CN106768933A - A kind of wind power generation unit blade method for diagnosing faults based on genetic algorithm - Google Patents

Abstract

The invention discloses a kind of variable K averages wind power generation unit blade method for diagnosing faults based on genetic algorithm, it is pre-processed mainly for typical failure situation, the wind power generation unit blade fault data to gathering, and obtains fault characteristic information；Data are standardized with PCA PCA, and are extracted and is caused the principal character information of wind power generation unit blade failure；The major failure characteristic information that will be extracted, using the variable K mean cluster algorithm based on genetic algorithm, cluster analysis is carried out to training sample.Present invention incorporates genetic algorithm, the calculating that K values are carried out on the basis of traditional K mean algorithms is selected, and shortens the fault information analysis time, improve data clusters analytical precision, cost is reduced, so that the process of fault diagnosis becomes simple and reliable, it is effective and feasible.

Description

A kind of wind power generation unit blade method for diagnosing faults based on genetic algorithm

Technical field

The present invention relates to method for diagnosing faults field, specifically, a kind of wind-powered electricity generation based on genetic algorithm is related specifically to Turbines vane method for diagnosing faults.

Background technology

In recent years, the situation of environmental pollution and energy crisis has growed in intensity, and is badly in need of a kind of environmental protection new energy Occur.Wind-powered electricity generation is increasingly paid attention to as a kind of renewable, without discharge new energy by world community, and this is also caused Wind generating technology is flourished in recent years.In China, a large amount of foundation of wind power plant and come into operation, will be loud Answering government improves energy resource structure, the call of reply climate change, realizes that progressively transition is embodied traditional energy to new energy.

In addition to renewable, pollution-free, wind-power electricity generation also has aboundresources in China, and floor space is wide, and unit holds The advantages of measuring small.But, generally all built in remote districts due to wind power plant, also have that technical conditions are poor, bad environments etc. are asked Topic.Meanwhile, the input application of large-scale wind power unit also makes its security and stability obtain the great attention of people.Due to wind-powered electricity generation Unit long-term work is in severe natural environment so that wind field wind regime is complicated and changeable, easily triggers the generation of various failures.

Therefore, fault diagnosis link is essential.And one of primary clustering that the blade of blower fan catches as wind energy, Except being influenceed by extreme natural environment, the fatigue damage for being subject to be produced by load fluctuation and quick change, leaf are also easy to The failures such as wheel unbalance loading, so as to cause catch wind efficiency to decline, severe patient even cannot continue to come into operation.

At present, for blade fault diagnosing, mostly using the method for neutral net.Although neural network algorithm has accurate Degree is high, and Serial Distribution Processing ability is strong, and distribution storage and learning ability are strong, can fully approach the non-linear relation of complexity etc. excellent Point, but obvious inferior position is there is also simultaneously, and if desired for a large amount of threshold values and weights are set in advance, learning time is long, sample Network excessively complexity etc. during substantial amounts.

The content of the invention

It is with low cost it is an object of the invention to be directed to deficiency of the prior art, there is provided a kind of simple and easy to apply, during process Between short novel wind power turbines vane method for diagnosing faults, i.e., the variable K averages wind power generation unit blade failure based on genetic algorithm Diagnostic method, so that the process of fault diagnosis becomes the safe and reliable property that simple possible improves wind power generation unit blade.

Technical problem solved by the invention can be realized using following technical scheme：

A kind of wind power generation unit blade method for diagnosing faults based on genetic algorithm, comprises the following steps：

1) fault data of wind power generation unit blade is gathered, then the fault data of the wind power generation unit blade to collecting is carried out Pretreatment, obtains fault characteristic information, and fault characteristic information is standardized；

2) fault characteristic information after standardization is analyzed with PCA PCA, extracting causes The principal character information of wind power generation unit blade failure；

3) the principal character information that will be extracted, using the variable K mean cluster algorithm based on genetic algorithm, to wanting feature Information carries out cluster analysis, obtains wind-powered electricity generation failure analysis result；

4) above-mentioned wind-powered electricity generation failure analysis result and Mishap Database or expert knowledge library are compared, obtains final failure The analysis result of type, finally includes on human-computer interaction interface analysis result.

Further, the step of PCA PCA is as follows：

A1) data normalization, formula isWhereinThen the data matrix of new input variable is

A2 correlation matrix) is sought,

Wherein, r_ij(i, j=1,2 ..., p) it is variable after standardizationWithCoefficient correlation, its computing formula is

A3 the characteristic value and characteristic vector of coefficient matrix R) are sought, λ is designated as_i=(i=1,2 ..., p), character pair vector It is e_i(i=1,2 ..., p)；

A4 main composition contribution rate, contribution rate of accumulative total) are calculated and principal component number is determined.Belong in population variance i-th it is main into Divide z_iRatio be referred to as contribution rate：The contribution rate sum of preceding i principal component is referred to as w₁, w₂..., w_iContribution rate of accumulative totalTypically taking contribution rate of accumulative total reaches 85% To 95% characteristic value corresponding to λ₁, λ₂..., λ_mThe 1st principal component, the 2nd principal component ... ... m (m≤p) individual principal component.

A5 the input variable data matrix z based on principal component) is calculated,

Further, the step of cluster analysis is as follows：

B1) using the process of genetic algorithm, initial population is generated, sets end condition, i.e. genetic algorithm iterations；

B2 the individuality for screening is optimized using K mean algorithms), and replaces original with the individuality after optimization Body；

B3) individuality in population is selected, is intersected, mutation operation, and recalculate K values in the completed；

B4) repeat step b2) and b3) to meeting end condition.

Compared with prior art, beneficial effects of the present invention are as follows：

1st, pretreatment is standardized to the wind power generation unit blade fault data for gathering, reduces fault characteristic information numerical value On difference, making the process of cluster analysis becomes easy.

2nd, the principal character information in fault message is extracted by PCA, the complexity of data is reduced, data are realized Uncorrelated and model simplification.

3rd, using the variable K mean cluster algorithm based on genetic algorithm, improve having for fault characteristic information cluster analysis Effect property, feasibility and efficiency.

Brief description of the drawings

Fig. 1 is wind power generation unit blade method for diagnosing faults schematic flow sheet of the present invention.

Fig. 2 is the schematic flow sheet of variable K mean algorithms of the present invention.

Specific embodiment

For technological means, creation characteristic, reached purpose and effect for making present invention realization are easy to understand, with reference to Specific embodiment, is expanded on further the present invention.

Referring to Fig. 1 and Fig. 2, the variable K averages wind power generation unit blade failure that yiz of the present invention is based on genetic algorithm is examined Disconnected method, specific implementation process is as follows：

Step one be fault data monitor with collection, i.e., advanced row data acquisition with standardization, i.e., for typical failure Situation, is monitored and gathers to wind power generation unit blade fault data, and fault data is standardized and is tentatively pre-processed, Obtain fault characteristic information.

Step 2 is to use PCA, i.e. principal component analysis, and extracting causes the principal character of wind power generation unit blade failure to be believed Breath.PCA concrete operation steps are：

Data normalization, formula isWherein

Then the data matrix of new input variable is

(2) correlation matrix is sought,

Wherein r_ij(i, j=1,2 ..., p) it is variable after standardizationWithCoefficient correlation, its computing formula is

(3) characteristic value and characteristic vector of coefficient matrix R are sought, λ is designated as_i=(i=1,2 ..., p), character pair vector It is e_i(i=1,2 ..., p)

(4) main composition contribution rate, contribution rate of accumulative total are calculated and principal component number is determined.Belong in population variance i-th it is main into Divide z_iRatio be referred to as contribution rate：The contribution rate sum of preceding i principal component is referred to as w₁, w₂... ..., w_iContribution rate of accumulative totalIt is general take contribution rate of accumulative total reach 85% to λ corresponding to 95% characteristic value₁, λ₂..., λ_mThe 1st principal component, the 2nd principal component ... ... m (m≤p) individual principal component.

(5) the input variable data matrix z based on principal component is calculated,

Step 3 is with fault cluster model, i.e., to carry out cluster point using the variable K mean algorithms based on genetic algorithm Analysis.

Step 4 be failure explanation facility, will wind-powered electricity generation failure analysis result carried out with Mishap Database or expert knowledge library Compare, obtain the analysis result of final fault type.

Step 5 is to include on human-computer interaction interface result.

Said process is to fault data cluster analysis main procedure.

Fig. 2 is fault message Clustering Model, i.e., the variable K mean algorithms concrete operations figure based on genetic algorithm.Step It is rapid as follows：

Step one is the data prediction that fault message is carried out with PCA.

Step 2 is designed to carry out chromosome coding design and fitness function.Because wind power generation unit blade failure is with more Dimension, the characteristics of quantity is big, therefore, it is to combine actual conditions, the present invention is using the floating-point encoding based on cluster centre.Dyeing Body structure isWherein K is the length of gene, is the number of cluster centre for randomly generating,It is dye The gene that colour solid is l+1, it has n dimension.x_l∈ (1,2 ..., c), l=1,2 ..., K.Fitness function is for commenting Valency is individual, the individual good and bad standard of difference.Fitness function of the invention is using the object function in cluster.

Step 3 sets for initialization information, including population generation and end condition setting.Wherein, end condition sets and is It is regulation genetic algorithm iterations, more than i.e. termination.

Step 4 is intersected and mutation operation, and carry out cluster analysis with reference to K mean algorithms to be selected.

(1) select.In order to ensure that it is bigger that the bigger chromosome of fitness function remains into follow-on probability, use herein Classical roulette algorithms selection operator in genetic algorithm.Specific practice is as follows：

1) K averaging operations are carried out, chromosome is updated；

2) according to fitness function, calculate the fitness of current population chromosome, and record wherein fitness it is maximum Body；

3) select probability of each individuality is calculated further according to individual fitness value.

(2) intersect.It is random to select M/2 to chromosome as parent from population, to each pair chromosome, produce random number P (0 ＜ P ＜ 1), as ＜ P_cWhen, in two chromosome i, crosspoint is randomly generated in j, intersection fortune is carried out to the gene behind crosspoint Calculate, and recalculate the chromosome length after intersecting, untill all individualities intersect.

(3) make a variation.Individual, each gene position to each, produces random number P, as P ＜ P_mWhen, the gene position is entered Row random variation computing, generation population of future generation.(wherein P_mIt is crossover probability).

Step 5 to check chromosome length, when chromosome length is more than maximum cluster number, by this chromosome from Leave out in population, otherwise perform step 7.

Step 6 be by make a variation and by after inspection produce chromogene centered on, true K values, carry out K averages again Cluster, then replaces the gene in original chromosome with the result of K mean cluster.Meanwhile, judge whether to meet termination bar Whether part, i.e., reach genetic algorithm iterations set in advance.If not up to, repeating to step 4 and being selected, intersect And variation, optimized with K averages individual.If reaching, output result.

General principle of the invention and principal character and advantages of the present invention has been shown and described above.The technology of the industry Personnel it should be appreciated that the present invention is not limited to the above embodiments, simply explanation described in above-described embodiment and specification this The principle of invention, without departing from the spirit and scope of the present invention, various changes and modifications of the present invention are possible, these changes Change and improvement all fall within the protetion scope of the claimed invention.The claimed scope of the invention by appending claims and its Equivalent thereof.

Claims (3)

1. a kind of wind power generation unit blade method for diagnosing faults based on genetic algorithm, it is characterised in that comprise the following steps：

1) fault data of wind power generation unit blade is gathered, then the fault data of the wind power generation unit blade to collecting carries out pre- place Reason, obtains fault characteristic information, and fault characteristic information is standardized；

2) fault characteristic information after standardization is analyzed with PCA PCA, extracting causes wind-powered electricity generation The principal character information of turbines vane failure；

3) the principal character information that will be extracted, using the variable K mean cluster algorithm based on genetic algorithm, to wanting characteristic information Cluster analysis is carried out, wind-powered electricity generation failure analysis result is obtained；

4) above-mentioned wind-powered electricity generation failure analysis result and Mishap Database or expert knowledge library are compared, obtains final fault type Analysis result, finally by analysis result include on human-computer interaction interface.

2. the wind power generation unit blade method for diagnosing faults based on genetic algorithm according to claim 1, it is characterised in that institute The step of stating PCA PCA is as follows：

A1) data normalization, formula isWhereinThen the data matrix of new input variable is

A2 correlation matrix) is sought,

$R=\left[\begin{array}{cccc}{r}_{11}& r_{12}& ...& r_{1p}\\ r_{21}& r_{22}& ...& r_{2p}\\ ...& ...& ...& ...\\ r_{n1}& r_{n2}& ...& r_{pp}\end{array}\right],$

Wherein, r_ij(i, j=1,2 ..., p) it is variable after standardizationWithCoefficient correlation, its computing formula is

$r_{ij}={\mathrm{\Σ}}_{k=1}^n(x_{kj}^{*}-{\stackrel{\‾}{x}}_j^{*})/\sqrt{{\mathrm{\Σ}}_{k=1}^n{\left(x_{ki}^{*}-{\stackrel{\‾}{x}}_i^{*}\right)}^2{\mathrm{\Σ}}_{k=1}^n{\left(x_{kj}^{*}-{\stackrel{\‾}{x}}_j^{*}\right)}^2}$

A3 the characteristic value and characteristic vector of coefficient matrix R) are sought, λ is designated as_i=(i=1,2 ..., p), character pair vector is e_i(i =1,2 ..., p)；

A4 main composition contribution rate, contribution rate of accumulative total) are calculated and principal component number is determined.Belong to i-th principal component z in population variance_i's Ratio is referred to as contribution rate：The contribution rate sum of preceding i principal component is referred to as w₁, w₂..., w_iContribution rate of accumulative totalTypically take the spy that contribution rate of accumulative total reaches 85% to 95% λ corresponding to value indicative₁,λ₂,...,λ_mThe 1st principal component, the 2nd principal component ... ... m (m≤p) individual principal component.

A5 the input variable data matrix z based on principal component) is calculated,

$z=\left[\begin{array}{cccc}{z}_{11}& z_{12}& ...& z_{1m}\\ z_{21}& z_{22}& ...& z_{2m}\\ ...& ...& ...& ...\\ z_{n1}& z_{n2}& ...& z_{nm}\end{array}\right].$

3. the wind power generation unit blade method for diagnosing faults based on genetic algorithm according to claim 1, it is characterised in that institute The step of stating cluster analysis is as follows：

B1) using the process of genetic algorithm, initial population is generated, sets end condition, i.e. genetic algorithm iterations；

B2 the individuality for screening is optimized using K mean algorithms), and replaces original individuality with the individuality after optimization；

B3) individuality in population is selected, is intersected, mutation operation, and recalculate K values in the completed；

B4) repeat step b2) and b3) to meeting end condition.