CN103117546B - A kind of Ultrashort-term slide prediction method for wind power - Google Patents
A kind of Ultrashort-term slide prediction method for wind power Download PDFInfo
- Publication number
- CN103117546B CN103117546B CN201310064615.6A CN201310064615A CN103117546B CN 103117546 B CN103117546 B CN 103117546B CN 201310064615 A CN201310064615 A CN 201310064615A CN 103117546 B CN103117546 B CN 103117546B
- Authority
- CN
- China
- Prior art keywords
- wind power
- prediction
- asd
- atom
- component
- 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
Landscapes
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The present invention relates to a kind of Ultrashort-term slide prediction method for wind power, adopt and a kind ofly have that very strong non-stationary signal is followed the tracks of, the atom Its Sparse Decomposition method of predictive ability, as the preposition decomposition method of neural net.Be atom component and residual component by wind power Time Series, carry out from prediction to atom component, residual component carries out neural network prediction, then upgrades the result of Atomic Decomposition by adding up-to-date wind power real time data, and then the wind power in slip heavy loads next moment.Verify with actual wind field data, demonstrating this model, can effectively to process wind power non-stationary, produces more sparse discomposing effect, between the Statistical Area that can reduce absolute average error, root-mean-square error calculated value significantly.Therefore, tool of the present invention has the following advantages: can effectively process wind power non-stationary, produces more sparse discomposing effect, between the Statistical Area that can reduce absolute average error, root-mean-square error calculated value significantly.
Description
Technical field
The present invention relates to a kind of super short-period wind power Forecasting Methodology, especially relate to a kind of Ultrashort-term slide prediction method for wind power.
Background technology
The new forms of energy that wind energy is lower as cost in regenerative resource, technology is more ripe, reliability is higher, development in recent years is very fast and start to play a significant role in energy supply.Along with the increase of wind energy turbine set scale, the fluctuation of wind speed and non-stationary become restriction wind-powered electricity generation on a large scale, efficiently grid-connected Tough questions.Wind power prediction technology is one of the key technology solving wind-powered electricity generation fluctuation, wind-electricity integration and dispatching of power netwoks, and this also accurately has higher requirement to the prediction of wind power.
For obtaining higher precision of prediction, lot of domestic and international research concentrates on the suitable forecast model of structure.According to the difference of input variable, existing forecast model can be divided into physical model, statistical model and physical statistics mixed model.Physical model use is if the information such as the technical characteristic (hub height, power curve and thrust coefficient) of meteorology (numerical weather forecast etc.), topology (orographic etc.) and Wind turbines are as mode input amount, object is the best estimate obtaining local wind speed, and then Land use models output statistical method (MOS) reduces prediction residual; Statistical model uses explanatory variable and On-line Measuring Method, usually uses as the recursive technique such as recurrent least square method and artificial neural network method; Mixed model, as optimal models, first obtains the physical quantitys such as the air-flow in Wind turbines region, re-uses advanced statistical model and supplements the information that physical model obtains, thus can obtain more accurate predicted value.Because the physical message etc. around wind energy turbine set in physical method has a significant impact the accuracy predicted the outcome, and statistical method according to the feature of wind energy turbine set self and position, can revise prediction model parameters at any time, can obtain higher accuracy.
Domestic observing and predicting wind power proposes up-to-date technical requirement with predicting, within 2009, State Grid Corporation of China issues clear stipulaties in company standard (Q/GDW392-2009) " wind energy turbine set access electric power network technique regulation detailed rules for the implementation (trying) ", wind power forecasting system should be able to report related data by private network to scheduling institution, at least should possess forecast function a few days ago and ultra-short term forecast function, forward direction scheduling institution when every day 12 declares the next day of wind power prediction curve a few days ago, predict the outcome according to ultra-short term, the wind power prediction curve that rolling adjustment 2 hours is later.The external research mainly concentrating on forecast model, more Zao than domestic starting, technology is relative maturity also.For obtaining the predicted value of 0.5 ~ 36 hour in advance, Denmark University of Science and Technology (DTU) proposes the wind power prediction model of the self adaptation recurrence least square estimation technique of a kind of note and forgetting factor; Carlos the third-largest in Madrid proposes Sipreolico model, and this model is made up of nine self adaptation nonparametric statistics models, uses recursive least squares or Kalman filtering algorithm cycle calculations; TrueWind company proposes a kind of EWIND model, and this model uses the local effect of disposable Parameters design research down wind NWP model output.
In existing wind power prediction modeling method, seldom there is the method for the non-stationary property considering original wind power sequence, neural net is a kind of widely used wind power prediction modeling method, but because the convergence of its adaptive training affects by factors such as step-length, hidden layer neuron number, hidden layer output function and output layer output functions, training time is longer, often can not the non-stationary property of Complete Mappings wind power.Therefore the present invention adopts and a kind ofly has that very strong non-stationary signal is followed the tracks of, New Methods of Signal Processing---the atom Its Sparse Decomposition method of predictive ability, as the preposition decomposition means of neural net.Wind power in reality has very strong non-stationary, and can regard as and have the atom component of multiple different parameters and the superposition of residual component, its non-stationary atomic parameter that causes constantly changes.Carry out from prediction atom component, residual component carries out neural network prediction, is finally predicted the outcome after superposition.
Existing atom sparse resolution theory proposes the concept of wind power signal being carried out to Its Sparse Decomposition, adopt Atomic Decomposition algorithm to carry out slip to wind power to decompose, and replace the residual signals of primary signal as the input variable prediction next moment of neural net with residual signals.Because the energy (relative to primary signal) of residual signals is very little, non-stationary neural network prediction is had an impact of the signal component (linear combination of atom) with Dominant energy greatly can be avoided like this.Therefore, than the neural net prediction method of routine, institute's extraction/prediction method has the ability processing non-stationary property better.
The basic ideas of atom Its Sparse Decomposition are: what take is a kind of adaptive decomposition strategy of greediness, its atom is high redundancy (excessively complete), to ensure that arbitrary signal therefrom can be selected one group of optimum Match atom adaptively and sparsely represent this signal.
(1) structure of wordbook
Atom is represented by general kernel function usually, and in signal transacting field, various kernel function can be used to represent atom, such as SIN function, Chirp function.The kernel function that the present invention adopts is Gaussian function, is shown below:
In formula: g (x) is gaussian kernel function, c and σ is center and scale parameter respectively.Select different centers and scale parameter, a series of different atom can be constructed.The set of these atoms is called wordbook.
(2) Atomic Decomposition of doubledictionary collection
In Breaking Recurrently process, the waiting of each iteration selects atom can be divided into two classes: previously by the old atom selected and the not yet selected new atom selected.Therefore, cross complete wordbook can be divided into two be separated wordbooks, one by old atomic building, one by new atomic building.
In the incipient stage, all atoms all belong to new wordbook, and before decomposable process several times in iteration, major part is belonged to new wordbook by the optimum atom selected, and along with the continuation of iteration, old wordbook slowly increases.When old wordbook is enough large, belonged to old wordbook by the optimum atom major part selected.From the angle of openness, choosing of new atom is unfavorable for the openness of decomposition, for the object reaching Its Sparse Decomposition should choose atom as far as possible from old wordbook.Therefore, propose a kind of openness optimum atom of decomposition that is conducive to and choose flow process, concrete kth step iteration is described below:
According to the decomposition result before kth step, cross complete wordbook and be divided into old and new two wordbooks.Inner product (each moment two product of signals sum) due to two signals describes their linear dependence: the absolute value of inner product is larger, and the correlation of two signals is stronger; Inner product is that zero, two linearly have nothing to do.So, calculate the inner product of each atom in residue signal and two wordbooks respectively, and select inner product maximum in each wordbook: c
oldand c
new, the atom in the old and new wordbook of its correspondence uses Φ respectively
oldand Φ
newrepresent.
If | c
old|>=| c
new|, select Φ
oldas the optimum atom in this iteration, i.e. Φ
opt=Φ
old, c
oldas the iteration coefficient of this optimum atom, i.e. c
opt=c
old.Obviously, the decomposition before k step makes the existing decomposition coefficient of each atom in old wordbook.Therefore, need by c
optadd selected atom Φ to
olddecomposition coefficient on add up.Finally, the residual signals of kth step is upgraded, i.e. R
(k)=R
(k-1)-c
optΦ
opt.
If | c
old| <|c
new|, so the optimum atom of kth step is selected in accordance with the following steps:
1) residual error on old and new wordbook is calculated respectively:
R
old=R
(k-1)-c
oldΦ
old,
R
new=R
(k-1)-c
newΦ
newformula two
2) relative error r is calculated:
In formula: || || represent the European norm of signal.
3) optimum atom is determined by given threshold value T:
If r≤T, select Φ
oldas the optimum atom in this iteration, follow-up computational process and | c
old|>=| c
new| identical in situation;
If r>T, select Φ
newas the optimum atom in this iteration, more new variables, even c
opt=c
new, Φ
opt=Φ
new, R
(k)=R
new.This atom is added in old wordbook, and deletes from new wordbook, coefficient c
optas the decomposition coefficient of this atom.
4) threshold values is upgraded
By given threshold value T, in old and new wordbook, select optimum atom.For ensureing convergence and stability, T is a function successively decreased along with iterative steps, and what the present invention adopted is Annealing function in simulated annealing:
In formula: 0.7≤α <1, T
0represent initial temperature, and setting to be less than 1, k be current iteration step number, N is the annealing speed factor.Along with the increase of iteration, T is tending towards 0.Shown in related algorithm flow process accompanying drawing 1.
But this area not yet has application atom sparse theory to occur to the technical scheme of wind power prediction at present.
Summary of the invention
The present invention mainly solves the technical problem existing for prior art; Providing one, can effectively to process wind power non-stationary, produces more sparse discomposing effect, a kind of Ultrashort-term slide prediction method for wind power between the Statistical Area that can reduce absolute average error, root-mean-square error calculated value significantly.
Above-mentioned technical problem of the present invention is mainly solved by following technical proposals:
A kind of Ultrashort-term slide prediction method for wind power, is characterized in that: comprise the following steps,
Step 1, adopts and carries out data prediction to wind power primary signal, and namely first in whole data area, determining that maximum and minimum value carry out unified normalization conversion process again, is the interval value of zero to by mode input output transform; Concrete normalization formula is as follows:
In formula: x be model input or output component;
for inputing or outputing component after normalized; x
maxand x
minbe respectively maximum and the minimum value of mode input or output variable;
Step 2, based on the data sample after the process of step 1 gained, adopt atom Its Sparse Decomposition method, be atom component and residual component by wind power Time Series, carry out from prediction to atom component, residual component carries out neural network prediction, then upgrades the result of Atomic Decomposition by adding up-to-date wind power real time data, and then the wind power in slip heavy loads next moment; Adopt linear regression method to correct again, obtain the wind power in next moment, repeat this step until reach the prediction total time of user's setting, namely obtain the final predicted value of wind power of the prediction total time of user's setting;
Step 3, based on the data sample after the process of step 1 gained and step 2 gained wind power end value, with normalization absolute average error and normalization root-mean-square error for foundation, the conventional normal distribution approximating method of statistics is adopted to carry out quantitative assessment to prediction effect.
In above-mentioned Ultrashort-term slide prediction method for wind power, described step 2 comprises following sub-step, and definition current time is t, and subsequent time is t+1;
Step 2.1, atom Its Sparse Decomposition also carries out from prediction:
Step 2.11, carries out n Atomic Decomposition to wind power data, and the time period of self-defined Atomic Decomposition is t-m to t, and wherein, t, m are positive integer:
In formula: r (t) is residual component; a
jt () is a jth atom component, equal the product of atom and its decomposition coefficient;
Step 2.12, residual signals carries out the residual prediction of subsequent time as the input variable in the t+1 moment of step 2;
Step 2.13, predicts the atom component value in t+1 moment certainly according to the expression formula of atom component;
Step 2.2, neural net carries out residual prediction:
According to step 2.1 gained residual component value, the input variable as neural net residual error prediction method carries out the residual prediction in t+1 moment;
Step 2.3, carries out slip heavy loads: predict the outcome neural net residual prediction result and atom Its Sparse Decomposition wind power prediction value certainly that superpose and namely obtain subsequent time; The time window adopting time scale to determine carries out the prediction in t+1 moment, after obtaining predicted value, time window slides and pushes ahead a moment, namely Atomic Decomposition time period be t-m+1 to t+1, repeated execution of steps 2.1 to step 2.3 until reach user setting prediction total time after terminate perform step 2.4;
Step 2.4, predict the outcome correction, and namely adopt linear regression method to correct, calibration model is as follows:
In formula: P
aSD, tthe predicted value of the wind power prediction t obtained for adopting the inventive method;
for the t predicted value after correction; e
aSD, t=a+bP
aSD, t, a and b is parameter, and adopt least square method to calculate, estimated by history wind power and error sample data thereof, method is as follows:
In formula: N
cfor sample size; e
aSD, i=P
aSD, i-P
meas, ifor history wind power prediction error; P
meas, ifor wind energy turbine set actual measurement wind power data;
The predicted value obtained in step 2.3 is carried out the correction that predicts the outcome, obtains final wind power prediction value.
In above-mentioned Ultrashort-term slide prediction method for wind power, described step 3 comprises following sub-step,
Step 3.1, based on the final wind power prediction value of step 2 gained, adopts normalization absolute average error e general in the world
nMAEwith normalization root-mean-square error e
nRMSEfor foundation, be defined as follows:
In formula: x (i) is actual value;
for predicted value; N is forecast sample number; P
cap.for the rated capacity of blower fan;
Step 3.2, based on the final wind power prediction value of step 2 gained, adopt probabilistic method to analyze, the normal distribution approximating method that namely statistics is conventional carries out quantitative assessment to prediction effect.
Therefore, tool of the present invention has the following advantages: can effectively process wind power non-stationary, produces more sparse discomposing effect, between the Statistical Area that can reduce absolute average error, root-mean-square error calculated value significantly.
Accompanying drawing explanation
Fig. 1 is the algorithm flow chart of atom sparse resolution theory of the present invention.
Fig. 2 is neural net residual error prediction method network configuration of the present invention.
Fig. 3 is 3 not homoatomic comparisons of the present invention.
Embodiment
Below by embodiment, and by reference to the accompanying drawings, technical scheme of the present invention is described in further detail.
Embodiment:
The present invention relates to a kind of Ultrashort-term slide prediction method for wind power.Because wind power sequence has stronger non-stationary property, neural net can not its characteristic of Complete Mappings, and the present invention adopts a kind ofly has that very strong non-stationary signal is followed the tracks of, the atom Its Sparse Decomposition method of predictive ability, as the preposition decomposition method of neural net.Be atom component and residual component by wind power Time Series, carry out from prediction to atom component, residual component carries out neural network prediction, then upgrades the result of Atomic Decomposition by adding up-to-date wind power real time data, and then the wind power in slip heavy loads next moment.Verify with actual wind field data, demonstrating this model, can effectively to process wind power non-stationary, produces more sparse discomposing effect, between the Statistical Area that can reduce absolute average error, root-mean-square error calculated value significantly.
Choose January 1 calendar year 2001 in 23 days June in 2008, abroad certain blower fan power output data is as sample, describes technical solution of the present invention in detail by reference to the accompanying drawings with case study on implementation.
The Ultrashort-term slide prediction method for wind power that case study on implementation provides can adopt computer software programs to realize automatic operational process.Contained by the flow process of case study on implementation, step is as follows:
Step 1, when carrying out model prediction, when each component dimension inputing or outputing vector is different or size differs greatly, tackles different components and being normalized respectively in its span.Adopt and carry out data prediction to wind power primary signal, namely first in whole data area, determining that maximum and minimum value carry out unified normalization conversion process again, is the interval value of zero to by mode input output transform; Concrete normalization formula is as follows:
In formula: x be model input or output component;
for inputing or outputing component after normalized; x
maxand x
minbe respectively maximum and the minimum value of mode input or output variable.
Step 2, based on the data sample after the process of step 1 gained, adopt and a kind ofly have that very strong non-stationary signal is followed the tracks of, the atom Its Sparse Decomposition method of predictive ability, be atom component and residual component by wind power Time Series, carry out from prediction to atom component, residual component carries out neural network prediction, then upgrades the result of Atomic Decomposition by adding up-to-date wind power real time data, and then the wind power in slip heavy loads next moment; Adopt linear regression method to correct again, obtain the final predicted value of wind power in next moment;
And step 2 comprises following sub-step,
Step 2.1, atom Its Sparse Decomposition also carries out from prediction:
1) n Atomic Decomposition is carried out to wind power data:
In formula: r (t) is residual component; a
jt () is a jth atom component, equal the product of atom and its decomposition coefficient.
2) residual signals carries out the residual prediction of subsequent time as the input variable of step 2;
3) the atom component value of subsequent time is certainly predicted according to the expression formula of atom component;
In atom Its Sparse Decomposition, atom is represented by general kernel function usually, and in signal transacting field, various kernel function can be used to represent atom, such as SIN function, Chirp function.The kernel function that the present invention adopts is Gaussian function, is shown below:
In formula: g (x) is gaussian kernel function, c and σ is center and scale parameter respectively.Select different centers and scale parameter, a series of different atom can be constructed.The set of these atoms is called wordbook.Three different atoms are listed in accompanying drawing 1.Wherein, g1 represents that center is 0, and yardstick is the atom of 2; G2 represents that center is 2, and yardstick is the atom of 2; G3 represents that center is 0, and yardstick is the atom of 3.
Step 2.2, neural net carries out residual prediction: according to step 2.1 gained residual component value, the input variable as neural net residual error prediction method carries out the residual prediction of subsequent time, shown in dependency structure accompanying drawing 2;
Step 2.3, carry out slip heavy loads: generally speaking, a wind power sequence determined after the preposition decomposition of forecast model, must obtain that one group of parameter is determined, stably and have the atom component of Dominant energy and non-stationary, the strong but residual component that energy is little of randomness.Because atom component occupies leading role, this method is used as wind power data stationary sequence to process in essence still unilaterally.Therefore, the present invention proposes a kind of slip heavy loads method, sets up 50 best forecast models by step 2.1 and step 2.2.Up-to-date input variable and corresponding different models are adopted to carry out slip heavy loads to next 15min wind power.The input variable that each model is corresponding is different: model i (M
i) using 400 wind powers before i point as input variable, to i point prediction; M
i+1utilize 399 wind powers before the measured value of i point and i point as input variable, to i+1 point slip heavy loads ..., the like.
The time window adopting time scale to determine carries out the prediction of subsequent time, and after obtaining predicted value, time window slides and pushes ahead a moment, continues similar prediction.The advantage of the method is, although what the decomposition in time window obtained is stable atom component, but along with time window constantly forward slip, atom component parameters between time window and time window is constantly change, when training sample is enough large, time window number is a lot, can be considered and has carried out non-stationary process to wind power data; Meanwhile, along with time window slide, atom component regulates self parameter adaptively, to adapt to the wind power data of non-stationary, significantly enhances the generalization ability of forecast model.
Step 2.4, predict the outcome correction:
Adopt linear regression method to correct, calibration model is as follows:
In formula: P
aSD, tthe predicted value of the wind power prediction t obtained for adopting the inventive method;
for the t predicted value after correction; e
aSD, t=a+bP
aSD, t, a and b is parameter, and least square method can be adopted to calculate, and estimated by history wind power and error sample data thereof, method is as follows:
In formula: N
cfor sample size; e
aSD, i=P
aSD, i-P
meas, ifor history wind power prediction error; P
meas, ifor wind energy turbine set actual measurement wind power data.
And step 3 comprises following sub-step,
Step 3.1, adopts normalization absolute average error e general in the world
nMAEwith normalization root-mean-square error e
nRMSEfor foundation, be defined as follows:
In formula: x (i) is actual value;
for predicted value; N is forecast sample number; P
cap.for the rated capacity of blower fan.
Step 3.2, adopts probabilistic method to analyze, and the normal distribution approximating method that namely statistics is conventional carries out quantitative assessment to prediction effect.
Specific embodiment described in the present invention is only to the explanation for example of the present invention's spirit.Those skilled in the art can make various amendment or supplement or adopt similar mode to substitute to described specific embodiment, but can't depart from spirit of the present invention or surmount the scope that appended claims defines.
Claims (2)
1. a Ultrashort-term slide prediction method for wind power, is characterized in that: comprise the following steps,
Step 1, adopts and carries out data prediction to wind power primary signal, and namely first in whole data area, determining that maximum and minimum value carry out unified normalization conversion process again, is the interval value of zero to by mode input output transform; Concrete normalization formula is as follows:
In formula: x be model input or output component;
for inputing or outputing component after normalized; x
maxand x
minbe respectively maximum and the minimum value of mode input or output variable;
Step 2, based on the data sample after the process of step 1 gained, adopt atom Its Sparse Decomposition method, be atom component and residual component by wind power Time Series, carry out from prediction to atom component, residual component carries out neural network prediction, then upgrades the result of Atomic Decomposition by adding up-to-date wind power real time data, and then the wind power in slip heavy loads next moment; Adopt linear regression method to correct again, obtain the wind power in next moment, repeat this step until reach the prediction total time of user's setting, namely obtain the final predicted value of wind power of the prediction total time of user's setting;
Step 3, based on the data sample after the process of step 1 gained and the final predicted value of step 2 gained wind power, with normalization absolute average error and normalization root-mean-square error for foundation, the conventional normal distribution approximating method of statistics is adopted to carry out quantitative assessment to prediction effect;
Described step 2 comprises following sub-step, and definition current time is t, and subsequent time is t+1;
Step 2.1, atom Its Sparse Decomposition also carries out from prediction:
Step 2.11, carries out n Atomic Decomposition to wind power data, and the time period of self-defined Atomic Decomposition is t-m to t, and wherein, t, m are positive integer:
In formula: r (t) is residual component; a
jt () is a jth atom component, equal the product of atom and its decomposition coefficient;
Step 2.12, residual component carries out the residual prediction of subsequent time as the input variable in the t+1 moment of step 2;
Step 2.13, predicts the atom component value in t+1 moment certainly according to the expression formula of atom component;
Step 2.2, neural net carries out residual prediction:
According to step 2.1 gained residual component value, the input variable as neural net residual error prediction method carries out the residual prediction in t+1 moment;
Step 2.3, carries out slip heavy loads: predict the outcome neural net residual prediction result and atom Its Sparse Decomposition wind power prediction value certainly that superpose and namely obtain subsequent time; The time window adopting time scale to determine carries out the prediction in t+1 moment, after obtaining predicted value, time window slides and pushes ahead a moment, namely Atomic Decomposition time period be t-m+1 to t+1, repeated execution of steps 2.1 to step 2.3 until reach user setting prediction total time after perform step 2.4;
Step 2.4, predict the outcome correction, and namely adopt linear regression method to correct, calibration model is as follows:
In formula: P
aSD, tthe predicted value of the wind power prediction t obtained for adopting the inventive method;
for the t predicted value after correction; e
aSD, t=a+bP
aSD, t, a and b is parameter, and adopt least square method to calculate, estimated by history wind power and error sample data thereof, method is as follows:
In formula: N
cfor sample size; e
aSD, i=P
aSD, i-P
meas, ifor history wind power prediction error; P
meas, ifor wind energy turbine set actual measurement wind power data;
The predicted value obtained in step 2.3 is carried out the correction that predicts the outcome, obtains final wind power prediction value.
2. Ultrashort-term slide prediction method for wind power according to claim 1, is characterized in that: described step 3 comprises following sub-step,
Step 3.1, based on the final wind power prediction value of step 2 gained, adopts normalization absolute average error e general in the world
nMAEwith normalization root-mean-square error e
nRMSEfor foundation, be defined as follows:
In formula: x (i) is actual value;
for predicted value; N is forecast sample number; P
cap.for the rated capacity of blower fan;
Step 3.2, based on the final wind power prediction value of step 2 gained, adopt probabilistic method to analyze, the normal distribution approximating method that namely statistics is conventional carries out quantitative assessment to prediction effect.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201310064615.6A CN103117546B (en) | 2013-02-28 | 2013-02-28 | A kind of Ultrashort-term slide prediction method for wind power |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201310064615.6A CN103117546B (en) | 2013-02-28 | 2013-02-28 | A kind of Ultrashort-term slide prediction method for wind power |
Publications (2)
Publication Number | Publication Date |
---|---|
CN103117546A CN103117546A (en) | 2013-05-22 |
CN103117546B true CN103117546B (en) | 2016-03-16 |
Family
ID=48415852
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201310064615.6A Expired - Fee Related CN103117546B (en) | 2013-02-28 | 2013-02-28 | A kind of Ultrashort-term slide prediction method for wind power |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN103117546B (en) |
Families Citing this family (19)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103425867A (en) * | 2013-06-24 | 2013-12-04 | 武汉大学 | Short-term wind power combination prediction method |
CN103473322A (en) * | 2013-09-13 | 2013-12-25 | 国家电网公司 | Photovoltaic generation power ultra-short term prediction method based on time series model |
CN103489041A (en) * | 2013-09-17 | 2014-01-01 | 国家电网公司 | Short-term wind power forecast method |
CN103617464B (en) * | 2013-12-10 | 2016-08-17 | 北方大贤风电科技(北京)有限公司 | The device predicted for the ultra-short term of wind power and Forecasting Methodology |
CN104376388A (en) * | 2014-12-08 | 2015-02-25 | 国家电网公司 | Ultra-short period wind power prediction method based on wind speed factor control model |
CN104571262B (en) * | 2015-01-16 | 2017-10-24 | 江南大学 | Short-term wind-electricity power interval probability Forecasting Methodology |
CN106709586B (en) * | 2015-07-15 | 2021-07-20 | 华北电力大学(保定) | Method and system for predicting short-term wind speed of wind power plant |
CN106203723A (en) * | 2016-07-19 | 2016-12-07 | 河海大学 | Wind power short-term interval prediction method based on RT reconstruct EEMD RVM built-up pattern |
CN106682760A (en) * | 2016-09-13 | 2017-05-17 | 上海电机学院 | Wind power climbing prediction method |
CN106712112B (en) * | 2017-02-17 | 2019-11-01 | 云南电网有限责任公司 | A kind of analysis method of the wind power plant cluster output smoothing effect based on probability distribution |
CN107798426A (en) * | 2017-10-16 | 2018-03-13 | 武汉大学 | Wind power interval Forecasting Methodology based on Atomic Decomposition and interactive fuzzy satisfying method |
CN109840613B (en) * | 2018-10-25 | 2021-02-23 | 浙江理工大学 | Short-term wind speed prediction method integrating coding and decoding and linear regression |
CN110175639B (en) * | 2019-05-17 | 2021-06-11 | 华北电力大学 | Short-term wind power prediction method based on feature selection |
CN110327043B (en) * | 2019-08-05 | 2022-04-01 | 哈尔滨工业大学 | Event-related potential waveform map solving method based on sparse modeling |
CN110826791A (en) * | 2019-10-31 | 2020-02-21 | 东北电力大学 | Hybrid wind power prediction method based on long-time and short-time memory neural network |
CN112036658B (en) * | 2020-09-09 | 2022-06-14 | 华东交通大学 | Wind power real-time prediction method based on forgetting long-term and short-term memory with difference |
CN114501010B (en) * | 2020-10-28 | 2023-06-06 | Oppo广东移动通信有限公司 | Image encoding method, image decoding method and related devices |
CN113193556B (en) * | 2021-05-12 | 2023-02-03 | 燕山大学 | Short-term wind power prediction method based on probability prediction model |
CN113723752A (en) * | 2021-07-26 | 2021-11-30 | 国网江苏省电力有限公司江阴市供电分公司 | Decomposition algorithm performance evaluation method in combined wind power prediction model |
Citations (2)
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 |
CN102184453A (en) * | 2011-05-16 | 2011-09-14 | 上海电气集团股份有限公司 | Wind power combination predicting method based on fuzzy neural network and support vector machine |
Family Cites Families (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2004289918A (en) * | 2003-03-20 | 2004-10-14 | Fujitsu Ltd | Power supply method |
-
2013
- 2013-02-28 CN CN201310064615.6A patent/CN103117546B/en not_active Expired - Fee Related
Patent Citations (2)
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 |
CN102184453A (en) * | 2011-05-16 | 2011-09-14 | 上海电气集团股份有限公司 | Wind power combination predicting method based on fuzzy neural network and support vector machine |
Non-Patent Citations (1)
Title |
---|
原子稀疏分解算法在电力系统扰动信号分析中的应用;贾清泉等;《电力系统保护与控制》;20101001;第38卷(第19期);第17-21页 * |
Also Published As
Publication number | Publication date |
---|---|
CN103117546A (en) | 2013-05-22 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN103117546B (en) | A kind of Ultrashort-term slide prediction method for wind power | |
Huang et al. | Multiple-input deep convolutional neural network model for short-term photovoltaic power forecasting | |
Miranda et al. | One-hour-ahead wind speed prediction using a Bayesian methodology | |
Hu et al. | New hybrid approach for short-term wind speed predictions based on preprocessing algorithm and optimization theory | |
CN108717579B (en) | Short-term wind power interval prediction method | |
CN102479347A (en) | Method and system for forecasting short-term wind speed of wind farm based on data driving | |
Yongsheng et al. | A Short‐Term Power Output Forecasting Model Based on Correlation Analysis and ELM‐LSTM for Distributed PV System | |
CN104036328A (en) | Self-adaptive wind power prediction system and prediction method | |
Zhang et al. | Predictive analytics for comprehensive energy systems state estimation | |
CN114091317A (en) | Photovoltaic power station power prediction method based on NWP irradiance correction and error prediction | |
Yang et al. | Short‐term photovoltaic power dynamic weighted combination forecasting based on least squares method | |
CN113991711A (en) | Capacity configuration method for energy storage system of photovoltaic power station | |
Alharbi et al. | Short-term wind speed and temperature forecasting model based on gated recurrent unit neural networks | |
Hu et al. | Adaptive corrected parameters algorithm applied in cooling load prediction based on black-box model: A case study for subway station | |
CN111967660B (en) | Ultra-short-term photovoltaic prediction residual error correction method based on SVR | |
CN116581755B (en) | Power prediction method, device, equipment and storage medium | |
Zhou et al. | A Hybrid Method for Ultrashort‐Term Wind Power Prediction considering Meteorological Features and Seasonal Information | |
CN113111592A (en) | Short-term wind power prediction method based on EMD-LSTM | |
Zhang | Short‐Term Power Load Forecasting Based on SAPSO‐CNN‐LSTM Model considering Autocorrelated Errors | |
CN112734073A (en) | Photovoltaic power generation short-term prediction method based on long and short-term memory network | |
Ehsan et al. | Artificial neural network predictor for grid-connected solar photovoltaic installations at atmospheric temperature | |
Nguyen et al. | A machine learning-based approach for the prediction of electricity consumption | |
Sarathkumar et al. | Renewable energy resources forecasting model for virtual power plant in the deregulated electricity market using machine learning | |
Liu et al. | A Self-Data-Driven Method for Lifetime Prediction of PV Arrays Considering the Uncertainty and Volatility | |
Di et al. | Ultra-short term load forecasting based on elman neural network |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
C14 | Grant of patent or utility model | ||
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: 20160316 Termination date: 20170228 |