A method for reducing continuous damage to a wind turbine generator unit after the blade of the wind turbine sweeps the tower

Abstract

The present invention relates to the field of wind power generation technology and discloses a method for reducing sustained damage to a wind turbine after a blade sweeps the tower. The method comprises the following steps: S1, installing sensors; S2, collecting sample data; S3, analyzing real-time status; and S4, controlling sustained damage. The present invention addresses the problem of prior art technologies that require immediate intervention in a sweep incident, resulting in sustained damage and serious accidents.

Description

Technical Field

[0001] The present invention relates to the technical field of wind power generation, in particular to a method for reducing continuous damage to a wind turbine generator set after blades sweep a tower. Background Art

[0002] With its outstanding resource endowment advantages and good development trends, such as abundant resources, environmental protection, high degree of automation in operation and management, and continuously decreasing cost per kilowatt-hour, wind energy has become one of the most widely developed and applied renewable energy sources. It is an important part of global renewable energy development and utilization. Its development is gradually shifting from supplementary energy to alternative energy, and its application is an important driving force for promoting energy structure optimization and low-carbon energy.

[0003] The trend toward larger wind turbine capacity in the wind power industry is certain. Large-megawatt, highly reliable, and cost-effective overall solutions for wind power projects are gaining widespread market recognition, and complete turbine manufacturers with the ability to produce large-megawatt turbines will be more competitive in the future. Advances in wind power technology are the foundation for larger turbine capacity. This increase in capacity will effectively improve the efficiency of wind energy resource utilization, enhance the overall economic viability of wind power project investment, development, and operations, improve land and sea area utilization, reduce the cost per kilowatt-hour (COP), increase returns on investment, and facilitate large-scale project development. The COP per kilowatt-hour (COP) of wind power is a key foundation for the steady progress of grid parity policies, which will also accelerate cost reductions in wind power and the development of large-megawatt turbines.

[0004] The increasing capacity of individual turbines is driving the development of technologies with longer blades and taller towers. These longer blades and taller towers increase the likelihood of blade sweeps. Once a blade sweep occurs, it can result in serious consequences. Therefore, immediate intervention is crucial to mitigate ongoing damage and minimize the severity of wind turbine losses. Summary of the Invention

[0005] In order to overcome the deficiencies of the prior art, the present invention provides a method for reducing the sustained damage to a wind turbine after a blade sweeps the tower, thereby solving the problem that the prior art cannot intervene at the moment of a tower sweep accident, thereby causing sustained damage and serious accidents.

[0006] The technical solution adopted by the present invention to solve the above problems is:

[0007] A method for reducing sustained damage to a wind turbine after blades sweep a tower comprises the following steps:

[0008] S1, sensor installation: installing sensors on wind turbines;

[0009] S2, collecting sample data: sensors collect status data of blades and towers and transmit it to the control system in real time;

[0010] S3, real-time status analysis: The control system analyzes the current status of the blades and tower based on the status data of the blades and tower;

[0011] S4, control of continuous damage: The control system compares the status signals in the sample in real time, and issues yaw, retract and / or shutdown commands in a timely manner once the blade sweep occurs.

[0012] As a preferred technical solution, in step S1, the sensor includes but is not limited to a vibration sensor and a sound sensor.

[0013] As a preferred technical solution, in step S1, a vibration sensor is installed on the tower wall, and the number of the vibration sensor is one or more.

[0014] As a preferred technical solution, in step S1, four vibration sensors are installed on the tower wall. The four vibration sensors are located on the same cross section of the tower wall, and the spacing between adjacent vibration sensors is equal.

[0015] As a preferred technical solution, step S3 includes the following steps:

[0016] SA31 calculates and extracts the characteristics of blade vibration time-domain data collected by the vibration sensor in real time to obtain the effective value and pulse factor of the vibration time-domain data over a period of time;

[0017] SA32, using the sample data of blade sweep tower, obtains the marginal spectrum energy of high-frequency IMF components;

[0018] SA33, establish an LSTM time series neural network model, use the vibration effective value, marginal spectrum energy of high-frequency IMF component, and pulse factor during normal operation of the unit to train the LSTM time series neural network model to obtain the model distribution relationship of the operating data; take the 3 sigma distribution range of the model curve of the trained LSTM time series neural network model as the threshold detection interval. If the real-time collected vibration effective value, marginal spectrum energy of high-frequency IMF component, and pulse factor values are all outside the threshold detection interval, it is considered that a vibration shock has occurred.

[0019] As a preferred technical solution, SA32 includes the following steps:

[0020] SA321, Data Decomposition: Perform empirical mode decomposition on the sample data of the blade sweep tower, decomposing the originally complex time domain signal into the sum of n intrinsic mode function components and trend terms, where n ≥ 2 and n is an integer;

[0021] SA322, Hilbert Transform: Perform Hilbert transform on all intrinsic mode function components after data decomposition to obtain a series of Hilbert spectra that can reflect the time, frequency, and energy distribution;

[0022] SA323, integrates the Hilbert spectrum on the time axis to obtain the marginal spectrum that describes the energy distribution on the frequency axis, thereby obtaining the marginal spectrum energy of the high-frequency IMF component.

[0023] As a preferred technical solution, in step S1, a sound sensor is installed on the tower or the bottom of the tower, and the number of the sound sensor is one or more.

[0024] As a preferred technical solution, in step S1, two sound sensors are installed, one of which is installed on the tower wall and the other is installed on the tower bottom platform.

[0025] As a preferred technical solution, step S3 includes the following steps:

[0026] S3B1, filtering processing: filtering the collected sound waveform data to remove environmental noise;

[0027] S3B2, joint time-frequency domain analysis: Perform a joint time-frequency domain analysis on the filtered data, transforming the one-dimensional time domain signal into a two-dimensional time-frequency plane to obtain the relationship between the sound frequency and time;

[0028] S3B3 performs differential analysis on adjacent time-frequency groups. Under normal conditions, the differential value changes continuously, and the range of change is calculated through historical data. When the differential value shows continuous mutations from one group to multiple groups of time-frequency, it is considered that a collision has occurred.

[0029] As a preferred technical solution, in step S3, the control system cleans and analyzes the status data of the blades and tower, and then analyzes the current status of the blades and tower.

[0030] Compared with the prior art, the present invention has the following beneficial effects:

[0031] (1) The present invention compares and analyzes sample data and processed data in real time, and determines the current status of the blades and tower. Once the control system detects that a blade sweep occurs, it promptly issues commands such as yaw, retract, and shut down. This solves the problem of the existing technology that it cannot intervene at the moment of a tower sweep accident, thus causing continuous damage and serious accidents.

[0032] (2) The present invention compares the current state of the blade tower with the sample data in real time, and through data analysis and other precautions, determines in real time whether a blade sweep occurs. Once a blade sweep is detected, the control system promptly issues commands such as yaw, retract the propeller, and shut down, thereby avoiding a secondary sweep accident and reducing secondary damage to the unit. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1 It is a schematic diagram of the steps of the present invention;

[0034] Figure 2 This is a schematic diagram of the installation positions of the vibration sensor and sound sensor of the present invention. DETAILED DESCRIPTION

[0035] The present invention will be further described in detail below with reference to the embodiments and the accompanying drawings, but the embodiments of the present invention are not limited thereto.

[0036] Example 1

[0037] like Figure 1 、 Figure 2 As shown, the present invention proposes a method for reducing the continuous damage to the wind turbine after the blades sweep the tower, which relates to the field of wind power generation technology. The method is: installing corresponding sensors at the corresponding positions of the wind turbine, including but not limited to vibration sensors, sound sensors, etc. Before starting the status monitoring of the wind turbine blades and tower, it is necessary to collect samples of the corresponding sensor blades hitting the tower. The sensor data for monitoring the status of the blades and tower are transmitted to the control system in real time, and the control system completes data cleaning and data analysis, and timely analyzes the current status of the blades and tower. The status signals in the samples are compared in real time. Once the control system detects that the blades sweep the tower, it will promptly issue commands such as yaw, blade retraction, and shutdown. Avoiding secondary damage to the unit due to tower sweeping, thereby reducing the degree of damage to the wind turbine.

[0038] in:

[0039] Install sensors at appropriate locations on the wind turbine to monitor the real-time status of the wind turbine blades or tower. This includes installing sensors on the tower and surrounding areas to monitor blade sweep status. Sensors used include, but are not limited to, vibration sensors and sound sensors.

[0040] Vibration sensor monitoring methods include: installing a required number of vibration sensors on the tower wall, with at least one or more. This article illustrates the installation of four vibration sensors, each spaced 90° apart. Installation can start at either the unit's 0° twist cable orientation or the unit's prevailing wind direction. The 0° twist cable orientation facilitates determining the impeller direction from the nacelle position and locking the corresponding vibration sensors. The prevailing wind direction also ensures that the sensors are close to the impact point during tower sweeps.

[0041] Acoustic sensor monitoring methods include: A minimum of one acoustic sensor can be installed near the tower or tower base. This article illustrates the installation of two acoustic sensors: one on the tower wall and one on the tower base platform.

[0042] Before initiating wind turbine blade and tower status monitoring, it's necessary to collect samples of blades striking the tower at the corresponding sensors. This involves leveraging the existing terminal or master control system on each wind turbine, adding software-enabled development, and enabling it after personnel are in place. A blade-like module or soft-packed weight is then used to rapidly scrape and strike the tower several times. During this process, raw data from sensors such as vibration and sound sensor channels is collected. To minimize data volume and facilitate analysis, a single experiment and collection time is limited to one minute.

[0043] During the collection process, raw data from sensors such as the vibration and sound sensor channels is collected. Considering actual tower sweeping conditions, the unit should be running. However, given the high noise level of the inverter, idling or emergency stopping may be considered for testing. In addition to the experimental data, several sets of data without the impact should be collected for comparative analysis. This data should also be kept within one minute.

[0044] Sensor data monitoring blade and tower status is transmitted in real time to the control system, which performs data cleaning and analysis, providing timely analysis of the current blade and tower status. Due to the complexities of wind turbine operation and the inherently complex environment, wind turbine blades and towers can also experience certain conditions during normal operation, including but not limited to vibration and sound. During normal wind turbine operation, raw sensor data is collected in real time and, through data cleaning and analysis, excluded from data generated during normal wind turbine operation. Sample data and processed data are compared and analyzed in real time to determine the current blade and tower status.

[0045] The control system compares status signals within the sample in real time. If blade sweep is detected, it promptly issues commands such as yaw, retract, and shut down. This includes real-time comparison of the current blade tower status with sample data, and through data analysis and other preventative measures, it determines in real time whether blade sweep is occurring. If a blade sweep is detected, the control system promptly issues commands such as yaw, retract, and shut down, thereby preventing a secondary sweep and minimizing secondary damage to the unit.

[0046] Example 2

[0047] like Figure 1 、 Figure 2As shown, a method for reducing the continuous damage to the wind turbine after the blades sweep the tower is provided, and corresponding sensors are installed at the corresponding positions of the wind turbine, including but not limited to vibration sensors, sound sensors, etc. Before starting the status monitoring of the wind turbine blades and tower, it is necessary to collect samples of the corresponding sensor blades hitting the tower. The sensor data for monitoring the status of the blades and tower is transmitted to the control system in real time, and the control system completes data cleaning and data analysis, and timely analyzes the current status of the blades and tower. The status signals in the samples are compared in real time. Once the control system detects that the blades sweep the tower, it will promptly issue commands such as yaw, blade retraction, and shutdown. This prevents the unit from suffering secondary damage due to tower sweeping, thereby reducing the degree of damage to the wind turbine.

[0048] It mainly includes the following specific implementation steps:

[0049] (1) Install corresponding sensors at the corresponding positions of the wind turbine generator set, including but not limited to vibration sensors, sound sensors, etc.

[0050] (2) Before starting the wind turbine blade and tower status monitoring, it is necessary to collect samples of the corresponding sensor blades hitting the tower.

[0051] (3) The sensor data that monitors the status of the blades and tower is transmitted to the control system in real time, and the control system completes data cleaning and data analysis, and timely analyzes the current status of the blades and tower.

[0052] (4) Real-time comparison of status signals within the sample. Once the control system detects that the blades are sweeping the tower, it will promptly issue commands such as yaw, retract the blades, and shut down.

[0053] The purpose of the present invention is to find a method for reducing the sustained damage of wind turbine blades after sweeping the tower. By using this method, the sustained damage of the wind turbine can be reduced when the wind turbine blades sweep the tower.

[0054] More specifically, the technical solution adopted by the present invention is as follows:

[0055] A method for reducing the sustained damage to a wind turbine after its blades sweep the tower, comprising: installing corresponding sensors at corresponding positions of the wind turbine, including but not limited to vibration sensors, sound sensors, etc. Before starting the status monitoring of the wind turbine blades and tower, it is necessary to collect samples of the corresponding sensor blades hitting the tower. The sensor data for monitoring the status of the blades and tower are transmitted to the control system in real time, and the control system completes data cleaning and data analysis, and timely analyzes the current status of the blades and tower. The status signals in the samples are compared in real time, and once the control system detects that a blade sweep has occurred, it promptly issues commands such as yaw, blade retraction, and shutdown. This prevents the unit from suffering secondary damage due to tower sweep, thereby reducing the degree of damage to the wind turbine.

[0056] in:

[0057] Install sensors at appropriate locations on the wind turbine to monitor the real-time status of the wind turbine blades or tower. This includes installing sensors on the tower and surrounding areas to monitor blade sweep status. Sensors used include, but are not limited to, vibration sensors and sound sensors.

[0058] Vibration sensor monitoring methods include: installing a required number of vibration sensors on the tower wall, with at least one or more. This article illustrates the installation of four vibration sensors, each spaced 90° apart. Installation can start at either the unit's 0° twist cable orientation or the unit's prevailing wind direction. The 0° twist cable orientation facilitates determining the impeller direction from the nacelle position and locking the corresponding vibration sensors. The prevailing wind direction also ensures that the sensors are close to the impact point during tower sweeps.

[0059] Acoustic sensor monitoring methods include: A minimum of one acoustic sensor can be installed near the tower or tower base. This article illustrates the installation of two acoustic sensors: one on the tower wall and one on the tower base platform.

[0060] Before initiating wind turbine blade and tower status monitoring, it's necessary to collect samples of blades striking the tower at the corresponding sensors. This involves leveraging the existing terminal or master control system on each wind turbine, adding software-enabled development, and enabling it after personnel are in place. A blade-like module or soft-packed weight is then used to rapidly scrape and strike the tower several times. During this process, raw data from sensors such as vibration and sound sensor channels is collected. To minimize data volume and facilitate analysis, a single experiment and collection time is limited to one minute.

[0061] During the collection process, raw data from sensors such as the vibration and sound sensor channels is collected. Considering actual tower sweeping conditions, the unit should be running. However, given the high noise level of the inverter, idling or emergency stopping may be considered for testing. In addition to the experimental data, several sets of data without the impact should be collected for comparative analysis. This data should also be kept within one minute.

[0062] Sensor data monitoring blade and tower status is transmitted in real time to the control system, which performs data cleaning and analysis, providing timely analysis of the current blade and tower status. Due to the complexities of wind turbine operation and the inherently complex environment, wind turbine blades and towers can also experience certain conditions during normal operation, including but not limited to vibration and sound. During normal wind turbine operation, raw sensor data is collected in real time and, through data cleaning and analysis, excluded from data generated during normal wind turbine operation. Sample data and processed data are compared and analyzed in real time to determine the current blade and tower status.

[0063] The control system compares status signals within the sample in real time. If blade sweep is detected, it promptly issues commands such as yaw, retract, and shut down. This includes real-time comparison of the current blade tower status with sample data, and through data analysis and other preventative measures, it determines in real time whether blade sweep is occurring. If a blade sweep is detected, the control system promptly issues commands such as yaw, retract, and shut down, thereby preventing a secondary sweep and minimizing secondary damage to the unit.

[0064] More specifically, the method for judging vibration shock, impact, and blade sweep is as follows:

[0065] 1. Impact will cause vibration shock. The entire tower plane is divided into four quadrants using four vibration sensors with a 45° range on each side. The quadrant in which the impeller is located is determined by the unit's yaw angle, and the vibration sensor data in that quadrant is used as the trusted data.

[0066] The real-time time domain data characteristics of blade vibration are calculated and extracted to obtain its effective value and pulse factor every 5 seconds.

[0067] Empirical mode decomposition (EMD) is performed on the sample data of the blade sweep tower. This decomposes the originally complex time-domain signal into the sum of n intrinsic mode function (IMF) components and a trend term. A Hilbert transform (HHT) is performed on all the IMF components after the data decomposition, resulting in a series of Hilbert spectra that reflect the time-frequency-energy distribution. The Hilbert spectra are then integrated along the time axis to obtain the marginal spectrum that describes the energy distribution along the frequency axis and the marginal spectrum energy of the high-frequency IMF components.

[0068] An LSTM time-series neural network model was established and trained using the effective vibration values, high-frequency energy, and pulse factors during normal operation of the unit to obtain a model distribution relationship for the operating data. The 3-sigma distribution range of the model curve was used as the threshold detection range. If the collected characteristic values exceeded the detection range, it was considered that a vibration shock had occurred.

[0069] 2. Impacts will produce an impact sound and multiple echoes within the tower cavity. This impact sound may not exceed the decibel level of the unit's operating noise, but it will cause changes in the sound pattern. First, the collected sound waveform data (sampling rate 25.6k / s) is filtered to remove ambient noise. A bandpass filter is preferred, and its high-pass and low-pass cutoff frequencies need to be dynamically adjusted. Then, by performing a joint time-frequency domain analysis on the filtered data, the one-dimensional time-domain signal is transformed into a two-dimensional time-frequency plane, and the relationship between the sound frequency and time is obtained. The time window is selected within a range of 5s. Next, due to the inertia of the unit, the unit's operating status does not change suddenly except for emergency shutdowns. Therefore, its time-frequency diagram should change smoothly. Differential analysis is performed on adjacent time-frequency groups. Under normal conditions, the differential value changes continuously. The range of variation can be statistically determined through historical data. When the differential value shows continuous mutations in one or more time-frequency groups, it can be considered that an impact has occurred.

[0070] 3. When the unit does not experience an emergency shutdown or other faults, if a special sound is detected by the soundprint analysis and a vibration impact is detected by the corresponding angle vibration sensor, it can be determined that the unit has a high probability of blade sweeping, and shutdown protection and inspection will be taken.

[0071] As described above, the present invention can be preferably implemented.

[0072] All features disclosed in all embodiments in this specification, or steps in all methods or processes implicitly disclosed, except for mutually exclusive features and / or steps, can be combined and / or expanded or replaced in any manner.

[0073] The above description is merely a preferred embodiment of the present invention and does not constitute any form of limitation to the present invention. Based on the technical essence of the present invention and within the spirit and principles of the present invention, any simple modification, equivalent replacement and improvement of the above embodiment shall still fall within the scope of protection of the technical solution of the present invention.

Claims

1. A method for reducing the damage sustained by wind turbine generator sets after blades sweep the tower, characterized in that: The following steps are involved: S1, installing sensors: installing vibration sensors and sound sensors on the wind turbine generator set, wherein the vibration sensors are four and are evenly distributed on the same cross section of the tower wall, with equal spacing between adjacent vibration sensors; the sound sensors are two and are installed on the tower wall and the tower bottom platform respectively; S2, collecting sample data: sensors collect status data of blades and towers and transmit it to the control system in real time; S3, analyze the real-time status: the control system analyzes the current status of the blades and tower according to the status data of the blades and tower, including: SA31, calculates and extracts the characteristics of the blade vibration time domain data collected by the vibration sensor in real time to obtain the effective value and pulse factor of the vibration time domain data over a period of time; SA32, uses the sample data of the blade scanning tower to obtain the marginal spectrum energy of the high-frequency IMF component; SA33, establishes an LSTM time series neural network model, uses the effective value of the vibration during normal operation of the unit, the marginal spectrum energy of the high-frequency IMF component, and the pulse factor to train the LSTM time series neural network model to obtain the model distribution relationship of the operating data; takes the 3 sigma distribution of the model curve of the trained LSTM time series neural network model The range is used as the threshold detection interval. If the real-time collected vibration effective value, marginal spectrum energy of high-frequency IMF component, and pulse factor value are all outside the threshold detection interval, it is considered that a vibration impact has occurred; S3B1, filtering processing: the collected sound waveform data is processed to achieve the purpose of removing environmental noise; S3B2, time-frequency domain joint analysis: the filtered data is subjected to time-frequency domain joint analysis, the one-dimensional time domain signal is transformed into a two-dimensional time-frequency plane, and the relationship between the sound frequency and time is obtained; S3B3, differential analysis is performed on adjacent time-frequency groups. In normal state, the differential value changes continuously. The range of change is calculated through historical data. When the differential value shows continuous mutations from one group to multiple groups of time-frequency, it is considered that an impact has occurred; S4, control continuous damage: the control system compares the status signals in the sample in real time, and when it is determined based on the detection of the sound sensor that a collision has occurred, and when it is confirmed based on the detection of the vibration sensor that a vibration shock has occurred, once the blade sweeping tower situation is monitored, a yaw, blade retraction and / or shutdown command is issued in time.

2. The method for reducing the sustained damage to a wind turbine after blades sweep the tower according to claim 1, characterized in that: SA32 includes the following steps: SA321, Data Decomposition: Perform empirical mode decomposition on the sample data of the blade sweep tower, decomposing the originally complex time domain signal into the sum of n intrinsic mode function components and trend terms, where n ≥ 2 and n is an integer; SA322, Hilbert Transform: Perform Hilbert transform on all intrinsic mode function components after data decomposition to obtain a series of Hilbert spectra that can reflect the time, frequency, and energy distribution; SA323, integrates the Hilbert spectrum on the time axis to obtain the marginal spectrum that describes the energy distribution on the frequency axis, thereby obtaining the marginal spectrum energy of the high-frequency IMF component.

3. The method for reducing the sustained damage to a wind turbine after blades sweep the tower according to claim 1, characterized in that: In step S3, the control system cleans and analyzes the status data of the blades and tower, and then analyzes the current status of the blades and tower.

Images