Offshore wind turbine adaptive fault early warning method

Abstract

This invention relates to an adaptive fault early warning method for offshore wind turbines, belonging to the field of offshore wind turbine fault monitoring and early warning technology. Addressing the technical problem that traditional fixed-threshold early warning methods struggle to adapt to dynamic changes in wind speed and wave height in the marine environment, leading to high false alarm and false alarm rates, this invention utilizes vibration sensors installed in the nacelle and tower, temperature sensors in the generator bearing housing, an ultrasonic anemometer on the top of the wind turbine, and a wave radar on the foundation structure to collect real-time effective values ​​of vibration velocity, bearing temperature, ten-minute average wind speed, and significant wave height. The average wind speed and significant wave height are input into an adaptive threshold calculation module with built-in vibration and temperature threshold response functions to obtain corresponding adaptive early warning thresholds. The real-time collected effective values ​​of vibration velocity and bearing temperature are then compared with the corresponding thresholds, enabling accurate and timely early warning of offshore wind turbine faults and ensuring the safe and reliable operation of the turbine.

Description

Technical Field

[0001] This invention relates to the field of wind turbine monitoring and fault early warning technology. More specifically, this invention relates to an adaptive fault early warning method for offshore wind turbines. Background Technology

[0002] Offshore wind power, as an important component of clean energy, operates its turbines in the marine environment for extended periods. The vibration status of the nacelle and tower, as well as the temperature changes in the generator bearings, are key indicators reflecting the turbine's operational health. Real-time monitoring of these parameters and timely fault warnings are crucial for ensuring safe and stable operation and reducing maintenance costs. Currently, the industry primarily uses fixed threshold monitoring methods for fault warnings of offshore wind turbines. This involves pre-setting fixed warning thresholds for the effective value of vibration velocity and bearing temperature. When the real-time collected parameter values ​​exceed these fixed thresholds, a fault warning signal is triggered.

[0003] The marine environment exhibits significant dynamic fluctuations, with wind speed and significant wave height constantly changing. These two environmental parameters are directly related to the operating load and stress state of wind turbines. When the average wind speed increases, the stress on the rotor increases, naturally leading to a rise in the vibration intensity of the nacelle and tower. Simultaneously, the increased generator load also increases the bearing cooling pressure, resulting in a rise in temperature. Changes in significant wave height are transmitted through the foundation structure to the turbine itself, further affecting the vibration characteristics of the nacelle and tower. Traditional fixed-threshold early warning methods do not consider the dynamic impact of environmental factors such as wind speed and wave height. Their warning thresholds are determined based on specific or average operating conditions and cannot be adaptively adjusted to changes in environmental parameters. This leads to a situation where, under normal operating conditions with high wind speeds and high wave heights, the effective value of the unit's vibration velocity and bearing temperature can easily reach or even exceed the fixed warning threshold, resulting in a large number of false alarm signals and increasing the ineffective workload of maintenance personnel. On the other hand, under operating conditions with low wind speeds and low wave heights, the fixed threshold may be set too high. Even if the unit experiences a minor fault that causes abnormal vibration or temperature rise, it may be difficult to reach the warning threshold, leading to missed alarms and making it impossible to detect faults in a timely manner. Summary of the Invention

[0004] Another objective of this invention is to provide an adaptive fault early warning method for offshore wind turbines.

[0005] To achieve these objectives and other advantages according to the present invention, an adaptive fault early warning method for offshore wind turbines is provided, comprising the following steps: Vibration sensors installed in the nacelle and tower of offshore wind turbines, and temperature sensors installed in the generator bearing housings, are used to collect the effective value of vibration velocity and bearing temperature in real time. The effective value of vibration velocity ranges from 0.5 to 20 mm / s, and the bearing temperature ranges from -10°C to 120°C. The average wind speed and significant wave height are collected in real time over ten minutes by an ultrasonic anemometer installed on the top of the wind turbine and a wave radar installed on the foundation structure. The average wind speed ranges from 3 to 30 m / s, and the significant wave height ranges from 0.3 to 8 m. The collected average wind speed and significant wave height are input into the processor's adaptive threshold calculation module. This module has built-in vibration threshold response function and temperature threshold response function trained based on the unit's historical operating data. The vibration threshold response function is a bivariate linear function with average wind speed and significant wave height as independent variables and an adaptive warning threshold for the effective value of vibration velocity as the dependent variable. Its function coefficients are determined by fitting historical data using the least squares method, and the calculated adaptive warning threshold range for the effective value of vibration velocity is 1.0-18 mm / s. The temperature threshold response function is a univariate linear function with average wind speed as the independent variable and an adaptive warning threshold for bearing temperature as the dependent variable. Its function coefficients are determined by fitting historical data using the least squares method, and the calculated adaptive warning threshold range for bearing temperature is 40°C-100°C. The real-time collected effective value of vibration velocity is compared with the adaptive warning threshold of effective vibration velocity, and the real-time collected bearing temperature is compared with the adaptive warning threshold of bearing temperature. When the real-time collected effective value of vibration velocity continues to exceed the adaptive warning threshold of effective vibration velocity for 60 seconds, or the real-time collected bearing temperature continues to exceed its corresponding adaptive warning threshold for 180 seconds, the processor's warning signal generation module sends a first-level fault warning signal to the wind farm monitoring system.

[0006] Preferably, the adaptive threshold calculation module has preset upper limits for environmental parameters such as average wind speed and significant wave height; When the real-time average wind speed or significant wave height exceeds the upper limit of its corresponding environmental parameter, the adaptive threshold calculation module will stop using the vibration threshold response function and the temperature threshold response function, and switch to a fixed limit warning threshold. The limit warning threshold for the effective value of vibration velocity is 10 mm / s, and the limit warning threshold for the bearing temperature is 85°C.

[0007] Preferably, the method further includes the following steps: Real-time acquisition of instantaneous vibration acceleration peak values ​​from vibration sensors; Set a fixed warning threshold for the peak value of instantaneous vibration acceleration; When the instantaneous vibration acceleration peak value exceeds its fixed warning threshold, and the over-limit event occurs at least 3 times in a row within 10 seconds, the warning signal generation module immediately sends a level-one fault warning signal to the wind farm monitoring system.

[0008] Preferably, when the instantaneous vibration acceleration peak value is detected to exceed its fixed warning threshold, the processor simultaneously starts to perform spectral analysis on the time domain signal of the impact event and calculates the proportion of impact energy in the low frequency band of 0.5-100 Hz; A low-frequency impact energy ratio threshold is preset. When more than half of the at least three consecutive over-limit events occur within 10 seconds, the low-frequency impact energy ratio of the events exceeds the preset low-frequency impact energy ratio threshold, the warning signal generation module sends a level-two fault warning signal to the wind farm monitoring system. The level-two fault warning signal has a higher warning level than the level-one fault warning signal, indicating a potential structural risk.

[0009] Preferably, the threshold for the proportion of low-frequency impact energy is generated by the dynamic threshold calculation module; The dynamic threshold calculation module takes the real-time collected effective wave height as input and establishes a mapping relationship between wave height and recommended warning threshold based on the normal impact energy distribution under different wave heights in historical data. When the effective wave height increases, the dynamic threshold calculation module raises the low-frequency impact energy ratio threshold to compensate for the normal low-frequency impact effect of the wave load itself on the foundation structure.

[0010] Preferably, while calculating the proportion of low-frequency impact energy, modal frequency identification is performed on the vibration signal to extract the first natural frequency of the tower structure at this time, and the value of the first natural frequency and the damping ratio are continuously monitored. The reference value of the first natural frequency and the reference value of the damping ratio of the tower under healthy conditions are pre-stored. When the conditions for sending a level 2 fault warning signal are met, the warning signal generation module further determines: if the decrease in the currently monitored first-order natural frequency value exceeds 2% of the first-order natural frequency reference value, or the increase in the currently monitored damping ratio exceeds 20% of the damping ratio reference value, then a structural performance degradation confirmation mark is added to the level 2 fault warning signal.

[0011] Preferably, the adaptive threshold calculation module is further configured to perform nonlinear correction on the vibration threshold response function and the temperature threshold response function based on the real-time operating status parameters of the wind turbine, wherein the operating status parameters include at least one of generator speed and power output; The calibration process includes: The real-time operating status parameters are input into the pre-stored correction factor table to obtain the corresponding correction factor. The correction factor table is generated by neural network training based on the unit's historical operating data. The output values ​​of the vibration threshold response function and the temperature threshold response function are multiplied by the correction factor to obtain the final adaptive warning threshold. When the generator speed exceeds 90% of the rated speed or the power output exceeds 85% of the rated power, the correction factor ranges from 0.9 to 1.1 to optimize the accuracy of the threshold under high load conditions.

[0012] Preferably, it also includes sensor health monitoring and redundancy management steps: The processor continuously monitors the data relationships of associated sensors and counts the number of times the data relationships exceed the baseline range established based on historical normal operation data within a set time window. When the number of times the limit is exceeded exceeds a preset threshold, the corresponding sensor data is marked as unreliable. The associated sensors include temperature sensors of the generator drive end and non-drive end bearings, as well as vibration sensors of the nacelle and tower. The processor judges the quality of sensor signals in parallel. If the vibration sensor signal is continuously saturated or does not change, or the temperature sensor reading changes beyond the physical limit, the sensor data is directly determined to be unreliable. When a sensor determines that the data is unreliable, perform the following actions: Isolate it from the early warning logic; Switch to the backup data source preset for this sensor. The backup data source is a backup sensor installed on the same component, or a virtual signal generated by model calculation based on data from other reliable sensors. Send an alert to the wind farm monitoring system indicating the abnormal sensor.

[0013] Preferably, in the process of determining whether the effective value of vibration velocity continuously exceeds the adaptive warning threshold, its upward gradient per unit time is calculated simultaneously. If the gradient of the effective value of vibration velocity exceeds the preset gradient threshold, the warning signal generation module will immediately send a first-level fault warning signal, without needing to meet the condition of 60 seconds.

[0014] Preferably, the warning signal generation module sends a Level 1 fault warning signal when the following combined conditions are met, without needing to meet the 60-second duration condition: The effective value of the vibration velocity collected in real time exceeds its adaptive warning threshold, and at the same time, the bearing temperature collected in real time exceeds its adaptive warning threshold.

[0015] The present invention has at least the following beneficial effects: First, this invention can dynamically adjust the early warning threshold according to the wind speed and wave height of the marine environment, adapt to the unit's operating characteristics under different sea conditions, reduce false alarms and missed alarms caused by environmental fluctuations; promptly detect abnormal vibration and temperature of the unit, provide accurate reference for operation and maintenance, and avoid the expansion of faults; ensure stable operation of the unit in complex marine environments and reduce the probability of unplanned shutdowns.

[0016] Secondly, this invention uses a fixed extreme warning threshold under extreme environmental conditions to avoid deviations in dynamic threshold calculation due to environmental extremes, thus ensuring the reliability of the warning logic; it prevents serious faults caused by sudden load changes in the unit under extreme conditions from going undetected, providing safety assurance for the unit's operation under extreme conditions; and it simplifies the threshold calculation logic under extreme sea conditions, improving the warning response speed.

[0017] Third, this invention can capture sudden impact faults of the unit, making up for the insufficient response of continuous over-limit warnings to instantaneous faults; by judging the number of times, it avoids false triggering caused by a single accidental impact, thus improving the accuracy of the warning; and it issues impact fault warnings in a timely manner to prevent the expansion of damage to unit components caused by impacts and extend the service life of components.

[0018] Fourth, this invention can distinguish the types of impact failures, accurately identify potential structural risks, and avoid confusing ordinary mechanical impacts with structural impacts; it provides more accurate fault level references for operation and maintenance, making it easier to prioritize the handling of serious hidden dangers; it provides early warning of structural problems, reduces the risk of structural failure, and ensures the overall structural safety of the unit.

[0019] Other advantages, objectives and features of the present invention will become apparent in part from the following description, and in part from those skilled in the art through study and practice of the invention. Detailed Implementation

[0020] The present invention will be further described in detail below with reference to embodiments, so that those skilled in the art can implement it based on the description.

[0021] An adaptive fault early warning method for offshore wind turbines includes the following steps: Vibration sensors installed in the nacelle and tower of offshore wind turbines, and temperature sensors installed in the generator bearing housings, are used to collect the effective values ​​of vibration velocity and bearing temperature in real time. The effective values ​​of vibration velocity range from 0.5 to 20 mm / s, and the effective values ​​of bearing temperature range from -10°C to 120°C. The average wind speed and significant wave height are collected in real time over ten minutes by an ultrasonic anemometer installed on the top of the wind turbine and a wave radar installed on the foundation structure. The average wind speed ranges from 3 to 30 m / s, and the significant wave height ranges from 0.3 to 8 m. The collected average wind speed and significant wave height are input into the processor's adaptive threshold calculation module. This module has built-in vibration threshold response function and temperature threshold response function trained based on the unit's historical operating data. The vibration threshold response function is a bivariate linear function with average wind speed and significant wave height as independent variables and an adaptive warning threshold for the effective value of vibration velocity as the dependent variable. Its function coefficients are determined by fitting historical data using the least squares method, and the calculated adaptive warning threshold range for the effective value of vibration velocity is 1.0-18 mm / s. The temperature threshold response function is a univariate linear function with average wind speed as the independent variable and an adaptive warning threshold for bearing temperature as the dependent variable. Its function coefficients are determined by fitting historical data using the least squares method, and the calculated adaptive warning threshold range for bearing temperature is 40°C-100°C. The real-time collected effective value of vibration velocity is compared with the adaptive warning threshold of effective vibration velocity, and the real-time collected bearing temperature is compared with the adaptive warning threshold of bearing temperature. When the real-time collected effective value of vibration velocity continues to exceed the adaptive warning threshold of effective vibration velocity for 60 seconds, or the real-time collected bearing temperature continues to exceed its corresponding adaptive warning threshold for 180 seconds, the processor's warning signal generation module sends a first-level fault warning signal to the wind farm monitoring system.

[0022] In this technical solution, vibration sensors can be piezoelectric or capacitive; temperature sensors can be PT100 platinum resistance or thermocouple; ultrasonic anemometers can be time-of-flight ultrasonic pulse anemometers; and wave radar can be high-frequency radar wave tide gauges. Multiple vibration sensors are mounted on the nacelle base, nacelle walls, and the upper, middle, and lower parts of the tower. Multiple temperature sensors are mounted on the drive and non-drive ends of the generator bearing housing. The ultrasonic anemometer is mounted on the top of the nacelle and hub of the wind turbine. The wave radar is mounted on the outer side of the jacket foundation and monopile foundation of the foundation structure. After the sensors are activated, they capture corresponding parameters in real time according to the set acquisition frequency. The vibration sensors convert vibration signals into electrical signals and calculate the effective value of vibration velocity. The temperature sensors directly sense the bearing temperature and output the value. The ultrasonic anemometer calculates the average wind speed through the ultrasonic wave propagation time difference. The wave radar obtains the effective wave height by transmitting and receiving radar waves. All data is transmitted to the processor in real time. The processor can be an industrial-grade microcontroller or PLC controller. The adaptive threshold calculation module can be a software module integrated into the processor or a separate digital signal processing module. The processor and adaptive threshold calculation module are installed in the control cabinet of the nacelle or the control box at the bottom of the tower. The coefficients of the vibration threshold response function and the temperature threshold response function are obtained by fitting historical operating data of the unit for no less than two years. The historical data should cover normal operating conditions with wind speeds of 3-25 m / s and wave heights of 0.5-6 m. Before fitting, the data needs to be preprocessed, including filtering using the moving average method and removing outliers according to the Laida criterion (3σ). To prevent overfitting, the Ridge Regression algorithm is used for function fitting, and its regularization strength parameter is determined through cross-validation. After receiving the real-time average wind speed and significant wave height, the processor inputs them into the adaptive threshold calculation module. The module calls the pre-stored vibration threshold response function and temperature threshold response function, substitutes the values, and calculates the corresponding adaptive warning threshold. The calculation result is stored in real time and transmitted to the warning judgment stage. The duration of exceeding the limit can be selected as 60s, 90s, 180s, or 240s. The wind farm monitoring system can be a centralized monitoring host or a distributed monitoring terminal. The warning signal generation module is assembled with the processor, and the wind farm monitoring system is installed in the central control room or remote monitoring center of the wind farm. The warning signal generation module receives the real-time collected effective value of vibration velocity, bearing temperature, and the corresponding adaptive warning threshold in real time, and compares the two one by one. When the effective value of vibration velocity exceeds the threshold and the duration reaches 60s, or the bearing temperature exceeds the threshold and the duration reaches 180s, the module immediately generates a first-level fault warning signal and sends it to the wind farm monitoring system via wired or wireless communication.

[0023] By adopting this technical solution, the present invention can dynamically adjust the early warning threshold according to the wind speed and wave height of the marine environment, adapt to the unit's operating characteristics under different sea conditions, reduce false alarms and missed alarms caused by environmental fluctuations, promptly detect abnormal vibration and temperature of the unit, provide accurate reference for operation and maintenance, and avoid the expansion of faults; ensure the stable operation of the unit in complex marine environments and reduce the probability of unplanned shutdowns.

[0024] In another technical solution, the adaptive threshold calculation module has preset upper limits for environmental parameters such as average wind speed and significant wave height. When the real-time average wind speed or significant wave height exceeds the upper limit of its corresponding environmental parameter, the adaptive threshold calculation module will stop using the vibration threshold response function and the temperature threshold response function, and switch to a fixed limit warning threshold. The limit warning threshold for the effective value of vibration velocity is 10 mm / s, and the limit warning threshold for the bearing temperature is 85°C. The upper limit for average wind speed can be selected as 30 m / s, 32 m / s, or 35 m / s, and the upper limit for significant wave height can be selected as 8 m, 9 m, or 10 m. The upper limit values ​​for environmental parameters such as average wind speed and significant wave height are pre-stored in the adaptive threshold calculation module. These upper limit values ​​are determined based on the unit's design operating limits and common extreme values ​​in the marine environment. The module calls upon these upper limit values ​​at any time during operation for subsequent judgments. The difference between the real-time collected average wind speed and the upper limit value can exceed 0.5 m / s or 1 m / s, and the difference between the real-time collected significant wave height and the upper limit value can exceed 0.3 m or 0. At 5m, the adaptive threshold calculation module will stop using the vibration threshold response function and the temperature threshold response function. The adaptive threshold calculation module receives real-time average wind speed and significant wave height, compares them with pre-stored upper limits for the corresponding environmental parameters, and determines whether the real-time values ​​exceed the upper limits. The comparison results are fed back to the module's control unit in real time. The effective value limit warning threshold for vibration velocity can be selected as 10mm / s, 9mm / s, or 11mm / s, and the limit warning threshold for bearing temperature can be selected as 85℃, 83℃, or 87℃. The adaptive threshold calculation module integrates a threshold switching module. When the judgment result indicates that the real-time average wind speed or significant wave height exceeds the upper limit, the control unit of the adaptive threshold calculation module issues a command to stop calling the vibration threshold response function and the temperature threshold response function, and simultaneously activates the threshold switching module to switch the warning threshold to a preset fixed limit warning threshold. The switched threshold is transmitted to the warning judgment stage in real time. By adopting this technical solution, the present invention uses a fixed extreme warning threshold under extreme environmental conditions to avoid deviations in dynamic threshold calculation due to environmental extremes, thus ensuring the reliability of the warning logic; it prevents serious faults caused by sudden load changes in the unit under extreme conditions from going undetected, providing safety assurance for the unit's operation under extreme conditions; and it simplifies the threshold calculation logic under extreme sea conditions, improving the warning response speed.

[0025] Another technical solution also includes the following steps: Real-time acquisition of instantaneous vibration acceleration peak values ​​from vibration sensors; Set a fixed warning threshold for the peak value of instantaneous vibration acceleration; When the peak instantaneous vibration acceleration exceeds its fixed warning threshold, and this over-limit event occurs at least three times consecutively within 10 seconds, the warning signal generation module immediately sends a Level 1 fault warning signal to the wind farm monitoring system. The acquisition range of the peak instantaneous vibration acceleration can be selected from 0-50 m / s². 2 0-100m / s 2 The acquisition accuracy can be selected to be 0.01m / s. 2 0.1m / s 2 The vibration sensor simultaneously captures instantaneous vibration acceleration signals while collecting the effective value of vibration velocity. The internal signal processing circuit converts these signals into peak data, i.e., the peak value of instantaneous vibration acceleration, which is transmitted to the processor in real time. The threshold storage unit can be the processor's built-in memory or an external memory card. The fixed warning threshold is determined based on the unit's structural strength design value and the peak impact acceleration data from historical fault data. By experimentally testing the impact of different impact intensities on the unit, a value that reflects potential faults is selected as the fixed warning threshold. A pre-set fixed warning threshold is stored in the threshold storage unit. During operation, the processor retrieves this threshold from the storage unit for comparison and judgment. The judgment time window can be selected as 10s, 8s, or 12s, and the over-limit threshold can be selected as 3 or 4 times. The counting module can be a built-in counter in the processor or an independent digital counter. Integrated into the processor, the counting module receives the instantaneous vibration acceleration peak value and compares it with the fixed warning threshold. If the threshold is exceeded, the counting module increments by 1, and the timing module starts timing. When the counting module records 3 over-limit times within 10 seconds, the warning signal generation module immediately generates a level-one fault warning signal and sends it to the wind farm monitoring system. Using this technical solution, the present invention can capture sudden impact faults of the unit, compensating for the insufficient response of continuous over-limit warnings to instantaneous faults; by judging the number of occurrences, it avoids false triggering caused by a single accidental impact, improving the accuracy of the warning; and promptly issues impact fault warnings to prevent the expansion of damage to unit components caused by impacts, extending the service life of components.

[0026] In another technical solution, when the instantaneous vibration acceleration peak is detected to exceed its fixed warning threshold, the processor simultaneously starts to perform spectral analysis on the time domain signal of the impact event and calculates the proportion of impact energy in the low frequency band of 0.5-100 Hz. A preset threshold for the proportion of low-frequency impact energy is established. When more than half of the at least three consecutive out-of-limit events occurring within 10 seconds have a low-frequency impact energy proportion exceeding the preset threshold, the warning signal generation module sends a secondary fault warning signal to the wind farm monitoring system. The secondary fault warning signal has a higher warning level than the primary fault warning signal, indicating a potential structural risk. The analysis frequency band is 0.5-100Hz, which can be further subdivided into 0.5-10Hz, 10-50Hz, and 50-100Hz. The energy proportion calculation accuracy can be selected as 0.1% or 1%. The spectrum analysis module can be a digital signal processor or a software module integrating spectrum analysis functions. The signal acquisition unit is paired with a vibration sensor. The spectrum analysis module is installed inside the processor in a separate signal processing box. When the instantaneous vibration acceleration peak exceeds a fixed warning threshold, the processor initiates the spectrum analysis module to perform a Fourier transform on the time-domain vibration signal of the impact event (the signal sampling frequency is set to 500 Hz, and the Hanning window is selected as the window function; the calculation process is specifically as follows: a Fast Fourier Transform (FFT) is performed on the windowed time-domain signal to obtain the power spectral density; firstly, the amplitude of all spectral lines in the 0.5 Hz to 100 Hz frequency band is integrated on the power spectrum to obtain the low-frequency impact energy). Low Then, the total impact energy is obtained by integrating the amplitudes of all spectral lines across the entire frequency band (0 Hz to the Nyquist frequency, i.e., 250 Hz). Etotal Ultimately, the proportion of low-frequency impact energy... R According to the formula R =( Low / EtotalThe impact energy is calculated by multiplying the total impact energy by 100%, then converted to a frequency domain signal. The impact energy in the 0.5-100Hz low-frequency band is then calculated, and divided by the total impact energy to obtain the low-frequency impact energy percentage. The calculation result is stored in real time. The low-frequency impact energy percentage threshold can be selected as 30%, 40%, or 50%. This threshold is determined based on the normal impact energy distribution data under the unit's healthy state. Normal impact signals under different operating conditions are collected, and their low-frequency energy percentage range is analyzed. The upper limit of this range is selected as the preset threshold. The low-frequency impact energy percentage threshold is pre-stored in the threshold storage unit, and the processor retrieves this threshold for subsequent judgment when needed. The warning signal generation module can generate a secondary warning signal. The statistics module counts at least three consecutive over-limit events occurring within 10 seconds, checking each event's low-frequency impact energy percentage against the preset threshold. If more than half of the events meet the condition, the warning signal generation module generates a secondary fault warning signal and sends it to the wind farm monitoring system. The secondary warning level is higher than the primary warning level. By adopting this technical solution, the present invention can distinguish the types of impact failures, accurately identify potential structural risks, and avoid confusing ordinary mechanical impacts with structural impacts; it provides more accurate fault level references for operation and maintenance, making it easier to prioritize the handling of serious hidden dangers; it provides early warning of structural problems, reduces the risk of structural failure, and ensures the overall structural safety of the unit.

[0027] In another technical solution, the threshold for the proportion of low-frequency impact energy is generated by a dynamic threshold calculation module; The dynamic threshold calculation module takes the real-time collected effective wave height as input and establishes a mapping relationship between wave height and recommended warning threshold based on the normal impact energy distribution at different wave heights in historical data. When the effective wave height increases, the dynamic threshold calculation module raises the low-frequency impact energy proportion threshold to compensate for the normal low-frequency impact impact on the foundation structure caused by the wave load itself. The calculation response time of the dynamic threshold calculation module can be selected as 0.1s, 0.5s, or 1s. The dynamic threshold calculation module can be an independent microprocessor or a software module integrated into the main processor. The dynamic threshold calculation module establishes a communication connection with the main processor, receives the effective wave height data transmitted by the main processor in real time, and has a preset calculation program inside the module, ready to adjust the low-frequency impact energy proportion threshold according to the effective wave height. The effective wave height division range can be selected as 0.3-2m, 2-5m, or 5-8m, and the corresponding threshold adjustment range can be selected as 5%, 8%, or 10% for every 1m increase in wave height. When the effective wave height increases from 0.3m to 8m, the threshold can be adjusted from 30% to 60% and from 35% to 65%. After receiving real-time effective wave height data, the dynamic threshold calculation module determines the corresponding low-frequency impact energy proportion threshold based on the mapping relationship. When the effective wave height increases, the threshold adjustment unit automatically raises the low-frequency impact energy proportion threshold to compensate for the normal low-frequency impact effects caused by wave loads. The adjusted threshold is transmitted to the early warning judgment stage in real time. Using this technical solution, the present invention can accurately compensate for the interference of the ocean wave environment on impact energy analysis, reduce false alarms caused by normal wave loads, improve the accuracy of structural risk identification under different wave height conditions, adapt to complex and changing ocean environments, and avoid misjudging normal wave-induced impacts as faults, reducing the ineffective workload of maintenance personnel.

[0028] In another technical solution, while calculating the proportion of low-frequency impact energy, modal frequency identification is performed on the vibration signal to extract the first natural frequency of the tower structure at this time, and the value of the first natural frequency and the damping ratio are continuously monitored. The reference value of the first natural frequency and the reference value of the damping ratio of the tower under healthy conditions are pre-stored. When the conditions for sending a Level 2 fault warning signal are met, the warning signal generation module further determines: if the decrease in the currently monitored first-order natural frequency value exceeds 2% of the first-order natural frequency reference value, or the increase in the currently monitored damping ratio exceeds 20% of the damping ratio reference value, then a structural performance degradation confirmation mark is added to the Level 2 fault warning signal. While calculating the proportion of low-frequency impact energy, modal frequency identification is performed on the vibration signal. The first-order natural frequency of the tower structure is extracted using a modal analysis algorithm, and the corresponding damping ratio is calculated. The changes in these two parameters are monitored in real time. The reference value for the first-order natural frequency under healthy tower conditions can be selected as 1.5Hz, 2Hz, or 2.5Hz, and the reference value for the damping ratio can be selected as 2%, 3%, or 4%. The threshold for the decrease in the first-order natural frequency is 2%, and the threshold for the increase in the damping ratio is 20%. When the conditions for sending a Level 2 fault warning signal are met, the comparison and judgment unit compares the real-time monitored first-order natural frequency with the reference value, calculates the decrease, and compares the increase in the damping ratio. If either parameter reaches the set threshold, a command is sent to the warning signal generation module to add a structural performance degradation confirmation mark to the Level 2 fault warning signal. By adopting this technical solution, the present invention can provide dual verification for structural risks, further improve the reliability of fault diagnosis and avoid misjudgment; clarify the specific situation of structural performance degradation, provide more accurate fault information for operation and maintenance, and facilitate the formulation of targeted maintenance plans; reduce under-maintenance or over-maintenance caused by unconfirmed structural performance degradation, and reduce operation and maintenance costs.

[0029] In another technical solution, the adaptive threshold calculation module is also configured to perform nonlinear correction on the vibration threshold response function and the temperature threshold response function based on the real-time operating status parameters of the wind turbine, wherein the operating status parameters include at least one of generator speed and power output; The calibration process includes: The real-time operating status parameters are input into the pre-stored correction factor table to obtain the corresponding correction factor. The correction factor table is generated by neural network training based on the unit's historical operating data. The output values ​​of the vibration threshold response function and the temperature threshold response function are multiplied by the correction factor to obtain the final adaptive warning threshold. When the generator speed exceeds 90% of the rated speed or the power output exceeds 85% of the rated power, the correction factor ranges from 0.9 to 1.1 to optimize the accuracy of the threshold under high load conditions. Generator speed is acquired by a speed sensor mounted on the generator's output shaft near the coupling, while power output is acquired by a power sensor mounted on the generator's power supply line inside the control cabinet. The speed and power sensors collect the generator's speed and power output data in real time, converting the data into electrical signals that are transmitted to the main processor as operating status parameters for threshold correction. The correction factor can be selected from 0.9, 0.95, 1.0, 1.05, and 1.1. Historical operating data of the unit is collected, including deviations between actual vibration and temperature values ​​and theoretical thresholds for different speeds and power outputs. This data is trained using a neural network to generate a correction factor table. The adaptive threshold calculation module receives the real-time operating status parameters, inputs them into the correction factor table, and queries the corresponding correction factor. For 90% of the rated speed and 85% of the rated power, the correction factor ranges from 0.9 to 1.1. The adaptive threshold calculation module first calculates the initial adaptive warning threshold using a response function, then multiplies the initial threshold by a correction factor obtained from the correction factor table to obtain the final adaptive warning threshold. When the generator speed exceeds 90% of the rated speed or the power output exceeds 85% of the rated power, a correction factor within the range of 0.9-1.1 is selected to optimize the threshold accuracy under high load conditions. Using this technical solution, the invention can adjust the warning threshold according to the actual operating load of the unit, improving the warning adaptability under all operating conditions; compensate for normal deviations in unit vibration and temperature under high load, further reducing the probability of false alarms and missed alarms; ensure the accuracy of warnings under critical operating conditions such as high load, avoid missed fault detection due to load changes, and extend the service life of the unit.

[0030] Another technical solution also includes sensor health monitoring and redundancy management steps: The processor continuously monitors the data relationships of associated sensors and counts the number of times the data relationships exceed the baseline range established based on historical normal operation data within a set time window. When the number of times the limit is exceeded exceeds a preset threshold, the corresponding sensor data is marked as unreliable. The associated sensors include temperature sensors of the generator drive end and non-drive end bearings, as well as vibration sensors of the nacelle and tower. The processor judges the quality of sensor signals in parallel. If the vibration sensor signal is continuously saturated or does not change, or the temperature sensor reading changes beyond the physical limit, the sensor data is directly determined to be unreliable. When a sensor determines that the data is unreliable, perform the following actions: Isolate it from the early warning logic; Switch to the backup data source preset for this sensor. The backup data source is a backup sensor installed on the same component, or a virtual signal generated by model calculation based on data from other reliable sensors. Send an alarm to the wind farm monitoring system indicating the abnormal sensor. The time window can be selected as 10min, 30min, or 60min, and the threshold for the number of times the limit is exceeded can be selected as 3, 5, or 8 times. Continuously monitor the data relationship between related sensors, establish a baseline range for the data relationship based on historical normal operation data, and count the number of times the data relationship exceeds the baseline range within the set time window. When the number of times the limit is exceeded exceeds the preset threshold, the corresponding sensor data is marked as unreliable. The vibration sensor signal saturation threshold can be selected as 95% or 98% of the sensor's full scale, and the physical limit of the temperature sensor reading change rate can be selected as 10℃ / min or 15℃ / min. The backup sensor switching response time can be selected as 0.1s, 0.5s, or 1s. When a sensor is determined to have unreliable data, isolate it from the alarm logic and switch to the preset backup data source to ensure the continuous operation of the alarm logic. At the same time, send an alarm to the wind farm monitoring system indicating the abnormal sensor. By adopting this technical solution, the present invention can ensure the continuous operation of the early warning system when the sensor fails, avoid the interruption of the early warning due to sensor failure; promptly prompt the sensor maintenance needs, reduce indirect losses caused by sensor failure; and improve the unit operating efficiency and reduce power generation loss by using a virtual signal replacement scheme, without the need for immediate shutdown to replace the sensor.

[0031] In another technical solution, while determining whether the effective value of vibration velocity continuously exceeds the adaptive warning threshold, its upward gradient per unit time is calculated simultaneously. If the gradient of the effective value of vibration velocity exceeds the preset gradient threshold, the warning signal generation module immediately sends a level-one fault warning signal, without needing to meet the 60-second duration condition. The gradient calculation time interval can be selected as 1s, 2s, or 5s, and the calculation accuracy can be selected as 0.1mm / s. 2 0.5mm / s 2 The gradient threshold can be selected as 1 mm / s. 2 2mm / s 2 3mm / s 2 If the vibration gradient exceeds the threshold, the warning signal generation module immediately generates a Level 1 fault warning signal and sends it to the wind farm monitoring system without waiting for the 60-second continuous over-limit period. Using this technical solution, the present invention can quickly respond to faults with rapidly increasing vibration, shorten the warning response time, and provide more buffer time for fault handling; it avoids missing faults where vibration parameters slowly exceed limits but rapidly deteriorate, improving the comprehensiveness of the warning; and it prevents faults from escalating into serious damage in a short period, reducing unit maintenance costs and downtime.

[0032] In another technical solution, the warning signal generation module sends a Level 1 fault warning signal when the following combined conditions are met, without needing to meet the 60-second duration condition: The real-time collected effective value of vibration velocity exceeds its adaptive early warning threshold, and simultaneously, the real-time collected bearing temperature exceeds its adaptive early warning threshold. Using this technical solution, the present invention can identify serious faults where vibration and temperature are abnormal simultaneously, reducing the limitations of single-parameter early warning; improve the early warning sensitivity of major faults, avoiding serious unit damage caused by coordinated abnormalities of multiple components; and facilitate maintenance personnel in prioritizing the handling of complex faults, reducing the risk of major safety accidents to the unit.

[0033] Although embodiments of the present invention have been disclosed above, they are not limited to the applications listed in the specification and embodiments. They can be applied to various fields suitable for the present invention. For those skilled in the art, other modifications can be easily made. Therefore, without departing from the general concept defined by the claims and their equivalents, the present invention is not limited to the specific details and embodiments shown and described herein.

Claims

1. An adaptive fault early warning method for offshore wind turbines, characterized in that, Includes the following steps: Vibration sensors installed in the nacelle and tower of offshore wind turbines, and temperature sensors installed in the generator bearing housings, are used to collect the effective value of vibration velocity and bearing temperature in real time. The effective value of vibration velocity ranges from 0.5 to 20 mm / s, and the bearing temperature ranges from -10°C to 120°C. The average wind speed and significant wave height are collected in real time over ten minutes by an ultrasonic anemometer installed on the top of the wind turbine and a wave radar installed on the foundation structure. The average wind speed ranges from 3 to 30 m / s, and the significant wave height ranges from 0.3 to 8 m. The collected average wind speed and significant wave height are input into the processor's adaptive threshold calculation module. This module has built-in vibration threshold response function and temperature threshold response function trained based on the unit's historical operating data. The vibration threshold response function is a bivariate linear function with average wind speed and significant wave height as independent variables and an adaptive warning threshold for the effective value of vibration velocity as the dependent variable. Its function coefficients are determined by fitting historical data using the least squares method, and the calculated adaptive warning threshold range for the effective value of vibration velocity is 1.0-18 mm / s. The temperature threshold response function is a univariate linear function with average wind speed as the independent variable and an adaptive warning threshold for bearing temperature as the dependent variable. Its function coefficients are determined by fitting historical data using the least squares method, and the calculated adaptive warning threshold range for bearing temperature is 40°C-100°C. The real-time collected effective value of vibration velocity is compared with the adaptive warning threshold of effective vibration velocity, and the real-time collected bearing temperature is compared with the adaptive warning threshold of bearing temperature. When the real-time collected effective value of vibration velocity continues to exceed the adaptive warning threshold of effective vibration velocity for 60 seconds, or the real-time collected bearing temperature continues to exceed its corresponding adaptive warning threshold for 180 seconds, the processor's warning signal generation module sends a first-level fault warning signal to the wind farm monitoring system.

2. The adaptive fault early warning method for offshore wind turbines as described in claim 1, characterized in that, The adaptive threshold calculation module has preset upper limits for environmental parameters such as average wind speed and significant wave height; When the real-time average wind speed or significant wave height exceeds the upper limit of its corresponding environmental parameter, the adaptive threshold calculation module will stop using the vibration threshold response function and the temperature threshold response function, and switch to a fixed limit warning threshold. The limit warning threshold for the effective value of vibration velocity is 10 mm / s, and the limit warning threshold for the bearing temperature is 85°C.

3. The adaptive fault early warning method for offshore wind turbines as described in claim 2, characterized in that, It also includes the following steps: Real-time acquisition of instantaneous vibration acceleration peak values ​​from vibration sensors; Set a fixed warning threshold for the peak value of instantaneous vibration acceleration; When the instantaneous vibration acceleration peak value exceeds its fixed warning threshold, and the over-limit event occurs at least 3 times in a row within 10 seconds, the warning signal generation module immediately sends a level-one fault warning signal to the wind farm monitoring system.

4. The adaptive fault early warning method for offshore wind turbines as described in claim 3, characterized in that, When the instantaneous vibration acceleration peak exceeds its fixed warning threshold, the processor simultaneously initiates spectral analysis of the time-domain signal of the impact event to calculate the proportion of impact energy in the low-frequency band of 0.5-100 Hz. A low-frequency impact energy ratio threshold is preset. When more than half of the at least three consecutive over-limit events occur within 10 seconds, the low-frequency impact energy ratio of the events exceeds the preset low-frequency impact energy ratio threshold, the warning signal generation module sends a level-two fault warning signal to the wind farm monitoring system. The level-two fault warning signal has a higher warning level than the level-one fault warning signal, indicating a potential structural risk.

5. The adaptive fault early warning method for offshore wind turbines as described in claim 4, characterized in that, The threshold for the proportion of low-frequency impact energy is generated by the dynamic threshold calculation module; The dynamic threshold calculation module takes the real-time collected effective wave height as input and establishes a mapping relationship between wave height and recommended warning threshold based on the normal impact energy distribution under different wave heights in historical data. When the effective wave height increases, the dynamic threshold calculation module raises the low-frequency impact energy ratio threshold to compensate for the normal low-frequency impact effect of the wave load itself on the foundation structure.

6. The adaptive fault early warning method for offshore wind turbines as described in claim 4, characterized in that, While calculating the proportion of low-frequency impact energy, modal frequency identification is performed on the vibration signal to extract the first natural frequency of the tower structure at this time, and the value of the first natural frequency and the damping ratio are continuously monitored. The reference values ​​of the first natural frequency and the damping ratio of the tower under healthy conditions are pre-stored. When the conditions for sending a level 2 fault warning signal are met, the warning signal generation module further determines: if the decrease in the currently monitored first-order natural frequency value exceeds 2% of the first-order natural frequency reference value, or the increase in the currently monitored damping ratio exceeds 20% of the damping ratio reference value, then a structural performance degradation confirmation mark is added to the level 2 fault warning signal.

7. The adaptive fault early warning method for offshore wind turbines as described in claim 1, characterized in that, The adaptive threshold calculation module is also configured to perform nonlinear correction on the vibration threshold response function and temperature threshold response function based on the real-time operating status parameters of the wind turbine, wherein the operating status parameters include at least one of generator speed and power output; The calibration process includes: The real-time operating status parameters are input into the pre-stored correction factor table to obtain the corresponding correction factor. The correction factor table is generated by neural network training based on the unit's historical operating data. The output values ​​of the vibration threshold response function and the temperature threshold response function are multiplied by the correction factor to obtain the final adaptive warning threshold. When the generator speed exceeds 90% of the rated speed or the power output exceeds 85% of the rated power, the correction factor ranges from 0.9 to 1.1 to optimize the accuracy of the threshold under high load conditions.

8. The adaptive fault early warning method for offshore wind turbines as described in claim 1, characterized in that, It also includes sensor health monitoring and redundancy management steps: The processor continuously monitors the data relationships of associated sensors and counts the number of times the data relationships exceed the baseline range established based on historical normal operation data within a set time window. When the number of times the limit is exceeded exceeds a preset threshold, the corresponding sensor data is marked as unreliable. The associated sensors include temperature sensors of the generator drive end and non-drive end bearings, as well as vibration sensors of the nacelle and tower. The processor judges the quality of sensor signals in parallel. If the vibration sensor signal is continuously saturated or does not change, or the temperature sensor reading changes beyond the physical limit, the sensor data is directly determined to be unreliable. When a sensor determines that the data is unreliable, perform the following actions: Isolate it from the early warning logic; Switch to the backup data source preset for this sensor. The backup data source is a backup sensor installed on the same component, or a virtual signal generated by model calculation based on data from other reliable sensors. Send an alert to the wind farm monitoring system indicating the abnormal sensor.

9. The adaptive fault early warning method for offshore wind turbines as described in claim 1, characterized in that, In the process of determining whether the effective value of vibration velocity continuously exceeds the adaptive warning threshold, its gradient per unit time is calculated simultaneously. If the gradient of the effective value of vibration velocity exceeds the preset gradient threshold, the warning signal generation module will immediately send a first-level fault warning signal, without needing to meet the condition of 60 seconds.

10. The adaptive fault early warning method for offshore wind turbines as described in claim 1, characterized in that, The warning signal generation module will send a Level 1 fault warning signal when the following combined conditions are met, without needing to meet the 60-second duration condition: The effective value of the vibration velocity collected in real time exceeds its adaptive warning threshold, and at the same time, the bearing temperature collected in real time exceeds its adaptive warning threshold.