CN107706938B - A kind of wind power waving interval analysis method returned based on quantile - Google Patents
A kind of wind power waving interval analysis method returned based on quantile Download PDFInfo
- Publication number
- CN107706938B CN107706938B CN201710793004.3A CN201710793004A CN107706938B CN 107706938 B CN107706938 B CN 107706938B CN 201710793004 A CN201710793004 A CN 201710793004A CN 107706938 B CN107706938 B CN 107706938B
- Authority
- CN
- China
- Prior art keywords
- wind power
- quantile
- regression
- waving interval
- function
- 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.)
- Expired - Fee Related
Links
- 238000004458 analytical method Methods 0.000 title claims abstract description 45
- 238000000034 method Methods 0.000 claims abstract description 23
- 230000009977 dual effect Effects 0.000 claims abstract description 15
- 238000012706 support-vector machine Methods 0.000 claims abstract description 14
- 230000001351 cycling effect Effects 0.000 claims abstract description 3
- 230000003252 repetitive effect Effects 0.000 claims abstract description 3
- 239000013598 vector Substances 0.000 claims description 13
- 238000013507 mapping Methods 0.000 claims description 10
- 238000012360 testing method Methods 0.000 claims description 10
- 238000000611 regression analysis Methods 0.000 claims description 5
- ONIBWKKTOPOVIA-BYPYZUCNSA-N L-Proline Chemical compound OC(=O)[C@@H]1CCCN1 ONIBWKKTOPOVIA-BYPYZUCNSA-N 0.000 claims description 4
- 238000013461 design Methods 0.000 claims description 3
- 239000011159 matrix material Substances 0.000 claims description 3
- 230000005611 electricity Effects 0.000 abstract description 13
- 230000003044 adaptive effect Effects 0.000 abstract description 2
- 230000006870 function Effects 0.000 description 37
- 238000004422 calculation algorithm Methods 0.000 description 4
- 238000005315 distribution function Methods 0.000 description 3
- 230000000694 effects Effects 0.000 description 3
- 230000002159 abnormal effect Effects 0.000 description 2
- 238000004364 calculation method Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000009434 installation Methods 0.000 description 2
- 238000012417 linear regression Methods 0.000 description 2
- 238000010606 normalization Methods 0.000 description 2
- 238000005457 optimization Methods 0.000 description 2
- 238000012795 verification Methods 0.000 description 2
- 230000007423 decrease Effects 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 235000013399 edible fruits Nutrition 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 238000009499 grossing Methods 0.000 description 1
- 238000007689 inspection Methods 0.000 description 1
- 230000008092 positive effect Effects 0.000 description 1
- 238000010248 power generation Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000007619 statistical method Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Classifications
-
- H02J3/386—
-
- H—ELECTRICITY
- H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
- H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
- H02J2203/00—Indexing scheme relating to details of circuit arrangements for AC mains or AC distribution networks
- H02J2203/20—Simulating, e g planning, reliability check, modelling or computer assisted design [CAD]
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02E—REDUCTION OF GREENHOUSE GAS [GHG] EMISSIONS, RELATED TO ENERGY GENERATION, TRANSMISSION OR DISTRIBUTION
- Y02E10/00—Energy generation through renewable energy sources
- Y02E10/70—Wind energy
- Y02E10/76—Power conversion electric or electronic aspects
Abstract
The invention discloses a kind of wind power waving interval analysis methods returned based on quantile, obtain Power Output for Wind Power Field data first；Determine the quantile of Power Output for Wind Power Field data；Regression function and regression model are established using support vector machines to each quantile；Then it uses prim al- dual interior point m ethod to solve the regression model of each quantile, calculates the quantile of subsequent time；S6:Obtain the measured value of subsequent time wind power data；It finally returns to repetitive cycling and obtains the waving interval of Power Output for Wind Power Field data.Method provided by the invention need not make any assumed condition to Disturbance, you can the adaptive selection that regression function is realized by support vector machines determines quantile regression model.The model is solved using the infeasible prim al- dual interior point m ethod of initial point, realizes the waving interval analysis to future time instance wind power.This method can not only obtain complete wind-powered electricity generation waving interval analysis result, while can also reflect newest situation of change in real time.
Description
Technical field
The present invention relates to technical field of new energy power generation, especially a kind of wind power wave zone returned based on quantile
Between analysis method.
Background technology
Currently, the research about wind-powered electricity generation prediction has focused largely on the prediction of wind power desired value, however, due to wind-powered electricity generation with
Machine rule has relatively strong dispersibility, in this, as prediction result, so that scheduling decision result is not met reality, or even can not
Row.For this reason, it is necessary to increase the analysis to wind power Possible waves section and corresponding probability during prediction (hereinafter referred to as
It is analyzed for waving interval).Because waving interval analysis can provide the probability distribution of future time instance wind power fluctuation, this is advantageous
Preferably recognize risk that is that may be present uncertain in change in future and facing in policymaker, more reasonably determines convenient for making
Plan such as assumes that prediction error obeys multivariate Gaussian distribution and beta distribution respectively, implements the estimation to distributed constant, to obtain
The probability-distribution function for predicting error, has positive effect to scheduling decision, but is actually difficult to seek as assumed consistent distribution
Function.
Therefore, it is necessary to a kind of wind power waving interval analysis methods returned based on quantile.
Invention content
The purpose of the present invention is to propose to a kind of wind power waving interval analysis methods returned based on quantile.
The purpose of the present invention is achieved through the following technical solutions：
The wind power waving interval analysis method provided by the invention returned based on quantile, is included the following steps：
S1:Obtain Power Output for Wind Power Field data；
S2:Determine the quantile of Power Output for Wind Power Field data；
S3:Regression function is established using support vector machines to each quantile；
S4:Regression model is established to each quantile according to regression function；
S5:The regression model that each quantile is solved using prim al- dual interior point m ethod calculates the quantile of subsequent time；
S6:Obtain the measured value of subsequent time wind power data；
S7:It returns to repetitive cycling and obtains the waving interval of Power Output for Wind Power Field data.
Further, the quantile of the Power Output for Wind Power Field data determines according to the following steps：
S21:Obtain Power Output for Wind Power Field data (vt(t), pt(t)), i=1 ... n；
Wherein, vt(t)∈Rd×l, indicate the input vector of (t-d)~(t-1) moment wind power values composition；
pt(t) ∈ R indicate t moment wind power value；
S22:Wind power non-linear relation between t moment and preceding d moment can always be described as following form：
Wherein,Indicate that the Nonlinear Mapping between input/output variable, θ areMiddle undetermined coefficient, b ∈ R are inclined
Shifting amount；
S23:In quantile recurrence, the estimation problem of τ quantiles regression parameter in model can be expressed as
Optimization problem：
S24:If there are θ and b, and following object function to be made to reach minimum, parameter θ and b are referred to as in nonlinear regression model (NLRM)
τ quantile regression coefficients, be denoted as respectivelyWithSubscript τ is to distinguish the regression estimates for different quantiles；Formula
In, ρ () is test function, is expressed as follows：
Wherein, (0,1) τ ∈；The test function ρ it can be seen from definitionτ(x) it is piecewise function, discontinuously may be used at x=0
It leads, and has ρτ(x)≥0。
Further, the regression function determines according to the following steps：
S31:Obtain Power Output for Wind Power Field data (vt(t), pt(t)), i=1 ... n；
S32:The non-linear relation of Power Output for Wind Power Field data is established according to support vector machines theory：
Pt=ωTφ(vt)+b
In formula,For the Nonlinear Mapping of input vector；ω∈Rd×lFor weight vectors；
S33 establishes Function Estimation problem optimal models according to empirical risk minimization, according to following formula：
In formula, θ=[θ1... θn]TFor Lagrange multiplier vector, etFor modeling error, i=1 ... n, γ are that normalization is joined
Number is controlled to the punishment degree beyond error range of operation sample；
S34:Following equation group is obtained according to Kuhn-Tucker optimal conditions：
Eliminate the ω and e in above formulatAnd simplification obtains：
In formula, P=[P1... Pn]T；Ie=[1 ... I]T；I is unit battle array,
S35:Using the inner product of the equivalent mapping space of the gaussian kernel function of the input spaceEstablish following formula：
S36:After being replaced using kernel function and solves above formula equation group and obtain coefficient θ*And b*；
S37:It obtains the measured value of subsequent time wind power data and wind power plant output work will be established according to following formula
Rate data (vt(t), pt(t)) non-linear relation of new input and output amount：
In formula, coefficient θi *Indicate influence degree of i-th group of sample to quantile regression result,
S38:By coefficient θi *Whether it is compared more than preset threshold value ξ, if it is greater than threshold value, is then used as back
Regression function during returning；If it is less than threshold value, then re-circulation is returned.
Further, the regression model prim al- dual interior point m ethod solution procedure of the quantile is as follows：
S41:Power Output for Wind Power Field data are obtained, the quantile estimate model of wind power is established according to following formula：
In formula, ViFor the input variable of regression function；
S42：Pass through slack variable U+And U-Following formula will be established：
s.t.K·θ+U+-U-=p
U+, U-≥0
Wherein, U+, U_∈RnFor the slack variable of introducing；K∈Rn+1For design matrix, KI, j∈k(viv)；
P=[P1... Pn]；θ∈R1；Ie=[1 ... I]T；I is unit battle array；
S43：Enable θ=β1-β2, β1, β1>=0, then establish standard linear programming model according to following formula：
min：cTx
S.t.Ax=p
x≥0
Wherein, A=[K ,-K, I ,-I], I are unit battle array；
S44：Dual form is established according to following formula：
maax pTd
s.t.ATD+s=c
s≥0
S45：Standard linear programming modular form is solved using the infeasible prim al- dual interior point m ethod of initial point, obtains τ recurrence
The estimated value of quantile parameter；
S46：The regression analysis model of subsequent time τ quantiles is worth to according to estimation.
Further, further comprising the steps of：
S71:Obtain the Power Output for Wind Power Field data for determining wind power waving interval；
S72:The output area of Power Output for Wind Power Field data is divided into several section D according to quantile0,…DL；
S73:Obtain the Probability p ro answered in each section0,…,proL；
S74:According to following formula Counting statistics amount χ2：
Wherein, 1/npiWeighted number as sum of square of deviations；N is number of samples；
S75:Judge statistic χ2Whether following condition is met：
If it is satisfied, then indicating that wind power waving interval is consistent with practical waving interval.
By adopting the above-described technical solution, the present invention has the advantage that：
The wind power waving interval analysis method provided by the invention returned based on quantile, by constructing the following wind-powered electricity generation
The new method of output scene, using regression analysis in statistics, quantile returns to realize the waving interval analysis of wind power
Support is provided, quantile recurrence can portray influence of the independent variable to dependent variable distribution characteristics more fully hereinafter, and only need to choose
Suitable regression function form is made any it is assumed that comparing classical least square regression without the Disturbance to model,
With to the more steady effect of abnormal point.How using quantile return to the possible waving interval of future time instance wind power into
Row comprehensively and effectively analyze, more valuable prediction result is provided for dispatching of power netwoks, control decision, be exactly this project to study with
The content of exploration.
This method need not make any assumed condition to Disturbance, you can realize regression function by support vector machines
Adaptive selection, determine quantile regression model.The model is solved using the infeasible prim al- dual interior point m ethod of initial point, it is real
The waving interval analysis to future time instance wind power is showed.This method can not only obtain complete wind-powered electricity generation waving interval analysis
As a result, can also reflect newest situation of change in real time simultaneously.Finally, using this project method to certain wind-powered electricity generation of Ji north area power grid
Field output power is analyzed, the results showed that this project method can portray the uncertain rule of wind power more fully hereinafter
Rule can provide foundation for the decision in the face of risk of scheduling, control.
Other advantages, target and the feature of the present invention will be illustrated in the following description to a certain extent, and
And to a certain extent, based on will be apparent to those skilled in the art to investigating hereafter, Huo Zheke
To be instructed from the practice of the present invention.The target and other advantages of the present invention can by following specification realizing and
It obtains.
Description of the drawings
The description of the drawings of the present invention is as follows.
Fig. 1 is the wind power waving interval analysis method flow chart of the present invention returned based on quantile.
Fig. 2 is wind power waving interval analysis result.
Specific implementation mode
The invention will be further described with reference to the accompanying drawings and examples.
Quantile is introduced first returns basic principle, it is specific as follows：
The concept of quantile：For a certain prediction object, at the moment to be predicted, value can regard a stochastic variable Y as,
Probability-distribution function is：
F (y)=P (Y≤y), then the τ quantiles of F (y) can be expressed as：
Wherein, 0 ＜ τ ＜ 1；
One group of quantile that prediction object can be calculated in future time instance t is returned by quantile
As long as the setting of quantile interval is appropriate, this group of quantile can be fully described by t moment prediction object wave zone
Between probability distribution, can effectively hold the changing condition of unascertained information；Wherein, ymaxThe value range of corresponding stochastic variable Y
The upper limit；
The wind-powered electricity generation waving interval analysis that quantile returns：Current ultra-short term Predicting Technique, such as Kalman filtering, in vacation
If under the premise of Disturbance is white noise, the quadratic sum to minimize residual error carries out sample curve as principle is returned
Fitting obtains the unbiased estimator of regression parameter, and here it is least square method (Least-Square), and on this basis to not
Carry out trend to be deduced, obtains the desired value for predicting object in future time instance.For wind power, due to random law
Diffusive, be actually unsatisfactory for Disturbance be white noise it is assumed that the information content that the prediction result obtained according to this simultaneously provides
It is few, it is difficult to which that complete description wind-powered electricity generation fluctuates situation.
Quantile returns the extension as least square method, using test function as loss function, instead of minimum two
Residual sum of squares (RSS) in multiplication, the weighted sum by minimizing residual absolute value obtain the estimates of parameters of regression function.The party
It is steady to abnormal data in sample that method not only has the advantages that, but also with statistical analysis energy more fully than least square method
Power is more suitable for the analysis of wind power waving interval.
As shown, the wind power waving interval analysis method provided in this embodiment returned based on quantile, specifically
It is as follows：
Quantile returns：
The total n groups of given wind power input and output sample data, (vt(t), pt(t)), i=1 ... n；Wherein, vt(t)∈Rd ×l, indicate the input vector of (t-d)~(t-1) moment wind power values composition, pt(t) ∈ R indicate t moment wind power value.Wind
Electrical power non-linear relation between t moment and preceding d moment can always be described as following form：
Wherein,Indicate that the Nonlinear Mapping between input/output variable, θ areMiddle undetermined coefficient, b ∈ R are offset
Amount.
In quantile recurrence, the estimation problem of τ quantiles regression parameter in model can be expressed as optimizing
Problem：
If there are θ and b, and following object function to be made to reach minimum, parameter θ and b are referred to as the τ in nonlinear regression model (NLRM)
Quantile regression coefficient, is denoted as respectivelyWithSubscript τ is to distinguish the regression estimates for different quantiles.In formula, ρ
() is test function, is expressed as follows：
Wherein, (0,1) τ ∈.The test function ρ it can be seen from definitionτ(x) it is piecewise function, discontinuously may be used at x=0
It leads, and has ρτ(x)≥0。
The selection of regression function：
Wind power waving interval is analyzed using quantile recurrence, the part of most critical is exactly regression function
Selection.Whether regression function is appropriate, determines whether final analysis result matches with true statistical distribution situation.Branch
It holds vector machine and uses structural risk minimization, while controlling empiric risk and learning machine capacity, not only there is preferably reflection
The characteristic of non-linear relation between different periods, it may have good generalization ability.
The present embodiment selects regression function using support vector machines technology, specific as follows：
For given n group input and output sample sets (vt(t), pt(t)), according to support vector machines theory, non-linear relation
It can be expressed as form：
pt=ωTφ(vt)+b
In formula,For the Nonlinear Mapping of input vector；ω∈Rd×1For weight vectors.According to structural risk minimization
Criterion, the Function Estimation problem in formula (3-18) can be expressed as optimal models：
Herein, θ=[θ1... θn]TFor Lagrange multiplier vector, etFor modeling error, i=1 ... n, γ are that normalization is joined
Number is controlled to the punishment degree beyond error range of operation sample.
According to Kuhn-Tucker optimal conditions, following equation group is obtained：
ω and e is eliminated in formulatAnd simplification obtains：
In formula, P=[P1... Pn]T；, Ie=[1 ... I]T；I is unit battle array,In order to keep away
Exempt from bySpace dimensionality after mapping increases, and calculation amount increases, empty using the equivalent mapping of the gaussian kernel function of the input space
Between inner product
After being replaced using kernel function, coefficient θ can be obtained by solving equation group*And b*.For new input and output amount (v, p),
For non-linear relation：
Coefficient θi *Influence degree of i-th group of sample to quantile regression result is reflected, by by θi *With it is preset
Threshold value ξ, which is compared, can decide whether to consider this group of sample in regression process, to obtain suitable regression function shape
Formula
The interior-point algohnhm that quantile returns：
From the foregoing, it will be observed that although regression function is for input variable ViFor be nonlinear function, but for of concern
It is still in a linear relationship for parameter θ and b.Assuming that according to the analysis result of support vector machines, shares l groups sample value and tied to returning
Fruit has obvious effect, then the quantile estimate model of wind power can be expressed as following form：
Linear quantile regression algorithm may be used and calculate regression quantile parameterWith
Linear quantile regression algorithm common at present includes mainly simplex method, interior-point algohnhm and smoothing algorithm etc..Its
In, for interior point method due to having operation efficiency high when handling large data, better numerical value stability is insensitive with calculation scale size
The advantages that, obtain most attention and extensive use.The present embodiment is solved using the infeasible prim al- dual interior point m ethod of initial point to be divided
Site regression model.
Due to test function ρ in quantile estimate modelτ(x)It can discontinuously be led at x=0, if directly using Kuhn-
It is relatively difficult that Tucker optimal conditions solves the model.By introducing slack variable U+And U-It converts optimization problem to：
s.t.K·θ+U+-U_=p
U+, U-≥0；
Wherein, U+,U-∈RnFor the slack variable of introducing；K∈Rn+lFor design matrix, Ki,j∈k(viv)；
P=[P1... Pn]；θ∈Rl；IeMeaning is same as above.
Enable θ=β1-β2, β1, β1>=0, then it can be expressed as standard linear programming model：
min：cT x
S.t.Ax=p
x≥0；
Wherein, A=[K ,-K, I ,-I], I are unit battle array；
Its dual form is expressed as：
max pTd
s.t.ATD+s=c
s≥0
It to standard linear programming modular form, is solved using the infeasible prim al- dual interior point m ethod of initial point, τ recurrence can be obtained
The estimated value of quantile parameter.It will be in estimated value substitution formula, you can obtain the regression analysis model of subsequent time τ quantiles.
The recursion analysis flow of wind power waving interval：
Described in comprehensive preceding two trifle, the detailed process of wind power waving interval recursion analysis is expressed as follows：
1) in t moment, to each quantile (τ1, τ2... τm), using the regression function of support vector machines；
2) regression model is established to each quantile；
3) it uses prim al- dual interior point m ethod to solve each quantile regression model, calculates the quantile at+1 moment of τ
4) at+1 moment of τ, after the measured value for obtaining the moment wind power, new sample set and return to step are formed
1)。
Finally, it obtains wind-powered electricity generation waving interval analysis result and credible verification is carried out to analysis result, it is specific as follows：
After the analysis result for obtaining wind power waving interval, it is important that a problem be discriminatory analysis result with
Whether actual conditions are consistent.For example, during longtime running, if there is about 80% measured value to be located at 0.8 quantile
Under regression curve.Otherwise electricity could be lost as the reliable basis of operation of power networks decision by only meeting actual analysis result
Net decision meaning.
In statistics, usually first assume that overall distribution is distributed for certain class, then examines this by way of sample drawn
Whether assuming that true, problems are known as test of fitness of fot problem.Wherein, for the totality of the Disturbances such as wind power
The case where being distributed not Normal Distribution commonly uses χ2Method is tested.
The output area of wind power is divided by L+1 section D using L quantile0,…DL, according to determining for quantile
Justice can obtain the corresponding Probability p ro in these regions0,…,proL.So when waving interval analysis result is consistent with actual conditions
When, in the n sample sampled to analysis result, it is located at DiMeasured value number n in sectioniIt should be close to ni.proi；
In order to weigh the difference degree of analysis result and actual conditions, frequently with statistic：
The output area of Power Output for Wind Power Field data is divided into several section D according to quantile0,…DL；
Obtain the Probability p ro answered in each section0,…,proL；
When waving interval analysis result is consistent with actual conditions, in the n sample sampled to analysis result,
Wherein, it is to embody the size of absolute value of the bias using square operation；Using 1/npiAs sum of square of deviations
Weighted number be the χ in order at n sufficiently large (generally higher than 50)2Distribution close toDistribution.
As statistic χ2Meet following condition：
I.e. it is believed that analysis result is consistent with actual conditions, there is credibility.In formula (3-29),It is notable
Property level be α, degree of freedom be L chi square distribution.In general, using χ2When method of inspection, the selection of α and L need to artificial experience,
This project takes α=0.05, L=4.
The wind power waving interval analysis method that basis provided in this embodiment is returned based on quantile, and verify the party
The validity of method analyzes Power Output for Wind Power Field by taking the wind power plant of Ji Bei actual motion as an example.The wind power plant is total
Installation 32, total installation of generating capacity 27.2MW, Power Output for Wind Power Field are sent into major network after the boosting of the stations 110kV.
Take 5 minutes average values of the June in 2008 of Power Output for Wind Power Field on the 2nd totally 288 groups of data as example data.Its
In, take historical data window width d=7, sample number n=30, after starting algorithm, remaining 251 groups of data is for verifying this method.By
Fig. 2 can be seen that obtain the waving interval analysis result of 72 periods of wind power.Wherein, dotted line be respectively 0.2,0.4,
0.6 and 0.8 quantile prediction result, solid line represent measured value.
As shown in Figure 2, the possible probability distribution range of wind power is not consistent in each period.When in upward period
Or when higher level, the fluctuation situation of wind-powered electricity generation is more complicated, and corresponding probability distribution range is also bigger, and maximum offset reaches
The 53.85% of measured value；When wind power is in downward period or reduced levels, distribution is much smaller, minimum
Offset is only the 3.27% of measured value, illustrates more stable when wind-powered electricity generation is in the decline stage, is not in significantly to fluctuate,
Distribution compares concentration.
To verify the credibility of waving interval analysis result, regression function and the use of support vector machines determination is respectively adopted
The linear regression function that least square method determines carries out quantile regression analysis, to preceding 100 periods, preceding 150 in analysis result
Distribution situation of the measured value of period, preceding 200 period and whole periods in quantile curve institute demarcation interval makes statistics, and
Carry out credible verification.
According to statistic χ2It is found that the chi square distribution that significance is 0.05, degree of freedom is 4
Statistic χ in each scope of statistics2Respectively less than 9.49, and with the increase χ of scope of statistics2It is gradually reduced, illustrates wave zone
Between analysis result be consistent with actual conditions.The regression function determined using support vector machines, effect will be significantly better than linear regression
Function illustrates that support vector machines can more reasonably describe wind-powered electricity generation Variation Features, therefore, is returned to wind power using quantile
It is feasible to carry out waving interval analysis.For decision of power system dispatching, effectively holding the probability distribution rule of randomness power supply is
Basis and premise.
Finally illustrate, the above examples are only used to illustrate the technical scheme of the present invention and are not limiting, although with reference to compared with
Good embodiment describes the invention in detail, it will be understood by those of ordinary skill in the art that, it can be to the skill of the present invention
Art scheme is modified or replaced equivalently, and without departing from the objective and range of the technical program, should all be covered in the present invention
Protection domain in.
Claims (4)
1. a kind of wind power waving interval analysis method returned based on quantile, it is characterised in that：Include the following steps：
S1:Obtain Power Output for Wind Power Field data；
S2:Determine the quantile of Power Output for Wind Power Field data；
S3:Regression function is established using support vector machines to each quantile；
S4:Regression model is established to each quantile according to regression function；
S5：The regression model that each quantile is solved using prim al- dual interior point m ethod calculates the quantile of subsequent time；
S6:Obtain the measured value of subsequent time wind power data；
S7:It returns to repetitive cycling and obtains the waving interval of Power Output for Wind Power Field data；
The regression model prim al- dual interior point m ethod solution procedure of the quantile is as follows：
S41:Power Output for Wind Power Field data are obtained, the quantile estimate model of wind power is established according to following formula：
In formula, ViFor the input variable of regression function；
S42:Pass through slack variable U+And U-Following formula will be established：
s.t.K·θ+U+-U-=p
U+, U-≥0
Wherein, U+,U-∈RnFor the slack variable of introducing；K∈Rn+lFor design matrix, Ki,j∈k(viv)；
P=[P1... Pn]；θ∈Rl；Ie=[1 ... I]T；I is unit battle array；
S43:Enable θ=β1-β2, β1, β1>=0, then establish standard linear programming model according to following formula：
min：cTx
S.t.Ax=p
x≥0
Wherein, A=[K ,-K, I ,-I], I are unit battle array；
S44:Dual form is established according to following formula：
max pTd
s.t.ATD+s=c
s≥0；
Wherein,
B is offset；
N indicates to give wind power input and output sample data volume；
S45:Standard linear programming modular form is solved using the infeasible prim al- dual interior point m ethod of initial point, τ is obtained and returns point position
The estimated value of point parameter；
S46:The regression analysis model of subsequent time τ quantiles is worth to according to estimation.
2. the wind power waving interval analysis method returned as described in claim 1 based on quantile, it is characterised in that：Institute
The quantile for stating Power Output for Wind Power Field data determines according to the following steps：
S21:Obtain Power Output for Wind Power Field data (vt(t), pt(t)), i=1 ... n；N is number of samples；
Wherein, vt(t)∈Rd×l, indicate the input vector of (t-d)~(t-1) moment wind power values composition；
pt(t) ∈ R indicate t moment wind power value；
S22:Wind power non-linear relation between t moment and preceding d moment can always be described as following form：
Wherein,Indicate that the Nonlinear Mapping between input/output variable, θ areMiddle undetermined coefficient, b ∈ R are offset；
S23:In quantile recurrence, the estimation problem of τ quantiles regression parameter in model can be expressed as optimizing
Problem：
Wherein, ρτ(x) test function is indicated；
S24:If there are θ and b, and following object function to be made to reach minimum, parameter θ and b are referred to as the τ in nonlinear regression model (NLRM)
Quantile regression coefficient, is denoted as respectivelyWithSubscript τ is to distinguish the regression estimates for different quantiles；In formula, ρ
() is test function, is expressed as follows：
Wherein, (0,1) τ ∈；Test function ρτ(x) it is piecewise function, can be discontinuously led at x=0, and have ρτ(x)≥0；
Pi indicates wind power non-linear relation between t moment and preceding d moment.
3. the wind power waving interval analysis method returned as described in claim 1 based on quantile, it is characterised in that：Institute
Regression function is stated to determine according to the following steps：
S31:Obtain Power Output for Wind Power Field data (vt(t), pt(t)), i=1 ... n；
S32:The non-linear relation of Power Output for Wind Power Field data is established according to support vector machines theory：
pi=ωTφ(vi)+b
In formula,For the Nonlinear Mapping of input vector；ω∈Rd×lFor weight vectors；
S33 establishes Function Estimation problem optimal models according to empirical risk minimization, according to following formula：
In formula, θ=[θ1... θn]TFor Lagrange multiplier vector, etFor modeling error, i=1 ... n, γ are regularization parameter, control
System is to the punishment degree beyond error range of operation sample；
S34:Following equation group is obtained according to Kuhn-Tucker optimal conditions：
ω in subtractive and etAnd simplification obtains：
In formula, P=[P1... Pn]T；Ie=[1 ... I]T；I is unit battle array,
S35:Using the inner product of the equivalent mapping space of the gaussian kernel function of the input spaceEstablish following formula：
S36:After being replaced using kernel function and solves equation group and obtain coefficient θ*And b*；
S37:It obtains the measured value of subsequent time wind power data and Power Output for Wind Power Field number will be established according to following formula
According to (vt(t), pt(t)) non-linear relation of new input and output amount：
In formula, coefficient θi *Indicate influence degree of i-th group of sample to quantile regression result,
S38:By coefficient θi *Whether it is compared more than preset threshold value ξ, if it is greater than threshold value, is then used as and returned
Regression function in journey；If it is less than threshold value, then re-circulation is returned.
4. the wind power waving interval analysis method returned as described in claim 1 based on quantile, it is characterised in that：Also
Include the following steps：
S71:Obtain the Power Output for Wind Power Field data for determining wind power waving interval；
S72:The output area of Power Output for Wind Power Field data is divided into several section D according to quantile0,…DL；
S73:Obtain the Probability p ro answered in each section0,…,proL；
S74:According to following formula Counting statistics amount χ2：
Wherein, 1/npiWeighted number as sum of square of deviations；
S75:Judge statistic χ2Whether following condition is met：
If it is satisfied, then indicating that wind power waving interval is consistent with practical waving interval；
Ni indicates to be located at DiMeasured value number in section.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710793004.3A CN107706938B (en) | 2017-09-05 | 2017-09-05 | A kind of wind power waving interval analysis method returned based on quantile |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710793004.3A CN107706938B (en) | 2017-09-05 | 2017-09-05 | A kind of wind power waving interval analysis method returned based on quantile |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN107706938A CN107706938A (en) | 2018-02-16 |
| CN107706938B true CN107706938B (en) | 2018-09-04 |
Family
ID=61172124
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710793004.3A Expired - Fee Related CN107706938B (en) | 2017-09-05 | 2017-09-05 | A kind of wind power waving interval analysis method returned based on quantile |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107706938B (en) |
Families Citing this family (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110555543B (en) * | 2018-05-31 | 2022-10-21 | 中国电力科学研究院有限公司 | Uncertainty analysis method and system for wind power plant prediction |
| CN108921339B (en) * | 2018-06-22 | 2021-08-10 | 南京工程学院 | Quantile regression-based photovoltaic power interval prediction method for genetic support vector machine |
| CN111612255B (en) * | 2020-05-23 | 2022-07-26 | 福州大学 | Wind power plant power curve modeling method based on support vector regression |
Family Cites Families (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101916998A (en) * | 2010-07-12 | 2010-12-15 | 东北电力科学研究院有限公司 | Support vector machine-based wind electric powder prediction device and method |
| CN103440541B (en) * | 2013-09-18 | 2016-03-30 | 山东大学 | The joint probability density Forecasting Methodology of windy electric field short-term output power |
| CN103530527A (en) * | 2013-10-30 | 2014-01-22 | 国家电网公司 | Wind power probability forecasting method based on numerical weather forecasting ensemble forecasting results |
| CN105608514B (en) * | 2016-03-31 | 2020-03-31 | 山东大学 | Overhead transmission line dynamic thermal definite value probability distribution prediction method based on quantile regression |
| CN106295899B (en) * | 2016-08-17 | 2017-08-11 | 合肥工业大学 | Wind power probability density Forecasting Methodology based on genetic algorithm Yu supporting vector quantile estimate |
| CN106778846A (en) * | 2016-12-02 | 2017-05-31 | 华北电力大学 | A kind of method for forecasting based on SVMs |
-
2017
- 2017-09-05 CN CN201710793004.3A patent/CN107706938B/en not_active Expired - Fee Related
Also Published As
| Publication number | Publication date |
|---|---|
| CN107706938A (en) | 2018-02-16 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US11409347B2 (en) | Method, system and storage medium for predicting power load probability density based on deep learning | |
| CN112465664B (en) | AVC intelligent control method based on artificial neural network and deep reinforcement learning | |
| CN113610303B (en) | Load prediction method and system | |
| CN107706938B (en) | A kind of wind power waving interval analysis method returned based on quantile | |
| CN111144663B (en) | Ultra-short-term wind power prediction method for offshore wind farm considering output fluctuation process | |
| CN106251027B (en) | Electric load probability density Forecasting Methodology based on fuzzy support vector quantile estimate | |
| CN104636985A (en) | Method for predicting radio disturbance of electric transmission line by using improved BP (back propagation) neural network | |
| CN106920014A (en) | A kind of short-term load forecasting method and device | |
| CN114362175B (en) | Wind power prediction method and system based on depth certainty strategy gradient algorithm | |
| CN104037761B (en) | AGC power multi-objective random optimization distribution method | |
| CN113471982B (en) | Cloud edge cooperation and power grid privacy protection distributed power supply in-situ voltage control method | |
| CN109840633A (en) | Photovoltaic output power predicting method, system and storage medium | |
| Lv et al. | Novel deterministic and probabilistic combined system based on deep learning and self-improved optimization algorithm for wind speed forecasting | |
| CN108964033A (en) | A kind of electric power system dispatching method, apparatus, computer installation and storage medium | |
| CN113300380A (en) | Load curve segmentation-based power distribution network reactive power optimization compensation method | |
| Gensler et al. | An analog ensemble-based similarity search technique for solar power forecasting | |
| CN115861671A (en) | Double-layer self-adaptive clustering method considering load characteristics and adjustable potential | |
| CN107947206A (en) | Three-phase distribution net Multipurpose Optimal Method based on improved NSGA II algorithms | |
| CN115060625A (en) | Method and device for obtaining slurry density, electronic equipment and storage medium | |
| Vukobratović et al. | Optimal Distributed Generation placement in distribution network | |
| CN108108837A (en) | A kind of area new energy power supply structure optimization Forecasting Methodology and system | |
| CN113298308B (en) | Parameter self-adaptive power consumption prediction method and system | |
| Perera et al. | Supply and demand planning of electricity power: a comprehensive solution | |
| CN110119842A (en) | A kind of micro-capacitance sensor short-term load forecasting method | |
| Basterrech et al. | Optimal design of neural tree for solar power prediction |
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 | ||
| TA01 | Transfer of patent application right | ||
| TA01 | Transfer of patent application right |
Effective date of registration: 20180315 Address after: 100084 Haidian District Tsinghua Yuan Beijing No. 1 Applicant after: TSINGHUA University Applicant after: STATE GRID JIBEI ELECTRIC POWER Co.,Ltd. Applicant after: BEIJING TSINGSOFT TECHNOLOGY CO.,LTD. Address before: 100084 Haidian District Tsinghua Yuan Beijing No. 1 Applicant before: Tsinghua University Applicant before: State Grid Corporation of China Applicant before: STATE GRID JIBEI ELECTRIC POWER Co.,Ltd. Applicant before: BEIJING TSINGSOFT TECHNOLOGY CO.,LTD. |
|
| TA01 | Transfer of patent application right | ||
| TA01 | Transfer of patent application right |
Effective date of registration: 20180727 Address after: 100084 Tsinghua Yuan, Haidian District, Beijing, No. 1 Applicant after: TSINGHUA University Applicant after: STATE GRID CORPORATION OF CHINA Applicant after: STATE GRID JIBEI ELECTRIC POWER Co.,Ltd. Applicant after: BEIJING TSINGSOFT TECHNOLOGY CO.,LTD. Address before: 100084 Tsinghua Yuan, Haidian District, Beijing, No. 1 Applicant before: Tsinghua University Applicant before: STATE GRID JIBEI ELECTRIC POWER Co.,Ltd. Applicant before: BEIJING TSINGSOFT TECHNOLOGY CO.,LTD. |
|
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20180904 Termination date: 20210905 |