An online monitoring and intelligent diagnosis system for wind turbine vibration state

Abstract

The present application relates to the technical field of wind turbine fault diagnosis, and more particularly to a wind turbine vibration state online monitoring and intelligent diagnosis system, comprising: a data acquisition module; a first diagnosis module for executing vibration state diagnosis of the target wind turbine; a state determination module for adjusting the diagnosis mode from vibration state diagnosis to sampling point disturbance diagnosis when the wind turbine is in an unstable working state; a second diagnosis module for determining whether the wind turbine is abnormal according to a sampling point proportion or determining whether to perform feature paragraph analysis based on phase difference accumulation and abnormal point proportion; a feature paragraph analysis module for determining whether the wind turbine is abnormal according to the phase difference oscillation coefficient of each feature paragraph; and an early warning output module. The present application can improve the accuracy of wind turbine abnormality recognition and the reliability of diagnosis.

Description

Technical Field

[0001] This invention relates to the field of wind turbine fault diagnosis technology, and in particular to an online monitoring and intelligent diagnosis system for the vibration status of wind turbines. Background Technology

[0002] The time-varying, non-stationary characteristics of wind turbine vibration signals and early, subtle fault features are easily masked by strong background noise, making it difficult for traditional monitoring methods to accurately capture anomalies. Furthermore, due to the wide distribution of wind farms and the random fluctuations in operating conditions, existing diagnostic methods have limitations in handling signal distortion and adapting to varying operating conditions, resulting in poor accuracy in fault diagnosis. Therefore, improving the accuracy of wind turbine anomaly identification and diagnostic reliability is a technical problem that urgently needs to be solved by those skilled in the art.

[0003] Chinese Patent Publication No. CN121854351A discloses a multi-component collaborative vibration monitoring and fault location system for wind turbine transmission chains. The system includes: acquiring full-frequency signals through differentiated sensor deployment; obtaining a stable order spectrum resistant to speed fluctuations using an improved order analysis method based on second-order fitting and absolute value nearest neighbor matching; constructing an observation feature matrix, introducing a SCAD penalty function to establish a sparse non-negative matrix decomposition model, and employing a near-end alternating linearization minimization algorithm for efficient and stable solution, achieving precise decoupling of fault characteristics from multiple components such as the main bearing, gearbox, and generator; and finally, achieving accurate fault location and identification through cosine similarity matching. However, the above technical solution suffers from the following problems: it cannot adaptively switch diagnostic methods according to actual application scenarios, resulting in poor diagnostic accuracy under complex operating conditions. Summary of the Invention

[0004] To address this issue, the present invention provides an online monitoring and intelligent diagnosis system for the vibration status of wind turbine units, which overcomes the problem in the prior art that the diagnosis method cannot be adaptively switched according to the actual application scenario, resulting in poor diagnosis accuracy under complex working conditions.

[0005] To achieve the above objectives, the present invention provides an online monitoring and intelligent diagnosis system for the vibration status of wind turbine generators, comprising: The data acquisition module is used to acquire monitoring information from the wind turbine. The first diagnostic module, which is connected to the data acquisition module, is used to perform a vibration state diagnosis for the target wind turbine, and in the vibration state diagnosis, it determines whether there is an abnormality in the wind turbine based on the vibration anomaly degree and the vibration change coefficient. The state determination module is connected to the data acquisition module and the first diagnosis module respectively. It is used to determine whether the wind turbine is in an unstable working state based on the pitch disturbance value and energy state offset. When it is in an unstable working state, the diagnosis method is changed from vibration state diagnosis to sampling point disturbance diagnosis. The second diagnostic module, which is connected to the state determination module, is used to determine the category of sampling points based on the instability characteristic value in the sampling point disturbance diagnosis, and to determine whether the wind turbine has an abnormality based on the proportion of a type of sampling point, or to determine whether to perform feature segment analysis based on the cumulative phase difference and the proportion of abnormal points. The feature segment analysis module, which is connected to the second diagnostic module, is used to determine the feature segment based on the density of feature points and the change in phase difference during feature segment analysis, and to determine whether there is an abnormality in the wind turbine based on the phase difference oscillation coefficient of each feature segment. The early warning output module is connected to the first diagnostic unit, the second diagnostic unit and the feature segment analysis unit respectively, and is used to send an early warning when it is determined that there is an abnormality in the target wind turbine.

[0006] Furthermore, the first diagnostic module determines that wind turbine units with a vibration anomaly degree greater than a preset vibration anomaly degree or a vibration change coefficient greater than a preset vibration change coefficient are abnormal.

[0007] Furthermore, the state determination module determines that a wind turbine generator whose pitch disturbance value is greater than a preset pitch disturbance value or whose energy state offset is greater than a preset energy state offset is in an unstable operating state.

[0008] Furthermore, the second diagnostic module determines that sampling points whose instability feature values ​​are greater than or equal to the first preset instability feature value are classified as a type of sampling point; The second diagnostic module determines that sampling points whose instability feature value is greater than or equal to the first preset instability feature value are classified as a type of sampling point; The second diagnostic module determines that sampling points whose instability feature value is less than the first preset instability feature value and greater than or equal to the second preset instability feature value are classified as Class II sampling points; The second diagnostic module determines that sampling points with an instability feature value less than the second preset instability feature value are classified as three types of sampling points; Among them, the first preset instability characteristic value is greater than the second preset instability characteristic value.

[0009] Furthermore, the second diagnostic module determines that the wind turbine has an abnormality if the proportion of a certain type of sampling point is greater than or equal to the preset proportion of a certain type of sampling point.

[0010] Furthermore, in response to the condition that the proportion of a certain type of sampling point is less than the preset proportion of a certain type of sampling point, the second diagnostic module determines whether to perform feature segment analysis based on the cumulative phase difference and the proportion of abnormal points.

[0011] Furthermore, the second diagnostic module performs characteristic segment analysis on wind turbines where the cumulative phase difference is less than the preset cumulative phase difference and the percentage of abnormal points is greater than or equal to the preset percentage of abnormal points. The second diagnostic module does not need to perform feature segment analysis on wind turbines whose cumulative phase difference is greater than or equal to the preset cumulative phase difference, and determines that the wind turbine has an anomaly. The second diagnostic module does not need to perform feature segment analysis on wind turbines whose cumulative phase difference is less than the preset cumulative phase difference and whose percentage of abnormal points is less than the preset percentage of abnormal points, and determines that the wind turbine does not have any abnormalities.

[0012] Furthermore, the feature paragraph analysis module determines that among any paragraph with a preset time length within a preset time window, the paragraph that satisfies the condition that the feature point density is greater than the preset feature point density or the phase difference change degree is greater than the preset phase difference change degree is the feature paragraph.

[0013] Furthermore, the feature segment analysis module determines that wind turbine units with a phase difference oscillation coefficient greater than or equal to a preset phase difference oscillation coefficient are abnormal.

[0014] Furthermore, the instability characteristic value is determined based on the vibration change value and the abnormal change coefficient.

[0015] Compared with the prior art, the beneficial effects of the present invention are that the technical solution of the present invention effectively reflects the degree of deviation of the current vibration state of the wind turbine from the historical normal benchmark and the degree of fluctuation relative to the previous normal window by using the vibration anomaly degree and vibration change coefficient. Then, based on the vibration anomaly degree and vibration change coefficient, the wind turbine is initially screened, which is conducive to quickly identifying abnormal units with significant deviation or severe fluctuation in vibration state at the global monitoring level, avoiding misjudgment and missed judgment caused by the traditional threshold early warning method due to ignoring changes in operating conditions or continuous interference from faults.

[0016] Furthermore, this invention effectively reflects the smoothness of the wind turbine's pitch action and the temporal coordination between vibration response and power output by using pitch disturbance value and energy state offset. This allows for the determination of whether the unit is in an unstable operating state. In the event of instability, the diagnostic method is automatically adjusted from vibration state diagnosis to sampling point disturbance diagnosis. This facilitates the dynamic switching of diagnostic granularity from the window level to the sampling point level, avoids the insufficient adaptability of fixed diagnostic modes under complex operating conditions, and improves the sensitivity and response time of early instability state capture.

[0017] Furthermore, in this invention, sampling points are classified by instability characteristic values, and anomaly is determined based on the proportion of a certain type of sampling point. This allows for the quantitative characterization of the relative degree of abnormality of vibration mutations relative to pitch angle and power mutations, achieving a precise transition from macroscopic anomaly screening to microscopic disturbance location, and overcoming the limitations of traditional methods in distinguishing between fault sources and normal disturbances.

[0018] Furthermore, this invention comprehensively captures the cumulative shift of the instantaneous phase difference between vibration and power and the instantaneous phase difference between vibration and pitch angle at anomaly points by accumulating phase difference. Based on the combined condition of this cumulative amount and the proportion of anomaly points, it decides whether to enter the characteristic segment analysis. This is beneficial for triggering fine-grained time series analysis when anomaly points are widely distributed but the phase shift is not significant, and for directly identifying anomalies when the phase shift is significant. This achieves an adaptive balance between computing resources and diagnostic accuracy, and avoids the waste of computing power caused by invalid analysis.

[0019] Furthermore, in this invention, the feature segment analysis module extracts feature segments that meet the requirements of feature point density or phase difference change, and further calculates the phase difference oscillation coefficient of abnormal points within the feature segment as the final basis for anomaly determination. This is beneficial for comprehensively assessing instability risk from two dimensions: the temporal aggregation degree of abnormal points and the intensity of phase difference change, effectively identifying oscillating and intermittent early faults, and significantly improving diagnostic accuracy. Attached Figure Description

[0020] Figure 1 This is a module connection diagram of the online monitoring and intelligent diagnosis system for wind turbine vibration status according to an embodiment of the present invention; Figure 2 This is a flowchart illustrating how to determine whether a wind turbine has an anomaly based on the vibration anomaly degree and vibration change coefficient, according to an embodiment of the present invention. Figure 3 This is a flowchart illustrating how the wind turbine is determined to be in an unstable operating state based on pitch disturbance value and operating stability, according to an embodiment of the present invention. Figure 4 This is a flowchart illustrating how an anomaly in a wind turbine is determined based on the proportion of a certain type of sampling point, or how to determine whether to perform feature segment analysis based on the cumulative phase difference, according to an embodiment of the present invention. Detailed Implementation

[0021] To make the objectives and advantages of the present invention clearer, the present invention will be further described below with reference to embodiments; it should be understood that the specific embodiments described herein are merely for explaining the present invention and are not intended to limit the present invention.

[0022] Preferred embodiments of the present invention will now be described with reference to the accompanying drawings. Those skilled in the art should understand that these embodiments are merely illustrative of the technical principles of the present invention and are not intended to limit the scope of protection of the present invention.

[0023] Please see Figure 1 The diagram shown is a module connection diagram of the online monitoring and intelligent diagnosis system for the vibration status of a wind turbine generator according to an embodiment of the present invention. The present invention provides an online monitoring and intelligent diagnosis system for the vibration status of a wind turbine generator, comprising: The data acquisition module is used to acquire monitoring information from the wind turbine. The first diagnostic module, which is connected to the data acquisition module, is used to perform a vibration state diagnosis for the target wind turbine, and in the vibration state diagnosis, it determines whether there is an abnormality in the wind turbine based on the vibration anomaly degree and the vibration change coefficient. The state determination module is connected to the data acquisition module and the first diagnosis module respectively. It is used to determine whether the wind turbine is in an unstable working state based on the pitch disturbance value and energy state offset. When it is in an unstable working state, the diagnosis method is changed from vibration state diagnosis to sampling point disturbance diagnosis. The second diagnostic module, which is connected to the state determination module, is used to determine the category of sampling points based on the instability characteristic value in the sampling point disturbance diagnosis, and to determine whether the wind turbine has an abnormality based on the proportion of a type of sampling point, or to determine whether to perform feature segment analysis based on the cumulative phase difference and the proportion of abnormal points. The feature segment analysis module, which is connected to the second diagnostic module, is used to determine the feature segment based on the density of feature points and the change in phase difference during feature segment analysis, and to determine whether there is an abnormality in the wind turbine based on the phase difference oscillation coefficient of each feature segment. The early warning output module is connected to the first diagnostic unit, the second diagnostic unit and the feature segment analysis unit respectively, and is used to send an early warning when it is determined that there is an abnormality in the target wind turbine.

[0024] In this embodiment of the invention, the monitoring information of the wind turbine is the monitoring parameters collected within a preset time window. The monitoring parameters include, but are not limited to, vibration acceleration, pitch angle, and power output. Vibration acceleration is detected by a piezoelectric accelerometer; the pitch angle is the angle between the airfoil chord at the tip of the wind turbine blade and the plane of rotation of the wind turbine rotor. The airfoil chord at the tip of the blade is the line connecting the leading edge and the trailing edge of the blade cross section. The pitch angle is detected by a pitch angle encoder installed on the pitch bearing; the power output is detected by a power transmitter installed at the output end of the wind turbine. In this embodiment of the invention, a fault diagnosis is performed once at the end of each preset time window. The duration of the preset time window and the sampling frequency of the monitoring parameters can be adaptively adjusted according to the characteristics of the monitored object and the frequency of fault characteristics to match the diagnostic needs of actual application scenarios. The duration of the preset time window and the sampling frequency of the monitoring parameters are not specifically limited in this embodiment; in actual applications, they can be flexibly configured according to system operating status, signal change rate, and computing resources to ensure the timeliness and accuracy of the diagnosis. The number of sampling points is dynamically determined according to the sampling frequency and the duration of the preset time window. Each sampling point corresponds to an instantaneous value of a monitoring parameter, used to constitute the monitoring data sequence within the current time window.

[0025] Please see Figure 2 The above is a flowchart of an embodiment of the present invention for determining whether a wind turbine has an anomaly based on the vibration anomaly degree and the vibration change coefficient.

[0026] Specifically, the first diagnostic module determines that wind turbines with a vibration anomaly degree greater than a preset vibration anomaly degree or a vibration change coefficient greater than a preset vibration change coefficient are abnormal.

[0027] The first diagnostic module determines that wind turbines with a vibration anomaly degree less than or equal to a preset vibration anomaly degree and a vibration change coefficient less than or equal to a preset vibration change coefficient do not have any anomalies.

[0028] In this embodiment, the first diagnostic module further includes a vibration anomaly calculation unit to determine the vibration anomaly degree. The vibration anomaly degree is used to determine the degree of deviation of the current vibration state of the wind turbine from the normal reference. The larger the vibration anomaly degree, the greater the deviation between the current vibration characteristic value and the reference vibration characteristic value, and the more abnormal the vibration state of the unit. In this embodiment, the vibration anomaly degree is determined based on the ratio of the absolute value of the characteristic difference to the reference vibration characteristic value. The absolute value of the characteristic difference is the absolute value of the difference between the vibration characteristic value and the reference vibration characteristic value. The vibration characteristic value is the root mean square of the vibration acceleration corresponding to each sampling point in a preset time window. The reference vibration characteristic value is the average value of the vibration characteristic values ​​of the unit in five consecutive historical time windows under normal operating conditions. In this embodiment, the first diagnostic module further includes a vibration change coefficient calculation unit to determine the vibration change coefficient. The vibration change coefficient is used to determine the degree of fluctuation of the current vibration state relative to the adjacent normal window. The larger the vibration change coefficient, the greater the deviation of the vibration characteristic value between the current window and the adjacent normal window, the more drastic the change in the unit's vibration state, and the more likely the wind turbine unit is to experience sudden or gradual vibration anomalies. In this embodiment, the vibration change coefficient is determined based on the ratio of the absolute value of the adjacent characteristic difference to the reference vibration characteristic value. The absolute value of the adjacent characteristic difference is the absolute value of the difference between the vibration characteristic value of the current preset time window and the vibration characteristic value of the adjacent window. The adjacent window is a preset time window that is adjacent to the current preset time window and located before the current preset time window. It should be noted that if the adjacent window determines that the wind turbine unit has an anomaly, the vibration characteristic value of the adjacent window is replaced with the vibration characteristic value of the most recent time window that was determined to be in a normal state before the abnormal window, so as to avoid the interference of the abnormal window on the calculation of the change coefficient and ensure that the vibration change coefficient can truly reflect the change in operating conditions rather than the deviation caused by the continuous fault.

[0029] In this embodiment of the invention, the degree of abnormal deviation and the intensity of fluctuation in the vibration state of the wind turbine are effectively reflected by the vibration anomaly degree and the vibration change coefficient. The greater the user's demand for early fault warning sensitivity, the smaller the values ​​of the preset vibration anomaly degree and the preset vibration change coefficient. In this embodiment, the values ​​of the preset vibration anomaly degree and the preset vibration change coefficient are determined based on the statistical results of historical operating conditions. Historical operating condition samples that meet the usage requirements and have been verified as abnormal are obtained. The average value of the vibration anomaly degree and the average value of the vibration change coefficient corresponding to the historical operating conditions are statistically calculated and denoted as the preset vibration anomaly degree and the preset vibration change coefficient, respectively. The historical operating conditions monitored that meet the usage requirements are the operating periods that have been manually verified and are consistent with the system's judgment results.

[0030] Please see Figure 3 As shown, it is a flowchart of an embodiment of the present invention for determining whether a wind turbine is in an unstable operating state based on pitch disturbance value and operating stability.

[0031] Specifically, the state determination module determines that a wind turbine is in an unstable operating state if the pitch disturbance value is greater than a preset pitch disturbance value or the energy state offset is greater than a preset energy state offset.

[0032] Specifically, the state determination module determines that wind turbines whose pitch disturbance value is less than or equal to a preset pitch disturbance value and whose energy state offset is less than or equal to a preset energy state offset are not in an unstable working state, and there is no need to change the diagnosis method from vibration state diagnosis to sampling point disturbance diagnosis.

[0033] In this embodiment, the state determination module further includes a disturbance value calculation unit to determine the pitch disturbance value. The pitch disturbance value is used to determine the severity of the wind turbine's pitch action. The larger the pitch disturbance value, the more frequent and drastic the pitch action of the wind turbine, indicating that the turbine is in an unstable pitch system or an unbalanced blade force state. In this case, the wind turbine is more likely to enter an unstable operating state. In this embodiment, the pitch disturbance value is determined based on the pitch angle corresponding to each sampling point, and the pitch disturbance value is the standard deviation of the pitch angle corresponding to each sampling point within a preset time window.

[0034] In this embodiment, the state determination module further includes an energy state offset calculation unit to determine the energy state offset. The energy state offset is used to determine the temporal synergy between the vibration and power of the wind turbine. The larger the energy state offset, the weaker the correlation between vibration acceleration and power generation, indicating that the mechanical energy transfer path of the unit is blocked or the energy conversion is abnormal, making the wind turbine more prone to entering an unstable operating state. In this embodiment, the energy state offset is determined based on the vibration acceleration sequence and the power generation sequence. The energy state offset is negatively correlated with the Pearson correlation coefficient of the vibration acceleration sequence and the power generation sequence. The energy state offset is the difference between 1 and the Pearson correlation coefficient of the vibration acceleration sequence and the power generation sequence. The vibration acceleration sequence is composed of the vibration acceleration values ​​of each sampling point within a preset time window, and the power generation sequence is composed of the power generation values ​​of each sampling point within a preset time window. The two sequences correspond one-to-one in time.

[0035] In this embodiment of the invention, the severity of pitch action is determined by the pitch disturbance coefficient, and the temporal synchronicity between vibration and power is determined by the energy state offset. Wind turbines with vibration anomalies exceeding a preset vibration anomaly or vibration change coefficients exceeding a preset vibration change coefficient are considered to have unstable factors in their pitch system or abnormal energy conversion paths, thus being determined to be in an unstable operating state. Therefore, the diagnostic method is changed from vibration state diagnosis to sampling point disturbance diagnosis to improve the sensitivity and diagnostic specificity for early instability faults. It can be understood that the larger the pitch disturbance value and energy state offset, the more severe the pitch action of the wind turbine and the more serious the temporal desynchronization between vibration and power, resulting in a higher risk of turbine instability. Therefore, the greater the requirement for early instability warning sensitivity for wind turbines, the smaller the preset pitch disturbance value and preset energy state offset. Specifically, wind turbine operation data segments covering different pitch change intensity levels and different vibration-power timing desynchronization levels are collected. Multiple sets of decreasing preset pitch disturbance thresholds and preset energy state offset thresholds are cross-combined to evaluate the accuracy of instability identification under each combination. The preset pitch disturbance threshold and preset energy state offset threshold corresponding to the highest identification accuracy are recorded as preset pitch disturbance and preset energy state offset, respectively.

[0036] Please continue reading. Figures 1 to 3 As shown in the embodiment of the present invention, the second diagnostic module determines that sampling points whose instability feature value is greater than or equal to the first preset instability feature value are a type of sampling point; The second diagnostic module determines that sampling points whose instability feature value is greater than or equal to the first preset instability feature value are classified as a type of sampling point; The second diagnostic module determines that sampling points whose instability feature value is less than the first preset instability feature value and greater than or equal to the second preset instability feature value are classified as Class II sampling points; The second diagnostic module determines that sampling points with an instability feature value less than the second preset instability feature value are classified as three types of sampling points; Among them, the first preset instability characteristic value is greater than the second preset instability characteristic value.

[0037] In this embodiment, the second diagnostic module further includes an instability characteristic value calculation unit to determine the instability characteristic value. The instability characteristic value is used to comprehensively characterize the instantaneous instability degree of the wind turbine at a single sampling point. The larger the instability characteristic value, the stronger the coupling between the vibration mutation and the pitch-power anomaly at the sampling point, and the higher the risk of the unit being in an unstable state at that moment. In this embodiment, the instability characteristic value is determined based on the vibration change value and the anomaly change coefficient. The instability characteristic value is positively correlated with the ratio of the vibration change value and the anomaly change coefficient. The specific method for determining the instability characteristic value is as follows: the vibration change value and the anomaly change coefficient corresponding to each sampling point within a preset time window are normalized using the minimum-maximum method. The instability characteristic value corresponding to a single sampling point is the ratio of the processed vibration change value to the processed anomaly change coefficient. Among them, the vibration change value corresponding to a single sampling point is the absolute value of the difference between the vibration acceleration of the sampling point and the vibration acceleration of the previous adjacent sampling point. It is used to characterize the instantaneous fluctuation amplitude of vibration acceleration between adjacent sampling points. The larger the vibration change value, the more violent the change in vibration acceleration of the wind turbine at the sampling point, indicating that the unit is subjected to instantaneous impact or has early fault signs. At this time, the sampling point is more likely to be identified as an abnormal point. The abnormal variation coefficient corresponding to a single sampling point is determined based on the pitch angle change rate and the power output, and is used to characterize the combined influence of the severity of pitch control and the degree of power fluctuation at that sampling point. The abnormal variation coefficient is positively correlated with both the pitch angle change rate and the power change rate. The pitch angle change rate corresponding to a single sampling point is determined by dividing the absolute value of the difference between the pitch angle of that sampling point and the previous adjacent sampling point by the average value of the pitch angles of all sampling points within a preset time window. The power change rate corresponding to a single sampling point is determined by dividing the absolute value of the difference between the power output of that sampling point and the previous adjacent sampling point by the average value of the power output of all sampling points within a preset time window. In this embodiment of the invention, the abnormal variation coefficient is obtained by adding the pitch angle change rate and the power output.

[0038] It should be noted that for the first sampling point, since there is no previous sampling point, its vibration change value and abnormal change coefficient can be set to 0 or directly discarded. This embodiment of the invention does not make specific limitations on this.

[0039] In this embodiment of the invention, the degree of instability of the wind turbine's operating state at a single sampling point is determined by the instability feature value. The larger the instability feature value, the stronger the coupling between the sudden vibration and the abnormal pitch-power change at that sampling point, and the higher the risk of the turbine being in an unstable state at that moment. Therefore, the greater the need for early instability fault monitoring sensitivity of the wind turbine, the smaller the values ​​of the first preset instability feature value and the second preset instability feature value. In this embodiment, the first preset instability feature value and the second preset instability feature value are determined by simulation. The acquisition process includes: setting multiple sets of candidate first preset instability feature values ​​and candidate second preset instability feature values; simulating different instability scenarios under each set of candidate value combinations and executing the corresponding sampling point classification judgment and instability early warning process; obtaining the instability state recognition accuracy corresponding to each set of thresholds; and recording the candidate first preset instability feature value and candidate second preset instability feature value corresponding to the highest recognition accuracy as the first preset instability feature value and the second preset instability feature value, respectively.

[0040] Please see Figure 4 As shown, it is a flowchart of an embodiment of the present invention for determining whether a wind turbine has an anomaly based on the proportion of a type of sampling point or for determining whether to perform feature segment analysis based on the cumulative phase difference.

[0041] Specifically, the second diagnostic module determines that the wind turbine has an abnormality when the proportion of a certain type of sampling point is greater than or equal to the proportion of a preset type of sampling point.

[0042] In this embodiment, the second diagnostic module further includes a sampling point proportion calculation unit to determine the proportion of a certain type of sampling point. This proportion characterizes the density of high-risk sampling points within a preset time window. A larger proportion indicates more sampling points indicating that the wind turbine is in an unstable state within that time window, resulting in a higher risk of continuous instability. In this case, the wind turbine is more likely to exhibit abnormal states requiring early warning. In this embodiment, the proportion of a certain type of sampling point is determined based on the number of such sampling points within the preset time window, which is the ratio of the number of such sampling points to the total number of sampling points within the preset time window.

[0043] The preset value of the first-class sampling point percentage can be determined based on the user's sensitivity requirements for wind turbine fault monitoring. This percentage effectively reflects the severity of the turbine's instability within a preset time window. The greater the user's need for early fault warning sensitivity, the smaller the preset first-class sampling point percentage should be. In this embodiment, the preset first-class sampling point percentage is determined through simulation optimization: multiple candidate preset first-class sampling point percentages are set, and different instability scenarios are simulated under each candidate value, with sampling point classification and instability warning judgment processes executed. The candidate value with the highest instability state identification accuracy is then used as the preset first-class sampling point percentage.

[0044] Specifically, the second diagnostic module responds to the condition that the proportion of a certain type of sampling point is less than the preset proportion of a certain type of sampling point, and determines whether to perform feature segment analysis based on the cumulative phase difference and the proportion of abnormal points.

[0045] Please continue reading. Figures 1 to 4 As shown, in this embodiment of the invention, the second diagnostic module performs feature segment analysis on wind turbines whose cumulative phase difference is less than a preset cumulative phase difference and whose percentage of abnormal points is greater than or equal to a preset percentage of abnormal points. The second diagnostic module does not need to perform feature segment analysis on wind turbines whose cumulative phase difference is greater than or equal to the preset cumulative phase difference, and determines that the wind turbine has an anomaly. The second diagnostic module does not need to perform feature segment analysis on wind turbines whose cumulative phase difference is less than the preset cumulative phase difference and whose percentage of abnormal points is less than the preset percentage of abnormal points, and determines that the wind turbine does not have any abnormalities.

[0046] In this embodiment, the second diagnostic module further includes a phase difference accumulation calculation unit to determine the phase difference accumulation. The phase difference accumulation is used to characterize the degree of temporal offset at abnormal sampling points. The larger the phase difference accumulation, the more severe the temporal synchronization disruption between the mechanical vibration response and power output or between the pitch adjustment action and vibration response of the wind turbine at the abnormal point. This increases the risk of early faults such as drivetrain wear, pitch jamming, or blade aerodynamic instability. In this embodiment, the phase difference accumulation is determined based on the absolute value of the first instantaneous phase difference and the absolute value of the second instantaneous phase difference corresponding to the abnormal point. The phase difference accumulation is the larger of the first accumulation and the second accumulation. The first cumulative amount is the sum of the absolute values ​​of the first instantaneous phase differences corresponding to each anomaly point after angle normalization, and the second cumulative amount is the sum of the absolute values ​​of the second instantaneous phase differences corresponding to each anomaly point after angle normalization. Anomalies include Class I sampling points and Class II sampling points. For each sampling point, the instantaneous phase of each monitoring parameter is constructed. For any monitoring parameter, the instantaneous phase of a single sampling point is obtained by dividing the monitoring parameter value corresponding to the sampling point by the same monitoring parameter value corresponding to the previous adjacent sampling point and taking the arctangent function of the resulting ratio. The first instantaneous phase difference corresponding to a single sampling point is the difference between the instantaneous phase of the vibration acceleration of the sampling point and the previous adjacent sampling point, and the second instantaneous phase difference corresponding to a single sampling point is the difference between the instantaneous phase of the pitch angle of the sampling point and the previous adjacent sampling point. The first instantaneous phase difference or the second instantaneous phase difference are respectively subjected to angle normalization processing, and their range is limited to between -π and +π to avoid phase jumps. The specific method is as follows: First, calculate the initial instantaneous phase difference, then divide it by 2π, round the result to the nearest integer, then multiply the rounded value by 2π, and finally subtract the product from the initial instantaneous phase difference to obtain the normalized instantaneous phase difference.

[0047] Understandably, only outliers are included when calculating the cumulative phase difference. This is because outliers directly reflect the unit's instability characteristics, while the instantaneous phase differences corresponding to the three types of sampling points (normal points) primarily reflect inherent, stable time-series lag relationships. Accumulating these outliers together would create strong background noise, diluting or even drowning out the weak but crucial phase shift signals at the outliers. Calculating the cumulative phase difference only for these outliers effectively eliminates irrelevant phase fluctuations caused by environmental factors such as wind speed fluctuations and short-term grid disturbances under normal operating conditions. This ensures the cumulative difference remains highly sensitive to early faults such as drivetrain wear and pitch jamming, while significantly reducing the computational burden and guaranteeing the engineering feasibility of real-time monitoring.

[0048] Understandably, during normal operation of a wind turbine, wind energy is converted into mechanical energy through the blades, and then transmitted to the generator through the transmission chain to be converted into electrical energy. Vibration acceleration serves as the mechanical response signal of the unit, and its changes lag behind the changes in the power output. The two are well synchronized in time, so the first instantaneous phase difference is within a stable small range, and the corresponding first cumulative value is small. When the unit experiences early faults such as transmission chain wear, blade aerodynamic instability, or abnormal pitch adjustment, the mechanical energy transmission path is blocked or the energy conversion efficiency is abnormal, which disrupts the synchronization between mechanical vibration and power output. The first instantaneous phase difference increases significantly, causing the first cumulative value to rise sharply, which can effectively detect such potential faults related to energy conversion. Meanwhile, as a core control parameter of wind turbines, the pitch angle directly affects the stress state of the blades and thus the vibration response of the turbine. The degree of timing matching between the two directly reflects the synergy between the pitch system and the mechanical system: during normal operation, the pitch adjustment action and the vibration response are synchronized and the second instantaneous phase difference is stable; when the pitch system is stuck, the blades are aerodynamically unstable, or the transmission chain is worn prematurely, the pitch adjustment action cannot respond to the changes in blade stress in time, resulting in a significant phase misalignment between the pitch adjustment and the vibration response. The second instantaneous phase difference continues to shift and its absolute value increases, and the corresponding second cumulative amount increases significantly. By taking the maximum value of the two cumulative values ​​as the phase difference cumulative value, the two core fault risks of energy conversion anomaly and pitch coordination anomaly can be captured in a comprehensive manner, so as to achieve specific identification of early faults of the wind turbine body, effectively avoid misjudgment caused by environmental interference, and improve the accuracy of fault diagnosis and anti-interference capability.

[0049] In this embodiment, the second diagnostic module further includes an anomaly point percentage calculation unit to determine the anomaly point percentage. The anomaly point percentage characterizes the density of anomaly point distribution within a preset time window. A higher anomaly point percentage indicates that more sampling points of the wind turbine are in an unstable or sub-unstable state within that time window, and the higher the risk of continuous instability of the unit. In this case, the wind turbine is more likely to have progressive fault hazards that require early warning. The anomaly point percentage is the ratio of the number of anomalies in the preset time window to the total number of sampling points in the preset time window.

[0050] The values ​​of the preset phase difference accumulation and the preset anomaly point ratio can be determined according to the user's requirements for monitoring the early instability faults of wind turbine units. The phase difference accumulation and the anomaly point ratio effectively reflect the temporal offset of abnormal sampling points and their overall distribution ratio, comprehensively characterizing the evolution risk of potential progressive faults of the unit. The greater the user's requirement for early fault warning sensitivity, the smaller the values ​​of the preset phase difference accumulation and the preset anomaly point ratio. In this embodiment, the preset phase difference accumulation and the preset anomaly point ratio are determined through simulation optimization: multiple sets of candidate preset phase difference accumulation and candidate preset anomaly point ratios are set. Under each set of candidate value combinations, different progressive instability scenarios are simulated and the feature segment analysis and judgment process is executed to obtain the fault identification accuracy corresponding to each set of thresholds. The candidate preset phase difference accumulation and candidate preset anomaly point ratio corresponding to the highest identification accuracy are recorded as the preset phase difference accumulation and the preset anomaly point ratio, respectively.

[0051] Specifically, the feature paragraph analysis module determines that among any paragraph with a preset time length within a preset time window, the paragraph that satisfies the condition that the feature point density is greater than the preset feature point density or the phase difference change is greater than the preset phase difference change is a feature paragraph.

[0052] The preset time length can be adjusted according to actual monitoring needs; the higher the required diagnostic precision, the smaller the preset time length. In this embodiment, the preset time length is a portion of the preset time window duration, meaning the preset time length dynamically adapts to changes in the preset time window duration to ensure the applicability of feature segment analysis across different monitoring time scales. It should be noted that the proportional relationship between the preset time length and the preset time window duration is not specifically limited in this embodiment; in practical applications, it can be adaptively adjusted based on diagnostic resolution requirements and computing resource availability.

[0053] The feature point density of a single paragraph with a preset time length is the ratio of the number of outliers in that paragraph to the number of outliers in the preset time window. In this embodiment, the feature paragraph analysis module further includes a phase difference change degree calculation unit to determine the phase difference change degree. The phase difference change degree characterizes the drastic change in the cumulative instantaneous phase difference of anomalies between adjacent paragraphs. The larger the phase difference change degree, the more significant the temporal shift abrupt change in the paragraph compared to the preceding paragraph, and the easier it is to identify the paragraph as a feature paragraph. In this embodiment, for a single paragraph with a preset time length, its corresponding phase difference change degree is the larger of the first change degree and the second change degree. The first change degree is the difference between the cumulative first change degree of the paragraph corresponding to the paragraph and the cumulative first change degree of the preceding paragraph corresponding to the paragraph. The cumulative first change degree of the paragraph corresponding to a single paragraph is the sum of the absolute values ​​of the angle-normalized first instantaneous phase differences corresponding to each anomaly point in the paragraph. The preceding paragraph corresponding to the paragraph is a paragraph with a preset time length and whose end time is the start time of the paragraph. The confirmation method of the second change degree is the same as that of the first change degree, except that the first instantaneous phase difference is replaced by the second instantaneous phase difference, which will not be described in detail here.

[0054] The values ​​of preset feature point density and preset phase difference change can be determined based on the user's sensitivity requirements for identifying early fault characteristic segments of wind turbines. The feature point density and phase difference change effectively reflect the degree of aggregation of abnormal points and the drastic temporal change of instantaneous phase difference within a specific time period. The greater the sensitivity requirement for early characteristic segment identification, the smaller the values ​​of preset feature point density and preset phase difference change. In this embodiment, the preset feature point density and preset phase difference change are determined through simulation optimization: multiple sets of candidate preset feature point density and candidate preset phase difference change are set, and under each set of candidate value combinations, different instability scenarios are simulated and the characteristic segment identification process is executed. The characteristic segment identification accuracy corresponding to each set of thresholds is obtained, and the candidate preset feature point density and candidate preset phase difference change corresponding to the highest identification accuracy are recorded as preset feature point density and preset phase difference change, respectively.

[0055] Specifically, the feature segment analysis module determines that wind turbines with a phase difference oscillation coefficient greater than or equal to a preset phase difference oscillation coefficient are abnormal.

[0056] In this embodiment, the feature segment analysis module further includes a phase difference oscillation coefficient calculation unit to determine the phase difference oscillation coefficient. The phase difference oscillation coefficient characterizes the relative volatility of the instantaneous phase difference at anomaly points within the feature segment. A larger phase difference oscillation coefficient indicates a higher degree of dispersion and more volatile instantaneous phase difference sequence within the feature segment, and a higher risk of the unit being in an anomaly. In this embodiment, the phase difference oscillation coefficient is the larger of a first coefficient of variation and a second coefficient of variation. The first coefficient of variation is the ratio of the standard deviation of the sequence formed by the absolute values ​​of the angle-normalized first instantaneous phase differences corresponding to each anomaly point within the feature segment to the average value of that sequence. The second coefficient of variation is the ratio of the standard deviation of the sequence formed by the absolute values ​​of the angle-normalized second instantaneous phase differences corresponding to each anomaly point within the feature segment to the average value of that sequence.

[0057] The value of the preset phase difference oscillation coefficient can be determined according to the user's monitoring sensitivity requirements for the timing coordination stability of wind turbine units. The phase difference oscillation coefficient effectively reflects the relative volatility of the instantaneous phase difference within the characteristic segment. The greater the requirement for early timing instability warning sensitivity, the smaller the value of the preset phase difference oscillation coefficient. In this embodiment, the preset phase difference oscillation coefficient is determined by simulation optimization: multiple sets of candidate preset phase difference oscillation coefficients are set, and instability scenarios with different oscillation degrees are simulated under each candidate value. The characteristic segment analysis and judgment process is executed to obtain the timing instability state recognition accuracy corresponding to each set of thresholds. The candidate preset phase difference oscillation coefficient corresponding to the highest recognition accuracy is recorded as the preset phase difference oscillation coefficient.

[0058] It is understandable that when the cumulative phase difference is less than the preset cumulative phase difference and the proportion of abnormal points is greater than or equal to the preset proportion of abnormal points, it indicates that there are many abnormal points in the current time window, but the instantaneous phase difference offset corresponding to each abnormal point has not yet reached the independent significance threshold. The global cumulative amount is difficult to effectively characterize the risk of instability accumulation in local time periods. Therefore, characteristic segment analysis is carried out. When the phase difference oscillation coefficient is greater than or equal to the preset phase difference oscillation coefficient, it indicates that the absolute value sequence of instantaneous phase difference corresponding to the abnormal points in the characteristic segment has a high degree of dispersion and violent fluctuations, and the unit is in an oscillating unstable working state.

[0059] Specifically, the instability characteristic value is determined based on the vibration change value and the abnormal change coefficient.

[0060] The technical solution of the present invention has been described above with reference to the preferred embodiments shown in the accompanying drawings. However, it will be readily understood by those skilled in the art that the scope of protection of the present invention is obviously not limited to these specific embodiments. Without departing from the principles of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will all fall within the scope of protection of the present invention.

Claims

1. An online monitoring and intelligent diagnosis system for the vibration status of a wind turbine generator, characterized in that, include: The data acquisition module is used to acquire monitoring information from the wind turbine. The first diagnostic module, which is connected to the data acquisition module, is used to perform a vibration state diagnosis for the target wind turbine, and in the vibration state diagnosis, it determines whether there is an abnormality in the wind turbine based on the vibration anomaly degree and the vibration change coefficient. The state determination module is connected to the data acquisition module and the first diagnosis module respectively. It is used to determine whether the wind turbine is in an unstable working state based on the pitch disturbance value and energy state offset. When it is in an unstable working state, the diagnosis method is changed from vibration state diagnosis to sampling point disturbance diagnosis. The second diagnostic module, which is connected to the state determination module, is used to determine the category of sampling points based on the instability characteristic value in the sampling point disturbance diagnosis, and to determine whether the wind turbine has an abnormality based on the proportion of a type of sampling point, or to determine whether to perform feature segment analysis based on the cumulative phase difference and the proportion of abnormal points. The feature segment analysis module, which is connected to the second diagnostic module, is used to determine the feature segment based on the density of feature points and the change in phase difference during feature segment analysis, and to determine whether there is an abnormality in the wind turbine based on the phase difference oscillation coefficient of each feature segment. The early warning output module is connected to the first diagnostic unit, the second diagnostic unit and the feature segment analysis unit respectively, and is used to send an early warning when it is determined that there is an abnormality in the target wind turbine.

2. The online monitoring and intelligent diagnosis system for the vibration state of wind turbine units according to claim 1, characterized in that, The first diagnostic module determines that wind turbines with vibration anomaly levels greater than preset vibration anomaly levels or vibration change coefficients greater than preset vibration change coefficients are abnormal.

3. The online monitoring and intelligent diagnosis system for the vibration state of wind turbine units according to claim 2, characterized in that, The state determination module determines that a wind turbine is in an unstable operating state if the pitch disturbance value is greater than a preset pitch disturbance value or the energy state offset is greater than a preset energy state offset.

4. The online monitoring and intelligent diagnosis system for the vibration status of wind turbine units according to claim 3, characterized in that, The second diagnostic module determines that sampling points whose instability feature value is greater than or equal to the first preset instability feature value are classified as a type of sampling point; The second diagnostic module determines that sampling points whose instability feature value is greater than or equal to the first preset instability feature value are classified as a type of sampling point; The second diagnostic module determines that sampling points whose instability feature value is less than the first preset instability feature value and greater than or equal to the second preset instability feature value are classified as Class II sampling points; The second diagnostic module determines that sampling points with an instability feature value less than the second preset instability feature value are classified as three types of sampling points; Among them, the first preset instability characteristic value is greater than the second preset instability characteristic value.

5. The online monitoring and intelligent diagnosis system for the vibration state of wind turbine units according to claim 4, characterized in that, The second diagnostic module determines that the wind turbine has an abnormality if the proportion of a certain type of sampling point is greater than or equal to the preset proportion of a certain type of sampling point.

6. The online monitoring and intelligent diagnosis system for the vibration state of wind turbine units according to claim 5, characterized in that, The second diagnostic module responds to the condition that the proportion of a certain type of sampling point is less than the preset proportion of a certain type of sampling point, and determines whether to perform feature segment analysis based on the cumulative phase difference and the proportion of abnormal points.

7. The online monitoring and intelligent diagnosis system for the vibration state of wind turbine units according to claim 6, characterized in that, The second diagnostic module performs characteristic segment analysis on wind turbines where the cumulative phase difference is less than the preset cumulative phase difference and the percentage of abnormal points is greater than or equal to the preset percentage of abnormal points. The second diagnostic module does not need to perform feature segment analysis on wind turbines whose cumulative phase difference is greater than or equal to the preset cumulative phase difference, and determines that the wind turbine has an anomaly. The second diagnostic module does not need to perform feature segment analysis on wind turbines whose cumulative phase difference is less than the preset cumulative phase difference and whose percentage of abnormal points is less than the preset percentage of abnormal points, and determines that the wind turbine does not have any abnormalities.

8. The online monitoring and intelligent diagnosis system for the vibration state of wind turbine units according to claim 7, characterized in that, The feature segment analysis module determines that among any segment with a preset time length within a preset time window, the segment that satisfies either a feature point density greater than a preset feature point density or a phase difference change greater than a preset phase difference change is a feature segment.

9. The online monitoring and intelligent diagnosis system for the vibration state of wind turbine units according to claim 8, characterized in that, The feature segment analysis module determines that wind turbines with a phase difference oscillation coefficient greater than or equal to a preset phase difference oscillation coefficient are abnormal.

10. The online monitoring and intelligent diagnosis system for the vibration state of wind turbine units according to claim 4, characterized in that, The instability characteristic value is determined based on the vibration change value and the abnormal change coefficient.

Images