CN110334847A - Based on the wind power prediction method for improving K-means cluster and support vector machines - Google Patents
Based on the wind power prediction method for improving K-means cluster and support vector machines Download PDFInfo
- Publication number
- CN110334847A CN110334847A CN201910437371.9A CN201910437371A CN110334847A CN 110334847 A CN110334847 A CN 110334847A CN 201910437371 A CN201910437371 A CN 201910437371A CN 110334847 A CN110334847 A CN 110334847A
- Authority
- CN
- China
- Prior art keywords
- wind power
- power prediction
- data
- support vector
- time period
- 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 66
- 238000012706 support-vector machine Methods 0.000 title claims abstract description 33
- 238000003064 k means clustering Methods 0.000 claims abstract description 9
- 230000003044 adaptive effect Effects 0.000 claims abstract description 8
- 230000008859 change Effects 0.000 claims abstract description 7
- 238000004364 calculation method Methods 0.000 claims description 5
- 238000012549 training Methods 0.000 claims description 4
- PEDCQBHIVMGVHV-UHFFFAOYSA-N Glycerine Chemical compound OCC(O)CO PEDCQBHIVMGVHV-UHFFFAOYSA-N 0.000 claims description 3
- 238000012216 screening Methods 0.000 claims description 3
- 241000208340 Araliaceae Species 0.000 claims 1
- 235000005035 Panax pseudoginseng ssp. pseudoginseng Nutrition 0.000 claims 1
- 235000003140 Panax quinquefolius Nutrition 0.000 claims 1
- 235000008434 ginseng Nutrition 0.000 claims 1
- ZFXYFBGIUFBOJW-UHFFFAOYSA-N theophylline Chemical compound O=C1N(C)C(=O)N(C)C2=C1NC=N2 ZFXYFBGIUFBOJW-UHFFFAOYSA-N 0.000 claims 1
- 230000006872 improvement Effects 0.000 abstract description 7
- 230000000694 effects Effects 0.000 abstract description 4
- 238000010276 construction Methods 0.000 abstract description 2
- 239000000523 sample Substances 0.000 description 33
- 230000006870 function Effects 0.000 description 7
- 238000013528 artificial neural network Methods 0.000 description 6
- 238000012731 temporal analysis Methods 0.000 description 4
- 238000000700 time series analysis Methods 0.000 description 4
- 230000008901 benefit Effects 0.000 description 3
- 238000010586 diagram Methods 0.000 description 3
- 241001269238 Data Species 0.000 description 2
- 230000005611 electricity Effects 0.000 description 2
- 238000010248 power generation Methods 0.000 description 2
- 238000004393 prognosis Methods 0.000 description 2
- 239000013074 reference sample Substances 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000013277 forecasting method Methods 0.000 description 1
- 238000009434 installation Methods 0.000 description 1
- 238000012423 maintenance Methods 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 230000008569 process Effects 0.000 description 1
Classifications
-
- 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
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Economics (AREA)
- General Physics & Mathematics (AREA)
- Human Resources & Organizations (AREA)
- Strategic Management (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Tourism & Hospitality (AREA)
- General Engineering & Computer Science (AREA)
- General Business, Economics & Management (AREA)
- Marketing (AREA)
- Evolutionary Computation (AREA)
- Evolutionary Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Artificial Intelligence (AREA)
- Entrepreneurship & Innovation (AREA)
- Development Economics (AREA)
- Probability & Statistics with Applications (AREA)
- Primary Health Care (AREA)
- General Health & Medical Sciences (AREA)
- Water Supply & Treatment (AREA)
- Public Health (AREA)
- Game Theory and Decision Science (AREA)
- Quality & Reliability (AREA)
- Operations Research (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- Wind Motors (AREA)
Abstract
The invention discloses a kind of based on the wind power prediction method for improving K-means cluster and support vector machines, step 1: the data sample of construction wind speed amplitude, step 2: normalized；Step 3: with K-means clustering method is improved, the classification sum C according to wind speed adaptive change is obtained；Step 4: scale counts according to set time, obtains L continuous time period sample；Step 5: after continuous time period sample is normalized, cluster obtains M class data set；Step 6: short-term wind power prediction model is established using support vector machines；Step 7: it introduces real time data and obtains ultra-short term wind power prediction value, introduce real-time wind power, obtain ultra-short term wind power prediction model；Step 8: input prediction day and its first several days wind power output data and numerical value data of weather forecast obtain wind power results.Wind power prediction method provided by the invention based on improvement K-means cluster and support vector machines, Clustering Effect is good, there is very strong generalization ability and robustness.
Description
Technical field
The present invention relates to wind power prediction technical fields, specifically, being related to a kind of based on improvement K-means cluster and branch
Hold the wind power prediction method of vector machine.
Background technique
Wind-power electricity generation has the advantages that pollution level is low, installation is flexible, operation maintenance cost is low etc., has been increasingly becoming using most
One of extensive distributed power generation type.However, wind-power electricity generation has the characteristics such as uncertain, fluctuation and intermittence, work as wind
When TV university scale accesses power grid, the unstability of wind power output power will bring huge challenge to system stable operation.
The precision of wind power prediction is of great significance for wind power system, improves the prediction essence of output power of wind power generation
Degree, can effectively reduce the stand-by time of system operation, to reduce the operating cost of wind power system, while guarantee that power grid is transported
Capable stability.Based on this, certain research is expanded to wind power prediction both at home and abroad.Wide prediction technique master is applied at present
Have: time series analysis method and artificial neural network method.Wherein, time series analysis method model basis, by external interference journey
Smaller, shorter suitable for lead prediction is spent, however as the increase of prediction lead, precision of prediction is often difficult to reach pre-
Phase；Artificial neural network method has stronger non-mapping ability and self study, adaptive ability, is suitble to description power prediction model
Complex nonlinear feature, but artificial neural network method net training time is long, and convergence rate is slower, and is easy to fall into part most
It is excellent.
Domestic existing literature is retrieved and is found, document " wind power time series analysis and prediction technique study " (Na
Benevolence figure is refined；University Of Tianjin) a kind of wind power prediction method based on improvement time series analysis method is proposed, simplifying operation
Precision of prediction, but its power prediction for not being suitable for large-scale wind power field are improved while journey to a certain extent.Document " base
In the wind power plant short term power forecasting research of BP neural network optimization " (Ma Cong；Kunming University of Science and Technology) propose it is a kind of based on changing
Into the neural network wind power Short-term Forecasting Model of BP algorithm, the prediction error of neural network is reduced in a certain range,
However its learning algorithm training time is longer, and easily falls into local optimum.
Summary of the invention
The purpose of the present invention is to provide a kind of based on the wind power prediction side for improving K-means cluster and support vector machines
Method, Clustering Effect is good, there is very strong generalization ability and robustness.
Skill used by wind power prediction method disclosed by the invention based on improvement K-means cluster with support vector machines
Art scheme is:
A kind of wind power prediction method based on improvement K-means cluster and support vector machines, comprising the following steps:
Step 1: the wind speed amplitude of each period in acquisition one day, and construct the data sample of wind speed amplitude；
Step 2: data sample is normalized；
Step 3: with K-means clustering method is improved, the data sample in one-year age section is clustered, is obtained
Classification sum C, the classification sum C is according to wind speed adaptive change；
Step 4: according to the time scale of setting, by the data sample after being clustered in step 3 by L continuous time period into
Row statistics, obtains L continuous time period sample；
Step 5: after continuous time period sample is normalized according to the method in step 2, according to step 3
In method continuous time period sample is clustered, obtain M class data set；
Step 6: corresponding model is trained for every a kind of data set using support vector machines, it is pre- to establish short-term wind power
Survey model；
Step 7: being introduced into real-time wind power output data and numerical value data of weather forecast into short-term wind power prediction model,
Ultra-short term wind power prediction value is obtained, introduces real-time wind power, and use to ultra-short term wind power prediction value and real-time wind power
The method of weight is modified, and obtains ultra-short term wind power prediction model；
Step 8: a few days ago to short-term wind power prediction model and ultra-short term wind power prediction mode input prediction day and prediction
Several days wind power output data and numerical value data of weather forecast obtain short-term wind power prediction and ultra-short term wind power prediction
Wind power results.
Preferably, the data sample constructed in the step 1 is as follows:
A=(as1,as2,…,asH,as max,as min,asmean,asstd)
In formula: as1,as2,…,asHFor the value of several node wind speed in one day；as maxFor day wind speed maximum value；as minFor
Day wind speed minimum value；asmeanFor day wind speed average value；asstdFor the standard deviation of day wind speed.
Preferably, in the step 2, the formula of the method for normalizing are as follows:
In formula, bijFor the ith sample value in j-th of physical quantity；bj minFor the minimum value in j-th of physical quantity；bj max
For the maximum value in j-th of physical quantity.
Preferably, the step of K-means cluster is improved in the step 3 is as follows:
3.1: setting initial category c=1, initialize the cluster centre of all data samples；
3.2: when clustering classification c=C', maximum one kind of Screening Samples capacity in cluster result is clustered
Heart point zwIt is split into symmetrical two points zw- ε and zw+ ε, it is c=C'+1 that the classification sum thus clustered, which increases,；
3.3: according to the cluster centre newly divided, data sample being reclassified；
3.4: according to reclassifying as a result, readjusting the position of each cluster centre using the method in step 3.2；
3.5: repeating step 3.3 to 3.4, until each cluster centre no longer moves；
3.6: repeating step 3.2 to 3.5, until all data samples are classified.
Preferably, in the step 3.5, as the central point z of clusterwThe distance of two points of division is less than setting
Positive number δ when, that is, meet | zw+ε-(zw- ε) | when the < δ condition of convergence, then assert that each cluster centre no longer moves.
Preferably, the time scale set in the step 4 is as week, in three time scales of half a month and the moon
One.
Preferably, the continuous time period sample constructed in the step 4 are as follows:
sl=(st1,st2,…,stC,ss max,ss min,ssmean,ssstd,ssmd1,ssmd2,ssmd3)
In formula, st1,st2,…stCFor the quantity on first of continuous time period all kinds of dates；smaxFor first of continuous time period
Interior wind speed maximum value；sminFor the wind speed minimum value in first of continuous time period；smeanFor being averaged for first continuous time period
Wind speed；ssstdFor the wind speed deviation in first of continuous time period；ssmd1It is absolutely flat for the first-order difference of first of continuous time period
Mean value；ssmd2For the second differnce absolute average of first of continuous time period；ssmd3For the third order difference of first of continuous time period
Absolute average.
Preferably, in the step 6, support vector machines method is by calling SVM-light and feature vector training
Model, target function type are as follows:
In formula, ξiWithFor relaxation factor；W and θ is band estimation parameter；ε is loss function；β is penalty factor.
Preferably, the physical quantity in described eigenvector includes wind speed, wind direction, temperature, humidity and air pressure.
Preferably, in the step 7, prediction a few days ago several days wind power output data and numerical weather forecasts are taken
It is established ultra-short term wind power prediction model as MC ° of prediction reference sample set by data in conjunction with air speed data ws, and is passed through
Sample attribute AT-1Predict the wind power plant generated output at T moment, calculation method is as follows:
F °=SVM (MC °, ws)
In formula, F ° is online wind power prediction model, xTFor the predicted value of T moment ultra-short term wind power；
Further according to a few days ago several days practical wind power of prediction, the prediction result of prediction day, calculating side are corrected using weight
Method is as follows:
In formula,For revised power prediction value；pD1、pD2、pD3…pDnRespectively predict a few days ago several days reality
Wind power sequence；ηTFor weight coefficient shared by predicted value, η1、η2、η3…ηnRespectively predict a few days ago several days weight coefficients.
Beneficial effect disclosed by the invention based on the wind power prediction method for improving K-means cluster and support vector machines
It is: the wind speed amplitude of each period in acquisition one day, and construct the data sample of wind speed amplitude.It then will be in one-year age section
Sample data clusters data sample, obtaining can be adaptive according to wind speed characteristics with K-means clustering method is improved
The change classification sum C answered.On this basis, the data sample after cluster is counted according to the time scale of setting, is obtained
It is same that continuous time period sample is clustered with improvement K-means clustering method to L continuous time period sample, obtain M
Class data set then trains corresponding model for every a kind of data set using support vector machines, establishes short-term wind power prediction
Model.Then real-time wind power output data and numerical value data of weather forecast are introduced into short-term wind power prediction model, are surpassed
Short-term wind power prediction value introduces real-time wind power, and to ultra-short term wind power prediction value and real-time wind power using weight
Method is modified, and obtains ultra-short term power prediction model.To short-term wind power prediction model and ultra-short term wind power prediction mould
Type input prediction day and its first several days wind power output data and numerical value data of weather forecast, obtain short-term wind power prediction and
The wind power results of ultra-short term wind power prediction.The advantage of this method is embodied in two aspects, first, by improving K-means
Clustering method enables classification sum according to wind speed characteristics adaptive change, substantially increases Clustering Effect；Second, support to
Amount machine is themselves based on empirical risk minimization, has good resolution ability to the problems such as non-linear and high dimension, has very strong
Generalization ability and robustness.
Detailed description of the invention
Fig. 1 is that the present invention is based on the process signals for improving K-means cluster with the wind power prediction method of support vector machines
Figure.
Fig. 2 is that the present invention is based on the short-term wind function for improving K-means cluster and the wind power prediction method of support vector machines
The analog result schematic diagram of rate type of prediction day 1.
Fig. 3 is that the present invention is based on the short-term wind function for improving K-means cluster and the wind power prediction method of support vector machines
The analog result schematic diagram of rate type of prediction day 2.
Fig. 4 is that the present invention is based on the ultra-short term wind for improving K-means cluster and the wind power prediction method of support vector machines
The analog result schematic diagram of power prediction typical case's day.
Specific embodiment
The present invention is further elaborated and is illustrated with Figure of description combined with specific embodiments below:
Referring to FIG. 1, a kind of, based on improving, K-means is clustered and the wind power prediction method of support vector machines includes following
Step:
Step 1: the wind speed amplitude of each period in acquisition one day, and construct the data sample of wind speed amplitude
Step 2: data sample is normalized；
Step 3: with K-means clustering method is improved, the data sample in one-year age section is carried out unsupervised poly-
Class obtains classification sum C, and the classification sum C is according to wind speed adaptive change；
Step 4: according to the time scale of setting, by the data sample after being clustered in step 3 by L continuous time period into
Row statistics, obtains L continuous time period sample；
Step 5: after continuous time period sample is normalized according to the method in step 2, according to step 3
In method to continuous time period sample carry out Unsupervised clustering, obtain M class data set；
Step 6: corresponding model is trained for every a kind of data set using support vector machines, it is pre- to establish short-term wind power
Survey model；
Step 7: being introduced into real-time wind power output data and numerical value data of weather forecast into short-term wind power prediction model,
Ultra-short term wind power prediction value is obtained, introduces real-time wind power, and use to ultra-short term wind power prediction value and real-time wind power
The method of weight is modified, and obtains ultra-short term wind power prediction model；
Step 8: to short-term wind power prediction model and ultra-short term wind power prediction mode input prediction day and its first three day
Wind power output data and numerical value data of weather forecast, obtain the wind power of short-term wind power prediction and ultra-short term wind power prediction
As a result.
The data sample constructed in step 1 is as follows:
A=(as1,as2,…,asH,as max,as min,asmean,asstd)
In formula: as1,as2,…,asHFor the value of several node wind speed in one day；as maxFor day wind speed maximum value；as minFor
Day wind speed minimum value；asmeanFor day wind speed average value；asstdFor the standard deviation of day wind speed.
In step 2, it is normalized using minimax method for normalizing, formula are as follows:
In formula, bijFor the ith sample value in j-th of physical quantity；bj minFor the minimum value in j-th of physical quantity；bj max
For the maximum value in j-th of physical quantity.
In step 3, primary condition c=1 is enabled, c=C is worked as, (2≤C and k ∈ N+) when, if the wind of different types can be distinguished
Speed, then c=C is cluster classification.If cluster sample set is X={ Xk|Xk∈Rp, k=1,2 ..., K, P ∈ N+, if having obtained class
Not and the quantity of cluster centre is C', Z={ Zc|Zc∈Rp, c=1,2 ..., C', P ∈ N+}。wcIndicate the C' class that cluster obtains
Not, then cluster centre is defined as:
Then objective function is defined as:
In formula, KcFor wcThe sample size that class is included；dck(xk,zk) indicate the kth point of c class to xkCluster centre zk's
Distance, the equal dimensionless of each component of X after normalization, using Euclidean distance as measuring, formula is as follows:
The step of K-means cluster is improved in step 3 is as follows:
3.1: setting initial category c=1, initialize the cluster centre of all sample datas；
3.2: as c=C', maximum one kind of Screening Samples capacity in cluster result, the central point z clusteredwPoint
It splits for symmetrical two points zw- ε and zw+ ε, it is c=C'+1 that the classification sum thus clustered, which increases,；
3.3: according to the cluster centre newly divided, data being reclassified；
3.4: according to classification results, the position of each cluster centre is readjusted using the method in step 3.2；
3.5: repeating step 3.3 to 3.4, until cluster centre no longer moves, as the central point z of clusterwTwo of division
When the distance of point is less than the positive number δ of setting, that is, meet | zw+ε-(zw- ε) | when the < δ condition of convergence, then assert each cluster centre
No longer move；
3.6: step 3.2 is repeated to 3.5, until all sample datas are classified, that is, can clearly will be different
Wind speed type distinguishes.
In step 4, the time scale set is one in three week, half a month and the moon time scales, the consecutive hours of construction
Between section sample are as follows:
sl=(st1,st2,…,stC,ss max,ss min,ssmean,ssstd,ssmd1,ssmd2,ssmd3)
In formula, st1,st2,…stCFor the quantity on first of continuous time period all kinds of dates；smaxFor first of continuous time period
Interior wind speed maximum value；sminFor the wind speed minimum value in first of continuous time period；smeanFor being averaged for first continuous time period
Wind speed；ssstdFor the wind speed deviation in first of continuous time period；ssmd1It is absolutely flat for the first-order difference of first of continuous time period
Mean value；ssmd2For the second differnce absolute average of first of continuous time period；ssmd3For the third order difference of first of continuous time period
Absolute average.
In step 6, support vector machines method is constructed using MATLAB, by calling SVM-light and feature vector to instruct
Practice model, wherein taking wind speed, wind direction, temperature, humidity, air pressure as the physical quantity in feature vector, target function type are as follows:
In formula, ξiWithFor relaxation factor；W and θ is band estimation parameter；ε is loss function；β is penalty factor.
In step 7, the wind power output data and numerical value data of weather forecast for taking prediction first three day of day are as prediction reference sample
MC ° of this collection, it is established to ultra-short term wind power prediction model in conjunction with air speed data ws, and pass through sample attribute AT-1When predicting T
The wind power plant generated output at quarter, calculation method are as follows:
F °=SVM (MC °, ws)
In formula, F ° is online wind power prediction model, xTFor the predicted value of T moment ultra-short term wind power；
Further according to the practical wind power of prediction day first three days, the prediction result of prediction day, calculation method are corrected using weight
It is as follows:
In formula,For revised power prediction value；pD1、pD2、pD3Respectively predict the practical wind power sequence of day first three days
Column；ηTFor weight coefficient shared by predicted value, η1、η2、η3Respectively predict the weight coefficient of day first three days.
The advantage of this method is embodied in two aspects, first, enabling classification sum by improving K-means clustering method
Enough according to wind speed characteristics adaptive change, Clustering Effect is substantially increased；Second, support vector machines is themselves based on structure risk most
Smallization criterion has good resolution ability to the problems such as non-linear and high dimension, there is very strong generalization ability and robustness.
The power forecasting method is illustrated by taking certain distributed wind power plant as an example.By its 2017 8,9, October
It is divided into two classes according to wind speed characteristics.The first kind is low wind speed characteristics day, and per day wind speed is lower, and minimum wind speed is lower than starting wind
Speed, and wind speed amplitude is relatively low；Second class is high wind speed characteristic day, and per day wind speed is low but wind speed amplitude is higher, wind speed deviation
It is larger.For 8,9, the historical data and meteorological data in October, short-term wind power prediction, two class date allusion quotations are carried out to wind power plant
Type power curve and prognosis modelling result are as shown in Figure 2 and Figure 3.For the historical data and meteorological data in July, to wind power plant into
The prediction of row super short-period wind power, the power curve and prognosis modelling result of typical day are as shown in Figure 4.
It can be seen that the curve of prediction can preferably track power output from Fig. 2,3,4.It is available by calculating
The predictablity rate of type day 1 is 96.38%, and the predictablity rate of type day 2 is 86.58%, the prediction of ultra-short term typical case's day
Accuracy rate is 95.40%, all meets the power prediction requirement of wind power plant；From the point of view of prediction result, by using improvement K-
Means clustering method and algorithm of support vector machine improve short-term and ultra-short term wind power prediction model, and precision of prediction obtains larger
Raising.
Finally it should be noted that the above embodiments are merely illustrative of the technical solutions of the present invention, rather than the present invention is protected
The limitation of range is protected, although explaining in detail referring to preferred embodiment to the present invention, those skilled in the art are answered
Work as understanding, it can be with modification or equivalent replacement of the technical solution of the present invention are made, without departing from the reality of technical solution of the present invention
Matter and range.
Claims (10)
1. a kind of based on the wind power prediction method for improving K-means cluster and support vector machines, which is characterized in that including following
Step:
Step 1: the wind speed amplitude of each period in acquisition one day, and construct the data sample of wind speed amplitude；
Step 2: data sample is normalized；
Step 3: with K-means clustering method is improved, the data sample in one-year age section is clustered, classification is obtained
Total C, the classification sum C is according to wind speed adaptive change；
Step 4: according to the time scale of setting, the data sample after clustering in step 3 is united by L continuous time period
Meter, obtains L continuous time period sample；
Step 5: after continuous time period sample is normalized according to the method in step 2, according in step 3
Method clusters continuous time period sample, obtains M class data set；
Step 6: corresponding model is trained for every a kind of data set using support vector machines, establishes short-term wind power prediction mould
Type；
Step 7: real-time wind power output data and numerical value data of weather forecast are introduced into short-term wind power prediction model, are obtained
Ultra-short term wind power prediction value introduces real-time wind power, and uses weight to ultra-short term wind power prediction value and real-time wind power
Method be modified, obtain ultra-short term wind power prediction model；
Step 8: a few days ago several to short-term wind power prediction model and ultra-short term wind power prediction mode input prediction day and prediction
It wind power output data and numerical value data of weather forecast, obtain the wind function of short-term wind power prediction and ultra-short term wind power prediction
Rate result.
2. as described in claim 1 based on the wind power prediction method for improving K-means cluster and support vector machines, feature
It is, the data sample constructed in the step 1 is as follows:
A=(as1,as2,…,asH,asmax,asmin,asmean,asstd)
In formula: as1,as2,…,asHFor the value of several node wind speed in one day；asmaxFor day wind speed maximum value；asminFor day wind speed
Minimum value；asmeanFor day wind speed average value；asstdFor the standard deviation of day wind speed.
3. as described in claim 1 based on the wind power prediction method for improving K-means cluster and support vector machines, feature
It is, in the step 2, the formula of the method for normalizing are as follows:
In formula, bijFor the ith sample value in j-th of physical quantity；bjminFor the minimum value in j-th of physical quantity；bjmaxFor jth
Maximum value in a physical quantity.
4. as described in claim 1 based on the wind power prediction method for improving K-means cluster and support vector machines, feature
The step of being, K-means cluster is improved in the step 3 is as follows:
3.1: setting initial category c=1, initialize the cluster centre of all data samples；
3.2: when clustering classification c=C', maximum one kind of Screening Samples capacity in cluster result, the central point clustered
zwIt is split into symmetrical two points zw- ε and zw+ ε, it is c=C'+1 that the classification sum thus clustered, which increases,；
3.3: according to the cluster centre newly divided, data sample being reclassified；
3.4: according to reclassifying as a result, readjusting the position of each cluster centre using the method in step 3.2；
3.5: repeating step 3.3 to 3.4, until each cluster centre no longer moves；
3.6: repeating step 3.2 to 3.5, until all data samples are classified.
5. as claimed in claim 4 based on the wind power prediction method for improving K-means cluster and support vector machines, feature
It is, in the step 3.5, as the central point z of clusterwWhen the distance of two points of division is less than the positive number δ of setting, that is, meet
|zw+ε-(zw- ε) | when the < δ condition of convergence, then assert that each cluster centre no longer moves.
6. as described in claim 1 based on the wind power prediction method for improving K-means cluster and support vector machines, feature
It is, the time scale set in the step 4 is one in three week, half a month and the moon time scales.
7. as described in claim 1 based on the wind power prediction method for improving K-means cluster and support vector machines, feature
It is, the continuous time period sample constructed in the step 4 are as follows:
sl=(st1,st2,…,stC,ssmax,ssmin,ssmean,ssstd,ssmd1,ssmd2,ssmd3)
In formula, st1,st2,…stCFor the quantity on first of continuous time period all kinds of dates；smaxFor in first of continuous time period
Wind speed maximum value；sminFor the wind speed minimum value in first of continuous time period；smeanFor the average wind of first of continuous time period
Speed；ssstdFor the wind speed deviation in first of continuous time period；ssmd1First-order difference for first of continuous time period is absolutely average
Value；ssmd2For the second differnce absolute average of first of continuous time period；ssmd3Third order difference for first of continuous time period is exhausted
To average value.
8. as described in claim 1 based on the wind power prediction method for improving K-means cluster and support vector machines, feature
It is, in the step 6, support vector machines method is by calling SVM-light and feature vector training pattern, objective function
Formula are as follows:
In formula, ξiWithFor relaxation factor；W and θ is band estimation parameter；ε is loss function；β is penalty factor.
9. as claimed in claim 8 based on the wind power prediction method for improving K-means cluster and support vector machines, feature
It is, the physical quantity in described eigenvector includes wind speed, wind direction, temperature, humidity and air pressure.
10. special as described in claim 1 based on the wind power prediction method for improving K-means cluster and support vector machines
Sign is, in the step 7, takes a few days ago several days wind power output data of prediction and numerical value data of weather forecast as prediction ginseng
Examine sample set MCo, it is established to ultra-short term wind power prediction model in conjunction with air speed data ws, and pass through sample attribute AT-1In advance
The wind power plant generated output at T moment is surveyed, calculation method is as follows:
Fo=SVM (MCo,ws)
In formula, FoFor online wind power prediction model, xTFor the predicted value of T moment ultra-short term wind power；
Further according to a few days ago several days practical wind power of prediction, the prediction result of prediction day is corrected using weight, calculation method is such as
Under:
In formula,For revised power prediction value；pD1、pD2、pD3…pDnRespectively predict a few days ago several days practical wind function
Rate sequence；ηTFor weight coefficient shared by predicted value, η1、η2、η3…ηnRespectively predict a few days ago several days weight coefficients.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910437371.9A CN110334847A (en) | 2019-05-24 | 2019-05-24 | Based on the wind power prediction method for improving K-means cluster and support vector machines |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910437371.9A CN110334847A (en) | 2019-05-24 | 2019-05-24 | Based on the wind power prediction method for improving K-means cluster and support vector machines |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110334847A true CN110334847A (en) | 2019-10-15 |
Family
ID=68139376
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910437371.9A Pending CN110334847A (en) | 2019-05-24 | 2019-05-24 | Based on the wind power prediction method for improving K-means cluster and support vector machines |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110334847A (en) |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110991743A (en) * | 2019-12-05 | 2020-04-10 | 河海大学 | Wind power short-term combination prediction method based on cluster analysis and neural network optimization |
| CN111008504A (en) * | 2019-12-18 | 2020-04-14 | 武汉大学 | Wind power prediction error modeling method based on meteorological pattern recognition |
| CN113188570A (en) * | 2021-04-27 | 2021-07-30 | 西南石油大学 | Attitude error calibration method of inclinometer while drilling based on support vector classifier and K-proximity method |
| CN114004276A (en) * | 2021-09-27 | 2022-02-01 | 中国电力工程顾问集团华东电力设计院有限公司 | Time feature processing method and processing device |
| CN117272827A (en) * | 2023-10-13 | 2023-12-22 | 辽宁天成联合科技有限公司 | Ultra-short-term wind speed prediction method of wind measuring system |
| CN117993519A (en) * | 2024-04-03 | 2024-05-07 | 长江三峡集团实业发展(北京)有限公司 | Training and predicting method and device for photovoltaic power station power prediction model |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102298707A (en) * | 2011-08-24 | 2011-12-28 | 辽宁力迅风电控制系统有限公司 | Wind power prediction method based on continuous time slice clustering and support vector machine (SVM) modeling |
| CN103617464A (en) * | 2013-12-10 | 2014-03-05 | 钱胜利 | Device for ultra-short-period predication of wind power and predication method |
| WO2018082132A1 (en) * | 2016-11-04 | 2018-05-11 | 江南大学 | Short-period prediction method for output power of energy system |
| CN108388962A (en) * | 2018-02-06 | 2018-08-10 | 北京天润新能投资有限公司 | A kind of wind power forecasting system and method |
-
2019
- 2019-05-24 CN CN201910437371.9A patent/CN110334847A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102298707A (en) * | 2011-08-24 | 2011-12-28 | 辽宁力迅风电控制系统有限公司 | Wind power prediction method based on continuous time slice clustering and support vector machine (SVM) modeling |
| CN103617464A (en) * | 2013-12-10 | 2014-03-05 | 钱胜利 | Device for ultra-short-period predication of wind power and predication method |
| WO2018082132A1 (en) * | 2016-11-04 | 2018-05-11 | 江南大学 | Short-period prediction method for output power of energy system |
| CN108388962A (en) * | 2018-02-06 | 2018-08-10 | 北京天润新能投资有限公司 | A kind of wind power forecasting system and method |
Non-Patent Citations (1)
| Title |
|---|
| 杨东方: "数学模型在生态学的应用及研究", 西南交通大学出版社 * |
Cited By (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110991743A (en) * | 2019-12-05 | 2020-04-10 | 河海大学 | Wind power short-term combination prediction method based on cluster analysis and neural network optimization |
| CN110991743B (en) * | 2019-12-05 | 2022-08-19 | 河海大学 | Wind power short-term combination prediction method based on cluster analysis and neural network optimization |
| CN111008504A (en) * | 2019-12-18 | 2020-04-14 | 武汉大学 | Wind power prediction error modeling method based on meteorological pattern recognition |
| CN111008504B (en) * | 2019-12-18 | 2022-03-15 | 武汉大学 | Wind power prediction error modeling method based on meteorological pattern recognition |
| CN113188570A (en) * | 2021-04-27 | 2021-07-30 | 西南石油大学 | Attitude error calibration method of inclinometer while drilling based on support vector classifier and K-proximity method |
| CN114004276A (en) * | 2021-09-27 | 2022-02-01 | 中国电力工程顾问集团华东电力设计院有限公司 | Time feature processing method and processing device |
| CN117272827A (en) * | 2023-10-13 | 2023-12-22 | 辽宁天成联合科技有限公司 | Ultra-short-term wind speed prediction method of wind measuring system |
| CN117272827B (en) * | 2023-10-13 | 2024-04-12 | 辽宁天成联合科技有限公司 | Ultra-short-term wind speed prediction method of wind measuring system |
| CN117993519A (en) * | 2024-04-03 | 2024-05-07 | 长江三峡集团实业发展(北京)有限公司 | Training and predicting method and device for photovoltaic power station power prediction model |
| CN117993519B (en) * | 2024-04-03 | 2024-06-11 | 长江三峡集团实业发展(北京)有限公司 | Training and predicting method and device for photovoltaic power station power prediction model |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110334847A (en) | Based on the wind power prediction method for improving K-means cluster and support vector machines | |
| Wang et al. | Deep belief network based k-means cluster approach for short-term wind power forecasting | |
| CN111369070B (en) | Multimode fusion photovoltaic power prediction method based on envelope clustering | |
| CN109002915B (en) | Photovoltaic power station short-term power prediction method based on Kmeans-GRA-Elman model | |
| CN110619360A (en) | Ultra-short-term wind power prediction method considering historical sample similarity | |
| CN111488896B (en) | Distribution line time-varying fault probability calculation method based on multi-source data mining | |
| CN114792156B (en) | Photovoltaic output power prediction method and system based on curve characteristic index clustering | |
| CN102478584B (en) | Wind power station wind speed prediction method based on wavelet analysis and system thereof | |
| CN106503867A (en) | A kind of genetic algorithm least square wind power forecasting method | |
| CN110766200A (en) | Method for predicting generating power of wind turbine generator based on K-means mean clustering | |
| CN105069521A (en) | Photovoltaic power plant output power prediction method based on weighted FCM clustering algorithm | |
| Zhang et al. | Wind speed prediction research considering wind speed ramp and residual distribution | |
| CN111461921B (en) | Load modeling typical user database updating method based on machine learning | |
| CN111915083B (en) | Wind power prediction method and prediction system based on time layered combination | |
| CN111626473A (en) | Two-stage photovoltaic power prediction method considering error correction | |
| CN116523142A (en) | Photovoltaic cluster interval prediction method considering space-time characteristic clustering | |
| CN114298140A (en) | Wind power short-term power prediction correction method considering unit classification | |
| CN110852492A (en) | Photovoltaic power ultra-short-term prediction method for finding similarity based on Mahalanobis distance | |
| Yu et al. | Load forecasting based on smart meter data and gradient boosting decision tree | |
| CN112819189A (en) | Wind power output prediction method based on historical predicted value | |
| CN116702937A (en) | Photovoltaic output day-ahead prediction method based on K-means mean value clustering and BP neural network optimization | |
| CN116454870A (en) | Regional distributed photovoltaic short-term power prediction method | |
| CN115545333A (en) | Method for predicting load curve of multi-load daily-type power distribution network | |
| CN113991711B (en) | Capacity configuration method for energy storage system of photovoltaic power station | |
| CN109978236A (en) | A kind of small power station's short term power fining prediction technique based on feature combination |
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 | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20191015 |