Wind turbine generator transmission chain anomaly detection method and system based on rotation speed guidance

Abstract

The invention discloses a wind turbine generator transmission chain anomaly detection method and system based on rotation speed guidance, a wind turbine generator uses a permanent magnet synchronous generator, and the method comprises the following steps: obtaining a vibration acceleration signal of a wind turbine generator transmission chain; calculating the rotating speed of a permanent magnet synchronous generator of the wind turbine generator according to the obtained vibration acceleration signal, and extracting multi-dimensional features from the vibration acceleration signal; taking the rotating speed of the permanent magnet synchronous generator as a modulation condition, and constructing a transmission chain anomaly detection model based on a conditional neural network; and inputting the multi-dimensional features into a transmission chain anomaly detection model, detecting the current state of the transmission chain of the wind turbine generator, and outputting a detection result. According to the method, a physical mechanism and data driving are combined, the wind turbine generator transmission chain anomaly detection performance in a real industrial environment is improved under the condition of not depending on an externally input rotating speed log, adaptability, real-time performance and high generalization are achieved, and meanwhile the transmission chain anomaly detection precision can be effectively improved.

Description

Technical Field

[0001] This invention relates to the technical field of abnormal state detection of wind turbine generators, and in particular to a method and system for detecting abnormalities in the transmission chain of wind turbine generators based on rotational speed guidance. Background Technology

[0002] Wind turbines using permanent magnet synchronous generators (PMSGs) include direct-drive and semi-direct-drive turbines. During long-term operation, their drivetrains are prone to various mechanical failures, such as bearing wear, gear tooth breakage, and generator bearing wear. To achieve early monitoring of the health status of wind turbines using PMSGs, the industry commonly employs a Condition Monitoring System (CMS) to collect vibration acceleration signals on the drivetrain of these turbines. Anomaly detection and fault diagnosis are then performed using spectral analysis or feature-based machine learning methods.

[0003] However, existing technologies have the following problems: 1. Strong dependence on operating conditions: Wind turbines using permanent magnet synchronous generators typically operate in variable speed and pitch mode, and their vibration characteristics dynamically evolve with fluctuations in rotational speed. See also Figure 1 The image shows a waterfall chart of "speed-frequency-amplitude" plotted using measured data from a semi-direct drive wind turbine. Figure 1 It can be intuitively observed that the spectral components and their amplitude energy distribution shift significantly with changes in rotational speed. However, traditional analysis methods are often based on the assumption of "approximate constant speed," which is severely out of touch with actual unsteady operating conditions. This results in inherent physical deficiencies in diagnostic models built upon such assumptions. Furthermore, in wind power systems, multiple factors such as wind speed, wind direction, and power curves collectively constitute complex operating conditions, ultimately mapping to the real-time rotational speed of the wind turbine.

[0004] 2. Speed ​​measurement relies on external signals: Most monitoring methods depend on control system logs or speed sensor data. However, these external signals are prone to synchronization delays or absences in actual operating conditions, resulting in inconsistencies between the characteristics and the actual speed, affecting the accuracy of anomaly detection.

[0005] 3. Deep learning methods that do not consider physical laws: In recent years, some studies have attempted to use deep learning methods to model vibration signals directly, but they lack explicit modeling of the coupling relationship between electromagnetic excitation and rotational speed, resulting in unstable performance when operating conditions change significantly, especially when deployed across multiple units.

[0006] Therefore, a method combining physical mechanisms and data-driven approaches is needed to improve the anomaly detection performance of wind turbine drivetrains in real industrial environments without relying on external speed logs. Summary of the Invention

[0007] The purpose of this invention is to overcome the shortcomings of existing technologies and propose a wind turbine drivetrain anomaly detection method and system based on rotational speed guidance. This addresses the problem of insufficient generalization ability of existing wind turbine anomaly detection methods under variable speed conditions. By directly extracting multidimensional features from vibration acceleration signals collected by the wind turbine condition monitoring system (CMS), and introducing a Conditional Neural Network (CondNN) for modeling, explicitly incorporating the permanent magnet synchronous generator (PMSG) rotational speed as a modulation condition during the modeling process, high-precision drivetrain anomaly detection is achieved across operating conditions and turbine units. Simultaneously, rotational speed is used as the core indicator characterizing the equipment's operating condition to more accurately reflect the characteristics of wind turbines using PMSGs in real-world complex environments.

[0008] The objective of this invention is achieved through the following technical solution: a method for detecting abnormalities in the transmission chain of a wind turbine based on rotational speed guidance, wherein the wind turbine is a wind turbine using a permanent magnet synchronous generator, and the method includes the following steps: S1. Obtain the vibration acceleration signal of the wind turbine drive train; S2. Based on the acquired vibration acceleration signal, calculate the RPM of the permanent magnet synchronous generator of the wind turbine and extract multi-dimensional features from the vibration acceleration signal; S3. Using the permanent magnet synchronous generator speed RPM as the modulation condition, a transmission chain anomaly detection model is constructed based on the conditional neural network CondNN. S4. Input the multi-dimensional features into the transmission chain anomaly detection model to detect the current state of the wind turbine transmission chain, output the detection results, and complete the anomaly detection of the wind turbine transmission chain.

[0009] Furthermore, step S1 includes: Vibration acceleration signals are obtained from the wind turbine drivetrain components through the wind turbine condition monitoring system (CMS).

[0010] Furthermore, step S2 includes: The vibration acceleration signal is analyzed by Fast Fourier Transform and preprocessed by taking the absolute value to obtain the amplitude spectrum. Frequency intervals are extracted within a preset frequency band of the amplitude spectrum, and the RPM of the permanent magnet synchronous generator is calculated using the following formula: f=p*n_s / 60 Where f is the stator current frequency of the permanent magnet synchronous generator in the wind turbine in one magnetic field cycle, p is the number of magnetic pairs of the permanent magnet synchronous generator, and n_s is the rotational speed of the permanent magnet synchronous motor. f_CMS=2*f=P*n_s / 60 Where f_CMS is the frequency interval captured within the preset frequency band of the amplitude spectrum, and P is the number of magnets in the permanent magnet synchronous generator.

[0011] Furthermore, step S2 includes: First, the acquired vibration acceleration signal is preprocessed by removing error signals, removing the mean, and multiplying by a window function. Then, the preprocessed vibration acceleration signal is subjected to multidimensional feature extraction, which includes time domain features, frequency domain features, statistical domain features, and dimensionless features.

[0012] Furthermore, step S3 includes: A transmission chain anomaly detection model is constructed by using a conditional neural network (CondNN) as a discriminator and the permanent magnet synchronous generator speed (RPM) as the modulation condition of the conditional neural network (CondNN). The conditional neural network (CondNN) includes a gated neural network (Gate MLP), a characteristic linear modulation neural network (FiLM MLP), and a conditional batch normalization neural network (CBN MLP).

[0013] Furthermore, step S4 includes: Multidimensional features are input into the transmission chain anomaly detection model to detect the current state of the wind turbine transmission chain. The multidimensional features include time domain features, frequency domain features, statistical domain features, and dimensionless features. If the detection result is normal, the detection result is output to a preset database. If the detection result is abnormal, the detection result is output to a preset database and an anomaly warning is issued to complete the anomaly warning for the wind turbine transmission chain.

[0014] Furthermore, the time-domain features include mean, variance, skewness, and kurtosis; the frequency-domain features include spectral mean, spectral root mean square value, and spectral skewness; the statistical-domain features include the third moment and the fourth moment of the time-domain signal; and the dimensionless features include peak factor, waveform factor, impulse factor, and margin factor.

[0015] A speed-guided wind turbine drivetrain anomaly detection system is provided to implement the aforementioned speed-guided wind turbine drivetrain anomaly detection method. The wind turbine is a wind turbine using a permanent magnet synchronous generator. The system includes: The data acquisition module obtains vibration acceleration signals from the wind turbine drivetrain components based on the wind turbine condition monitoring system (CMS). The rotational speed calculation module calculates the rotational speed (RPM) of the permanent magnet synchronous generator of the wind turbine based on the vibration acceleration signal obtained by the data acquisition module. The signal preprocessing and feature extraction module performs preprocessing on the vibration acceleration signal acquired by the data acquisition module, including removing error signals, removing the mean, and multiplying by a window function. Then, it performs multidimensional feature extraction on the preprocessed vibration acceleration signal. The multidimensional features include time domain features, frequency domain features, statistical domain features, and dimensionless features. The conditional modeling module uses the conditional neural network CondNN as the discriminator and the permanent magnet synchronous generator speed RPM as the modulation condition of the conditional neural network CondNN to construct a transmission chain anomaly detection model. The alarm module inputs the multi-dimensional features extracted by the signal preprocessing and feature extraction module into the transmission chain anomaly detection model of the condition modeling module to detect the current state of the wind turbine transmission chain, output the detection results, and complete the anomaly detection of the wind turbine transmission chain.

[0016] Furthermore, the rotational speed calculation module performs the following operations: The vibration acceleration signal is analyzed by Fast Fourier Transform and preprocessed by taking the absolute value to obtain the amplitude spectrum. Frequency intervals are extracted within a preset frequency band of the amplitude spectrum, and the RPM of the permanent magnet synchronous generator is calculated using the following formula: f=p*n_s / 60 Where f is the stator current frequency of the permanent magnet synchronous generator in the wind turbine in one magnetic field cycle, p is the number of magnetic pairs of the permanent magnet synchronous generator, and n_s is the rotational speed of the permanent magnet synchronous motor. f_CMS=2*f=P*n_s / 60 Where f_CMS is the frequency interval captured within the preset frequency band of the amplitude spectrum, and P is the number of magnets in the permanent magnet synchronous generator.

[0017] Furthermore, it includes a database module for storing the detection results output by the alarm module; The database module is communicatively connected to the alarm module, and the alarm module performs the following operations: The time-domain features, frequency-domain features, statistical domain features, and dimensionless features extracted by the signal preprocessing and feature extraction module are input into the transmission chain anomaly detection model to detect the current state of the wind turbine transmission chain. If the detection result is normal, the detection result is output to the database module; if the detection result is abnormal, the detection result is output to the database module, and an anomaly warning is issued, thus completing the anomaly warning for the wind turbine transmission chain. Among them, the time-domain features include mean, variance, skewness, and kurtosis; the frequency-domain features include spectral mean, spectral root mean square value, and spectral skewness; the statistical domain features include the third moment and the fourth moment of the time-domain signal; and the dimensionless features include peak factor, waveform factor, impulse factor, and margin factor.

[0018] Compared with the prior art, the present invention has the following advantages and beneficial effects: 1. Speed ​​Adaptive: This invention uses the speed of the permanent magnet synchronous generator for conditional modeling, which can handle the changes in spectrum mode under different operating conditions.

[0019] 2. Higher generalization ability: This invention constructs a transmission chain anomaly detection model through a conditional neural network (CondNN). During the modeling process, the speed of the permanent magnet synchronous generator is explicitly introduced as a modulation condition, which improves the generalization ability of transmission chain anomaly detection while maintaining high-precision detection. It is applicable to various types of wind turbine units.

[0020] 3. Low dependence: This invention does not require external input speed monitoring signals or synchronization with external speed logs. The speed of the permanent magnet synchronous generator can be directly estimated from the spectrum of the vibration signal, which has higher real-time performance and accuracy.

[0021] 4. Easy to deploy: No need to modify the wind turbine condition monitoring system (CMS) hardware. The existing sensors in the wind turbine condition monitoring system (CMS) can be used directly to detect the drive train, saving costs and making it easy to implement. Attached Figure Description

[0022] Figure 1 A waterfall chart of "speed-frequency-amplitude" plotted using measured data for a semi-direct drive wind turbine.

[0023] Figure 2 This is a schematic diagram of the transmission chain structure of a semi-direct drive wind turbine generator set.

[0024] Figure 3 This is a flowchart and architecture diagram of the present invention. Detailed Implementation

[0025] The present invention will be further described below with reference to specific embodiments.

[0026] Example 1 Taking a semi-direct drive wind turbine in a wind farm as an example, abnormal transmission chain detection is performed. (See...) Figure 2 As shown, the selected semi-direct drive wind turbine transmission chain components in this embodiment include the main bearing 1, the gearbox 2, and the permanent magnet synchronous generator 3.

[0027] See Figure 3 As shown, the wind turbine drivetrain anomaly detection method based on rotational speed guidance provided in this embodiment includes the following steps: S1. Vibration acceleration signals are obtained from the drive chain components of the semi-direct drive wind turbine through the wind turbine condition monitoring system (CMS).

[0028] S2. Based on the acquired vibration acceleration signal, calculate the RPM of the permanent magnet synchronous generator of the semi-direct drive wind turbine, and extract multi-dimensional features from the vibration acceleration signal, including: The vibration acceleration signal is analyzed by Fast Fourier Transform (FTT) and preprocessed by taking the absolute value to obtain the amplitude spectrum. Frequency intervals are extracted within a preset frequency band of the amplitude spectrum, and the RPM of the permanent magnet synchronous generator is calculated using the following formula: f=p*n_s / 60 Where f is the stator current frequency of the permanent magnet synchronous generator 3 in one magnetic field cycle in the semi-direct drive wind turbine, p is the number of magnetic pairs of the permanent magnet synchronous generator 3, and n_s is the rotational speed of the permanent magnet synchronous motor 3. f_CMS=2*f=P*n_s / 60 Where f_CMS is the frequency interval captured within the preset frequency band of the amplitude spectrum, and P is the number of magnets in the permanent magnet synchronous generator 3.

[0029] First, the acquired vibration acceleration signal undergoes preprocessing, including error signal removal, mean removal, and multiplication by a window function. Next, the preprocessed vibration acceleration signal is subjected to multidimensional feature extraction, which includes time-domain features, frequency-domain features, statistical domain features, and dimensionless features. The time-domain features include mean, variance, skewness, and kurtosis. The frequency-domain features include spectral mean, spectral root mean square value, and spectral skewness. The statistical domain features include the third and fourth moments of the time-domain signal. The dimensionless features include peak factor, waveform factor, impulse factor, and margin factor.

[0030] S3. Using a conditional neural network (CondNN) as a discriminator, and using the permanent magnet synchronous generator speed (RPM) as the modulation condition of the conditional neural network (CondNN), a transmission chain anomaly detection model is constructed; the conditional neural network (CondNN) includes a gated neural network (Gate MLP), a characteristic linear modulation neural network (FiLM MLP), and a conditional batch normalization neural network (CBNMLP).

[0031] S4. Input the multi-dimensional features into the transmission chain anomaly detection model to detect the current state of the semi-direct drive wind turbine transmission chain. If the detection result is normal, output the detection result to the preset database; if the detection result is abnormal, output the detection result to the preset database and issue an anomaly warning to complete the anomaly warning of the semi-direct drive wind turbine transmission chain.

[0032] Example 2 The speed-guided wind turbine drivetrain anomaly detection system provided in this embodiment is used to implement the speed-guided wind turbine drivetrain anomaly detection method described in Embodiment 1. Similarly, a semi-direct-drive wind turbine in a wind farm is used as an example for drivetrain anomaly detection. (See [link to documentation]). Figure 2 As shown, the selected semi-direct drive wind turbine transmission chain components in this embodiment include the main bearing 1, the gearbox 2, and the permanent magnet synchronous generator 3.

[0033] See Figure 3 As shown, the wind turbine drivetrain anomaly detection system provided in this embodiment includes: 1) Data acquisition module: Based on the wind turbine condition monitoring system (CMS), it obtains vibration acceleration signals from the drive chain components of the semi-direct drive wind turbine.

[0034] 2) The speed calculation module calculates the RPM of the permanent magnet synchronous generator of the semi-direct drive wind turbine based on the vibration acceleration signal obtained by the data acquisition module, and performs the following operations: The vibration acceleration signal is analyzed by Fast Fourier Transform and preprocessed by taking the absolute value to obtain the amplitude spectrum. Frequency intervals are extracted within a preset frequency band of the amplitude spectrum, and the RPM of the permanent magnet synchronous generator is calculated using the following formula: f=p*n_s / 60 Where f is the stator current frequency of the permanent magnet synchronous generator 3 in one magnetic field cycle in the semi-direct drive wind turbine, p is the number of magnetic pairs of the permanent magnet synchronous generator 3, and n_s is the rotational speed of the permanent magnet synchronous motor 3. f_CMS=2*f=P*n_s / 60 Where f_CMS is the frequency interval captured within the preset frequency band of the amplitude spectrum, and P is the number of magnets in the permanent magnet synchronous generator 3.

[0035] 3) The signal preprocessing and feature extraction module sequentially preprocesses the vibration acceleration signal acquired by the data acquisition module by removing error signals, removing the mean, and multiplying by a window function. Then, it performs multidimensional feature extraction on the preprocessed vibration acceleration signal. The multidimensional features include time-domain features, frequency-domain features, statistical domain features, and dimensionless features. The time-domain features include mean, variance, skewness, and kurtosis. The frequency-domain features include spectral mean, spectral root mean square value, and spectral skewness. The statistical domain features include the third moment and the fourth moment of the time-domain signal. The dimensionless features include peak factor, waveform factor, impulse factor, and margin factor.

[0036] 4) Conditional modeling module: using a conditional neural network (CondNN) as a discriminator, and using the permanent magnet synchronous generator speed (RPM) as the modulation condition of the conditional neural network (CondNN) to construct a transmission chain anomaly detection model; the conditional neural network (CondNN) includes a gated neural network (Gate MLP), a characteristic linear modulation neural network (FiLM MLP), and a conditional batch normalization neural network (CBN MLP).

[0037] 5) Database module (not shown in the figure), used to store the detection results output by the alarm module. The database module is communicatively connected to the alarm module.

[0038] 6) The alarm module inputs the time-domain features, frequency-domain features, statistical domain features, and dimensionless features extracted by the signal preprocessing and feature extraction module into the transmission chain anomaly detection model to detect the current state of the semi-direct drive wind turbine transmission chain. If the detection result is normal, the detection result is output to the database module; if the detection result is abnormal, the detection result is output to the database module and an anomaly warning is issued, thus completing the anomaly warning for the semi-direct drive wind turbine transmission chain.

[0039] Example 3 Unlike Embodiment 1, the wind turbine transmission chain anomaly detection method based on rotational speed guidance provided in this embodiment is applied to direct-drive wind turbines, and the transmission chain anomaly detection of the direct-drive wind turbine is performed based on the permanent magnet synchronous generator in the direct-drive wind turbine.

[0040] Example 4 This embodiment discloses a non-transitory computer-readable medium storing instructions that, when executed by a processor, perform the steps of the wind turbine drivetrain anomaly detection method based on rotational speed guidance as described in Embodiment 1.

[0041] In this embodiment, the non-transitory computer-readable medium can be a disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), USB flash drive, portable hard drive, etc.

[0042] Example 5 This embodiment discloses a computing device, including a processor and a memory for storing processor-executable programs. When the processor executes the program stored in the memory, it implements the wind turbine transmission chain anomaly detection method based on rotational speed guidance described in Embodiment 1.

[0043] The computing device described in this embodiment may be a desktop computer, laptop computer, smartphone, PDA handheld terminal, tablet computer, programmable logic controller (PLC), or other terminal device with processor function.

[0044] The above-described embodiments are merely preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Therefore, any changes made in accordance with the shape and principle of the present invention should be covered within the protection scope of the present invention.

Claims

1. A method for detecting abnormalities in the transmission chain of a wind turbine based on rotational speed guidance, characterized in that, The wind turbine is a wind turbine using a permanent magnet synchronous generator, and the method includes the following steps: S1. Obtain the vibration acceleration signal of the wind turbine drive train; S2. Based on the acquired vibration acceleration signal, calculate the RPM of the permanent magnet synchronous generator of the wind turbine and extract multi-dimensional features from the vibration acceleration signal; S3. Using the permanent magnet synchronous generator speed RPM as the modulation condition, a transmission chain anomaly detection model is constructed based on the conditional neural network CondNN. S4. Input the multi-dimensional features into the transmission chain anomaly detection model to detect the current state of the wind turbine transmission chain, output the detection results, and complete the anomaly detection of the wind turbine transmission chain.

2. The method for detecting abnormalities in the transmission chain of a wind turbine based on rotational speed guidance according to claim 1, characterized in that, Step S1 includes: Vibration acceleration signals are obtained from the wind turbine drivetrain components through the wind turbine condition monitoring system (CMS).

3. The method for detecting abnormalities in the transmission chain of a wind turbine based on rotational speed guidance according to claim 1, characterized in that, Step S2 includes: The vibration acceleration signal is analyzed by Fast Fourier Transform and preprocessed by taking the absolute value to obtain the amplitude spectrum. Frequency intervals are extracted within a preset frequency band of the amplitude spectrum, and the RPM of the permanent magnet synchronous generator is calculated using the following formula: f=p*n_s / 60 Where f is the stator current frequency of the permanent magnet synchronous generator in the wind turbine in one magnetic field cycle, p is the number of magnetic pairs of the permanent magnet synchronous generator, and n_s is the rotational speed of the permanent magnet synchronous motor. f_CMS=2*f=P*n_s / 60 Where f_CMS is the frequency interval captured within the preset frequency band of the amplitude spectrum, and P is the number of magnets in the permanent magnet synchronous generator.

4. The method for detecting abnormalities in the transmission chain of a wind turbine based on rotational speed guidance according to claim 1, characterized in that, Step S2 includes: First, the acquired vibration acceleration signal is preprocessed by removing error signals, removing the mean, and multiplying by a window function. Then, the preprocessed vibration acceleration signal is subjected to multidimensional feature extraction, which includes time domain features, frequency domain features, statistical domain features, and dimensionless features.

5. The method for detecting abnormalities in the transmission chain of a wind turbine based on rotational speed guidance according to claim 3, characterized in that, Step S3 includes: A transmission chain anomaly detection model is constructed by using a conditional neural network (CondNN) as a discriminator and the permanent magnet synchronous generator speed (RPM) as the modulation condition of the conditional neural network (CondNN). The conditional neural network (CondNN) includes a gated neural network (Gate MLP), a characteristic linear modulation neural network (FiLM MLP), and a conditional batch normalization neural network (CBN MLP).

6. The method for detecting abnormalities in the transmission chain of a wind turbine based on rotational speed guidance according to claim 5, characterized in that, Step S4 includes: Multidimensional features are input into the transmission chain anomaly detection model to detect the current state of the wind turbine transmission chain. The multidimensional features include time domain features, frequency domain features, statistical domain features, and dimensionless features. If the detection result is normal, the detection result is output to a preset database. If the detection result is abnormal, the detection result is output to a preset database and an anomaly warning is issued to complete the anomaly warning for the wind turbine transmission chain.

7. A method for detecting abnormalities in the transmission chain of a wind turbine based on rotational speed guidance, as described in claim 4 or 6, characterized in that: The time-domain features include mean, variance, skewness, and kurtosis; the frequency-domain features include spectral mean, spectral root mean square value, and spectral skewness; the statistical-domain features include the third moment and the fourth moment of the time-domain signal; and the dimensionless features include peak factor, waveform factor, impulse factor, and margin factor.

8. A wind turbine transmission chain anomaly detection system based on rotational speed guidance, characterized in that, For implementing the wind turbine drivetrain anomaly detection method based on rotational speed guidance as described in claim 1, wherein the wind turbine is a wind turbine using a permanent magnet synchronous generator, the system comprises: The data acquisition module obtains vibration acceleration signals from the wind turbine drivetrain components based on the wind turbine condition monitoring system (CMS). The rotational speed calculation module calculates the rotational speed (RPM) of the permanent magnet synchronous generator of the wind turbine based on the vibration acceleration signal obtained by the data acquisition module. The signal preprocessing and feature extraction module performs preprocessing on the vibration acceleration signal acquired by the data acquisition module, including removing error signals, removing the mean, and multiplying by a window function. Then, it performs multidimensional feature extraction on the preprocessed vibration acceleration signal. The multidimensional features include time domain features, frequency domain features, statistical domain features, and dimensionless features. The conditional modeling module uses the conditional neural network CondNN as the discriminator and the permanent magnet synchronous generator speed RPM as the modulation condition of the conditional neural network CondNN to construct a transmission chain anomaly detection model. The alarm module inputs the multi-dimensional features extracted by the signal preprocessing and feature extraction module into the transmission chain anomaly detection model of the condition modeling module to detect the current state of the wind turbine transmission chain, output the detection results, and complete the anomaly detection of the wind turbine transmission chain.

9. A wind turbine transmission chain anomaly detection system based on rotational speed guidance according to claim 8, characterized in that, The rotational speed calculation module performs the following operations: The vibration acceleration signal is analyzed by Fast Fourier Transform and preprocessed by taking the absolute value to obtain the amplitude spectrum. Frequency intervals are extracted within a preset frequency band of the amplitude spectrum, and the RPM of the permanent magnet synchronous generator is calculated using the following formula: f=p*n_s / 60 Where f is the stator current frequency of the permanent magnet synchronous generator in the wind turbine in one magnetic field cycle, p is the number of magnetic pairs of the permanent magnet synchronous generator, and n_s is the rotational speed of the permanent magnet synchronous motor. f_CMS=2*f=P*n_s / 60 Where f_CMS is the frequency interval captured within the preset frequency band of the amplitude spectrum, and P is the number of magnets in the permanent magnet synchronous generator.

10. A wind turbine transmission chain anomaly detection system based on rotational speed guidance according to claim 8, characterized in that: It includes a database module for storing the detection results output by the alarm module; The database module is communicatively connected to the alarm module, and the alarm module performs the following operations: The time-domain features, frequency-domain features, statistical domain features, and dimensionless features extracted by the signal preprocessing and feature extraction module are input into the transmission chain anomaly detection model to detect the current state of the wind turbine transmission chain. If the detection result is normal, the detection result is output to the database module; if the detection result is abnormal, the detection result is output to the database module, and an anomaly warning is issued, thus completing the anomaly warning for the wind turbine transmission chain. Among them, the time-domain features include mean, variance, skewness, and kurtosis; the frequency-domain features include spectral mean, spectral root mean square value, and spectral skewness; the statistical domain features include the third moment and the fourth moment of the time-domain signal; and the dimensionless features include peak factor, waveform factor, impulse factor, and margin factor.

Images