Offshore wind power early fault feature extraction method

Abstract

The invention discloses an offshore wind power early-stage fault feature extraction method. The method comprises the specific steps that vibration signals in an offshore wind power unit are collected through a motion sensor; performing wavelet decomposition on the vibration signal, and decomposing the vibration signal into four sub-signal components according to the frequency from high to low; processing the vibration signal based on the continuous kurtosis to obtain a continuous kurtosis value of the signal; the wavelet instantaneous energy and the continuous kurtosis are fused, a fault feature signal is reconstructed, a local cross correlation coefficient of the wavelet instantaneous energy and the continuous kurtosis is calculated, and fused feature instantaneous energy is obtained; and calculating an envelope spectrum, and extracting a fault feature frequency. According to the method, the kurtosis index and the wavelet transform are combined, the kurtosis index is extremely sensitive to impact components in the signals, the wavelet transform has good time-frequency localization characteristics, the time-varying characteristics of the signals can be depicted in detail, the defect of a single method is effectively overcome through combination of the kurtosis index and the wavelet transform, and the extraction precision of weak fault characteristics is improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of offshore wind power, and in particular relates to a method for extracting early fault features of offshore wind power. Background Art

[0002] Offshore wind turbines are exposed to high humidity and highly corrosive marine environments for a long time. This harsh working environment poses a severe challenge to the safe and stable operation of the transmission system. Compared with the relatively stable factory environment, the key components of the transmission system of offshore wind turbines need to withstand greater environmental pressure. Any minor defects may deteriorate rapidly in such a harsh environment, thereby causing major safety hazards. Therefore, for offshore wind turbines, it is far from enough to rely solely on traditional regular maintenance. Timely and effective fault detection must be achieved in the early stages of the fault so that measures can be taken quickly to prevent the further deterioration of the defects, thereby avoiding the catastrophic consequences of late-stage failures.

[0003] As an important reflection of the operating status of mechanical equipment, vibration signals contain rich information and can be regarded as a generalized high-dimensional information. However, starting from a single-dimensional observation method, only some features of the signal can be captured, which is a direct but limited way to extract fault features. In practical applications, each signal processing method is equivalent to a specific observation direction in the generalized high-dimensional information space. It can only observe and extract the fault features in the signal from a certain dimensional direction, but it cannot guarantee that this direction is the best projection direction of the fault features in the generalized high-dimensional space. Therefore, when faced with more complex signals or weaker fault features, these methods may show poor results or even fail completely, and cannot meet the needs of high-precision fault detection. Summary of the invention

[0004] In order to make up for the deficiencies of the prior art, the present invention aims to provide a method for extracting early fault features of offshore wind power, so as to effectively extract early weak fault features. To achieve this purpose, the specific technical solution of this application is as follows:

[0005] The method for extracting early fault features of offshore wind power generation specifically comprises the following steps:

[0006] S1. Collect vibration signals from offshore wind turbines through motion sensors;

[0007] S2. Perform wavelet decomposition on the vibration signal and decompose it into four sub-signal components from high to low frequency;

[0008] S3. Based on the continuous kurtosis, the vibration signal is processed to obtain the continuous kurtosis value of the signal;

[0009] S4. Fuse the wavelet instantaneous energy and continuous kurtosis, reconstruct the fault characteristic signal, calculate the local mutual correlation coefficient of the wavelet instantaneous energy and continuous kurtosis, and obtain the fused characteristic instantaneous energy;

[0010] S5. Calculate the envelope spectrum and extract the fault characteristic frequency.

[0011] Furthermore, in step S2, for the vibration signal x(t) with limited energy, its wavelet transform can be expressed as:

[0012]

[0013] Where a is the scale expansion parameter, u is the time translation parameter, ψ(t) is the mother wavelet function, and a series of related wavelet basis functions can be obtained through the corresponding scale expansion and time translation transformation:

[0014]

[0015] That is, the inner product operation of the vibration signal x(t) and the wavelet basis function ψ(t) is used to obtain the first four sub-signal components with the highest frequency center in the vibration signal.

[0016] Furthermore, in step S2, the Morlet wavelet is used to replace the wavelet basis function, and the Morlet mother wavelet function is:

[0017]

[0018] Its wavelet instantaneous energy can be defined as:

[0019] E(a,u)=|f(a,u) 2 ,

[0020] That is, the instantaneous energy of the wavelet at time u is obtained through the scale parameter a.

[0021] Furthermore, in step S3, the kurtosis of the vibration signal in the offshore wind turbine is defined as:

[0022]

[0023] Where p(x) is the probability density function, σ is the standard deviation, is the signal mean;

[0024] The signal kurtosis in the t-th local observation window is defined as:

[0025]

[0026] In the formula, s is the scale parameter, which indicates the window width; v is the translation parameter, which indicates the step size of the window translation; is the mean value of the signal in the window, σt is the standard deviation of the signal in the window; 0<v<s, the continuous kurtosis of the original signal can be calculated based on the parameters of the local observation window.

[0027] Furthermore, in step S4, the local correlation coefficient between the instantaneous wavelet energy and the continuous kurtosis is:

[0028]

[0029] Where M is the width of the localized observation window, t is the center of the localized observation window, and n is the delay parameter. The continuous kurtosis and wavelet instantaneous energy are fused according to the mutual correlation coefficient:

[0030] E k (a, t) = E(a, t)kurt(t)R(t),

[0031] According to the energy of the vibration signal component, the instantaneous energy of the fusion feature is adjusted:

[0032]

[0033] In the formula, E k (a, t) is the instantaneous energy of the fused feature, E(a) is the total energy of the signal component corresponding to scale a, and E k (a) is the total energy of the fused features, E r (a, t) is the adjusted instantaneous energy of the fusion feature, and the reconstructed fault feature signal f(t) is:

[0034]

[0035] Among them, s(t) is the original signal, sgn[·] is the sign function, expressed as:

[0036]

[0037] Furthermore, in step S4, the envelope signal is extracted according to Hilbert transform, and the envelope signal e(t) of the fault characteristic signal f(t) can be expressed as:

[0038]

[0039] in,

[0040]

[0041] Therefore, through envelope spectrum analysis, obvious impact characteristics and noise components can be identified to determine whether there is a fault.

[0042] Compared with the prior art, the present invention has the following advantages:

[0043] (1) The method of the present invention combines the kurtosis index and the wavelet transform, making full use of their respective advantages in reflecting the transient impact characteristics of the signal. The kurtosis index is extremely sensitive to the impact component in the signal, while the wavelet transform has good time-frequency localization characteristics and can carefully depict the time-varying characteristics of the signal. The combination of the two effectively makes up for the shortcomings of a single method and improves the extraction accuracy of weak fault features. By integrating the observation results of the kurtosis index and the wavelet transform, the present application can approach the optimal observation direction of the fault feature in the generalized high-dimensional space to the greatest extent, thereby realizing the effective extraction of early weak fault features.

[0044] (2) Faced with the complex and ever-changing actual working environment, this technical solution has demonstrated strong robustness and adaptability. Regardless of the changes in the signal noise level or the fluctuations in the equipment status, it can maintain a high fault detection accuracy by adjusting the fusion strategy. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 The present invention is a flow chart of the method. DETAILED DESCRIPTION

[0046] In order to enable those skilled in the art to more clearly understand the technical solution of the present application, a method for extracting early fault features of offshore wind power according to the present invention is further described below in conjunction with the accompanying drawings.

[0047] like Figure 1 As shown, a method for extracting early fault features of offshore wind power is described, and the specific steps are as follows:

[0048] (a) Collect vibration signals from offshore wind turbines through motion sensors.

[0049] The vibration signals monitored by the vibration monitoring system, SCADA, CMS or other motion sensors in the offshore wind turbine are collected through motion sensors.

[0050] (ii) Perform wavelet decomposition on the vibration signal and decompose it into four sub-signal components from high to low frequency.

[0051] For a vibration signal x(t) with limited energy, its wavelet transform can be expressed as:

[0052]

[0053] Where a is the scale expansion parameter, u is the time translation parameter, ψ(t) is the mother wavelet function, and a series of related wavelet basis functions can be obtained through the corresponding scale expansion and time translation transformation:

[0054]

[0055] Wavelet transform is the inner product operation of the vibration signal x(t) and the wavelet basis function ψ(t), which is used to obtain the first four sub-signal components with the highest frequency center in the vibration signal. Specifically: wavelet transform is used to decompose the vibration signal into sub-signals with different frequency centers. The selection of different wavelet basis functions directly determines the characteristics of the signal components obtained by wavelet transform. After multi-level wavelet decomposition of the original vibration signal, the signal is decomposed into an approximate part (low-frequency part) and a detail part (high-frequency part) at each level of decomposition. As the number of decomposition levels increases, the frequency center of the detail part gradually increases. Determine how many levels of wavelet decomposition to perform based on the sampling frequency of the signal and the frequency range to be analyzed (generally not more than 4 levels to avoid the signal strength being too low and losing the observation significance), and then select the first four sub-signal components with the highest frequency center based on the results.

[0056] For offshore wind power transmission systems, the Morlet wavelet with the highest similarity to the characteristic waveform of the fault vibration state is used as an application example, which can be replaced by other wavelet functions. The Morlet wavelet is used to replace the wavelet basis function, and the Morlet mother wavelet function is:

[0057]

[0058] Its wavelet instantaneous energy can be defined as:

[0059] E(a,u)=|f(a,u) 2 ,

[0060] That is, the instantaneous energy of the wavelet at time u is obtained through the scale parameter a.

[0061] (iii) Based on the continuous kurtosis, the vibration signal is processed to obtain the continuous kurtosis value of the signal.

[0062] The kurtosis of the vibration signal in an offshore wind turbine is defined as:

[0063]

[0064] Where p(x) is the probability density function, σ is the standard deviation, is the signal mean.

[0065] Since the components of the unit vibration signal are complex, using kurtosis as a macro index cannot reflect the detailed characteristics of the signal. The signal kurtosis in the t-th local observation window is defined as:

[0066]

[0067] In the formula, s is the scale parameter, which indicates the window width; v is the translation parameter, which indicates the step size of the window translation; is the mean value of the signal in the window, σ tis the standard deviation of the signal in the window; 0<v<s, the continuous kurtosis of the original signal can be calculated based on the parameters of the local observation window.

[0068] (iv) Fuse the wavelet instantaneous energy and continuous kurtosis (i.e., steps 2 and 3 are equivalent to two parallel steps, processing the original vibration signal respectively), reconstruct the fault characteristic signal, calculate the local mutual correlation coefficient of the wavelet instantaneous energy and the continuous kurtosis, and obtain the fused characteristic instantaneous energy.

[0069] Local cross-correlation coefficient of wavelet instantaneous energy and continuous kurtosis:

[0070]

[0071] Where M is the width of the localized observation window, t is the center of the localized observation window, and n is the delay parameter. The continuous kurtosis and wavelet instantaneous energy are fused according to the mutual correlation coefficient:

[0072] E k (a, t) = E(a, t)kurt(t)R(t),

[0073] According to the energy of the vibration signal component, the instantaneous energy of the fusion feature is adjusted:

[0074]

[0075] In the formula, E k (a, t) is the instantaneous energy of the fused feature, E(a) is the total energy of the signal component corresponding to scale a, and E k (a) is the total energy of the fused features, E r (a, t) is the adjusted instantaneous energy of the fusion feature, and the reconstructed fault feature signal f(t) is:

[0076]

[0077] Among them, s(t) is the original signal, sgn[·] is the sign function, expressed as:

[0078]

[0079] (V) Calculate the envelope spectrum and extract the fault characteristic frequency.

[0080] According to the Hilbert transform to extract the envelope signal, the envelope signal e(t) of the fault characteristic signal f(t) can be expressed as:

[0081]

[0082] in,

[0083]

[0084] Therefore, envelope spectrum analysis can identify obvious impact characteristics and noise components, thereby determining whether there is a fault. Specifically, after the vibration signal is processed, envelope spectrum analysis can amplify the fault signal and reduce the noise, thereby identifying the fault signal in a noisy environment. If there is a fault, the peak value can be clearly seen in the frequency spectrum.

[0085] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for extracting early fault features of offshore wind power, characterized in that: The specific steps include: S1. Collect vibration signals from offshore wind turbines through motion sensors; S2. Perform wavelet decomposition on the vibration signal and decompose it into four sub-signal components from high to low frequency; S3. Based on the continuous kurtosis, the vibration signal is processed to obtain the continuous kurtosis value of the signal; S4. Fuse the wavelet instantaneous energy and continuous kurtosis, reconstruct the fault characteristic signal, calculate the local mutual correlation coefficient of the wavelet instantaneous energy and continuous kurtosis, and obtain the fused characteristic instantaneous energy; S5. Calculate the envelope spectrum and extract the fault characteristic frequency.

2. The offshore wind power early fault feature extraction method according to claim 1 is characterized in that: In step S2, for the vibration signal x(t) with limited energy, its wavelet transform can be expressed as: Where a is the scale expansion parameter, u is the time translation parameter, ψ(t) is the mother wavelet function, and a series of related wavelet basis functions can be obtained through the corresponding scale expansion and time translation transformation: That is, the inner product operation of the vibration signal x(t) and the wavelet basis function ψ(t) is used to obtain the first four sub-signal components with the highest frequency center in the vibration signal.

3. The offshore wind power early fault feature extraction method according to claim 2 is characterized in that: In step S2, the Morlet wavelet is used to replace the wavelet basis function, and the Morlet mother wavelet function is: Its wavelet instantaneous energy can be defined as: E(a,u)=|f(a,u) 2 , That is, the instantaneous energy of the wavelet at time u is obtained through the scale parameter a.

4. The offshore wind power early fault feature extraction method according to claim 3 is characterized in that: In step S3, the kurtosis of the vibration signal in the offshore wind turbine is defined as: Where p(x) is the probability density function, σ is the standard deviation, is the signal mean; The signal kurtosis in the t-th local observation window is defined as: In the formula, s is the scale parameter, which indicates the window width; v is the translation parameter, which indicates the step size of the window translation; is the mean value of the signal in the window, σ t is the standard deviation of the signal in the window; 0<v<s, the continuous kurtosis of the original signal can be calculated based on the parameters of the local observation window.

5. The offshore wind power early fault feature extraction method according to claim 4 is characterized in that: In step S4, the local correlation coefficient between the instantaneous wavelet energy and the continuous kurtosis is: Where M is the width of the localized observation window, t is the center of the localized observation window, and n is the delay parameter. The continuous kurtosis and wavelet instantaneous energy are fused according to the mutual correlation coefficient: E k (a,t)=E(a,t)kurt(t)R(t), According to the energy of the vibration signal component, the instantaneous energy of the fusion feature is adjusted: In the formula, E k (a, t) is the instantaneous energy of the fused feature, E(a) is the total energy of the signal component corresponding to scale a, and E k (a) is the total energy of the fused features, E r (a, t) is the adjusted instantaneous energy of the fusion feature, and the reconstructed fault feature signal f(t) is: Among them, s(t) is the original signal, sgn[·] is the sign function, expressed as:

6. The offshore wind power early fault feature extraction method according to claim 5 is characterized in that: In step S4, the envelope signal is extracted according to Hilbert transform, and the envelope signal e(t) of the fault characteristic signal f(t) can be expressed as: in, The obvious impact characteristics and noise components are identified through envelope spectrum analysis to determine whether there is a fault.

Images