CN114065618A - Method for fitting fan power curve based on parameter model of differential evolution - Google Patents

Method for fitting fan power curve based on parameter model of differential evolution Download PDF

Info

Publication number
CN114065618A
CN114065618A CN202111313404.2A CN202111313404A CN114065618A CN 114065618 A CN114065618 A CN 114065618A CN 202111313404 A CN202111313404 A CN 202111313404A CN 114065618 A CN114065618 A CN 114065618A
Authority
CN
China
Prior art keywords
fan
power curve
parameter
model
power
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN202111313404.2A
Other languages
Chinese (zh)
Inventor
孙勇
陈棋
傅凌焜
王琳
杨秦敏
陈积明
孟文超
方静宜
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Zhejiang University ZJU
Zhejiang Windey Co Ltd
Original Assignee
Zhejiang University ZJU
Zhejiang Windey Co Ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Zhejiang University ZJU, Zhejiang Windey Co Ltd filed Critical Zhejiang University ZJU
Priority to CN202111313404.2A priority Critical patent/CN114065618A/en
Publication of CN114065618A publication Critical patent/CN114065618A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/20—Design optimisation, verification or simulation
    • G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/004—Artificial life, i.e. computing arrangements simulating life
    • G06N3/006—Artificial life, i.e. computing arrangements simulating life based on simulated virtual individual or collective life forms, e.g. social simulations or particle swarm optimisation [PSO]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2111/00—Details relating to CAD techniques
    • G06F2111/04—Constraint-based CAD
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2111/00—Details relating to CAD techniques
    • G06F2111/08—Probabilistic or stochastic CAD
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2113/00—Details relating to the application field
    • G06F2113/06—Wind turbines or wind farms
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2119/00—Details relating to the type or aim of the analysis or the optimisation
    • G06F2119/02—Reliability analysis or reliability optimisation; Failure analysis, e.g. worst case scenario performance, failure mode and effects analysis [FMEA]

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Evolutionary Computation (AREA)
  • Software Systems (AREA)
  • Artificial Intelligence (AREA)
  • General Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • General Health & Medical Sciences (AREA)
  • Mathematical Physics (AREA)
  • Computational Linguistics (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • Biophysics (AREA)
  • Biomedical Technology (AREA)
  • Data Mining & Analysis (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Health & Medical Sciences (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Medical Informatics (AREA)
  • Computer Hardware Design (AREA)
  • Geometry (AREA)
  • Wind Motors (AREA)

Abstract

The invention discloses a method for fitting a fan power curve based on a differential evolution parameter model. The method comprises the steps of aiming at data collected by a fan data collecting and monitoring system, and cleaning the data based on wind speed information; a five-parameter logistic model is established to fit the cleaned power curve data set, and the logistic parameter model has the characteristic of high similarity with a standard fan power curve, so that the accuracy of curve fitting is ensured; and identifying the parameters of the five-parameter logistic model by using a differential evolution algorithm for adaptively controlling parameter change, ensuring the smoothness of the fitting curve, and improving the fitting speed at the same time, thereby finally obtaining the five-parameter logistic model for fitting the power curve of the fan. The method is a parameter modeling method based on data, the data cleaning steps are complete, the selected parameter model takes the actual operation data characteristics and the standard fan power curve characteristics into consideration, and certain theoretical value and actual engineering significance are achieved.

Description

Method for fitting fan power curve based on parameter model of differential evolution
Technical Field
The invention relates to an algorithm for fitting a fan power curve, in particular to a method for fitting a fan power curve based on a parameter model of differential evolution.
Background
The fossil energy is limited in storage amount and non-renewable, and meanwhile, serious environmental pollution can be caused, so that the renewable energy is gradually favored by the public, wherein the wind energy is renewable energy with extremely small environmental pollution, has extremely large storage amount and is high-quality energy in the renewable energy. In recent years, the installed capacity of wind power in the world is increased year by year, the permeability of wind power is increased year by year, the installed capacity of onshore wind turbines in China is the first in the world, the offshore wind power is also developed vigorously, and the wind power industry becomes a new energy industry which is vigorously developed at home and abroad.
However, when the wind power industry is vigorously developed, a series of problems of the fan operation and maintenance process are generated due to the continuous degradation of the fan. When the wind turbine generator set actually operates, the natural wind speed has the characteristics of intermittence and uncertainty, the actual wind speed is not accurately measured, so that the operation and maintenance personnel of the wind turbine generator set can evaluate the operation state of the wind turbine generator set to a certain delay, and the operation and maintenance personnel can control the wind turbine generator set and make misjudgment on fault diagnosis. The method has the advantages that the running state and the health condition of the wind turbine generator are correctly evaluated and diagnosed, so that reasonable planning in the aspect of operation and maintenance of the wind turbine generator is facilitated, and meanwhile, the method has practical significance for prolonging the service life of a fan.
In the running process of the wind turbine, a plurality of running state information and fault state information can be intuitively reflected in a power-wind speed curve of the wind turbine, so that accurate fitting of the wind turbine power curve is the primary basis for researching the running state and the health state of the wind turbine. For a data set generated during the actual operation of the fan, how to clean abnormal data, selecting a proper parameter model for fitting the data set, identifying parameters related to the parameter model according to data information, and further analyzing the actual physical significance of the real power curve of the fan is the key content of research in the wind power industry at present. The existing fan power curve fitting method mainly faces the following challenges:
(1) the method for fitting the fan power curve requires that original operation data of the wind turbine generator and reasonable and standard data cleaning are carried out;
(2) the fan power curve fitting method needs to consider the characteristics of a standard power curve and the characteristics of actual operation data at the same time;
(3) the fan power curve fitting method needs to ensure fitting accuracy and fitting speed simultaneously so as to realize online fan state monitoring based on a fan power curve;
therefore, based on the challenges faced by the existing wind turbine generator power curve fitting method, a set of standardized wind turbine generator power curve fitting method needs to be designed to improve the curve fitting process.
Disclosure of Invention
The invention aims to perfect and standardize the defects of the existing research and technology, and provides a method for fitting a fan power curve based on a differential evolution parameter model. According to the method, a set of standardized wind turbine generator power curve fitting method is designed, and the accuracy and reliability of power curve fitting can be improved by performing data cleaning on initial data; the method has the advantages that the parameter model based on the operation data is established, and the characteristics of the standard power curve model and the characteristics of the operation data are considered, so that the model result has certain theoretical value and practical application significance.
The purpose of the invention is realized by the following technical scheme: a method for fitting a fan power curve based on a differential evolution parameter model comprises the following steps:
1) according to the total n pieces of fan operation data information in the corresponding demand period acquired by a supervisory control and data acquisition (SCADA) system of the fan to be evaluated, the wind speed { v ] in the operation data information is converted into the wind speed { viActive power { P }iJ and pitch angle [ beta ]iThe information constitutes a power curve information data set to be preprocessed, which is recorded as
Figure BDA0003342673100000021
Wherein i is 1,2,3, …, n;
2) according to the normal value ranges of the corresponding active power and the pitch angle under different wind speed conditions in the running process of the wind turbine generator, the information data set in the step 1) is compared with the information data set in the step 1)
Figure BDA0003342673100000022
Cleaning the original data set, removing obvious abnormal points to obtain n which is not in accordance with the abnormal point characteristicsnormRunning data is recorded and recorded as a standard running data set of the fan
Figure BDA0003342673100000023
Wherein i is 1,2,3, …, nnorm；
3) Collecting the standard operation data set of the fan cleaned in the step 2)
Figure BDA0003342673100000024
Medium wind speed
Figure BDA0003342673100000025
And active power
Figure BDA0003342673100000026
Establishing a five-parameter logistic model, and recording a power curve model of the fan as P5logistic(v, θ) wherein the wind speed
Figure BDA0003342673100000027
Being independent variable of the power curve model, active power
Figure BDA0003342673100000028
The dependent variable of the power curve model is theta, a set (A, B, C, D, G) of five parameters of the five-parameter logistic model is theta, wherein A represents the lowest power of the fan, B represents the efficiency of the fan for capturing wind energy to the maximum extent, C represents the median of the maximum generating power of the fan, D represents the maximum power of the fan, and G represents the symmetry of the power curve;
4) initial values theta of five parameter sets theta of the five parameter logistic model in the step 3) (A, B, C, D, G)0Setting, namely setting an initial value of a parameter set by combining the meaning of each parameter of the five-parameter logistic model and the physical meaning of a fan power curve, wherein:
setting the initial value of A as active power in fan standard operation data set
Figure BDA0003342673100000029
Minimum value of (d);
setting the initial value of B as the slope of a connecting line between two data points with the maximum Euclidean distance in the fan standard operation data set;
setting the initial value of C as active power in fan standard operation data set
Figure BDA00033426731000000210
The mean of the maximum and minimum values;
setting the initial value of D as active power in fan standard operation data set
Figure BDA00033426731000000211
Maximum value of (d);
the initial value of G is set to 1, which represents that the initial power curve is symmetrical;
5) root of herbaceous plantInitial values theta of five parameter sets theta of the five parameter logistic model according to step 4)0Using a differential evolution method to carry out the wind turbine power curve model P in the step 3)5logisticAnd (v, theta) performing parameter identification on five parameters (A, B, C, D and G) and recording a standard operation data set of the fan
Figure BDA0003342673100000031
Each data point in the data points is an individual and is integrally a population, five parameter sets corresponding to each individual are used as chromosome vectors, difference calculation is carried out by randomly extracting three original chromosome vectors to obtain donor vectors, a binomial crossover operator is adopted to carry out crossover operation to generate test vectors, the two original chromosome vectors and the chromosome vector closest to the optimal chromosome vector in the test vectors are selected as the individuals of a new generation of the population, iteration is repeated until the set times are limited, and then the optimal five parameter sets theta are obtainedbest＝(Abest,Bbest,Cbest,Dbest,Gbest)；
6) Obtaining an optimal fan power curve model according to the optimal five parameter sets obtained in the step 5)
Figure BDA0003342673100000032
Further, in the step 2), different wind speed ranges are divided, and abnormal point detection is carried out on a fan power curve according to abnormal characteristics of the fan running state; the specific criteria for data cleaning based on wind speed information are as follows:
a) when the fan is started, if the information data set is available
Figure BDA0003342673100000033
Medium wind speed viGreater than cut-in wind speed vcut_inActive power PiBelow active power threshold PthresOr the value is negative, the abnormal data point is judged, and the criterion formula is listed as follows:
Figure BDA0003342673100000034
b) if the fan is operated at low wind speed, the information data set
Figure BDA0003342673100000035
Medium wind speed viGreater than cut-in wind speed vcut_inAnd is less than the rated wind speed v of the fanratedAngle of pitch betaiGreater than pitch angle threshold betathresIf the data points are judged to be abnormal data points, the criterion formula is listed as follows:
Figure BDA0003342673100000036
c) if the fan is operated under high wind speed condition, the information data set
Figure BDA0003342673100000037
Medium wind speed viGreater than rated wind speed v of fanratedAnd is less than cut-out wind speed vcut_offActive power PiGreater than active power threshold PthresAnd is less than the rated active power and active power threshold value P of the fanthresThe difference value is determined as an abnormal data point, and the criterion formula is listed as follows:
Figure BDA0003342673100000038
further, in the step 3), the data set is operated according to the standard of the fan
Figure BDA0003342673100000039
Established fan power curve model P5logistic(v, θ) is specifically described as follows:
Figure BDA00033426731000000310
further, in the step 5), a fan power curve model P is determined5logisticWhen five parameters (A, B, C, D and G) in (v and theta) are identified, converting the identification optimization problem of the parameters theta (A, B, C, D and G) into a nonlinear programming problem, wherein the target function is an error square function, and the constraint condition is the interval constraint of the five parameters (A, B, C, D and G) according to expert experience and fan delivery specification parameters; the formula of the objective function is as follows:
Figure BDA0003342673100000041
wherein, Pe 5logistic(vi) For active power estimated using a fan power curve model, Pa(vi) Is the actual active power.
Further, in the step 5), a differential evolution method is used for a fan power curve model P5logisticThe specific steps of parameter identification of five parameters (A, B, C, D, G) in (v, theta) are as follows:
a) standard operation data set of fan
Figure BDA0003342673100000042
Medium wind speed
Figure BDA0003342673100000043
Each data point of (a) is recorded as an individual viWill total nnormThe individuals form an initial population, namely a data point population with the iteration number of 0, and the initial population is recorded as
Figure BDA0003342673100000044
Wherein the content of the first and second substances,
Figure BDA0003342673100000045
for one individual in the initial population, each individual has five parameters of a five-parameter logistic model as its chromosome vector, denoted as
Figure BDA0003342673100000046
b) According to population size nnormSetting and selecting maximum iteration number NiterInitial scaling factor F0Initial cross probability
Figure BDA0003342673100000047
And the probability of cross probability adjustment τ.
c) When the number of iterations is N, the nth generation mutation process is described as: in population PopNIn extracting two parameter vectors at random, i.e. chromosome vectors
Figure BDA0003342673100000048
And
Figure BDA0003342673100000049
simultaneously randomly extracting a parameter vector as a reference chromosome vector
Figure BDA00033426731000000410
And k, i and j are different integers; making the N generation population individual vary by a scaling factor FNControlling the degree of variation, and differentially mutating to obtain the (N +1) th generation donor vector
Figure BDA00033426731000000411
Record as
Figure BDA00033426731000000412
Wherein the scaling factor FNThe method is a self-adaptive fine-tuning numerical value, the scaling factor is adjusted according to the iteration number N, the variation degree is controlled, the scaling factor is large in the initial iteration process, and the diversity of population individuals is reserved; the scaling factor is smaller when approaching the optimal solution, the global convergence is improved, and the scaling factor FNIs recorded as a specific adjustment strategy
Figure BDA00033426731000000413
d) In order to improve the diversity of population individuals, a binomial crossover operator is adopted to carry out crossover operation to generate an (N +1) th generation of test vectors
Figure BDA00033426731000000414
Figure BDA00033426731000000415
Each element in the vector
Figure BDA00033426731000000416
Can be obtained from the following formula
Figure BDA00033426731000000417
Wherein rnbriIs [1, D ]]D is the dimension of the chromosome vector, randjIs [0, 1 ]]Is determined by the random number of (1),
Figure BDA00033426731000000418
the cross probability is an adaptive random value, keeps the diversity of the iterative population, and is
Figure BDA0003342673100000051
Is recorded as
Figure BDA0003342673100000052
e) After the cross operation, the population of the (N +1) th generation is selected, and the value of an objective function to be optimized is calculated, and the vector is tested
Figure BDA0003342673100000053
And a target vector
Figure BDA0003342673100000054
The first generation of the population is formed by preferential selection
Figure BDA0003342673100000055
Figure BDA0003342673100000056
f) Repeating the steps c) to e) until the iteration number N reaches the maximum iteration number Niter(ii) a Finally obtaining the parameters of the optimal five-fan power curve fitting model
Figure BDA0003342673100000057
Further, the optimal fan power curve model obtained in the step 6) according to the optimal five parameter sets
Figure BDA0003342673100000058
The method can be used as a reference for subsequent fan state monitoring and fault detection; the real-time wind turbine generator power fitting curve is drawn by using the real-time fan SCADA data, so that the operation and maintenance personnel of the fan are helped to judge the running state of the fan; various fault detection index values can be calculated according to the fitted power curve, and the real-time monitoring of the fault indexes can be used for prejudging the faults of the fan in advance.
Compared with the prior art, the invention has the following innovative advantages and remarkable effects:
1) aiming at the accuracy requirement of the wind turbine generator power curve fitting, a five-parameter logistic model which is in accordance with the standard fan power curve characteristic is selected, the model parameter physical significance is given to the model parameter, the initial value of the model parameter is selected, the characteristic of the standard fan power curve and the actual operation data characteristic are considered, and the accuracy of the wind turbine generator power curve fitting is improved.
2) Aiming at the rapidity requirement of the wind turbine generator power curve fitting, a differential evolution method for adaptively controlling parameter change is selected to identify parameters of a five-parameter logistic model, so that the overall convergence is improved while the diversity of optimization parameters is maintained, the parameter optimization efficiency is improved, and the rapidity of the wind turbine generator power curve fitting is ensured.
Drawings
FIG. 1 is a flow chart of a method of fitting a fan power curve based on a parametric model of differential evolution in accordance with the present invention;
FIG. 2 is a raw data set wind speed-power scatter plot of an embodiment of the present invention;
FIG. 3 is a graph showing the results of data washing in step 2 according to the embodiment of the present invention;
FIG. 4 is a schematic diagram of a five-parameter logistic power curve of a wind turbine obtained by fitting in steps 3 and 4 according to an embodiment of the present invention.
Detailed Description
In order to make the aforementioned objects, features and advantages of the present invention comprehensible, embodiments accompanied with figures are described in detail below.
In the following description, numerous specific details are set forth in order to provide a thorough understanding of the present invention, but the present invention may be practiced in other ways than those specifically described and will be readily apparent to those of ordinary skill in the art without departing from the spirit of the present invention, and therefore the present invention is not limited to the specific embodiments disclosed below.
Examples
In the embodiment, the modeling of the parameter power curve fitting model is performed based on the operation data of one month collected by the SCADA system of a certain wind driven generator of a certain wind power plant within the period from 5 months to 6 months in 2018. The data sampling time interval of the SCADA system of the fan is 5min, the data information period is 27 days, and the time range is 2018/5/1610: 55: 00-2018/6/1115: 05: 00. The specific data variables and related data information included in the data set are shown in tables 1 and 2:
TABLE 1 partial data of SCADA system data set of certain wind turbine of certain wind power plant
Data sequence number Time of acquisition Actual value of pitch angle Average wind speed of 10min Average active power 10min
1 2018/5/16 10:55:00 -0.029 7.235 522
2 2018/5/16 11:00:00 -0.029 7.546 561
… … … … …
3178 2018/5/27 12:05:00 8.69 11.636 1483
3179 2018/5/27 12:10:00 3.29 11.452 1471
3180 2018/5/27 12:15:00 9.98 11.5 1509
… … … … …
TABLE 2 variable information of SCADA system data set of certain wind turbine of certain wind power plant
Variable names Meaning of variables Variable unit
Time of acquisition Data acquisition real-time point Year/month/day hour, minute/second
Actual value of Pitch Angle (β)i) Current real-time pitch angle of fan degree
Average wind speed 10min (v)i) At presentAverage wind speed of fan for 10min m/s
Average active power 10min (p)i) Average active power of current fan for 10min kW
In this embodiment, the implementation data set of the default power curve data obtaining method is the operation data of the certain fan within 27 days, and the method result is the obtained wind turbine generator power curve fitting model
Figure BDA0003342673100000061
The detailed implementation steps are as follows:
1) according to the total n pieces of fan operation data information in the corresponding demand period acquired by a supervisory control and data acquisition (SCADA) system of the fan to be evaluated, the average wind speed { v ] in the operation data information is calculatediMean active power { P }iJ and pitch angle [ beta ]iThe information constitutes a power curve information data set to be preprocessed, and the information data set is recorded as
Figure BDA0003342673100000062
Wherein i is 1,2,3, …, n. According to the data information provided in tables 1 and 2, all necessary information (collection time, pitch angle, average wind speed 10min, and average active power 10min) required in this step is included, and as shown in fig. 2, a scatter diagram of the raw data of the wind speed-power curve of the wind turbine in this step is shown.
2) According to the normal value ranges of the corresponding active power and the pitch angle under different wind speed conditions in the running process of the wind turbine generator, the information data set in the step 1) is compared with the information data set in the step 1)
Figure BDA0003342673100000071
Cleaning the original data set, removing obvious abnormal points and obtaining the characteristics of the residual abnormal points which are not in accordance with the abnormal pointsN of (A) to (B)normRunning data is recorded and recorded as a standard running data set of the fan
Figure BDA0003342673100000072
Wherein i is 1,2,3, …, nnorm(ii) a The specific criteria for data cleaning based on wind speed information are as follows:
a) when the fan is started, if the information data set is available
Figure BDA0003342673100000073
Medium wind speed viGreater than cut-in wind speed vcut_inActive power PiBelow active power threshold PthresOr the value is negative, the abnormal data point is judged, and the criterion formula is listed as follows:
Figure BDA0003342673100000074
b) if the fan is operated at low wind speed, the information data set
Figure BDA0003342673100000075
Medium wind speed viGreater than cut-in wind speed vcut_inAnd is less than the rated wind speed v of the fanratedAngle of pitch betaiGreater than pitch angle threshold betathresIf the data points are judged to be abnormal data points, the criterion formula is listed as follows:
Figure BDA0003342673100000076
c) if the fan is operated under high wind speed condition, the information data set
Figure BDA0003342673100000077
Medium wind speed viGreater than rated wind speed v of fanratedAnd is less than cut-out wind speed vcut_offActive power PiGreater than active power threshold PthresAnd is less than the rated active power and active power threshold value P of the fanthresIs determined to be abnormalData points, the criterion formula, are listed below:
Figure BDA0003342673100000078
in this embodiment, PratedIs 1560kW, vcut_inIs 3m/s, vratedIs 12m/s, vcut_offIs 28m/s, PthresIs 10kW, betathresIs 4 deg.. Fig. 3 shows the result of data cleaning in this embodiment after this step, and it can be seen from the figure that the data cleaning effect is obvious, and most of the outliers are detected and subjected to the culling operation.
3) Cleaning the data in the step 2) to obtain a fan standard operation data set
Figure BDA0003342673100000079
Performing a five-parameter logistic parametric power curve fitting model P5logistic(v, θ) modeling:
Figure BDA00033426731000000710
wherein the wind speed
Figure BDA00033426731000000711
Being independent variable of the power curve model, active power
Figure BDA00033426731000000712
And theta is a set of five parameters (A, B, C, D, G) of the five-parameter logistic model, wherein A represents the lowest power of the fan, B represents the efficiency of the fan for capturing wind energy to the maximum extent, C represents the median of the maximum generated power of the fan, D represents the maximum power of the fan, and G represents the symmetry of the power curve.
4) For the fan power curve model P in the step 3)5logisticWhen five parameters (A, B, C, D, G) in (v, theta) are identified, the identification optimization problem of the parameters theta (A, B, C, D, G) is converted into a nonlinear programming problem, and the purpose of the nonlinear programming problem is to solve the problemThe standard function is an error square function, the constraint condition is interval constraint of five parameters (A, B, C, D and G) according to expert experience and factory specification parameters of the fan, and the formula of the target function is as follows:
Figure BDA0003342673100000081
wherein, Pe 5logistic(vi) For active power estimated using a fan power curve model, Pa(vi) Is the actual active power.
5) Initial values theta of five parameter sets theta of the five parameter logistic model in the step 3) (A, B, C, D, G)0When setting is carried out, setting initial values of parameter sets by combining the meaning of each parameter of the five-parameter logistic model and the physical meaning of a fan power curve, wherein:
setting the initial value of A as active power in fan standard operation data set
Figure BDA0003342673100000082
For the present embodiment, the parameter A0＝0；
The initial value of B is set as the slope of the connection line between the two data points with the maximum euclidean distance in the fan standard operation data set, and for this embodiment, the parameters
Figure BDA0003342673100000083
Setting the initial value of C as active power in fan standard operation data set
Figure BDA0003342673100000084
Mean of maximum and minimum, for the present example, parameter
Figure BDA0003342673100000085
Setting the initial value of D as active power in fan standard operation data set
Figure BDA0003342673100000086
Maximum value of (D) for the present embodiment0＝1560；
The initial value of G is set to 1, which means that the initial power curve is symmetrical, i.e. for this embodiment, the parameter G0＝1。
6) Initial values theta of five parameter sets theta based on the nonlinear programming problem in step 4) and the five-parameter logistic model in step 5)0Using a differential evolution method to carry out the wind turbine power curve model P in the step 3)5logisticAnd (v, theta) performing parameter identification on five parameters (A, B, C, D and G), wherein the specific implementation steps are as follows:
a) standard operation data set of fan
Figure BDA0003342673100000087
Medium wind speed
Figure BDA0003342673100000088
Each data point of (a) is recorded as an individual viWill total nnormThe individuals form an initial population, namely a data point population with the iteration number of 0, and the initial population is recorded as
Figure BDA0003342673100000089
Wherein the content of the first and second substances,
Figure BDA00033426731000000810
for one individual in the initial population, each individual has five parameters of a five-parameter logistic model as its chromosome vector, denoted as
Figure BDA00033426731000000811
In this embodiment, the initial chromosome vector can be written as
Figure BDA00033426731000000812
b) According to population size nnormSetting and selecting maximum iteration number NiterInitial scaling factor F0Initial cross probability
Figure BDA00033426731000000813
And the probability tau of cross probability adjustment, in this embodiment, the maximum number of iterations N is selectediter100, initial scaling factor F0Initial crossover probability of 0.6
Figure BDA0003342673100000091
And the probability τ of the cross probability adjustment is 0.2.
c) When the number of iterations is N, the nth generation mutation process is described as: in population PopNIn extracting two parameter vectors at random, i.e. chromosome vectors
Figure BDA0003342673100000092
And
Figure BDA0003342673100000093
simultaneously randomly extracting a parameter vector as a reference chromosome vector
Figure BDA0003342673100000094
And k, i and j are different integers. Making the N generation population individual vary by a scaling factor FNControlling the degree of variation, and differentially mutating to obtain the (N +1) th generation donor vector
Figure BDA0003342673100000095
Record as
Figure BDA0003342673100000096
Wherein the scaling factor FNThe method is a self-adaptive fine-tuning numerical value, the scaling factor is adjusted according to the iteration number N, the variation degree is controlled, the scaling factor is large in the initial iteration process, and the diversity of population individuals is reserved; the scaling factor is smaller when approaching the optimal solution, the global convergence is improved, and the scaling factor FNIs recorded as a specific adjustment strategy
Figure BDA0003342673100000097
d) In order to improve the diversity of population individuals, a binomial crossover operator is adopted to carry out crossover operation to generate an (N +1) th generation of test vectors
Figure BDA0003342673100000098
Figure BDA0003342673100000099
Each element in the vector
Figure BDA00033426731000000910
Can be obtained from the following formula
Figure BDA00033426731000000911
Wherein rnbriIs [1, D ]]D is the dimension of the chromosome vector, randjIs [0, 1 ]]Is determined by the random number of (1),
Figure BDA00033426731000000912
the cross probability is an adaptive random value, keeps the diversity of the iterative population, and is
Figure BDA00033426731000000913
Is recorded as
Figure BDA00033426731000000914
e) After the cross operation, the population of the (N +1) th generation is selected, and the value of an objective function to be optimized is calculated, and the vector is tested
Figure BDA00033426731000000915
And a target vector
Figure BDA00033426731000000916
The first generation of the population is formed by preferential selection
Figure BDA00033426731000000917
Figure BDA00033426731000000918
f) Repeating the steps c) to e) until the iteration number N reaches the maximum iteration number Niter. Finally obtaining the parameters of the optimal five-fan power curve fitting model
Figure BDA00033426731000000919
For this embodiment, the parameters of the optimal five fan power curve fitting model
Figure BDA00033426731000000920
Obtaining the final optimal wind turbine generator power curve fitting model
Figure BDA00033426731000000921
Writing:
Figure BDA00033426731000000922
fig. 4 shows a power curve obtained by fitting the data set corresponding to this embodiment after this step, where the power curve obtained by fitting is a thick solid line and the data points are labeled ".
The invention provides a method for fitting a fan power curve by a parameter model based on differential evolution, which mainly comprises the steps of cleaning an initial data set, establishing a five-parameter logistic power curve fitting model aiming at wind speed and power data, performing parameter identification on the parameter model by using the differential evolution method and the like. Fig. 1 is a specific flow of implementation and application of a method for fitting a fan power curve based on a parametric model of differential evolution. In the whole embodiment, according to the flow shown in fig. 1, the fan power curve data set is preprocessed and finally fitted to obtain a parametric model power curve. Fig. 2-4 show results of each link of fan power curve fitting modeling by using the method for fitting a fan power curve based on a parameter model of differential evolution, and parameters of the parameter model are optimized by establishing the parameter model and using the differential evolution method, so that the characteristics of a standard power curve model of a wind turbine generator are considered, and parameter identification based on operating data is considered. Therefore, the method standardizes the flow of fan power curve fitting and can obtain a more accurate and real fan power curve. The method has guiding significance for the follow-up research on the running state, fault detection and early warning of the wind turbine generator.
The foregoing is only a preferred embodiment of the present invention, and although the present invention has been disclosed in the preferred embodiments, it is not intended to limit the present invention. Those skilled in the art can make numerous possible variations and modifications to the present teachings, or modify equivalent embodiments to equivalent variations, without departing from the scope of the present teachings, using the methods and techniques disclosed above. Therefore, any simple modification, equivalent change and modification made to the above embodiments according to the technical essence of the present invention are still within the scope of the protection of the technical solution of the present invention, unless the contents of the technical solution of the present invention are departed.

Claims (6)

1. A method for fitting a fan power curve based on a differential evolution parameter model is characterized by comprising the following steps:
1) according to the total n pieces of fan operation data information in the corresponding demand period acquired by a supervisory control and data acquisition (SCADA) system of the fan to be evaluated, the wind speed { v ] in the operation data information is converted into the wind speed { viActive power { P }iJ and pitch angle [ beta ]iThe information constitutes a power curve information data set to be preprocessed, which is recorded as
Figure FDA0003342673090000011
Wherein i is 1,2,3, …, n;
2) according to the normal value ranges of the corresponding active power and the pitch angle under different wind speed conditions in the running process of the wind turbine generator, the information data set in the step 1) is compared with the information data set in the step 1)
Figure FDA0003342673090000012
Cleaning the original data set, removing obvious abnormal points to obtain n which is not in accordance with the abnormal point characteristicsnormRunning data is recorded and recorded as a standard running data set of the fan
Figure FDA0003342673090000013
Wherein i is 1,2,3, …, nnorm；
3) Collecting the standard operation data set of the fan cleaned in the step 2)
Figure FDA0003342673090000014
Medium wind speed
Figure FDA0003342673090000015
And active power
Figure FDA0003342673090000016
Establishing a five-parameter logistic model, and recording a power curve model of the fan as P5logistic(v, θ) wherein the wind speed
Figure FDA0003342673090000017
Being independent variable of the power curve model, active power
Figure FDA0003342673090000018
The dependent variable of the power curve model is theta, a set (A, B, C, D, G) of five parameters of the five-parameter logistic model is theta, wherein A represents the lowest power of the fan, B represents the efficiency of the fan for capturing wind energy to the maximum extent, C represents the median of the maximum generating power of the fan, D represents the maximum power of the fan, and G represents the symmetry of the power curve;
4) for the five-parameter logistic model in the step 3)Is equal to the initial value of (a, B, C, D, G)0Setting, namely setting an initial value of a parameter set by combining the meaning of each parameter of the five-parameter logistic model and the physical meaning of a fan power curve, wherein:
setting the initial value of A as active power in fan standard operation data set
Figure FDA0003342673090000019
Minimum value of (d);
setting the initial value of B as the slope of a connecting line between two data points with the maximum Euclidean distance in the fan standard operation data set;
setting the initial value of C as active power in fan standard operation data set
Figure FDA00033426730900000110
The mean of the maximum and minimum values;
setting the initial value of D as active power in fan standard operation data set
Figure FDA00033426730900000111
Maximum value of (d);
the initial value of G is set to 1, which represents that the initial power curve is symmetrical;
5) initial values theta of five parameter sets theta of the five-parameter logistic model according to the step 4)0Using a differential evolution method to carry out the wind turbine power curve model P in the step 3)5logisticAnd (v, theta) performing parameter identification on five parameters (A, B, C, D and G) and recording a standard operation data set of the fan
Figure FDA00033426730900000112
Each data point in the data points is an individual and the whole is a population, five parameter sets corresponding to each individual are used as chromosome vectors, donor vectors are obtained by randomly extracting three original chromosome vectors and carrying out differential calculation, a binomial crossover operator is adopted to carry out crossover operation to generate test vectors, and the two original chromosome vectors and the dye closest to the optimal chromosome vector in the test vectors are selectedThe color body vector is used as an individual of a new generation of population, and the optimal five parameter sets theta can be obtained after repeated iteration is carried out until the set times are limitedbest＝(Abest，Bbest，Cbest，Dbest，Gbest)；
6) Obtaining an optimal fan power curve model according to the optimal five parameter sets obtained in the step 5)
Figure FDA0003342673090000021
2. The method for fitting the fan power curve based on the parameter model of differential evolution according to claim 1, wherein in the step 2), different wind speed ranges are divided, and abnormal point detection is performed on the fan power curve according to abnormal characteristics of the fan operation state; the specific criteria for data cleaning based on wind speed information are as follows:
a) when the fan is started, if the information data set is available
Figure FDA0003342673090000022
The wind speed in the wind power system is greater than the cut-in wind speed, the active power is lower than the active power threshold value or the value is a negative number, and the wind power system is judged to be an abnormal data point;
b) the fans operating at low wind speed, i.e. information data sets
Figure FDA0003342673090000023
The wind speed in the process is greater than the cut-in wind speed and less than the rated wind speed of the fan, if the pitch angle is greater than the pitch angle threshold value, an abnormal data point is determined;
c) the fans operating at high wind speeds, i.e. information data sets
Figure FDA0003342673090000024
If the active power is greater than the active power threshold value and less than the difference value between the rated active power of the fan and the active power threshold value, judging as an abnormal data point。
3. The method for fitting the fan power curve based on the parametric model of differential evolution as claimed in claim 1, wherein in the step 3), the fan power curve is fitted according to a fan standard operation data set
Figure FDA0003342673090000025
Established fan power curve model P5logistic(v, θ) is specifically described as follows:
Figure FDA0003342673090000026
4. the method for fitting the fan power curve based on the parameter model of differential evolution as claimed in claim 1, wherein in the step 5), the fan power curve model P is subjected to5logisticWhen five parameters (A, B, C, D and G) in (v and theta) are identified, converting the identification optimization problem of the parameters theta (A, B, C, D and G) into a nonlinear programming problem, wherein the target function is an error square function, and the constraint condition is the interval constraint of the five parameters (A, B, C, D and G) according to expert experience and fan delivery specification parameters; the formula of the objective function is as follows:
Figure FDA0003342673090000027
wherein, Pe 5logistic(vi) For active power estimated using a fan power curve model, Pa(vi) Is the actual active power.
5. The method for fitting the fan power curve based on the parameter model of differential evolution as claimed in claim 1, wherein in the step 5), the fan power curve model P is subjected to a differential evolution method5logisticFive of (v, theta)The specific steps of parameter identification for numbers (A, B, C, D, G) are as follows:
a) standard operation data set of fan
Figure FDA0003342673090000028
Medium wind speed
Figure FDA0003342673090000029
Each data point of (a) is recorded as an individual viWill total nnormThe individuals form an initial population, namely a data point population with the iteration number of 0, and the initial population is recorded as
Figure FDA0003342673090000031
Wherein the content of the first and second substances,
Figure FDA0003342673090000032
for one individual in the initial population, each individual has five parameters of a five-parameter logistic model as its chromosome vector, denoted as
Figure FDA0003342673090000033
b) According to population size nnormSetting and selecting maximum iteration number NiterInitial scaling factor F0Initial cross probability
Figure FDA0003342673090000034
And the probability of cross probability adjustment τ.
c) When the number of iterations is N, the nth generation mutation process is described as: in population PopNIn extracting two parameter vectors at random, i.e. chromosome vectors
Figure FDA0003342673090000035
And
Figure FDA0003342673090000036
simultaneously randomly extracting a parameter vector as a reference chromosome vector
Figure FDA0003342673090000037
And k, i and j are different integers; making the N generation population individual vary by a scaling factor FNControlling the degree of variation, and differentially mutating to obtain the (N +1) th generation donor vector
Figure FDA0003342673090000038
Record as
Figure FDA0003342673090000039
Wherein the scaling factor FNIs a self-adaptive fine-tuning numerical value, and the scaling factor is adjusted according to the iteration number N to control the variation degree, and the scaling factor FNIs recorded as a specific adjustment strategy
Figure FDA00033426730900000310
d) In order to improve the diversity of population individuals, a binomial crossover operator is adopted to carry out crossover operation to generate an (N +1) th generation of test vectors
Figure FDA00033426730900000311
Figure FDA00033426730900000312
Each element in the vector
Figure FDA00033426730900000313
Can be obtained from the following formula
Figure FDA00033426730900000314
Wherein the content of the first and second substances,rnbriis [1, D ]]D is the dimension of the chromosome vector, randjIs [0, 1 ]]Is determined by the random number of (1),
Figure FDA00033426730900000315
the cross probability is an adaptive random value, keeps the diversity of the iterative population, and is
Figure FDA00033426730900000316
Is recorded as
Figure FDA00033426730900000317
e) After the cross operation, the population of the (N +1) th generation is selected, and the value of an objective function to be optimized is calculated, and the vector is tested
Figure FDA00033426730900000318
And a target vector
Figure FDA00033426730900000319
The first generation of the population is formed by preferential selection
Figure FDA00033426730900000320
Figure FDA00033426730900000321
f) Repeating the steps c) to e) until the iteration number N reaches the maximum iteration number Niter(ii) a Finally obtaining the parameters of the optimal five-fan power curve fitting model
Figure FDA0003342673090000041
6. According to the rightThe method for fitting the fan power curve based on the parameter model of differential evolution according to claim 1, wherein the optimal fan power curve model obtained in the step 6) is obtained according to the optimal five parameter sets
Figure FDA0003342673090000042
The method can be used as a reference for subsequent fan state monitoring and fault detection; the real-time wind turbine generator power fitting curve is drawn by using the real-time fan SCADA data, so that the operation and maintenance personnel of the fan are helped to judge the running state of the fan; various fault detection index values can be calculated according to the fitted power curve, and the real-time monitoring of the fault indexes can be used for prejudging the faults of the fan in advance.
CN202111313404.2A 2021-11-08 2021-11-08 Method for fitting fan power curve based on parameter model of differential evolution Pending CN114065618A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202111313404.2A CN114065618A (en) 2021-11-08 2021-11-08 Method for fitting fan power curve based on parameter model of differential evolution

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202111313404.2A CN114065618A (en) 2021-11-08 2021-11-08 Method for fitting fan power curve based on parameter model of differential evolution

Publications (1)

Publication Number Publication Date
CN114065618A true CN114065618A (en) 2022-02-18

Family

ID=80274444

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202111313404.2A Pending CN114065618A (en) 2021-11-08 2021-11-08 Method for fitting fan power curve based on parameter model of differential evolution

Country Status (1)

Country Link
CN (1) CN114065618A (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
EP4254083A1 (en) * 2022-03-30 2023-10-04 Siemens Aktiengesellschaft Method for parameterizing a monitoring system, parameter device, and a monitoring system

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
EP4254083A1 (en) * 2022-03-30 2023-10-04 Siemens Aktiengesellschaft Method for parameterizing a monitoring system, parameter device, and a monitoring system

Similar Documents

Publication Publication Date Title
CN110685857B (en) Mountain wind turbine generator behavior prediction model based on ensemble learning
CN111539553B (en) Wind turbine generator fault early warning method based on SVR algorithm and off-peak degree
CN111275367B (en) Regional comprehensive energy system energy efficiency state evaluation method
CN110362045B (en) Marine doubly-fed wind turbine generator fault discrimination method considering marine meteorological factors
CN107944175B (en) Method for acquiring real power curve of fan by considering wind turbulence intensity
CN111260503B (en) Wind turbine generator power curve outlier detection method based on cluster center optimization
CN114215706B (en) Early warning method and device for cracking faults of wind turbine generator blades
CN106570790B (en) Wind power plant output data restoration method considering wind speed data segmentation characteristics
CN110009135B (en) Wind power prediction method based on width learning
CN111522808B (en) Abnormal operation data processing method for wind turbine generator
CN111881420A (en) Wind turbine generator set operation data interpolation method
CN107728059A (en) A kind of pitch-controlled system state evaluating method
CN110991701A (en) Wind power plant fan wind speed prediction method and system based on data fusion
CN111814848B (en) Self-adaptive early warning strategy design method for temperature faults of wind turbine generator
CN114065618A (en) Method for fitting fan power curve based on parameter model of differential evolution
CN116771610A (en) Method for adjusting fault evaluation value of variable pitch system of wind turbine
CN112598148A (en) Fan variable pitch motor temperature fault early warning method based on collaborative expression and LightGBM algorithm
CN114595762A (en) Photovoltaic power station abnormal data sequence extraction method
CN114281846B (en) New energy power generation prediction method based on machine learning
CN114912807A (en) Method and system for evaluating generated energy improving effect of technically improved wind turbine generator
CN115450864A (en) Wind turbine blade icing diagnosis method based on synthesis of minority samples
Souza et al. Evaluation of data based normal behavior models for fault detection in wind turbines
Bao et al. Iterative modeling of wind turbine power curve based on least‐square B‐spline approximation
CN115271423A (en) Wind generating set running state evaluation method based on hybrid similarity
CN110334951B (en) Intelligent evaluation method and system for high-temperature capacity reduction state of wind turbine generator

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination