Method for monitoring a variable pitch system, system and computer readable storage medium

Abstract

The embodiment of the application provides a kind of variable pitch system monitoring method and its system and computer readable storage medium.The monitoring method includes: selecting multiple monitoring variables of the variable pitch system of wind turbine generator set;Obtain the monitoring data of multiple monitoring variables;Based on the monitoring data of multiple monitoring variables, Cox proportional hazards model is established and trained, to obtain the Cox proportional hazards model after training;Obtain the current monitoring data of multiple monitoring variables in the actual operation process of wind turbine generator set;And based on the current monitoring data of multiple monitoring variables and the Cox proportional hazards model after training, the current risk rate of variable pitch system is obtained to monitor variable pitch system.The embodiment of the application can better monitor the failure of variable pitch system, reduce false positive rate.

Description

TECHNICAL FIELD

[0001] The embodiment of the present application relates to the technical field of wind power generation, in particular to a monitoring method of a variable pitch system, a system thereof and a computer readable storage medium. BACKGROUND

[0002] With the gradual exhaustion of energy such as coal and oil, human beings pay more and more attention to the use of renewable energy. As a clean renewable energy, wind energy is increasingly valued by countries around the world. With the continuous development of wind power technology, wind turbine generators are increasingly used in power systems. The wind turbine generator is a large device for converting wind energy into electric energy, and is usually set in an area rich in wind energy resources.

[0003] The variable pitch system, as an important part of the wind turbine generator, plays a crucial role in ensuring the normal operation of the wind turbine generator. Hydraulic variable pitch is one of the commonly used variable pitch methods of the variable pitch system. Due to the harsh operating environment of the wind turbine generator and improper maintenance and protection, the variable pitch system will frequently malfunction, which is not conducive to the stable operation of the wind turbine generator.

[0004] At present, the biggest problem of the oil pressure fault diagnosis of the variable pitch system is that the false positive rate of the wind turbine generator is high. During the actual operation of the wind turbine generator, there will be a lot of uncertainties, such as sensor abnormalities, external environmental interference, etc., which will cause fluctuations in the data. SUMMARY

[0005] The purpose of the embodiment of the present application is to provide a monitoring method of a variable pitch system, a system thereof and a computer readable storage medium, which can better monitor the fault of the variable pitch system and reduce the false positive rate.

[0006] One aspect of the embodiment of the present application provides a monitoring method of a variable pitch system, which is applied to a wind turbine generator. The monitoring method comprises: selecting a plurality of monitoring variables affecting the variable pitch system of the wind turbine generator; obtaining monitoring data of the plurality of monitoring variables; establishing and training a Cox proportional hazards model based on the monitoring data of the plurality of monitoring variables to obtain a trained Cox proportional hazards model; obtaining current monitoring data of the plurality of monitoring variables in the actual operation process of the wind turbine generator; and obtaining a current hazard rate of the variable pitch system based on the current monitoring data of the plurality of monitoring variables and the trained Cox proportional hazards model to monitor the variable pitch system.

[0007] Further, the obtaining of the monitoring data of the plurality of monitoring variables comprises: extracting the monitoring data of the plurality of monitoring variables from a SCADA system of the wind turbine generator.

[0008] Further, the wind turbine generator set comprises a faulty wind turbine generator set and a healthy wind turbine generator set, and the extracting the monitoring data of the plurality of monitoring variables from the SCADA system of the wind turbine generator set comprises: extracting the monitoring data of the plurality of monitoring variables of the faulty wind turbine generator set and the monitoring data of the plurality of monitoring variables of the healthy wind turbine generator set from the SCADA system of the wind turbine generator set respectively.

[0009] Further, the selecting the plurality of monitoring variables affecting the pitch system of the wind turbine generator set comprises: selecting the plurality of monitoring variables affecting the oil pressure state of the pitch system of the wind turbine generator set.

[0010] Further, the plurality of monitoring variables comprises the pressure of the plurality of blades, the position angle of the plurality of blades, the generator speed and the wind speed of the wind turbine generator set.

[0011] Further, the establishing and training the Cox proportional hazards model based on the monitoring data of the plurality of monitoring variables comprises: integrating the pressure of the plurality of blades and the position angle of the plurality of blades respectively to obtain integrated variables; taking the integrated variables, the generator speed and the wind speed as modeling input variables of the Cox proportional hazards model; and training the Cox proportional hazards model based on the data of the modeling input variables.

[0012] Further, the integrated variables comprise the mean of the sum of the pressure of the plurality of blades, the mean of the sum of the position angle of the plurality of blades, the mean of the difference of the pressure of the plurality of blades and the mean of the difference of the position angle of the plurality of blades.

[0013] Further, the wind turbine generator set comprises a faulty wind turbine generator set and a healthy wind turbine generator set, and the training the Cox proportional hazards model based on the data of the modeling input variables comprises: training the Cox proportional hazards model based on the data of the modeling input variables of the faulty wind turbine generator set and the data of the modeling input variables of the healthy wind turbine generator set.

[0014] Further, the training the Cox proportional hazards model based on the data of the modeling input variables comprises: performing a single factor analysis on the Cox proportional hazards model to determine the p-value of each modeling input variable; eliminating the modeling input variable with a p-value greater than a predetermined threshold; continuing to train the Cox proportional hazards model based on the data of the remaining modeling input variables until the p-value of each modeling input variable is less than or equal to the predetermined threshold; and taking the modeling input variable with a p-value less than or equal to the predetermined threshold as a modeling input variable of the trained Cox proportional hazards model.

[0015] Further, the training of the Cox proportional hazards model based on the data of the modeling input variables further comprises: introducing key variables of failure mechanisms of the variable pitch system, the key variables being modeling input variables with p values greater than the predetermined threshold; and taking the key variables and modeling input variables with p values less than or equal to the predetermined threshold as final modeling input variables of the trained Cox proportional hazards model.

[0016] Further, the training of the Cox proportional hazards model based on the data of the modeling input variables further comprises: solving partial regression coefficients of each modeling input variable in the trained Cox proportional hazards model by establishing a partial likelihood function and using an iterative method.

[0017] Further, the predetermined threshold is 0.05.

[0018] Further, the monitoring method further comprises: determining a predetermined early warning threshold based on a hazard rate of the variable pitch system when a hazard actually occurs; and monitoring the variable pitch system based on the obtained current hazard rate and the predetermined early warning threshold.

[0019] Another aspect of the embodiments of the present application also provides a monitoring system of a variable pitch system. The monitoring system comprises one or more processors for implementing the monitoring method of the variable pitch system as described in the above various embodiments.

[0020] Another aspect of the embodiments of the present application also provides a computer readable storage medium having a program stored thereon, the program being executed by a processor to implement the monitoring method of the variable pitch system as described in the above various embodiments.

[0021] The monitoring method of the variable pitch system, the system thereof and the computer readable storage medium of one or more embodiments of the present application can improve the recognition accuracy of the model and reduce the false positive rate of the variable pitch system by introducing the Cox proportional hazards model into the monitoring of the variable pitch system and using the Cox proportional hazards model to establish the influence of multiple factors of the variable pitch system on the variable pitch failure. BRIEF DESCRIPTION OF DRAWINGS

[0022] Figure 1 A flowchart of the monitoring method of the variable pitch system of one embodiment of the present application;

[0023] Figure 2 A monitoring effect diagram of the Cox proportional hazards model in actual wind turbine data of one embodiment of the present application;

[0024] Figure 3 A schematic block diagram of the monitoring system of the variable pitch system of one embodiment of the present application. DETAILED DESCRIPTION

[0025] The exemplary embodiments will be described in detail herein with reference to the attached drawings. The description of the exemplary embodiments is intended to apply to all alternative embodiments, as would be understood by persons skilled in the art. To that end, the following description is not intended to limit the exemplary embodiments to a particular hardware or software configuration. Rather, the following description is intended to describe the exemplary embodiments in sufficient detail to enable persons skilled in the art to practice the exemplary embodiments.

[0026] The terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the present application. Unless otherwise defined, technical terms or scientific terms used herein are intended to have the meanings commonly understood by one of ordinary skill in the art to which this application pertains. The detailed description includes specific details for the purpose of providing a thorough understanding of the exemplary embodiments. However, it will be apparent to those skilled in the art that the exemplary embodiments can be practiced without these specific details. In some instances, well-known structures and components are shown in block diagram form in order to avoid obscuring the concepts of the exemplary embodiments. Like numbers refer to like elements throughout.The use of "first", "second", and "third" or similar terminology does not imply an order or sequence unless specifically stated. The terms "first", "second", and "third" are used herein merely as labels to avoid confusion with one another, but are not intended to impose or imply any order, quantity, or importance in the specification or claims. Similarly, the terms "one", "another", and "an" do not imply one or more than one, unless specifically stated. The use of "a" or "an" does not restrict the meaning to "one", unless specifically stated. The terminology of "front", "rear", "upper", "lower", and the like, merely describe the relative positions of the components, and are not meant to limit the orientation of the components. The terminology of "include", "including", and "includes" or the like means encompassing or covering, but does not exclude other components or steps. The terminology of "connect", "connected", and "connecting" or the like means an electrical connection, either direct or indirect, and is not limited to a physical or mechanical connection. The use of the singular herein, e.g., "a", "said", and "the", means "one or more" unless specifically stated otherwise. The use of the term "and / or" includes any and all combinations of one or more of the associated listed items.

[0027] The present embodiments provide a monitoring method of a variable pitch system, which is applied to a wind turbine generator system. Figure 1 A flowchart of the monitoring method of the variable pitch system of one embodiment of the present application is disclosed. As shown in FIG. 1, the monitoring method of the variable pitch system of one embodiment of the present application can include steps S11 to S15. Figure 1

[0028] In step S11, a plurality of monitoring variables affecting the variable pitch system of the wind turbine generator system are selected.

[0029] In some embodiments, the oil pressure state of the variable pitch system of the wind turbine generator system can be monitored. Therefore, a plurality of monitoring variables affecting the oil pressure state of the variable pitch system of the wind turbine generator system can be selected.

[0030] In one embodiment, the plurality of monitoring variables affecting the pitch system oil pressure state may, for example, include the pressure of a plurality of blades of the wind turbine generator unit, the position angle of the plurality of blades, the generator speed, and the wind speed. For example, the wind turbine generator unit generally includes three blades, blade A, blade B, and blade C, and thus, the plurality of monitoring variables affecting the pitch system oil pressure state of the wind turbine generator unit are as shown in Table 1 below:

[0031] Table 1

[0032] Monitoring variable name Monitoring variable description PresA Pressure at blade A PresB Pressure at blade B PresC Pressure at blade C PitcPosA Position angle at blade A PitcPosB Position angle at blade B PitcPosC Position angle at blade C GenRpm Generator speed wind Wind speed

[0033] In step S12, monitoring data of the plurality of monitoring variables is acquired.

[0034] The monitoring data of the plurality of monitoring variables can be extracted from a SCADA (Supervisory Control And Data Acquisition) system of the wind turbine generator unit. The extracted monitoring data of the plurality of monitoring variables includes monitoring data of the plurality of monitoring variables of a faulty wind turbine generator unit (referred to as a faulty unit) and monitoring data of the plurality of monitoring variables of a healthy wind turbine generator unit (referred to as a healthy unit).

[0035] For the faulty unit, monitoring data of the plurality of monitoring variables at the time of the fault occurrence can be extracted as fault data, and data of the plurality of monitoring variables at a random time within one year before the fault occurrence can be extracted as health data. For the healthy unit, the end time of the monitoring data is consistent, and if only the monitoring data of the monitoring variables at this time is extracted as health data, the monitoring data of the monitoring variables changes monotonically, and overfitting phenomenon is likely to occur, i.e., the health status of the wind turbine generator unit identified by the subsequently established model is too single. Therefore, the monitoring data of the monitoring variables at a random time within a health time period can be additionally extracted as health data for the healthy unit.

[0036] In step S13, a Cox proportional hazards model is established and trained based on the monitoring data of the plurality of monitoring variables to obtain a trained Cox proportional hazards model.

[0037] The Cox proportional hazards model is a semi-parametric regression model proposed by D.R. Cox, a British statistician, in 1972. The Cox proportional hazards model takes survival outcome and survival time as the dependent variable, and can analyze the influence of numerous factors on survival period. The basic form of the Cox proportional hazards model is as follows:

[0038] h(t,X)=h0(t)g(X)

[0039] To ensure g(X)>0, g(X) is usually taken as Thus, the common Cox proportional hazards model is obtained as follows:

[0040] h(t,X) = h0(t)exp(β1X1+β2X2+…β k X k )

[0041] where β1, β2, β k are the partial regression coefficients of the independent variables, which are parameters to be estimated from the sample data; h0(t) is the baseline hazard rate of h(t,X), which is a quantity to be estimated from the sample data. The above formula is referred to as the Cox proportional hazards model.

[0042] The monitoring method of the pitch system according to the embodiments of the present application can introduce the Cox proportional hazards model into the hydraulic fault diagnosis of the pitch system, and use the Cox proportional hazards model to establish the influence of multiple factors of the pitch system on the low-pressure fault of the pitch blade.

[0043] The following will describe in detail how to establish and train the Cox proportional hazards model based on the monitoring data of multiple monitoring variables, so as to obtain the trained Cox proportional hazards model.

[0044] Considering the similarity of the operating conditions of multiple (for example, three) blades of the wind turbine generator, the pressure of the multiple blades and the position angle of the multiple blades can be integrated respectively, and the same variables of the multiple blades are averaged into a single variable, so that the integrated variables can be obtained. The integrated variables may, for example, include the average of the sum of the pressures of the multiple blades, the average of the sum of the position angles of the multiple blades, the average of the difference between the pressures of the multiple blades, and the average of the difference between the position angles of the multiple blades.

[0045] The integrated variables, the generator speed and the wind speed can be used as the modeling input variables of the Cox proportional hazards model. For example, the average of the sum of the pressures of three blades A, B and C, the average of the sum of the position angles of the three blades A, B and C, the average of the difference between the pressures of the three blades A, B and C, the average of the difference between the position angles of the three blades A, B and C, the generator speed and the wind speed can be used as the modeling input variables of the Cox proportional hazards model.

[0046] The Cox proportional hazards model can be trained based on the data of the modeling input variables. The hazard rate of a wind turbine at a certain time is estimated according to the failure data and health data in the historical data of the wind turbine. Therefore, the data of the modeling input variables can include the data of the modeling input variables of the failed wind turbines and the data of the modeling input variables of the healthy wind turbines. In addition, in order to enable the Cox proportional hazards model to accurately analyze the differences between the variable characteristics at the failure time and the variable characteristics at the healthy time, the failure data and the health data can be marked respectively, that is, the data of the modeling input variables of the failed wind turbines and the data of the modeling input variables of the healthy wind turbines can be marked respectively, wherein the data of the modeling input variables of the failed wind turbines can be marked as 1, and the data of the modeling input variables of the healthy wind turbines can be marked as 0. The Cox proportional hazards model can be trained based on the data of the modeling input variables of the failed wind turbines and the data of the modeling input variables of the healthy wind turbines as a training set.

[0047] How to train the Cox proportional hazards model based on the data of the modeling input variables will be described in detail below.

[0048] In order to compare the influence of the introduction of different modeling input variables on the Cox proportional hazards model, a single factor analysis can be performed on the Cox proportional hazards model to determine the p value of each modeling input variable, and the p value can be used to evaluate the performance of a single modeling input variable.

[0049] By comparing the p values of the modeling input variables, if the p values are large, the introduction of these modeling input variables is likely to significantly increase the uncertainty of the Cox proportional hazards model. Therefore, the modeling input variables with p values greater than a predetermined threshold value can be removed, and the modeling input variables with small p values can be used as the modeling input variables of the subsequent Cox proportional hazards model, which will be more conducive to determining the improvement direction of the Cox proportional hazards model. The predetermined threshold value can be 0.05, for example.

[0050] Specifically, after removing the modeling input variables with p values greater than 0.05 each time, the inputs of the Cox proportional hazards model are updated, and the Cox proportional hazards model is continuously trained based on the data of the remaining modeling input variables, until all the modeling input variables meet the requirement that the p values are less than or equal to the predetermined threshold value, for example, the p values are less than or equal to 0.05. The modeling input variables with p values less than or equal to the predetermined threshold value, for example, 0.05, are used as the modeling input variables of the trained Cox proportional hazards model.

[0051] In the training process of the above Cox proportional hazards model, some key variables related to the failure mechanism of the variable pitch system can be excluded from the modeling input variables because their p-value is greater than 0.05, but these key variables can be related to the failure mechanism according to engineering experience, so in some embodiments, some key variables of the failure mechanism of the variable pitch system can be introduced according to engineering experience, which are modeling input variables whose p-value is greater than a predetermined threshold, for example 0.05, in the training process, and the modeling input variables whose p-value is less than or equal to the predetermined threshold, for example 0.05, are finally retained, and the key variables and the finally retained modeling input variables whose p-value is less than or equal to the predetermined threshold, for example 0.05, are used as the final modeling input variables of the trained Cox proportional hazards model.

[0052] After determining the final modeling input variables of the trained Cox proportional hazards model, the partial regression coefficients of each modeling input variable in the trained Cox proportional hazards model can be solved by establishing a partial likelihood function and using an iterative method.

[0053] Specifically, the Cox proportional hazards model has two basic assumptions:

[0054] (1) Proportional hazards assumption: the influence of each factor on the survival life does not change with time, that is, h(t,X) / h0(t) does not change with time. This assumption is a prerequisite for establishing the Cox proportional hazards model, and also provides greater flexibility for model processing problems.

[0055] (2) Log-linear assumption: taking the logarithm of both sides of the equation, we get

[0056] lnh(t,X)=lnh0(t)+X1β1+X2β2+…+X k β k

[0057] It can be seen that each factor in the Cox proportional hazards model is linearly related to the logarithmic risk rate.

[0058] The Cox proportional hazards model does not assume any distribution of the result variable (survival time), so it cannot establish a likelihood function based on the distribution of the result variable like a parametric model. Therefore, for its parameter estimation, the method of partial likelihood estimation is introduced, and the construction of the Cox likelihood function is determined by the events and their occurrence order, so the data needs to be sorted by survival time before the likelihood function is constructed, that is, t1<t2<…<t m The Cox likelihood function is expressed as:

[0059]

[0060] where m represents that there are m different survival times in the data sample, t iSurvival time of the i th sample, R(t i ) is a hazard set composed of samples with survival time greater than or equal to t i , δ i is a mark variable, δ i = 0 indicates that the sample data is a censored data (i.e. normal data), δ i = 1 indicates that the sample event occurs (i.e. failure data), L i (β) represents the probability of the j th sample occurring at the risk at time t i .

[0061] It can be seen that the likelihood function L(β) is the cumulative multiplication of the probabilities L i (β) at all survival time points. Taking the logarithm of the likelihood function L(β) obtains:

[0062]

[0063] Taking the partial derivative of β obtains:

[0064]

[0065] If taking , then

[0066]

[0067] wherein k is the number of variables.

[0068] By solving the k nonlinear equations, the maximum likelihood estimation of the partial regression coefficient β of each variable can be obtained, and the solving process is usually carried out by using an iterative method for numerical solution.

[0069] Therefore, through the above process, the partial regression coefficient of each modeling input variable in the trained Cox proportional risk model can be solved.

[0070] Continuing to refer to Figure 1 , in step S14, current monitoring data of multiple monitoring variables in the actual operation process of the wind turbine generator set is obtained.

[0071] In step S15, the current hazard rate of the variable pitch system is obtained based on the current monitoring data of the multiple monitoring variables and the trained Cox proportional risk model to monitor the variable pitch system.

[0072] In the monitoring method of the variable pitch system of the embodiment of the application, the SCADA operation data of the wind turbine generator set is taken as input, a multivariate parameter estimation Cox proportional risk model is established, that is, by analyzing the wind turbine generator set variables monitored by the SCADA, the internal relationship between the variables is deduced, so as to represent the main characteristics of the sample, and then the general characteristics and trends of the sample are analyzed.

[0073] In some embodiments, the monitoring method of the pitch system of the embodiments of the present application can further include: determining a predetermined early warning threshold based on a risk rate when the pitch system actually occurs in danger; and monitoring the pitch system based on the obtained current risk rate and the predetermined early warning threshold.

[0074] The embodiments of the present application can improve the identification accuracy of the model and reduce the false alarm rate of the pitch system by introducing the Cox proportional hazards model into the monitoring method of the pitch system, such as the diagnosis of the oil pressure state of the pitch system, and using the Cox proportional hazards model to establish the influence of multiple factors of the pitch system on the low pressure failure of the pitch blade.

[0075] Figure 2 A monitoring effect schematic diagram of the Cox proportional hazards model of one embodiment of the present application in actual wind turbine data is disclosed. Figure 2 In the figure, the abscissa represents time, Figure 2 The data of the wind turbine from 2018-07 to 2020-01 is monitored, and the ordinate represents the risk rate corresponding to each time point. As Figure 2 shown, D1 and D2 are two thresholds set, two levels of alarms are set, wherein D1 represents the early warning line, D2 represents the alarm line, the severity of D2 is higher than that of D1, h represents the actual risk rate obtained by using the Cox proportional hazards model of the embodiments of the present application, and T represents the actual alarm time when the pitch system of the wind turbine actually fails. From Figure 2 it can be seen that the actual risk rate obtained by using the Cox proportional hazards model of the embodiments of the present application can produce an alarm in advance before the actual alarm time T when the pitch system actually fails. Therefore, the Cox proportional hazards model of the embodiments of the present application has high identification accuracy.

[0076] The embodiments of the present application also provide a monitoring system 200 of a pitch system, which is applied to a wind turbine. Figure 3 A schematic block diagram of the monitoring system 200 of the pitch system of one embodiment of the present application is disclosed. As Figure 3 shown, the monitoring system 200 of the pitch system can include one or more processors 201 for implementing the monitoring method of the pitch system described in any of the above embodiments. In some embodiments, the monitoring system 200 of the pitch system can include a computer readable storage medium 202, which can store programs that can be called by the processor 201, and can include a non-volatile storage medium. In some embodiments, the monitoring system 200 of the pitch system can include a memory 203 and an interface 204. In some embodiments, the monitoring system 200 of the pitch system of the embodiments of the present application can further include other hardware according to actual application.

[0077] The monitoring system 200 of the embodiment of the variable pitch system has similar beneficial technical effects to the monitoring method of the variable pitch system described above, and thus will not be described here again.

[0078] The embodiment of the present application also provides a computer readable storage medium. The computer readable storage medium stores a program, and the program is executed by a processor to implement the monitoring method of the variable pitch system according to any one of the above embodiments.

[0079] The embodiment of the present application can be in the form of a computer program product implemented on one or more storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing program codes. The computer readable storage medium includes permanent and non-permanent, removable and non-removable media, and can be implemented by any method or technology to store information. The information can be computer readable instructions, data structures, program modules or other data. Examples of computer readable storage media include, but are not limited to: new memory such as phase change memory / resistive memory / magnetic memory / ferroelectric memory (PRAM / RRAM / MRAM / FeRAM), 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 technology, compact disc read only memory (CD-ROM), digital versatile disc (DVD) or other optical storage, magnetic cassette, magnetic tape disk storage or other magnetic storage device, or any other non-transmission medium that can be used to store information accessible by a computing device.

[0080] The above describes the monitoring method of the variable pitch system, the system thereof and the computer readable storage medium provided by the embodiment of the present application in detail. The specific examples are applied in this paper to describe the monitoring method of the variable pitch system, the system thereof and the computer readable storage medium of the embodiment of the present application. The above embodiment is only used to help understand the core idea of the present application, and does not limit the present application. It should be pointed out that, for those skilled in the art, without departing from the spirit and principles of the present application, some improvements and modifications can be made to the present application, and these improvements and modifications should also fall within the protection scope of the appended claims of the present application.

Claims

1. A method for monitoring a variable pitch system applied to a wind turbine generator system, characterized in that: It comprises: selecting a plurality of monitoring variables affecting the oil pressure state of a variable pitch system of a wind turbine generator set, the plurality of monitoring variables including pressures of a plurality of blades of the wind turbine generator set, position angles of the plurality of blades, a generator speed, and a wind speed; obtaining monitoring data of the plurality of monitoring variables; establishing and training a Cox proportional hazards model based on the monitoring data of the plurality of monitoring variables to obtain a trained Cox proportional hazards model, the establishing and training of the Cox proportional hazards model based on the monitoring data of the plurality of monitoring variables comprising: integrating the pressures of the plurality of blades and the position angles of the plurality of blades respectively to obtain integrated variables; based on the integrated variables, the generator speed, and the wind speed as modeling input variables of the Cox proportional hazards model; and training the Cox proportional hazards model based on data of the modeling input variables, wherein the training of the Cox proportional hazards model based on data of the modeling input variables comprises: performing a single factor analysis on the Cox proportional hazards model to determine a p-value of each modeling input variable; eliminating a modeling input variable with a p-value greater than a predetermined threshold; continuing to train the Cox proportional hazards model based on data of the remaining modeling input variables until the p-values of the modeling input variables are all less than or equal to the predetermined threshold; and taking the modeling input variables with p-values less than or equal to the predetermined threshold as modeling input variables of the trained Cox proportional hazards model; the training of the Cox proportional hazards model based on data of the modeling input variables further comprises: introducing a key variable of a failure mechanism of the variable pitch system, the key variable being a modeling input variable with a p-value greater than the predetermined threshold; and taking the key variable and the modeling input variables with p-values less than or equal to the predetermined threshold as final modeling input variables of the trained Cox proportional hazards model; obtaining current monitoring data of the plurality of monitoring variables in an actual operation process of the wind turbine generator set; and obtaining a current hazard rate of the variable pitch system based on the current monitoring data of the plurality of monitoring variables and the trained Cox proportional hazards model to monitor the variable pitch system.

2. The monitoring method of claim 1, wherein: The obtaining of the monitoring data of the plurality of monitoring variables comprises: extracting the monitoring data of the plurality of monitoring variables from a SCADA system of the wind turbine generator set.

3. The monitoring method of claim 2, wherein: The wind turbine generator set comprises a failure unit and a healthy unit, and the extracting of the monitoring data of the plurality of monitoring variables from the SCADA system of the wind turbine generator set comprises: extracting the monitoring data of the plurality of monitoring variables of the failure unit and the monitoring data of the plurality of monitoring variables of the healthy unit from the SCADA system of the wind turbine generator set respectively.

4. The monitoring method of claim 1, wherein: The integrated variables comprise a mean value of a sum of the pressures of the plurality of blades, a mean value of a sum of the position angles of the plurality of blades, a mean value of a difference between the pressures of the plurality of blades, and a mean value of a difference between the position angles of the plurality of blades.

5. The monitoring method of claim 4, wherein: The wind turbine generator set comprises a failure unit and a healthy unit, and the training of the Cox proportional hazards model based on data of the modeling input variables comprises: training the Cox proportional hazards model based on data of modeling input variables of failed units and data of modeling input variables of healthy units.

6. The monitoring method of claim 1, wherein: training the Cox proportional hazards model based on data of the modeling input variables further comprises: solving the partial regression coefficients of each modeling input variable in the trained Cox proportional hazards model by an iterative method through establishing a partial likelihood function.

7. The monitoring method of claim 1, wherein: the predetermined threshold is 0.

05.

8. The monitoring method of claim 1, wherein: further comprising: determining a predetermined early warning threshold based on the hazard rate when the risk actually occurs to the pitch system; and monitoring the pitch system based on the obtained current hazard rate and the predetermined early warning threshold.

9. A monitoring system of a variable pitch system, characterized by: one or more processors for implementing the monitoring method of the pitch system according to any one of claims 1-8.

10. A computer-readable storage medium, characterized in that, a program stored thereon, which, when executed by a processor, implements the monitoring method of the pitch system according to any one of claims 1-8.

Images