turbine data includes vibration data, temperature data, and rotational speed data at each time period;
[0045] Specifically, in step S1, preprocessing the first wind turbine data and the first environmental parameters includes:
[0046] Denoise, clean the data, and normalize the wind speed parameters, humidity parameters, air pressure parameters, vibration data, temperature data, and rotational speed data. Use the sliding window method to align the processed wind speed parameters, humidity parameters, air pressure parameters, vibration data, temperature data, and rotational speed data for each time period.
[0047] Among them, for denoising the data, denoising methods such as Fourier transform or wavelet transform can be selected to improve the data quality and enhance the data representativeness. For data cleaning, it can unify the data format and standard to ensure the analysis results. For data normalization, it can eliminate the differences between different data, facilitate data processing, and also includes filling in missing values and deleting outliers in the data.
[0048] Furthermore, the principle of using the sliding window method is to define a window by using two pointers and gradually process the data by moving these two pointers to find a continuous subsequence that meets specific conditions.
[0049] Using the sliding window method to align the processed wind speed parameters, humidity parameters, air pressure parameters, vibration data, temperature data, and rotational speed data for each time period includes the following steps:
[0050] S101. Determine the size and step size of the window, and initialize the window, where the step size is a preset time interval;
[0051] S102. Extract the data points within the window and calculate the statistics of the data within the window, where the statistics include the mean, median, and standard deviation;
[0052] S103. Slide the window forward by one step along the time axis and repeat step S102;
[0053] S104. Store the calculation results of each window to obtain the aligned result.
[0054] S2. Extract features from the second wind turbine data to obtain wind turbine feature data, perform correlation screening on the wind turbine feature data and the second environmental parameters, and construct a training set based on the screened data;
[0055] Specifically, in step S2, extracting features from the second wind turbine data to obtain wind turbine feature data includes:
[0056] S201. Centralize the second wind turbine data to obtain the centralized second wind turbine data;
[0057] The principle of centralization is to subtract the mean of the second wind turbine data from the data. Centralizing the data eliminates the errors caused by the self-variation or large numerical differences of the second wind turbine data and ensures the consistency of the data.
[0058] S202. Construct a data matrix based on the second wind turbine data after centralization, calculate the data matrix, and obtain a covariance matrix;
[0059] S203. Perform eigenvalue decomposition on the covariance matrix to obtain the eigenvalues of the covariance matrix and the eigenvectors corresponding to the eigenvalues of the covariance matrix;
[0060] S204. Extract the corresponding eigenvectors according to the magnitudes of the eigenvalues to obtain the wind turbine characteristic data.
[0061] Further, in step S2, perform correlation screening on the wind turbine characteristic data and the second environmental parameters, including:
[0062] S205. Use kernel density estimation to calculate the joint probability distribution and the marginal probability distribution;
[0063] Further, use kernel density estimation to calculate the probability density functions of the wind turbine characteristic data and the second environmental parameters. By integrating the probability density function of the wind turbine characteristic data, the joint probability distribution is obtained. By integrating the probability density function of the second environmental parameters, the marginal probability distribution is obtained. The working principle of kernel density estimation (KDE) is to place a kernel function around each data point and then superimpose these kernels to obtain the overall probability density function. The probability density function (PDF) is a function that describes the probability density of a continuous random variable near a certain determined value point. According to the integration of the probability density function, the probability distribution can be obtained.
[0064] S206. Calculate the mutual information between the wind turbine characteristic data and the second environmental parameters based on the joint probability distribution and the marginal probability distribution. The mutual information is used to characterize the degree of association between the wind turbine characteristic data and the second environmental parameters;
[0065] Mutual Information is used to measure the degree of association between two random variables and can be regarded as the amount of information contained in one random variable about another random variable.
[0066] S207. Use the recursive feature elimination method to iterate on the wind turbine characteristic data, remove the wind turbine characteristic data with the smallest mutual information until the number of remaining wind turbine characteristic data reaches a preset value.
[0067] Among them, the working principle of the Recursive Feature Elimination (RFE) method is to repeatedly train the model and eliminate the features that have less impact on the prediction results.
[0068] S3. Build a fault diagnosis model based on a machine learning model, input the training set into the fault diagnosis model for training, and obtain a trained fault diagnosis model;
[0069] Specifically, the fault diagnosis model is built based on a convolutional neural network and a long short-term memory network. Among them, the fault diagnosis model includes a CNN layer and an LSTM layer. The CNN layer of the fault diagnosis model is used to extract spatial features from the input data according to the time series; the LSTM layer is used to input the features extracted in the CNN layer and perform analysis and diagnosis based on the time series information prediction ability.
[0070] Among them, the Convolutional Neural Networks (CNN) is a deep learning model. Its core principle is to automatically extract local features in the data and combine complex features layer by layer by simulating the mechanism of the biological visual system; the Long Short-Term Memory (LSTM) is a recurrent neural network model designed to solve the problem of gradient disappearance or gradient explosion encountered by traditional RNNs when processing long sequence data. Through three gating mechanisms, namely the forget gate, the input gate, and the output gate, it can effectively handle long-term dependencies and predict and analyze time series data.
[0071] S4. Obtain the real-time operation data of the wind turbine, use the trained fault diagnosis model to diagnose the real-time operation data of the wind turbine, obtain the diagnosis result, and predict whether there will be a future fault in the wind turbine according to the diagnosis result. Among them, the real-time operation data of the wind turbine includes real-time vibration data, real-time temperature data, and real-time speed data.
[0072] Specifically, in step S4, predicting whether there will be a future fault in the wind turbine according to the diagnosis result includes:
[0073] S401. Obtain the historical fault data of the wind turbine and determine the fault threshold based on the historical fault data of the wind turbine;
[0074] S402. Judge the diagnosis result based on the fault threshold. If the diagnosis result is greater than the fault threshold, it is determined that there is a future fault.
[0075] Further, after step S402, if the diagnosis result is greater than the fault threshold and it is determined that there is a future fault, the method further includes:
[0076] S403. Classify the determined future faults, where the levels include red warnings and yellow warnings;
[0077] In a possible implementation, a red warning indicates that the predicted future fault is relatively serious and urgent and requires maintenance within 24 hours, and a yellow warning indicates that the predicted future fault requires maintenance within 72 hours.
[0078] S404. Generate a fault warning message according to the level of the determined future fault, and visually display the fault warning message.
[0079] In a possible design, before the real-time operation data of the wind turbine is transmitted, an encryption operation is performed on the real-time operation data of the wind turbine. Encryption methods such as asymmetric key encryption, hash function, or digital signature can be used for encryption. The above encryption methods are prior arts and can be selected according to specific usage scenarios and will not be elaborated here in detail. Performing the encryption operation can ensure information security and the security during data transmission.
[0080] In summary, in the first aspect of this embodiment, a method for predicting onshore wind power faults based on machine learning is provided. By collecting the first wind turbine data and the first environmental parameters, and preprocessing the first wind turbine data and the first environmental parameters to obtain the second wind turbine data and the second environmental parameters, feature extraction is performed on the second wind turbine data to obtain the wind turbine feature data, correlation screening is performed on the wind turbine feature data and the second environmental parameters, and a training set is constructed based on the screened data; a fault diagnosis model is constructed based on a machine learning model, the training set is input into the fault diagnosis model for training to obtain a trained fault diagnosis model; the real-time operation data of the wind turbine is obtained, and the trained fault diagnosis model is used to diagnose the real-time operation data of the wind turbine to obtain a diagnosis result, and based on the diagnosis result, it is predicted whether there will be a future fault in the wind turbine. In this embodiment, a fault diagnosis model is constructed using a machine learning model, the real-time operation data is used to predict the future faults of the wind turbine, early warnings are given for abnormal data, assisting maintenance personnel to detect potential hazards in time, improving the maintenance efficiency of the wind turbine, and ensuring the normal operation of the wind turbine.
[0081] As Figure 2 shown, in the second aspect of this embodiment, a system for predicting onshore wind power faults based on machine learning is provided, which is used to implement the method for predicting onshore wind power faults based on machine learning in the first aspect. The onshore wind power fault prediction system includes:
[0082] A data acquisition module, configured to collect first wind turbine data and first environmental parameters, preprocess the first wind turbine data and the first environmental parameters to obtain second wind turbine data and second environmental parameters. The first environmental parameters include wind speed parameters, humidity parameters, and air pressure parameters for each time period, and the first wind turbine data includes vibration data, temperature data, and rotational speed data for each time period.
[0083] A feature extraction module, configured to extract features from the second wind turbine data to obtain wind turbine feature data, perform correlation screening on the wind turbine feature data and the second environmental parameters, and construct a training set based on the screened data.
[0084] A model training module, configured to construct a fault diagnosis model based on a machine learning model, input the training set into the fault diagnosis model for training, and obtain a trained fault diagnosis model.
[0085] A fault prediction module, configured to obtain real-time operation data of the wind turbine, use the trained fault diagnosis model to diagnose the real-time operation data of the wind turbine to obtain a diagnosis result, and predict whether there is a future fault in the wind turbine according to the diagnosis result. The real-time operation data of the wind turbine includes real-time vibration data, real-time temperature data, and real-time rotational speed data.
[0086] Specifically, install multiple sensors on the wind turbine to collect the first wind turbine data. For example, install temperature sensors, vibration sensors, and acceleration sensors at the blades, gearbox, and generator.
[0087] In summary, a land-based wind power fault prediction system based on machine learning provided in this embodiment includes a data acquisition module, a feature extraction module, a model training module, and a fault prediction module. The data acquisition module collects the first wind turbine data and the first environmental parameters, preprocesses the first wind turbine data and the first environmental parameters to obtain the second wind turbine data and the second environmental parameters. The feature extraction module extracts features from the second wind turbine data to obtain wind turbine feature data, performs correlation screening on the wind turbine feature data and the second environmental parameters, and constructs a data set based on the screened data. The model training module constructs a fault diagnosis model based on a machine learning model, inputs the training set into the fault diagnosis model for training, and obtains a trained fault diagnosis model. The fault prediction module obtains the real-time operation data of the wind turbine, uses the trained fault diagnosis model to diagnose the real-time operation data of the wind turbine to obtain a diagnosis result, and predicts whether there is a future fault in the wind turbine according to the diagnosis result. Using the trained fault diagnosis model to diagnose the real-time operation data, timely diagnose abnormal data, assist maintenance personnel in timely troubleshooting potential hazards, improve maintenance efficiency, ensure the normal operation of the wind turbine, and reduce maintenance costs.
[0088] In the third aspect of this embodiment, a computer-readable storage medium is provided. Instructions are stored on the computer-readable storage medium. When the instructions run on a computer, they are used to execute the machine learning-based onshore wind power fault prediction method described in the first aspect of the embodiment. Among them, the computer-readable storage medium refers to a carrier for storing data, and may include, but is not limited to, computer-readable storage media such as floppy disks, optical discs, hard disks, flash memories, USB flash drives, and / or memory sticks. The computer may be a general-purpose computer, a dedicated computer, a computer network, or other programmable devices.
[0089] For the working process, working details, and technical effects of the aforementioned computer-readable storage medium provided in the third aspect of this embodiment, reference may be made to the machine learning-based onshore wind power fault prediction method described in the first aspect, and details will not be elaborated herein.
[0090] In the fourth aspect of this embodiment, a computer device is provided, including a memory, a processor, and a transceiver that are communicatively connected in sequence. Among them, the memory is used to store a computer program, the transceiver is used to send and receive messages, and the processor is used to read the computer program and execute the machine learning-based onshore wind power fault prediction method described in the first aspect of the embodiment.
[0091] Specifically, the memory may include, but is not limited to, random access memory (RAM), read-only memory (ROM), flash memory, first-in-first-out memory (FIFO), and / or first-in-last-out memory (FILO), etc.; the processor may not be limited to using microprocessors of the STM32F105 series, architectures such as ARM (Advanced RISC Machines), X86, or processors integrated with NPU (neural-network processing units); the transceiver may include, but is not limited to, WiFi (Wireless Fidelity) wireless transceivers, Bluetooth wireless transceivers, General Packet Radio Service (GPRS) wireless transceivers, ZigBee (a low-power local area network protocol based on the IEEE802.15.4 standard) wireless transceivers, 3G transceivers, 4G transceivers, and / or 5G transceivers, etc.
[0092] For the working process, working details, and technical effects of the device provided in the fifth aspect of this embodiment, reference may be made to the machine learning-based onshore wind power fault prediction method described in the first aspect of the embodiment, and details will not be elaborated herein.
[0093] Finally, it should be noted that the above are only preferred embodiments of the present invention and are not intended to limit the protection scope of the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.
Claims
1. A method for predicting onshore wind power faults based on machine learning, characterized in that: include: Collecting first wind turbine data and first environmental parameters, preprocessing the first wind turbine data and the first environmental parameters to obtain second wind turbine data and second environmental parameters, wherein the first environmental parameters include wind speed parameters, humidity parameters, and air pressure parameters in each time period, and the first wind turbine data include vibration data, temperature data, and speed data in each time period; Extracting features from the data of the second wind turbine generator set to obtain characteristic data of the wind turbine generator set, screening the correlation between the characteristic data of the wind turbine generator set and the second environmental parameter, and constructing a training set based on the screened data; Building a fault diagnosis model based on the machine learning model, inputting the training set into the fault diagnosis model for training, and obtaining a trained fault diagnosis model; The real-time operation data of the wind turbine is obtained, and the real-time operation data of the wind turbine is diagnosed using the trained fault diagnosis model to obtain the diagnosis result, and the whether the wind turbine will have a fault in the future is predicted based on the diagnosis result. The real-time operation data of the wind turbine includes real-time vibration data, real-time temperature data and real-time speed data.
2. The onshore wind power fault prediction method based on machine learning according to claim 1, characterized in that: Preprocessing the first wind turbine generator set data and the first environmental parameter includes: The wind speed parameters, humidity parameters, air pressure parameters, vibration data, temperature data and speed data were denoised, cleaned and normalized, and the sliding window method was used to align the processed wind speed parameters, humidity parameters, air pressure parameters, vibration data, temperature data and speed data for each time period.
3. The onshore wind power fault prediction method based on machine learning according to claim 1 is characterized in that: Feature extraction is performed on the second wind turbine generator set data to obtain wind turbine generator set feature data, including: Centralizing the data of the second wind turbine generator set to obtain centralized data of the second wind turbine generator set; A data matrix is constructed based on the centralized data of the second wind turbine generator set, and the data matrix is calculated to obtain a covariance matrix; Perform eigendecomposition on the covariance matrix to obtain the eigenvalues of the covariance matrix and the eigenvectors corresponding to the eigenvalues of the covariance matrix; According to the size of the eigenvalue, the corresponding eigenvector is extracted to obtain the characteristic data of the wind turbine.
4. The onshore wind power fault prediction method based on machine learning according to claim 1, characterized in that: The correlation between the characteristic data of wind turbines and the second environmental parameters is screened, including: Compute joint and marginal probability distributions using kernel density estimation; Calculate mutual information between the characteristic data of the wind turbine generator set and the second environmental parameter based on the joint probability distribution and the marginal probability distribution, wherein the mutual information is used to characterize the degree of association between the characteristic data of the wind turbine generator set and the second environmental parameter; The wind turbine characteristic data are iterated using a recursive feature elimination method to remove the wind turbine characteristic data with the smallest mutual information until a preset number of remaining wind turbine characteristic data is reached.
5. The onshore wind power fault prediction method based on machine learning according to claim 1, characterized in that: The fault diagnosis model is constructed based on convolutional neural networks and long short-term memory networks, wherein the fault diagnosis model includes a CNN layer and an LSTM layer. The CNN layer of the fault diagnosis model is used to extract spatial features of input data according to time series; the LSTM layer is used to input the features extracted in the CNN layer and perform analysis and diagnosis based on the prediction capability of time series information.
6. The onshore wind power fault prediction method based on machine learning according to claim 1, characterized in that: Based on the diagnostic results, predict whether the wind turbine will have future faults, including: Acquire historical fault data of the wind turbine generator set, and determine a fault threshold based on the historical fault data of the wind turbine generator set; The diagnosis result is determined based on the fault threshold. If the diagnosis result is greater than the fault threshold, it is determined that there is a future fault.
7. The onshore wind power fault prediction method based on machine learning according to claim 6, characterized in that: If the diagnosis result is greater than the fault threshold, after determining that there is a future fault, the method further includes: Classifying the determined future faults into levels, wherein the levels include red warning and yellow warning; Fault warning information is generated according to the determined level of future faults, and the fault warning information is visualized.
8. An onshore wind power fault prediction system based on machine learning, characterized in that: include: a data acquisition module, used for collecting first wind turbine data and first environmental parameters, preprocessing the first wind turbine data and the first environmental parameters to obtain second wind turbine data and second environmental parameters, wherein the first environmental parameters include wind speed parameters, humidity parameters and air pressure parameters in each time period, and the first wind turbine data include vibration data, temperature data and rotation speed data in each time period; A feature extraction module is used to extract features from the second wind turbine generator set data to obtain wind turbine generator set feature data, perform correlation screening on the wind turbine generator set feature data and the second environmental parameter, and construct a training set based on the screened data; A model training module is used to build a fault diagnosis model based on the machine learning model, input the training set into the fault diagnosis model for training, and obtain a trained fault diagnosis model; The fault prediction module is used to obtain the real-time operation data of the wind turbine, diagnose the real-time operation data of the wind turbine using the trained fault diagnosis model, obtain the diagnosis result, and predict whether the wind turbine will have a fault in the future based on the diagnosis result. The real-time operation data of the wind turbine includes real-time vibration data, real-time temperature data and real-time speed data.
9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores instructions, and when the instructions are executed on a computer, the onshore wind power fault prediction method based on machine learning according to any one of claims 1 to 7 is executed.
10. A computer device, characterized in that: It comprises a memory, a processor and a transceiver which are communicatively connected in sequence, wherein the memory is used to store a computer program, the transceiver is used to send and receive messages, and the processor is used to read the computer program to execute the onshore wind power fault prediction method based on machine learning as described in any one of claims 1 to 7.