CN103759891B - A kind of uneven on-line fault diagnosis method of double-fed wind power generator blade - Google Patents

Abstract

The invention provides the uneven on-line fault diagnosis method of a kind of double-fed wind power generator blade belonged in technical field of wind power generation.The power signal P that the rotating speed of the method synchronized sampling double-fed generator and generator export; To sample the generated output power and rotating speed that obtain according to axial fan hub moment of inertia J and sampling period inter-sync, estimate the input pneumatic torque T of axial fan hub; Order ratio analysis obtains resampling pneumatic torque Ta, eliminates the frequency ambiguity phenomenon that rotation speed change is brought; Power spectrumanalysis is carried out to the signal after order ratio analysis reconstruct, carries out feature extraction; Relatively 1 double frequency power spectral amplitude ratio Q (1) and 3 double frequency power spectral amplitude ratio Q (3), if Q (1) is greater than Q (3), judges that imbalance fault occurs.The present invention is according to double-fed wind power generator rotating speed, voltage, the current signal quick diagnosis wind power generation unit blade imbalance fault of synchronized sampling, and the normal work of the safety for wind power generating set provides safeguard.The sensor used is simple, easy for installation, is well suited for onsite application.

Description

A kind of uneven on-line fault diagnosis method of double-fed wind power generator blade

Technical field

The present invention relates to a kind of uneven on-line fault diagnosis method of double-fed wind power generator blade, belong to technical field of wind power generation.

Background technology

The meaning of wind-powered electricity generation---wind-power electricity generation is that country advocates the renewable green energy resource of one greatly developed, and can effectively save water, coal resource, alleviate atmospheric pollution, and to profound significance of having preserved the ecological environment.There is abundant wind resource in China, government using wind-power electricity generation as improving energy structure, reply climate change, improves one of main substitute energy technology of energy security.Have data to estimate the year two thousand twenty, the electric power supply had more than 10% is derived from wind-force by the whole world.Wind-power electricity generation single-machine capacity compared with thermoelectricity, nuclear power is little, and floor area is wide, and wind power plant is located in an outlying district usually, and technical conditions, service condition are generally poor, therefore the reliability of wind generator system and security most important.Blade catches the critical component of wind energy as blower fan, and its performance directly affects overall performance and the power generating quality of blower fan.Under blade is exposed to harsh climate environment for a long time, be vulnerable to blast, lightning, the attack of ice and snow; Meanwhile, long-term work, under the condition of load wide fluctuations, very easily causes crackle, damaged, icing, the generation of the various fault such as bolt looseness.When blade construction hydraulic performance decline or when there is minor failure, current wind energy turbine set SCADA system for want of cannot Timeliness coverage take appropriate counter-measure to the monitoring of necessity of blade.Have research report to-2010 years nineteen ninety-fives Germany and Denmark occur various blower fan accidents done statistical study, find due to fan blade fault cause or accident quantity directly related with it maximum.

Due to the overwhelming majority blade fault as: the faults such as breakage, icing, thunderbolt, blades installation error all will show as quality or pneumatic imbalance, if accurately and timely detect the imbalance fault of blade run duration, to the early detection of various blade fault and state-detection particularly important, can prevent fault pervasion from avoiding large loss to a great extent.

There were some research institutions and companies to develop some wind energy turbine set blade watch-dogs in recent years abroad, mainly contained: infrared imaging analysis, acoustic analysis, video analysis etc.; Ye You blower fan manufacturer is pre-buried Fibre Optical Sensor when production blade, to obtain the running state information of blade.At present, most detection means is expensive but also installation complexity not only, is difficult to apply; Particularly as Fibre Optical Sensor, detection means is difficult to be applied to the blower fan put into operation, can not meet the monitoring to whole wind energy turbine set blade operation conditions.

In order to solve online fan blade fault diagnosis and condition monitoring gordian technique, domestic many research institutions and blower fan manufacturer have also carried out the correlative study for blade fault diagnosing and condition monitoring, also some patents for fan blade on-line monitoring are disclosed recently: as " wind power generating motor and blade state on-line computing model " (CN202305007U), " wind generator set blade real-time state monitoring and fault diagnosis system and method " (CN102539438A), " a kind of blade crack of wind driven generator automatic Testing Alarm System " (CN202533424U).Substantially adopt video in above-mentioned patent, foil gauge, the modes such as sound detect blade and whether catastrophic failure occur, these modes or installing complexity is not suitable for on-line fault diagnosis, or accuracy in detection is not high, is difficult to detect blade noncritical failure." the blade imbalance fault diagnostic method based on the Wind turbines of current signal " (201210396295.X) utilizes the three-phase current signal of generator, the vibration frequency signal extracted wherein carries out the diagnosis of blade imbalance fault, the method can solve straight drive blower blade imbalance fault preferably, but for double-fed unit owing to there is excitation con-trol, the output signal fluctuation of generator will greatly reduce, and be difficult to directly utilize this signal to carry out fault diagnosis.

Summary of the invention

For the problems referred to above, this invention exploits a kind of uneven on-line fault diagnosis method of double-fed wind power generator blade, export in conjunction with wheel speed and generator, carry out the diagnosis of blade imbalance fault online.

Technical scheme of the present invention is,

A kind of uneven on-line fault diagnosis method of double-fed wind power generator blade, the method step is as follows:

1, the rotating speed (ω) of synchronized sampling double-fed generator and the power signal P of generator output.Because double-fed wind power generator group is that variable speed constant frequency runs, so it is non-stationary signal that its generator exports, in order to solve non-stationary problem, the data that the present invention uses are taked to utilize synchronized sampler to sample.

2, to sample the generated output power and rotating speed that obtain according to axial fan hub moment of inertia J and sampling period inter-sync, estimate the input pneumatic torque T of axial fan hub.

The input pneumatic torque T method of estimation of axial fan hub is:

$T=(P\×r)/(\mathrm{\ω}\×\mathrm{\η})+J\×\left(\frac{d\mathrm{\ω}}{dt}\right)/r,---\left(1\right)$

Wherein P is that generator exports instantaneous power, and ω is rotating speed or the high speed shaft of gearbox rotating speed of double-fed generator, and r is gear case raising speed ratio. for generator angular acceleration, η is generator conversion efficiency.

Therefore, at certain sampling instant i, the generator that synchronized sampling obtains this moment i exports instantaneous power P _i, the high speed shaft of gearbox rotational speed omega in this moment _i, then the blower fan input pneumatic torque T in this moment _ifor:

T _i＝(P _i×r)/(ω _i×η)+J×(ω _i-ω _i-1)/r，(2)

3, according to rotary speed data, order ratio analysis is carried out to the pneumatic torque data estimating in the sampling period to obtain and obtain resampling pneumatic torque Ta, eliminate the frequency ambiguity phenomenon that rotation speed change is brought;

Described order ratio analysis method is:

Measure engine speed, obtain rotational speed pulse signal; Suppose that certain pulse moment is t ₀, the moment collecting pulse in co-located after engine turns over one week is t ₁, now turning over angle is 2 π; The moment collecting pulse in co-located after turning over two weeks is t ₂, now turning over angle is 4 π; Here, assuming that angular acceleration is 0; That is,

θ(t)＝b ₀+b ₁t+b ₂t ²，(3)

Coefficient can be separated and be:

$\left(\begin{array}{c}{b}_0\\ b_1\\ b_2\end{array}\right)={\left(\begin{array}{ccc}1& t_0& t_0^2\\ 1& t_1& t_1^2\\ 1& t_2& t_2^2\end{array}\right)}^{-1}\left(\begin{array}{c}0\\ 2\mathrm{\π}\\ 4\mathrm{\π}\end{array}\right),---\left(4\right)$

Order ratio analysis requires equiangular sampling, needs calculating generator often to turn over the output valve of specified angle Δ θ;

Suppose within the sampling time, the total angle that generator turns over is Φ, then can obtain the time that generator in the sampling time turns over any k Δ θ angle to be:

$t_k=\frac{1}{2b_2}\[\sqrt{2b_2^2(k\·\mathrm{\Δ}\mathrm{\θ}-b_0)+b_1^2}-b_1\],---\left(5\right)$

K Δ θ is the k times of angle of specified angle Δ θ, k=1...M, and M Δ θ≤Φ;

Signal after reconstruct is:

$T_{ak}=\underset{n=1}{\overset{N}{\Σ}}{T}_i\×Ts\×sin\[{\mathrm{\ω}}_c\×(t_k-n\×Ts)\]/\[\mathrm{\π}\×(t_k-n\×Ts)\]---\left(6\right)$

Wherein: Ts is sampling time interval, N is counting of sampling, ω _cfor the low pass filter cutoff frequency set as required;

4, spectra calculation is carried out to the signal after order ratio analysis reconstruct, and carry out feature extraction.

Described spectra calculation method is:

First discrete Fourier transformation is carried out to Ta,

$y\left(m\right)=\frac{1}{M}\underset{k=1}{\overset{M}{\Σ}}{T}_{ak}\·e^{-jk\left(\frac{2\mathrm{\π}}{N}\right)m}---\left(7\right)$

Wherein: 0≤m≤M-1, j represents and get imaginary number.

In order to the main frequency part in outstanding signal, use power spectrum,

Q(m)＝y(m)·conj(y(m))(8)

Wherein, conj (y (the m)) conjugation that is y (m);

In extraction power spectrum, 1 frequency multiplication Q (1) and 3 frequency multiplication Q (3) amplitude are as characteristic signal.

In order to improve fault diagnosis robustness, Q (1) is desirable centered by 1 frequency multiplication, with the maximum amplitude of specifying radius region σ internal power to compose.Q (3) is desirable centered by 3 frequencys multiplication, with the maximum amplitude of specifying radius region σ internal power to compose.Generally σ gets 0.1.

Q(1)＝MAX[Q(1-σ),Q(1+σ)]，(9)

Q(3)＝MAX[Q(3-σ),Q(3+σ)]，(10)

5, compare the feature under normal condition and the feature under malfunction, obtain diagnostic result;

Q (1) <Q (3) under nominal situation, and Q (1) is far longer than Q (3) under imbalance fault, the present invention utilizes Q (1) whether to be greater than Q (3) as judging the mark whether imbalance fault occurs, namely 1 double frequency power spectral amplitude ratio Q (1) and 3 double frequency power spectral amplitude ratio Q (3) are compared, if Q (1) is greater than Q (3), judge that imbalance fault occurs.

The beneficial effect of the invention is:

1. the present invention can according to the double-fed wind power generator rotating speed of synchronized sampling, voltage, current signal quick diagnosis wind turbine blade imbalance fault, and the normal work of the safety for wind power generating set provides safeguard.

2. the sensor of the present invention's use is simple, easy for installation, is well suited for onsite application.

Accompanying drawing explanation

Fig. 1 is process flow diagram of the present invention.

Fig. 2 is order ratio analysis signal reconstruction process flow diagram.

Fig. 3 is generating unit speed and generated output power time-domain diagram under nominal situation.

Fig. 4 is generating unit speed and generated output power time-domain diagram under single blade icing operating mode.

Fig. 5 is generating unit speed and generated output power time-domain diagram under biplate blade icing operating mode.

Fig. 6 is the pneumatic Driving Torque time-domain diagram of unit under nominal situation.

Fig. 7 is the pneumatic Driving Torque time-domain diagram of unit under monolithic icing operating mode.

Fig. 8 is the pneumatic Driving Torque time-domain diagram of unit under biplate icing operating mode.

Fig. 9 is unit pneumatic Driving Torque rank specific power spectrogram under nominal situation.

Figure 10 is unit pneumatic Driving Torque rank specific power spectrogram under monolithic icing operating mode.

Figure 11 is unit pneumatic Driving Torque rank specific power spectrogram under biplate icing operating mode.

Embodiment

Below in conjunction with accompanying drawing, by instantiation, the present invention is described in further detail.It is emphasized that following explanation is only exemplary, instead of in order to limit the scope of the invention and apply.

Fig. 1 is process flow diagram of the present invention.A kind of uneven on-line fault diagnosis method of double-fed wind power generator blade, the method step is as follows:

(1) rotational speed omega of synchronized sampling double-fed generator and the power signal P of generator output;

(2) to sample the generated output power and rotating speed that obtain according to axial fan hub moment of inertia J and sampling period inter-sync, estimate the input pneumatic torque T of axial fan hub;

The input pneumatic torque T method of estimation of axial fan hub is:

$T=(P\&times;r)/(\mathrm{\&omega;}\&times;\mathrm{\&eta;})+J\&times;\left(\frac{d\mathrm{\&omega;}}{dt}\right)/r,---\left(1\right)$

Wherein, P is that generator exports instantaneous power, and ω is the rotating speed of double-fed generator, and r is gear case raising speed ratio, and J is wheel rotation inertia, for generator angular acceleration, η is generator conversion efficiency;

Therefore, at certain sampling instant i, synchronized sampling obtains generator and exports instantaneous power P _i, the high speed shaft of gearbox rotational speed omega in this moment _i, then the blower fan input pneumatic torque T in this moment _ifor:

T _i＝(P _i×r)/(ω _i×η)+J×(ω _i-ω _i-1)/r，(2)

(3) according to rotary speed data, order ratio analysis is carried out to the pneumatic torque data estimating in the sampling period to obtain and obtain resampling pneumatic torque Ta, eliminate the frequency ambiguity phenomenon that rotation speed change is brought;

Described order ratio analysis method is:

Measure engine speed, obtain rotational speed pulse signal; Suppose that certain pulse moment is t ₀, the moment collecting pulse in co-located after engine turns over one week is t ₁, now turning over angle is 2 π; The moment collecting pulse in co-located after turning over two weeks is t ₂, now turning over angle is 4 π; Here, assuming that angular acceleration is 0; That is,

θ(t)＝b ₀+b ₁t+b ₂t ²，(3)

Coefficient can be separated and be:

$\left(\begin{array}{c}{b}_0\\ b_1\\ b_2\end{array}\right)={\left(\begin{array}{ccc}1& t_0& t_0^2\\ 1& t_1& t_1^2\\ 1& t_2& t_2^2\end{array}\right)}^{-1}\left(\begin{array}{c}0\\ 2\mathrm{\&pi;}\\ 4\mathrm{\&pi;}\end{array}\right),---\left(4\right)$

Order ratio analysis requires equiangular sampling, needs calculating generator often to turn over the output valve of specified angle Δ θ;

Suppose within the sampling time, the total angle that generator turns over is Φ, then can obtain the time that generator in the sampling time turns over any k Δ θ angle to be:

$t_k=\frac{1}{2b_2}\&lsqb;\sqrt{2b_2^2(k\&CenterDot;\mathrm{\&Delta;}\mathrm{\&theta;}-b_0)+b_1^2}-b_1\&rsqb;,---\left(5\right)$

K Δ θ is the k times of angle of specified angle Δ θ, k=1...M, and M Δ θ≤Φ.

Signal after reconstruct is:

$T_{ak}=\underset{m=1}{\overset{N}{\&Sigma;}}{T}_i\&times;Ts\&times;sin\&lsqb;{\mathrm{\&omega;}}_c\&times;(t_k-m\&times;Ts)\&rsqb;/\&lsqb;\mathrm{\&pi;}\&times;(t_k-m\&times;Ts)\&rsqb;,---\left(6\right)$

Wherein: Ts is sampling time interval, N is counting of sampling, ω _cfor the low pass filter cutoff frequency set as required; t _kfor the sampling time of specified angle;

(4) spectra calculation is carried out to the signal after order ratio analysis reconstruct, and carry out feature extraction;

Described spectra calculation method is:

First discrete Fourier transformation is carried out to Ta,

$y\left(m\right)=\frac{1}{M}\underset{k=1}{\overset{M}{\&Sigma;}}{T}_{ak}\&CenterDot;e^{-jk\left(\frac{2\mathrm{\&pi;}}{N}\right)m},---\left(7\right)$

Wherein: 0≤m≤M-1, j represents and get imaginary number.

In order to the main frequency part in outstanding signal, use power spectrum,

Q(m)＝y(m)·conj(y(m))，(8)

Wherein, conj (y (the m)) conjugation that is y (m);

In extraction power spectrum, 1 frequency multiplication Q (1) and 3 frequency multiplication Q (3) amplitude are as characteristic signal.

(5) compare the feature under normal condition and the feature under malfunction, obtain diagnostic result;

Relatively 1 double frequency power spectral amplitude ratio Q (1) and 3 double frequency power spectral amplitude ratio Q (3), if Q (1) is greater than Q (3), judges that imbalance fault occurs.

Described synchronized sampling is sampled for utilizing synchronized sampler.

With certain 2MW double-fed wind power generator for research object, its hub rotation inertia J=4.15 × 10 ⁸kgm ², gear case raising speed is than being r=83.3, generator efficiency eta=95%.Rotating speed in synchronized sampling unit running process and generator export, and sample frequency is 250Hz, and the sampling time is 180 seconds.Experiment, wind speed about 8 meters, has been carried out the experiment under three kinds of situations respectively, has been respectively normal running experiment, and monolithic icing 170Kg off-center operation is tested, and biplate icing 170Kg off-center operation is tested.

Fig. 3 is rotating speed under nominal situation and power stage time-domain signal, and Fig. 4 is rotating speed under monolithic icing operating mode and power stage time-domain signal, and Fig. 5 is rotating speed under biplate icing operating mode and power stage time-domain signal.Upper as can be seen from figure, the operation curve under several operating mode is similar, is difficult to direct differentiation and whether occurs imbalance fault.

Utilize T _i=(P _i× r)/(ω _i× η)+J × (ω _i-ω _i-1)/r estimates the pneumatic input torque under each sampling instant.Fig. 6 is pneumatic input torque time-domain diagram under nominal situation, can find out, due to tower shadow effect, periodic fluctuation has appearred in pneumatic input torque.Fig. 7 is pneumatic input torque time-domain diagram under monolithic icing operating mode, and due to tower shadow effect and imbalance fault impact, periodic fluctuation has appearred in pneumatic input torque.Fig. 8 is pneumatic input torque time-domain diagram under biplate icing operating mode, and due to tower shadow effect and imbalance fault impact, periodic fluctuation has appearred in pneumatic input torque.Can find out, the periodicity that under several operating mode, pneumatic input torque occurs is different.The input torque that tower shadow effect causes periodically fluctuates closely related with rotating speed, and due to time-domain signal be wait the sampling period sampling, its spectrogram there will be the phenomenon of frequency ambiguity.

Carry out order ratio analysis to the pneumatic input torque time-domain signal under three kinds of operating modes respectively, Fig. 2 is order ratio analysis signal reconstruction process flow diagram, obtains reconstruction signal.In general, blower fan has three blades, and because tower shadow effect can cause 3 times to the fluctuation of rotating speed on input pneumatic torque, this is the major frequency components of pneumatic input torque, and the frequency amplitude of other compositions will be far smaller than this frequency.In order to the major frequency components of outstanding input torque, spectra calculation is carried out to the pneumatic input torque of reconstruction signal.

Fig. 9 is pneumatic input torque power spectrum chart under nominal situation, can find out, due to tower shadow effect, near 3 frequencys multiplication, occurred the frequency content of higher magnitude, near other frequencys multiplication, frequency content amplitude is all less.Figure 10 is pneumatic input torque time-domain diagram under monolithic icing operating mode, due to tower shadow effect and imbalance fault impact, near 1 frequency multiplication and 3 frequencys multiplication, occurred the frequency content of higher magnitude, and 1 frequency multiplication amplitude composition is especially large, near other frequencys multiplication, frequency content amplitude is all less.Figure 11 is pneumatic input torque time-domain diagram under biplate icing operating mode, due to tower shadow effect and imbalance fault impact, near 1 frequency multiplication and 3 frequencys multiplication, occurred the frequency content of higher magnitude, and 1 frequency multiplication amplitude composition is especially large, near other frequencys multiplication, frequency content amplitude is all less.Therefore, judge that the amplitude size near power spectrum 1 frequency multiplication is the important symbol judging whether wind power generating set breaks down.

Q (1) and Q (3) is obtained by formula (9) and formula (10) according to power spectrum.Through great many of experiments, Q (1) <Q (3) under nominal situation, and Q (1) is far longer than Q (3) under imbalance fault.In the present embodiment, the imbalance fault that Figure 10 and Figure 11 reflects is all the power spectrum amplitude Q's (3) being greater than 3 frequencys multiplication at the power spectrum amplitude Q (1) of 1 frequency multiplication.The present invention utilizes Q (1) whether to be greater than Q (3) as judging the mark whether imbalance fault occurs, namely 1 double frequency power spectral amplitude ratio Q (1) and 3 double frequency power spectral amplitude ratio Q (3) are compared, if Q (1) is greater than Q (3), judge that imbalance fault occurs.

The present invention utilizes double-fed generating unit speed and alternator output signal to carry out the diagnosis of blade of wind-driven generator imbalance fault, and accuracy rate is high, simple, is a kind of reliable double-fed wind power generator blade fault diagnosing method efficiently.

Claims (4)

1. the uneven on-line fault diagnosis method of double-fed wind power generator blade, it is characterized in that, the method step is as follows:

(1) rotational speed omega of synchronized sampling double-fed generator and generator export instantaneous power P;

(2) to sample the generated output power and rotating speed that obtain according to axial fan hub moment of inertia J and sampling period inter-sync, estimate the input pneumatic torque T of axial fan hub;

The input pneumatic torque T method of estimation of axial fan hub is:

$T=(P\&times;r)/(\mathrm{\&omega;}\&times;\mathrm{\&eta;})+J\&times;\left(\frac{d\mathrm{\&omega;}}{dt}\right)/r,---\left(1\right)$

Wherein, P is that generator exports instantaneous power, and ω is the rotating speed of double-fed generator, and r is gear case raising speed ratio, for generator angular acceleration, η is generator conversion efficiency;

Therefore, at certain sampling instant i, the generator that synchronized sampling obtains this moment i exports instantaneous power P _i, the high speed shaft of gearbox rotational speed omega in this moment _i, then the blower fan input pneumatic torque T in this moment _ifor:

T _i＝(P _i×r)/(ω _i×η)+J×(ω _i-ω _i-1)/r，(2)

(3) according to rotary speed data, order ratio analysis is carried out to the pneumatic torque data estimating in the sampling time to obtain and obtain resampling pneumatic torque Ta, eliminate the frequency ambiguity phenomenon that rotation speed change is brought;

Described order ratio analysis method is:

Measure generator speed, obtain rotational speed pulse signal; Suppose that certain pulse moment is t ₀, the moment collecting pulse in co-located after generator turns over one week is t ₁, now turning over angle is 2 π; The moment collecting pulse in co-located after turning over two weeks is t ₂, now turning over angle is 4 π; Here, assuming that angular acceleration is 0; That is,

θ(t)＝b ₀+b ₁t+b ₂t ²，(3)

Coefficient can be separated and be:

$\left(\begin{array}{c}{b}_0\\ b_1\\ b_2\end{array}\right)={\left(\begin{array}{ccc}1& t_0& t_0^2\\ 1& t_1& t_1^2\\ 1& t_2& t_2^2\end{array}\right)}^{-1}\left(\begin{array}{c}0\\ 2\mathrm{\&pi;}\\ 4\mathrm{\&pi;}\end{array}\right),---\left(4\right)$

Order ratio analysis requires equiangular sampling, needs calculating generator often to turn over the output valve of specified angle Δ θ;

Suppose within the sampling time, the total angle that generator turns over is Φ, then obtain the time that generator in the sampling time turns over any k Δ θ angle to be:

$t_k=\frac{1}{2b_2}\&lsqb;\sqrt{2b_2^2(k\&CenterDot;\mathrm{\&Delta;}\mathrm{\&theta;}-b_0)+b_1^2}-b_1\&rsqb;,---\left(5\right)$

Wherein, k Δ θ is the k times of angle of specified angle Δ θ, k=1...M, and M Δ θ≤Φ;

Signal after reconstruct is:

$T_{ak}=\underset{n=1}{\overset{N}{\&Sigma;}}{T}_i\&times;Ts\&times;sin\&lsqb;{\mathrm{\&omega;}}_c\&times;(t_k-n\&times;Ts)\&rsqb;/\&lsqb;\mathrm{\&pi;}\&times;(t_k-n\&times;Ts)\&rsqb;,---\left(6\right)$

Wherein: Ts is sampling time interval, N is counting of sampling, ω _cfor the low pass filter cutoff frequency set as required;

(4) spectra calculation is carried out to the signal after order ratio analysis reconstruct, and carry out feature extraction;

Described spectra calculation method is:

First discrete Fourier transformation is carried out to Ta,

$y\left(m\right)=\frac{1}{M}\underset{k=1}{\overset{M}{\&Sigma;}}{T}_{ak}\&CenterDot;e^{-jk\left(\frac{2\mathrm{\&pi;}}{N}\right)m},---\left(7\right)$

Wherein: 0≤m≤M-1, j represents and get imaginary number;

In order to the main frequency part in outstanding signal, use power spectrum,

Q(m)＝y(m)·conj(y(m))，(8)

Wherein, conj (y (the m)) conjugation that is y (m);

In extraction power spectrum, 1 double frequency power spectral amplitude ratio Q (1) and 3 double frequency power spectral amplitude ratio Q (3) is as characteristic signal;

(5) compare the feature under normal condition and the feature under malfunction, obtain diagnostic result;

Relatively 1 double frequency power spectral amplitude ratio Q (1) and 3 double frequency power spectral amplitude ratio Q (3), if Q (1) is greater than Q (3), judges that imbalance fault occurs.

2. the uneven on-line fault diagnosis method of a kind of double-fed wind power generator blade according to claim 1, it is characterized in that, described synchronized sampling is sampled for utilizing synchronized sampler.

3. the uneven on-line fault diagnosis method of a kind of double-fed wind power generator blade according to claim 1, it is characterized in that, in order to improve fault diagnosis robustness, Q (1) gets centered by 1 frequency multiplication, to specify the maximum amplitude of radius sigma region internal power spectrum; Q (3) gets centered by 3 frequencys multiplication, with the maximum amplitude of specifying radius region σ internal power to compose;

Q(1)＝MAX[Q(1-σ),Q(1+σ)]，(9)

Q(3)＝MAX[Q(3-σ),Q(3+σ)]，(10)。

4. the uneven on-line fault diagnosis method of a kind of double-fed wind power generator blade according to claim 3, it is characterized in that, described σ gets 0.1.