CN105225252A - Particle clouds motion Forecasting Methodology - Google Patents

Abstract

The invention discloses a kind of particle clouds motion Forecasting Methodology, by first obtaining the pixel set at cloud cluster edge in not sky image in the same time, solved the corresponding relation of the pixel at cloud cluster edge in not sky image in the same time by optimization algorithm based on predetermined optimization aim, based on the position of the pixel of this corresponding relation and correspondence, the subsequent motion position for cloud cluster image border is predicted.Thus, can predict comparatively accurately for particle clouds motion, for a photovoltaic generation power minute level prediction provides Data support.

Description

Particle clouds motion Forecasting Methodology

Technical field

The present invention relates to image procossing and photovoltaic power electric powder prediction, be specifically related to a kind of particle clouds motion Forecasting Methodology, more specifically, relate to the particle clouds motion Forecasting Methodology under a kind of Optimum Theory framework.

Background technology

Photovoltaic generation is the same with wind-power electricity generation all belongs to undulatory property and intermittent power supply, because photovoltaic generating system is by the impact of the climatic factor such as intensity of illumination and environment temperature, the change of its output power has uncertainty, the disturbance of output power will likely affect the stable of electrical network, therefore, need the research strengthening photovoltaic power generation power prediction, obtain the daily generation curve of photovoltaic generating system in advance, thus coordinate electric system formulation generation schedule, reduce the Randomization of photovoltaic generation to the impact of electric system.Earth's surface irradiance prediction is the primary link of photovoltaic generation power substep prediction, and its accuracy is the key of guaranteed output precision of prediction.But earth's surface irradiance under cloudy weather disappears and the impact of motion by cloud cluster is raw, its change in the feature such as random, quick, violent, has had a strong impact on the precision of Classical forecast algorithm sometimes.Therefore, in order to improve the precision of prediction of earth's surface irradiance under cloudy weather, directly observation must be carried out to the cloud cluster that it is aerial to obtain corresponding data, studying the method for cloud cluster displacement identification and motion prediction.

Summary of the invention

In view of this, the invention provides a kind of particle clouds motion Forecasting Methodology, to predict comparatively accurately particle clouds motion according to sky image.

A kind of particle clouds motion Forecasting Methodology of the present invention comprises:

Obtain first day null images in the first moment, obtain the second sky image in the second moment, described second moment is relative to delayed first schedule time in described first moment;

Obtain first set characterizing described first day null images medium cloud group image border and the second point set characterizing described second sky image medium cloud group image border;

Calculate the distance matrix that described first point set is incorporated into described second point set, the element D of described distance matrix _i,jfor described first i-th point gathered is to the distance of a jth point of described second point set;

With minimize the shift length sum of all pixels of cloud cluster image border and the shift length variance minimizing all pixels of cloud cluster image border for optimization aim, ask for transposed matrix based on optimization algorithm;

Described transposed matrix, for characterizing the corresponding relation of described first set and described second point set, is two-valued function matrix; The element E of described transposed matrix _i,jvalue shows when being 1 that i-th point of described first set is corresponding with a jth point of described second point set, the element E of described transposed matrix _i,jvalue shows when being 0 that i-th point of described first set and the jth of described second point set are put without corresponding relation;

Obtain according to described first set, described second point set and described transposed matrix and thirdly gather, the described predicted position thirdly gathered for characterizing the 3rd moment cloud cluster image border, described 3rd moment is relative to delayed second schedule time in described second moment.

Preferably, obtain first set characterizing described first day null images medium cloud group image border to comprise:

First day null images binaryzation is obtained the first day null images of binaryzation;

To first day null images smoothing process acquisition first smoothed image of binaryzation;

For the first smoothed image, travel through image edge pixels point in a predefined manner to obtain edge pixel point coordinate and to count first set, until all pixels are traversed or point in first set forms closed figures, described edge pixel point is that gray-scale value is 0 and surrounding at least exists a gray-scale value is the pixel of 1.

Preferably, obtain the second point set characterizing described second sky image medium cloud group image border to comprise:

Second sky image binaryzation is obtained the second sky image of binaryzation;

To the second sky image smoothing process acquisition second smoothed image of binaryzation;

For the second smoothed image, travel through image edge pixels point in a predefined manner to obtain edge pixel point coordinate and to count second point set, until all pixels are traversed or first set forms closed figures, described edge pixel point is that gray-scale value is 0 and surrounding at least exists a gray-scale value is the pixel of 1.

Preferably, described optimization algorithm is genetic algorithm, and its objective function is:

$\min =\mathrm{\α}\left(\underset{i=1}{\overset{m}{\Σ}}\underset{j=1}{\overset{n}{\Σ}}{E}_{i,j}\·D_{i,j}\right)+\mathrm{\β}\·\mathrm{var}\left(E\·D|{E_{i,j}\·D_{i,j}\≠0}_{}\right)$

Wherein, E is described transposed matrix, E _i,jfor the element of described transposed matrix, D is described distance matrix, D _i,jfor the element of described distance matrix, for { E _i,jd _i,j| i=1,2 ..., m; J=1,2 ..., all in n} is not the variance of the number of 0 value, α and β is weight coefficient, and m is the quantity of pixel in first set, and n is the quantity of pixel in second point set.

Preferably, describedly ask for transposed matrix based on optimization algorithm and comprise:

If with E _i,j=1, then E _{i+1, j}+ E _{i, j+1}+ E _{i+1, j+1}=1, for constraint condition asks for transposed matrix based on genetic algorithm, wherein, if i=m, make i+1=1, if j=n, make j+1=1.

Preferably, obtain thirdly set based on described first set, described second point set and described transposed matrix to comprise:

The corresponding relation of the pixel in described first set and second point set is obtained according to transposed matrix;

Travel through the pixel in described second point set, calculate the position obtaining corresponding predict pixel point, and described predict pixel point is counted thirdly gather;

Wherein, if the pixel of second point set only corresponds to the pixel of one first set, then based on following formulae discovery predict pixel point position:

$B_k^3=B_j^2+(B_j^2-B_i^1)\frac{{\mathrm{\Δt}}_2}{{\mathrm{\Δt}}_1};$

Wherein, for the pixel coordinate in second point set, for with pixel coordinate in first corresponding set, for the described predict pixel point coordinate in described thirdly set, i, j, k are the sequence number of each pixel coordinate in corresponding point set; Δ t ₁for described first schedule time, Δ t ₂for described second schedule time;

If the pixel of this second point set corresponds to the pixel at least two first set, then based on following formulae discovery predict pixel point position:

$B_k^3=round(2B_j^2-\frac{\underset{l=1}{\overset{L}{\Σ}}{B}_{i,l}^1}{L});$

Wherein, L be with the quantity of the pixel of first corresponding set, for with the coordinate of i-th pixel in L first corresponding set pixel, round () is round function.

Preferably, described method also comprises:

The 3rd sky image of binaryzation is generated to characterize the cloud cluster edge of prediction according to described thirdly set.

By first obtaining the pixel set at cloud cluster edge in not sky image in the same time, solved the corresponding relation of the pixel at cloud cluster edge in not sky image in the same time by optimization algorithm based on predetermined optimization aim, based on the position of the pixel of this corresponding relation and correspondence, the subsequent motion position for cloud cluster image border is predicted.Thus, can predict comparatively accurately for particle clouds motion, for a photovoltaic generation power minute level prediction provides Data support.

Accompanying drawing explanation

By referring to the description of accompanying drawing to the embodiment of the present invention, above-mentioned and other objects, features and advantages of the present invention will be more clear, in the accompanying drawings:

Fig. 1 is the process flow diagram of the particle clouds motion Forecasting Methodology of the embodiment of the present invention;

Fig. 2 is the first day null images of the embodiment of the present invention and the schematic diagram of the second sky image;

Fig. 3 is the first smoothed image of the embodiment of the present invention and the schematic diagram of the second smoothed image;

Fig. 4 is the comparison diagram of the 3rd sky image that obtains based on the embodiment of the present invention and the mutually sky image profile of actual measurement in the same time.

Embodiment

Based on embodiment, present invention is described below, but the present invention is not restricted to these embodiments.In hereafter details of the present invention being described, detailedly describe some specific detail sections.Do not have the description of these detail sections can understand the present invention completely for a person skilled in the art yet.In order to avoid obscuring essence of the present invention, known method, process, flow process, element and circuit do not describe in detail.

In addition, it should be understood by one skilled in the art that the accompanying drawing provided at this is all for illustrative purposes, and accompanying drawing is not necessarily drawn in proportion.

Unless the context clearly requires otherwise, similar words such as " comprising ", " comprising " otherwise in whole instructions and claims should be interpreted as the implication that comprises instead of exclusive or exhaustive implication; That is, be the implication of " including but not limited to ".

In describing the invention, it is to be appreciated that term " first ", " second " etc. are only for describing object, and instruction or hint relative importance can not be interpreted as.In addition, in describing the invention, except as otherwise noted, the implication of " multiple " is two or more.

In the present invention, the nonlinear motion of cloud cluster entirety is decomposed into the combination of the linear movement of each point forming cloud cluster set, by the corresponding relation of optimized Algorithm for Solving cloud cluster edge each point, based on this corresponding relation and existing motion conditions predict cloud cluster next step movement and situation of change.

Fig. 1 is the process flow diagram of the particle clouds motion Forecasting Methodology of the embodiment of the present invention.

As shown in Figure 1, described particle clouds motion Forecasting Methodology comprises:

Step 100, at the first moment t ₁obtain first day null images, at the second moment t ₂obtain the second sky image, described second moment t ₂relative to described first moment t ₁delayed first schedule time Δ t ₁.

Should be understood that the sky image do not obtained in the same time is obtained by ground facilities for observation based on identical position and parameter, so just likely carry out subsequent treatment.

Described first day null images and the second sky image are gray level image.Preferably, the second moment is relative to Δ t retardation time in the first moment ₁within being limited in 10 minutes.

Step 200, acquisition characterize first set B of described first day null images medium cloud group image border ¹with the second point set B characterizing described second sky image medium cloud group image border ².

First described set B can be obtained respectively by carrying out image procossing to first day null images and the second sky image ¹with second point set B ².

Particularly, the process of step 200 pair first day null images can comprise:

Step 210, first day null images binaryzation is obtained the first day null images of binaryzation.

Preferably, binary conversion treatment is carried out by the following method.

If image resolution ratio is M × N, gray-scale value matrix is F, and image binaryzation parameter is T, and the bianry image matrix after conversion is F ', then

$F_{x,y}^{\′}=\left\{\begin{array}{cc}1& F_{x,y}>T\\ 0& F_{x,y}\≤T\end{array}\right.,x=1,2,\mathrm{...},M;y=1,2,\mathrm{...},N$

In the present embodiment, setting binaryzation parameter T=100, through the sky image of binaryzation, cloud cluster part is black (gray scale is 0), and sky portion is white (gray scale is 1).

Step 220, first day null images smoothing process acquisition first smoothed image to binaryzation.

Preferably, described smoothing processing can be carried out in the following way:

First carry out the opening operation in morphological image to image with identical structural element, and then carry out closed operation, the dimension of structural element matrix is determined according to sky image resolution, and shape is generally circular.In the present embodiment, setting structure element is

$se=\left[\begin{array}{ccccc}0& 0& 1& 0& 0\\ 0& 1& 1& 1& 0\\ 1& 1& 1& 1& 1\\ 0& 1& 1& 1& 0\\ 0& 0& 1& 0& 0\end{array}\right]$

First smoothed image is binary image, as shown in Figure 3.

Should be understood that and other modes well known to those skilled in the art also can be adopted to carry out binaryzation and smoothing processing.

Step 230, for the first smoothed image, travel through image edge pixels point in a predefined manner to obtain edge pixel point coordinate and to count first set B ¹, until all pixels are traversed or first set B ¹in point form closed figures, described edge pixel point is that gray-scale value is 0 and surrounding at least exists a gray-scale value is the pixel of 1.

Particularly, from the matrix upper left corner of the first smoothed image, first from left to right order by row, then by row order from top to bottom, judges according to the value of each coordinate points on image, confirms that whether it is the pixel at cloud cluster edge.

If a pixel (x ₀, y ₀) meet: $F_{x_0,y_0}^{\′}=0$ And then this pixel is labeled as edge pixel point, is counted first set B ¹.Also namely, edge pixel point is that gray-scale value is 0 and surrounding at least exists a gray-scale value is the pixel of 1.

Coordinate figure being exceeded to the situation of image range, the pixel gray-scale value at this coordinate place can be ignored when calculating.

Preferably, the edge that can be labeled from first is lighted (on the left of cloud cluster the top first marginal point), with this edge pixel point for benchmark, if its coordinate is (x _a, y _a), then by (x _a-1, y _a), (x _a-1, y _a+1), (x _a, y _a+1), (x _a+1, y _a+1), (x _a+1, y _a), (x _a+1, y _a-1), (x _a, y _a-1), (x _a-1, y _a-1) sequentially judging whether successively to meet: a, this point are marginal point; B, this pixel coordinate are not also counted into first set.If meet a, b two condition, extract this pixel coordinate and enter first set, and using this point as new reference point, repeat above-mentioned steps, until all pixels are traversed or first set forms closed figures.

Similarly, step 200 can comprise the process of the second sky image:

Step 240, the second sky image binaryzation is obtained the second sky image of binaryzation.

Step 250, the second sky image smoothing process acquisition second smoothed image to binaryzation.

Second smoothed image as shown in Figure 3.

Step 260, for the second smoothed image, travel through image edge pixels point in a predefined manner to obtain edge pixel point and to count second point set B ², until all pixels are traversed or second point set B ²in point form closed figures, described edge pixel point is that gray-scale value is 0 and surrounding at least exists a gray-scale value is the pixel of 1.

Above-mentioned steps 210-230 and step 240-260 can carry out simultaneously, also can carry out in a predetermined order.

Thus, namely all pixel coordinates that can obtain the first smoothed image medium cloud group image border and the second smoothed image medium cloud group image border by step 210-230 and step 240-260, also obtain first set B ¹with second point set B ².

Step 300, calculating first set B ¹to described second point set B ²distance matrix D, the element D of described distance matrix _i,jfor described first set B ¹i-th point to described second point set B ²the distance of jth point.

Particularly, initial pictures and displacement diagram are respectively B as the set of cloud cluster edge pixel point coordinate ¹, B ², wherein B ¹middle element number is m, B ²middle element number is n, then

$D_{i,j}=dis\tan ce(B_i^1,B_j^2)$ (distance represent the two Euclidean distance)

i＝1,2,...,m；j＝1,2,...,n

Wherein $B_i^1=(x_i^1,y_i^1),B_j^2=(x_j^2,y_j^2)$

Then $dis\tan ce(B_i^1,B_j^2)=\sqrt{{(x_i^1-x_j^2)}^2+{(y_i^1-y_j^2)}^2}$

Thus, the line number of Distance matrix D is m, and columns is n.

Should be understood that Distance matrix D is not limited to adopt Euclidean distance, also can adopt other known distance account form.

Step 400, with minimize the shift length sum of all pixels of cloud cluster image border and the shift length variance minimizing all pixels of cloud cluster image border for optimization aim, ask for transposed matrix based on optimization algorithm, described transposed matrix is for characterizing described first set B ¹with described second point set B ²corresponding relation.

The corresponding relation computation model of the embodiment of the present invention is set up based on considering as follows, namely, for the transposed matrix E of cloud cluster edge pixel point, because inter-picture temporal interval is extremely short, be not enough to allow cloud cluster carry out complicated motion, therefore the transposed matrix that final mask is tried to achieve answers corresponding the most simple and quick particle clouds motion process, namely requires that cloud cluster edge each point shift length sum is minimum.On the other hand, the motion of cloud cluster is subject to the wind field in region and the impact of cloud cluster self inertia residing for it, under minute level time scale, between image, the possible range of movement of cloud cluster is also restricted, thus can think that the direction of wind field and size are substantially consistent within the scope of the time and space that this is less, now in image, on cloud cluster edge, the motion of each point should be consistent as far as possible, namely requires that the variance of cloud cluster edge each point shift length is minimum.

Wherein, the dimension of transposed matrix E is identical with Distance matrix D, and wherein element is 0 or 1, and

i＝1,2,...,m；j＝1,2,...,n

Based on above-mentioned consideration determine optimization aim be: a, minimize cloud cluster image border shift length sum a little; B, minimize cloud cluster image border shift length variance a little.

Preferably, for the weight coefficient that above-mentioned two optimization aim are different, establishing target function is in a linear fashion:

$\min =\mathrm{\α}\left(\underset{i=1}{\overset{m}{\Σ}}\underset{j=1}{\overset{n}{\Σ}}{E}_{i,j}\·D_{i,j}\right)+\mathrm{\β}\·\mathrm{var}\left(E\·D|{E_{i,j}\·D_{i,j}\≠0}_{}\right)$

Wherein, E is described transposed matrix, E _i,jfor the element of described transposed matrix, D is described distance matrix, D _i,jfor the element of described distance matrix, for { E _i,jd _i,j| i=1,2 ..., m; J=1,2 ..., all in n} is not the variance of the number of 0 value, α and β is weight coefficient.In the present embodiment, setting α=1, β=10.

In embodiments of the present invention, described optimization algorithm can be genetic algorithm.Genetic algorithm (GeneticAlgorithm) is the computation model of the simulation natural selection of Darwinian evolutionism and the biological evolution process of genetic mechanisms, is a kind of method by simulating nature evolutionary process search optimum solution.Genetic algorithm is that a population is then made up of the individuality (individual) of the some of encoding through gene (gene) from representing a population (population) of the potential disaggregation of problem possibility.Each individuality is actually chromosome (chromosome) and is with characteristic entity.Chromosome is as the main carriers of inhereditary material, the i.e. set of multiple gene, its inner performance (i.e. genotype) is certain assortment of genes, which determines the external presentation of individual shape, and the feature as dark hair is determined by certain assortment of genes controlling this feature in chromosome.Therefore, need to realize from phenotype to genotypic mapping and coding work at the beginning.Owing to copying the work of gene code very complicated, we often simplify, as binary coding, after just producing for population, according to the principle of the survival of the fittest and the survival of the fittest, develop by generation (generation) and produce the approximate solution of becoming better and better, in every generation, select (selection) individual according to fitness (fitness) size individual in Problem Areas, and carry out combination intersection (crossover) and variation (mutation) by means of the genetic operator (geneticoperators) of natural genetics, produce the population representing new disaggregation.This process is more adapted to environment for population than former generation by causing the same rear life of kind of images of a group of characters natural evolution, and the optimum individual in last reign of a dynasty population, can as problem approximate optimal solution through decoding (decoding).Genetic algorithm can select the solution making optimization aim optimum in a fairly large number of possibility.

The transposed matrix E obtained based on optimization algorithm can characterize first set B ¹to second point set B ²mapping, also, corresponding relation, each second point set B ²in pixel coordinate can corresponding one or more first set B ¹pixel coordinate

Preferably, constraint condition can also be added in the solution procedure of optimization algorithm, to make solving result more close to the actual movement rule of cloud cluster.

Because particle clouds motion is a continuous deformation process, its edge wheel profile should keep continuous in motion process, the situations such as twisting and breaking of intersecting can not be there is, because cloud cluster edge wheel profile obtains by edge respectively being pressed permanent order connection, therefore require that on edge, the relative position of each point remains constant.Therefore, constraint condition can be set to: if E _i,j=1, then E _{i+1, j}+ E _{i, j+1}+ E _{i+1, j+1}=1.Wherein, if i+1 or j+1 exceeds index range, return the 1st index.

Thus, based on this displacement relation, be linearly moving prerequisite at short notice based on each marginal point, the prediction of cloud cluster movement can be carried out.

Step 500, according to described first set B ¹, described second point set B ²thirdly set B is obtained with described transposed matrix E ³, described thirdly set B ³for characterizing the 3rd moment t ₃the predicted position of cloud cluster image border.Described t ₃moment is relative to described t ₂moment delayed second schedule time Δ t ₂.

Particularly, step 500 can comprise:

Step 510, according to transposed matrix obtain described first set B ¹with second point set B ²in the corresponding relation of pixel.

Step 520, travel through described second point set B ²in pixel, calculate the position obtaining corresponding predict pixel point, and described predict pixel point counted thirdly set B ³.

Wherein, if second point set B ²pixel only correspond to first set B ¹pixel, then based on following formulae discovery predict pixel point position:

$B_k^3=B_j^2+(B_j^2-B_i^1)\frac{{\mathrm{\Δt}}_2}{{\mathrm{\Δt}}_1};$

Wherein, for the pixel coordinate of second point set, for with the pixel coordinate of first corresponding set, for thirdly set B ³in predict pixel point coordinate, i, j, k are the sequence number of each coordinate figure in corresponding point set.

With rectangular coordinate be example, if $B_j^2=(x_j^2,y_j^2),B_i^1=(x_i^1,y_i^1),$ Then $B_k^3=(x_k^3,y_k^3)=(x_j^2+(x_j^2-x_i^1)\frac{{\mathrm{\Δt}}_2}{{\mathrm{\Δt}}_1},y_j^2+(y_j^2-y_i^1\left)\frac{{\mathrm{\Δt}}_2}{{\mathrm{\Δt}}_1}\right).$

If this second point set B ²pixel correspond at least two first set B ¹pixel, then based on following formulae discovery predict pixel point position:

$B_k^3=round\[B_j^2+(B_j^2-\frac{\underset{l=1}{\overset{L}{\Σ}}{B}_{i,l}^1}{L})\frac{{\mathrm{\Δt}}_2}{{\mathrm{\Δt}}_1}\].$

Wherein, L be with first corresponding set B ¹the quantity of pixel, for with the coordinate of i-th pixel in L first corresponding set pixel, round () is round function.

Thus, in traversal second point set B ²in all pixels after, the predict pixel point obtaining all correspondences can be calculated.The thirdly set B be made up of predict pixel point ³define the prediction for particle clouds motion image.

Preferably, the method for the embodiment of the present invention can also comprise step 600, also, according to described thirdly set B ³generate the 3rd sky image of binaryzation to characterize the cloud cluster edge of prediction.3rd sky image (also namely, the 3rd moment t of the binaryzation generated ₃predicted picture for sky image) and at the 3rd moment t ₃the comparison diagram at the sky image edge of actual measurement as shown in Figure 4.

Thus, by first obtaining the pixel set at cloud cluster edge in not sky image in the same time, solve the corresponding relation of the pixel at cloud cluster edge in not sky image in the same time based on optimization algorithm based on predetermined optimization aim, based on the position of the pixel of this corresponding relation and correspondence, subsequent motion position, cloud cluster image border is predicted.Thus, can predict comparatively accurately for particle clouds motion, for a photovoltaic generation power minute level prediction provides Data support.

Predict according to the particle clouds motion Forecasting Methodology of the embodiment of the present invention, in step 100, first day null images as shown in Figure 2 and the second sky image can be obtained

By step 200, edge acquisition is carried out to this image of the first sky and the second sky image and can obtain first following set B ¹with second point set B ²:

B ¹for (abscissa value of the corresponding each point of odd-numbered line, the ordinate value of the corresponding each point of even number line):

B ²for (abscissa value of the corresponding each point of odd-numbered line, the ordinate value of the corresponding each point of even number line):

By step 300-500, the thirdly set B that prediction obtains ³as follows (abscissa value of the corresponding each point of odd-numbered line, the ordinate value of the corresponding each point of even number line):

Comparison diagram based on Fig. 4 is known, and the detection of the embodiment of the present invention has good accuracy.Sky image particle clouds motion Forecasting Methodology in the present invention is by being decomposed into the combination of the linear movement of each marginal point by the nonlinear motion of cloud cluster entirety, achieve sky image particle clouds motion nonlinear prediction, compared with the existing forecast model based on linear extrapolation, the change of shape of cloud cluster in motion process can be identified better.

The foregoing is only the preferred embodiments of the present invention, be not limited to the present invention, to those skilled in the art, the present invention can have various change and change.All do within spirit of the present invention and principle any amendment, equivalent replacement, improvement etc., all should be included within protection scope of the present invention.

Claims (7)

1. a particle clouds motion Forecasting Methodology, comprising:

Obtain first day null images in the first moment, obtain the second sky image in the second moment, described second moment is relative to delayed first schedule time in described first moment;

Obtain first set characterizing described first day null images medium cloud group image border and the second point set characterizing described second sky image medium cloud group image border;

Calculate the distance matrix that described first point set is incorporated into described second point set, the element D of described distance matrix _i,jfor described first i-th point gathered is to the distance of a jth point of described second point set;

With minimize the shift length sum of all pixels of cloud cluster image border and the shift length variance minimizing all pixels of cloud cluster image border for optimization aim, ask for transposed matrix based on optimization algorithm;

Described transposed matrix, for characterizing the corresponding relation of described first set and described second point set, is two-valued function matrix; The element E of described transposed matrix _i,jvalue shows when being 1 that i-th point of described first set is corresponding with a jth point of described second point set, the element E of described transposed matrix _i,jvalue shows when being 0 that i-th point of described first set and the jth of described second point set are put without corresponding relation;

Obtain according to described first set, described second point set and described transposed matrix and thirdly gather, the described predicted position thirdly gathered for characterizing the 3rd moment cloud cluster image border, described 3rd moment is relative to delayed second schedule time in described second moment.

2. particle clouds motion Forecasting Methodology according to claim 1, is characterized in that, obtains first set characterizing described first day null images medium cloud group image border and comprises:

First day null images binaryzation is obtained the first day null images of binaryzation;

To first day null images smoothing process acquisition first smoothed image of binaryzation;

For the first smoothed image, travel through image edge pixels point in a predefined manner to obtain edge pixel point coordinate and to count first set, until all pixels are traversed or point in first set forms closed figures, described edge pixel point is that gray-scale value is 0 and surrounding at least exists a gray-scale value is the pixel of 1.

3. particle clouds motion Forecasting Methodology according to claim 2, is characterized in that, obtains the second point set characterizing described second sky image medium cloud group image border and comprises:

Second sky image binaryzation is obtained the second sky image of binaryzation;

To the second sky image smoothing process acquisition second smoothed image of binaryzation;

For the second smoothed image, travel through image edge pixels point in a predefined manner to obtain edge pixel point coordinate and to count second point set, until all pixels are traversed or first set forms closed figures, described edge pixel point is that gray-scale value is 0 and surrounding at least exists a gray-scale value is the pixel of 1.

4. particle clouds motion Forecasting Methodology according to claim 1, is characterized in that, described optimization algorithm is genetic algorithm, and its objective function is:

$\min =\mathrm{\α}\left(\underset{i=1}{\overset{m}{\Σ}}\underset{j=1}{\overset{n}{\Σ}}{E}_{i,j}\·D_{i,j}\right)+\mathrm{\β}\·\mathrm{var}(E\·D{|}_{E_{i,j}\·D_{i,j}\≠0})$

Wherein, E is described transposed matrix, E _i,jfor the element of described transposed matrix, D is described distance matrix, D _i,jfor the element of described distance matrix, for { E _i,jd _i,j| i=1,2 ..., m; J=1,2 ..., all in n} is not the variance of the number of 0 value, α and β is weight coefficient, and m is the quantity of pixel in first set, and n is the quantity of pixel in second point set.

5. particle clouds motion Forecasting Methodology according to claim 4, is characterized in that, describedly asks for transposed matrix based on optimization algorithm and comprises:

If with E _i,j=1, then E _{i+1, j}+ E _{i, j+1}+ E _{i+1, j+1}=1, for constraint condition asks for transposed matrix based on genetic algorithm, wherein, if i=m, make i+1=1, if j=n, make j+1=1.

6. particle clouds motion Forecasting Methodology according to claim 1, is characterized in that, obtains thirdly set comprise based on described first set, described second point set and described transposed matrix:

The corresponding relation of the pixel in described first set and second point set is obtained according to transposed matrix;

Travel through the pixel in described second point set, calculate the position obtaining corresponding predict pixel point, and described predict pixel point is counted thirdly gather;

Wherein, if the pixel of second point set only corresponds to the pixel of one first set, then based on following formulae discovery predict pixel point position:

$B_k^3=B_j^2+(B_j^2-B_i^1)\frac{{\mathrm{\Δt}}_2}{{\mathrm{\Δt}}_1};$

Wherein, for the pixel coordinate in second point set, for with pixel coordinate in first corresponding set, for the described predict pixel point coordinate in described thirdly set, i, j, k are the sequence number of each pixel coordinate in corresponding point set; Δ t ₁for described first schedule time, Δ t ₂for described second schedule time;

If the pixel of this second point set corresponds to the pixel at least two first set, then based on following formulae discovery predict pixel point position:

$B_k^3=round\[B_j^2+(B_j^2-\frac{\underset{l=1}{\overset{L}{\Σ}}{B}_{i,l}^1}{L})\frac{{\mathrm{\Δt}}_2}{{\mathrm{\Δt}}_1}\];$

Wherein, L be with the quantity of the pixel of first corresponding set, for with the coordinate of i-th pixel in L first corresponding set pixel, round () is round function.

7. particle clouds motion Forecasting Methodology according to claim 1, is characterized in that, described method also comprises:

The 3rd sky image of binaryzation is generated to characterize the cloud cluster edge of prediction according to described thirdly set.