CN105320809B - A kind of wind speed forecasting method for wind power plant spatial coherence - Google Patents
A kind of wind speed forecasting method for wind power plant spatial coherence Download PDFInfo
- Publication number
- CN105320809B CN105320809B CN201510640990.XA CN201510640990A CN105320809B CN 105320809 B CN105320809 B CN 105320809B CN 201510640990 A CN201510640990 A CN 201510640990A CN 105320809 B CN105320809 B CN 105320809B
- Authority
- CN
- China
- Prior art keywords
- data
- wind
- prediction
- wind power
- kalman filtering
- 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.)
- Active
Links
Classifications
-
- 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
- Y02E60/00—Enabling technologies; Technologies with a potential or indirect contribution to GHG emissions mitigation
Landscapes
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a kind of prediction techniques for wind farm wind velocity.It is main to consider the problems such as not having to consider well the optimization of spatial coherence and Unscented kalman filtering between each wind power plant in existing method.Main method is for giving 22 wind power plants, calculate the rank correlation coefficient of remaining 21 wind power plants and target wind farm, size according to related coefficient goes judgement for the wind power plant of prediction, and Kendall rank correlation coefficient is selected to be greater than the wind power plant that 0.55 and Spearman rank correlation coefficient is greater than 0.75.It goes to establish Nonlinear state space model using Support vector regression later, carries out Unscented kalman filtering prediction using established Nonlinear state space model.It goes to optimize according to prediction error minimum principle for the scale parameter of Unscented kalman filtering.The wind farm wind velocity data for finally selecting synchronization in 4 years carry out grey relational grade analysis with the air speed data of the First Year synchronization of itself and No. 9 machines of target fan.
Description
Technical field
Predicting wind speed of wind farm field of the present invention, especially in the more intensive area of wind-powered electricity generation field distribution to a certain wind power plant
Wind speed predicted, for solve to ignore during forecasting wind speed wind field position and caused by the insufficient problem of precision of prediction.
Background technique
Wind-power electricity generation rapidly develops in recent years because having many advantages, such as environmental protection, renewable, has become whole world public affairs
The ideal energy recognized.However after wind-powered electricity generation penetrates power and reaches certain value, randomness, fluctuation and the unstability of wind-powered electricity generation will
The operation of electric system is had a huge impact.Related scholar has carried out many correlative studys to this aspect, has obtained one
The conclusion being just of practical significance a bit.And to reduce the influence with regard to it is main seek to make comparison to wind speed or wind power it is accurate
Prediction.
The most important determinant of wind turbine power generation power is wind speed, so can be used as prediction for being effectively predicted for wind speed
The key component of wind power.Currently, being generally possible to be divided into two classes both at home and abroad about the method for forecasting wind speed: one kind is to utilize to go through
The statistical method of history data modeling, another kind of is to utilize numerical weather forecast and topographic physical method.The former includes the time
Sequence hair, Kalman filtering method, neural network etc..Physical method generally comprise the prediction model using numerical weather forecast,
Spatial coherence method etc..But a kind of wind the characteristics of having oneself as natural phenomena, existing method cannot well to wind speed into
Row prediction.For the prediction of wind speed, not only to consider single wind series, need wind power plant historical wind speed sequence and in real time
Data, it is necessary to consider geographical location, roughness, wind direction, air pressure, the physical factors such as temperature and spatial position.
Therefore a good forecasting wind speed model, will not only consider the temporal correlation between wind speed, it is also contemplated that ambient wind
The spatial coherence of the wind speed of electric field could improve precision of prediction in this way, improve prediction model.
In document Short-term wind prediction using an unscented Kalman filter
In based state-space support vector regression approach, author gives a kind of based on non-thread
Property the electrodeless Kalman filtering of state-space model wind speed forecasting method, in order to establish the state equation of state-space model, text
In use Support vector regression method, this method can predict wind speed in real time, and can reduce wind
The influence of fast randomness, still, this method only considered single wind speed time series, not have for surrounding wind field wind speed
There is research, is not a kind of complete prediction technique.
In document Application of artificial neural networks for the wind speed
In prediction of station using reference stations data, author, which proposes, utilizes neural network
Model predicts wind speed, the selection this process employs the air speed data of Wind Field as input data, it is contemplated that empty
Between correlation, but do not referred to for how the power of such as correlation goes to differentiate in text, and be not a kind of prediction side in real time
Method.
In document A hybrid statistical method to predictwind speed and wind power
In, author proposes a kind of hybrid prediction model, and in order to reduce influence of the fluctuations in wind speed to prediction result, author is proposed to original
Beginning data carry out wavelet decomposition, and sub-sequences carry out time series forecasting, finally carry out again to the prediction result of each subsequence
Merge.This method does not account for its spatial coherence still.
In document Wind power prediction based on numerical and statistical model
In, this method is predicted that wind speed, author considers meteorology according to numerical weather forecast and Kalman prediction model
Influence of the parameter to wind speed, and wind speed is predicted in real time using Kalman filter model, however due to this state space
Model is linear, so the actual conditions of wind speed cannot be portrayed well.
During forecasting wind speed, time and spatial coherence between wind speed are and deposit, and not with distance
It is same and different, in addition to document Application ofartificial neural networksfor in above-mentioned document
Thewind speed prediction ofstation using reference stations data considers ambient wind
The correlation of wind speed between, remaining does not all consider the spatial coherence between wind speed, certainly exists to precision of prediction larger
Influence, so prediction model can be further perfect.
Summary of the invention
It is an object of the invention to not consider the deficiency of its spatial coherence for most of forecasting wind speed model at present, mention
A kind of wind speed hybrid prediction model out, the model are primarily based on copula function to compare spatial coherence between its each wind field
Power establishes Nonlinear state space model followed by support vector machines, finally utilizes the Unscented kalman filtering of optimization
Wind speed is predicted.Real-time prediction of the model realization to wind speed, and consider the temporal correlation and space phase of wind speed
Guan Xing.Before introducing the present invention, 2 definition are first provided.
Calculating the general index of the spatial coherence between two wind fields using copula function is kendall rank correlation system
Several and Spearman rank correlation coefficient.We provide the definition and calculation method of the two rank correlation coefficients below.
It defines 1 (Kendall rank correlation coefficient) and enables (x1,y1) and (x2,y2) it is independent identically distributed stochastic variable, definition
τ≡P[(x1-x2)(y1-y2) > 0]-P [(x1-x2)(y1-y2) < 0]
=2P [(x1-x2)(y1-y2) > 0] -1
For kendall rank correlation coefficient, it is denoted as τ.
If the edge distribution of stochastic variable X, Y are respectively F (x), G (y), corresponding copula function is C (u, v), wherein u
=F (x), v=G (y), u, v ∈ [0,1], then kendall rank correlation coefficient τ can have corresponding copula function C (u, v) to provide
It defines 2 (Spearman rank correlation coefficients) and enables (x1,y1),(x2,y2) and (x3,y3) it is independent identically distributed random change
Amount, definition
ρ≡3{P[(x1-x2)(y1-y3) > 0]-P [(x1-x2)(y1-y3) < 0]
For Spearman rank correlation coefficient, it is denoted as ρ.
If the edge distribution of stochastic variable X, Y are respectively F (x), G (y), corresponding copula function is C (u, v), wherein u
=F (x), v=G (y), u, v ∈ [0,1], then kendall rank correlation coefficient τ can have corresponding copula function C (u, v) to provide
Note: when calculating the two rank correlation coefficients, the copula function that the present invention uses is distributed for normal distyribution function and T
Function.Data set used in the present invention is from public data collection Wisconsin, USA (Wisconsin state, WI) 22
The air speed data of wind power plant (1 day to 2009 January in 2006, on December 31, primary every 1 hour record).
The step of when with rank correlation coefficient selection input data, is as follows:
1: on December 30,24 days to 2009 December in 2009 is selected from data set totally 7 days data.
2: the validity of detection data is filled missing data, and the present invention was replaced using 1 day around it average value
Missing data.
3: test of normality being carried out to the data detected, then seeks its empirical distribution function, its initial data is carried out
Kernel distribution estimator can carry out parameter Estimation to copula function later.
4: its corresponding kendall rank correlation system is calculated according to calculated Norm copula and t copula function
Several and Spearman rank correlation coefficient.
5: being chosen according to the size of calculated kendall rank correlation coefficient and Spearman rank correlation coefficient suitable defeated
Enter data.
The rank correlation coefficient table that we calculate according to the above method is shown in Table 1.
Kendall_norm | Spearman_norm | Kendallt | Spearman_t | Kendall | Spearman | |
0.5031 | 0.6937 | 0.5530 | 0.7480 | 0.5901 | 0.7566 | |
0.5548 | 0.7499 | 0.5970 | 0.7924 | 0.6090 | 0.7592 | |
0.4550 | 0.6377 | 0.4890 | 0.6776 | 0.5021 | 0.6463 | |
0.5110 | 0.7026 | 0.5497 | 0.7445 | 0.5813 | 0.7365 | |
0.5333 | 0.7271 | 0.5738 | 0.7695 | 0.5713 | 0.7307 | |
0.5528 | 0.7478 | 0.5874 | 0.7831 | 0.6226 | 0.7778 | |
0 | 0.5532 | 0.7483 | 0.5915 | 0.7870 | 0.6162 | 0.7742 |
1 | 0.4354 | 0.6139 | 0.4715 | 0.6572 | 0.4978 | 0.6693 |
4 | 0.5304 | 0.7239 | 0.5700 | 0.7656 | 0.5798 | 0.7382 |
6 | 0.3280 | 0.4754 | 0.3731 | 0.5352 | 0.3750 | 0.4666 |
7 | 0.5528 | 0.7478 | 0.6042 | 0.7994 | 0.6370 | 0.7837 |
8 | 0.1956 | 0.3564 | 0.3215 | 0.4235 | 0.1674 | 0.2014 |
9 | 0.5767 | 0.7724 | 0.6185 | 0.8128 | 0.6456 | 0.7900 |
0 | 0.0987 | 0.1477 | 0.2347 | 0.3460 | 0.0966 | 0.0972 |
1 | 0.5238 | 0.6937 | 0.5984 | 0.7689 | 0.6015 | 0.7215 |
2 | 0.5598 | 0.7551 | 0.5998 | 0.7952 | 0.6341 | 0.8010 |
Table 1
No. 3, No. 4, No. 5, No. 6, No. 7, No. 11, No. 14 and No. 19 blowers are filtered out as data by selecting to compare us
The blower of input.
Support vector machines is a kind of machine learning techniques, and support vector regression is exclusively with Non-linear Kernel function and branch
The prediction regression model for holding vector to establish.In support vector regression, most crucial step is to find a function f ∈ F (F
It is a collection of functions), allow corresponding expected risk function to reach minimum value, it may be assumed that R [f]=∫ l (y-f (x)) dP (x, y).Wherein l
() indicates loss function, represents the deviation between y and f (x), common type be l ()=| y-f (x) |p, wherein p is some
Positive integer.
The basic thought of SVR is as follows:
For given training sample { (x1,y1),(x2,y2),…,(xN,yN) whereinIf returning
Function is
Above-mentioned regression problem can be in the hope of according to Lagrange's theorem and KKT condition:
WhereinFor kernel function.xr,xsTo recognize vector.
Number when 2009 year December 24 day 00 of this method according to 8 blowers selecting to 24 days 08 December in 2009
It is trained according to the air speed data for No. 9 blowers, obtained α1:8,It is shown in Table 2.
α | α* |
0.3536 | 0 |
0.3424 | 0.3424 |
0 | 0.3536 |
0 | 0.3536 |
0 | 0.3536 |
0.2779 | 0.2779 |
0.3536 | 0 |
0.3536 | 0 |
Table 2
Optimization process following steps of this method to Unscented kalman filtering:
1: the feasible set of a specified scale parameter κ, minimum value 0, maximum value are selected according to document.
2: one initial κ of selection0, this method takes
3: enablingPrediction result is calculated in the step of bringing Unscented kalman filtering into.
4: if
Then take κj+1=κj+, otherwise κj+1=κj。
5: circulation 3-4 step, until the threshold value that prediction error reaches setting stops.
The present invention finally again further optimizes input data using grey incidence coefficient, and optimization method is as follows:
1: being directed to selected 8 wind fields, select 2006-2009 annual December 24 to 30 number of days in December respectively
According to each wind field selects 4 groups of data, every group of 168 data.
2: the group data of each wind field and No. 9 wind fields data on December 30 in 24 days to 2009 December in 2009 are subjected to ash
Color calculation of relationship degree.
3: according to the size of grey incidence coefficient, selecting corresponding input data.
What the present invention selected is Absolute Correlation Analysis, and the calculation of the degree of association is as follows:
1: by original series X0={ x0(k), k=1,2 ..., n } and Xi={ xi(k), k=1,2 ..., n } carry out just value
Processing:
2 calculate x0With xiAbsolute Correlation Analysis are as follows:
Wherein,
Calculated result is shown in Table 3.
0.6736 | 0.7399 | 0.7726 | 0.7307 | 0.7348 | 0.6674 | 0.7134 | 0.6789 |
0.7371 | 0.7769 | 0.7848 | 0.7788 | 0.7946 | 0.7660 | 0.7765 | 0.7068 |
0.7151 | 0.8041 | 0.7995 | 0.7697 | 0.8238 | 0.7582 | 0.7598 | 0.7411 |
0.7805 | 0.8056 | 0.7801 | 0.7783 | 0.7749 | 0.7752 | 0.8033 | 0.7201 |
Table 3
Compared with the existing technology, the present invention has following advantage:
(1) present invention can effectively improve precision of prediction compared to more general traditional prediction method.
(2) present invention not only realizes the real-time prediction of wind speed, but also considers the spatial coherence between wind field.
(3) present invention not only allows for the spatial coherence of same year wind field data, and considers between annual each wind speed
Temporal correlation.Processing and the utilization degree of association for shortage of data judge different year with the similitude of time segment data
It calculates, compares the similitude between same time segment data, improve precision of prediction.
(4) present invention not only allows for the correlation between data during prediction, and what is established is nonlinear model
Type preferably reduces the problem of the deficiency of the precision of prediction as caused by wind speed randomness, fluctuation.
(5) it has fully considered the spatial coherence between wind field, has been more in line with actual conditions.
(6) it goes to predict wind speed by a nonlinear prediction technique, be gone using Unscented kalman filtering to wind
Speed is predicted, it is ensured that its real-time predicted, and it is more in line with the practical rule of wind speed, it is as a result more accurate.
(7) grey Absolute data relating extent meter is carried out to 4 years air speed datas of selected wind field and target wind field air speed data
It calculates, influence of the wind speed mutation to prediction result can be reduced, that is, use and go to carry out with the stronger data of prediction period similitude
Prediction, precision of prediction are higher.
(8) pre-processing to data can be to avoid the predicted impact as caused by terms of data in subsequent steps.
Detailed description of the invention
Fig. 1 is general flow chart of the invention;
Fig. 2 is the scale parameter Optimization Steps figure of Unscented kalman filtering;
Fig. 3 is to select optimal input data process figure using grey incidence coefficient;
Fig. 4 is the location map of each wind power plant;
Fig. 5 is the wind speed curve variation diagram of 3 wind power plants;
Fig. 6 is the fitted figure of Ken Deer rank correlation coefficient and selected wind field to target wind field distance;
Fig. 7 is the fitted figure of this Pierre's rank correlation coefficient;
Fig. 8 is not select by grey relational grade, the wind speed result figure predicted with same annual data;
Fig. 9 is the histogram of Fig. 8 prediction residual;
Figure 10 is the prediction result figure of hybrid algorithm;
Figure 11 is to utilize AR-KALMAN filter forecasting result figure;
Figure 12 is wavelet neural network method prediction result figure;
Figure 13 is hybrid algorithm prediction result residual plot;
Figure 14 is AR-KALMAN filter forecasting result residual plot;
Figure 15 is wavelet neural network method prediction result residual plot;
Specific embodiment
To enable target of the invention, technical solution and the more explicit expression of advantage, in conjunction with above-mentioned attached drawing pair
Detailed implementation steps of the invention are described further.
With reference to Fig. 1, the specific steps of the present invention are as follows:
Step 1. selects initial data, concentrates from public data and chooses each wind field data, data acquisition 2006~2009 this 4
Start on year annual December 24, December 30 end of day.
Step 2. carries out missing inspection to all collected data, directly deletes the wind field more than missing data, is left
Utilization averaging method be filled, it may be assumed that assuming that aiFor missing data, then the data for being filled into the position are
If aiThe data of front and back 12 are drawn back forward recursion when also having missing, until obtaining the data of front and back 12;If aiIt is
One data, at this timeSimilarly, aiWhen for the last one data,If aiIt is in data set preceding 12
With rear 12 data when, for convenience of calculation, we still select consistent with the fill method of head and the tail.As a result it can be shown that it is right
The influence of prediction error is not too big.
Step 3. is obtained according to the padding scheme of top except No. 1, No. 8, No. 12, No. 13, other outer wind fields of No. 15 wind fields
Complete data set.
Step 4. selects input data, specific as follows: calculating target wind field 9 and other 16 wind according to copula function
The rank correlation coefficient of the 4th annual data of field, the size selection input wind field according to rank correlation coefficient.By relatively selecting 8 groups of wind
?.
Step 5. selects selected 2006~2009 data of wind field, and data are acquired from annual December 24 to December 30,
Totally 32 groups of data, every group of 168 data.It is on December 30,24 days to 2009 December in 2009 that target wind field, which acquires data, altogether
168 data.
Step 6. calculates the grey relational grade of the annual data of each wind field Yu the 4th annual data of target wind field.
Step 7. selects final input data according to the size relation of grey relational grade.
The input data that step 8. pair is chosen carries out modeling analysis, establishes Nonlinear state space model according to SVR.
Step 9. carries out Unscented kalman filtering prediction according to established state-space model.
Step 10. optimizes the scale parameter in Unscented kalman filtering according to error minimum principle
Step 11. obtains optimal prediction result.
Effect of the invention can be further described by following emulation:
1. simulated conditions
The present invention is by the experiment simulation to same data set progress distinct methods, to illustrate the validity of algorithm.Emulation
Laboratory is carried out in a 4G memory, Celeron double-core 2.6Hz, 32 win7 operating systems using MATLAB2012b.
2. emulation content
Emulation 1, the data set for choosing 22 wind power plants of public data collection Wisconsin, USA are tested.The data
Collecting acquisition time is, time from 2006 year January 1 day to 2009 year December 31 day primary every 1 hour.Data have 4 attributes, point
It Wei not T, dir, spd and airmp.Based on methods herein, the attribute that the present invention chooses in emulation only has wind speed.Data set D1
For the longitude and dimension of 22 wind power plants.
Fig. 4 is the 3 dimension location maps drawn according to the longitude and dimension of each wind power plant.The figure is the centre of sphere with the earth
For spherical coordinates origin, earth radius is the radius of a ball, according to the location map drawn through dimension, can clearly be found out in figure
The range distribution of each wind power plant and target wind field.Fig. 5 is the change curve of the same period wind speed of 3 wind power plants, wherein
Star line is No.1 wind field, and dotted line is No. 2 wind fields, and dotted line is No. 3 wind fields, and data amount check is 92.It can be seen that each wind field is same
The wind speed of period has very strong similitude.
Emulation 2, data set D2 are filled complete data set, including 17 2006~2009 annual December 24 of wind field
To data on the 30th in December.Data set D3 is each wind power plant December 30 24 days to 2009 December in 2009 extracted from D2
The air speed data of day.Fig. 6 is the fitted figure of Ken Deer rank correlation coefficient and selected wind field to target wind field distance, and straight line is represented as
Matched curve, scatterplot are true value.Fig. 7 is the fitted figure of this Pierre's rank correlation coefficient.
Emulation 3 detects algorithm of the invention using data set D3, and Fig. 8 is not select by grey relational grade, directly
The wind speed result figure predicted with same annual data, Fig. 9 are the histograms of prediction residual.Prediction step number is 24 steps, i.e. target
The wind speed in wind field on December 31st, 2009, dotted line are true wind speed, and star line is predicted value.
Emulation 4, in order to preferably improve precision of prediction, it would be desirable to similitude selection be carried out to data, that is, utilize grey
The degree of association selects data and the 4th year strongest number that year of the same period relevance of wind field of target wind field in each wind field 4 years
According to as input data, data set by this method selection we be known as data set D3 '.Figure 10 is D3 ' as input data
When, the prediction result figure of the hybrid algorithm, Figure 11 and Figure 12 are respectively to utilize AR-KALMAN filtering and wavelet neural network method pair
The result figure that same data are predicted, Figure 13, Figure 14 and Figure 15 are the histograms of the prediction residual of above-mentioned three kinds of methods.
Symbol description
SVR: Support vector regression
T: temperature
Dir: wind direction
Spd: wind speed
Airmp: air pressure
D1: emulation data set 1
D2: emulation data set 2
D3: emulation data set 3
Claims (1)
1. a kind of wind speed forecasting method for wind power plant spatial coherence, it is characterised in that: itself the following steps are included:
S1: selection initial data acquires data in disclosed wind farm data collection, carries out shortage of data inspection to the data being collected into
It tests and supplements complete;
S2: its rank correlation coefficient is calculated according to copula function to complete data are supplemented, is selected according to the size of rank correlation coefficient
Select the wind field of suitable input data;
S3: 4 years air speed datas and target wind field air speed data to selected wind field carry out the calculating of grey Absolute data relating extent, according to
Data are finally entered according to what the size relation of grey Absolute data relating extent determined model;
S4: modeling analysis is carried out to the data that finally enter of selection, establishes non-linear state space according to support vector regression
The state equation and measurement equation of model, and estimation prediction is carried out by state of the Unscented kalman filtering to model;
S5: according to the selection criteria of setting, the scale parameter in Unscented kalman filtering is optimized and is updated, is predicted
As a result;
The detailed process of the step S1 are as follows: the data for the same period for selecting each wind power plant annual;22 groups of selection altogether,
Then every group of 192 data carry out missing to all collected data and check that the wind field for being more than 30 for missing data is straight
Deletion is connect, it is remaining to be filled using averaging method, that is, assuming that aiFor missing data, then the data of the position are filled into
If aiWhen the data of front and back 12 also have missing, recursion of drawing back forward, until obtaining the data of front and back 12;
If aiFor first data, then at this timeSimilarly, aiWhen for the last one data,If aiFor
In data set when preceding 12 and rear 12 data, still selection and the identical fill method of head and the tail;And then it obtains except No. 1, No. 8,12
Number, No. 13, the complete data set of other outer wind fields of No. 15 wind fields;
The step S5 is to be optimized to be predicted to the scale parameter in Unscented kalman filtering according to error minimum principle
As a result;
The detailed process of the step S5 is as follows:
The feasible set λ ∈ [0,12] of a specified scale parameter λ, update method is as follows:
(1) initial λ is selected;
Take it to beWherein λmax=12, λmin=0
(2) when updating, every time in original λjOn the basis of be added a random value ej, the value meet normal state variation, be desired for
0, variance very little;
κj+=κj+ej ej~N (0, Pe j), enable j=0,1,
(3) by above-mentioned λj+And λjIt substitutes into Unscented kalman filtering respectively and carries out prediction calculating;
(4) the prediction error for calculating the two takes the small person of error to enter and updates in next step;
If
Then take λj+1=λj+
Otherwise λj+1=λj
(5) circulation 2-4 step, when prediction error reaches the threshold value of setting or update times reach established standards, update stops
Only, optimal scale parameter λ is obtained.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510640990.XA CN105320809B (en) | 2015-09-30 | 2015-09-30 | A kind of wind speed forecasting method for wind power plant spatial coherence |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510640990.XA CN105320809B (en) | 2015-09-30 | 2015-09-30 | A kind of wind speed forecasting method for wind power plant spatial coherence |
Publications (2)
Publication Number | Publication Date |
---|---|
CN105320809A CN105320809A (en) | 2016-02-10 |
CN105320809B true CN105320809B (en) | 2019-01-25 |
Family
ID=55248189
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201510640990.XA Active CN105320809B (en) | 2015-09-30 | 2015-09-30 | A kind of wind speed forecasting method for wind power plant spatial coherence |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN105320809B (en) |
Families Citing this family (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105824987A (en) * | 2016-03-09 | 2016-08-03 | 浙江大学 | Wind field characteristic statistical distributing model building method based on genetic algorithm |
CN109102101B (en) * | 2017-06-21 | 2021-01-26 | 北京金风科创风电设备有限公司 | Wind speed prediction method and system for wind power plant |
CN108090614B (en) * | 2017-12-18 | 2021-05-18 | 哈尔滨工业大学 | Method for establishing space wind field prediction model based on correlation coefficient |
CN109032845A (en) * | 2018-05-04 | 2018-12-18 | 马晨光 | A kind of circuit and method of intelligent control power-off |
CN109166110A (en) * | 2018-08-15 | 2019-01-08 | 安徽省农业科学院土壤肥料研究所 | A kind of intelligence vegetables nursery breeding automatic load system and packing method |
CN109726802B (en) * | 2018-12-29 | 2020-11-20 | 中南大学 | Machine learning prediction method for wind speed in railway and wind farm environment |
CN109902863B (en) * | 2019-02-15 | 2021-08-10 | 浙江财经大学 | Wind speed prediction method and device based on multi-factor time-space correlation |
CN110704406B (en) * | 2019-08-30 | 2020-12-15 | 珠海格力电器股份有限公司 | Energy data processing method, device and equipment |
CN111027852B (en) * | 2019-12-09 | 2023-04-18 | 华南理工大学广州学院 | Method for analyzing wind power generation benefits based on big data processing and grey theory |
CN112052466B (en) * | 2020-08-28 | 2022-04-19 | 西安电子科技大学 | Support vector machine user data prediction method based on multi-party secure computing protocol |
CN113591957B (en) * | 2021-07-21 | 2023-10-27 | 国网上海市电力公司 | Wind power output short-term rolling prediction and correction method based on LSTM and Markov chain |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101871948A (en) * | 2010-06-09 | 2010-10-27 | 中国科学院深圳先进技术研究院 | System and method for predicting wind speed of wind farm |
CN102768738A (en) * | 2012-07-31 | 2012-11-07 | 华锐风电科技(集团)股份有限公司 | Method and device for power predication of wind power station |
KR101214841B1 (en) * | 2011-11-22 | 2012-12-24 | 서강대학교산학협력단 | Real time optimization system for a fixed-bed adsorption process and method therefor |
CN103605908A (en) * | 2013-12-03 | 2014-02-26 | 天津工业大学 | Wind speed sequence forecasting method based on Kalman filtering |
-
2015
- 2015-09-30 CN CN201510640990.XA patent/CN105320809B/en active Active
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101871948A (en) * | 2010-06-09 | 2010-10-27 | 中国科学院深圳先进技术研究院 | System and method for predicting wind speed of wind farm |
KR101214841B1 (en) * | 2011-11-22 | 2012-12-24 | 서강대학교산학협력단 | Real time optimization system for a fixed-bed adsorption process and method therefor |
CN102768738A (en) * | 2012-07-31 | 2012-11-07 | 华锐风电科技(集团)股份有限公司 | Method and device for power predication of wind power station |
CN103605908A (en) * | 2013-12-03 | 2014-02-26 | 天津工业大学 | Wind speed sequence forecasting method based on Kalman filtering |
Non-Patent Citations (3)
Title |
---|
A Hybrid Model for Wind Speed Prediction based on Spatial Correlation;Dong Wang .etal;《International Conference on Modelling, Simulation and Applied Mathematics》;20150131;第124-126页 |
基于支持向量回归机的无迹卡尔曼滤波设计与应用;金瑶;《中国博士学位论文全文数据库 信息科技辑》;20140415(第4期);第6,21-24页 |
基于时间序列分析和卡尔曼滤波算法的风电场风速预测优化模型;潘迪夫等;《电网技术》;20080430;第32卷(第7期);第82-86页 |
Also Published As
Publication number | Publication date |
---|---|
CN105320809A (en) | 2016-02-10 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN105320809B (en) | A kind of wind speed forecasting method for wind power plant spatial coherence | |
Wang et al. | Comparison of seven methods for determining the optimal statistical distribution parameters: A case study of wind energy assessment in the large-scale wind farms of China | |
CN105426970B (en) | A kind of meteorological intimidation estimating method based on discrete dynamic Bayesian network | |
CN114092832B (en) | High-resolution remote sensing image classification method based on parallel hybrid convolutional network | |
CN109146162B (en) | A kind of probability wind speed forecasting method based on integrated Recognition with Recurrent Neural Network | |
CN105574615B (en) | wavelet-BP neural network wind power prediction method based on spatial correlation and GA | |
CN106815850A (en) | The method that canopy density forest reserves very high is obtained based on laser radar technique | |
CN104951834A (en) | LSSVM (least squares support vector machine) wind speed forecasting method based on integration of GA (genetic algorithm) and PSO (particle swarm optimization) | |
CN102495919B (en) | Extraction method for influence factors of carbon exchange of ecosystem and system | |
CN113762603B (en) | Photovoltaic base station short-term photovoltaic power prediction method based on improved sparrow algorithm optimization | |
CN110942194A (en) | Wind power prediction error interval evaluation method based on TCN | |
CN104091216A (en) | Traffic information predication method based on fruit fly optimization least-squares support vector machine | |
CN113657662B (en) | Downscaling wind power prediction method based on data fusion | |
CN105260786A (en) | Comprehensive optimization method of simulation credibility evaluation model of electric propulsion system | |
CN109934422A (en) | Neural network wind speed prediction method based on time series data analysis | |
CN104657584A (en) | Lorenz-system-based wind speed prediction method | |
CN109299208A (en) | Transmission tower intelligent visual methods of risk assessment under a kind of typhoon disaster | |
Sun et al. | A powerful Bio-Inspired optimization algorithm based PV cells diode models parameter estimation | |
CN106786499B (en) | Based on the short-term wind power forecast method for improving AFSA optimization ELM | |
CN104714977B (en) | A kind of correlating method and device of entity and knowledge library item | |
CN111192158A (en) | Transformer substation daily load curve similarity matching method based on deep learning | |
CN110289987A (en) | Multi-agent system network resilience appraisal procedure based on representative learning | |
CN111914488B (en) | Data area hydrologic parameter calibration method based on antagonistic neural network | |
CN106529075B (en) | A kind of non-linear simulation wind speed method considered at times | |
CN116526478B (en) | Short-term wind power prediction method and system based on improved snake group optimization algorithm |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |