CN113297291A

CN113297291A - Monitoring method, monitoring system, readable storage medium and wind driven generator

Info

Publication number: CN113297291A
Application number: CN202110501853.3A
Authority: CN
Inventors: 顾浩天; 洪文钟; 巩源泉; 蒋勇; 王莉娟
Original assignee: Shanghai Electric Wind Power Group Co Ltd
Current assignee: Shanghai Electric Wind Power Group Co Ltd
Priority date: 2021-05-08
Filing date: 2021-05-08
Publication date: 2021-08-24

Abstract

The application provides a monitoring method, a monitoring system, a readable storage medium and a wind driven generator. The monitoring method comprises the following steps: acquiring working condition data of the working condition of the wind driven generator to be monitored and corresponding state characteristic parameter data; inputting the working condition data into a threshold value determination model to output a target threshold value of the working condition of the wind driven generator, wherein the threshold value determination model is obtained by training by utilizing a parameter data sample of the working condition of the wind driven generator; and determining the fault condition of the wind driven generator according to the state characteristic parameter data and the target threshold value. This allows for more accurate diagnosis of fault conditions.

Description

Monitoring method, monitoring system, readable storage medium and wind driven generator

Technical Field

The present application relates to the field of wind power generation, and in particular, to a monitoring method, a monitoring system, a readable storage medium, and a wind power generator.

Background

The wind power generator mainly depends on wind power to realize power generation. During the actual operation of a wind turbine, various types of faults may occur in the wind turbine. Therefore, the fault of the wind turbine needs to be monitored by the monitoring device to realize the maintenance of the wind turbine.

In the related art, a monitoring device sets a threshold value according to manual experience, and then compares data to be monitored in the actual operation process of the wind turbine with the threshold value to monitor the fault of the wind turbine. Such a threshold value is preset, i.e. fixed. However, the actual working condition environment of the wind driven generator is complex, and the preset threshold value cannot accurately diagnose the fault of the wind driven generator.

Disclosure of Invention

The application provides a monitoring method, a monitoring system, a readable storage medium and a wind driven generator.

The application provides a monitoring method, wherein the method comprises the following steps:

acquiring working condition data of the working condition of the wind driven generator to be monitored and corresponding state characteristic parameter data;

inputting the working condition data into a threshold value determination model to output a target threshold value of the working condition of the wind driven generator, wherein the threshold value determination model is obtained by training by utilizing a parameter data sample of the working condition of the wind driven generator; and

and determining the fault condition of the wind driven generator according to the state characteristic parameter data and the target threshold value.

Optionally, the method includes:

classifying the parameter data samples of the working conditions of the wind driven generator to obtain a plurality of classes;

and determining the target threshold values of various types according to the parameter data samples in various types to obtain the threshold value determination model.

Optionally, the classifying the parameter data samples of the wind turbine operating conditions to obtain a plurality of classes includes:

classifying the parameter data samples of the working conditions of the wind driven generator by adopting a K-mean clustering method to obtain a plurality of classes; each parameter data sample comprises a plurality of working condition parameters, and the value of K of the K-mean clustering method is determined according to the value attributes of the working condition parameters; and the value attribute represents that the working condition parameter is a continuous value or a discrete value.

Optionally, the plurality of working condition parameters include continuous working condition parameters, and the value attribute represents a working condition parameter in which the working condition parameter value is a continuous value;

the determining the value of K of the K-mean clustering method according to the value attributes of the working condition parameters comprises the following steps:

segmenting the value range interval of the continuous working condition parameters to obtain a plurality of sections of intervals, wherein the working conditions of the wind driven generator in different intervals are different;

and determining the value of the K according to the number of the sections of the multi-section interval.

Optionally, the plurality of operating condition parameters include a plurality of continuous operating condition parameters;

the determining the value of K according to the number of segments of the multi-segment interval includes:

and determining the value of K according to the product of the number of the sections of the multi-section interval corresponding to the continuous working condition parameters.

Optionally, the plurality of operating condition parameters include discrete operating condition parameters, and the discrete operating condition parameters are operating condition parameters whose value attribute represents that the operating condition parameters take discrete values;

and determining the value of K according to the number of the discrete values which can be obtained by the discrete working condition parameters.

Optionally, the plurality of operating condition parameters include a plurality of discrete operating condition parameters;

determining the value of K according to the number of discrete values which can be obtained by the discrete working condition parameters, wherein the determining comprises the following steps:

and determining the value of K according to the product of the number of discrete values which can be obtained by the plurality of discrete working condition parameters.

and determining the value of K according to the product of the number of the sections of the multi-section interval corresponding to the continuous working condition parameter and the number of the discrete values which can be obtained by the discrete working condition parameter.

Optionally, the method includes:

acquiring a test sample of a label marked with a known fault mode and corresponding test state characteristic parameter data;

inputting the test sample into the threshold determination model to output a test target threshold of the working condition of the wind driven generator;

determining a test fault mode of the wind driven generator according to the test state characteristic parameter data and the test target threshold;

and under the condition that the accuracy of the test fault mode is not up to the standard compared with the known fault mode, carrying out interval segmentation on the value domain interval of the continuous working condition parameters again to obtain a new multi-segment interval so as to re-determine the value of the K, reclassifying the parameter data samples, and re-determining the target threshold values of various types so as to modify the threshold value determination model.

Optionally, the determining the target threshold of each class according to the parameter data samples in each class includes:

determining initial threshold values corresponding to a plurality of parameter data samples in each class respectively;

weighting a plurality of the initial threshold values in each class, and determining the target threshold value of each class.

Optionally, the plurality of parameter data samples in each category include a cluster center sample serving as a cluster center and a cluster element sample outside the cluster center;

the determining initial thresholds corresponding to a plurality of parameter data samples in each class includes:

determining the distance between the clustering element sample in each class and the clustering center sample in the class;

and determining the initial threshold value of the clustering element sample in each class according to the distance between the clustering element sample in each class and the clustering center sample in the class.

Optionally, the determining the distance between the cluster element sample in each class and the cluster center sample in this class includes:

determining a first distance between the clustering element sample in each class and the clustering center sample in the class;

normalizing the first distance to obtain a normalized second distance;

the determining an initial threshold of the clustering element samples in each class according to the distance between the clustering element samples in each class and the clustering center samples in the class comprises:

and determining an initial threshold value of the clustering element samples in each class according to the second distance.

Optionally, the weighting the plurality of initial thresholds in each class to determine the target threshold of each class includes:

weighting a plurality of initial thresholds in each class by adopting weighted weights, and determining the target threshold of each class, wherein the weighted weights are the distances from the parameter data samples in each class to the clustering center in the class.

Optionally, before classifying the parameter data samples of the wind turbine operating conditions to obtain a plurality of classes, the method further includes:

acquiring a parameter data set of the wind driven generator; wherein the parameter data set comprises: a plurality of parameter data, each parameter data comprising: data of a plurality of initial operating condition parameters;

and selecting the first n initial working condition parameters of which the cumulative variance contribution rate of the corresponding data is greater than a contribution rate threshold from the data of the initial working condition parameters of the parameter data by adopting a principal component analysis method, wherein n is a natural number greater than 1, and the initial working condition parameters are taken as the working condition parameters of the parameter data sample.

The present application also provides a readable storage medium having a computer program stored thereon, wherein the program, when executed by a processor, implements the monitoring method described above.

The application also provides a wind driven generator state monitoring system, which comprises one or more processors and is used for realizing the monitoring method.

The present application further provides a wind power generator, wherein the wind power generator includes:

a tower;

a nacelle mounted to the tower;

a wind wheel assembled to the nacelle;

a wind driven generator state monitoring system is used for realizing the monitoring method.

According to the technical scheme provided by the embodiment of the application, the state parameter data of the working condition of the wind driven generator is input into the threshold value determination model so as to output the target threshold value of the working condition of the wind driven generator and further determine the fault condition of the wind driven generator. The target threshold is determined by using the state parameter data through the threshold determination model, and the threshold determination model is obtained through parameter data sample training, so that even if the working condition environment of the wind driven generator is complex and changeable, the target threshold can be dynamically and adaptively adjusted through the threshold determination model.

Drawings

FIG. 1 illustrates a schematic structural view of an embodiment of a wind turbine of the present application;

FIG. 2 is a flow chart illustrating one embodiment of a monitoring method of the present application;

FIG. 3 is a schematic flow diagram illustrating one embodiment of the trained threshold determination model of the deterministic monitoring method shown in FIG. 2;

FIG. 4 is a flow diagram illustrating one embodiment for determining the value of K;

FIG. 5 is a schematic flow chart diagram illustrating one embodiment of FIG. 3 in connection with determining a target threshold in the threshold determination model;

FIG. 6 is a flow diagram illustrating one embodiment of determining initial thresholds for cluster element samples in a thresholding model;

FIG. 7 is a detailed flow diagram illustrating one embodiment of determining initial thresholds for cluster element samples in a thresholding model;

FIG. 8 is a schematic flow diagram illustrating another embodiment of the threshold determination model of the deterministic monitoring method shown in FIG. 2;

FIG. 9 is a block diagram of a wind turbine condition monitoring system according to the present application.

Detailed Description

Reference will now be made in detail to the exemplary embodiments, examples of which are illustrated in the accompanying drawings. When the following description refers to the accompanying drawings, like numbers in different drawings represent the same or similar elements unless otherwise indicated. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with one or more embodiments of the present specification. Rather, they are merely examples of apparatus and methods consistent with certain aspects of one or more embodiments of the specification, as detailed in the claims which follow.

It should be noted that: in other embodiments, the steps of the corresponding methods are not necessarily performed in the order shown and described herein. In some other embodiments, the method may include more or fewer steps than those described herein. Moreover, a single step described in this specification may be broken down into multiple steps for description in other embodiments; multiple steps described in this specification may be combined into a single step in other embodiments.

FIG. 1 illustrates a schematic structural view of an embodiment of a wind turbine 10 of the present application. As shown in fig. 1, a wind turbine 10 includes a tower 12 extending from a support surface 11, a nacelle 13 mounted on the tower 12, and a rotor structure 14 assembled to the nacelle 13. The wind rotor structure 14 comprises a rotatable hub 15 and at least one blade 16, the blade 16 being connected to the hub 15 and extending outwardly from the hub 15. In the embodiment shown in FIG. 1, the wind turbine structure 14 includes three blades 16. In some other embodiments, the wind turbine structure 14 may include more or fewer blades 16. A plurality of blades 16 may be spaced about hub 15 to facilitate rotating rotor structure 14 to enable wind energy to be converted into usable mechanical energy, and subsequently, electrical energy.

In some embodiments, wind turbine 10 includes: the wind driven generator state monitoring system can monitor the state of the wind driven generator 10 to monitor the fault of the wind driven generator 10, so as to realize the advance prediction and active alarm of the fault of the wind driven generator 10, and further realize the maintenance of the wind driven generator 10.

FIG. 2 is a flow chart illustrating one embodiment of a monitoring method 20 of the present application. Referring to fig. 2, the monitoring method of the present application includes steps 201 to 203.

Step 201, obtaining working condition data of the working condition of the wind driven generator to be monitored and corresponding state characteristic parameter data.

The operating condition data may be data of operating condition parameters of the wind turbine, which reflect operating conditions (external environment operating conditions and conditions) of the wind turbine, such as power, rotation speed, and temperature. The working condition data may include data of a plurality of working condition parameters (working condition parameters of a plurality of dimensions), and the combination of the data of the plurality of working condition parameters reflects the working condition of the wind turbine generator. For example, each condition data is data of n condition parameters, which can be represented by n-dimensional vectors, and n is a positive integer greater than 1.

In some embodiments, the wind turbine operating condition may include an operating condition of a component of the wind turbine, such as an operating condition of a generator bearing, and the operating condition data may be operating condition data of the component of the wind turbine, and may reflect the operating condition of the component through a combination of a plurality of operating condition parameters. In some embodiments, the condition of the component may include a vibration condition of the component, such as a vibration condition of a generator bearing, and the plurality of condition parameters may include a generator bearing temperature, a generator power, and a generator speed, and a combination of the generator bearing temperature, the generator power, and the generator speed may reflect the vibration condition of the generator bearing, and the condition data is a combination of data of the condition parameters. Therefore, the working condition data can reflect the working conditions of the parts. Different parts of the wind driven generator can correspond to different working condition data and different working condition parameters. Therefore, the wind driven generator state monitoring system can collect a part of data in the data of the working condition parameters of the actual operation of the wind driven generator and can be used as the working condition data of the working conditions of the parts of the wind driven generator. The working condition data are used for assistance and reference, and the threshold value of the state characteristic parameter data can be determined in a self-adaptive mode according to different working condition data.

In some embodiments, the state characteristic parameter data of the wind turbine operating condition may refer to a characteristic parameter of the wind turbine state, so that whether the wind turbine fails or not may be monitored through the state characteristic parameter data. In some embodiments, the state characteristic parameter data for wind turbine operating conditions may include: vibration state characteristic parameter data, displacement state characteristic parameter data and inclination angle state characteristic parameter data.

The working condition data of the working condition of the wind driven generator and the corresponding state characteristic parameter data can be data which are collected at the same time and reflect the same working condition of the wind driven generator, and the data are collected at present. For example, the working condition of the wind driven generator is the vibration working condition of the generator bearing, and the state characteristic parameter data corresponding to the vibration working condition of the generator bearing is the vibration characteristic parameter data of the generator bearing. The vibration characteristic parameter data may include: the method includes the steps of measuring one or more of a jitter value, an amplitude value, a statistic and an energy value of the vibration signal, wherein the statistic can be a first-order statistic, a second-order statistic and a high-order statistic. State characteristic parameter data (e.g., one or more of a jitter value, an amplitude value, a statistic, and an energy value of the vibration signal) may be extracted from the state signal (e.g., the vibration signal).

Step 202, inputting the working condition data into a threshold value determination model to output a target threshold value of the working condition of the wind driven generator, wherein the threshold value determination model is obtained by training parameter data samples of the working condition of the wind driven generator.

The parameter data samples may be historical condition data of wind turbine conditions. Each parameter data sample may include a plurality of operating condition parameters, and the operating condition parameters corresponding to the operating condition data are consistent. The parametric data samples may also be represented by n-dimensional vectors, i.e. each parametric data sample is represented as an n-dimensional vector x₁,x₂,…,x_n]Elements of the n-dimensional vector are working condition parameters, wherein x₁Is the 1 st operating condition parameter, x₂Is the 2 nd operating condition parameter, x_nIs the nth working condition parameter.

The threshold determination model is obtained by training parameter data samples of the working condition of the wind driven generator, so that the machine learning can be used for learning the parameter data samples and learning the working condition of the wind driven generator. In some embodiments, the parameter data samples may also reflect operating conditions of the wind turbine components. Therefore, the machine learning can be used for learning the parameter data samples and learning the working conditions of the parts of the wind driven generator. Different operating condition data are input into the threshold determination model, and different target thresholds can be output. The threshold determination model may output a target threshold corresponding to the state characteristic parameter data. One threshold determination model corresponds to the working condition of one part, and different threshold determination models can be obtained by training for different parts.

The parameter data samples are obtained in various manners, and in some embodiments, a parameter data set of the wind turbine is obtained; wherein the parameter data set comprises: a plurality of parameter data, each parameter data comprising: the data of the plurality of initial operating condition parameters may be represented as a u-dimensional vector, for example, where u is a positive integer greater than or equal to n. And selecting the first n initial working condition parameters of which the cumulative variance contribution rate of corresponding data is greater than a contribution rate threshold value from a plurality of initial working condition parameters of the parameter data set by adopting a Principal Component Analysis (PCA) method as the working condition parameters of the parameter data sample. For example, the first n elements whose cumulative variance contribution ratio of the values of the elements is greater than the contribution ratio threshold may be selected from the u-dimensional vector, and an n-dimensional vector may be obtained, so as to obtain the parameter data sample, where n is the total number of the condition parameters of the parameter data sample, and n is a natural number greater than 1. Therefore, working condition parameters more relevant to the monitored working condition of the wind driven generator can be selected, the working condition parameters with smaller relevance to the working condition are eliminated, the influence of the parameter data samples on the fault of the wind driven generator is larger, and then the data volume processed by the threshold value determination model is reduced when the parameter data samples are processed by the threshold value determination model subsequently, so that the processing speed of the threshold value determination model is increased, and the determination efficiency of the target threshold value is correspondingly improved.

In some embodiments, a working condition parameter matrix is constructed by obtaining initial working condition parameters in a parameter data set of the working conditions of the wind driven generator; and performing dimensionality reduction mapping on the working condition parameter matrix by adopting a PCA (principal component analysis) method, determining the cumulative variance contribution rate of the corresponding data, selecting the first n working condition parameters of which the cumulative variance contribution rate of the corresponding data is greater than a contribution rate threshold value from a plurality of initial working condition parameters of the parameter data set, and using the working condition parameters as parameter data samples, so that the cumulative variance contribution rates of the working condition parameters of multiple dimensions can be calculated by adopting the PCA method. Therefore, the accumulated variance contribution rate of the working condition parameters of each dimension can be accurately mastered by the PCA method. And the working condition parameters of the selected parameter data samples have larger correlation with the working condition of the wind driven generator. In some embodiments, for example, the working condition of the wind turbine may be a vibration working condition, and the correlation between the working condition parameters of the selected parameter data sample and the vibration working condition is large, and the influence on the change of the vibration working condition is also large.

Wherein, the contribution rate threshold value can be set according to the industrial demand. For example, the contribution rate threshold may be 95%, but is not limited thereto. The contribution rate threshold value is positively correlated with the working condition of the parts of the wind driven generator. The higher the cumulative variance contribution rate of the working condition parameters is, the greater the correlation between the working condition parameters and the working conditions of the parts of the wind driven generator is.

The following description takes a parameter data sample of the working condition of the generator bearing of the wind driven generator as an example:

the initial operating condition parameters of the generator bearings may include power, generator speed, hub speed, wind speed, yaw angle, gearbox oil temperature, main shaft bearing temperature, high speed shaft bearing temperature, and generator bearing temperature, and the parameter data set may include data of the above initial operating condition parameters. The working condition parameters with small correlation such as the rotating speed of a hub, the wind speed, the yaw angle, the oil temperature of a gear box, the temperature of a main shaft bearing and the temperature of a high-speed shaft bearing can be eliminated by a PCA method, and the working condition parameters with large correlation such as the temperature of a generator bearing, the power and the rotating speed of a generator are selected, so that the working condition parameters of a parameter data sample comprise: generator bearing temperature, power, generator speed.

And step 203, determining the fault condition of the wind driven generator according to the state characteristic parameter data and the target threshold value.

In some embodiments, a fault condition of the wind turbine may be determined based on a comparison of the state characteristic parameter data to a target threshold value, and subsequent alerts may be made. It may be determined that the wind turbine is not malfunctioning or is malfunctioning, and in some embodiments, upon the occurrence of the malfunction, a degree of the malfunction may be determined.

In some embodiments, the fault condition of the wind turbine may include: non-failed conditions and failed conditions. In some embodiments, the failed condition may include a pattern of different failure degrees. For example, the degree of failure may include: the first fault degree is higher than the second fault degree. Corresponding to the modes with different fault degrees, different warning modes can be used for warning or alarming subsequently, so that maintenance personnel can be reminded to maintain the wind driven generator in time. In some embodiments, the warning corresponds to a lower degree of failure than the warning corresponds to.

In some embodiments, the state characteristic parameter data is compared to a target threshold value to determine a fault condition of the wind turbine. In some embodiments, the state characteristic parameter value is above a target threshold value, then a wind turbine fault is determined. For example, if the state characteristic parameter value is a tower barrel inclination angle value, and the tower barrel inclination angle value is higher than a target threshold value, it is determined that a tower barrel of the wind driven generator fails.

And comparing the extracted characteristic parameter data with the target threshold value of the corresponding class to determine the fault condition of the wind driven generator. In some embodiments, the state characteristic parameter data may comprise data for a plurality of state characteristic parameters. The target threshold may include a corresponding plurality of dimensions of the threshold corresponding to the plurality of dimensions of the comparison state characteristic parameter data and the plurality of dimensions of the target threshold, respectively. For example, the multiple dimensions of the state characteristic parameter data are four dimensions of a jitter value, an amplitude value, a statistic and an energy value of the vibration signal, the target threshold includes thresholds respectively corresponding to the four dimensions, and the four dimensions of the state characteristic parameter data are correspondingly compared with the four dimensions of the target threshold. In other embodiments, the state characteristic parameter data may be data in one dimension.

In some embodiments, the fault condition of the wind turbine is used to reflect a fault with a component of the wind turbine. Fault conditions such as wind turbines may include: the wind driven generator comprises one or more of a failed wind driven generator transmission chain system, a failed wind driven generator yaw system and a failed wind driven generator pitch system. When the wind driven generator drive chain system fails, the wind driven generator drive chain system can comprise: the transmission blades, the inner and outer rings, the abrasion degree, the rotating shaft and other parts of the wind driven generator can be analyzed through the method for judging whether the transmission blades, the inner and outer rings, the abrasion degree, the rotating shaft and other parts of the wind driven generator have faults or not. Therefore, the fault of the drive chain system of the wind driven generator can be analyzed, the faults of other systems of the wind driven generator set can be further analyzed, and the faults of other system parts of the wind driven generator set can be further analyzed.

In the embodiment of the application, the working condition data of the working condition of the wind driven generator is input into the threshold value determination model so as to output the target threshold value of the working condition of the wind driven generator, and further determine the fault condition of the wind driven generator. The target threshold is determined by using the working condition data through the threshold determination model, the threshold determination model is obtained through parameter data sample training, and therefore under the condition that the working condition environment of the wind driven generator is continuously changeable and complex, the target threshold can be dynamically and adaptively adjusted through the threshold determination model.

FIG. 3 is a schematic flow chart diagram illustrating one embodiment of the threshold determination model trained in the determination monitoring method 20 shown in FIG. 2. In some embodiments, the following steps 301 to 302 may be adopted in the embodiments of the present application to obtain the threshold determination model:

step 301, classifying the parameter data samples of the wind driven generator working conditions to obtain a plurality of classes.

The parameter data samples are more relevant to the working conditions of the wind driven generator, and have larger influence on the faults of the wind driven generator. Therefore, in the classification process of step 301, the parameter data samples of the working condition parameters with relatively high correlation are used, and the parameter data samples of the working condition parameters with relatively low correlation with the working condition of the wind driven generator are not used, so that not only can the data amount in the classification process be reduced, but also the influence of the parameter data samples of the working condition parameters with relatively low correlation on the classification can be reduced, and the overfitting phenomenon in the classification process is avoided. And finally, the obtained parameter data samples in the class are more relevant, so that the classification accuracy is improved, and the accuracy of the target threshold is further improved.

There are various embodiments of obtaining multiple classes in step 301, and in one embodiment, a clustering method may be adopted to classify the parameter data samples of the wind turbine operating condition to obtain multiple classes. In this way, parameter data samples of the working conditions of the wind driven generator are classified through a clustering method, and characteristic parameter data of the parameter data samples are extracted, so that the working conditions of the wind driven generator are self-learned and self-classified, and a plurality of classes are obtained.

In some embodiments, the clustering method described above may include, but is not limited to, a K-means clustering method. The classification process is described below by taking the K-means clustering method as an example: classifying the parameter data samples of the working conditions of the wind driven generator by adopting a K-mean clustering method to obtain a plurality of classes; each parameter data sample comprises a plurality of working condition parameters, and the value of K of the K-mean clustering method is determined according to the value attributes of the working condition parameters; and the value attribute represents that the working condition parameter is a continuous value or a discrete value.

The continuous value may be a numerical value taken by the operating condition parameter, and is continuous process quantity data. For example, the operating condition parameter is the generator bearing temperature, and the operating condition parameter is a value taken from the value range of [0,80 ]. The discrete value can be a numerical value taken by the working condition parameter and is discrete process quantity data. For example, the operating condition parameter is power, and the value of the operating condition parameter is 100kw, 200kw or 300 kw.

In the embodiment, a K-means clustering method is used, parameter data samples of the working conditions of the wind driven generator do not need to be marked in advance, and multiple classes are obtained through unsupervised learning. Therefore, the workload of pre-marking is reduced, and the working condition of the wind driven generator which is more consistent with the actual working condition can be learned. And when the value of the value attribute characterization working condition parameter is a continuous value and/or when the value of the value attribute characterization working condition parameter is a discrete value, the value of K in the clustering method can be determined. Therefore, the classification quantity can be effectively determined, the data samples of each parameter under the working condition of the wind driven generator can be effectively analyzed, and the accuracy of fault diagnosis of the wind driven generator is improved.

FIG. 4 is a first flow diagram illustrating one embodiment for determining the value of K. In some embodiments, the plurality of operating condition parameters include a continuous operating condition parameter, and the continuous operating condition parameter is an operating condition parameter whose value attribute represents that the operating condition parameter has a continuous value, and the following steps 411 to 412 may be adopted to determine the value of K in the K-means clustering method:

and 411, performing interval segmentation on the value range interval of the continuous working condition parameters to obtain multiple segments of intervals, wherein the working conditions of the wind driven generators in different segments of intervals are different.

When the value attribute of the working condition parameter is a continuous value, the working condition parameter at this time can be called as a continuous working condition parameter. The value range interval is a value interval which can be obtained by the working condition parameters. For example, the 1 st continuous operating condition parameter is the generator temperature, the value range interval that the 1 st continuous operating condition parameter can get is [20 ℃, 60 ℃), and the [20 ℃, 60 ℃/is divided. The working conditions of different section intervals are different, and the working conditions of the same section interval are close.

In step 412, the value of K is determined based on the number of segments in the multi-segment interval. Therefore, the interval segmentation is carried out on the value range interval of the continuous working condition parameters, so that the working condition difference of the wind driven generator between the sections of each interval is large, the classification is facilitated, and the classification effectiveness is improved. And the value range interval of the continuous working condition parameters is segmented, the sections of the sections have similar working conditions of the wind driven generator, and the classification efficiency is improved.

In some embodiments, the plurality of operating condition parameters includes a plurality of continuous operating condition parameters; determining the value of K according to the number of segments of the multi-segment interval, comprising: and determining the value of K according to the product of the number of the sections of the multi-section interval corresponding to the continuous working condition parameters. In some embodiments, the plurality of operating condition parameters are continuous operating condition parameters, the value range interval of each continuous operating condition parameter is respectively segmented, and the product of the segment numbers of the intervals corresponding to the plurality of operating condition parameters is used as the value of K. Therefore, the value of K can be accurately determined under the condition that the working condition parameters are continuous working condition parameters, classification is convenient, and various target thresholds are further determined.

In some embodiments, the plurality of continuous operating condition parameters are continuous operating condition parameters, and the value range interval is subjected to interval segmentation, for example, the 1 st continuous operating condition parameter is subjected to interval segmentation, and the interval segmentation can be divided into m₁Segment, the 2 nd continuous working condition parameter is segmented into m₂The nth one is continued by analogyThe working condition parameters are segmented into m sections_nSegment, then the value of K for the K-means clustering method is determined as in equation (1):

∏m_i1≤i≤n (1)

wherein m is₁Number of stages, m, of the 1 st continuous operating mode parameter₂Number of stages, m, of the 2 nd continuous operating mode parameter_nThe number of the nth continuous working condition parameter segments is shown, i is the serial number of any continuous working condition parameter, and n is the total number of the continuous working condition parameters.

In some embodiments, the plurality of operating condition parameters includes discrete operating condition parameters, and the discrete operating condition parameters are operating condition parameters whose value attributes characterize the operating condition parameters whose values are discrete values. Determining the value of K of the K-mean clustering method according to the value attributes of a plurality of working condition parameters, wherein the method comprises the following steps: and determining the value of K according to the number of the discrete values which can be obtained by the discrete working condition parameters. The value of K thus determined is more accurate.

In some embodiments, when the value attributes of the plurality of operating condition parameters are discrete values, the operating condition parameters may be referred to as discrete operating condition parameters. The plurality of operating condition parameters comprises a plurality of discrete operating condition parameters; determining the value of K according to the number of discrete values which can be obtained by the discrete working condition parameters, wherein the determining comprises the following steps: and determining the value of K according to the product of the number of discrete values which can be obtained by the plurality of discrete working condition parameters. In some embodiments, the plurality of operating condition parameters are discrete operating condition parameters, and the product of the number of discrete values that can be obtained by the plurality of discrete operating condition parameters may be determined as the value of K. This improves the accuracy of the classification.

In some embodiments, the plurality of operating condition parameters are discrete operating condition parameters, and the number of discrete values that can be obtained by each operating condition parameter, for example, the 1 st discrete operating condition parameter, is a₁The number of the discrete values which can be obtained by the 2 nd discrete working condition parameter is a₂By analogy, the number of discrete values that the nth discrete operating condition parameter can take is a_nThen, the value of K in the K-means clustering method is determined as shown in formula (2):

∏a_i1≤i≤n (2)

wherein, a_iFor the ith discrete operating condition parameterThe number of discrete values that can be taken.

In other embodiments, the plurality of operating condition parameters includes at least one discrete operating condition parameter and at least one continuous operating condition parameter. And determining the value of K according to the product of the number of the sections of the multi-section interval corresponding to the continuous working condition parameters and the number of the discrete values which can be obtained by the discrete working condition parameters. The product of the number of segments of the multi-segment interval corresponding to the continuous working condition parameter and the number of discrete values that can be obtained by the discrete working condition parameter can be used as the value of K. Therefore, continuous working condition parameters and discrete working condition parameters are fully considered, and the classification accuracy is improved.

For example, the first k working condition parameters are discrete working condition parameters, the last n-k working condition parameters are continuous working condition parameters, and further: the number of discrete values which can be obtained by the 1 st discrete working condition parameter is a₁The number of the discrete values which can be obtained by the 2 nd discrete working condition parameter is a₂By analogy, the number of discrete values that the k-th discrete operating condition parameter can take is a_k(ii) a The k +1 continuous working condition parameters are segmented into m_k+1Segment, the k +2 continuous working condition parameters are segmented into intervals which can be divided into m_k+2And by analogy, the nth continuous working condition parameter is subjected to interval segmentation and can be divided into m_nSegment, then the value of K for the K-means clustering method is determined as in equation (3):

continuing with FIG. 3, step 302, determining target thresholds for the classes based on the parametric data samples in the classes to obtain a threshold determination model. Therefore, various target thresholds are adaptively adjusted through various parameter data samples, so that the obtained various target thresholds can better accord with the actual working condition of the wind driven generator, and further the fault condition can be diagnosed more accurately.

FIG. 5 is a flow diagram illustrating one embodiment of determining a target threshold in the threshold determination model shown in FIG. 3. In some embodiments, the target threshold values for each class may be determined from the parametric data samples in each class by:

step 501, determining initial thresholds corresponding to a plurality of parameter data samples in each class.

Wherein the plurality of parameter data samples in each class may include a cluster center sample as a cluster center and a cluster element sample outside the cluster center. In some embodiments, a K-means algorithm (K-means clustering algorithm) is performed on the parametric data samples, classified into K classes, and the clustering centers of the classes are determined. When the clustering element sample is closer to the clustering center sample, the reflected working condition is more similar to the working condition reflected by the clustering element sample, and the initial threshold of the clustering element sample is closer to the initial threshold of the clustering center sample.

The initial threshold may be a threshold corresponding to each of the multiple parameter data samples in each category, and may include: and the initial threshold value corresponds to the clustering center sample and the initial threshold value corresponds to the clustering element sample outside the clustering center. The initial threshold value may be a threshold value for the wind turbine operating condition, for example for a vibration condition of the generator bearing, the initial threshold value may be a threshold value corresponding to a characteristic parameter of the vibration signal. The dimension of the initial threshold coincides with the dimension of the target threshold.

In some embodiments, the initial threshold for this cluster center sample may be determined experimentally. Wherein, the process of the experiment is as follows: synchronously acquiring actual working condition data and state characteristic parameter data through a wind driven generator, and clustering the working condition data to obtain a plurality of classes; each type is provided with a clustering center, and working condition parameter vectors of the clustering centers in each type are obtained. Each element of the working condition parameter vector of the clustering centers in each type is a working condition parameter, the data of the working condition parameters are known determination data, and the data of the required working condition parameters on the experiment table are set through the known determination data. Then, the data of the condition parameters required on the experiment table are set through the data of the condition parameter vectors of the clustering centers. Then, under the specific condition of the data of the working condition parameters, state signals (such as vibration signals) are acquired through experiments, and state characteristic parameter data (such as one or more of jitter values, amplitude values, statistics and energy values of the vibration signals) are extracted from the state signals. And then, determining the change trend of the state characteristic parameter data according to the state characteristic parameter data to monitor the fault state of the wind driven generator, wherein the fault of the wind driven generator is serious, and an inflection point appears in the change trend of the state characteristic parameter data when an alarm is required. And determining the inflection point of the state characteristic parameter data from the variation trend of the state characteristic parameter data. And finally, taking the inflection point of the state characteristic parameter data as an initial threshold value of the clustering center sample.

In some embodiments, from category 1 to category K, the cluster centers of the categories are sequentially selected, and the initial threshold values of the cluster center samples are respectively obtained. In some embodiments, the initial threshold for clustering center samples may include two different initial thresholds, such as with a first initial threshold Warning_thresholdAnd a second initial threshold Alarming_thresholdAnd (4) showing. This may reflect different failure levels through two different initial thresholds. For example, the degree of failure of the second initial threshold may be more severe than the degree of failure of the first initial threshold.

FIG. 6 is a flow diagram illustrating one embodiment of determining initial thresholds for cluster element samples in a thresholding model. In some embodiments, the initial threshold of the cluster element sample may be obtained by using the following steps 511 and 512:

and 511, determining the distance between the clustering element samples in each class and the clustering center samples in the class.

And step 512, determining initial threshold values of the clustering element samples in each class according to the distances between the clustering element samples in each class and the clustering center samples in the class.

The classification can be to classify more relevant parameter data samples into one class, and the working condition environments corresponding to the classes are close. And the closer the distance between the clustering element sample in each class and the clustering center sample in the class is, the more relevant the clustering element sample and the clustering center sample are to the class.

FIG. 7 is a detailed flow diagram illustrating one embodiment of determining initial thresholds for cluster element samples in a thresholding model. In some embodiments, the initial threshold for the cluster element sample may be obtained as follows:

step 521, determining a first distance between the cluster element sample in each class and the cluster center sample in the class.

In some embodiments, the euclidean distance between the cluster element samples in each class and the cluster center sample of that class is determined as the first distance using a euclidean distance formula, where the euclidean distance formula is formula (4):

wherein d is the distance between the clustering element sample and the clustering center sample, x_iIs the i-th operating condition parameter, y, of the sample of the cluster element_iIs the ith working condition parameter of the cluster center sample.

Step 522, normalize the first distance to obtain a normalized second distance.

In some embodiments, a normalization formula of the first distance is used to normalize the first distance between the cluster element samples in each class and the cluster center sample in the class, so as to obtain a normalized second distance. Wherein the normalized formula of the first distance is formula (5):

wherein, d'_iIs the normalized second distance, d_iAnd p is the total number of various types of clustering element samples.

Step 523, determining an initial threshold of the clustering element samples in each class according to the second distance. The normalized second distance obtained in this way facilitates unified and rapid obtaining of the initial threshold of the clustering element sample.

In some embodiments, the initial threshold of the cluster element samples in each class is determined according to the distance between the cluster element sample in each class and the cluster center sample in the class and the initial threshold of the cluster center sample. In some embodiments, the initial threshold of the clustering element samples in each class is determined according to the normalized second distance by using the following threshold determination formula, such as formulas (6) to (9):

WarningThreshold1_i＝Warning_threshold(1+d′_i) (6)

WarningThreshold2_i＝Warning_threshold(1-d′_i) (7)

AlarmingThreshold1_i＝Alarming_threshold(1+d′_i) (8)

AlarmingThreshold2_i＝Alarming_threshold(1-d′_i) (9)

among them, WarningThreshold1_i、WarningThreshold2_i、AlarmingThreshold1_iAnd AlarmingThreshold2_iThe initial threshold values of the ith cluster element sample with different fault degrees are obtained by performing threshold value amplification on normalized distances between cluster elements and a cluster center and performing calculation on the initial threshold values of the formula (6) and the formula (8) respectively, and the initial threshold values of the formula (7) and the formula (9) are obtained by performing threshold value reduction on the normalized distances between the cluster elements and the cluster center and performing calculation on the initial threshold values of the ith cluster element sample with different fault degrees.

With continued reference to FIG. 5, at step 502, a plurality of initial thresholds in each class are weighted to determine a target threshold for each class. The target threshold value determined in this way is more accurate, and the accuracy of fault diagnosis of the wind driven generator is improved.

The "initial" of the initial thresholds and the "target" of the target thresholds are used to distinguish the two thresholds.

In some embodiments, this step 502 may determine various types of target thresholds as follows: and weighting a plurality of initial thresholds in each type by adopting weighted weights to determine target thresholds of each type, wherein the weighted weights are the distances from parameter data samples of each type to the clustering center in the type. Thus, the distance from each type of parameter data sample to the clustering center in the class is used as the weight, and the target threshold value of the actual class is better met, so that the obtained target threshold value of the class is more accurate.

In some embodiments, the target threshold determination formula is formulas (10) and (11):

among them, WarnngThreshold_totalAnd AlarmingThreshold_totalTarget thresholds for two different fault levels for each class; WarnngThreshold_iAnd AlarmingThreshold_iInitial thresholds, warnengthreshold, for two different fault levels of the sample_iIs WarningThreshold1_iOr WarningThreshold2_i，AlarmingThreshold_iIs AlarmingThreshold1_iOr AlarmingThreshold2_i。

When the target threshold is determined for the first time, it is not known whether the target threshold obtained for the first time is too large or too small, and therefore, any one of the above formulas (6) and (7) and any one of the formulas (8) and (9) may be used to participate in the operation of the formulas (10) and (11) to obtain the target threshold. After the target threshold is obtained, whether the target threshold is larger or smaller can be judged, and then the target threshold is adjusted next time. For example, when the target threshold is determined for the first time, the target threshold is obtained by using the formula (6) and the formula (8), and the target threshold is larger, then, when the target threshold is determined for the next time, the formula (7) and the formula (9) are used, so that the target threshold can be reduced. In the following, the same principle is applied when the target threshold obtained this time needs to be adjusted, and the target threshold can be determined according to the formula used in the previous time.

In the embodiment of the application, a plurality of working condition parameters are classified by using a K-means clustering algorithm, and a weighting idea is introduced to calculate the target threshold, so that the initial threshold characteristics of each parameter data sample are balanced, the accuracy of determining the target threshold is improved, and the method can be more suitable for complex and changeable working condition environments.

FIG. 8 is a schematic flow chart diagram illustrating another embodiment of the threshold determination model of the determination monitoring method 20 shown in FIG. 2. In some embodiments, the following steps 601 to 604 are performed, modifying the threshold determination model:

step 601, obtaining a test sample of a label marked with a known fault mode and corresponding test state characteristic parameter data.

Wherein the test samples identify known failure modes, i.e., known failure conditions, which may include: non-failing modes and failing modes, in some embodiments, failing modes may include modes of different failure degrees. The test specimen includes: and the working condition parameters of the test sample are consistent with the working condition parameters of the parameter data sample used for training.

Step 602, inputting the test sample into a threshold determination model to output a test target threshold of the wind turbine working condition.

In one embodiment, step 602 inputs the test sample into a threshold determination model, which may determine a class to which the test sample belongs and a target threshold for the class, which serves as a test target threshold.

Step 603, determining a test fault mode of the wind driven generator according to the test state characteristic parameter data and the test target threshold.

The test failure mode may include, but is not limited to: modes of different failure degrees.

In one embodiment, a test failure mode of the wind turbine may be determined based on a comparison of the test state characteristic parameter data with a test target threshold. An accurate result is determined if the test failure mode of the wind turbine corresponds to a known failure mode, otherwise an inaccurate result is determined.

In some embodiments, the test condition characteristic parameter data is compared to a test target threshold to determine a fault condition of the wind turbine. The dimension of the test state characteristic parameter data is consistent with the dimension of the state characteristic parameter data to be monitored, the characteristic parameter corresponding to the test state characteristic parameter data is consistent with the characteristic parameter corresponding to the state characteristic parameter data to be monitored, for example, one or more historical data of a jitter value, an amplitude value, a statistic and an energy value of a vibration signal is used as the test state characteristic parameter data.

Step 604, determining whether the accuracy of the test failure mode compared with the known failure mode is up to standard, if not, indicating that the accuracy of the test failure mode compared with the known failure mode is not up to standard, executing step 605; if yes, it indicates that the accuracy of the test failure mode is up to standard compared to the known failure mode, go to step 606. Therefore, the accuracy is used as a target function for feedback control, the accuracy of the threshold value determination model can be improved, and the effectiveness of the threshold value determination model is ensured.

And 605, if the accuracy of the test fault mode is not up to the standard compared with the known fault mode, re-segmenting the value domain interval of the continuous working condition parameters to obtain a new multi-segment interval, re-determining the value of K, re-classifying the parameter data samples, and re-determining various target thresholds so as to modify the threshold determination model. This accuracy can be set according to the industrial requirements. The higher the accuracy of the general industry needs, the larger the accuracy value. For example, the accuracy may be 80%. Such that the failure of the accuracy of the test failure mode compared to the known failure mode may be an accuracy of the test failure mode compared to the known failure mode of less than or equal to 80%.

Step 606, a threshold determination model is obtained. Therefore, the threshold determination model is obtained by self-adapting various target thresholds, the working condition of the wind driven generator can be more accurately reflected, the fault condition of the wind driven generator is analyzed, and the accuracy of fault diagnosis of the wind driven generator is further improved.

In some embodiments, the accurate results may be counted, and the ratio of the count of accurate results to the total number of test samples determined as the accuracy of the test failure mode compared to the known failure mode.

In some embodiments, the accuracy of the test failure mode as compared to the known failure modes is determined as follows, as in equations (12) and (13):

t＝t₁+t₂ (13)

where η is the accuracy of the test failure mode compared to the known failure mode, t is the total number of test samples, t is₁To test the failure mode for compliance with a known failure mode, t₂To test failure modes that do not correspond to known failure modes.

In embodiments of the present application, the adjusted threshold determination model is verified as being optimal by testing the accuracy of the failure mode as compared to known failure modes. If the accuracy of the test fault mode is not up to standard compared with the accuracy of the known fault mode, the distance from the parameter data sample to the clustering center in the class can be shown, and the distance from the parameter data sample to the clustering center is different compared with the distance from the actual sample, so that the threshold determination model can be adjusted, new multi-segment intervals are obtained by executing interval segmentation again on the value domain intervals of the continuous working condition parameters, the threshold determination model is modified until the accuracy of the test fault mode is up to standard compared with the accuracy of the known fault mode, and the threshold determination model is obtained. Therefore, the threshold determination model can be corrected through testing the sample, and the optimal threshold determination model is obtained.

FIG. 9 is a block diagram of a wind turbine condition monitoring system 70 according to the present application. The wind turbine condition monitoring system 70 includes one or more processors 701 for implementing the method as described above.

The wind turbine condition monitoring system 70 includes one or more processors 701 for implementing the monitoring method as described above. In some embodiments, wind turbine condition monitoring system 70 may include a readable storage medium 705, where readable storage medium 705 may store a program that may be invoked by processor 701, and may include a non-volatile storage medium.

In some embodiments, wind turbine condition monitoring system 70 may include a memory 704 and an interface 703.

In some embodiments, the wind turbine condition monitoring system 70 may also include other hardware depending on the application.

The readable storage medium 705 of the embodiment of the present application stores thereon a program, which when executed by the processor 701, is used to implement the monitoring method as described above.

This application may take the form of a computer program product that is embodied on one or more readable storage media 705 (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) having program code embodied therein. Readable storage media 705, which include both permanent and non-permanent, removable and non-removable media, may implement information storage by any method or technology. The information may be computer readable instructions, data structures, modules of a program, or other data. Examples of readable storage medium 705 include, but are not limited to: phase change memory (PRAM), Static Random Access Memory (SRAM), Dynamic Random Access Memory (DRAM), other types of Random Access Memory (RAM), Read Only Memory (ROM), Electrically Erasable Programmable Read Only Memory (EEPROM), flash memory or other memory technologies, compact disc read only memory (CD-ROM), Digital Versatile Discs (DVD) or other optical storage, magnetic tape storage or other magnetic storage devices, or any other non-transmission medium, may be used to store information that may be accessed by a computing device. The above description is only a preferred embodiment of the present disclosure, and should not be taken as limiting the present disclosure, and any modifications, equivalents, improvements, etc. made within the spirit and principle of the present disclosure should be included in the scope of the present disclosure.

The above description is only a preferred embodiment of the present disclosure, and should not be taken as limiting the present disclosure, and any modifications, equivalents, improvements, etc. made within the spirit and principle of the present disclosure should be included in the scope of the present disclosure.

It should also be noted that the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising an … …" does not exclude the presence of other like elements in a process, method, article, or apparatus that comprises the element.

Claims

1. A method of monitoring, the method comprising:

2. The method of claim 1, wherein the method comprises:

3. The method of claim 2, wherein classifying the parametric data samples for the wind turbine operating conditions into a plurality of classes comprises:

4. The method of claim 3, wherein the plurality of operating condition parameters include continuous operating condition parameters, and the continuous operating condition parameters are operating condition parameters whose value attribute characterizes the operating condition parameters' value as a continuous value;

5. The method of claim 4, wherein the plurality of operating condition parameters includes a plurality of the continuous operating condition parameters;

6. The method of claim 3, wherein the plurality of operating condition parameters include discrete operating condition parameters, and the discrete operating condition parameters are operating condition parameters whose value attribute characterizes the operating condition parameters as discrete values;

7. The method of claim 6, wherein the plurality of operating condition parameters includes a plurality of the discrete operating condition parameters;

8. The method of claim 6, wherein the plurality of operating condition parameters include continuous operating condition parameters, and the continuous operating condition parameters are operating condition parameters whose value attribute characterizes the operating condition parameters' value as a continuous value;

9. The method of claim 4 or 8, wherein the method comprises:

10. The method of claim 2, wherein said determining said target threshold values for each class from said parametric data samples in each class comprises:

11. The method of claim 10, wherein a plurality of said parametric data samples in each class comprise cluster center samples as cluster centers and cluster element samples outside of cluster centers;

12. The method of claim 11, wherein said determining distances between said cluster element samples in each class to said cluster center samples in a class comprises:

normalizing the first distance to obtain a normalized second distance;

13. The method of any of claims 11-12, wherein weighting a plurality of the initial thresholds in each class, determining the target threshold for each class, comprises:

14. The method of claim 2, wherein prior to classifying the parametric data samples for the wind turbine operating conditions into a plurality of classes, the method further comprises:

15. A readable storage medium, on which a computer program is stored, which, when being executed by a processor, carries out the monitoring method according to any one of claims 1 to 14.

16. A wind turbine condition monitoring system comprising one or more processors configured to implement a monitoring method according to any one of claims 1 to 14.

17. A wind power generator, characterized in that it comprises:

a tower;

a nacelle mounted to the tower;

a wind wheel assembled to the nacelle;

wind turbine condition monitoring system for implementing a monitoring method according to any of claims 1 to 14.