A method, system, and related device for extracting the speed reliability of wind turbine units based on multi-point vibration signals.

Abstract

This invention provides a method, system, and related apparatus for extracting the reliability of wind turbine rotational speed based on multi-point vibration signals, comprising the following steps: Step 1, preprocessing the vibration signals of key parts of the wind turbine under test to obtain preprocessed vibration signals; Step 2, performing frequency domain transformation on the preprocessed vibration signals to obtain frequency sequences and corresponding amplitude sequences; Step 3, obtaining the frequency intervals corresponding to each key part based on the frequency sequences; Step 4, selecting candidate fundamental frequencies based on the frequency intervals; Step 5, calculating the significance and confidence based on the candidate fundamental frequencies; Step 6, calculating candidate rotational speeds based on the significance and confidence; Step 7, clustering the candidate rotational speeds corresponding to multiple key parts to obtain the final rotational speed. This invention not only enhances adaptability to complex operating conditions and low signal-to-noise ratio environments but also automatically identifies the non-started state, avoiding false rotational speed outputs.

Description

Technical Field

[0001] This invention belongs to the field of wind turbine generator condition monitoring technology, and particularly relates to a method, system and related device for extracting the speed reliability of wind turbine generators based on multi-point vibration signals. Background Technology

[0002] During the operation of a wind turbine generator, the rotational speed of its drivetrain components is one of the fundamental parameters for fault diagnosis. In vibration signal analysis, various mechanical faults (such as bearing failure, gear meshing damage, imbalance, and misalignment) exhibit characteristic frequencies in the frequency spectrum that are proportional to the rotational speed. Therefore, accurately obtaining the real-time rotational speed of the drivetrain is crucial for calculating these fault characteristic frequencies, establishing the correspondence between spectral components and mechanical components, and achieving fault location. If rotational speed information is missing or significantly inaccurate, the characteristic frequency calculation will be inaccurate, thus affecting the reliability of fault identification and the correctness of the diagnostic results.

[0003] In existing technologies, wind turbine condition monitoring systems (CMS) typically measure speed signals using speed sensors (such as proximity switches or encoders) installed on the high-speed shaft or generator end to assist in the demodulation and feature extraction of vibration signals. However, in actual operating environments, these sensors are often affected by factors such as oil contamination, dust, high temperatures, electromagnetic interference, and limited installation space, leading to problems such as signal loss, waveform distortion, or measurement errors. This makes it difficult for the system to continuously obtain accurate speed data, thereby affecting the accuracy of vibration analysis and fault identification. Furthermore, the interface standards of speed sensors are not standardized across different turbine models and manufacturers, making on-site installation and maintenance difficult. Long sensor replacement cycles and low reliability increase system maintenance costs.

[0004] More significantly, during the analysis of existing CMS data from various wind farms, it was discovered that many third-party monitoring systems lacked effective rotational speed data, leading to the malfunction of automated centralized diagnostic algorithms or a significant decrease in diagnostic effectiveness. The main reason for this is the lack of standardized design for CMS systems from different manufacturers. Some systems are only designed for manual analysis scenarios and do not simultaneously collect or upload rotational speed signals. While manual diagnosis can rely on experience to identify spectral features, rotational speed is not essential; however, automated diagnostic algorithms depend on accurate rotational speed parameters to calculate fault characteristic frequencies and achieve spectral normalization and pattern recognition. The lack of accurate rotational speed signals leads to difficulties in characteristic frequency identification, incorrect fault location, and algorithm model failure, severely restricting the widespread application of intelligent diagnostic systems at the group level and the full utilization of monitoring data value.

[0005] On the other hand, the vibration signals of wind turbines naturally contain characteristic information such as rotational frequency and its harmonics, and theoretically, the rotational speed can be deduced from the vibration data using signal processing methods. However, existing algorithms for extracting rotational speed based on vibration signals still have many shortcomings: (1) The limitations of single measurement point are prominent. Most methods only use the signal from a single measurement point for spectral peak identification, which is easily affected by gear meshing, structural resonance and environmental noise, making it difficult to obtain stable results; when the measurement point is installed in a different position or there is local damage, the algorithm's misjudgment rate increases significantly.

[0006] (2) It relies on single-peak characteristics and has weak noise resistance. Traditional algorithms often take only the single peak with the largest amplitude in the spectrum as the switching frequency, or extract the instantaneous frequency through short-time Fourier transform, Hilbert transform, energy operator, autocorrelation and other methods. It is extremely sensitive to noise and harmonics, and its accuracy drops sharply when the rotation speed fluctuates or the signal-to-noise ratio is low.

[0007] (3) Insufficient utilization of harmonic information. The harmonics in vibration signals are usually close to or even more significant than the fundamental frequency amplitude. However, existing algorithms do not integrate the energy distribution relationship between harmonic amplitudes, which can easily lead to the misidentification of harmonics, gear meshing frequency or damage frequency as the fundamental frequency, resulting in distorted extraction results.

[0008] (4) FFT resolution effect is not compensated. Since the spectrum is discrete, the peak of the harmonics often falls between the sampling points. The existing method does not introduce a compensation mechanism, which causes the harmonic energy to be underestimated or omitted.

[0009] (5) Lack of multi-point consistency verification. Most existing algorithms are single-channel calculations and have not established a fusion and consistency verification mechanism for multi-point results. They cannot effectively eliminate abnormal points and have insufficient overall robustness.

[0010] (6) Unable to identify the off-state. Most methods assume that the equipment is in operation and do not introduce a significance or confidence judgment mechanism, which can easily lead to misjudgment under shutdown or low-speed conditions.

[0011] (7) The algorithms are complex and have poor real-time performance. Some methods employ complex processes such as matching pursuit, time-frequency ridge tracking, envelope cepstral analysis, and deep learning, which involve large computational loads and poor real-time performance, making them difficult to deploy in the edge devices of wind turbine units.

[0012] Therefore, there is an urgent need for a speed extraction method with a simple algorithm structure, strong noise resistance, multi-point fusion capability, and start / stop determination function. This method should be able to accurately deduce the speed from the vibration signal even when the speed sensor fails or the measurement is inaccurate, thus providing a reliable foundation for automated diagnosis. Summary of the Invention

[0013] The purpose of this invention is to provide a method, system, and related device for extracting the speed reliability of wind turbines based on multi-point vibration signals, which solves the problem of inaccurate condition monitoring and fault diagnosis caused by missing speed signals or sensor failure during wind turbine operation.

[0014] To achieve the above objectives, the technical solution adopted by the present invention is as follows: In a first aspect, the present invention provides a method for extracting the speed reliability of wind turbine generators based on multi-point vibration signals, comprising the following steps: Step 1: Preprocess the vibration signals of the key parts of the wind turbine under test to obtain the preprocessed vibration signals. Step 2: Perform frequency domain transformation on the preprocessed vibration signal to obtain the frequency sequence and the corresponding amplitude sequence; Step 3: Based on the frequency sequence, obtain the frequency range corresponding to each key part; Step 4: Select candidate fundamental frequencies based on frequency ranges; Step 5: Calculate the significance and confidence based on the candidate fundamental frequencies; Step 6: Calculate candidate rotational speeds based on significance and confidence. Step 7: Cluster the candidate rotation speeds corresponding to multiple key components to obtain the final rotation speed.

[0015] Preferably, the key components of the wind turbine under test include the high-speed shaft of the gearbox, the generator drive end bearing, and the generator non-drive end bearing.

[0016] Preferably, candidate fundamental frequencies are selected based on frequency ranges, specifically by: Select local extrema within the frequency range; The frequency points that satisfy the preset amplitude threshold among multiple local extrema are selected as candidate fundamental frequencies.

[0017] Preferably, the significance is calculated based on the candidate fundamental frequency, specifically by: Based on the amplitude of each candidate fundamental frequency and the amplitudes of multiple harmonics of the candidate fundamental frequency, a significance is calculated according to a preset weight. Each harmonic is placed within a corresponding frequency range, and each harmonic is the maximum value of the sampling point and multiple frequency points on both sides or the average value of the sampling point and multiple frequency points on both sides.

[0018] Preferably, the confidence level is calculated based on the candidate fundamental frequency, specifically by: The confidence level is calculated based on each candidate fundamental frequency and its corresponding harmonics.

[0019] Preferably, candidate rotational speeds are calculated based on significance and confidence levels, specifically by: Candidate rotational speeds are calculated based on the candidate fundamental frequency corresponding to the maximum significance value. The obtained confidence level is compared with a preset threshold, and the candidate rotation speed is determined based on the comparison result.

[0020] Preferably, the candidate rotational speeds corresponding to multiple key components are clustered to obtain the final rotational speed. The specific method is as follows: By utilizing the rotational speed tolerance, candidate rotational speeds corresponding to multiple key components are clustered to obtain multiple sets of data; The final rotational speed is obtained by weighting the rotational speed estimates based on the significance or confidence level of the data set with the largest number of samples.

[0021] Secondly, the present invention provides a wind turbine speed reliability extraction system based on multi-point vibration signals, comprising: The vibration signal preprocessing unit is used to preprocess the vibration signals of the key parts of the wind turbine under test to obtain the preprocessed vibration signal. The frequency domain conversion unit is used to convert the preprocessed vibration signal into a frequency domain to obtain a frequency sequence and a corresponding amplitude sequence. The frequency range acquisition unit is used to acquire the frequency range corresponding to each key part based on the frequency sequence. The candidate fundamental frequency selection unit is used to select candidate fundamental frequencies based on a frequency range; The confidence calculation unit is used to calculate the significance and confidence based on the candidate fundamental frequency; The candidate rotational speed calculation unit is used to calculate candidate rotational speeds based on significance and confidence. The rotational speed clustering unit is used to cluster candidate rotational speeds corresponding to multiple key components to obtain the final rotational speed.

[0022] Thirdly, the present invention provides an electronic device including a processor and a memory, wherein the memory stores computer instructions, and when the computer instructions are executed by the processor, the electronic device performs the method described thereon.

[0023] Fourthly, the present invention provides a computer program product, the computer program product including computer-executable instructions, which, when executed, implement the method described.

[0024] Compared with the prior art, the beneficial effects of the present invention are: This invention provides a wind turbine speed reliability extraction method based on multi-point vibration signals. By introducing a series of structured steps, including multi-point collaborative analysis, frequency domain range limitation, candidate fundamental frequency screening, dual-index evaluation of significance and confidence, and multi-point clustering fusion, the method significantly improves the reliability, anti-interference capability, and engineering applicability of speed extraction. This method effectively overcomes the problems of traditional single-point speed extraction methods, such as susceptibility to noise and harmonic interference, insufficient utilization of harmonic information, lack of start-stop status judgment, and inconsistency verification of multi-point results. By synchronously collecting vibration signals from multiple key components such as the high-speed shaft of the gearbox, the generator drive end, and the non-drive end, and extracting candidate fundamental frequencies within a predefined frequency range, the method calculates significance by combining harmonic energy accumulation, judges the operating status through confidence, and finally integrates the multi-point results using a clustering fusion mechanism, achieving high-precision speed inversion even in the absence of speed sensors or sensor failure. This method not only enhances adaptability to complex operating conditions and low signal-to-noise ratio environments, but also automatically identifies the non-started state, avoiding false speed output. At the same time, the algorithm has a clear structure and moderate computational load, making it suitable for real-time deployment in the edge devices of wind turbine units. It provides a reliable and adaptive speed data foundation for wind turbine unit condition monitoring and intelligent diagnosis, and has important engineering application value. Attached Figure Description

[0025] Figure 1 This is a schematic diagram of the system structure according to an embodiment of the present invention; Figure 2 This is a schematic diagram of the vibration signal preprocessing process according to an embodiment of the present invention; Figure 3 This is a schematic diagram of the candidate frequency band determination process according to an embodiment of the present invention; Figure 4 This is a schematic diagram of the candidate fundamental frequency extraction process according to an embodiment of the present invention; Figure 5 This is a schematic diagram of the confidence calculation process according to an embodiment of the present invention; Figure 6 This is a schematic diagram of the candidate rotation speed processing flow according to an embodiment of the present invention. Detailed Implementation

[0026] In the following description, specific details such as particular system architectures and techniques are set forth for illustrative purposes and not for limitation, in order to provide a thorough understanding of the embodiments of this application. However, those skilled in the art will understand that this application may also be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, apparatuses, circuits, and methods have been omitted so as not to obscure the description of this application with unnecessary detail.

[0027] It should be understood that, when used in this application specification and the appended claims, the term "comprising" indicates the presence of the described features, integrals, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or a collection thereof.

[0028] It should also be understood that the term “and / or” as used in this application specification and the appended claims means any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.

[0029] As used in this application specification and the appended claims, the term "if" may be interpreted, depending on the context, as "when," "once," "in response to determination," or "in response to detection." Similarly, the phrase "if determined" or "if detected [the described condition or event]" may be interpreted, depending on the context, as meaning "once determined," "in response to determination," "once detected [the described condition or event]," or "in response to detection [the described condition or event]."

[0030] Furthermore, in the description of this application and the appended claims, the terms "first," "second," "third," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0031] References to "one embodiment" or "some embodiments" as described in this specification mean that one or more embodiments of this application include a specific feature, structure, or characteristic described in connection with that embodiment. Therefore, the phrases "in one embodiment," "in some embodiments," "in other embodiments," "in still other embodiments," etc., appearing in different parts of this specification do not necessarily refer to the same embodiment, but rather mean "one or more, but not all, embodiments," unless otherwise specifically emphasized. The terms "comprising," "including," "having," and variations thereof mean "including but not limited to," unless otherwise specifically emphasized.

[0032] Example 1 like Figure 1 As shown in the figure, this embodiment provides a wind turbine speed reliability extraction system based on multi-point vibration signals, including a vibration sensor group, a data acquisition unit, and a signal processing module, wherein: The vibration sensor group includes vibration sensors installed at key parts of the wind turbine to collect vibration signals. In this embodiment, three locations are selected: the radial direction of the high-speed shaft of the gearbox, the radial direction of the generator drive-end bearing, and the radial direction of the generator non-drive-end bearing, which are typically standard on wind turbines; the measuring point on the high-speed shaft of the gearbox contains both rotational frequency and meshing frequency information.

[0033] The data acquisition unit is electrically connected to the vibration sensor group to digitize the analog vibration signal; it has independent power supply and protection functions.

[0034] The signal processing module is used to perform preprocessing, spectrum analysis, frequency band filtering, significance and confidence calculation, and multi-point fusion of the collected vibration data to extract and save the rotational speed.

[0035] It can be deployed on the data acquisition unit on the generator side or the host computer server on the station side.

[0036] Example 2 Based on Example 1, this example provides a wind turbine speed reliability extraction system based on multi-point vibration signals, wherein the signal processing module includes: The vibration signal preprocessing unit is used to preprocess the vibration signals of the key parts of the wind turbine under test to obtain the preprocessed vibration signal. The frequency domain conversion unit is used to convert the preprocessed vibration signal into a frequency domain to obtain a frequency sequence and a corresponding amplitude sequence. The frequency range acquisition unit is used to acquire the frequency range corresponding to each key part based on the frequency sequence. The candidate fundamental frequency selection unit is used to select candidate fundamental frequencies based on a frequency range; The confidence calculation unit is used to calculate the significance and confidence based on the candidate fundamental frequency; The candidate rotational speed calculation unit is used to calculate candidate rotational speeds based on significance and confidence. The rotational speed clustering unit is used to cluster candidate rotational speeds corresponding to multiple key components to obtain the final rotational speed. Example 3 like Figures 2 to 6 As shown in the figure, this embodiment provides a method for extracting the speed reliability of wind turbine units based on multi-point vibration signals, which includes the following steps: Step 1: Simultaneously collect vibration signals for a certain duration at three measurement points: the radial direction of the high-speed shaft of the gearbox, the radial direction of the generator drive end bearing, and the radial direction of the generator non-drive end bearing. The sampling duration and sampling rate are set according to the site requirements and are used for speed extraction.

[0037] Step 2, DC component removal and window function processing: Subtract the DC average value from each acquired waveform sequence to eliminate sensor bias; then multiply by a Hanning window or other window function to reduce spectral leakage caused by truncation and ensure the accuracy of subsequent frequency domain analysis.

[0038] Step 3: Frequency domain transformation and determination of fundamental frequency candidate bands The preprocessed time-domain signal is converted to the frequency domain using a fast Fourier transform to obtain a frequency sequence and a corresponding amplitude sequence, using only the positive frequency portion.

[0039] Based on the frequency sequence, the frequency range corresponding to each key part is obtained.

[0040] In this embodiment, for the high-speed shaft, the stable operating speed range N of the high-speed shaft of the fan is considered. min N max (For example, 1100~1800 RPM) Calculate the corresponding rotational frequency range [ min , max ].

[0041] Based on the number of teeth of the gear connected to the high-speed shaft of the wind turbine (For example, 23 teeth), calculate the gear meshing frequency range [ min × , max × ].

[0042] The rotation range is used to identify the rotational speed of the high-speed shaft body; the meshing range is used to identify the rotational speed characteristics generated by gear meshing.

[0043] Subsequent candidate fundamental frequency extraction is performed within the above-mentioned range to avoid erroneously extracting the peak value of the rotational speed and other high-amplitude signals (such as damage signals and noise) in other frequency bands.

[0044] For the measuring points of the generator drive-end bearing and the generator non-drive-end bearing, since the generator rotor is rigidly connected to the high-speed shaft of the gearbox through a coupling, its mechanical speed is consistent with that of the high-speed shaft of the gearbox. This embodiment is based on the stable speed range of the high-speed shaft given in the unit's technical manual. min Max (unit: RPM) is divided by 60 to obtain the high-speed shaft rotation frequency range. min, [max] (unit: Hz), and this frequency range is used as the candidate fundamental frequency search range for both the generator drive end and non-drive end measurement points. Step 4, Candidate Fundamental Frequency Extraction Select local extrema within the frequency range; The frequency points that satisfy the preset amplitude threshold among multiple local extrema are selected as candidate fundamental frequencies.

[0045] Step 5: Significance and Confidence Calculation Saliency Definition and Calculation: For each candidate fundamental frequency, the amplitude of the candidate fundamental frequency and the amplitudes of its corresponding harmonics are combined. Weights are assigned to the candidate fundamental frequency and its harmonics, and a combined amplitude value is calculated according to the preset weights. The significance is obtained by dividing the sum of the weights by the sum of the combined amplitude values. .

[0046] In this embodiment, the comprehensive amplitude value E is defined as follows: E = w0·A0 + w1·A1 + w2·A2 + … + wM·AM Where A0 is the compensated amplitude of the candidate fundamental frequency; AM is the compensated amplitude of the Mth harmonic; and wM is the corresponding preset weight.

[0047] The significance level S is: S = E / (w0 + w1 + … + wM). In this embodiment, the candidate fundamental frequency and its first 3 to 5 harmonics are selected for significance calculation. Considering the structural characteristics of different models, the harmonic order can be adjusted according to actual needs.

[0048] In this embodiment, the preset weights of the candidate fundamental frequency and several harmonics are set as follows: the weights of each order can be set according to the experience of typical vibration energy distribution, and the weights of each order decrease with each order (e.g., if the preset weight of the candidate fundamental frequency is 1, then the preset weight of the second harmonic is 0.8, the preset weight of the third harmonic is 0.6, etc.), in order to highlight the dominant role of the candidate fundamental frequency on the rotational speed, while retaining the auxiliary contribution of information such as the second and third harmonics.

[0049] In this embodiment, for the rotational frequency range corresponding to the high-speed shaft of the gearbox, the frequency range of the generator drive end, and the frequency range of the generator non-drive end, the harmonic amplitude of the candidate fundamental frequency is the maximum value of the harmonic frequency and the amplitude of multiple frequency points on both sides of the harmonic frequency, in order to compensate for the peak misalignment caused by frequency drift. For the meshing frequency range corresponding to the high-speed shaft of the gearbox, the harmonic amplitude of the candidate fundamental frequency is the average of the harmonic frequency and the amplitudes of multiple frequency points on both sides of the harmonic frequency, in order to smooth the spectrum splitting and sideband effect caused by gear meshing.

[0050] Candidate rotational speed calculation: Candidate rotational speeds are calculated based on saliency.

[0051] In this embodiment, for the rotational frequency range corresponding to the high-speed shaft of the gearbox, the frequency range at the generator drive end, and the frequency range at the generator non-drive end, the candidate rotational speed corresponding to the candidate fundamental frequency with the highest significance is equal to the fundamental frequency. 60 (unit: RPM); For the meshing frequency range corresponding to the high-speed shaft of the gearbox, the candidate rotational speed corresponding to the candidate fundamental frequency with the highest significance is equal to the fundamental frequency. 60 / Z (unit: RPM).

[0052] Confidence calculation: The combined amplitude of the candidate fundamental frequency corresponding to the candidate rotational speed is summed with the combined amplitude of several harmonics corresponding to the candidate fundamental frequency. The summation result is then compared with the average energy of the corresponding frequency range or the normalized difference to calculate the confidence level. .

[0053] The confidence level reflects the prominence of the candidate fundamental frequency and is further used to determine the start-up and shutdown status. For the non-started or low-speed operation, although a "highest point" may appear in the significance sequence, its energy is very different from the background energy, and the confidence level is lower than the threshold. Therefore, it is determined to be a state where the operating speed has not been reached (i.e., non-started or low-speed operation).

[0054] Candidate speed output: When the confidence level is less than the threshold, output an "Not Started" or "Micro-Spin" flag. When the confidence level is greater than or equal to the threshold, the output rotational speed is the calculated candidate rotational speed.

[0055] Step 6: Consistency determination of multiple measurement points and result fusion Results Summary and Clustering: The candidate speed results obtained from each measuring point (including the high-speed shaft rotation frequency band of the gearbox, the meshing frequency band, and four candidate speeds for the generator drive end and non-drive end) are summarized into a list. These values ​​are clustered according to the preset "speed tolerance" (e.g., ±30 rpm) to obtain several consistent groups. When the speed difference between different measuring points is less than the tolerance, they are considered to come from the same real rotation state.

[0056] Exclusion of outlier measurement points and weighted average: If cluster analysis finds that the result of a certain measurement point deviates significantly from other measurement points, then the result of that measurement point will be removed in the current period; For the consensus group with the largest number of samples among the obtained consensus groups, the candidate rotation speeds are weighted and averaged according to the significance or confidence of the measurement points to obtain the final output rotation speed.

[0057] Step 7, Output and Recording Output results: The system outputs the final rotational speed value at each moment, as well as the significance, confidence level, and filtering reasons of each measurement point, to the log file and host computer software for operation and maintenance personnel to query and perform subsequent diagnosis.

[0058] Additional features: This embodiment can also generate spectrum diagrams and saliency distribution diagrams for engineers to analyze; at the same time, it counts the occurrence of various anomalies to provide a basis for subsequent threshold adjustments.

[0059] The working principle and effects of this embodiment: This embodiment utilizes techniques such as multi-point data acquisition, multi-frequency energy aggregation, frequency compensation, dual-index evaluation of significance and confidence, and multi-point consistency fusion to accurately identify the high-speed shaft speed of a wind turbine without a speed sensor, achieving the following significant effects: (1) Cumulative analysis of multi-frequency rotational speed characteristics This embodiment fully utilizes the energy distribution patterns of the fundamental frequency and multiple harmonics in the vibration signals of rotating machinery, treating these frequency components as multiple physical bases reflecting rotational speed characteristics. By accumulating the harmonic amplitudes with weights as a significance index, the ability to identify the true rotational frequency can be significantly enhanced, avoiding misjudgments caused by the traditional "single-peak maximization" algorithm in the presence of gear meshing, resonance, or stray noise. The harmonic energy aggregation mechanism employed in this embodiment makes the algorithm more sensitive to weaker but multi-order distributed harmonic components in the signal, thereby improving robustness in low signal-to-noise ratio environments.

[0060] (2) Confidence measurement Based on the calculation of significance, this embodiment introduces a confidence index, which quantifies the reliability of the identification results by using the ratio of the combined amplitude of the candidate rotational speed frequency and its harmonics to the background energy of the surrounding spectrum. This index not only reflects the prominence of a single identification result but can also be used to evaluate the stability of the identification in conjunction with time series trends. When the wind turbine is not running or operating at low energy, although spurious peaks with the highest significance may still appear in the spectrum, the difference between the combined amplitude of the candidate frequency and the background energy is extremely small, and the confidence level is significantly lower.

[0061] By setting a joint threshold for saliency and confidence, the zero-speed (not started) state can be automatically identified and invalid results can be masked, preventing the system from outputting false speeds under static or low-vibration conditions. This significantly improves the algorithm's identification accuracy and self-correction capability in complex field environments. Furthermore, this embodiment can dynamically compare the current identification results with historical operating data or the confidence change trend of continuous time series. By comparing the energy distribution characteristics of historical start-up and shutdown phases, when the system detects that the overall energy level of the spectrum matches known shutdown characteristics, the not-start state can be further confirmed.

[0062] This mechanism not only utilizes instantaneous features for identification but also incorporates trend information over time, giving the algorithm a near-empirical judgment capability. As a result, the system can maintain stable judgments under noise disturbances or sudden energy fluctuations, significantly improving the accuracy of non-startup identification and its long-term adaptability. This mechanism not only provides interpretability to the speed extraction process but also endows the system with adaptive quality control capabilities, enabling it to continuously output stable and reliable identification results without human intervention.

[0063] (3) Frequency harmonic neighborhood energy compensation To address the inherent discretization problem of Fast Fourier Transform (FFT) and the spectral ambiguity caused by speed changes during wind turbine operation, this invention proposes a frequency-doubled neighborhood energy compensation method. This method involves selecting optimal values ​​(choosing the point with the largest amplitude in the neighborhood) or averaging the neighborhood values ​​at the nominal frequency-doubled position and several adjacent discrete frequency points to recover the amplitude loss caused by frequency misalignment or energy diffusion.

[0064] This method is particularly crucial for situations where the sampling rate is limited (not higher than 25600Hz), the signal acquisition time is long (e.g., 5 seconds or more), and the rotation speed fluctuates slightly during the sampling period. It can effectively improve the integrity of the frequency multiplication energy accumulation and prevent the saliency calculation from being distorted due to energy loss, thereby significantly improving the speed extraction accuracy and algorithm stability under non-steady-state conditions.

[0065] (4) Consistency determination of multiple measurement points To avoid misjudgment of the most significant signal (candidate speed signal) extracted from a single measuring point due to factors such as local damage to the monitored structure and environmental noise, this invention proposes a fusion strategy based on multi-measuring point consistency verification. This strategy comprehensively utilizes the vibration signal results from three typical monitoring points: the high-speed shaft of the gearbox, the drive end of the generator, and the non-drive end of the generator, to achieve cross-validation of the speed values ​​identified at different locations.

[0066] First, the system calculates the independent speed identification result for each measuring point. Then, using the "speed tolerance range" as the criterion, results with similar values ​​are automatically grouped together. When the difference between the identification results of different measuring points is less than a set threshold (e.g., ±30 RPM), they are considered to come from the same actual rotation state; if the difference exceeds this range, it is judged as an abnormal result. This threshold is not a fixed physical constant, but is determined based on the typical speed fluctuations and signal consistency of wind turbine units, which can accommodate natural differences between measuring points while effectively identifying abnormal deviations.

[0067] Based on this, the system prioritizes the consistent group containing the largest number of measurement points as the valid result, and calculates the weighted average rotational speed using the signal quality indicators (such as significance or confidence) of each measurement point as weights. Through this dynamic clustering and adaptive weighting mechanism, the system can automatically identify and eliminate local abnormal results while retaining complementary information from multiple measurement points. This allows it to output stable and reliable global rotational speed results even in noisy environments or when some measurement points fail. This mechanism effectively improves the algorithm's anti-interference capabilities and long-term stability in actual wind farm operations.

[0068] (5) Start-stop status recognition By utilizing a combined judgment mechanism of significance and confidence thresholds, this embodiment can automatically identify the start-up, shutdown, and low-energy states of the unit. When the overall energy characteristics are insufficient to support reliable identification, the system actively outputs an "invalid" or "zero speed" flag to prevent false low-speed results from occurring during the stationary phase.

[0069] (6) Lightweight implementation This embodiment employs a computational framework based on FFT and frequency domain energy distribution, avoiding the high-complexity computations of time-frequency ridge tracking, empirical mode decomposition, matching pursuit, or deep learning models used in existing technologies. The algorithm structure consists of lightweight operators such as weighted calculation, threshold determination, and cluster fusion, which can run in real time at the acquisition end or edge computing unit to achieve edge-side state sampling (such as uploading vibration data at high speeds).

[0070] (7) Dual-frequency band speed recognition mechanism for high-speed shaft of gearbox To address the characteristic that high-speed shaft vibration signals in gearboxes simultaneously contain both bearing rotational characteristic frequencies and gear meshing characteristic frequencies, this embodiment proposes a frequency band partitioning identification strategy. The algorithm independently calculates significance and confidence levels within the "high-speed shaft rotational frequency range" and the "gear meshing frequency range," extracting the corresponding rotational speed results for each. The former reflects the rotational state of the bearing and shaft system; the latter reflects the transmission characteristics of the gear meshing stage and is converted into high-speed shaft rotational speed using the number of teeth. Both candidate rotational speeds participate in the multi-measurement point consistency determination.

[0071] This method fully explores the frequency band information (rotation frequency and meshing frequency band) of the high-speed shaft measuring points of the gearbox, and forms two independent candidate rotational speed results within a single measuring point, thereby maximizing the use of vibration data from the same measuring point and increasing the number of samples participating in multi-measuring-point clustering analysis.

[0072] This embodiment aims to accurately identify the high-speed shaft speed of the wind turbine by using vibration signals collected by the wind turbine condition monitoring system (CMS) in the event of failure or absence of the speed sensor, and ensures stable and reliable results through multi-point, multi-frequency band analysis and consistency fusion.

[0073] In summary, the vibration signal-based wind turbine speed extraction method provided by this invention is practically sound, has a simple structure, and can accurately extract speed even in the absence of a speed sensor or when the sensor signal is abnormal. It has broad engineering application value and has been successfully applied to nearly 14,000 units within the Huaneng Group.

[0074] Example 4 This embodiment also provides a computing device. The computing device includes a bus, a processor, a memory, and a communication interface. The processor, memory, and communication interface communicate with each other via the bus. The computing device can be a server or a terminal device. It should be understood that this application does not limit the number of processors and memory in the computing device.

[0075] A bus can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. Buses can be categorized as address buses, data buses, control buses, etc. For ease of representation, a bus can include a path for transmitting information between various components of a computing device (e.g., memory, processor, communication interfaces).

[0076] The processor may include any one or more of the following: central processing unit (CPU), graphics processing unit (GPU), tensor processing unit (TPU), application specific integrated circuit (ASIC), field-programmable gate array (FPGA), microprocessor (MP), or digital signal processor (DSP).

[0077] Memory can include volatile memory, such as random access memory (RAM). Processors can also include non-volatile memory. volatile memory, such as read-only memory (ROM). ROM (memory only), flash memory, hard disk drive (HDD), or solid state drive (SSD).

[0078] The memory stores executable program code, which the processor executes to implement the functions of the aforementioned units, thereby achieving, for example, the method described in Embodiment 1. That is, the memory may store instructions for the methods and functions relating to the computing device in any of the above embodiments.

[0079] The communication interface uses transceiver modules such as, but not limited to, network interface cards and transceivers to enable communication between computing devices and other devices or communication networks.

[0080] Example 5 This embodiment also provides a computer-readable storage medium storing computer instructions that, when executed by a processor, cause the processor to perform the methods and functions of the computing device involved in any of the above embodiments.

[0081] Generally, the various embodiments of this disclosure can be implemented in hardware or dedicated circuitry, software, logic, or any combination thereof. Some aspects can be implemented in hardware, while others can be implemented in firmware or software, which can be executed by a controller, microprocessor, or other computing device. Although various aspects of the embodiments of this disclosure are shown and described as block diagrams, flowcharts, or represented using some other illustration, it should be understood that the blocks, apparatuses, systems, techniques, or methods described herein can be implemented as, as non-limiting examples, in hardware, software, firmware, dedicated circuitry or logic, general-purpose hardware or controllers or other computing devices, or some combination thereof.

[0082] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application, and should all be included within the protection scope of this application.

Claims

1. A method for extracting the reliability of the rotating speed of a wind turbine based on multi-point vibration signals, characterized in that, Includes the following steps: Step 1: Preprocess the vibration signals of the key parts of the wind turbine under test to obtain the preprocessed vibration signals. Step 2: Perform frequency domain transformation on the preprocessed vibration signal to obtain the frequency sequence and the corresponding amplitude sequence; Step 3: Based on the frequency sequence, obtain the frequency range corresponding to each key part; Step 4: Select candidate fundamental frequencies based on frequency ranges; Step 5: Calculate the significance and confidence based on the candidate fundamental frequencies; Step 6: Calculate candidate rotational speeds based on significance and confidence. Step 7: Cluster the candidate rotation speeds corresponding to multiple key components to obtain the final rotation speed.

2. The wind turbine rotor speed reliability extraction method based on multi-point vibration signals according to claim 1, characterized in that, The key components of the wind turbine under test include the high-speed shaft of the gearbox, the generator drive end bearing, and the generator non-drive end bearing.

3. The wind turbine rotor speed reliability extraction method based on multi-point vibration signals according to claim 1, characterized in that, The method for selecting candidate fundamental frequencies based on frequency ranges is as follows: Select local extrema within the frequency range; The frequency points that satisfy the preset amplitude threshold among multiple local extrema are selected as candidate fundamental frequencies.

4. The wind turbine rotor speed reliability extraction method based on multi-point vibration signals according to claim 1, characterized in that, The significance is calculated based on the candidate fundamental frequency. The specific method is as follows: Based on the amplitude of each candidate fundamental frequency and the amplitudes of multiple harmonics of the candidate fundamental frequency, a significance is calculated according to a preset weight. Each harmonic is placed within a corresponding frequency range, and each harmonic is the maximum value of the sampling point and multiple frequency points on both sides or the average value of the sampling point and multiple frequency points on both sides.

5. The wind turbine rotor speed reliability extraction method based on multi-point vibration signals according to claim 1, characterized in that, The confidence score is calculated based on the candidate fundamental frequency. The specific method is as follows: The confidence level is calculated based on each candidate fundamental frequency and its corresponding harmonics.

6. The wind turbine rotor speed reliability extraction method based on multi-point vibration signals according to claim 1, characterized in that, Candidate rotational speeds are calculated based on significance and confidence levels, specifically using the following method: Candidate rotational speeds are calculated based on the candidate fundamental frequency corresponding to the maximum significance value. The obtained confidence level is compared with a preset threshold, and the candidate rotation speed is determined based on the comparison result.

7. The wind turbine rotor speed reliability extraction method based on multi-point vibration signals according to claim 1, characterized in that, The final rotational speed is obtained by clustering candidate rotational speeds corresponding to multiple key components. The specific method is as follows: By utilizing the rotational speed tolerance, candidate rotational speeds corresponding to multiple key components are clustered to obtain multiple sets of data; The final rotational speed is obtained by weighting the rotational speed estimates based on the significance or confidence level of the data set with the largest number of samples. 8.A system for extracting wind turbine rotating speed reliability based on multi-point vibration signals, characterized in that, include: The vibration signal preprocessing unit is used to preprocess the vibration signals of the key parts of the wind turbine under test to obtain the preprocessed vibration signal. The frequency domain conversion unit is used to convert the preprocessed vibration signal into a frequency domain to obtain a frequency sequence and a corresponding amplitude sequence. The frequency range acquisition unit is used to acquire the frequency range corresponding to each key part based on the frequency sequence. The candidate fundamental frequency selection unit is used to select candidate fundamental frequencies based on a frequency range; The confidence calculation unit is used to calculate the significance and confidence based on the candidate fundamental frequency; The candidate rotational speed calculation unit is used to calculate candidate rotational speeds based on significance and confidence. The rotational speed clustering unit is used to cluster candidate rotational speeds corresponding to multiple key components to obtain the final rotational speed.

9. An electronic device, characterized in that, It includes a processor and a memory, the memory storing computer instructions that, when executed by the processor, cause the electronic device to perform the method of any one of claims 1 to 7.

10. A computer program product, characterized in that, The computer program product includes computer-executable instructions that, when executed, implement the method of any one of claims 1 to 7.

Images