A method for monitoring abnormal state of wind turbine bearing

Abstract

The present invention relates to the technical field of bearing abnormal state monitoring, and discloses a method for monitoring the abnormal state of a wind turbine bearing, including: constructing a wind turbine monitoring model based on historical operation data of the wind turbine; setting a monitoring time node of the wind turbine, obtaining the operation data of the wind turbine at the current monitoring time node, generating the abnormal type of the wind turbine bearing through the wind turbine monitoring model; and generating a maintenance strategy based on the abnormal type of the engine bearing. The monitoring model of the present invention can be adaptively adjusted according to the actual operation of the wind turbine and environmental changes to ensure the accuracy and reliability of monitoring; according to the urgency of the abnormal state of the bearing, the system can automatically generate a corresponding maintenance strategy, from simple adjustment of the monitoring cycle to emergency shutdown and maintenance, realizing the intelligentization of maintenance decision-making.

Description

Technical Field

[0001] The invention relates to the technical field of bearing abnormal state monitoring, and in particular to a method for monitoring the abnormal state of a bearing of a wind turbine generator set. Background Art

[0002] In recent years, with the advancement of sensor technology, data acquisition technology, and data analysis methods, maintenance strategies based on condition monitoring have gradually become the mainstream of wind turbine maintenance. This strategy monitors the operating status of wind turbines in real time and uses data analysis technology to predict and diagnose potential faults, thereby achieving more accurate maintenance plans and resource allocation. In terms of bearing abnormal condition monitoring, researchers have developed a variety of technologies and methods. These methods usually involve monitoring and analyzing multiple parameters such as bearing vibration signals, temperature, and oil.

[0003] There are many types of faults in wind turbine bearings, including fatigue spalling, deformation, indentation and local corrosion. The characteristic signals of each fault type may overlap, increasing the difficulty of fault diagnosis. In addition, early fault signals are often weak and difficult to be identified by traditional methods. Summary of the invention

[0004] The purpose of the present invention is to integrate multiple sensor data, such as vibration, temperature, oil, etc., taking into account the complexity of the operating environment of wind turbines to provide a more comprehensive bearing status assessment; to collect and process data in real time, to promptly detect abnormal bearing conditions, prevent failures and reduce downtime.

[0005] In order to achieve the above object, the present invention provides a method for monitoring abnormal state of a wind turbine bearing, comprising:

[0006] Based on the historical operation data of wind turbines, a wind turbine monitoring model is constructed;

[0007] Set the monitoring time node of the wind turbine, obtain the operation data of the wind turbine at the current monitoring time node, and generate the abnormality type of the wind turbine bearing through the wind turbine monitoring model;

[0008] Generate maintenance strategies based on the anomaly types of engine bearings.

[0009] Wherein, the construction of the wind turbine monitoring model includes:

[0010] Obtain historical operation data of wind turbines;

[0011] and dividing the historical operation data of the wind turbine generator set into first characteristic data and second characteristic data;

[0012] Constructing a first monitoring model for the wind turbine bearing based on the first characteristic data;

[0013] Generate an abnormality type pool of wind turbine bearings through a first monitoring model, wherein the abnormality type pool of wind turbine bearings includes abnormalities of multiple types of wind turbine bearings;

[0014] A type selection model is constructed based on the second characteristic data, and the abnormal type pool of the wind turbine bearing is screened through the type selection model to generate the abnormal type of the wind turbine.

[0015] In some embodiments of the present invention, the step of constructing a first monitoring model for a wind turbine bearing based on the first characteristic data includes:

[0016] The first characteristic data is a vibration signal;

[0017] Get the historical vibration signal D of the current wind turbine, D = [d 1 ,d 2 …d i …d n ];

[0018] Among them, d i is the acceleration of the vibration signal at the i-th sampling moment, and n is the total number of sampling moments of the vibration signal;

[0019] Perform equal gradient division on the historical vibration signal D, generate the acceleration evaluation value of the vibration signal, and obtain the peak value of the acceleration evaluation value;

[0020] and determine the frequency, amplitude and duration of the corresponding peak;

[0021] The frequency, amplitude and duration of the peak are standardized to obtain the frequency evaluation value f in the same value range. 1i , amplitude evaluation value f 2i And the duration evaluation value f3 i ;

[0022] Based on the frequency evaluation value f 1i , amplitude evaluation value f 2i And the duration evaluation value f 3i Generate an abnormal state evaluation value F;

[0023] Classify abnormal types of wind turbines based on historical abnormal data of wind turbines;

[0024] The abnormality type of the wind turbine is determined based on the abnormal state evaluation value F, and an abnormality type pool of the wind turbine bearing is generated.

[0025] In some embodiments of the present invention, the generating of the abnormal state evaluation value F includes:

[0026] Based on the gradient of wind turbine operating load, a multi-gradient load E is generated, where E = [E 1 ,E2 …E i …E k ];

[0027] E i represents the i-th gradient load, k represents the number of gradient loads;

[0028] Calculate the average value of multiple gradient loads E, and select the gradient load closest to the average value as the average gradient load Ea;

[0029] Calculate the frequency evaluation value f of the average gradient load Ea 1i , amplitude evaluation value f 2i and duration evaluation value f 3i The correlation between the abnormal state and the correlation coefficients r1, r2, and r3 under gradient load were obtained;

[0030]

[0031] Among them, w 1 Indicates the frequency evaluation value f 1i The initial weight, w 2 Indicates the amplitude evaluation value f 2i The initial weight, w 3 Indicates the duration evaluation value f 3i The initial weight of

[0032] The initial weight is adjusted according to the gradient load of the engine to generate the abnormal state evaluation value F.

[0033] In some embodiments of the present invention, the initial weight is adjusted according to the gradient load of the engine, including:

[0034] Get the i-th gradient load Ei of the engine and calculate the frequency evaluation value f of the i-th gradient load Ei 1i , amplitude evaluation value f 2i and duration evaluation value f 3i The correlation between the abnormal state and the gradient load is obtained. 1i 、r 2i 、r 3i ;

[0035]

[0036] Among them, a 1i represents the frequency evaluation value f of the i-th gradient load 1i The adjustment factor of 2i represents the magnitude evaluation value f of the i-th gradient load 2i The adjustment factor of 3i represents the duration evaluation value f of the i-th gradient load3i The adjustment factor of

[0037] Generate abnormal state evaluation value F, F = w 1 *a 1i *f 1i +w 2 *a 2i *f 2i +w 3 *a 3i *f 3i .

[0038] In some embodiments of the present invention, the step of constructing a type selection model based on the second feature data includes:

[0039] Based on historical data, obtain the abnormal state evaluation value F when all abnormal types occur and the second feature data change set R at the same time, R = [T, V, Q, S];

[0040] Among them, T represents the temperature change data set, V represents the noise data set, Q represents the bearing deformation data set, and S represents the lubricating oil pollution data set;

[0041] The data change degrees of the temperature change data set T, the noise data set V, the bearing deformation data set Q, and the lubricating oil pollution data set S are respectively divided into thresholds to generate data change scores for the temperature change data set T, the noise data set V, the bearing deformation data set Q, and the lubricating oil pollution data set S;

[0042] Based on the data change scores of the temperature change data set T, the noise data set V, the bearing deformation data set Q, and the lubricating oil pollution data set S, an abnormal type in the abnormal type pool of the wind turbine bearing corresponding to the abnormal state evaluation value F is generated.

[0043] In some embodiments of the present invention, the generating of the wind turbine bearing abnormality type by the wind turbine monitoring model includes:

[0044] Acquire a second characteristic data set at a current monitoring time node and data on changes in the surrounding environment of the wind turbine generator set, and correct the second characteristic data set based on the data on changes in the surrounding environment of the wind turbine generator set to generate a corrected second characteristic data set;

[0045] Generate an abnormal type pool of the current wind turbine bearing through the first monitoring model;

[0046] Based on the abnormal type pool of the current wind turbine bearing, the abnormal type of the current wind turbine bearing is generated through the type selection model combined with the corrected second feature data set.

[0047] In some embodiments of the present invention, the generation of a maintenance strategy based on the abnormality type of the engine bearing includes:

[0048] Generate the urgency of abnormality processing based on the abnormality type of the current wind turbine bearing and the second feature set data;

[0049] Develop maintenance strategies based on the urgency of the exception.

[0050] In some embodiments of the present invention, the generating of the urgency of the exception handling includes:

[0051] Building an exception handling urgency evaluation model based on the second feature set data;

[0052] The emergency evaluation value θ of exception handling is generated by the exception handling urgency evaluation model. The emergency evaluation value θ of exception handling is compared with the preset value (θ 1 ,θ 2 ) to generate the urgency of exception handling, where θ 1 is the first preset value, θ 2 is a second preset value;

[0053] If θ 2 <θ, then the current exception type is level one emergency;

[0054] If θ 1 ≤θ≤θ 2 , then the current exception type is level 2 emergency;

[0055] If θ<θ 1 , the current exception type is level 3 emergency.

[0056] In some embodiments of the present invention, the maintenance strategy is formulated based on the abnormal urgency, including:

[0057] If the current abnormality type is level 1 emergency, a shutdown instruction is generated and maintenance is performed;

[0058] If the current abnormality type is level 2 emergency, the maintenance time is set according to the unit's operation plan, and the monitoring cycle of the unit's operating status is adjusted within a period of time after the unit is restarted;

[0059] If the current abnormality type is level 3 emergency, the unit will continue to operate and the monitoring cycle of the unit operating status will be adjusted. Based on the change in the emergency evaluation value of the abnormality processing, it is determined whether to adjust the urgency of the current abnormality type.

[0060] Compared with the prior art, the method for monitoring abnormal state of a wind turbine bearing according to the embodiment of the present invention has the following beneficial effects:

[0061] By dividing the historical operation data into the first characteristic data and the second characteristic data and constructing the corresponding monitoring model, the operation status of the wind turbine can be analyzed from multiple angles. This method can more accurately detect abnormal conditions of the bearings and reduce the possibility of false alarms and missed alarms.

[0062] By considering the characteristics of multiple dimensions such as the frequency, amplitude and duration of peak and valley values, the operating status of the bearing can be evaluated more comprehensively, avoiding misjudgment of a single factor. By analyzing historical abnormal data and establishing an abnormality type pool containing multiple abnormal types, a more comprehensive reference can be provided for subsequent fault diagnosis and maintenance.

[0063] Determining the initial weights w1, w2, and w3 by the proportion of the correlation coefficients of the vibration characteristics can enable the abnormal state evaluation value F to more accurately reflect the actual abnormal condition of the bearing.

[0064] The initial weight is adjusted according to the gradient load of the engine, so that the abnormal state evaluation value F can dynamically adapt to the change of the operating load of the wind turbine, and can ensure that under different load conditions, the abnormal state evaluation value F can accurately reflect the operating state of the bearing.

[0065] By acquiring data from multiple sources, it is possible to more comprehensively consider the factors that affect the operating status of wind turbine bearings. The addition of factors such as temperature, noise, bearing deformation and lubricant contamination can provide richer information and make the analysis of bearing abnormalities more comprehensive.

[0066] By monitoring these environmental factors and correcting the second characteristic data set, the actual status of the bearing can be reflected more accurately, ensuring that the accuracy of the data can be guaranteed under different environmental conditions, thereby providing a reliable data basis for subsequent accurate judgment of the abnormal type of wind turbine bearings.

[0067] By combining the type selection model with the corrected second feature data set to generate the abnormal type of the current wind turbine bearing, the problem can be located more quickly, which helps to take corresponding maintenance measures in time, reduce the downtime of the wind turbine, and improve its operating efficiency.

[0068] The second feature set data reflects the current actual operating status of the bearing. Building an evaluation model based on these data can accurately evaluate the urgency of abnormal handling. Quantifying the urgency of abnormal handling helps to classify abnormal situations more accurately, avoids the subjectivity of human judgment, and improves the accuracy and reliability of the evaluation. BRIEF DESCRIPTION OF THE DRAWINGS

[0069] Figure 1 The present invention provides a flow chart of a method for monitoring abnormal state of a wind turbine bearing according to an embodiment of the present invention. DETAILED DESCRIPTION

[0070] The specific implementation of the present invention is further described in detail below in conjunction with the accompanying drawings and examples. The following examples are used to illustrate the present invention, but are not intended to limit the scope of the present invention.

[0071] In the description of the present invention, it should be understood that the terms "center", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside", etc., indicating the orientation or position relationship are based on the orientation or position relationship shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation on the present invention.

[0072] The terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the features. In the description of the present invention, unless otherwise specified, "plurality" means two or more.

[0073] In the description of the present invention, it should be noted that, unless otherwise clearly specified and limited, the terms "installed", "connected", and "connected" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal communication of two components. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0074] Embodiment 1:

[0075] A preferred method for monitoring abnormal state of a wind turbine bearing in an embodiment of the present invention is as follows: Figure 1 As shown, including:

[0076] Based on the historical operation data of wind turbines, a wind turbine monitoring model is constructed;

[0077] Set the monitoring time node of the wind turbine, obtain the operation data of the wind turbine at the current monitoring time node, and generate the abnormality type of the wind turbine bearing through the wind turbine monitoring model;

[0078] Generate maintenance strategies based on the anomaly types of engine bearings.

[0079] In this embodiment, a model capable of identifying the type of bearing abnormality is constructed by processing historical operation data of the wind turbine generator set.

[0080] The model building process includes data acquisition, data division, and building sub-models based on different data (the first monitoring model and the type selection model), and finally generates a comprehensive model that can be used to monitor the current abnormal type of wind turbine bearings.

[0081] Reasonable monitoring time nodes are set to obtain the operation data of wind turbines at each node. These data will be input into the established wind turbine monitoring model to obtain the abnormal type of wind turbine bearings.

[0082] According to the identified engine bearing abnormality type, the relevant second feature set data is considered to determine the urgency of abnormality handling, and then formulate a corresponding maintenance strategy to ensure the normal operation of the wind turbine.

[0083] The construction of the wind turbine monitoring model includes:

[0084] Obtain historical operation data of wind turbines;

[0085] and dividing the historical operation data of the wind turbine generator set into first characteristic data and second characteristic data;

[0086] Constructing a first monitoring model for the wind turbine bearing based on the first characteristic data;

[0087] Generate an abnormality type pool of wind turbine bearings through a first monitoring model, wherein the abnormality type pool of wind turbine bearings includes abnormalities of multiple types of wind turbine bearings;

[0088] A type selection model is constructed based on the second characteristic data, and the abnormal type pool of the wind turbine bearing is screened through the type selection model to generate the abnormal type of the wind turbine.

[0089] In this embodiment, building a wind turbine monitoring model is the basis of the entire monitoring method. By processing the historical operation data of the wind turbine, a model capable of identifying the type of bearing abnormality is built.

[0090] Analysis of factors affecting wind turbine bearings:

[0091] Abnormal vibration signal: When the spindle bearing fails, the vibration signal will change significantly. For example, the vibration amplitude may increase, the vibration frequency may shift, and the vibration waveform may be distorted.

[0092] Temperature rise: When a bearing fails, the bearing temperature may rise due to increased friction and other reasons. Therefore, monitoring the temperature change of the bearing is also an important means of detecting bearing failure.

[0093] Increased noise: When a bearing fails, abnormal noise may be generated. This noise may be caused by wear and cracks inside the bearing.

[0094] Lubricant contamination: When a bearing fails, it may cause lubricant contamination. For example, metal particles produced by bearing wear may mix into the lubricant, causing the color of the lubricant to become black or turbid.

[0095] Bearing deformation or damage: Severe bearing failure may cause bearing deformation or damage. For example, the inner and outer rings of the bearing may be cracked, worn, etc., and the rolling elements may be damaged.

[0096] Reduced operating efficiency: When a bearing fails, the operating efficiency of the wind turbine may be reduced due to increased friction, etc. For example, the output power of the generator may decrease and the overall performance of the wind turbine may be affected.

[0097] In summary, when a wind turbine main shaft bearing fails, it will show abnormalities in vibration signals, temperature, noise, lubricating oil status, bearing appearance, and operating efficiency. Therefore, by monitoring and analyzing these parameters, bearing failures can be discovered in a timely manner, and appropriate measures can be taken to repair or replace them to ensure the normal operation of the wind turbine.

[0098] The model building process includes data acquisition, data division, and building sub-models based on different data (the first monitoring model and the type selection model), and finally generates a comprehensive model that can be used to monitor the current abnormal type of wind turbine bearings.

[0099] Set monitoring time nodes and obtain operating data for monitoring. Reasonably set monitoring time nodes and obtain wind turbine operating data at each node. These data will be input into the constructed wind turbine monitoring model to obtain the abnormal type of wind turbine bearing.

[0100] A maintenance strategy is generated based on the abnormality type. According to the identified engine bearing abnormality type, the relevant second feature set data is considered to determine the urgency of abnormality processing, and then a corresponding maintenance strategy is formulated to ensure the normal operation of the wind turbine.

[0101] Embodiment 2:

[0102] The first monitoring model of the wind turbine bearing is constructed based on the first characteristic data, comprising:

[0103] The first characteristic data is a vibration signal;

[0104] Get the historical vibration signal D of the current wind turbine, D = [d 1 ,d 2 …d i …d n ];

[0105] Among them, di is the acceleration of the vibration signal at the i-th sampling moment, and n is the total number of sampling moments of the vibration signal;

[0106] Perform equal gradient division on the historical vibration signal D, generate the acceleration evaluation value of the vibration signal, and obtain the peak value of the acceleration evaluation value;

[0107] and determine the frequency, amplitude and duration of the corresponding peak;

[0108] The frequency, amplitude and duration of the peak are standardized to obtain the frequency evaluation value f in the same value range. 1i , amplitude evaluation value f 2i And the duration evaluation value f3 i ;

[0109] Based on the frequency evaluation value f 1i , amplitude evaluation value f 2i And the duration evaluation value f 3i Generate an abnormal state evaluation value F;

[0110] Classify abnormal types of wind turbines based on historical abnormal data of wind turbines;

[0111] The abnormality type of the wind turbine is determined based on the abnormal state evaluation value F, and an abnormality type pool of the wind turbine bearing is generated.

[0112] Embodiment 3:

[0113] The generating of the abnormal state evaluation value F includes:

[0114] Based on the gradient of wind turbine operating load, a multi-gradient load E is generated, where E = [E 1 ,E 2 …E i …E k ];

[0115] E i represents the i-th gradient load, k represents the number of gradient loads;

[0116] Calculate the average value of multiple gradient loads E, and select the gradient load closest to the average value as the average gradient load Ea;

[0117] Calculate the frequency evaluation value f of the average gradient load Ea 1i , amplitude evaluation value f 2i and duration evaluation value f 3i The correlation between the abnormal state and the correlation coefficients r1, r2, and r3 under gradient load were obtained;

[0118]

[0119] Among them, w 1 Indicates the frequency evaluation value f 1i The initial weight, w 2 Indicates the amplitude evaluation value f 2i The initial weight, w 3 Indicates the duration evaluation value f 3i The initial weight of

[0120] The initial weight is adjusted according to the gradient load of the engine to generate the abnormal state evaluation value F.

[0121] In this embodiment, the first characteristic data mainly focuses on the characteristics that directly reflect the state of the wind turbine bearing. In addition to the vibration signal mentioned above, it can also include the temperature gradient of the bearing (the change of temperature at different positions of the bearing), the pressure distribution inside the bearing and other characteristics. These characteristics are closely related to the physical structure and operation mechanism of the bearing, and can directly reflect whether the bearing is in an abnormal state.

[0122] The second characteristic data considers more external factors and auxiliary judgment factors. In addition to data related to temperature, noise, bearing deformation, and lubricant contamination, it can also include data such as the altitude of the wind turbine location (which may affect air density and thus heat dissipation), local air quality (dust and other pollutants may affect the operation of bearings and other components), etc.

[0123] When performing the division, the principal component analysis (PCA) method can be used to analyze the variance contribution rate of the data, and the components that mainly reflect the bearing status can be divided into the first characteristic data, and the other components can be divided into the second characteristic data.

[0124] Embodiment 4: The adjusting of the initial weight according to the gradient load of the engine comprises:

[0125] Get the i-th gradient load Ei of the engine and calculate the frequency evaluation value f of the i-th gradient load Ei 1i , amplitude evaluation value f 2i and duration evaluation value f 3i The correlation between the abnormal state and the gradient load is obtained. 1i 、r 2i 、r 3i ;

[0126]

[0127] Among them, a 1i represents the frequency evaluation value f of the i-th gradient load 1i The adjustment factor of 2i represents the magnitude evaluation value f of the i-th gradient load 2iThe adjustment factor of 3i represents the duration evaluation value f of the i-th gradient load 3i The adjustment factor of

[0128] Generate abnormal state evaluation value F, F = w 1 *a 1i *f 1i +w 2 *a 2i *f 2i +w 3 *a 3i *f 3i .

[0129] In this embodiment, the vibration signal processing details, for the historical vibration signal D of the current wind turbine, when performing equal gradient division, the gradient can be determined according to the normal operating vibration range of the bearing and actual engineering experience. For example, if the normal vibration acceleration range of the bearing is [a, b], this range can be divided into m equal gradient intervals, and the width of each interval is (ba) / m.

[0130] After determining the frequency, amplitude, and duration of the peak, the normalization method can use Z-score normalization. 1i , the calculation formula is f 1i =(x-μ) / σ, where x is the original frequency value, μ is the mean value of the frequency, and σ is the standard deviation of the frequency. 2i and duration evaluation value f 3i A similar standardization approach was also used.

[0131] When classifying the abnormal types of wind turbines based on the historical abnormal data of wind turbines, a cluster analysis method can be used. For example, using the K-means clustering algorithm, the historical abnormal data is divided into different clusters according to the vibration signal characteristics (frequency, amplitude, duration, etc.), and each cluster represents an abnormal type.

[0132] When determining the abnormal type of the wind turbine and generating the abnormal type pool of the wind turbine bearing, the size of the abnormal state evaluation value F is compared with the central value of each cluster, and those that meet a certain threshold range are classified into the corresponding abnormal type, thereby constructing the abnormal type pool.

[0133] Embodiment 5:

[0134] The step of constructing a type selection model based on the second feature data includes:

[0135] Based on historical data, obtain the abnormal state evaluation value F when all abnormal types occur and the second feature data change set R at the same time, R = [T, V, Q, S];

[0136] Among them, T represents the temperature change data set, V represents the noise data set, Q represents the bearing deformation data set, and S represents the lubricating oil pollution data set;

[0137] The data change degrees of the temperature change data set T, the noise data set V, the bearing deformation data set Q, and the lubricating oil pollution data set S are respectively divided into thresholds to generate data change scores for the temperature change data set T, the noise data set V, the bearing deformation data set Q, and the lubricating oil pollution data set S;

[0138] Based on the data change scores of the temperature change data set T, the noise data set V, the bearing deformation data set Q, and the lubricating oil pollution data set S, an abnormal type in the abnormal type pool of the wind turbine bearing corresponding to the abnormal state evaluation value F is generated.

[0139] In this embodiment, the data change score is generated, and for the temperature change data set T, when dividing the threshold, it can be determined according to the material characteristics of the bearing and the normal operating temperature range. For example, if the normal operating temperature range of the bearing is [c, d], the temperature change degree can be divided into three threshold intervals: when the temperature change is less than e (e is a small value, indicating normal fluctuation), the data change score is 1; when the temperature change is between [e, f], the data change score is 2; when the temperature change is greater than f, the data change score is 3.

[0140] A similar threshold division method based on actual engineering experience and equipment characteristics is also used for the noise dataset V, bearing deformation dataset Q, and lubricant contamination dataset S to generate their respective data change scores.

[0141] A decision tree model can be established based on the relationship between the data change scores of the temperature change data set T, the noise data set V, the bearing deformation data set Q, and the lubricating oil contamination data set S and the abnormal state evaluation value F. For example, the data change score is used as the branch condition of the decision tree, and the abnormal state evaluation value F is used as the judgment basis of the leaf node, and finally the abnormal type in the abnormal type pool of the wind turbine bearing is determined.

[0142] By analyzing the association between the change score of each data and the type of anomaly, we can better utilize the patterns in historical data. The historical data records the changes in each data when different types of anomalies occur. By analyzing and modeling these data, a more accurate mechanism for determining the type of anomaly can be established. For example, if historical data shows that when a certain anomaly occurs, the change score of the lubricating oil contamination dataset S is always at a high level, and the score of the temperature change dataset T also changes accordingly, then in actual monitoring, when a similar score combination appears again, the corresponding type of anomaly can be determined more accurately.

[0143] Generate anomaly types based on the change scores of multi-source data, which can adapt to a variety of abnormal situations. There are many types of abnormalities that may occur in wind turbine bearings, such as fatigue, wear, insufficient lubrication, etc. Each abnormal situation may have different effects on temperature, noise, bearing deformation and lubricant contamination. By comprehensively considering the change scores of these factors, more possible anomaly types can be covered, so that the anomaly types in the generated anomaly type pool are more comprehensive and can better cope with various situations that may occur in actual operation.

[0144] Embodiment 6:

[0145] When the wind turbine bearing abnormality type is generated by the wind turbine monitoring model, it includes:

[0146] Acquire a second characteristic data set at a current monitoring time node and data on changes in the surrounding environment of the wind turbine generator set, and correct the second characteristic data set based on the data on changes in the surrounding environment of the wind turbine generator set to generate a corrected second characteristic data set;

[0147] Generate an abnormal type pool of the current wind turbine bearing through the first monitoring model;

[0148] Based on the abnormal type pool of the current wind turbine bearing, the abnormal type of the current wind turbine bearing is generated through the type selection model combined with the corrected second feature data set.

[0149] In this embodiment, the data of the current monitoring time node is obtained and the second characteristic data set is corrected. When the second characteristic data set of the current monitoring time node and the environmental change data around the wind turbine are obtained, the collection of environmental change data is more accurate. For example, the environmental temperature data should take into account the influence of factors such as solar radiation and shading of surrounding buildings. Multiple temperature sensors can be set at different positions around the wind turbine, and the average value is taken as the environmental temperature data.

[0150] When the second characteristic data set is corrected based on the data of the change of the surrounding environment of the generator, a relationship model between the environmental factors and the second characteristic data is established. For example, if the ambient temperature rises, it may have a certain impact on the bearing deformation data set Q (due to thermal expansion and contraction). According to physical parameters such as the thermal expansion coefficient, a correction formula is established to correct the data in Q, thereby generating a corrected second characteristic data set.

[0151] When the abnormal type pool of the current wind turbine bearing is generated by the first monitoring model, the first characteristic data of the current monitoring time node, ie, the vibration signal, is processed according to the method of constructing the first monitoring model to obtain the current abnormal type pool.

[0152] Based on the abnormal type pool of the current wind turbine bearing, when the abnormal type of the current wind turbine bearing is generated by combining the type selection model with the corrected second feature data set, the corrected second feature data set is input into the type selection model, and the matching degree is calculated with each abnormal type in the abnormal type pool. For example, the distance between the corrected second feature data set and the second feature data template corresponding to each abnormal type can be calculated by using methods such as Euclidean distance, and the one with the smallest distance is the abnormal type of the current wind turbine bearing.

[0153] Embodiment 7:

[0154] The generation of the maintenance strategy based on the abnormal type of the engine bearing includes:

[0155] Generate the urgency of abnormality processing based on the abnormality type of the current wind turbine bearing and the second feature set data;

[0156] Develop maintenance strategies based on the urgency of the exception.

[0157] Embodiment 8:

[0158] The urgency of generating the exception handling includes:

[0159] Building an exception handling urgency evaluation model based on the second feature set data;

[0160] The emergency evaluation value θ of exception handling is generated by the exception handling urgency evaluation model. The emergency evaluation value θ of exception handling is compared with the preset value (θ 1 ,θ 2 ) to generate the urgency of exception handling, where θ 1 is the first preset value, θ 2 is a second preset value;

[0161] If θ 2 <θ, then the current exception type is level one emergency;

[0162] If θ 1 ≤θ≤θ 2 , then the current exception type is level 2 emergency;

[0163] If θ<θ 1 , the current exception type is level 3 emergency.

[0164] In this embodiment, constructing an exception handling urgency evaluation model includes:

[0165] When constructing an exception handling urgency evaluation model based on the second feature set data, a neural network model, such as a multi-layer perceptron (MLP), can be used. The second feature set data is used as the input layer, and after being processed by the hidden layer, an exception handling urgency evaluation value θ is output.

[0166] When training this neural network model, the historical second feature set data and the corresponding exception handling urgency (manually labeled according to the actual processing situation) are used as training data. The weights of the neural network are adjusted through the back-propagation algorithm so that the model can accurately generate the emergency evaluation value θ based on the input second feature set data.

[0167] By comparing the emergency evaluation value θ of the abnormality processing with the preset value (θ 1 ,θ 2 ) to generate the urgency of abnormal handling, the θ should be determined according to the actual operation requirements and safety standards of the wind turbine. 1 and θ 2 For example, for some key wind turbines, θ 1 and θ 2 The value of may be set more strictly to ensure that effective maintenance measures can be taken in time when bearing abnormalities occur.

[0168] Embodiment 9:

[0169] The maintenance strategy formulated based on the degree of emergency includes:

[0170] If the current abnormality type is level 1 emergency, a shutdown instruction is generated and maintenance is performed;

[0171] If the current abnormality type is level 2 emergency, the maintenance time is set according to the unit's operation plan, and the monitoring cycle of the unit's operating status is adjusted within a period of time after the unit is restarted;

[0172] If the current abnormality type is level 3 emergency, the unit will continue to operate and the monitoring cycle of the unit operating status will be adjusted. Based on the change in the emergency evaluation value of the abnormality processing, it is determined whether to adjust the urgency of the current abnormality type.

[0173] In this embodiment, when the current abnormality type is the first-level emergency, a shutdown command is generated and maintenance is performed. During the shutdown process, the operation should be carried out in accordance with the safety operating procedures. First, the connection between the wind turbine and the power grid is cut off, and then the operation of each component of the unit is gradually stopped to avoid sudden shutdown causing additional damage to the unit. During maintenance, the bearings and related components should be fully inspected, including disassembling the bearings to check the internal wear, checking the quality and quantity of the lubricating oil, and checking the concentricity of the transmission shaft connected to the bearings.

[0174] When the current abnormality type is the second level of emergency, set the maintenance time according to the unit's operation plan. When setting the maintenance time, take into account the needs of the current power generation task and the risks that may be caused by the continued operation of the unit. Adjust the monitoring cycle of the unit's operating status within a period of time after the unit is restarted. For example, the original monitoring once an hour can be adjusted to monitoring once every 15 minutes, so as to promptly detect whether the abnormal situation has deteriorated.

[0175] When the current abnormality type is level 3 emergency, continue to keep the unit running and adjust the monitoring cycle of the unit's operating status. When adjusting the monitoring cycle, it can be determined based on the changing trend of the emergency evaluation value of abnormal handling. If the emergency evaluation value has an upward trend, the monitoring cycle should be appropriately shortened; if the emergency evaluation value remains stable or decreases, the monitoring cycle can be appropriately extended. At the same time, continue to pay attention to the operating status of the unit and determine whether to adjust the urgency of the current abnormality type based on relevant indicators.

[0176] Finally, it should be noted that it is obvious that those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the present invention and its equivalent technology, the present invention is also intended to include these modifications and variations.

[0177] The above description is only a preferred embodiment of the present invention and is not intended to limit the protection scope of the present invention.

Claims

1. A method for monitoring abnormal state of a wind turbine bearing, characterized in that: include: Based on the historical operation data of wind turbines, a wind turbine monitoring model is constructed; Set the monitoring time node of the wind turbine, obtain the operation data of the wind turbine at the current monitoring time node, and generate the abnormality type of the wind turbine bearing through the wind turbine monitoring model; Generate maintenance strategies based on the abnormality types of engine bearings; Among them, when building a wind turbine monitoring model, it includes: Dividing the historical operation data of the wind turbine generator set into first characteristic data and second characteristic data; Constructing a first monitoring model for the wind turbine bearing based on the first characteristic data; Generate an abnormality type pool of wind turbine bearings through a first monitoring model, wherein the abnormality type pool of wind turbine bearings includes abnormalities of multiple types of wind turbine bearings; A type selection model is constructed based on the second characteristic data, and the abnormal type pool is screened through the type selection model to generate the abnormal type of the wind turbine.

2. The method for monitoring abnormal state of a wind turbine bearing according to claim 1, characterized in that: The first monitoring model of the wind turbine bearing is constructed based on the first characteristic data, comprising: The first characteristic data is a vibration signal; Get the historical vibration signal D of the current wind turbine, D = [d1, d2…d i …d n ]; Among them, d i is the acceleration of the vibration signal at the i-th sampling moment, and n is the total number of sampling moments of the vibration signal; Perform equal gradient division on the historical vibration signal D, generate the acceleration evaluation value of the vibration signal, and obtain the peak value of the acceleration evaluation value; and determine the frequency, amplitude and duration of the corresponding peak; The frequency, amplitude and duration of the peak are standardized to obtain the frequency evaluation value f in the same value range. 1i , amplitude evaluation value f 2i And the duration evaluation value f3 i ; Based on the frequency evaluation value f 1i , amplitude evaluation value f 2i And the duration evaluation value f 3i Generate an abnormal state evaluation value F; Classify abnormal types of wind turbines based on historical abnormal data of wind turbines; The abnormality type of the wind turbine is determined based on the abnormal state evaluation value F, and an abnormality type pool of the wind turbine bearing is generated.

3. The method for monitoring abnormal state of a wind turbine bearing according to claim 2, characterized in that: The generating of the abnormal state evaluation value F includes: Based on the gradient of wind turbine operating load, a multi-gradient load E is generated, where E = [E1, E2…E i …E k ]; E i represents the i-th gradient load, k represents the number of gradient loads; Calculate the average value of multiple gradient loads E, and select the gradient load closest to the average value as the average gradient load Ea; Calculate the frequency evaluation value f of the average gradient load Ea 1i , amplitude evaluation value f 2i and duration evaluation value f 3i The correlation between the abnormal state and the correlation coefficients r1, r2, and r3 under gradient load were obtained; Among them, w1 represents the frequency evaluation value f 1i The initial weight of w2 represents the amplitude evaluation value f 2i The initial weight of w3 represents the duration evaluation value f 3i The initial weight of The initial weight is adjusted according to the gradient load of the engine to generate the abnormal state evaluation value F.

4. The method for monitoring abnormal state of a wind turbine bearing according to claim 3, characterized in that: The step of adjusting the initial weight according to the gradient load of the engine includes: Get the i-th gradient load Ei of the engine and calculate the frequency evaluation value f of the i-th gradient load Ei 1i , amplitude evaluation value f 2i and duration evaluation value f 3i The correlation between the abnormal state and the gradient load is obtained. 1i 、r 2i 、r 3i ; Among them, a 1i represents the frequency evaluation value f of the i-th gradient load 1i The adjustment factor of 2i represents the magnitude evaluation value f of the i-th gradient load 2i The adjustment factor of 3i represents the duration evaluation value f of the i-th gradient load 3i The adjustment factor of Generate abnormal state evaluation value F, F = w1*a 1i *f 1i +w2*a 2i *f 2i +w3*a 3i *f 3i .

5. The method for monitoring abnormal state of a wind turbine bearing according to claim 4, characterized in that: The step of constructing a type selection model based on the second feature data includes: Based on historical data, obtain the abnormal state evaluation value F when all abnormal types occur and the second feature data change set R at the same time, R = [T, V, Q, S]; Among them, T represents the temperature change data set, V represents the noise data set, Q represents the bearing deformation data set, and S represents the lubricating oil pollution data set; The data change degrees of the temperature change data set T, the noise data set V, the bearing deformation data set Q, and the lubricating oil pollution data set S are respectively divided into thresholds to generate data change scores for the temperature change data set T, the noise data set V, the bearing deformation data set Q, and the lubricating oil pollution data set S; Based on the data change scores of the temperature change data set T, the noise data set V, the bearing deformation data set Q, and the lubricating oil pollution data set S, an abnormal type in the abnormal type pool of the wind turbine bearing corresponding to the abnormal state evaluation value F is generated.

6. The method for monitoring abnormal state of a wind turbine bearing according to claim 5, characterized in that: When the wind turbine bearing abnormality type is generated by the wind turbine monitoring model, it includes: Acquire a second characteristic data set at a current monitoring time node and data on changes in the surrounding environment of the wind turbine generator set, and correct the second characteristic data set based on the data on changes in the surrounding environment of the wind turbine generator set to generate a corrected second characteristic data set; Generate an abnormal type pool of the current wind turbine bearing through the first monitoring model; Based on the abnormal type pool of the current wind turbine bearing, the abnormal type of the current wind turbine bearing is generated through the type selection model combined with the corrected second feature data set.

7. The method for monitoring abnormal state of a wind turbine bearing according to claim 6, characterized in that: The generation of the maintenance strategy based on the abnormal type of the engine bearing includes: Generate the urgency of abnormality processing based on the abnormality type of the current wind turbine bearing and the second feature set data; Develop maintenance strategies based on the urgency of the exception.

8. The method for monitoring abnormal state of a wind turbine bearing according to claim 7, characterized in that: The urgency of generating the exception handling includes: Building an exception handling urgency evaluation model based on the second feature set data; Generate an emergency evaluation value θ of the exception handling through the exception handling urgency evaluation model, and generate the urgency of the exception handling by comparing the emergency evaluation value θ of the exception handling with a preset value (θ1, θ2), wherein θ1 is a first preset value and θ2 is a second preset value; If θ2<θ, the current exception type is level 1 emergency; If θ1≤θ≤θ2, the current exception type is level 2 urgency; If θ<θ1, the current exception type is level 3 emergency.

9. The method for monitoring abnormal state of a wind turbine bearing according to claim 8, characterized in that: The maintenance strategy formulated based on the degree of emergency includes: If the current abnormality type is level 1 emergency, a shutdown instruction is generated and maintenance is performed; If the current abnormality type is level 2 emergency, the maintenance time is set according to the unit's operation plan, and the monitoring cycle of the unit's operating status is adjusted within a period of time after the unit is restarted; If the current abnormality type is level 3 emergency, the unit will continue to operate and the monitoring cycle of the unit operating status will be adjusted. Based on the change in the emergency evaluation value of the abnormality processing, it is determined whether to adjust the urgency of the current abnormality type.

Images