Wind turbine generator fault detection method based on machine learning and related device

Abstract

The invention discloses a wind turbine generator fault detection method based on machine learning and a related device. The method comprises the following steps: acquiring SCADA data of a wind power plant by using an SCADA system; the SCADA data of the wind power plant are input into the trained asymmetric variational automatic coding Gaussian mixture model based on the long-short-term memory network, whether the wind turbine generator fails or not is detected, and the method and the related device can accurately detect and predict the failure of the wind turbine generator.

Description

Technical Field

[0001] The present invention belongs to the technical field of wind power generation and relates to a wind turbine fault detection method based on machine learning and a related device. Background Art

[0002] With the increasing global demand for renewable energy, wind power, as an important green energy source, has become a key component of the power system. The operation and maintenance of wind turbines are crucial factors in ensuring the efficiency of wind power generation. Traditional wind turbine maintenance methods rely primarily on manual monitoring and empirical maintenance, which cannot effectively provide early warning and real-time monitoring of faults, resulting in long downtime and high maintenance costs. In recent years, with the rapid development of big data, artificial intelligence (AI), and machine learning technologies, the intelligent operation and maintenance of wind turbines has become a research hotspot.

[0003] Although there are currently some wind turbine health monitoring methods based on data analysis, they have the following shortcomings in actual applications: First, existing methods mostly rely on simple statistical analysis, which makes it difficult to capture complex nonlinear relationships; second, the operating environment of wind turbines is changeable, and the time-varying and nonlinear characteristics of the data make it difficult for traditional models to cope with it; third, there is a lack of effective root cause analysis methods, which leads to inaccurate maintenance decisions after a fault occurs.

[0004] Therefore, there is an urgent need for an intelligent operation and maintenance strategy for wind turbines based on advanced machine learning technology to improve the fault detection, diagnosis and prevention capabilities of wind turbines and optimize the overall operation and maintenance efficiency of wind turbines. Summary of the Invention

[0005] The purpose of the present invention is to overcome the shortcomings of the above-mentioned prior art and provide a wind turbine fault detection method and related devices based on machine learning, which can accurately detect and predict wind turbine faults.

[0006] To achieve the above objectives, the present invention discloses a wind turbine fault detection method based on machine learning, comprising:

[0007] Use the SCADA system to obtain SCADA data of wind farms;

[0008] The SCADA data of the wind farm is input into a trained asymmetric variational autoencoding Gaussian mixture model based on a long short-term memory network to detect whether a wind turbine generator set has failed.

[0009] The further improvement of the wind turbine fault detection method based on machine learning described in the present invention is:

[0010] Furthermore, the SCADA data includes at least one of the rotational speed, wind speed, temperature, vibration, generated power and output power of the wind turbine.

[0011] Furthermore, it also includes:

[0012] The causes of wind turbine failure are analyzed using the SHAP method.

[0013] Furthermore, the process of analyzing the cause of wind turbine failure using the SHAP method is as follows:

[0014] Calculate the Shapley value of different data types in SCADA data for wind turbine faults;

[0015] The root cause of wind turbine failure is determined based on the Shapley value of different data types in SCADA data.

[0016] Furthermore, the calculation formula of the Shapley value is:

[0017]

[0018] Among them, φ i is the Shapley value of the i-th feature; S is the feature subset, excluding feature i; f(S) is the model output of predicting the sample using only the features of subset S; |N| is the total number of features.

[0019] Furthermore, it also includes:

[0020] According to the root cause of the wind turbine failure, a maintenance suggestion is generated, wherein the maintenance suggestion includes maintenance components, maintenance timing and maintenance method.

[0021] The present invention discloses a wind turbine fault detection system based on machine learning, comprising:

[0022] An acquisition module, used to acquire SCADA data of the wind farm using the SCADA system;

[0023] The detection module is used to input the SCADA data of the wind farm into a trained asymmetric variational autoencoding Gaussian mixture model based on a long short-term memory network to detect whether a wind turbine generator set has failed.

[0024] The further improvement of the wind turbine fault detection system based on machine learning described in the present invention is:

[0025] Furthermore, it also includes:

[0026] The causes of wind turbine failure are analyzed using the SHAP method.

[0027] The present invention discloses a computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the steps of the wind turbine fault detection method based on machine learning are implemented.

[0028] The present invention discloses a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, the steps of the wind turbine fault detection method based on machine learning are implemented.

[0029] The present invention has the following beneficial effects:

[0030] During specific operation, the wind turbine fault detection method and related devices based on machine learning described in the present invention input the SCADA data of the wind farm into a trained asymmetric variational auto-encoding Gaussian mixture model based on a long short-term memory network to detect whether the wind turbine has a fault. The asymmetric variational auto-encoding Gaussian mixture model (LSTM-AVAGMM) is used to perform early anomaly detection of the wind turbine. The LSTM model can effectively capture the time series characteristics of the wind turbine during operation, and the combination of the variational autoencoder (VAE) and the Gaussian mixture model (GMM) can further enhance the anomaly detection capability of the model. The model achieves real-time monitoring of the operating status of the wind turbine by fitting the distribution of SCADA data and performing anomaly detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] The accompanying drawings, which constitute part of the present invention, are intended to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are intended to explain the present invention and do not constitute an undue limitation of the present invention. In the accompanying drawings:

[0032] Figure 1 is a flow chart of the method of the present invention;

[0033] Figure 2 This is the flow chart of the wind turbine SCADA system;

[0034] Figure 3 This is a structural diagram of the LSTM neural network;

[0035] Figure 4 Schematic diagram of the structure of the variational autoencoder;

[0036] Figure 5 The schematic diagram of the SHAP method. DETAILED DESCRIPTION

[0037] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of them. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0038] In the description of the present invention, it is to be understood that the terms “include” and “comprise” indicate the presence of the described features, wholes, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, wholes, steps, operations, elements, components and / or collections thereof.

[0039] It should also be understood that the terms used in the present specification are only for the purpose of describing particular embodiments and are not intended to limit the present invention. As used in the present specification and the appended claims, the singular forms "a", "an", and "the" are intended to include the plural forms unless the context clearly indicates otherwise.

[0040] It should be further understood that the term "and / or" as used in the present specification and the appended claims refers to and includes any and all possible combinations of one or more of the associated listed items. For example, A and / or B may represent: A exists alone, A and B exist simultaneously, and B exists alone. In addition, the character " / " in the present invention generally indicates that the associated objects are in an "or" relationship.

[0041] It should be understood that although the terms "first," "second," and "third" may be used to describe preset ranges in embodiments of the present invention, these preset ranges should not be limited to these terms. These terms are merely used to distinguish one preset range from another. For example, without departing from the scope of embodiments of the present invention, the first preset range may also be referred to as the second preset range, and similarly, the second preset range may also be referred to as the first preset range.

[0042] The word "if," as used herein, may be interpreted as "at the time of" or "when" or "in response to determining" or "in response to detecting," depending on the context. Similarly, the phrases "if it is determined" or "if (stated condition or event) is detected" may be interpreted as "when it is determined" or "in response to the determination" or "when detecting (stated condition or event)" or "in response to detecting (stated condition or event)," depending on the context.

[0043] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Generally, the components of the embodiments of the present invention described and shown in the drawings herein can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the drawings is not intended to limit the scope of the claimed invention, but merely represents selected embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0044] The accompanying drawings illustrate various schematic diagrams of structures according to embodiments disclosed herein. These figures are not drawn to scale; for clarity, some details are exaggerated and some details may be omitted. The shapes of the various regions and layers shown in the figures, as well as their relative sizes and positional relationships, are merely exemplary and may deviate in practice due to manufacturing tolerances or technical limitations. Those skilled in the art may design regions / layers with different shapes, sizes, and relative positions as needed.

[0045] Example 1

[0046] refer to Figure 1 The wind turbine fault detection method based on machine learning of the present invention comprises the following steps:

[0047] 1) Use the SCADA system to obtain SCADA data of the wind farm;

[0048] It should be noted that the main functions of the SCADA system are:

[0049] 1a) Collection of on-site unit operation data.

[0050] 2a) Storage of unit operation historical data.

[0051] 3a) A database that provides information about all processes.

[0052] 4a) Provide unit operation fault status and related data display.

[0053] 5a) Perform remote control of field devices.

[0054] 6a) Perform system status monitoring and diagnosis and take appropriate measures.

[0055] The structural hierarchy of the SCADA system is mainly divided into three parts, namely the wind farm monitoring part, the central monitoring part and the remote monitoring part. Among them, the wind farm monitoring part is located in the control cabinet of the unit's cabin or tower, and uses the measurement points distributed in each system to comprehensively monitor the unit's operating data and status. The data sampling time is generally 1 second, and it is fed back to other parts of the SCADA system; the central monitoring part is located in the wind farm's monitoring room, which monitors the real-time operating status of the unit through the information provided by the remote monitoring part, and can adjust the status of the unit according to various needs; the remote monitoring part is generally located in the wind farm operator, which is convenient for managing multiple wind farms at the same time. The operation and management personnel can remotely connect to the wind farm monitoring room and control the host to remotely monitor and adjust the wind turbines.

[0056] 2) Preprocessing of the collected SCADA data requires data elimination, correction and cleaning, and normalization to ensure data stability and accuracy.

[0057] 3) Construct an asymmetric variational autoencoding Gaussian mixture model (LSTM-AVAGMM);

[0058] refer to Figure 2 、 Figure 3 、 Figure 4 and Figure 5 , LSTM network is used for time series feature extraction.

[0059] In wind turbines, many parameters (such as speed, power, and temperature) have temporal dependencies, and LSTM can extract these dependencies well.

[0060] It should be noted that LSTM is a time series algorithm commonly used to process and predict time-based sequence data. LSTM has two key variables: the hidden state h, which is primarily responsible for memorizing short-term information, especially the current time step, and the cell state C, which is primarily responsible for long-term memory. It has three key gating units: the input gate, the forget gate, and the output gate. These gating units control the flow of information through learning, helping the LSTM network better handle long-term dependencies. To further improve prediction accuracy, the TPE algorithm is also used to optimize the LSTM network's hyperparameters, including the number of neurons, learning rate, and activation function.

[0061] Anomaly detection is performed using the Asymmetric Variational Autoencoder Gaussian Mixture Model (AVAGMM). AVAGMM is a variational autoencoder (VAE)-based model that uses a Gaussian Mixture Model (GMM) as an approximation of the probability distribution for data generation and anomaly detection. Its asymmetric design means that the model uses different network structures during encoding and decoding. The encoder may focus on capturing a characteristic representation of the data, while the decoder attempts to reconstruct the original data from this representation, which helps better capture the complex distribution of the data.

[0062] The core formula of the variational autoencoder (VAE) is:

[0063] L VAE =E q(z|x) [logp(x|z)]-KL[q(z|x)||p(z)]

[0064] Where x is the input data; z is the latent variable; q(z|x) is the encoder, used to estimate the latent distribution; p(x|z) is the decoder, used to generate data from the latent variable; p(z) is the prior distribution of the latent variable (usually assumed to be a standard normal distribution); KL[q(z|x)║p(z)] is the Kullback-Leibler divergence, which is used to measure the difference between the two distributions.

[0065] Standard VAE usually assumes that the latent distributions q(z|x) and p(z) are symmetric distributions, such as Gaussian distributions. In AVAGMM, asymmetric latent distribution modeling is introduced by using skewed distributions or parameterized asymmetric distributions for modeling.

[0066] The asymmetric latent distribution model is:

[0067] q(z|x)=N(z;μ(x),Σ(x)).f(z;λ)

[0068] Where μ(x) is the mean of the encoder output; Σ(x) is the covariance matrix of the encoder output; f(z;λ) is the asymmetric correction function, and λ is a parameter that controls asymmetry (such as the coefficient of skewness or higher-order moments).

[0069] In general, f(z;λ) can be taken as a correction term of the skew Gaussian distribution or other asymmetric functions. By adjusting f(z;λ), the potential distribution can flexibly adapt to the asymmetry of the data.

[0070] A Gaussian mixture model (GMM) is a probabilistic model that represents a data point as belonging to a mixture of multiple Gaussian distributions:

[0071]

[0072] Where K is the number of Gaussian distributions; π k is the weight of the kth Gaussian distribution, satisfying μ k ,Σ k is the mean and covariance matrix of the kth Gaussian component.

[0073] Extract samples from the latent variable z output by GMM, and use kernel density estimation (KDE) to fit its probability density function as follows:

[0074]

[0075] Among them, K(t) is the kernel function; h is the bandwidth in the kernel density estimation, which controls the smoothness of the kernel.

[0076] Based on the probability density function fitted by KDE, the threshold corresponding to the 99.7% confidence interval is calculated, and samples below this quantile are set as outliers. When the anomaly probability exceeds the set threshold, the system will issue a fault warning, notifying operations and maintenance personnel in advance to address the problem, thereby avoiding major failures.

[0077] 4) Input the SCADA data into the trained asymmetric variational autoencoder Gaussian mixture model (LSTM-AVAGMM) to detect whether the wind turbine has any abnormality.

[0078] After anomaly detection and fault warning, root cause analysis is performed using the SHAP method. By calculating the SHAP value of each input feature (such as wind speed, temperature, and rotational speed), the factors that have the greatest impact on the fault are analyzed, providing operations and maintenance personnel with a clear and explainable decision-making basis to optimize maintenance strategies and resource scheduling.

[0079] The calculation formula of Shapley value is:

[0080]

[0081] Among them, φ i is the Shapley value of the i-th feature; S is the feature subset, excluding feature i; f(S) is the model output of predicting the sample using only the features of subset S; |N| is the total number of features.

[0082] During operation, new SCADA system data is sampled regularly. When new data is input, the system automatically updates the training data and performs incremental learning, enabling the model to adapt to new wind turbine operating conditions and continuously improve the accuracy of detection and prediction.

[0083] Based on the results of root cause analysis and fault prediction, intelligent maintenance recommendations are generated, including recommended repair parts, repair timing, and repair methods. These recommendations will serve as an important basis for wind turbine operation and maintenance decisions, helping operators improve maintenance efficiency and reduce turbine downtime.

[0084] It should be noted that the present invention utilizes a SCADA system that integrates data monitoring, information collection, and fault alarms, with wide coverage and stable performance. Through the LSTM-AVAGMM model and the 99.7% confidence interval threshold setting, it can accurately detect early abnormalities during the operation of wind turbines, thereby avoiding unplanned shutdowns or performance degradation of wind turbines. The introduction of the SHAP method can deeply analyze the root causes behind the abnormalities and provide accurate fault diagnosis for operation and maintenance personnel, thereby optimizing maintenance strategies. The incremental training mechanism ensures that the system can adapt to changes in wind turbines in real time, improving the long-term effectiveness and accuracy of the model. Through effective anomaly detection and prediction, the present invention can detect potential faults in advance, reduce maintenance costs and unit downtime, and improve the overall operating efficiency of wind turbines.

[0085] Example 2

[0086] The wind turbine fault detection system based on machine learning of the present invention comprises:

[0087] An acquisition module, used to acquire SCADA data of the wind farm using the SCADA system;

[0088] The detection module is used to input the SCADA data of the wind farm into a trained asymmetric variational autoencoding Gaussian mixture model based on a long short-term memory network to detect whether a wind turbine generator set has failed.

[0089] In this embodiment, it also includes:

[0090] The causes of wind turbine failure are analyzed using the SHAP method.

[0091] The division of modules in the embodiments of the present application is illustrative and is merely a logical functional division. In actual implementation, other division methods may be used. Furthermore, the functional modules in the various embodiments of the present application may be integrated into a single processor, or may exist physically separately, or two or more modules may be integrated into a single module. The aforementioned integrated modules may be implemented in the form of hardware or software functional modules.

[0092] Example 3

[0093] A computer device includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the steps of the machine learning-based wind turbine fault detection method are implemented, for example, including: obtaining SCADA data from a wind farm using a SCADA system; inputting the SCADA data from the wind farm into a trained asymmetric variational autoencoder Gaussian mixture model based on a long short-term memory network to detect whether a wind turbine fault has occurred. The SCADA data includes at least one of the wind turbine's rotational speed, wind speed, temperature, vibration, generated power, and output power. The memory may include a RAM, such as a high-speed random access memory (RAM), or may also include a non-volatile memory, such as at least one disk drive. The processor, network interface, and memory are interconnected via an internal bus, which may be an industry standard architecture bus, a peripheral component interconnect standard bus, an extended industry standard architecture bus, or the like. The bus may be classified as an address bus, a data bus, a control bus, or the like. The memory is used to store programs. Specifically, the programs may include program code, and the program code includes computer operating instructions. The memory may include both RAM and non-volatile memory, and provides instructions and data to the processor.

[0094] Example 4

[0095] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of the wind turbine fault detection method based on machine learning, for example, including: using a SCADA system to obtain SCADA data of a wind farm; inputting the SCADA data of the wind farm into a trained asymmetric variational autoencoding Gaussian mixture model based on a long short-term memory network to detect whether a wind turbine fault occurs, wherein the SCADA data includes at least one of the rotational speed, wind speed, temperature, vibration, generated power and output power of the wind turbine. Specifically, the computer-readable storage medium includes, but is not limited to, volatile memory and / or non-volatile memory. The volatile memory may include random access memory and / or cache memory, etc. The non-volatile memory may include a read-only memory, a hard disk, a flash memory, an optical disk, a magnetic disk, etc.

[0096] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Furthermore, the present application may take the form of a computer program product implemented on one or more computer-usable storage media containing computer-usable program code.

[0097] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the steps in the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0098] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0099] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 The steps for the function specified in one or more boxes.

[0100] Those skilled in the art will readily identify other embodiments of the present invention after considering the specification and disclosure of the invention. This application is intended to cover any variations, uses, or adaptations of the present invention that follow the general principles of the invention and include common knowledge or customary techniques in the art not disclosed herein. The description and examples are to be considered as exemplary only, with the true scope and spirit of the invention being indicated by the following claims.

[0101] It should be understood that the present invention is not limited to the exact construction described above and shown in the drawings, and that various modifications and changes may be made without departing from the scope thereof. The scope of the present invention is limited only by the appended claims.

[0102] The above description is only a preferred embodiment of the present invention and does not limit the present invention in any way. Any simple modification, change and equivalent structural change made to the above embodiment based on the technical essence of the present invention shall still fall within the scope of protection of the technical solution of the present invention.

Claims

1. A wind turbine fault detection method based on machine learning, characterized in that: include: Use the SCADA system to obtain SCADA data of wind farms; The SCADA data of the wind farm is input into a trained asymmetric variational autoencoding Gaussian mixture model based on a long short-term memory network to detect whether a wind turbine generator set has failed.

2. The wind turbine fault detection method based on machine learning according to claim 1, characterized in that: The SCADA data includes at least one of the rotation speed, wind speed, temperature, vibration, generated power and output power of the wind turbine generator set.

3. The wind turbine fault detection method based on machine learning according to claim 2, characterized in that: Also includes: The causes of wind turbine failure are analyzed using the SHAP method.

4. The wind turbine fault detection method based on machine learning according to claim 3, characterized in that: The process of analyzing the cause of wind turbine failure using the SHAP method is as follows: Calculate the Shapley value of different data types in SCADA data for wind turbine faults; The root cause of wind turbine failure is determined based on the Shapley value of different data types in SCADA data.

5. The wind turbine fault detection method based on machine learning according to claim 4, characterized in that: The calculation formula of the Shapley value is: Among them, φ i is the Shapley value of the i-th feature; S is the feature subset, excluding feature i; f(S) is the model output of predicting the sample using only the features of subset S; |N| is the total number of features.

6. The wind turbine fault detection method based on machine learning according to claim 4, characterized in that: Also includes: According to the root cause of the wind turbine failure, a maintenance suggestion is generated, wherein the maintenance suggestion includes maintenance components, maintenance timing and maintenance method.

7. A wind turbine fault detection system based on machine learning, characterized in that: include: An acquisition module, used to acquire SCADA data of the wind farm using the SCADA system; The detection module is used to input the SCADA data of the wind farm into a trained asymmetric variational autoencoding Gaussian mixture model based on a long short-term memory network to detect whether a wind turbine generator set has failed.

8. The wind turbine fault detection system based on machine learning according to claim 7, characterized in that: Also includes: The causes of wind turbine failure are analyzed using the SHAP method.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the steps of the wind turbine fault detection method based on machine learning as described in any one of claims 1 to 6 are implemented.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the wind turbine fault detection method based on machine learning as described in any one of claims 1 to 6 are implemented.

Images