Torque fluctuation identification early warning method and system

Abstract

This invention provides a method and system for early warning of torque fluctuations, belonging to the field of intelligent wind power generation technology. The method includes: continuously acquiring real-time data of the wind turbine's main state parameters and torque; integrating the acquired real-time data into time data of a specific length for torque fluctuation identification and analysis; performing empirical mode decomposition on the time data of the specific length to obtain modal components; filtering and recombining these components to obtain recombined component data; then performing envelope spectrum decomposition to obtain the frequency and amplitude data corresponding to the wind turbine at this time; filtering characteristic frequencies and harmonics from the frequency and amplitude data; and performing cluster analysis on the filtered characteristic frequencies and harmonics based on the empirical torque fluctuation data to obtain the analysis results and push them out. This invention is easy to implement and cost-effective, achieving good real-time early warning results. Based on the early warning, timely follow-up measures are taken, improving the safety of the drivetrain and the entire wind turbine unit.

Description

Technical Field

[0001] This invention belongs to the field of intelligent wind power generation technology and is applied to the intelligent early warning process of wind turbine units. Specifically, it is an early warning method and system for torque fluctuation identification. Background Technology

[0002] With the technological advancements in offshore wind turbines, wind turbine capacities are increasing, turbine blade lengths are growing, and tower heights are rising, leading to a rapid increase in the demand for turbine safety. The drivetrain is a core mechanical component of a wind turbine, its importance self-evident, as it bears the crucial responsibility of converting wind energy into electrical energy. In industrial applications, frequent torque fluctuations in the wind turbine can damage the generator and potentially cause damage to the drivetrain as well.

[0003] The negative impacts of torque fluctuations on generators mainly include the following:

[0004] 1. Generators generate vibration and noise during normal operation, and large torque fluctuations will exacerbate vibration and noise, affecting the lifespan and normal operating condition of the equipment;

[0005] 2. When the generator torque fluctuates greatly, it will cause the grid frequency to be unstable. The frequency deviating from the standard value will lead to the loss of electrical energy.

[0006] 3. Large torque fluctuations can lead to stress concentration in the generator's internal components, increasing the failure rate and affecting operational reliability.

[0007] In existing technologies, to monitor, identify, and warn of torque fluctuations, accelerometers are typically installed in the drivetrain and generator components. These sensors detect drivetrain vibration in real time and collect vibration data. However, the collected data is transmitted offline periodically for remote offline analysis to determine if the vibration data is abnormal, thereby identifying drivetrain anomalies and temporarily analyzing torque fluctuations. This method only provides acceleration data, the analysis process is offline with low real-time accuracy, and it cannot confirm the cause of the anomaly. In practical applications, when a warning occurs, further onboard inspection is required, resulting in significant drawbacks and limitations. Summary of the Invention

[0008] The purpose of this invention is to overcome the deficiencies and shortcomings of existing technologies by promptly detecting abnormal torque fluctuations in wind turbine generators, thereby reducing problems in the drivetrain and generator caused by torque fluctuations. This invention proposes an easy-to-implement and cost-effective early warning method and system for identifying torque fluctuations, which has good real-time early warning performance. Based on the early warning, timely follow-up measures can reduce abnormal vibrations in the drivetrain and generator caused by torque fluctuations, thus improving the safety of the drivetrain and the entire wind turbine generator.

[0009] The present invention employs the following technical solutions to achieve its objective:

[0010] A method for early warning of torque fluctuation identification, the method comprising the following steps:

[0011] S1. Continuously acquire real-time data of the main status parameters and torque of the fan;

[0012] S2. Integrate the acquired real-time data into time data of a specific length for subsequent torque fluctuation identification and analysis.

[0013] S3. Perform empirical mode decomposition on time data of a specific length to obtain modal components; filter the modal components and reassemble them to obtain reassembled component data;

[0014] S4. Perform envelope spectrum decomposition on the recombined component data to obtain the frequency and amplitude data of the wind turbine at this time.

[0015] S5. Filter out characteristic frequencies and harmonics from the frequency and amplitude data;

[0016] S6. Based on the empirical data of torque fluctuation, cluster analysis is performed on the selected characteristic frequencies and harmonic frequencies to obtain the analysis results; the analysis results include the type of torque fluctuation and the cause of the anomaly.

[0017] Furthermore, in step S3, the basis for screening and recombining modal components is the correlation between the modal components and data of a specific length of time.

[0018] Furthermore, in step S5, characteristic frequencies and harmonic data are filtered out by determining a reference frequency from the frequency and amplitude data; the reference frequency is the frequency corresponding to the first amplitude in the frequency and amplitude data.

[0019] Furthermore, in step S6, the characteristic frequency and harmonic frequency data are clustered to obtain clustering results; based on the location of the clustering results in the region of the torque fluctuation experience data, the corresponding analysis results are obtained; different analysis results are assigned to different early warning triggering conditions, and when the analysis results are generated, the corresponding early warning triggering conditions are transmitted to the wind turbine control system in real time to change the operating status of the unit.

[0020] This invention also provides an early warning system for torque fluctuation identification, the system comprising an edge device platform, a data integration module, an empirical mode decomposition module, an envelope spectrum analysis module, a feature frequency screening module, and a cluster analysis and result push module;

[0021] The edge device platform is used to continuously collect and acquire real-time data of the wind turbine's main state parameters and wind turbine torque, obtaining a series of 10ms data points.

[0022] The data integration module is used to integrate multiple 10ms data points into time data of a specific length, providing a data foundation for torque fluctuation identification and analysis in subsequent modules;

[0023] The empirical mode decomposition module is used to perform empirical mode decomposition on time data of a specific length to obtain modal components; it is also used to filter and reorganize the modal components based on the correlation between them and the time data of a specific length to obtain reorganized component data.

[0024] The envelope spectrum analysis module is used to decompose the recombined component data into envelope spectra to obtain the frequency and amplitude data of the wind turbine at this time.

[0025] The characteristic frequency filtering module is used to filter characteristic frequencies and harmonic data from frequency and amplitude data; the filtering process is based on a determined reference frequency, which is the frequency corresponding to the first amplitude in the frequency and amplitude data;

[0026] The clustering analysis and result push module is used to perform clustering analysis on characteristic frequency and harmonic data, and push the analysis results to the wind turbine control system according to different early warning triggering conditions.

[0027] In summary, due to the adoption of this technical solution, the beneficial effects of this invention are as follows:

[0028] 1. The method of the present invention acquires and processes wind turbine data in real time, which can effectively identify torque fluctuations. This method can help optimize the corresponding torque fluctuations in the turbine control system during the wind turbine development and design phase, thereby improving turbine safety.

[0029] 2. In the later actual wind farm operation, the method of the present invention can transmit the early warning trigger result to the unit control system in a timely manner, control the wind turbine to shut down or reduce power as soon as possible, avoid the unit vibration exceeding the limit due to torque fluctuation, and thus ensure the safe and stable operation of the wind turbine. Attached Figure Description

[0030] Figure 1 This is a schematic flowchart of the method of the present invention;

[0031] Figure 2 This is a schematic diagram of the data integration process in this invention;

[0032] Figure 3 This is a schematic diagram of empirical mode decomposition in this invention;

[0033] Figure 4 This is a schematic diagram of the envelope spectrum after decomposition and recombination in this invention. Detailed Implementation

[0034] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.

[0035] Therefore, the following detailed description of the embodiments of the invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the invention without inventive effort are within the scope of protection of the invention.

[0036] Example 1

[0037] like Figure 1 As shown, a method for early warning of torque fluctuation identification includes the following steps:

[0038] S1. Continuously acquire real-time data of the main status parameters and torque of the fan;

[0039] S2. Integrate the acquired real-time data into time data of a specific length for subsequent torque fluctuation identification and analysis.

[0040] S3. Perform empirical mode decomposition on time data of a specific length to obtain modal components; filter the modal components and reassemble them to obtain reassembled component data;

[0041] S4. Perform envelope spectrum decomposition on the recombined component data to obtain the frequency and amplitude data of the wind turbine at this time.

[0042] S5. Filter out characteristic frequencies and harmonics from the frequency and amplitude data;

[0043] S6. Based on the empirical data of torque fluctuation, cluster analysis is performed on the selected characteristic frequencies and harmonic frequencies to obtain the analysis results; the analysis results include the type of torque fluctuation and the cause of the anomaly.

[0044] This embodiment will describe in detail the specific content of each step in the method. First, in step S1, real-time data of the wind turbine is acquired through the edge device platform. The real-time data acquired in this embodiment is a series of consecutive 10ms data points, with a sampling frequency of 100Hz, that is, 100 consecutive 10ms data points are collected per second.

[0045] Based on this, you can refer to Figure 2The illustration shows that in step S2, multiple 10ms data points are integrated into time data of a specific length. In this embodiment, the duration of the specific length time data is 30 seconds, meaning a total of 30 consecutive time data points per second will be acquired and integrated. During this process, all 10ms data points per second are acquired sequentially and continuously by means of data length discrimination, and data deduplication is performed after the data per second is acquired. When the total length of the acquired time data per second meets the preset time threshold, that is, when the total length reaches 30 seconds, the data is output, completing the data integration process.

[0046] The empirical mode decomposition process in step S3 follows immediately; see [link / reference]. Figure 3 The decomposition process is illustrated. In this step, the basis for screening and recombining modal components is the correlation between the modal components and data over a specific time period; that is, components with high correlation are screened out and recombined.

[0047] In this embodiment, the specific process is as follows:

[0048] S31. Perform empirical mode decomposition on the integrated time data of a specific length to obtain a continuous series of intrinsic mode components.

[0049] S32. Preset a correlation threshold, compare the correlation between the intrinsic modal components and the original content of data of a specific length of time, and filter out the modal components that are greater than the correlation threshold.

[0050] S33. Recombinate the selected modal components to obtain recombined component data.

[0051] In step S4, the envelope spectrum analysis of the recombinant component data is performed in a conventional manner to complete the decomposition of the envelope spectrum, such as... Figure 4 As shown, the corresponding frequency and amplitude data can be obtained directly.

[0052] In step S5, characteristic frequencies and harmonics are selected by determining a reference frequency from the frequency and amplitude data; in this embodiment, the reference frequency is the frequency corresponding to the first amplitude value in the frequency and amplitude data; the specific process is as follows:

[0053] S51. The frequency and amplitude data are compiled into a dictionary, with the key value being frequency and the attribute value being amplitude, to obtain the correspondence between frequency and amplitude and determine the frequency-amplitude data points.

[0054] S52. Sort the amplitude values ​​in descending order to obtain the sorted frequency-amplitude data points;

[0055] S53. Preset a specific filtering value n, in this embodiment n = 500; filter out the first 500 frequency-amplitude data points in the sorting queue;

[0056] S54. Take the frequency corresponding to the first amplitude among the first 500 frequency-amplitude data points in the sorted queue as the reference frequency;

[0057] S55. Based on the determined reference frequency, the amplitude of the corresponding harmonic is further obtained, thereby completing the screening of characteristic frequency and harmonic data.

[0058] In this embodiment, the specific process of step S55 is as follows:

[0059] S551. Based on the reference frequency, construct a reference list. The reference list stores multiple frequency reference values ​​corresponding to different integer multiples of the reference frequency. In this embodiment, the reference list is: [1×base frequency, 2×base frequency, 3×base frequency, 4×base frequency, 5×base frequency].

[0060] S552. Compare each amplitude data point in the first 500 frequency-amplitude data points with its adjacent amplitude data points; when the amplitude data point is greater than all its adjacent amplitude data points, determine that point as the peak point, and store the frequency-amplitude data point corresponding to the peak point in the filtering dictionary; after completing the comparison and filtering of all n frequency-amplitude data points, obtain the complete filtering dictionary.

[0061] S553. Traverse and filter the frequency value data in the dictionary; when the deviation of the frequency value data from any frequency reference value in the reference list is within ±0.01, store the frequency value data in the multiplication table; and so on until the traversal is completed, all characteristic frequencies and multiplication table data containing amplitude information are obtained.

[0062] Finally, in step S6, the characteristic frequency and harmonic frequency data are clustered to obtain clustering results; based on the location of the clustering results in the region of the torque fluctuation experience data, the corresponding analysis results are obtained; different analysis results are assigned to different early warning triggering conditions, and when the analysis results are generated, the corresponding early warning triggering conditions are transmitted to the wind turbine control system in real time to change the operating status of the unit.

[0063] In this embodiment, different warning triggering conditions can be preset with warning thresholds. When a warning triggering condition is generated but the value is less than the warning threshold, the wind turbine is operated at reduced capacity. When a warning triggering condition is generated and the value is greater than the warning threshold, the wind turbine is shut down and relevant maintenance personnel are notified in a timely manner to conduct on-site inspection of the wind turbine.

[0064] Example 2

[0065] Based on Example 1, this example provides an early warning system for torque fluctuation identification corresponding to the method of Example 1. The system includes an edge device platform, a data integration module, an empirical mode decomposition module, an envelope spectrum analysis module, a characteristic frequency screening module, and a cluster analysis and result push module.

[0066] The edge device platform is used to continuously collect and acquire real-time data of the wind turbine's main status parameters and torque, obtaining a series of 10ms data points.

[0067] The data integration module is used to integrate multiple 10ms data points into time data of a specific length, providing a data foundation for torque fluctuation identification and analysis in subsequent modules;

[0068] The Empirical Mode Decomposition (EMD) module is used to perform EMD on time data of a specific length to obtain modal components; it is also used to filter and reorganize the modal components based on their correlation with the time data of a specific length to obtain reorganized component data.

[0069] The envelope spectrum analysis module is used to decompose the recombined component data into envelope spectra to obtain the frequency and amplitude data of the wind turbine at this time.

[0070] The characteristic frequency filtering module is used to filter out characteristic frequencies and harmonics from frequency and amplitude data; the filtering process is based on a determined reference frequency, which is the frequency corresponding to the first amplitude in the frequency and amplitude data;

[0071] The clustering analysis and result push module is used to perform clustering analysis on characteristic frequency and harmonic data, and push the analysis results to the wind turbine control system according to different early warning trigger conditions.

Claims

1. A method for early warning of torque fluctuations, characterized in that, The method includes the following steps: S1. Continuously acquire real-time data of the main status parameters and torque of the fan; S2. Integrate the acquired real-time data into time data of a specific length for subsequent torque fluctuation identification and analysis. S3. Perform empirical mode decomposition on time data of a specific length to obtain modal components; Based on the correlation between modal components and time data of a specific length, modal components are screened and recombined to obtain recombined component data; S4. Perform envelope spectrum decomposition on the recombined component data to obtain the frequency and amplitude data of the wind turbine at this time. S5. By determining the reference frequency in the frequency and amplitude data, the characteristic frequency and harmonic data are filtered out; the reference frequency is the frequency corresponding to the first amplitude in the frequency and amplitude data; S6. Based on the empirical data of torque fluctuations, cluster analysis is performed on the selected characteristic frequencies and harmonic frequencies to obtain the analysis results; the analysis results include the type of torque fluctuations and the cause of the anomalies. The specific process of step S3 is as follows: S31. Perform empirical mode decomposition on the integrated time data of a specific length to obtain a continuous series of intrinsic mode components. S32. Preset a correlation threshold, compare the correlation between the intrinsic modal components and the original content of data of a specific length of time, and filter out the modal components that are greater than the correlation threshold. S33. Recombinate the selected modal components to obtain recombined component data; The specific process of step S5 is as follows: S51. The frequency and amplitude data are compiled into a dictionary, with the key value being frequency and the attribute value being amplitude, to obtain the correspondence between frequency and amplitude and determine the frequency-amplitude data points. S52. Sort the amplitude values ​​in descending order to obtain the sorted frequency-amplitude data points; S53. Preset a specific filter value n to filter out the first n frequency-amplitude data points in the sorting queue; S54. Take the frequency corresponding to the first amplitude in the first n frequency-amplitude data points of the sorted queue as the reference frequency; S55. Based on the determined reference frequency, the amplitude of the corresponding harmonic is further obtained, thereby completing the screening of characteristic frequency and harmonic data.

2. The early warning method for torque fluctuation identification according to claim 1, characterized in that: In step S1, real-time data of the wind turbine is acquired through the edge device platform; the real-time data is a continuous series of multiple 10ms data points with a sampling frequency of 100Hz.

3. The early warning method for torque fluctuation identification according to claim 2, characterized in that: In step S2, multiple 10ms data points are integrated into time data of a specific length. By judging the data length, all 10ms data points per second are acquired one by one, and data deduplication is performed after the data per second is acquired. When the total length of the acquired time data per second meets the preset time threshold, the data is output, and the data integration process is completed.

4. The early warning method for torque fluctuation identification according to claim 1, characterized in that, The specific process of step S55 is as follows: S551. Construct a reference list based on the reference frequency; the reference list stores multiple frequency reference values ​​corresponding to different integer multiples of the reference frequency; S552. Compare each amplitude data point in the first n frequency-amplitude data points with its adjacent amplitude data points before and after it. When the amplitude data at a certain point is greater than the amplitude data before and after it, the point is determined to be the peak point, and the frequency-amplitude data point corresponding to the peak point is stored in the filtering dictionary; after completing the comparison and filtering of all n frequency-amplitude data points, the complete filtering dictionary is obtained. S553. Traverse and filter the frequency value data in the dictionary; when the deviation of the frequency value data from any frequency reference value in the reference list is within ±0.01, store the frequency value data in the multiplication table; after completing the traversal, all characteristic frequencies and multiplication table data containing amplitude information are obtained.

5. The early warning method for torque fluctuation identification according to claim 1, characterized in that: In step S6, the characteristic frequency and harmonic frequency data are clustered to obtain clustering results; based on the location of the clustering results in the region of the torque fluctuation experience data, the corresponding analysis results are obtained; different analysis results are assigned to different early warning triggering conditions, and when the analysis results are generated, the corresponding early warning triggering conditions are transmitted to the wind turbine control system in real time to change the operating status of the unit.

6. An early warning system for identifying torque fluctuations using the method of claim 1, characterized in that: The system includes an edge device platform, a data integration module, an empirical mode decomposition module, an envelope spectrum analysis module, a feature frequency screening module, and a cluster analysis and result push module. The edge device platform is used to continuously collect and acquire real-time data of the wind turbine's main state parameters and wind turbine torque, obtaining a series of 10ms data points. The data integration module is used to integrate multiple 10ms data points into time data of a specific length, providing a data foundation for torque fluctuation identification and analysis in subsequent modules; The empirical mode decomposition module is used to perform empirical mode decomposition on time data of a specific length to obtain modal components; It is also used to filter and reorganize modal components based on their correlation with time data of a specific length to obtain reorganized component data; The envelope spectrum analysis module is used to decompose the recombined component data into envelope spectra to obtain the frequency and amplitude data of the wind turbine at this time. The characteristic frequency filtering module is used to filter out characteristic frequencies and harmonics data from frequency and amplitude data; the filtering process is based on a determined reference frequency, which is the frequency corresponding to the first amplitude in the frequency and amplitude data; The clustering analysis and result push module is used to perform clustering analysis on characteristic frequency and harmonic data, and push the analysis results to the wind turbine control system according to different early warning triggering conditions.

Images