Fault judgment method, device and equipment for yaw system of wind turbine generator and storage medium

Abstract

The invention provides a wind turbine generator yaw system fault diagnosis method, device and equipment and a storage medium, and the method specifically comprises the steps: collecting an axial vibration signal and a transverse vibration signal of a cabin under the operation state of a wind turbine generator, firstly employing an improved adaptive noise complete set empirical mode decomposition algorithm, and carrying out the fault diagnosis of the wind turbine generator yaw system; decomposing the two vibration signals to obtain a plurality of intrinsic mode function components; secondly, selecting components rich in fault information from the IMF components by using an energy ratio method; hilbert transform is carried out on the IMF components, integration is carried out on the IMF components, and Hilbert marginal spectrums corresponding to the axial and transverse vibration signals of the cabin are obtained; and finally, based on the fault corresponding characteristics, whether a fault characteristic frequency corresponding peak value exists in the Hilbert marginal spectrum is judged, and whether the yaw system has a fault is judged. According to the invention, vibration signal actual measurement and fault diagnosis of the actual wind turbine generator can be realized.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of wind power generation, in particular to a wind turbine yaw system fault judgment method, device, equipment and storage medium. BACKGROUND

[0002] The yaw system is one of the main structures of the wind turbine, and the yaw load thereof mainly comes from the wind wheel. The load generated by the wind wheel is transmitted to the yaw system through the hub, the main shaft and the main frame, and thus the imbalance of the wind wheel will have a direct impact on the yaw system. The yaw system failure will not only cause the wind turbine to fail to achieve the maximum energy conversion efficiency, but also may cause damage to other components, and thus it is necessary to diagnose the yaw system failure.

[0003] With the continuous development of fault diagnosis and early warning technology of wind turbines, advanced monitoring technology and diagnosis methods are applied to the analysis of wind turbines. Xu Shen et al. proposed a two-stage Attention-Informer algorithm to solve the problem of yaw position error accumulation. By using the ReliefF feature algorithm based on standardized interaction gain to identify feature parameters, a yaw system position prediction model was established, and accurate diagnosis of yaw position abnormalities was achieved. Scholars such as Zhao Hongshan established a three-closed-loop yaw motor servo system model including position loop, current loop and speed loop, calculated the residual error between the actual collected yaw system data and the output of the established dynamic model, used appropriate variable residual error as fault detection quantity, and verified the effectiveness of the method in detecting three typical faults of yaw system transmission equipment jamming, yaw motor failure and sensor failure through Simulink simulation. Marcantonio Catelani et al. proposed a data-driven state monitoring and diagnosis method, which effectively monitors the health status of the yaw system of wind turbines and detects damage by measuring key parameters of the target system. Researchers such as Ali Ouanas designed a hybrid signal processing method, which significantly reduces the interference components in non-stationary signals by using a frequency domain filter preprocessor and a multi-scale decomposition module (DWT-EMD) method to process the output signals of the yaw drive motor inverter, successfully achieving fault diagnosis of the yaw drive induction motor. Li Yongzhan et al. used machine learning methods combined with support vector machine (SVM) models to extract key fault feature parameters such as time-frequency energy entropy and Mel cepstral coefficients from acoustic signals collected from the yaw system, successfully constructed an SVM classifier and conducted model training, and the experimental results showed that this method could effectively identify yaw system faults such as mechanical jamming and gear wear. Researchers such as Li Bin collected high-precision time series data of yaw system electrical parameters, established a dynamic torque fluctuation feature analysis model, used wavelet transform and long short-term memory network (LSTM) fusion algorithm, constructed a multi-parameter coupling analysis framework based on entropy method, and developed a yaw system health degree evaluation tool containing 12 core indicators, significantly improving the accuracy of technical improvement risk assessment. Scholars such as ZHANG Le arranged laser radar type anemometers to collect output power and yaw angle data of wind turbines, and divided them into 22 power intervals and 12 yaw angle sectors, based on which a three-dimensional yaw error map was constructed for fine adjustment of yaw angle, which could significantly improve the annual equivalent power generation (AEP) of wind turbines.Xu Jianei proposed a power trend detection method based on tip speed ratio optimization, through analyzing the dynamic range of unit power curve and tip speed ratio, locking the effective wind speed interval under the maximum wind energy capture mode, constructing the yaw angle-wind speed two-dimensional cumulative power model, using the Bessy theory to deduce that the optimal output power corresponds to the yaw angle, that is, the sensor zero offset, the model can still maintain high error detection precision under complex turbulent conditions. Wang Xiaohai et al. verified by large eddy simulation that with the increase of yaw angle, the negative pressure and laminar resistance coefficient in the tower interference zone present nonlinear change, and the flow influence is significantly weakened. This research result provides important theoretical support for the design of tower wind load under yaw state. Bo Jing proposed an autonomous detection method based on the maximum power capture theory in order to solve the problem of insufficient anti-interference ability of external data acquisition of the yaw system, defined two typical yaw modes of static and dynamic, and built an MPC detection framework based on SCADA data, realizing the synchronous identification of yaw angle deviation order and direction. Xie Baoyu et al. proposed an improved SOM discretization method, which effectively reduces the information loss in the discretization process by optimizing the structure and training mechanism of SOM, and combines the improved method with Bayesian network to improve the accuracy and reliability of fault diagnosis in the fault diagnosis of wind turbine yaw system.

[0004] At present, there are many diagnosis methods for the yaw system of the unit at home and abroad, but most of them only analyze the fault position, few methods study the root cause of the fault, and the load of the yaw system is from the wind wheel. Few methods are used to diagnose the fault caused by the wind wheel. SUMMARY

[0005] The purpose of the present application is to provide a wind turbine yaw system fault judgment method, which aims to realize the fault diagnosis of the actual wind turbine and realize the real measurement of the vibration signal.

[0006] The present application discloses a wind turbine yaw system fault judgment method, comprising:

[0007] Collecting the axial vibration signal and the transverse vibration signal of the wind wheel of the wind turbine in the running state;

[0008] Based on the improved adaptive noise complete ensemble empirical mode decomposition algorithm (ICEEMDAN), the axial vibration signal and the transverse vibration signal are decomposed to obtain a plurality of intrinsic mode functions (IMF);

[0009] Selecting the component rich in fault information from the plurality of IMF components by the energy ratio method;

[0010] The Hilbert spectrum is obtained by performing Hilbert transform on the IMF component, and then performing integration, so as to obtain the Hilbert marginal spectrum corresponding to the axial vibration signal and the transverse vibration signal of the wind turbine nacelle;

[0011] Based on the fault corresponding feature, it is judged whether there is a peak value corresponding to the fault feature frequency in the Hilbert marginal spectrum, and whether the yaw system of the wind turbine exists a fault.

[0012] The axial vibration signal and the transverse vibration signal of the wind turbine rotor under the running state are collected, including:

[0013] In the wind turbine nacelle, the first vibration sensor and the second vibration sensor are installed near the yaw motor and near the top of the nacelle and the electrical control cabinet to collect the axial and transverse vibration signals;

[0014] In the wind turbine, the remaining vibration sensors are installed in the low-speed shaft, the high-speed shaft and the gear box to collect the vibration signals of the low-speed shaft, the high-speed shaft and the gear box.

[0015] The axial vibration signal and the transverse vibration signal are decomposed based on the improved adaptive noise complete ensemble empirical mode decomposition algorithm (ICEEMDAN) to obtain a plurality of intrinsic mode functions (IMF), including:

[0016] The initial obtained axial vibration signal and transverse vibration signal are decomposed based on the adaptive noise complete ensemble empirical mode decomposition algorithm (CEEMDAN) to obtain a plurality of first IMF components and a first residual;

[0017] The standard deviation of the first residual is calculated, and if the first residual standard deviation is greater than a preset standard deviation threshold, the iterative decomposition is continued to obtain a plurality of second IMF components and a second residual;

[0018] The iteration is continued until the standard deviation of the Nth residual calculated is less than the preset standard deviation threshold, and the iteration is completed;

[0019] The plurality of Nth IMF components obtained by the Nth iteration are taken as the final intrinsic mode function (IMF).

[0020] The intrinsic mode function containing fault information is selected from the plurality of intrinsic mode functions (IMF) by the energy ratio method, including:

[0021] Suppose that the original signal x(t) is processed by the decomposition method to obtain N IMF components and a residual component r(t), and the expression is:

[0022]

[0023] On this basis, the total energy E of the signal total is defined as the sum of the energy of each IMF and the residual energy, that is

[0024]

[0025] The energy E of each IMF component is i calculated by the following integral formula:

[0026]

[0027] The energy E of the residual component is r then

[0028]

[0029] The energy ratio ER of the i-th IMF component is i defined as the ratio of its energy to the total energy, that is

[0030]

[0031] In the formula, x(t) is the original signal, representing the time series data to be analyzed, and N represents the total number of IMF components.

[0032] Among them, based on the Hilbert marginal spectrum, the intrinsic mode function screened is subjected to Hilbert transform to obtain the Hilbert marginal spectrum of the axial vibration signal and the transverse vibration signal corresponding to the wind turbine rotor, including

[0033] The conjugate signal of the i-th IMF component subjected to Hilbert transform is defined as

[0034]

[0035] Among them, P represents the Cauchy principal value integral. In this way, a complex analytic signal can be established.

[0036] Among them, the instantaneous amplitude a i (t) is defined as

[0037]

[0038] The instantaneous phase θ i (t) is defined as

[0039]

[0040] Further, by taking the derivative of the instantaneous phase, the instantaneous frequency can be obtained:

[0041]

[0042] The instantaneous amplitude and instantaneous frequency information obtained after the Hilbert transform is completed is used to construct a Hilbert spectrum of the signal, and the Hilbert spectrum H(w, t) is defined as:

[0043]

[0044] Wherein, δ is Dirac impact function, ensure that at the instantaneous frequency w i (t) has a non-zero contribution; the spectrum function can describe the instantaneous energy distribution of each frequency component of the signal in the time-frequency plane;

[0045] Integrating H(w, t) obtains a marginal spectrum h(w):

[0046]

[0047] The marginal spectrum converts the original three-dimensional time-frequency energy distribution into a two-dimensional frequency energy distribution, so that the energy distribution of the signal at different frequencies is more directly reflected.

[0048] Wherein, based on the Hilbert marginal spectrum of the axial vibration signal and the transverse vibration signal corresponding to the wind wheel of the wind turbine, it is judged whether the yaw system of the wind turbine is faulty, including:

[0049] Obtaining the fault feature of the yaw system of the wind turbine, and extracting the fault feature frequency information therefrom;

[0050] Judging whether there is a peak value corresponding to the fault feature frequency in the Hilbert marginal spectrum of the axial vibration signal and the transverse vibration signal of the wind wheel of the wind turbine;

[0051] If there is, it is determined that there is a fault.

[0052] Wherein, in the step of obtaining the fault feature of the yaw system of the wind turbine, the fault feature is extracted from the historical data in the SCADA database, or is obtained by simulating the simulation model of the yaw system of the wind turbine.

[0053] The application discloses a kind of wind turbine yaw system fault judging devices, including:

[0054] Signal acquisition module, for collecting the axial vibration signal and the transverse vibration signal of the wind wheel of wind turbine under operating state;

[0055] Signal decomposition module, for decomposing the axial vibration signal and the transverse vibration signal based on improved adaptive noise complete set empirical mode decomposition algorithm (ICEEMDAN), to obtain multiple intrinsic mode function (IMF);

[0056] Screening module, for selecting the component rich in fault information from multiple IMF components by energy ratio method;

[0057] a transformation module configured to perform Hilbert transformation on the screened IMF component to obtain a Hilbert spectrum, and then perform integration on the Hilbert spectrum to obtain a Hilbert marginal spectrum corresponding to the axial vibration signal and the lateral vibration signal of the nacelle of the wind turbine generator;

[0058] a fault judgment module configured to judge whether a peak value corresponding to a fault feature frequency exists in the Hilbert marginal spectrum based on the fault corresponding feature, and judge whether the yaw system of the wind turbine generator has a fault.

[0059] The application discloses a computer device, which comprises an input and output unit, a memory and a processor, the memory stores computer readable instructions, and the computer readable instructions are executed by the processor to make the processor execute steps in the method of the foregoing embodiments.

[0060] The application discloses a storage medium storing computer readable instructions, and the computer readable instructions are executed by one or more processors to make the one or more processors execute steps in the method of the foregoing embodiments.

[0061] Different from the prior art, the wind turbine generator yaw system fault judgment method of the application comprises the following steps: collecting axial vibration signals and lateral vibration signals of a wind wheel of a wind turbine generator in a running state; decomposing the axial vibration signals and the lateral vibration signals based on an improved adaptive noise complete ensemble empirical mode decomposition algorithm (ICEEMDAN) to obtain a plurality of intrinsic mode functions (IMFs); screening intrinsic mode functions containing fault information from the plurality of intrinsic mode functions through an energy ratio method; performing Hilbert transformation on the screened intrinsic mode functions based on a Hilbert marginal spectrum to obtain a Hilbert marginal spectrum corresponding to the axial vibration signals and the lateral vibration signals of the wind wheel of the wind turbine generator; and judging whether the yaw system of the wind turbine generator has a fault based on the Hilbert marginal spectrum corresponding to the axial vibration signals and the lateral vibration signals of the wind wheel of the wind turbine generator. Through the application, fault diagnosis of an actual wind turbine can be realized, and vibration signals can be measured. BRIEF DESCRIPTION OF DRAWINGS

[0062] The application will be further described below in combination with the drawings and embodiments, and the drawings are as follows:

[0063] Figure 1 Fig. 1 is a flowchart of a wind turbine generator yaw system fault judgment method of the application.

[0064] Figure 2 Fig. 2 is a logic diagram of an ICEEMDAN algorithm in the wind turbine generator yaw system fault judgment method of the application.

[0065] Figure 3It is a fault judgment schematic diagram of a wind turbine yaw system fault judgment method of the application.

[0066] Figure 4 It is a measurement point arrangement schematic diagram of vibration signal collection in a wind turbine yaw system fault judgment method of the application.

[0067] Figure 5 It is a wind turbine coordinate schematic diagram set in a wind turbine yaw system fault judgment method of the application.

[0068] Figure 6 It is a wind turbine blade force schematic diagram under different attack angles in a wind turbine yaw system fault judgment method of the application.

[0069] Figure 7 It is an ICEEMDAN decomposition schematic diagram of an axial vibration signal in a wind turbine yaw system fault judgment method of the application.

[0070] Figure 8 It is an ICEEMDAN decomposition schematic diagram of a transverse vibration signal in a wind turbine yaw system fault judgment method of the application.

[0071] Figure 9 It is a structure schematic diagram of a wind turbine yaw system fault judgment device of the application.

[0072] Figure 10 It is a structure schematic diagram of a non-transient computer readable storage medium storing computer instructions provided by the application. DETAILED DESCRIPTION

[0073] In order to have a clearer understanding of the technical features, objects and effects of the application, the specific embodiments of the application will be described in detail with reference to the drawings.

[0074] Please refer to Figure 1 The application discloses a wind turbine yaw system fault judgment method, which comprises the following steps.

[0075] Step S110: collecting axial vibration signals and transverse vibration signals of a wind turbine rotor in a running state.

[0076] In the wind turbine cabin, first and second vibration sensors are installed near the yaw motor and near the top of the cabin and the electrical control cabinet to collect the axial and transverse vibration signals; in the wind turbine, the remaining vibration sensors are installed in the low-speed shaft, high-speed shaft and gear box to collect the low-speed shaft, high-speed shaft and gear box vibration signals.

[0077] S120: Based on the improved adaptive noise complete ensemble empirical mode decomposition algorithm (ICEEMDAN), the axial vibration signal and the transverse vibration signal are decomposed to obtain a plurality of intrinsic mode functions (IMF).

[0078] Based on the adaptive noise complete ensemble empirical mode decomposition algorithm (CEEMDAN), the initially obtained axial vibration signal and transverse vibration signal are decomposed to obtain a plurality of first IMF components and a first residual; the standard deviation of the first residual is calculated, and if the first residual standard deviation is greater than a preset standard deviation threshold, iterative decomposition is continued to obtain a plurality of second IMF components and a second residual; the iteration is continued until the standard deviation of the Nth residual calculated is less than the preset standard deviation threshold, and the iteration is completed; the plurality of Nth IMF components obtained by the Nth iteration are taken as the final intrinsic mode function (IMF).

[0079] The vibration signal of a wind turbine presents significant nonlinear and non-stationary characteristics due to the complex and variable operating conditions. Although the traditional empirical mode decomposition (EMD) algorithm can decompose the signal into a plurality of intrinsic mode functions (IMFs), the modal aliasing phenomenon often occurs in the vibration signal processing of the wind turbine, resulting in the simultaneous inclusion of high-frequency noise and low-frequency trends in a single modal function (IMF), which greatly reduces the accuracy of signal analysis. To solve this problem, the EMD algorithm is improved, and the adaptive noise complete ensemble empirical mode decomposition (CEEMDAN) algorithm is proposed. CEEMDAN effectively utilizes the characteristics of noise, and performs adaptive adjustment at each step, which not only suppresses modal aliasing but also reduces noise residue, thereby improving the accuracy and reliability of signal decomposition, and is particularly suitable for the analysis of nonlinear and non-stationary signals such as vibration acceleration.

[0080] And ICEEMDAN has two optimizations on the basis of the original. First, although CEEMDAN uses an adaptive noise strategy in the noise addition process, the noise intensity adjustment is relatively rough. In the initial stage, the noise intensity coefficient ε1 is usually set to a fixed value, which cannot be flexibly adapted according to the actual characteristics of the signal. ICEEMDAN introduces a more refined noise intensity dynamic adjustment mechanism. In each iteration, the local characteristics of the current signal and the characteristics of the decomposed IMF components are comprehensively considered. After the signal is decomposed into k-1 IMF components and the residual r k-1 (t), the standard deviation of the residual is calculated. If is large, it indicates that there are large fluctuations in the residual, which may contain signal components that have not been fully decomposed. At the same time, the energy distribution of the decomposed IMF components is calculated. Through a complex function f, the two characteristics are combined to determine the new noise intensity coefficient Secondly, ICEEMDAN adopts a multi-criteria joint stopping condition. In addition to considering the monotonicity of the residual and the error threshold, it also introduces the frequency domain feature analysis of the IMF components. When calculating each IMF component, its frequency bandwidth and frequency center are calculated simultaneously. The frequency bandwidth reflects the frequency range contained in the IMF component, and the frequency center represents the main frequency position of the IMF component. When the frequency bandwidth and the frequency center change of the adjacent IMF components are both very small, it means that the IMF components obtained by subsequent decomposition have little characteristic change, and the signal has been basically decomposed sufficiently. The specific decomposition process is shown in Figure 2 .

[0081] S130: Select the component rich in fault information from the multiple IMF components by the energy ratio method.

[0082] After modal decomposition, it is usually necessary to analyze each IMF component to select representative low-frequency signal characteristics. The energy ratio method provides a normalized and objective evaluation index by quantifying the proportion of each modal component in the total energy. In particular, in the analysis of low-frequency vibration of wind turbine units, this method can effectively highlight the key dynamic information carried by low-frequency signals while effectively suppressing the interference of high-frequency noise. This screening method helps to accurately extract fault characteristics and provides a reliable basis for subsequent diagnosis.

[0083] The basic principle is to use a modal decomposition method (such as ICEEMDAN) to split the mixed signal into multiple intrinsic mode functions (IMF) and a residual component. Each IMF can reflect the sample characteristics of the signal in different frequency intervals. By calculating the energy of each IMF, the energy distribution of the signal in the time-frequency domain can be reconstructed.

[0084] Let the original signal x(t) be processed by the decomposition method to obtain N IMF components and a residual component r(t), which can be expressed as:

[0085]

[0086] On this basis, the total energy E total of the signal is defined as the sum of the energy of each IMF and the energy of the residual, that is,

[0087]

[0088] Among them, the energy E i of each IMF component is calculated by the following integral:

[0089]

[0090] And the energy E r of the residual component is

[0091]

[0092] Energy ratio ER of the i-th IMF component i defined as the ratio of its energy to the total energy, i.e.

[0093]

[0094] where x(t) is the original signal, representing the time series data to be analyzed, and N represents the total number of IMF components.

[0095] S140: Perform Hilbert transform on the screened IMF components to obtain Hilbert spectrum, and then integrate the Hilbert spectrum to obtain Hilbert marginal spectrum corresponding to the axial vibration signal and the lateral vibration signal of the nacelle of the wind turbine generator.

[0096] As an adaptive time-frequency analysis method, Hilbert marginal spectrum can perform local decomposition on signals and obtain instantaneous frequency information, thereby realizing fine analysis of complex signals. Hilbert marginal spectrum is the result of integrating Hilbert spectrum (HHT) in time, which represents the cumulative component of a certain frequency over the entire time. Since it is based on local instantaneous frequency and instantaneous amplitude, it may appear more blurred in frequency distribution due to local changes and instantaneous noise, which is not as sharp as Fourier transform. However, when dealing with nonlinear and non-stationary signals such as vibration, direct Fourier transform may lose time-varying information, while HHT can provide joint description of time and frequency.

[0097] In the initial Hilbert marginal spectrum, empirical mode decomposition (EMD) is first used, and in this paper, the modes after ICEEMDAN decomposition are taken, and then Hilbert transform is performed on each component. Transform the i-th IMF, i.e. i (t), whose conjugate signal is defined as follows:

[0098]

[0099] where P represents the Cauchy principal value integral. By this means, a complex analytic signal can be established.

[0100] where the instantaneous amplitude a i (t) is defined as:

[0101]

[0102] The instantaneous phase θ i (t) is defined as:

[0103]

[0104] Further, by taking the instantaneous phase derivative, the instantaneous frequency can be obtained:

[0105]

[0106] This local time-frequency analysis method based on Hilbert transform can overcome the limitations of traditional global time-frequency analysis method and is suitable for describing the energy change of the signal in the instantaneous scale. The instantaneous amplitude and instantaneous frequency information obtained after completing the Hilbert transform can construct the Hilbert spectrum of the signal, and the Hilbert spectrum H(w, t) is defined as:

[0107]

[0108] Where δ is the Dirac impulse function, which ensures that there is a non-zero contribution at the instantaneous frequency w i (t) place. This spectrum function can describe the instantaneous energy distribution of each frequency component of the signal in the time-frequency plane.

[0109] Compared with the traditional Fourier spectrum, HHT can not only reveal the real existence of the signal at each frequency component, but also intuitively reflect the energy proportion occupied by different frequency components. Based on the calculation of H(w, t) in the foregoing, the marginal spectrum h(w) is obtained by integrating:

[0110]

[0111] The marginal spectrum converts the original three-dimensional time-frequency energy distribution into two-dimensional frequency energy distribution, so as to more intuitively reflect the energy distribution of the signal at different frequencies.

[0112] S150: Based on the fault corresponding feature, it is judged whether there is a peak value corresponding to the fault feature frequency in the Hilbert marginal spectrum, and whether the yaw system of the wind turbine exists a fault.

[0113] Obtaining the fault feature of the yaw system of the wind turbine, and extracting the fault feature frequency information from the fault feature;

[0114] Judging whether there is a peak value corresponding to the fault feature frequency in the Hilbert marginal spectrum of the axial vibration signal and the transverse vibration signal of the wind turbine rotor;

[0115] If it exists, it is determined that there is a fault.

[0116] In the embodiments of the present application, the fault feature is extracted from the historical data in the SCADA database, or obtained by simulating the simulation model of the yaw system of the wind turbine. The fault judgment process of the present application is as shown in Figure 3 .

[0117] The following is a specific embodiment provided by the present application:

[0118] In order to find the reason of yaw system frequent failure of wind turbine, the 2MW wind turbine with high yaw failure rate is selected for data collection. The yaw system failure of the unit occurs many times, in order to find the reason of the phenomenon, according to the Bladed simulation scheme and the actual failure of the unit, vibration acceleration sensor and yaw current transformer are arranged, and MODBUS interface is connected to wind turbine point table information, SCADA data of the unit in the ring network of the wind farm is collected, which provides reliable data support for the research in the following text.

[0119] In view of the yaw system failure problem, in order to realize the data collection of the key components of the unit, a multi-physical quantity joint measurement scheme is proposed in this paper.

[0120] The main measurement points are as follows: cabin X, Y direction vibration acceleration signal; yaw motor current signal; low speed shaft horizontal and vertical direction vibration acceleration signal; gear box vibration acceleration signal; high speed shaft vibration acceleration signal, as shown in Figure 4

[0121] The field arrangement process is as follows: in the arrangement of vibration sensor measurement point, in the X direction of the cabin, the vibration sensor is installed on the side of the cabin wall close to the yaw motor, in the Y direction, the vibration sensor is installed close to the top of the cabin and the electrical control cabinet, which can effectively monitor the vibration of the cabin in Y direction. In addition, corresponding installation is carried out on the low speed shaft, high speed shaft and gear box. In the installation of current sensor, the wiring direction of yaw motor power supply line in electrical cabinet is determined, and the position with relatively independent line and convenient sensor installation is selected for installation.

[0122] After completing the installation of vibration sensor and current transformer, the collection cabinet is fixed and the whole measurement system is comprehensively debugged. It is verified that each sensor can collect accurate and reliable data under different working conditions, and can transmit the data to the host computer in time and accurately.

[0123] The unbalance response of wind turbine rotor of wind turbine causes the yaw deceleration motor flange cracking and yaw gear ring cracking, etc. In order to diagnose the root cause of the failure, GH Bladed software is used to build 2WM wind turbine in this embodiment, and signal analysis is carried out on various wind turbine unbalance faults to find out the fault characteristics. First, the mechanism analysis of wind turbine stress is carried out, and the theoretical response under fault state is deduced, second, Bldaded is used to simulate the wind turbine unbalance fault under each wind condition, and the fault characteristics are obtained after analyzing the vibration signal.

[0124] ​This embodiment analyzes the two failure mechanisms of aerodynamic imbalance and mass imbalance, and describes the load and stress variation of wind turbine under aerodynamic imbalance and mass imbalance. For the convenience of subsequent stress analysis in each direction, a three-dimensional rectangular coordinate system is established for the nacelle as shown in Figure 5 . Figure 5 The X direction is the axial direction of the unit, and the Y direction is the lateral direction of the unit.

[0125] When the wind turbine runs under ideal conditions, the three blades should receive uniform aerodynamic force. However, due to the influence of the actual environment, there is a difference in air flow rate at different heights. This phenomenon makes the lift generated by the blade close to the ground smaller than that of other blades due to the lower wind speed, and the difference in pitch angle or blade shape of each blade also leads to the emergence of aerodynamic imbalance phenomenon. The axial and radial directions of the wind turbine under this operating state produce periodic oscillation. Taking the pitch angle error as an example, the blade stress under different attack angles is analyzed from the wind wheel lateral direction as shown in Figure 6 .

[0126] Under standard operating conditions, the three-blade unit exhibits balanced aerodynamic characteristics, with consistent aerodynamic load distribution. At this time, the torque transmitted from the wind wheel to the drive system remains stable, and the transverse and axial bending moment components are both maintained at zero. When a specific blade has a pitch angle deviation, the aerodynamic load characteristics will change significantly, specifically in the form of abnormal changes in axial thrust and circumferential force. This abnormality not only causes a decrease in output torque, resulting in power loss, but more critically, it forms periodic alternating bending moment loads in the transverse and axial planes. The excitation frequency of this alternating load is synchronized with the wind wheel rotation frequency (1P), which induces a characteristic 1P vibration response in the drive system. When aerodynamic imbalance leads to pitch angle errors in the blades, the differences in blade surface flow field distribution cause significant changes in the force, forming a periodic bending moment in the rotation plane, and ultimately causing a resonance response with the same rotation frequency between the nacelle and the tower. Since the effect of aerodynamic imbalance on axial thrust is much greater than that on lateral force, its impact is more significant, and the axial vibration amplitude of the nacelle and tower caused by it is much higher than the lateral vibration.

[0127] During the normal operation of a wind turbine, the three blades are usually in a state of uniform stress. When the blades have additional attachments or their own weight changes, leading to an imbalance in mass distribution, the wind wheel system will have a mass imbalance problem. This imbalance is usually manifested as a shift in the center of rotation of the wind wheel and the center of gravity, causing abnormal vibration between the wind wheel and the main shaft. This vibration is transmitted to the subsequent drive system through the main shaft, adversely affecting key components such as the generator, and ultimately affecting the overall performance of the unit. This vibration phenomenon caused by differences in blade mass density is referred to as wind wheel mass imbalance failure in wind turbine operation.

[0128] Research shows that the presence of additional mass block of wind turbine will produce centrifugal force, thereby stimulating the unit in the horizontal and vertical direction of the two-way vibration. Research has confirmed that the vibration frequency and the wind wheel rotation frequency have a significant correlation. It is worth noting that the wind wheel system has a high structural stiffness and rotational inertia in the vertical direction. Based on the principles of mechanics and engineering experience, the vertical vibration will encounter greater resistance in the process of transmission, resulting in a magnitude of amplitude decay characteristics. From the point of view of engineering practice, a large number of experimental results show that the amplitude after the magnitude attenuation has little effect on the overall operation of the unit, so it can be ignored in practical application.

[0129] Under the condition of mass imbalance, the periodic change of additional load will cause the system to produce speed fluctuation phenomenon: when the additional mass moves to the windward side of the rotation track, the centrifugal force component and the rotation direction form a positive superposition, causing instantaneous speed increase; Conversely, it produces a deceleration effect in the leeward side, so under the condition of mass imbalance, the lateral vibration 1P frequency of the wind wheel is greater than the vibration amplitude of the same frequency in the axial direction. Research shows that the vibration of the system caused by mass imbalance is mainly derived from the complex effect of centrifugal force vector and gravity vector. This composite excitation not only significantly affects the output power stability of the unit, but also aggravates the fatigue damage of the key components of the yaw system, which poses a potential threat to the operation reliability of the wind turbine.

[0130] In order to be as close as possible to the actual running state of the wind turbine under the condition of imbalance of the wind wheel, the present application constructs an imbalance simulation model of a doubly-fed wind turbine consistent with the actual wind turbine model based on GHBladed simulation software. The model mainly consists of three sub-modules of aerodynamics, transmission chain and vibration: the aerodynamic model is set as the basic module of the simulation, which accurately sets the shape parameters, geometric structure and aerodynamic characteristic parameters of the wind turbine; the transmission chain model is constructed according to the characteristics of the doubly-fed wind turbine, which details the mechanical, electrical parameters and various losses; the vibration model reflects the vibration characteristics of the unit under the condition of imbalance fault by determining the modal order of the blade and the tower.

[0131] The constructed 2MW wind turbine wind wheel imbalance model, the parameters are determined according to the actual data obtained in the simulation stage of the field wind turbine. After setting the parameters of each module, the simulation length and the operation state of the unit are set. Three different wind conditions and two fault modes, i.e. aerodynamic imbalance and mass imbalance, are set in the test process. Data collection starts after 30 seconds of unit operation and ends at 630 seconds, with a total sampling period of 10 minutes and a sampling frequency of 20 Hz to ensure that continuous and sufficient running information is captured.

[0132] Firstly, the data is collected when the unit is in a healthy operating state, which can be used as a benchmark for the balance state of the wind wheel; secondly, in order to simulate the aerodynamic imbalance fault, the pitch angle of blade 1 is adjusted to deviate, and the operating data in this state is recorded on the premise of ensuring that the other blades remain in the standard installation state; finally, after restoring blade 1 to the standard state, a multi-parameter mass block is installed at a distance of 30 m from the blade root to simulate the mass imbalance fault, and the corresponding operating data is collected.

[0133] The specific parameters of the wind wheel, the nacelle and the tower are shown in Table 1.

[0134] Table 1ParameterTable.ofWindTurbineModel

[0135]

[0136]

[0137] A 2MW doubly-fed wind turbine is simulated by using GH Bladed software according to the actual wind turbine model. According to IEC61400-1 document, the wind condition of the generator is the normal turbulent model. According to the actual operating state of the wind turbine, the working condition is divided into three kinds: below rated wind speed, rated wind speed and above rated wind speed. The nacelle axial and lateral vibration data near the wind wheel are analyzed respectively.

[0138] The model parameters and controllers of the unit are derived from the simulation model and parameters established at the beginning of the actual wind turbine. The simulation analysis is carried out for 10 minutes under the condition of 6m / s wind speed and 6% turbulence intensity. The simulation study covers two states: balanced state and aerodynamic imbalance state. In the aerodynamic imbalance state, the pitch angle deviation of 2°, 4° and 8° is set respectively to evaluate the influence of different pitch angle errors on the performance of the wind turbine.

[0139] Under the condition of average wind speed of 6m / s, the rotating speed is 9.01RPM. According to the relationship between rotating speed and rotating frequency, the one rotating frequency (1P) at this time is about 0.15Hz, and the natural frequency of the tower is about 0.33Hz.

[0140] In the axial direction, the vibration amplitude in the balanced state is basically maintained at 0.036m / s 2 With the increase of pitch angle deviation to 2°, 4° and 8°, the vibration value increases obviously, and the maximum vibration values are 0.075m / s 2 , 0.101m / s 2 and 0.14m / s 2 respectively. From the lateral direction, it can be seen that the amplitude in the balanced state is basically lower than 0.036m / s 2, with the increase of the pitch angle deviation, the overall vibration amplitude increases, and the amplitude is maximum 0.04 m / s at the pitch angle deviation of 2° 2 , the amplitude is maximum 0.044 m / s at the pitch angle deviation of 4° 2 , and the amplitude is maximum 0.061 m / s at the pitch angle deviation of 8° 2 When the unit is aerodynamically unbalanced due to the pitch angle error, the axial and lateral vibration accelerations both increase, and the increase of the axial vibration amplitude is much higher than the change of the lateral vibration amplitude.

[0141] This shows that the pitch angle error not only causes the amplification of the overall vibration of the unit, but also leads to the sharp rise of the axial vibration, which may have a more serious impact on the safety of the unit structure.

[0142] Through the comparison of the lateral and axial vibrations under the balanced state and each angle deviation, it can be seen that there is no obvious difference in the amplitudes of the two directions under the balanced state, and with the increase of the pitch angle deviation, the axial amplitude is larger than the lateral amplitude, and the difference increases with the unbalance degree.

[0143] To analyze the frequency domain characteristics, frequency domain analysis is performed, and the amplitude is not normalized after Fourier transform, which does not affect the amplitude comparison. Through the frequency domain analysis of the data, there are three prominent frequency points in the axial frequency spectrum, which are one times rotational frequency (1P), two times rotational frequency (2P) and three times rotational frequency (3P). At 1P, the amplitudes of the balanced state and the pitch angle deviations of 2°, 4° and 8° are 15.230, 73.54, 237.22 and 525.77 respectively; at 2P, they are 16.5, 74.1, 175.07 and 352.3 respectively; at 3P, the amplitudes of the four are all around 101.5, with no obvious change. According to the comparison of the spectrum and data, with the increase of the deviation angle, the increase of the amplitude at 1P is the most obvious in the axial direction, the increase of the amplitude at 2P is affected by 1P, and the change of the amplitude at 3P is not obvious.

[0144] From the lateral spectrum, 1P, 2P, natural frequency and 3P are more prominent. At 1P, the amplitudes of the balanced state and the pitch angle deviations of 2°, 4° and 8° are 11.26, 17.32, 37.59 and 80 respectively; at 2P, they are 6.16, 20.32, 42.95 and 60 respectively; at the natural frequency, they are 43.02, 48.57, 57.21 and 85 respectively; at 3P, the amplitudes of the four are all around 48.35, with no obvious difference. Through the analysis, it is shown that the aerodynamic unbalance causes the increase of the 1P amplitude in the axial and lateral directions, and the response at 1P in the axial direction is higher.

[0145] To further observe the amplitude change at each characteristic frequency, the frequency domain data under balanced state and each deviation angle are compared and analyzed. In the balanced state, the dominant frequency of the wind wheel side is 3P, and the amplitude of both sides decreases in turn. At 1P and 2P, there is a small amplitude, which is not obvious compared with the dominant frequency. With the increase of aerodynamic imbalance, the axial vibration at 1P is more and more prominent, and its amplitude is much larger than that of the transverse direction; at 2P frequency, the axial is slightly larger than the transverse; at 3P frequency, the difference remains basically unchanged.

[0146] In order to conform to the actual operation condition of the wind turbine as much as possible, under the condition of wind speed of 10 m / s and turbulence intensity of 6%, 10 minutes of simulation analysis is carried out, which is divided into balanced state and aerodynamic imbalance state. In the aerodynamic imbalance state, the pitch angle deviation of 2°, 4° and 8° is set respectively. First, near the rated wind speed, its operating power is close to full load, and the rotating speed is maintained near 13.7 RPM.

[0147] The analysis result shows that in the axial balanced state, the vibration amplitude is basically maintained at 0.12 m / s 2 ; when the pitch angle deviation is expanded to 2°, 4° and 8° respectively, the vibration amplitude rises significantly, and the maximum value reaches 0.227 m / s 2 , 328 m / s 2 and 0.513 m / s 2 respectively. At the same time, after the aerodynamic imbalance caused by the pitch angle error, the axial and transverse vibration acceleration of the wind turbine is improved, but the amplitude growth of the axial vibration far exceeds that of the transverse direction. Specifically, in the transverse balanced state, the amplitude is usually lower than 0.036 m / s 2 , and with the increase of the pitch angle deviation, the maximum amplitude rises to 0.04 m / s 2 , 0.044 m / s 2 and 0.061 m / s 2 respectively at 2°, 4° and 8° deviation. The result shows that the aerodynamic imbalance not only aggravates the overall vibration of the wind turbine, but also more easily leads to the sharp rise of the axial vibration.

[0148] Under the influence of turbulence, the wind speed fluctuates around 10 m / s, and the rotating speed fluctuates around 13.75 RPM. According to the relationship between rotating speed and rotating frequency, the one rotating frequency (1P) at this time is about 0.228 Hz, and the natural frequency of the tower is about 0.33 Hz. The aerodynamic imbalance of the unit near the rated wind speed is analyzed in the frequency domain.

[0149] From the axial frequency domain spectrum, there are three relatively prominent frequency points, which are 1P, 2P and tower natural frequency. At 1P, the amplitudes of aerodynamic unbalance 8°, 4°, 2° and balanced state are 1837.38, 1253.67, 237.22 and 183.95 respectively; at natural frequency, they are 408.65, 282.86, 175.58 and 172.35 respectively; at 2P, the amplitude of aerodynamic unbalance 8° is larger than the other three, which is 544.32, and the amplitude increment of the rest is smaller. According to the comparison of the spectrum, with the increase of the deviation angle, the amplitude increment at one times rotational frequency in the axial direction is the most obvious, the increment at two times rotational frequency is increased, which may be affected by one times rotational frequency, the amplitude of tower natural frequency is also affected and increased under the condition of extreme aerodynamic unbalance, and the change of three times rotational frequency is not obvious.

[0150] The amplitudes at 1P, 2P and natural frequency are relatively prominent, and the amplitudes of balanced state and pitch angle deviation 2°, 4°, 8° at 1P are 80.94, 149.51, 396.66 and 482.08 respectively; at 2P, they are 7.55, 25.07, 106.9 and 222.19 respectively; at natural frequency, they are 224.39, 242.98, 351.61 and 643.96 respectively. The amplitudes of the four at 3P are all around 72.65, and there is no obvious difference. The analysis shows that there is a significant difference in the vibration changes caused by aerodynamic unbalance in the axial and lateral directions.

[0151] In order to further observe the amplitude change at each characteristic frequency, the frequency domain data of balanced state and each aerodynamic unbalance angle are compared and analyzed.

[0152] From the comparison of the lateral and axial frequency domain of balanced state and each aerodynamic unbalance state near the rated wind speed, it can be seen that in the balanced state, the dominant frequency of the wind wheel side is the natural frequency of the tower, and the axial and lateral vibrations at natural frequency, 1P, 2P and 3P have little difference, and the amplitudes on both sides decrease in turn. Compared with the dominant frequency, the amplitudes at 1P and 3P are not obvious. With the increase of aerodynamic unbalance angle, the axial vibration at 1P has the most obvious amplitude increment, which is much larger than the lateral vibration. The amplitude at natural frequency increases a little, but there is no obvious difference between the two directions. At 2P and 3P frequencies, the increment is smaller compared with the increment at one times rotational frequency, and the amplitudes in the two directions are basically the same.

[0153] In order to explore the actual operation condition of wind turbine above the rated wind speed, a simulation analysis of 10 minutes was carried out under the condition of wind speed of 14 m / s and turbulence intensity of 6%. The rest of the parameters are the same as above.

[0154] When the wind speed is above the rated wind speed, the unit runs at full power and the rotating speed is maintained at 13.7 RPM, the rotating frequency (1P) is 0.23 Hz, the blade passing frequency (3P) is 0.69 Hz, and the natural frequency is 0.33 Hz. The characteristic frequency point is near the above frequencies, which is affected by the rotating speed fluctuation.

[0155] The axial and lateral vibrations of the wind turbine under high wind speed are compared and analyzed in time domain. From the time domain graph, the basic amplitude trend in the axial and lateral directions is consistent with the performance near the rated wind speed. However, the aerodynamic imbalance degree increases significantly with the increase of the wind speed, and the amplitudes in both directions increase. From the axial direction, the maximum amplitudes under the balanced state, aerodynamic imbalance of 2°, 4°, and 8° are 0.147 m / s 2 , 0.256 m / s 2 , 0.416 m / s 2 , and 0.818 m / s 2 , respectively. In the lateral direction, the maximum amplitudes under the balanced state, aerodynamic imbalance of 2°, 4°, and 8° are 0.1346 m / s 2 , 0.2377 m / s 2 , 0.264 m / s 2 , and 0.491 m / s 2 , respectively. The lateral vibration is larger after the fan starts, and the vibration value decreases after a period of time. Under the same aerodynamic imbalance angle, the axial amplitude is higher than the lateral amplitude.

[0156] It is obvious that the axial amplitude exceeds the lateral amplitude with the increase of the pitch angle deviation, which is the same as the analysis result above. To study the characteristic response in the frequency domain under high wind speed, the axial and lateral vibrations under this working condition are analyzed in the frequency domain.

[0157] Under high wind speed, the wind turbine needs to cope with larger load, and the vibration value increases. Under the 1P frequency in the axial direction, the amplitudes under aerodynamic imbalance of 8°, 4°, and 2° are above 3000, 1500, and 1000, respectively, and the amplitude at 2P frequency is 900. In the lateral direction, the amplitudes under aerodynamic imbalance of 8°, 4°, and 2° at 1P and natural frequency are below 700, and the amplitudes at 3P are almost the same. It can be seen that the aerodynamic imbalance has a serious impact on the vibration at 1P in the axial direction, and the amplitude at 1P in the axial direction reflects the important characteristics of the aerodynamic imbalance.

[0158] Under the balanced state of high wind speed, the axial and lateral vibration amplitudes are generally consistent, except that the axial amplitude at 1P is slightly higher than the lateral amplitude, and there is no obvious difference. Under the imbalance angles, the amplitude at 1P in the axial direction is much higher than that in the lateral direction, and there is no obvious difference at the natural frequency, 2P, and 3P.

[0159] To explore the fault characteristics of wind turbine rotor mass imbalance, the 2MW wind turbine model established above was simulated and analyzed. The working state under three wind conditions below the rated wind speed, near the rated wind speed and above the rated wind speed was distinguished, and the balance state of the blade, the mass imbalance state of 1 / 10, 1 / 15 and 1 / 20 of the total mass of a single blade under each working condition was analyzed to explore the fault characteristics under the condition of mass imbalance.

[0160] The average wind speed was set to 6m / s and the turbulence intensity was set to 6% in Bladed. Under this working condition, the speed of the unit and the wind speed fluctuated greatly under the influence of mass imbalance, and the average value of the speed was calculated to be 9.26RPM, and the rotor rotation frequency(1P) was 0.154Hz.

[0161] With the increase of mass imbalance, the transverse vibration value increased. From the axial direction, the maximum amplitude of the balance state, mass imbalance 1 / 10, 1 / 15 and 1 / 20 was basically within the range of plus or minus 0.085m / s 2 . In the transverse direction, the maximum amplitude of the balance state, mass imbalance 1 / 10, 1 / 15 and 1 / 20 was 0.076m / s 2 , 0.0837m / s 2 , 0.0702m / s 2 , 0.656m / s 2 respectively. The simulation data under the condition of 6m / s wind showed that under the condition of low wind speed, the axial vibration caused by mass imbalance was small, basically the same as the balance state, while the transverse vibration caused a large amplitude increment.

[0162] The axial amplitude was slightly higher than the transverse amplitude in the balance state, and with the increase of mass imbalance, the transverse amplitude was higher than the axial amplitude and increased continuously.

[0163] The specific frequency characteristics need to be analyzed from the frequency domain, and from the frequency spectrum, mass imbalance has little effect on the axial direction, with a small amount of amplitude increase at 1P. In the transverse direction, it causes the amplitude to increase at 1P, 2P and near the natural frequency.

[0164] The increase of mass imbalance caused the transverse vibration amplitude near 1P(0.154Hz) and 2P(0.308Hz) to be much larger than the axial vibration, and the amplitude difference at 3P remained basically the same.

[0165] The unit was simulated under the rated wind speed, the average wind speed was set to 10m / s and the turbulence intensity was set to 6%. The results showed that the speed fluctuated greatly under the influence of mass imbalance, and the average value of the speed was calculated to be 13.7RPM, and the rotor rotation frequency(1P) was 0.2283Hz. The axial and transverse vibration accelerations after simulation were compared and analyzed in time domain.

[0166] At the rated wind speed, the amplitude in both axial and lateral directions increases compared with the low wind speed condition. In the axial direction, the maximum amplitude of the balanced state, mass imbalance 1 / 20, 1 / 15, 1 / 10 is 0.165 m / s 2 , 0.236 m / s 2 , 0.337 m / s 2 , 0.47 m / s 2 , respectively. It can be seen that the axial vibration is intensified under the influence of mass imbalance. In the lateral direction, the maximum amplitude of the balanced state, mass imbalance 1 / 20, 1 / 15, 1 / 10 is 0.132 m / s 2 , 0.291 m / s 2 , 0.509 m / s 2 , 0.827 m / s 2 , respectively. The amplitude is greater than the axial direction, which has the same trend as the axial direction.

[0167] Through the comparative analysis of the lateral and axial directions under each mass imbalance state in the time domain, the lateral vibration is slightly greater than the axial vibration, and the amplitude difference between the two is not large overall. It can be seen that when working near the rated wind speed, mass imbalance will increase the axial and lateral vibration amplitudes of the fan, but there is no large difference between the two.

[0168] Mass imbalance mainly causes the vibration of the axial and lateral directions to increase at 1P frequency. The amplitude at 1P is 1833.65 for the axial direction and 3618.36 for the lateral direction. As the mass imbalance degree deepens, both have a large amplitude at 1P. Mass imbalance causes the lateral vibration to be slightly greater than the axial vibration. To further verify, the frequency spectrum of the axial and lateral directions under each mass imbalance parameter is compared.

[0169] There is no large difference between the two in the balanced state. As the mass imbalance degree increases, the lateral vibration at 1P is higher than the axial vibration, which is consistent with the analysis result above.

[0170] Bladed is used to simulate the operating data of the wind turbine above the rated wind speed. The average wind speed is set to 14 m / s, and the turbulence intensity is 6%, which meets the actual operating conditions of the fan. In this environment, it can be seen that the speed of the unit is affected by mass imbalance and fluctuates. The average speed is 13.75 RPM, and the 1P is 0.229 Hz.

[0171] Comparative analysis of the axial and lateral directions under each mass imbalance state shows that the amplitude of both directions increases under the influence of mass imbalance. The maximum amplitude of the axial balanced state and mass imbalance is 0.1289 m / s 2 , 0.367 m / s 2 , and the lateral direction is 0.1582 m / s 2 , 0.645 m / s2 The transverse amplitude is slightly larger than the axial.

[0172] The mass imbalance parameter results at 14m / s wind speed are similar to the operating results near the rated wind speed, i.e., the axial and transverse amplitudes are basically consistent in the balanced state, and the transverse amplitude is slightly higher than the axial amplitude under the mass imbalance condition.

[0173] The axial vibration amplitude at 1P is 2588.28, and the transverse vibration amplitude is 4267, and the transverse amplitude at 1P is higher than the axial amplitude, and the amplitudes at 2P and 3P are not obvious compared with 1P.

[0174] The time domain and frequency domain characteristics of the unit under each wind condition in the balanced state are analyzed, and it can be seen that the transverse vibration at 1P is greater than the axial vibration as the mass imbalance increases.

[0175] In order to more clearly compare the fault characteristics in the foregoing, the time domain and frequency domain characteristics of the unit under each wind condition in the balanced state are compared and analyzed. With the increase of wind speed, the vibration of the nacelle increases, and when running at low wind speed, 3P is the main vibration frequency, and when running near and above the rated wind speed, 1P, 3P and the natural frequency are the main vibration frequencies.

[0176] The present application proposes a signal analysis method based on ICEEMDAN and Hilbert marginal spectrum fusion, which is used for analysis and fault diagnosis of wind turbine measured data. First, the original signal is adaptively decomposed by ICEEMDAN algorithm, and the multi-scale intrinsic mode function is extracted. After the decomposition is completed, the modes are screened according to the energy ratio, and the noise and invalid modes are effectively removed. Then, the Hilbert transform is performed on the screened modal samples to obtain the instantaneous frequency and amplitude data. On this basis, the Hilbert spectrum and its marginal spectrum are constructed to intuitively show the energy distribution characteristics of the signal in the time-frequency domain. The method fully utilizes the advantages of ICEEMDAN in signal decomposition accuracy and noise immunity, and combines the ability of Hilbert marginal spectrum in revealing local time-frequency characteristics of the signal, and realizes the fault diagnosis of the wind turbine running state by comparing and analyzing the measured signals.

[0177] Through data acquisition on a 2WM wind turbine, the measured vibration data of the nacelle and transmission shaft are obtained. In this section, the measured vibration data are taken as the analysis object, the relevant data of the unit near the average wind speed of 10m / s are extracted through SCADA wind speed information division, the fault characteristics are extracted through modal decomposition and marginal spectrum analysis, and finally the fault diagnosis is realized through the fault diagnosis model process.

[0178] According to the SCADA data, a 2.5-minute-long operation data near the rated wind speed of the unit is intercepted, and the corresponding data of the cabin axial and lateral vibration data is extracted. First, the ICEEMDAN decomposition is performed on the signal, in which the ICEEMDAN decomposition of the axial vibration signal is as shown in Figure 7 , and the ICEEMDAN decomposition of the lateral vibration signal is as shown in Figure 8 .

[0179] From the decomposition effect of the vibration in two directions, the ICCEMDAN algorithm has decomposed the signals in each frequency band into each IMF component, and from each component, there is no obvious modal aliasing phenomenon. After obtaining the decomposed signals, in order to obtain the IMF modal component with more fault information, the energy ratio is used to select the appropriate IMF modal. The energy ratio values of each IMF component in two directions are shown in Table 2.

[0180] Table 2 Numerical Values of the Energy Ratio of Each Order IMF

[0181]

[0182] Through energy ratio calculation, the energy ratio of each modal is arranged from large to small, and when the cumulative threshold of the energy ratio reaches 0.9 or more from large to small, the first five IMF components that meet the requirements are selected in the axial direction, and the first four IMF components are selected in the lateral direction, and then data reconstruction is performed. Due to the nonlinearity and non-stationarity of the vibration data, the traditional Fourier transform cannot well analyze the energy distribution characteristics in the frequency domain, so the Hilbert transform is performed on the IMF components selected in the foregoing to obtain the Hilbert spectrum, which reflects the distribution of the signal on the frequency axis, and the marginal spectrum is obtained by integrating the Hilbert spectrum.

[0183] From the marginal spectrum of the measured data and the balanced state fan data, it can be seen that in the axial direction, the amplitude of the measured data at 1P increases by 6.81 times the amplitude at 1P frequency in the balanced state, which has a significant gap. In the lateral direction, the measured data has a significant amplitude increase at the natural frequency compared with the balanced state. The axial amplitude increase is significantly higher than the lateral amplitude increase. From the simulation analysis in the foregoing, when there is aerodynamic imbalance near the rated wind speed, the axial amplitude increase is greater than the lateral amplitude increase, and the axial vibration has a large increase at 1P and a small increase near the natural frequency compared with the balanced state. According to the change characteristics of the vibration data in the time domain and the frequency domain, the diagnosis model is diagnosed, and it is found that the measured data of the unit has a certain degree of aerodynamic imbalance vibration response, which is the reason for the frequent failure of the yaw system.

[0184] In order to verify the diagnosis result, the staff cooperated with the test fan blade pitch angle to find that one blade and the other two blades pitch angle deviated 2.5°, which caused the existence of large aerodynamic imbalance of the unit, caused the yaw system load too large, and frequent failure.

[0185] As Figure 9 shown, the present application provides a wind turbine yaw system fault judgment device 900, comprising:

[0186] The signal acquisition module 910 is used for acquiring the axial vibration signal and the transverse vibration signal of the wind wheel of the wind turbine in the running state;

[0187] The signal decomposition module 920 is used for decomposing the axial vibration signal and the transverse vibration signal based on the improved adaptive noise complete set empirical mode decomposition algorithm (ICEEMDAN), to obtain a plurality of intrinsic mode functions (IMF);

[0188] The screening module 930 is used for selecting a component rich in fault information from the plurality of IMF components by the energy ratio method;

[0189] The transformation module 940 is used for performing Hilbert transformation on the screened IMF component to obtain a Hilbert spectrum, and then integrating the Hilbert spectrum to obtain a Hilbert marginal spectrum corresponding to the axial vibration signal and the transverse vibration signal of the wind turbine nacelle;

[0190] The fault judgment module 950 is used for judging whether there is a peak value corresponding to a fault characteristic frequency in the Hilbert marginal spectrum based on the fault corresponding feature, and judging whether the wind turbine yaw system has a fault.

[0191] In order to realize the embodiment, the present application further provides an electronic device, comprising: at least one processor; and a memory connected with the at least one processor in communication; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform each step in the method of the foregoing technical solution.

[0192] As Figure 10 shown, the non-transitory computer readable storage medium includes a memory 810 storing instructions, and an interface 830, and the instructions can be executed by a processor 820 according to the blade flutter suppression processing to complete the method. Optionally, the storage medium can be a non-transitory computer readable storage medium, for example, the non-transitory computer readable storage medium can be a ROM, a random access memory (RAM), a CD-ROM, a magnetic tape, a floppy disk, and an optical data storage device, etc.

[0193] In order to achieve the embodiments, the application further provides a non-transitory computer readable storage medium, which stores a computer program, and the computer program is executed by a processor to realize the operation optimization for blade flutter suppression according to the embodiments of the application.

[0194] In the description of the specification, the description of the terms "one embodiment", "some embodiments", "an example", "a specific example", or "some examples" and the like means that the specific features, structures, materials or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the application. In the specification, the illustrative description of the terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any appropriate manner in any one or more embodiments or examples. In addition, the person skilled in the art can combine and combine the different embodiments or examples described in the specification and the features of the different embodiments or examples without contradiction.

[0195] In addition, the terms "first", "second" are only for descriptive purposes and cannot be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined with "first", "second" can explicitly or implicitly include at least one of the features. In the description of the application, the meaning of "multiple" is at least two, for example, two, three, etc., unless otherwise specifically limited.

[0196] Any process or method descriptions in flow charts or otherwise described herein can be understood as representing code modules, segments, or portions of code that include one or more executable instructions for implementing the specified logic functions or processes, and the preferred embodiments of the application also include additional implementation examples, in which the functions can be performed in different orders, including substantially simultaneously, or in reverse order, depending on the functionality involved, which should be understood by those skilled in the art of the embodiments of the application.

[0197] The logic and / or steps represented in the flowcharts and / or described herein, for example, can be considered as a sequence of executable instructions stored in a computer readable medium, which can be executed by an instruction execution system, apparatus or device, such as a computer-based system, processor- based system, or other system that can fetch the instructions from the instruction execution system, apparatus, or device and execute the instructions, or a combination thereof. For the purposes of this specification, a "computer readable medium" can be any apparatus that can contain, store, communicate, propagate, or transport the program for use by or in connection with the instruction execution system, apparatus, or device. The computer readable medium can be a computer readable storage medium or a computer readable signal medium. The computer readable storage medium can include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, a portable computer diskette (magnetic), a RAM (random access memory), a ROM (read only memory), an EPROM (erasable programmable ROM), EEPROM (electrically erasable programmable ROM), a storage

[0198] It should be understood that aspects of the application can be implemented in hardware, software, firmware or a combination thereof. In this embodiment, various steps or methods can be embodied in software or firmware stored in a memory and executed by a suitable instruction execution system. As such, it should be understood that aspects of the application can be implemented in a variety of ways, including as a computer readable medium comprising code or instructions to be executed by an instruction execution system, or as a computer readable medium comprising code or instructions to be executed by an instruction execution system.

[0199] It should be understood that aspects of the application can be implemented in hardware, software, firmware or a combination thereof. In this embodiment, various steps or methods can be embodied in software or firmware stored in a memory and executed by a suitable instruction execution system. As such, it should be understood that aspects of the application can be implemented in a variety of ways, including as a computer readable medium comprising code or instructions to be executed by an instruction execution system, or as a computer readable medium comprising code or instructions to be executed by an instruction execution system.

[0200] In addition, each function unit in each embodiment of the present application can be integrated in one processing module, or each unit can be physically present separately, or two or more units can be integrated in one module. The integrated module can be realized in the form of hardware or in the form of a software function module. When the integrated module is realized in the form of a software function module and sold or used as an independent product, it can also be stored in a computer readable storage medium.

[0201] The mentioned storage medium can be a read-only memory, a magnetic disk or an optical disk, etc. Although the embodiments of the present application have been shown and described above, it should be understood that the embodiments are exemplary and should not be construed as limiting the present application, and those skilled in the art can make changes, modifications, replacements and variations to the embodiments within the scope of the present application.

Claims

1. A method for diagnosing faults in the yaw system of a wind turbine, the process of which includes: The axial and lateral vibration signals of the wind turbine nacelle were collected during operation. Based on the improved adaptive noise complete set empirical mode decomposition (ICEEMDAN) algorithm, the axial vibration signal and the transverse vibration signal are decomposed to obtain multiple intrinsic mode function (IMF) components; The energy ratio method is used to select components rich in fault information from multiple IMF components; The selected IMF components are subjected to Hilbert transform to obtain the Hilbert spectrum, which is then integrated to obtain the Hilbert marginal spectrum corresponding to the axial vibration signal and lateral vibration signal of the wind turbine nacelle. Based on the fault correspondence characteristics, it is determined whether there is a peak value corresponding to the fault characteristic frequency in the Hilbert marginal spectrum, and thus whether there is a fault in the yaw system of the wind turbine.

2. The method for diagnosing faults in the yaw system of a wind turbine according to claim 1, characterized in that, The axial and lateral vibration signals of the wind turbine nacelle were collected during operation, including: In the nacelle of the wind turbine, a first vibration sensor and a second vibration sensor are installed near the yaw motor and near the top of the nacelle and the electrical control cabinet to collect the axial and lateral vibration signals. In the wind turbine, additional vibration sensors are installed in the low-speed shaft, high-speed shaft, and gearbox to collect vibration signals from the low-speed shaft, high-speed shaft, and gearbox.

3. The method for diagnosing faults in the yaw system of a wind turbine according to claim 1, characterized in that, Based on the improved adaptive noise complete ensemble empirical mode decomposition (ICEEMDAN) algorithm, the axial vibration signal and the transverse vibration signal are decomposed to obtain multiple intrinsic mode function (IMF) components, including: Based on the adaptive noise complete set empirical mode decomposition (CEEMDAN) algorithm, the initially acquired axial vibration signal and the transverse vibration signal are decomposed to obtain multiple first IMF components and first residuals; Calculate the standard deviation of the first residual. If the standard deviation of the first residual is greater than a preset standard deviation threshold, continue the iterative decomposition to obtain multiple second IMF components and second residuals. The iteration continues until the standard deviation of the calculated Nth residual is less than the preset standard deviation threshold, at which point the iteration is complete. The multiple NIMF components obtained in the Nth iteration are used as the final intrinsic mode functions (IMFs).

4. The method for diagnosing faults in the yaw system of a wind turbine according to claim 1, characterized in that, Using the energy ratio method, intrinsic mode functions (IMFs) containing fault information are selected from multiple IMFs, including: Suppose that the original signal x(t) is processed by a decomposition method to obtain N IMF components and one residual component r(t), and its expression is: Based on this, the total energy E of the signal total Defined as the sum of the energies of each IMF and the residual energy, i.e. Among them, the energy E of each IMF component i The following integral formula is used to calculate: The energy E of the residual component r Then it is The energy ratio of the i-th IMF component to ER i Defined as the ratio of its energy to the total energy, i.e. In the formula, x(t) is the original signal, representing the time series data to be analyzed, and N represents the total number of IMF components.

5. The method for diagnosing faults in the yaw system of a wind turbine according to claim 4, characterized in that, Based on the Hilbert marginal spectrum, a Hilbert transform is performed on the selected intrinsic mode functions to obtain the Hilbert marginal spectra of the axial vibration signal and the lateral vibration signal corresponding to the wind turbine rotor, including: The conjugate signal of the i-th IMF component is defined as follows: Here, P represents the Cauchy principal value integral. This allows the construction of a complex analytic signal. Where the instantaneous amplitude a i (t) is defined as: instantaneous phase angle θ i (t) is defined as: By further differentiating the instantaneous phase, the instantaneous frequency can be obtained: The instantaneous amplitude and frequency information obtained after performing the Hilbert transform are used to construct the Hilbert spectrum of the signal. The Hilbert spectrum H(w,t) is defined as follows: Where δ is the Dirac impulse function, ensuring that at the instantaneous frequency w i The spectral function has a non-zero contribution at (t); it can describe the instantaneous energy distribution of each frequency component of the signal in the time-frequency plane. Integrating H(w,t), we obtain the marginal spectrum h(w): Marginal spectrum transforms the original three-dimensional time-frequency energy distribution into a two-dimensional frequency energy distribution, thus more intuitively reflecting the energy distribution of the signal at different frequencies.

6. The method for diagnosing faults in the yaw system of a wind turbine according to claim 5, characterized in that, Based on the Hilbert marginal spectra of the axial and lateral vibration signals corresponding to the wind turbine rotor, the determination of whether the wind turbine's yaw system has a fault includes: Obtain the fault characteristics of the wind turbine yaw system and extract the fault characteristic frequency information from it; Determine whether there are peak values ​​corresponding to fault characteristic frequencies in the Hilbert marginal spectra of the axial vibration signal and the lateral vibration signal of the wind turbine rotor; If it exists, then a fault is determined to exist.

7. The method for diagnosing faults in the yaw system of a wind turbine according to claim 6, characterized in that, In the step of obtaining the fault characteristics of the wind turbine yaw system, the fault characteristics are extracted from historical data in the SCADA database or obtained by simulating the wind turbine yaw system by constructing a simulation model.

8. A fault diagnosis device for the yaw system of a wind turbine, characterized in that, include: The signal acquisition module is used to collect the axial and lateral vibration signals of the wind turbine rotor during operation. The signal decomposition module is used to decompose the axial vibration signal and the transverse vibration signal based on the improved adaptive noise complete set empirical mode decomposition algorithm (ICEEMDAN) to obtain multiple intrinsic mode functions (IMFs). The screening module is used to select components rich in fault information from multiple IMF components using the energy ratio method; The transformation module is used to perform Hilbert transformation on the screened IMF components to obtain the Hilbert spectrum, and then integrate it to obtain the Hilbert marginal spectrum corresponding to the axial vibration signal and lateral vibration signal of the wind turbine nacelle. The fault diagnosis module is used to determine whether there is a peak value corresponding to the fault characteristic frequency in the Hilbert marginal spectrum based on the fault correspondence characteristics, and to determine whether there is a fault in the yaw system of the wind turbine.

9. A computer device, characterized in that, The method includes an input / output unit, a memory, and a processor, wherein the memory stores computer-readable instructions that, when executed by the processor, cause the processor to perform the steps of the method as described in any one of claims 1 to 7.

10. A storage medium storing computer-readable instructions, characterized in that, When the computer-readable instructions are executed by one or more processors, they cause the one or more processors to perform the steps in the method as described in any one of claims 1 to 7.

Images