CN113312836A - Short-term wind speed prediction method - Google Patents
Short-term wind speed prediction method Download PDFInfo
- Publication number
- CN113312836A CN113312836A CN202110506041.8A CN202110506041A CN113312836A CN 113312836 A CN113312836 A CN 113312836A CN 202110506041 A CN202110506041 A CN 202110506041A CN 113312836 A CN113312836 A CN 113312836A
- Authority
- CN
- China
- Prior art keywords
- wind speed
- equivalent
- value
- wind
- fan
- 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
Links
- 238000000034 method Methods 0.000 title claims abstract description 46
- 241000039077 Copula Species 0.000 claims abstract description 53
- 238000009826 distribution Methods 0.000 claims abstract description 45
- 238000012549 training Methods 0.000 claims abstract description 27
- 230000002068 genetic effect Effects 0.000 claims abstract description 21
- 238000012360 testing method Methods 0.000 claims description 18
- 238000005315 distribution function Methods 0.000 claims description 6
- 230000001419 dependent effect Effects 0.000 claims description 4
- 150000001875 compounds Chemical class 0.000 claims description 3
- 230000001186 cumulative effect Effects 0.000 claims description 3
- 238000012545 processing Methods 0.000 claims description 2
- 230000007774 longterm Effects 0.000 abstract description 2
- 230000006870 function Effects 0.000 description 46
- 238000010248 power generation Methods 0.000 description 3
- 230000006403 short-term memory Effects 0.000 description 3
- 238000013528 artificial neural network Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 230000007787 long-term memory Effects 0.000 description 2
- 238000010801 machine learning Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 230000015654 memory Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000005855 radiation Effects 0.000 description 1
- 230000000306 recurrent effect Effects 0.000 description 1
- 230000001105 regulatory effect Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
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
- G01—MEASURING; TESTING
- G01W—METEOROLOGY
- G01W1/00—Meteorology
- G01W1/10—Devices for predicting weather conditions
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
- G06F18/232—Non-hierarchical techniques
- G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
- G06F18/23213—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions with fixed number of clusters, e.g. K-means clustering
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/044—Recurrent networks, e.g. Hopfield networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/12—Computing arrangements based on biological models using genetic models
- G06N3/126—Evolutionary algorithms, e.g. genetic algorithms or genetic programming
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/06—Energy or water supply
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Business, Economics & Management (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Physics & Mathematics (AREA)
- Health & Medical Sciences (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Biophysics (AREA)
- Economics (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Strategic Management (AREA)
- Software Systems (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Human Resources & Organizations (AREA)
- Evolutionary Biology (AREA)
- General Health & Medical Sciences (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Molecular Biology (AREA)
- Computational Linguistics (AREA)
- General Business, Economics & Management (AREA)
- Tourism & Hospitality (AREA)
- Environmental & Geological Engineering (AREA)
- Marketing (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Biomedical Technology (AREA)
- Medical Informatics (AREA)
- Operations Research (AREA)
- Development Economics (AREA)
- Ecology (AREA)
- Game Theory and Decision Science (AREA)
- Biodiversity & Conservation Biology (AREA)
- Probability & Statistics with Applications (AREA)
- Entrepreneurship & Innovation (AREA)
- Atmospheric Sciences (AREA)
- Environmental Sciences (AREA)
Abstract
The invention discloses a short-term wind speed prediction method, which comprises the following steps: carrying out scene clustering on the wind turbine generator by using a K-medoids clustering algorithm to obtain a plurality of equivalent fans; constructing an LSTM model, and training a plurality of equivalent fans through the LSTM model to obtain an initial predicted value of the wind speed; establishing a functional relation among a plurality of equivalent fans by using a hybrid Copula function, and solving parameters in the hybrid Copula function by using a genetic algorithm; solving a residual value of the wind speed through the wind speed edge distribution values of the equivalent fans and a mixed Copula function; obtaining an accurate predicted value of the wind speed according to the initial predicted value and the residual value of the wind speed; the fans in the wind power plant are classified by using a K-medoids clustering algorithm to obtain equivalent fans, the randomness of the wind speed is reduced, the equivalent fans are modeled by using a mixed Copula function and then combined with a long-term network and a short-term network, and the precision of wind speed prediction is improved.
Description
Technical Field
The invention belongs to the technical field of wind power plant power generation, and particularly relates to a short-term wind speed prediction method.
Background
The wind energy resource storage is huge, and no pollution is caused to the environment, so that the wind power generation is greatly advocated. Accurate wind speed prediction is very important to the scheduling and planning of a power system and is a very critical link in wind power generation.
The traditional wind speed prediction mostly adopts one representative wind motor data to predict the whole wind power plant, and at present, a method combining machine learning is mostly adopted. Due to the influence of the position of the fan and the wake flow, the difference of the operation data of each wind turbine in the wind power plant is large, and the accuracy is not high because only the group data of one wind turbine is used for prediction.
Disclosure of Invention
In order to solve the problems, the invention provides a short-term wind speed prediction method which can obtain a relatively accurate wind speed prediction value.
A short term wind speed prediction method comprising the steps of:
carrying out scene clustering on the wind turbine generator by using a K-medoids clustering algorithm to obtain a plurality of equivalent fans;
constructing an LSTM model, and training a plurality of equivalent fans through the LSTM model to obtain an initial predicted value of the wind speed;
establishing a functional relation among a plurality of equivalent fans by using a hybrid Copula function, and solving parameters in the hybrid Copula function by using a genetic algorithm;
solving a residual value of the wind speed through the wind speed edge distribution values of the equivalent fans and a mixed Copula function;
and obtaining an accurate predicted value of the wind speed through the initial predicted value and the residual value of the wind speed.
Further, carrying out scene clustering on the wind turbine generator by using a K-medoids clustering algorithm to obtain a plurality of equivalent fans, which specifically comprises the following steps:
acquiring historical wind speed data of each wind turbine generator from a wind power plant;
acquiring historical data of wind speed, power, generator rotating speed and blade angle of a region from a meteorological station, analyzing and processing the data, and determining model input and output variables;
carrying out scene clustering on the wind turbine generator by adopting K-medoids clustering, respectively taking the wind speed and the power of the wind turbine generator as horizontal and vertical coordinate values, constructing a rectangular coordinate system, and establishing a criterion function Fs, wherein the method specifically comprises the following steps:
in the formula, j represents the number of the fan variables in each class, n represents the total number of the fan variables in each class, and xi、yiAs a coordinate of the center point, xj、yjCoordinates in class except the central point;
solving a criterion function Fs, and determining k wind turbine generator clusters;
and determining each type of wind turbine generator as an equivalent fan.
Further, solving a criterion function Fs to determine k wind turbine clusters, specifically as follows:
s11, randomly selecting k points from the n samples as central points medoids;
s12, distributing the rest points to the current best cluster according to the principle of being nearest to the center;
s13, sequentially calculating the value of the criterion function Fs when the points are medoids for all other points except the medoids in each class, traversing all possibilities, and selecting the point corresponding to the minimum criterion function Fs as a new medoid;
s14, repeating S12 and S13 until all medoids are not changed;
and S15, finally determining k wind turbine generator clusters through continuous iterative operation.
Further, the method for determining the equivalent fan specifically comprises the following steps:
the power of the equivalent fan is the sum of the powers of the clustered wind turbine generators;
the wind speed of the equivalent fan is the average value of the wind speeds of the clustered wind generation sets;
the generator rotating speed of the equivalent fan is the average value of the generator rotating speeds of the clustered wind turbine generators;
and the blade angle of the equivalent fan is the average value of the blade angles of the clustered wind turbine generators.
Further, an LSTM model is constructed, a plurality of equivalent fans are trained through the LSTM model, and an initial predicted value of the wind speed is obtained, wherein the initial predicted value is as follows:
and (3) the calculated wind speed, power, generator rotating speed and blade angle of the equivalent fan are arranged into a new data set, the new data set is divided into a training set and a testing set, and the training set is input into an LSTM model for training to obtain a model parameter and a preliminary predicted value of the wind speed.
Further, the data ratio of the training set to the test set was 8: 2.
Further, a function relation among a plurality of equivalent fans is established by using a hybrid Copula function, and parameters in the hybrid Copula function are solved by using a genetic algorithm, which is specifically as follows:
let x be1,x2,…xnN random wind speed variables, the edge distribution of the n random wind speed variables is respectively f1(x1),f2(x2),…,fn(xn) The combined distribution is f (x)1,x2,…xn) Let there be a function C, such that:
f(x1,x2,…xn)=C(f1(x1),f2(x2),…,fn(xn)) (2)
let u1=f1(x1),u2=f2(x2),…,un=fn(xn) Wherein u is1,u2…unFor the edge distribution of wind speed, a mathematical expression of the Copula function is obtained: c (u)1,u2,…un);
Establishing a mixed Copula function, wherein the concrete form C is as follows:
C(u,v;θ)=ω1Cg(u,v;θ1)+ω2Cc(u,v;θ2)+ω3Cf(u,v;θ3) (3)
ω1+ω2+ω3=1,ω1,ω2,ω3≥0 (4)
wherein u is the wind speed edge distribution value of the first equivalent fan, v is the wind speed edge distribution value of the second equivalent fan, omega1,ω2,ω3As a weight parameter of the model, θ1,θ2,θ3Is a dependent parameter of the model;
solving parameter omega of mixed Copula in the formula (3) by adopting genetic algorithm1、ω2、ω3、θ1、θ2、θ3。
Further, solving the parameter omega of the mixed Copula in the formula (3) by adopting a genetic algorithm1、ω2、ω3、θ1、θ2、θ3The method comprises the following steps:
s21, calculating edge distribution values u, v of the wind speeds of the first equivalent fan and the second equivalent fan by a Gaussian kernel density estimation method;
s22, determining a penalty factor gamma of a fitness function of a Genetic Algorithm (GA), and establishing a fitness function f by combining constraint conditions in a formula (4)sThe method comprises the following steps:
wherein, Ce(u, v) is an empirical Copula function, ui,viThe ith value represents the edge distribution of the first and second equivalent fans, i represents the number of the edge distribution of the fan, Ce(u, v) are specifically as follows:
Ce(u,v)=P{U<u,V<v} (6)
in the formula, U and V are edge distribution of historical wind speed;
s23, determining the maximum number G of genetic iterationsmAnd a crossing rate pcThe maximum number of genetic iterations represents the number of iterations, the crossover rate pcIs the probability of individual variation in the population, GmSet to 200-500, pcSet to 0.9-0.97;
s24, solving omega by combining the formulas (4), (5) and (6)1、ω2、ω3、θ1、θ2、θ3。
Further, the residual value of the wind speed is obtained through the wind speed edge distribution values of the equivalent fans and the mixed Copula function, and the method specifically comprises the following steps:
s31, dividing a data set of the first equivalent fan into a training set and a testing set, inputting the training set into an LSTM model for training to obtain an LSTM wind speed prediction model, solving a preliminary prediction value of the test set LSTM model wind speed, and setting the total number of samples of the test set as n;
s32, solving the edge distribution value v of the second equivalent fan at the corresponding moment, wherein the number of the edge distribution value v is the same as that of the first equivalent fan test sets;
s33, subtracting the actual value of the first equivalent fan at the corresponding moment from the predicted value of the LSTM model wind speed to obtain a residual sequence, recording the residual sequence as y, and obtaining an edge distribution function w of the residual sequence, wherein the method specifically comprises the following steps:
w=G(y)
s34, establishing a mixed Copula function of the wind speed of the equivalent fan 2 and the residual value of the equivalent fan 1;
s35, when the wind speed value of the second equivalent fan at the moment of t-1 is known, obtaining the corresponding edge distribution value v of the second equivalent fant-1V is to bet-1Substituting into the mixed Copula function expression C as follows:
C=C(vt-1,w) (7)
s36, pairing the numbers obtained aboveFitting to obtain C (v)t-1W) and w are approximated by the following equation:
C=a1wm+a2wm-1+…amw+am+1 (8)
in the formula, a1,a2,…am+1Is a polynomial coefficient, and m is the order of w;
s37, simultaneous solution of formula (7) and formula (8), which can be solved to obtain w*The variable y is at time t-1The predicted value is:
in the formula, G-1Is the inverse of the cumulative distribution function of the residual error w;
s38, calculating the residual value of the first equivalent fan at the time tUsing residual error at time t-1Approximately replacing the residual at time tObtaining a residual error value of the first equivalent fan at the time tThe method comprises the following specific steps:
in the formula (I), the compound is shown in the specification,is the predicted value of the variable y at the time t.
Further, an accurate predicted value of the wind speed is obtained through the preliminary predicted value and the residual value of the wind speed, and the method specifically comprises the following steps:
inputting indexes such as wind speed, power, generator rotating speed and blade angle by using the LSTM wind speed prediction model to obtain a preliminary predicted value f of the wind speed at the time tt;
Obtaining a wind speed predicted value with higher precision, which is specifically as follows:
in the formula (f)tIs the preliminary predicted value of the wind speed of the LSTM wind speed prediction model at the time t,is the wind speed residual prediction value at the time t,the predicted value of the wind speed at the time t is high in precision.
The invention has the beneficial effects that: the fans in the wind power plant are classified by using a K-medoids clustering algorithm to obtain equivalent fans, the randomness of the wind speed is reduced, the equivalent fans are modeled by using a mixed Copula function and then combined with a long-term network and a short-term network, and the precision of wind speed prediction is improved.
Additional features and advantages of the invention will be set forth in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. The objectives and other advantages of the invention will be realized and attained by the structure particularly pointed out in the written description and claims hereof as well as the appended drawings.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, and it is obvious that the drawings in the following description are some embodiments of the present invention, and those skilled in the art can also obtain other drawings according to the drawings without creative efforts.
FIG. 1 shows a flow diagram of a short term wind speed prediction method according to an embodiment of the invention;
fig. 2 is a schematic flow chart illustrating a criterion function Fs for determining k wind turbine clusters according to the short-term wind speed prediction method in the embodiment of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are some, but not all, embodiments of the present invention. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
Referring to fig. 1, fig. 1 is a flow chart illustrating a short-term wind speed prediction method according to an embodiment of the invention, the short-term wind speed prediction method includes the following steps:
carrying out scene clustering on the wind turbine generator by using a K-medoids clustering algorithm to obtain a plurality of equivalent fans;
constructing an LSTM (Long Short-Term Memory, Long and Short Term Memory network) model, and training a plurality of equivalent fans through the LSTM model to obtain a preliminary predicted value of the wind speed;
establishing a functional relation among a plurality of equivalent fans by using a hybrid Copula function, and solving parameters in the hybrid Copula function by using a Genetic Algorithm (GA);
solving a residual value of the wind speed by utilizing the wind speed edge distribution value and a mixed Copula function;
and summing the preliminary predicted value and the residual value of the wind speed to obtain an accurate predicted value of the wind speed.
The fans in the wind power plant are classified by using a K-medoids clustering algorithm, then the classes are regarded as equivalent fans, modeling is carried out on the equivalent fans by using a hybrid Copula, and then the equivalent fans are combined with a long-term and short-term memory network to obtain a more accurate wind speed predicted value.
Long-short term memory is a special recurrent neural network, in which a concept called "gate" is introduced, which can play a role in regulating information. Through the construction of individual gates, useful information of previous sequences can be well "remembered". The LSTM model can be easily constructed using the Keras (Keras is an open source artificial neural network library written by Python) framework.
Carrying out scene clustering on the wind turbine generator by using a K-medoids clustering algorithm to obtain a plurality of equivalent fans, which comprises the following steps:
and acquiring historical wind speed data of each wind turbine generator from the wind power plant.
For a wind power plant, indexes such as wind speed, wind power and the like have certain regularity. Since the wind speed is due to the air flow caused by the change of solar thermal radiation, the wind speed has an integral characteristic that the wind speed at a future time is related to the wind speed at the present time and the historical wind speed. Therefore, according to historical relevant data, the rule is learned by using a deep learning or machine learning method such as LSTM, and the like, and a wind speed prediction model can be obtained.
Historical wind speed, power, generator speed and blade angle data of the region are obtained from a meteorological station, and are analyzed and processed to reduce random noise generated in wind speed measurement and signal acquisition and determine model input and output variables.
Carrying out scene clustering on the wind turbine generator by adopting K-medoids clustering, taking the wind speed and the power of a fan as horizontal and vertical coordinate values respectively, constructing a rectangular coordinate system, and establishing a criterion function Fs, wherein the criterion function Fs is the Euclidean distance from all other points in the current class to the medoids of the central point, and the expression of the ith class (i is more than or equal to 1 and less than or equal to K) is as follows:
in the formula, j represents the number of the fan variables in each class, n represents the total number of the fan variables in each class, and xi、yiAs a coordinate of the center point, xj、yjCoordinates in the class other than the center point.
Referring to fig. 2, fig. 2 shows a flow diagram of determining k wind turbine clusters according to a criterion solving function Fs of a short-term wind speed prediction method according to an embodiment of the present invention, where the criterion solving function Fs determines the k wind turbine clusters, and the specific steps are as follows:
and S11, randomly selecting k points from the n samples as central points medoids.
And S12, distributing the rest points to the current best cluster according to the principle of being closest to the center.
S13, sequentially calculating the value of the criterion function Fs when the points are medoids for all other points except the medoids in each class, traversing all the possibilities, and selecting the point corresponding to the minimum criterion function Fs as a new medoid.
S14, repeating S2 and S3 until all medoids are not changing.
And S15, finally determining k wind turbine generator clusters through continuous iterative operation.
Determining each type of wind turbine generator as an equivalent fan, wherein the method for determining each type of wind turbine generator as the equivalent fan specifically comprises the following steps:
the power of the equivalent fan is the sum of the powers of the clustered wind turbine generators;
the wind speed of the equivalent fan is the average value of the wind speeds of the clustered wind generation sets;
the generator rotating speed of the equivalent fan is the average value of the generator rotating speeds of the clustered wind turbine generators;
and the blade angle of the equivalent fan is the average value of the blade angles of the clustered wind turbine generators.
Constructing an LSTM model, training a plurality of equivalent fans through the LSTM model, and obtaining an initial predicted value of the wind speed, wherein the method specifically comprises the following steps:
and (3) the calculated wind speed, power, generator rotating speed and blade angle of the equivalent fan are arranged into a new data set, the new data set is divided into a training set and a testing set, and the training set is input into an LSTM model for training to obtain a model parameter and a preliminary predicted value of the wind speed.
For example, the training set and test set have a data ratio of 8: 2.
Establishing a functional relation among a plurality of equivalent fans by using a hybrid Copula function, and solving parameters in the hybrid Copula function by using a Genetic Algorithm (GA), wherein the parameters are as follows:
for the classified scenes, according to the sklar theorem, the joint distribution of N random variables can be decomposed into N univariate distributions and a Copula function.
Let x be1,x2,…xnN random wind speed variables, the edge distribution of the n random wind speed variables is respectively f1(x1),f2(x2),…,fn(xn) The combined distribution is f (x)1,x2,…xn) Let there be a function C, such that:
f(x1,x2,…xn)=C(f1(x1),f2(x2),…,fn(xn)) (2)
let u1=f1(x1),u2=f2(x2),…,un=fn(xn) Wherein u is1,u2…unObtaining a mathematical table of Copula function for the edge distribution of wind speedThe expression is as follows: c (u)1,u2,…un)。
The Copula functions with different characteristics are adopted to form a mixed Copula to establish a dependent structure among the wind speeds, so that the prediction accuracy can be improved. Gumbel, Clayton and Frank are combined according to a certain proportion, wherein Cg、CcAnd CfAbbreviations for Gumbel, Clayton, Frank Copula functions, respectively. The model integrates the advantages of each single Copula and can better reflect data characteristics.
Taking two-dimensional Copula as an example, a hybrid Copula function is established, and the specific form C is as follows:
C(u,v;θ)=ω1Cg(u,v;θ1)+ω2Cc(u,v;θ2)+ω3Cf(u,v;θ3) (3)
ω1+ω2+ω3=1,ω1,ω2,ω3≥0 (4)
wherein u is the wind speed edge distribution value of the first equivalent fan, v is the wind speed edge distribution value of the second equivalent fan, omega1,ω2,ω3As a weight parameter of the model, θ1,θ2,θ3Are dependent parameters of the model.
From the above equation (3), it can be known that the mixed Copula has six parameters, and compared with a single Copula, the mixed Copula can describe the correlation degree between variables more easily, and the parameters thereof more specifically show the dependency relationship between the variables, and a Genetic Algorithm (GA) is used to solve the parameter ω of the mixed Copula in the above equation (3)1、ω2、ω3、θ1、θ2、θ3The method comprises the following specific steps:
s21, edge distribution values u and v of the wind speeds of the first equivalent fan and the second equivalent fan are obtained by a Gaussian kernel density estimation method, wherein u and v represent the probability that the wind speeds of the first equivalent fan and the second equivalent fan are smaller than a specific value.
S22, determining a penalty factor gamma of a fitness function of a Genetic Algorithm (GA), and combining constraint conditions in the formula (4) to establishVertical fitness function fsThe method comprises the following steps:
wherein, Ce(u, v) is an empirical Copula function, ui,viAnd the ith value represents the edge distribution of the first and second equivalent fans, and i represents the number of the edge distribution of the fan.
Taking a two-dimensional Copula function as an example, Ce(u, v) the specific formula is as follows:
Ce(u,v)=P{U<u,V<v} (6)
in the formula, U and V are edge distributions of the historical wind speed.
S23, determining the maximum number G of genetic iterationsmAnd a crossing rate pcThe maximum number of genetic iterations represents the number of iterations, the crossover rate pcIs the probability of individual variation in the population, GmSet to 200-500, pcSet to 0.9-0.97.
S24, solving omega by combining the formulas (4), (5) and (6)1,ω2,ω3And theta1,θ2,θ3。
And solving a residual value of the wind speed through the wind speed edge distribution values of the equivalent fans and a mixed Copula function, wherein the residual value is as follows:
and S31, dividing the obtained data set of the first equivalent fan into a training set and a testing set according to the ratio of 8:2, inputting the training set into an LSTM model for training to obtain an LSTM wind speed prediction model, obtaining a preliminary predicted value of the wind speed of the LSTM model in the testing set, and setting the total number of samples in the testing set to be n.
And S32, obtaining the edge distribution value v of the second equivalent fan at the corresponding moment, wherein the number of the edge distribution value v is the same as that of the first equivalent fan test sets.
S33, subtracting the actual value of the first equivalent fan at the corresponding moment from the predicted value of the LSTM model wind speed to obtain a residual sequence, recording the residual sequence as y, and obtaining an edge distribution function w of the residual sequence, wherein the method specifically comprises the following steps:
w=G(y)
and S34, establishing a mixed Copula function of the wind speed of the equivalent fan 2 and the residual value of the equivalent fan 1.
S35, when the wind speed value of the second equivalent fan at the moment of t-1 is known, obtaining the corresponding edge distribution value v of the second equivalent fant-1V is to bet-1Substituting into the mixed Copula function expression C as follows:
C=C(vt-1,w) (7)
s36, pairing the numbers obtained above(n is the total number of samples) and fitting to obtain C (v)t-1W) and w are approximated by the following equation:
C=a1wm+a2wm-1+…amw+am+1 (8)
in the formula, a1,a2,…am+1Is a polynomial coefficient, and m is the order of w.
S37, simultaneous solution of formula (7) and formula (8), which can be solved to obtain w*The variable y is at time t-1The predicted value is:
in the formula, G-1Is the inverse of the cumulative distribution function of the residuals w.
Since the residual error has differential characteristics, the residual error at the time t-1 is usedApproximately replacing the residual at time tObtaining the residual error value of the first equivalent fan at the time tThe method comprises the following specific steps:
in the formula (I), the compound is shown in the specification,is the predicted value of the variable y at the time t.
Summing the preliminary predicted value and the residual value of the wind speed to obtain an accurate predicted value of the wind speed, which is as follows:
inputting indexes such as wind speed, power, generator rotating speed and blade angle by using the obtained LSTM wind speed prediction model to obtain a preliminary predicted value f of the wind speed at the time tt。
Summing the residual value and the corresponding preliminary predicted value of the wind speed to obtain a wind speed predicted value with higher precision, which is as follows:
in the formula (f)tIs the preliminary predicted value of the wind speed of the LSTM wind speed prediction model at the time t,is the wind speed residual prediction value at the time t,the predicted value of the wind speed at the time t is high in precision.
Although the present invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some technical features may be equivalently replaced; and such modifications or substitutions do not depart from the spirit and scope of the corresponding technical solutions of the embodiments of the present invention.
Claims (10)
1. A short-term wind speed prediction method is characterized by comprising the following steps:
carrying out scene clustering on the wind turbine generator by using a K-medoids clustering algorithm to obtain a plurality of equivalent fans;
constructing an LSTM model, and training a plurality of equivalent fans through the LSTM model to obtain an initial predicted value of the wind speed;
establishing a functional relation among a plurality of equivalent fans by using a hybrid Copula function, and solving parameters in the hybrid Copula function by using a genetic algorithm;
solving a residual value of the wind speed through the wind speed edge distribution values of the equivalent fans and a mixed Copula function;
and obtaining an accurate predicted value of the wind speed through the initial predicted value and the residual value of the wind speed.
2. The short term wind speed prediction method of claim 1,
carrying out scene clustering on the wind turbine generator by using a K-medoids clustering algorithm to obtain a plurality of equivalent fans, which comprises the following steps:
acquiring historical wind speed data of each wind turbine generator from a wind power plant;
acquiring historical data of wind speed, power, generator rotating speed and blade angle of a region from a meteorological station, analyzing and processing the data, and determining model input and output variables;
carrying out scene clustering on the wind turbine generator by adopting K-medoids clustering, respectively taking the wind speed and the power of the wind turbine generator as horizontal and vertical coordinate values, constructing a rectangular coordinate system, and establishing a criterion function Fs, wherein the method specifically comprises the following steps:
in the formula, j represents the number of the fan variables in each class, n represents the total number of the fan variables in each class, and xi、yiAs a coordinate of the center point, xj、yjCoordinates in class except the central point;
solving a criterion function Fs, and determining k wind turbine generator clusters;
and determining each type of wind turbine generator as an equivalent fan.
3. The short term wind speed prediction method of claim 2,
solving a criterion function Fs, and determining k wind turbine generator clusters, specifically as follows:
s11, randomly selecting k points from the n samples as central points medoids;
s12, distributing the rest points to the current best cluster according to the principle of being nearest to the center;
s13, sequentially calculating the value of the criterion function Fs when the points are medoids for all other points except the medoids in each class, traversing all possibilities, and selecting the point corresponding to the minimum criterion function Fs as a new medoid;
s14, repeating S12 and S13 until all medoids are not changed;
and S15, finally determining k wind turbine generator clusters through continuous iterative operation.
4. The short-term wind speed prediction method according to claim 2 or 3, wherein the method for determining an equivalent fan is specifically as follows:
the power of the equivalent fan is the sum of the powers of the clustered wind turbine generators;
the wind speed of the equivalent fan is the average value of the wind speeds of the clustered wind generation sets;
the generator rotating speed of the equivalent fan is the average value of the generator rotating speeds of the clustered wind turbine generators;
and the blade angle of the equivalent fan is the average value of the blade angles of the clustered wind turbine generators.
5. The short term wind speed prediction method of claim 4,
constructing an LSTM model, training a plurality of equivalent fans through the LSTM model, and obtaining an initial predicted value of the wind speed, wherein the method specifically comprises the following steps:
and (3) the calculated wind speed, power, generator rotating speed and blade angle of the equivalent fan are arranged into a new data set, the new data set is divided into a training set and a testing set, and the training set is input into an LSTM model for training to obtain a model parameter and a preliminary predicted value of the wind speed.
6. The short term wind speed prediction method of claim 5, wherein the training set and the test set have a data ratio of 8: 2.
7. The short term wind speed prediction method of claim 6,
establishing a functional relation among a plurality of equivalent fans by using a hybrid Copula function, and solving parameters in the hybrid Copula function by using a genetic algorithm, wherein the parameters are as follows:
let x be1,x2,...xnN random wind speed variables, the edge distribution of the n random wind speed variables is respectively f1(x1),f2(x2),...,fn(xn) The combined distribution is f (x)1,x2,...xn) Let there be a function C, such that:
f(x1,x2,...xn)=C(f1(x1),f2(x2),...,fn(xn)) (2)
let u1=f1(x1),u2=f2(x2),...,un=fn(xn) Wherein u is1,u2...unFor the edge distribution of wind speed, a mathematical expression of the Copula function is obtained: c (u)1,u2,...un);
Establishing a mixed Copula function, wherein the concrete form C is as follows:
C(u,v;θ)=ω1Cg(u,v;θ1)+ω2Cc(u,v;θ2)+ω3Cf(u,v;θ3) (3)
ω1+ω2+ω3=1,ω1,ω2,ω3≥0 (4)
wherein u is the wind speed edge distribution value of the first equivalent fan, v is the wind speed edge distribution value of the second equivalent fan, omega1,ω2,ω3As a weight parameter of the model, θ1,θ2,θ3Is a dependent parameter of the model;
solving parameter omega of mixed Copula in the formula (3) by adopting genetic algorithm1、ω2、ω3、θ1、θ2、θ3。
8. The short term wind speed prediction method of claim 7,
solving parameter omega of mixed Copula in the formula (3) by adopting genetic algorithm1、ω2、ω3、θ1、θ2、θ3The method comprises the following steps:
s21, calculating edge distribution values u, v of the wind speeds of the first equivalent fan and the second equivalent fan by a Gaussian kernel density estimation method;
s22, determining a penalty factor gamma of a fitness function of a Genetic Algorithm (GA), and establishing a fitness function f by combining constraint conditions in a formula (4)sThe method comprises the following steps:
wherein, Ce(u, v) is an empirical Copula function, ui,viThe ith value represents the edge distribution of the first and second equivalent fans, i represents the number of the edge distribution of the fan, Ce(u, v) are specifically as follows:
Ce(u,v)=P{U<u,V<v} (6)
In the formula, U and V are edge distribution of historical wind speed;
s23, determining the maximum number G of genetic iterationsmAnd a crossing rate pcThe maximum number of genetic iterations represents the number of iterations, the crossover rate pcIs the probability of individual variation in the population, GmSet to 200-500, pcSet to 0.9-0.97;
s24, solving omega by combining the formulas (4), (5) and (6)1、ω2、ω3、θ1、θ2、θ3。
9. The short term wind speed prediction method of claim 8,
and solving a residual value of the wind speed through the wind speed edge distribution values of the equivalent fans and a mixed Copula function, wherein the residual value is as follows:
s31, dividing a data set of the first equivalent fan into a training set and a testing set, inputting the training set into an LSTM model for training to obtain an LSTM wind speed prediction model, solving a preliminary prediction value of the test set LSTM model wind speed, and setting the total number of samples of the test set as n;
s32, solving the edge distribution value v of the second equivalent fan at the corresponding moment, wherein the number of the edge distribution value v is the same as that of the first equivalent fan test sets;
s33, subtracting the actual value of the first equivalent fan at the corresponding moment from the predicted value of the LSTM model wind speed to obtain a residual sequence, recording the residual sequence as y, and obtaining an edge distribution function w of the residual sequence, wherein the method specifically comprises the following steps:
w=G(y)
s34, establishing a mixed Copula function of the wind speed of the equivalent fan 2 and the residual value of the equivalent fan 1;
s35, when the wind speed value of the second equivalent fan at the moment of t-1 is known, obtaining the corresponding edge distribution value v of the second equivalent fant-1V is to bet-1Substituting into the mixed Copula function expression C as follows:
C=C(vt-1,w) (7)
s36, pairing the numbers obtained aboveFitting to obtain C (v)t-1W) and w are approximated by the following equation:
C=a1wm+a2wm-1+…amw+am+1 (8)
in the formula, a1,a2,...am+1Is a polynomial coefficient, and m is the order of w;
s37, simultaneous solution of formula (7) and formula (8), which can be solved to obtain w*The variable y is at time t-1The predicted value is:
in the formula, G-1Is the inverse of the cumulative distribution function of the residual error w;
s38, calculating the residual value of the first equivalent fan at the time tUsing residual error at time t-1Approximately replacing the residual at time tObtaining a residual error value of the first equivalent fan at the time tThe method comprises the following specific steps:
10. The short term wind speed prediction method of claim 9,
obtaining an accurate predicted value of the wind speed through the initial predicted value and the residual value of the wind speed, which is as follows:
inputting indexes such as wind speed, power, generator rotating speed and blade angle by using the LSTM wind speed prediction model to obtain a preliminary predicted value f of the wind speed at the time tt;
Obtaining a wind speed predicted value with higher precision, which is specifically as follows:
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110506041.8A CN113312836A (en) | 2021-05-10 | 2021-05-10 | Short-term wind speed prediction method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110506041.8A CN113312836A (en) | 2021-05-10 | 2021-05-10 | Short-term wind speed prediction method |
Publications (1)
Publication Number | Publication Date |
---|---|
CN113312836A true CN113312836A (en) | 2021-08-27 |
Family
ID=77372823
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202110506041.8A Pending CN113312836A (en) | 2021-05-10 | 2021-05-10 | Short-term wind speed prediction method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN113312836A (en) |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107292439A (en) * | 2017-06-23 | 2017-10-24 | 广东工业大学 | A kind of method and apparatus for the short-term wind speed forecasting that Copula functions are mixed based on time-varying |
CN108615097A (en) * | 2018-05-10 | 2018-10-02 | 广东工业大学 | A kind of wind speed forecasting method, system, equipment and computer readable storage medium |
CN109272139A (en) * | 2018-07-24 | 2019-01-25 | 东华大学 | It is a kind of based on Nonlinear Set at the short-term wind speed forecasting method of deep learning |
CN111159900A (en) * | 2019-12-30 | 2020-05-15 | 国核电力规划设计研究院有限公司 | Method and device for predicting wind speed of fan |
CN111353641A (en) * | 2020-02-26 | 2020-06-30 | 西南交通大学 | Modeling method based on wind speed and wind direction combined distribution along high-speed rail |
CN112597620A (en) * | 2020-09-24 | 2021-04-02 | 新天绿色能源股份有限公司 | Wind turbine generator power curve modeling method based on k-medoids algorithm |
-
2021
- 2021-05-10 CN CN202110506041.8A patent/CN113312836A/en active Pending
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107292439A (en) * | 2017-06-23 | 2017-10-24 | 广东工业大学 | A kind of method and apparatus for the short-term wind speed forecasting that Copula functions are mixed based on time-varying |
CN108615097A (en) * | 2018-05-10 | 2018-10-02 | 广东工业大学 | A kind of wind speed forecasting method, system, equipment and computer readable storage medium |
CN109272139A (en) * | 2018-07-24 | 2019-01-25 | 东华大学 | It is a kind of based on Nonlinear Set at the short-term wind speed forecasting method of deep learning |
CN111159900A (en) * | 2019-12-30 | 2020-05-15 | 国核电力规划设计研究院有限公司 | Method and device for predicting wind speed of fan |
CN111353641A (en) * | 2020-02-26 | 2020-06-30 | 西南交通大学 | Modeling method based on wind speed and wind direction combined distribution along high-speed rail |
CN112597620A (en) * | 2020-09-24 | 2021-04-02 | 新天绿色能源股份有限公司 | Wind turbine generator power curve modeling method based on k-medoids algorithm |
Non-Patent Citations (3)
Title |
---|
刘立燕: "基于Copula函数和神经网络模型的洪水预测", 中国优秀硕士论文电子期刊网 工程科技Ⅱ辑, no. 2, 15 February 2019 (2019-02-15), pages 037 - 3 * |
王红军 等: "M-Copula模型在金融时间序列分析中的研究与应用", 系统科学与数学, vol. 40, no. 6, 15 June 2020 (2020-06-15), pages 1117 - 1132 * |
邵煦晴: "基于Copula函数探究内蒙古自治区风速关系", 中国优秀硕士论文电子期刊网 基础科学辑, no. 2, 15 February 2021 (2021-02-15), pages 009 - 48 * |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Wang et al. | Wind power interval prediction based on improved PSO and BP neural network | |
CN108428017B (en) | Wind power interval prediction method based on nuclear extreme learning machine quantile regression | |
CN112488395A (en) | Power distribution network line loss prediction method and system | |
CN110910004A (en) | Reservoir dispatching rule extraction method and system with multiple uncertainties | |
CN110969290A (en) | Runoff probability prediction method and system based on deep learning | |
Shao et al. | Wind speed forecast based on the LSTM neural network optimized by the firework algorithm | |
CN106651007A (en) | Method and device for GRU-based medium and long-term prediction of irradiance of photovoltaic power station | |
CN112884236B (en) | Short-term load prediction method and system based on VDM decomposition and LSTM improvement | |
CN112241836B (en) | Virtual load leading parameter identification method based on incremental learning | |
Wu et al. | Optimal economic dispatch model based on risk management for wind‐integrated power system | |
Yi et al. | Intelligent prediction of transmission line project cost based on least squares support vector machine optimized by particle swarm optimization | |
CN116702937A (en) | Photovoltaic output day-ahead prediction method based on K-means mean value clustering and BP neural network optimization | |
CN116187835A (en) | Data-driven-based method and system for estimating theoretical line loss interval of transformer area | |
Wibawa et al. | Long Short-Term Memory to Predict Unique Visitors of an Electronic Journal | |
CN116523142A (en) | Photovoltaic cluster interval prediction method considering space-time characteristic clustering | |
Zhu et al. | Research on PSO‐ARMA‐SVR Short‐Term Electricity Consumption Forecast Based on the Particle Swarm Algorithm | |
CN117408394B (en) | Carbon emission factor prediction method and device for electric power system and electronic equipment | |
Guo et al. | Research on short-term forecasting method of photovoltaic power generation based on clustering SO-GRU method | |
CN110675276B (en) | Method and system for inversion droop control of direct current power transmission system | |
CN112633591A (en) | Space search method and device based on deep reinforcement learning | |
CN117077546A (en) | Power system load modeling method and system based on data driving | |
Sun et al. | Support vector machines with PSO algorithm for short-term load forecasting | |
CN113312836A (en) | Short-term wind speed prediction method | |
CN109615142A (en) | A kind of wind farm wind velocity combination forecasting method based on wavelet analysis | |
CN115034159A (en) | Power prediction method, device, storage medium and system for offshore wind farm |
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 |