CN104657791A - Wind power plant group wind speed distribution prediction method based on correlation analysis - Google Patents

Abstract

The invention relates to a wind power plant group wind speed distribution prediction method based on correlation analysis and belongs to the field of electric power system operation and control. A wind power plant group correlation area division method based on a correction experience variation function is given with a space down-scaling thought, and a wind power plant group is divided into a plurality of correlation areas; on the basis, a space up-scaling thought is adopted, an experience accumulation distribution function is used, correlations between reference wind power plants and target wind power plants in the correlation areas are considered, and wind speeds of the target wind power plants are calculated by the aid of wind speeds of the reference wind power plants, and finally, wind speed curves of all the wind power plants in the wind power plant group are obtained. Therefore, the method has the advantages as follows: (1) the effect caused by the time delay factor is introduced, characteristics of the wind power plant group during actual operation are more conformed, and the prediction effect is better; (2) wind speed distribution of all the wind power plants in the wind power plant group can be effectively described with the space down-scaling and up-scaling thoughts.

Description

A kind of wind farm group wind speed profile Forecasting Methodology based on correlation analysis

Technical field

The present invention relates to a kind of wind speed profile Forecasting Methodology, belong to operation and control of electric power system field, be specifically related to a kind of wind farm group wind speed profile Forecasting Methodology based on correlation analysis.

Background technology

Along with the large-scale development of wind-powered electricity generation, wind power integration all brings huge challenge to electric power netting safe running and scheduling controlling etc.Along with increasing wind-powered electricity generation accesses electrical network in the mode of field group, how to improve the wind speed profile of wind farm group and wind power prediction ability and precision, produce most important to electric power safety.

For single predicting wind speed of wind farm, due to uncertain factor impacts such as weather, temperature and humidity, along with predetermined period increases, its precision significantly declines, the forecasting wind speed error being the cycle with one day is usually comparatively large, to predict that the multiple wind energy turbine set the obtained field group's wind power error characteristics obtained that add up are difficult to do labor.And directly utilize the statistics of wind farm group output power to predict, owing to affecting by net side Power Limitation, be difficult to the real wave characteristic of reflection wind farm group power.

Ask for wind farm group wind speed profile based on correlation analysis, and then predict that the overall power of wind farm group exports, be a kind of new Research Thinking proposed in recent years, Chinese scholars has carried out large quantity research, and research method is broadly divided into three classes:

(1) Pearson correlation coefficient method, the method utilizes two wind energy turbine set historical datas to carry out linear dependence analysis, and weighs correlativity size between two wind energy turbine set with this.The method can only reflect simple linear correlativity between variable, does not introduce space length factor.

(2) based on copula Function Fitting, carry out matching with copula function to wind energy turbine set historical data, searching optimized parameter weighs the correlativity between wind energy turbine set.These class methods are compared with conventional linear correlativity, feature correlativity between variable more neatly, but copula function kind is a lot, that relatively commonly uses at present remains normal state copula function, and the copula Function Optimization parameter of realistic wind farm data be found very difficult.

(3) based on the correlation analysis of experience variation figure, the method utilizes fluctuations in wind speed difference to weigh the correlative character between wind energy turbine set.Under present stage, based on the correlative character between the wind energy turbine set that experience variation function is portrayed, give concrete correlativity regional extent and divide, but wind farm wind velocity delay characteristics under not considering actual conditions.

Summary of the invention

The present invention mainly solves the above-mentioned technical matters existing for prior art, provides a kind of wind farm group wind speed profile Forecasting Methodology based on correlation analysis.The method, compared to single predicting wind speed of wind farm, considers the relevance factors between wind energy turbine set in the group of field, and introduce Delay Factor impact, wind energy turbine set character under more realistic operation, prediction effect is better; And by space NO emissions reduction and the thinking rising yardstick, each wind farm wind velocity distribution in wind farm group is described effectively, under large-scale wind power field group accesses the background of electric system at home, for prediction wind farm group power output capacity provides important information source.

Above-mentioned technical matters of the present invention is mainly solved by following technical proposals:

Based on a wind farm group wind speed profile Forecasting Methodology for correlation analysis, comprise the following steps:

Correlation analysis step, for reading in wind farm group geography information and historical wind speed data, in conjunction with Delay, carries out the wind energy turbine set correlation analysis under correction experience variation function;

Relevant range partiting step: choose suitable reference wind energy turbine set, divides wind farm group correlativity region based on correction experience variation figure;

Correlativity describes step, in the correlativity region that divides in the partiting step of relevant range, utilizes empirical distribution function to describe correlativity between wind energy turbine set.

Scene generates and reduction step, and sampling for utilizing inverse transformation generates target wind farm wind velocity scene, and carries out scene reduction;

Wind speed curve obtaining step, for asking for target wind energy turbine set optimal wind speed curve to reduction scene, thus obtains each wind farm wind velocity curve in wind farm group.

Optimize, above-mentioned a kind of wind farm group wind speed profile Forecasting Methodology based on correlation analysis, described correlation analysis step comprises following sub-step further:

Optimum time delay determination sub-step, calculate optimum time delay based on formula 1, in formula 1, d represents the space length between wind energy turbine set i and wind energy turbine set j, w (f _i, t)=Z (f _i, t+1) and-Z (f _i, t): Z (f _i, t) represent the wind speed of i-th wind energy turbine set at moment t, Δ t represents decay time: γ (d) is correction experience variation function, and wherein, described optimum time delay meets the time delay making γ (d) minimum;

${\mathrm{\γ}}^{\′}\left(d\right)=\frac{1}{2(T-1)}\underset{t=1}{\overset{T-1}{\mathrm{\Σ}}}{\{w(f_i,t+\mathrm{\Δt})-w(f_j,t)\}}^2$ Formula 1

Variation function matching step is rapid, and having in the wind farm group of n wind energy turbine set at one, is one group with two wind energy turbine set, asks for the correction experience variation function under optimum time delay, utilizes formula 2 right individual correction experience variation function point carries out exponential function matching, determines the critical distance r in correlativity region,

G (d)=Nu+s (1-exp (-3d/r)) formula 2

In formula, block gold number N _ufor observational error and the combination apart from interval minimum lower variation intensity, r is critical distance, as d≤r, has stronger spatial coherence between wind energy turbine set, as d>r, does not substantially have spatial coherence between wind energy turbine set.

Optimize, above-mentioned a kind of wind farm group wind speed profile Forecasting Methodology based on correlation analysis, described relevant range partiting step comprises further:

Choose sub-step with reference to electric field, choose and there is stronger forecasting wind speed ability and with the stronger electric field of ambient wind electric field dependencies for reference to wind energy turbine set;

Region dividing sub-step, with reference wind energy turbine set for the center of circle, r is that radius divides wind farm group correlativity region.

Optimize, above-mentioned a kind of wind farm group wind speed profile Forecasting Methodology based on correlation analysis, described correlativity describes in step and adopts " branch mailbox " theory to analyze with reference to the target wind farm wind velocity probability distribution of wind energy turbine set under different wind speed, and further comprises following sub-step:

Branch mailbox sub-step, with reference to wind farm wind velocity data preparation in the chest of equal length, makes have several to comprise the data group of " with reference to wind farm wind velocity, target wind farm wind velocity " in each chest;

Descriptor step, asks for target wind farm wind velocity probability distribution curve corresponding with reference wind energy turbine set in each chest with empirical distribution function.

Optimize, above-mentioned a kind of wind farm group wind speed profile Forecasting Methodology based on correlation analysis, in described descriptor step, describes the probability distribution of target wind farm wind velocity in each chest based on the empirical distribution function in formula 3,

$F_l\left(X\right)=\frac{1}{l}\underset{i=1}{\overset{l}{\mathrm{\Σ}}}\mathrm{\θ}(X-x_i)$ Formula 3

Wherein:

$\mathrm{\&theta;}(X-x_i)=\left\{\begin{array}{cc}1,& \mathrm{forX}\&GreaterEqual;x_i\\ 0,& \mathrm{forX}<x_i\end{array}\right.,$

X is stochastic variable, (x ₁, x ₂..., x _n) be data sample.

Optimize, above-mentioned a kind of wind farm group wind speed profile Forecasting Methodology based on correlation analysis, in described branch mailbox sub-step, data assigned in 25 chests, and each chest data length is 0.04p.u..

Optimize, above-mentioned a kind of wind farm group wind speed profile Forecasting Methodology based on correlation analysis, scene generates and in abatement step, the step of generating scene specifically comprises following sub-step:

Function generates sub-step, and utilize matlab tool box to generate first normal distyribution function Z ~ N (μ 0, ∑) of d T, discontinuity surface number when wherein T is, μ 0 is average, covariance matrix ∑ to be diagonal element be 1 positive definite matrix, d is generating scene quantity;

Inverse transformation sub-step, to discontinuity surface t time each, uses d T unit's normal distyribution function Z ~ N (μ 0, ∑) to carry out inverse transformation sampling to target wind energy turbine set probability distribution curve, just can obtain d wind speed scene of target wind energy turbine set.

Optimize, above-mentioned a kind of wind farm group wind speed profile Forecasting Methodology based on correlation analysis, described scene generates and in abatement step, the step of abatement scene specifically comprises following sub-step:

Cut down scene determination sub-step, determine need the scene of reduction and cut down data ds, wherein, described reduction scene meets two conditions below: (1) is minimum with other scene probability metricses; 2. all scenes of scene likelihood ratio are little;

Scene sum changes sub-step, makes scene sum Ns=NS-1, meanwhile, selects that scene ω s2 nearest with disallowable scene ω s1;

Probability adjustment sub-step, after ensureing to reject scene, remaining scene probability sum is 1, changes the probability of ω s2, makes π (ω s2)=π (ω s2)+π (ω s1);

Cycle criterion sub-step, if remanent field shadow Ns is greater than given scenario quantity ds, then repeats above sub-step successively.

Optimize, above-mentioned a kind of wind farm group wind speed profile Forecasting Methodology based on correlation analysis, in described wind speed curve obtaining step, ask for target wind energy turbine set optimal wind speed curve based on formula 4:

$v_j=\underset{i=1}{\overset{d_s}{\mathrm{\&Sigma;}}}{p}_i{\mathrm{\&omega;}}_{i,j}$ Formula 4

In formula, vj is target wind energy turbine set optimal wind speed curve j moment wind speed, and pi is i-th scene probability, and ω i, j are i-th scene j moment wind speed, and ds is the scene quantity after cutting down.

Therefore, tool of the present invention has the following advantages: (1) predicts based on the wind farm group wind speed profile of correlation analysis, compared to single predicting wind speed of wind farm, consider the relevance factors between wind energy turbine set in the group of field, introducing Delay Factor affects, wind energy turbine set character under more realistic operation, prediction effect is better; (2) by space NO emissions reduction and the thinking rising yardstick, each wind farm wind velocity distribution in wind farm group can be described effectively, under large-scale wind power field group accesses the background of electric system at home, for prediction wind farm group power output capacity provides important information source.

Accompanying drawing explanation

Fig. 1: the inverse transformation sampling schematic diagram being the embodiment of the present invention.

Fig. 2: the wind farm group wind speed profile prediction and calculation process flow diagram being the embodiment of the present invention.

Fig. 3: the wind energy turbine set distribution schematic diagram being the embodiment of the present invention.

Fig. 4-1: be that reaching of the embodiment of the present invention manages wind energy turbine set and western field wind energy turbine set delay character figure.

Fig. 4-2: be the flood pineapple wind energy turbine set of the embodiment of the present invention and western field wind energy turbine set delay character figure.

Fig. 5: the correction experience variation function figure for correlativity Region dividing being the embodiment of the present invention.

Fig. 6: the original experience variation function figure being the embodiment of the present invention.

Fig. 7: the wind farm group correlativity zoning plan being the embodiment of the present invention.

Fig. 8-1: No. 8 chest internal object wind farm wind velocity probability distribution graph being the embodiment of the present invention.

Fig. 8-2: No. 16 chest internal object wind farm wind velocity probability distribution graph being the embodiment of the present invention.

Fig. 8-3: No. 20 chest internal object wind farm wind velocity probability distribution graph being the embodiment of the present invention.

Fig. 9-1: the wind energy turbine set 2 wind speed scene generation figure being the embodiment of the present invention.

Fig. 9-2: the wind energy turbine set 3 wind speed scene generation figure being the embodiment of the present invention.

Figure 10-1: the wind energy turbine set 2 optimal wind speed curve map being the embodiment of the present invention.

Figure 10-2: the wind energy turbine set 3 optimal wind speed curve map being the embodiment of the present invention.

Figure 11: the wind energy turbine set 4 optimal wind speed curve map being the embodiment 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:

1, based on the wind farm group correlativity Region dividing of space NO emissions reduction

So-called wind farm group space NO emissions reduction, is exactly by the problem of each wind farm wind velocity in research wind farm group, is converted into the problem of correlativity region wind speed in research wind farm group.The major issue that wind farm group space NO emissions reduction needs solve is exactly the division in wind farm group correlativity region, method proposes and considers that the correction experience variation function of Delay is to divide the method in wind farm group correlativity region.

1.1, the correction experience variation function of Delay is considered

Experience variation function is in order to the means of survey region variable space variation characteristic and intensity and instrument in geostatistics.

The process of time domain trend term is gone to n wind farm wind velocity, that is:

w(f _i,t)＝Z(f _i,t+1)-Z(f _i,t)

i＝1,2,...,n；t＝1,2,...,T-1 (1)

In formula: Z (fi, t) represents the wind speed of i-th wind energy turbine set at moment t.

For wind energy turbine set i and wind energy turbine set j, their experience variation function can be expressed as:

$\mathrm{\&gamma;}\left(d\right)=\frac{1}{2(T-1)}\underset{t=1}{\overset{T-1}{\mathrm{\&Sigma;}}}{\{w(f_i,t)-w(f_j,t)\}}^2---\left(2\right)$

In formula, d represents the space length between wind energy turbine set i and wind energy turbine set j, and experience variation function value is less, then two wind energy turbine set correlativitys are larger.

Owing to having certain space length between the wind energy turbine set in reality, so wind is propagated between wind energy turbine set need the time, cause the appearance of time delay.Therefore, the correction carrying out considering time delay to the experience of correction variation function is needed:

${\mathrm{\&gamma;}}^{\&prime;}\left(d\right)=\frac{1}{2(T-1)}\underset{t=1}{\overset{T-1}{\mathrm{\&Sigma;}}}{\{w(f_i,t+\mathrm{\&Delta;t})-w(f_j,t)\}}^2---\left(3\right)$

In formula, Δ t represents decay time, and meeting the time delay making γ ' (d) minimum is optimum time delay.

1.2 based on revise experience variation figure wind farm group correlativity Region dividing

For n wind energy turbine set in some wind farm groups, be one group with two wind energy turbine set, C2n group altogether.Ask for the correction experience variation function under optimum time delay, exponential function matching carried out to C2n correction experience variation function point:

g(d)＝Nu+s(1-exp(-3d/r))

(4)

In formula, r is critical distance, as d≤r, has stronger spatial coherence between wind energy turbine set, as d>r, does not substantially have spatial coherence between wind energy turbine set.

When wind energy turbine set is comparatively concentrated, with reference to wind energy turbine set for the center of circle, r is that radius divides wind farm group correlativity region, namely thinks that other wind energy turbine set (being called target wind energy turbine set) have stronger correlativity with reference wind energy turbine set in region.Selection principle with reference to wind energy turbine set is: 1. have stronger forecasting wind speed ability; 2. stronger with ambient wind electric field dependencies.

2, the correlativity regional wind power wind speed rising yardstick based on space is asked for

In correlativity region, obtain each wind farm wind velocity curve in region, be space in correlativity region and rise yardstick.This method, using correlativity region internal reference wind farm wind velocity as input, asks for region internal object wind farm wind velocity curve, to realize the target that correlativity regional space rises yardstick.

2.1 raw data arrange

For the reference wind farm wind velocity that discontinuity surface time certain is given, need to know corresponding target wind farm wind velocity probability distribution.

First " branch mailbox " theory is adopted to analyze with reference to wind energy turbine set under different wind speed, target wind farm wind velocity probability distribution.First, be organized in the chest of equal length with reference to wind farm wind velocity data.In this method, 25 chests and each chest data length is adopted to be 0.04p.u..Each has a time upper corresponding target wind farm wind velocity with reference to wind farm wind velocity, is belonged to by target wind farm wind velocity in the chest at corresponding reference wind farm wind velocity place.Therefore, several data groups [with reference to wind farm wind velocity, target wind farm wind velocity] are had in each chest.

Thus, empirical distribution function can be adopted to describe the probability distribution of target wind farm wind velocity in each chest.For a wind speed stochastic variable X, if having N number of data x1, x2 ..., xN, so the empirical cumulative distribution function of stochastic variable X is:

$F_l\left(X\right)=\frac{1}{l}\underset{i=1}{\overset{l}{\mathrm{\&Sigma;}}}\mathrm{\&theta;}(X-x_i)$

$\mathrm{\&theta;}(X-x_i)=\left\{\begin{array}{cc}1,& \mathrm{forX}\&GreaterEqual;x_i\\ 0,& \mathrm{forX}<x_i\end{array}\right.$

$\left(5\right)$

Sample size N is larger, and empirical cumulative distribution is more tending towards true.

2.2 target wind energy turbine set scenes generate

With reference to wind farm wind velocity V={vi, i ∈ T}T can be considered as a stochastic variable, as known reference wind farm wind velocity vi, by judging which chest this wind speed belongs to and carry out evaluating objects wind farm wind velocity probability curve, thus target wind farm wind velocity scene can be obtained.

2.2.1 inverse transformation sampling

Inverse transformation method has been widely used in Monte Carlo sampling, and this method uses inversion to bring the target wind farm wind velocity obtaining and obey particular probability distribution.Inverse transformation concrete grammar is as follows.

For some stochastic variable Vi, it obeys Pr (Vi≤v)=Fl (v) distribution, so carries out inverse transformation sampling to variable V i and is expressed as:

$v_i=F_l^{-1}\left(U\right),U~\mathrm{Unif}\left[\mathrm{0,1}\right]---\left(6\right)$

Unif [0,1] is being uniformly distributed on [0,1] interval.

Because the accumulated probability distribution function value of standardized normal distribution is that it is equally distributed to obey between [0,1].So above formula U can substitute with Standard Normal Distribution value Φ (Zt):

$\mathrm{\&Phi;}\left(Z_t\right)=\underset{-\&infin;}{\overset{Z_t}{\&Integral;}}\frac{1}{\sqrt{2\mathrm{\&pi;}}}{e}^{-x^2/2}\mathrm{dx}---\left(7\right)$

$v_i=F_l^{-1}\left(\mathrm{\&Phi;}\left(Z_t\right)\right)---\left(8\right)$

Above-mentioned inverse transformation can display simply, as shown in Figure 1.Arrow represents the travel direction of inverse transformation: starting point is standardized normal distribution random quantity Zt, obtains empirical distribution function value Fl (v) corresponding with Standard Normal Distribution value Φ (Zt), final output wind speed v.Know that as long as visible target wind farm wind velocity experience distributes, just can obtain target wind farm wind velocity.

2.2.2 scene generation step

Given with reference to wind farm wind velocity vi, i=1,2 ..., T is as input, and the concrete steps of target wind farm wind velocity being carried out to scene generation are as follows:

(1) empirical distribution function is utilized to ask for target wind farm wind velocity probability distribution curve corresponding with reference wind energy turbine set in each chest.

(2) for discontinuity surface t time each, judge to belong in which chest with reference to wind farm wind velocity vi, thus obtain the probability distribution curve of this chest internal object wind farm wind velocity.

(3) matlab tool box is utilized to generate d T unit's normal distyribution function Z ~ N (μ 0, ∑), discontinuity surface number when wherein T is, μ 0 is average, can be taken as 0, covariance matrix ∑ to be diagonal element be 1 positive definite matrix, d is generating scene quantity, is generally taken as 500.

(4) to discontinuity surface t time each, use d T unit's normal distyribution function Z ~ N (μ 0, ∑) to carry out inverse transformation sampling to target wind energy turbine set probability distribution curve, just can obtain d wind speed scene of target wind energy turbine set.

2.2.2 scene is cut down

D scene of target wind farm wind velocity can be generated by said process, in order to improve computing velocity, needing d the scene to generating to cut down, namely under the prerequisite ensureing precision, providing as far as possible few scene.

This patent uses synchronous back substitution null method, and concrete steps are:

(1) determine the scene needing to cut down, cut down scene and meet two conditions below:

1. very near with other scene probability metricses; 2. scene probability is very little;

(2) scene sum is changed, that is: Ns=Ns-1.Meanwhile, that scene ω s2 nearest with disallowable scene ω s1 is selected.

(3) reject after scene to ensure, remaining scene probability sum is 1.Change the probability of ω s2, that is: π (ω s2)=π (ω s2)+π (ω s1)

(4) as long as Ns is greater than given scenario quantity ds, repeat to step 1.

2.3 target wind energy turbine set optimal wind speed curves

In electric system actual schedule is run, often need to know an optimal wind speed curve, the probability that namely it occurs is maximum.The concept average according to statistically probability weight, ask for target wind energy turbine set optimal wind speed curve with ds scene after cutting down:

$v_j=\underset{i=1}{\overset{d_s}{\mathrm{\&Sigma;}}}{p}_i{\mathrm{\&omega;}}_{i,j}---\left(9\right)$

In formula, vj is target wind energy turbine set optimal wind speed curve j moment wind speed, and pi is i-th scene probability, and ω i, j are i-th scene j moment wind speed.

3 wind farm group wind speed profile prediction and calculation steps

1. to sum up, based on correlation analysis wind farm group wind speed profile Forecasting Methodology process flow diagram as shown in Figure 2, comprise the following steps:

Correlation analysis step, for reading in wind farm group geography information and historical wind speed data, in conjunction with Delay, carries out the wind energy turbine set correlation analysis under correction experience variation function;

Relevant range partiting step: choose suitable reference wind energy turbine set, divides wind farm group correlativity region based on correction experience variation figure;

Correlativity describes step, in the correlativity region that divides in the partiting step of relevant range, utilizes empirical distribution function to describe correlativity between wind energy turbine set.

Scene generates and reduction step, and sampling for utilizing inverse transformation generates target wind farm wind velocity scene, and carries out scene reduction;

Wind speed curve obtaining step, for asking for target wind energy turbine set optimal wind speed curve to reduction scene, thus obtains each wind farm wind velocity curve in wind farm group.

4, example and emulation

This patent simulation calculation adopts 6 wind energy turbine set, 2 months air speed datas near Chifeng, and the time interval is 15min point.6 wind energy turbine set location distribution as shown in Figure 3.

4.1 based on the correlativity Region dividing revising experience variation function

4.1.1 the correction experience variation function under delay character

For 3 wind farm data, be respectively in reaching, Xi Chang and flood pineapple, the correction experience variation function between them under delay character as shown in Figure 3.

As shown in Figure 4: curve minimum point characterizes two wind energy turbine set correction experience variation function minimum value, and the corresponding horizontal ordinate time is optimum time delay.Da Li and Xi Chang is at a distance of 124.8km, and optimum time delay is-12.7min (namely the former is more delayed than the latter); Xi Chang and flood pineapple are at a distance of 142.9km, and optimum time delay is+29.8min (namely the former is more advanced than the latter).

4.1.2 correlativity Region dividing

Utilize correction experience variation figure to divide correlativity region, correction experience variation figure can be obtained, as shown in Figure 5.

The correction experience variation function figure of MATLAB fitting function " fit " to wind farm wind velocity is utilized to carry out exponential function matching, the optimized parameter of exponential fitting can be obtained: Nu=0.6059, s=0.2631, r=189.1, namely the wind energy turbine set of space length within the scope of 189.1km has stronger correlativity.

Utilize and do not consider that the original experience variation figure of Delay carries out exponential fitting result, if shown in 5.The optimized parameter of exponential fitting can be obtained: Nu=0.6383, s=0.2717, r=109.6, can find that correlativity regional extent has obviously to reduce, namely adopt original experience variation function to divide correlativity region simply, weaken the correlativity between wind energy turbine set, and fitting effect is poor.

4.2 wind farm group wind speed profile are asked for

In this patent use correlativity region, 3 wind energy turbine set and the extra-regional wind farm wind velocity data of correlativity are as analytic target, as shown in Figure 7.In order to understand conveniently, this patent is numbered wind energy turbine set 1,2,3,4 to it, wherein wind energy turbine set 1 is as reference wind energy turbine set, centered by reference wind energy turbine set, radius is that the border circular areas of r=189.1km is divided into correlativity region, wind energy turbine set 2 and 3 is the target wind energy turbine set in correlativity region, and wind energy turbine set 4 is the extra-regional wind energy turbine set of correlativity.

4.2.1 the correction experience variation function under delay character

Wind energy turbine set 1 and wind energy turbine set 2 have 3264 data groups, and they are assigned in 25 chests.In No. 10 chest, have 235 data groups (the reference wind farm wind velocity in chest changes between 0.36p.u. to 0.4p.u.), in No. 16 chest, have 125 data groups (the reference wind farm wind velocity in chest changes between 0.6p.u. to 0.64p.u.).Each chest internal reference wind farm wind velocity is very close to (between 0.04p.u.), but target wind farm wind velocity but has very large difference.

Fig. 8-1,8-2,8-3 are respectively target wind farm wind velocity probability distribution in No. 8, No. 16 and No. 20 chest (why ordinate scope is greater than 1, be because horizontal ordinate scope is less than 1, and curvilinear integral is the cause of 1).

As shown in Figure 8, when different with reference to wind farm wind velocity, target wind farm wind velocity probability distribution also changes thereupon, and increases along with reference to wind farm wind velocity, the peak point of target wind farm wind velocity probability distribution also becomes large, characterizes between wind energy turbine set and there is certain correlativity.

4.2.2 correlativity region internal object wind farm wind velocity is asked for

Using the wind energy turbine set 1 actual measurement wind speed of certain day as input, generate d=500 wind energy turbine set 2,3 wind speed scene, and original scene d is reduced to ds=10, as shown in Figure 9, p is scene probable value.

10 scene curve are carried out probability weight average, merge into an optimal wind speed curve, as shown in Figure 10.

4.2.3 outside correlativity region, wind farm wind velocity is asked for

With certain sky of wind energy turbine set 1 actual measurement wind speed for input, scene generation is carried out to the extra-regional wind energy turbine set 4 of correlativity, and 10 scene curve are merged into an optimal wind speed curve, as shown in figure 11.

4.2.4 wind farm group wind speed profile predicated error is analyzed

The interpretation of result that table 1 gives context of methods and adopts neural net method to predict wind farm wind velocity.

The wind speed of contrast actual measurement wind speed, context of methods and Neural Network model predictive can be found out, using reference wind energy turbine set day measured data as input, the correlating region internal object wind energy turbine set wind speed of a day, gained wind speed curve can reflect target wind energy turbine set actual wind speed situation, substantially realistic wind speed variation tendency.Compare and utilize neural network to carry out wind farm group forecast of distribution, substantially increase precision of prediction, can describe better, correlativity between wind energy turbine set.Definition precision improves formula:

Compared with neural network model, adopt the wind speed profile prediction that context of methods carries out wind energy turbine set 2,3,4, its precision of prediction increasing amount is respectively 48.9%, 49.9% and 62.1%.

It should be noted that, when correlating extra-regional wind farm wind velocity, due to more weak with reference to wind energy turbine set correlativity, only can reflect the situation of actual wind speed mean value, cannot truly reflect wind speed variation tendency.

The present invention for thinking, on the basis analyzing Delay Factor between wind energy turbine set, proposes the wind farm group correlativity region partitioning method based on the experience of correction variation function with space NO emissions reduction; In some correlativity regions, the thinking rising yardstick with space, to ask for wind farm wind velocity distribution in region, with reference wind farm wind velocity as input, obtains target wind farm wind velocity curve.Simulation result shows:

(1) the wind farm group wind speed profile based on correlation analysis is predicted, compared to single predicting wind speed of wind farm, consider the relevance factors between wind energy turbine set in the group of field, introduce Delay Factor impact, wind energy turbine set character under more realistic operation, prediction effect is better.

(2) by space NO emissions reduction and the thinking rising yardstick, each wind farm wind velocity distribution in wind farm group can be described effectively.Under large-scale wind power field group accesses the background of electric system at home, for prediction wind farm group power output capacity provides important information source.

Specific embodiment described herein 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 (9)

1., based on a wind farm group wind speed profile Forecasting Methodology for correlation analysis, it is characterized in that, comprise the following steps:

Correlation analysis step, for reading in wind farm group geography information and historical wind speed data, in conjunction with Delay, carries out the wind energy turbine set correlation analysis under correction experience variation function;

Relevant range partiting step: choose suitable reference wind energy turbine set, divides wind farm group correlativity region based on correction experience variation figure;

Correlativity describes step, in the correlativity region that divides in the partiting step of relevant range, utilizes empirical distribution function to describe correlativity between wind energy turbine set;

Scene generates and reduction step, and sampling for utilizing inverse transformation generates target wind farm wind velocity scene, and carries out scene reduction;

Wind speed curve obtaining step, for asking for target wind energy turbine set optimal wind speed curve to reduction scene, thus obtains each wind farm wind velocity curve in wind farm group.

2. a kind of wind farm group wind speed profile Forecasting Methodology based on correlation analysis according to claim 1, it is characterized in that, described correlation analysis step comprises following sub-step further:

Optimum time delay determination sub-step, calculate optimum time delay based on formula 1, in formula 1, d represents the space length between wind energy turbine set i and wind energy turbine set j, w (f _i, t)=Z (f _i, t+1) and-Z (f _i, t): Z (f _i, t) represent the wind speed of i-th wind energy turbine set at moment t, Δ t represents decay time: γ (d) is correction experience variation function, and wherein, described optimum time delay meets the time delay making γ (d) minimum;

${\mathrm{\&gamma;}}^{\&prime;}\left(d\right)=\frac{1}{2(T-1)}\underset{t=1}{\overset{T-1}{\mathrm{\&Sigma;}}}{\{w(f_i,t+\mathrm{\&Delta;t})-w(f_j,t)\}}^2$ Formula 1

Variation function matching step is rapid, and having in the wind farm group of n wind energy turbine set at one, is one group with two wind energy turbine set, asks for the correction experience variation function under optimum time delay, utilizes formula 2 right individual correction experience variation function point carries out exponential function matching, determines the critical distance r in correlativity region,

G (d)=Nu+s (1-exp (-3d/r)) formula 2

In formula, block gold number N _ufor observational error and the combination apart from interval minimum lower variation intensity, r is critical distance, as d≤r, has stronger spatial coherence between wind energy turbine set, as d>r, does not substantially have spatial coherence between wind energy turbine set.

3. a kind of wind farm group wind speed profile Forecasting Methodology based on correlation analysis according to claim 2, it is characterized in that, described relevant range partiting step comprises further:

Choose sub-step with reference to electric field, choose and there is stronger forecasting wind speed ability and with the stronger electric field of ambient wind electric field dependencies for reference to wind energy turbine set;

Region dividing sub-step, with reference wind energy turbine set for the center of circle, r is that radius divides wind farm group correlativity region.

4. a kind of wind farm group wind speed profile Forecasting Methodology based on correlation analysis according to claim 3, it is characterized in that, described correlativity describes in step and adopts " branch mailbox " theory to analyze with reference to the target wind farm wind velocity probability distribution of wind energy turbine set under different wind speed, and further comprises following sub-step:

Branch mailbox sub-step, with reference to wind farm wind velocity data preparation in the chest of equal length, makes have several to comprise the data group of " with reference to wind farm wind velocity, target wind farm wind velocity " in each chest;

Descriptor step, asks for target wind farm wind velocity probability distribution curve corresponding with reference wind energy turbine set in each chest with empirical distribution function.

5. a kind of wind farm group wind speed profile Forecasting Methodology based on correlation analysis according to claim 4, it is characterized in that, in described descriptor step, the probability distribution of target wind farm wind velocity in each chest is described based on the empirical distribution function in formula 3

$F_l\left(X\right)=\frac{1}{l}\underset{i=1}{\overset{l}{\mathrm{\&Sigma;}}}\mathrm{\&theta;}(X-x_i)$ Formula 3

Wherein:

$\mathrm{\&theta;}(X-x_i)=\left\{\begin{array}{cc}1,& \mathrm{for\; X}\&GreaterEqual;x_i\\ 0,& \mathrm{for\; X}<x_i,\end{array}\right.$

X is stochastic variable, (x ₁, x ₂..., x _n) be data sample.

6. a kind of wind farm group wind speed profile Forecasting Methodology based on correlation analysis according to claim 4, is characterized in that, in described branch mailbox sub-step, data assigned in 25 chests, and each chest data length is 0.04p.u..

7. a kind of wind farm group wind speed profile Forecasting Methodology based on correlation analysis according to claim 4, is characterized in that, scene generates and in abatement step, the step of generating scene specifically comprises following sub-step:

Function generates sub-step, and utilize matlab tool box to generate first normal distyribution function Z ~ N (μ 0, ∑) of d T, discontinuity surface number when wherein T is, μ 0 is average, covariance matrix ∑ to be diagonal element be 1 positive definite matrix, d is generating scene quantity;

Inverse transformation sub-step, to discontinuity surface t time each, uses d T unit's normal distyribution function Z ~ N (μ 0, ∑) to carry out inverse transformation sampling to target wind energy turbine set probability distribution curve, just can obtain d wind speed scene of target wind energy turbine set.

8. a kind of wind farm group wind speed profile Forecasting Methodology based on correlation analysis according to claim 4, is characterized in that, described scene generates and in abatement step, the step of abatement scene specifically comprises following sub-step:

Cut down scene determination sub-step, determine need the scene of reduction and cut down data ds, wherein, described reduction scene meets two conditions below: condition 1, minimum with other scene probability metricses; Condition 2, all scenes of scene likelihood ratio are little;

Scene sum changes sub-step, makes scene sum Ns=NS-1, meanwhile, selects that scene ω s2 nearest with disallowable scene ω s1;

Probability adjustment sub-step, after ensureing to reject scene, remaining scene probability sum is 1, changes the probability of ω s2, makes π (ω s2)=π (ω s2)+π (ω s1);

Cycle criterion sub-step, if remanent field shadow Ns is greater than given scenario quantity ds, then repeats above sub-step successively.

9. a kind of wind farm group wind speed profile Forecasting Methodology based on correlation analysis according to claim 8, is characterized in that, in described wind speed curve obtaining step, asks for target wind energy turbine set optimal wind speed curve based on formula 4:

$v_j=\underset{i=1}{\overset{d_s}{\mathrm{\&Sigma;}}}{p}_i{\mathrm{\&omega;}}_{i,j}$ Formula 4

In formula, vj is target wind energy turbine set optimal wind speed curve j moment wind speed, and pi is i-th scene probability, and ω i, j are i-th scene j moment wind speed, and ds is the scene quantity after cutting down.