CN110009035A - A kind of air measuring station group space clustering method based on images match - Google Patents
A kind of air measuring station group space clustering method based on images match Download PDFInfo
- Publication number
- CN110009035A CN110009035A CN201910266398.6A CN201910266398A CN110009035A CN 110009035 A CN110009035 A CN 110009035A CN 201910266398 A CN201910266398 A CN 201910266398A CN 110009035 A CN110009035 A CN 110009035A
- Authority
- CN
- China
- Prior art keywords
- wind speed
- measuring station
- air measuring
- indicatrix
- pixel
- 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.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
Landscapes
- Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- Business, Economics & Management (AREA)
- Physics & Mathematics (AREA)
- Economics (AREA)
- Human Resources & Organizations (AREA)
- Strategic Management (AREA)
- General Physics & Mathematics (AREA)
- Entrepreneurship & Innovation (AREA)
- Artificial Intelligence (AREA)
- Tourism & Hospitality (AREA)
- Operations Research (AREA)
- General Business, Economics & Management (AREA)
- Marketing (AREA)
- Game Theory and Decision Science (AREA)
- Development Economics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Quality & Reliability (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Paper (AREA)
- Length Measuring Devices By Optical Means (AREA)
- Investigating Materials By The Use Of Optical Means Adapted For Particular Applications (AREA)
- Indicating Or Recording The Presence, Absence, Or Direction Of Movement (AREA)
Abstract
The invention discloses a kind of air measuring station group space clustering method based on images match, comprising: step 1, for air measuring station each in target area, obtain historical wind speed data to construct wind speed sample；Step 2, multiple and different sample moments is obtained in historical time section, by calculating the wind speed incidence coefficient of the target area in each sample moment section, select the highest 3 sample moment sections of wind speed incidence coefficient as RGB component to construct the wind speed characteristic pattern of air measuring station；Step 3, gray scale wind speed characteristic pattern is obtained to the processing of wind speed characteristic pattern gray processing；Step 4, gray scale wind speed characteristic pattern is divided into several mutually disjoint regions, extracts the wind speed indicatrix in each region；Step 5, realize that air measuring station clusters using the wind speed indicatrix of each air measuring station.The present invention sufficiently excavates the conjunction coupling relationship between wind speed big data to air measuring station clustering class, reduces the mottled degree of data, improves the science and practicability of air measuring station clustering method.
Description
Technical field
The invention belongs to air measuring station clustering class fields, and in particular to a kind of air measuring station group space cluster based on images match
Method.
Background technique
Strong wind is common one of extreme weather.Burst strong wind seriously affects the stabilization and access electricity of Power Output for Wind Power Field
The safety of net.Wind speed advanced prediction is one of the effective ways for solving burst strong wind.The research of current wind speed forecasting method is main
Intelligent algorithm and Time Series Method are concentrated on, needs that a large amount of air measuring stations are arranged in target prediction region, acquires wind speed big data
Carry out the training and foundation of forecasting wind speed model.
But wind-powered electricity generation field areas occupied area is wide, terrain is changeable, and atmospheric flow field is complicated.Discrete point of each air measuring station
Cloth leads to the mottled of wind speed big data collected in region, and overall relevance is low, can not uniformly establish high-precision wind speed
Prediction model.The air measuring station of high correlation can be assembled by the air measuring station cluster of distribution discrete in region, be conducive to subsequent point
Safety pin uniformly establishes forecasting wind speed model to each air measuring station under each classification, and each classification correspondence obtains a forecasting wind speed mould
Type, each forecasting wind speed model carries out forecasting wind speed to the air measuring station point under respective classes, pre- so as to effectively promote wind speed
Survey the generalization ability and precision of prediction of model.
The research of air measuring station clustering method at present, merely with the association journey between air measuring station between historical wind speed time series
Degree carries out air measuring station cluster, and method is single, and research face is narrow, less effective.
Summary of the invention
In view of air measuring station cluster presently, there are the above problem, the present invention provides a kind of air measuring station group based on images match
Spatial clustering method merges the variation in target prediction regional atmospheric flow field, forms wind speed characteristic image, utilizes the side of image procossing
Method sufficiently excavates the conjunction coupling relationship between wind speed big data, by air measuring station clustering class, reduces the number under each cluster classification
According to mottled degree, improve the science of air measuring station clustering method, with further improve it is subsequent for each cluster classification divide
The forecasting wind speed model that do not establish carries out the precision of prediction of forecasting wind speed to the air measuring station point under corresponding cluster classification.
To realize the above-mentioned technical purpose, the present invention adopts the following technical scheme:
A kind of air measuring station group space clustering method based on images match, comprising the following steps:
Step 1, for air measuring station each in target area, historical wind speed data is obtained to construct wind speed sample；
Obtain the Num of each air measuring station in target area1When each sample of a elevational position in identical historical time section
The air speed data at quarter constructs each air measuring station in the wind speed sample sequence of each elevational position；Each air measuring station whole height above sea level position
The wind speed sample sequence set constitutes wind speed sample corresponding with air measuring station；
Step 2, for each air measuring station, all in accordance with wind speed sample acquisition wind speed characteristic pattern；
Step 2.1, different sample moment sections is obtained from historical time section, when each sample moment section is by history
Between continuous Num in section1A sample moment is constituted；
Step 2.2, the wind speed incidence coefficient of the target area in each sample moment section is calculated, selects wind speed incidence coefficient most
3 high sample moment section c1、c2、c3；
Step 2.3, by j-th of elevational position of air measuring station in first sample moment section c1K-th of sample moment wind
Fast data are as the pixel pixel in the air measuring station wind speed characteristic patternj,kR component, j-th of elevational position of air measuring station is existed
Second sample moment section c2K-th of sample moment air speed data as the pixel in the air measuring station wind speed characteristic pattern
pixelj,kG component, by j-th of elevational position of air measuring station in third sample moment section c3K-th of sample moment wind speed
Data are as the pixel pixel in the air measuring station wind speed characteristic patternj,kB component；
Step 3, for the wind speed characteristic pattern of each air measuring station, progress gray processing processing obtains gray scale wind speed characteristic pattern；
Step 4, for the gray scale wind speed characteristic pattern of each air measuring station, it is divided into num2A mutually disjoint region, mentions
The wind speed indicatrix for taking each region constitutes the wind speed indicatrix of air measuring station；
Step 5, using the wind speed indicatrix of each air measuring station, air measuring station cluster is realized；
Step 5.1, for the wind speed indicatrix of each air measuring station, the wind speed feature histogram of air measuring station is obtained；
Wind speed indicatrix is divided into 4 classes by direction, is divided into 3 classes by length, so that building obtains 12 wind speed feature classes,
Respectively feam, m=1,2,3 ... 12；Then the wind speed indicatrix of air measuring station is distributed to corresponding wind speed feature class, and
The quantity for recording the wind speed indicatrix that each wind speed feature class includes constitutes the wind speed feature histogram of air measuring station；
Step 5.2, using each air measuring station as element to be clustered, using wind speed feature histogram as the generalized coordinates of element,
Simultaneously by the quantity and cluster centre of preset rules setting clustering cluster；For each element to be clustered, element to be clustered is calculated
The distance between cluster centre of each clustering cluster, according to the distance between element to be clustered and each cluster centre to each air measuring station
It is clustered.
This programme is by calculating the wind speed incidence coefficient of target area, when selecting highest 3 samples of wind speed incidence coefficient
Carve section, using 3 sample moment sections in each elevational position of air measuring station wind speed sample sequence as RGB component, build
Vertical fusion strong wind is special in the wind speed of the elevational position Spatial Dimension represented and the changing rule of the time dimension of wind speed timing representative
Sign figure, is equivalent to and the flow path of strong wind and changing rule is fully recorded in wind speed characteristic pattern；Further to wind
Fast characteristic pattern carries out gray processing processing and extraction indicatrix obtains the wind speed feature histogram of air measuring station, to realize similar wind speed
The cluster of air measuring station group is realized in the matching of characteristic pattern.Therefore, this programme is changed carries out currently with single historical wind speed sequence
The method of air measuring station cluster is excavated the association syncretic relation between wind speed big data sufficiently to cluster to each air measuring station, is subtracted
The mottled degree of a small number of evidences improves the science and practicability of air measuring station clustering method.
In addition, being clustered by this programme to air measuring station group, high degree is improved between same category air measuring station
Overall relevance reduces the mottled degree of wind speed big data, and the high air measuring station of matching degree uniformly establishes forecasting wind speed model, drops
Low calculating cost, improves the generalization ability and precision of prediction of prediction model.
Further, the specific of the wind speed incidence coefficient of the target area in this moment of various kinds section is calculated in step 2.2
Process are as follows:
Step A1 successively calculates each survey using the wind speed sample sequence of the coefficient of variation and each elevational position of all air measuring stations
The most stable state elevational position at wind station；Wherein, most stable state elevational position refers to corresponding to the smallest wind speed sample sequence of the coefficient of variation
Elevational position, the calculation formula of coefficient of variation CV are as follows:
In formula, xi, i=1,2,3 ..., n indicates the air speed value of i-th of sample moment in wind speed sample sequence；N indicates wind
The number of sample moment in fast sample sequence；Indicate the average value of all sample moment air speed values in wind speed sample sequence；
The most stable state elevational position ascending order of each air measuring station is arranged, selects intermediate value as target area most stable state by step A2
Elevational position is denoted as pos1Position；
Step A3, using each air measuring station in current sample in the wind speed sample sequence of target area most stable state elevational position
Air speed data in moment section is calculated the wind speed Pearson came relative coefficient between any two air measuring station, will own
The wind speed Pearson came relative coefficient summation of any two air measuring station is averaged in air measuring station, as the sample moment currently calculated
The wind speed incidence coefficient of the target area in section.
This programme quantifies each air measuring station in the stability of Different Altitude position using the coefficient of variation, and it is most steady to obtain target area
The wind series sample of state elevational position can reduce the influence of wind speed random disturbances as processing data, and it is poly- to improve air measuring station
The precision of prediction of forecasting wind speed in the accuracy and clustering cluster of class；Using wind speed Pearson came relative coefficient as target area
Wind speed incidence coefficient, and the high sample moment section of wind speed incidence coefficient indicates this period region of interest within that there are wind speed and winds
To apparent strong wind, therefore 3 sample moment sections for selecting wind speed incidence coefficient high are logical respectively as the RGB of wind speed characteristic pattern
Road is able to ascend the accuracy of air measuring station cluster, improves the precision of prediction of forecasting wind speed to carry out air measuring station cluster；On the contrary,
The low air measuring station cluster result that will lead to of wind speed incidence coefficient is vulnerable to random disturbances such as measurement errors.
Further, the method for the wind speed indicatrix in each region in gray feature figure is extracted in step 4 specifically:
Step B1 randomly chooses arbitrary characteristics pixel, the starting point as a wind speed indicatrix；
Step B2 judges in 8 pixels adjacent with the feature pixel of wind speed indicatrix least significant end, if contain
It is not belonging to the feature pixel of current wind speed indicatrix, if so, the selection highest feature pixel of gray value, as current wind
Next point of fast indicatrix, enters step B2, if it is not, obtaining a wind speed indicatrix；
Whether step B3 judges containing the feature pixel for being not belonging to any wind speed indicatrix in region, if so, selection
Wherein starting point of the arbitrary characteristics pixel as a wind speed indicatrix repeats step B2 and step B3, if it is not, entering step
Rapid B4；
Second threshold σ is arranged in step B42, reject of length no more than second threshold σ2Wind speed indicatrix；
Third threshold value σ is arranged in step B53, successively select any 2 wind speed features containing same characteristic features pixel bent
The length of line, 2 wind speed indicatrixes is denoted as length respectively1And length2, and length1≤length2, same characteristic features picture
The quantity of vegetarian refreshments is denoted as Num3If meeting Num3≥length1*σ3, rejecting length is length1Wind speed indicatrix；
The length of the wind speed indicatrix is the quantity for the feature pixel that wind speed indicatrix includes.
Further, in the step 5.2, using Pasteur's distance and k-means means clustering algorithm to each air measuring station into
Row cluster, method particularly includes:
Step C1, using each air measuring station as element to be clustered, using wind speed feature histogram as the generalized coordinates of element, with
Machine selects Num4For a element as cluster centre, each cluster centre represents 1 clustering cluster；
Step C2 randomly selects a certain element, calculates its Pasteur's distance with each cluster centre, and by the Elemental partition
To Pasteur's clustering cluster representated by the nearest cluster centre between the element, immediately by the poly- of the increased clustering cluster of element
Class center is updated to the average coordinates of all elements in clustering cluster, randomly selects next element, repeats step C2, until completing
The distribution of all elements.
Further, in the quantity Num that step C1 is cluster centre4Initial value 3 is set, is also set after executing step C2
Set step C3:
Judge current Num4Whether a clustering cluster is otherwise stable state clustering cluster if then clustering completion enables Num4=
Num4+ 1, and repeat C1- step C3；Wherein, stable state clustering cluster refers to, the Pasteur in clustering cluster between each element and cluster centre
The maximum value of distance is no more than the 30% of the intermediate value of Pasteur's distance between each element and cluster centre.
Further, calculation formula of the air measuring station with Pasteur's distance of cluster centre are as follows:
In formula, Pasteur's distance between Dist (coor1, coor2) indicates coordinate coor1 and coordinate coor2, pcoor1
(feam) and pcoor2(feam) successively indicate wind speed feature class feamIt is general in the element that coordinate coor1 and coordinate coor2 is represented
Rate, and
Further, 4 class directions of wind speed indicatrix are respectively as follows: 0 °, 45 °, 90 °, 135 ° in the step 5.1；Institute
The direction for stating wind speed indicatrix refers to: in wind speed indicatrix, the mode in the direction between all adjacent feature pixels；Institute
It states in wind speed indicatrix, the direction between the adjacent feature pixel of any two is, straight line where two feature pixels the
Angle between one quadrant or the second quadrantal heading and horizontal line right direction；
3 class length of wind speed indicatrix are respectively low length, middle length, Gao Changdu, the low length in the step 5.1
Degree refers to indicatrix length in section [6,10], and middle length refers to indicatrix length in section [11,15], and middle length refers to
Indicatrix length is at section [16 ,+∞].
Further, the step 3 carries out gray processing processing as follows:
hj,k=r 'j,k=g'j,k=b'j,k=rj,k*wr+gj,k*wg+bj,k*wb,
wr+wg+wb=1,
In formula, hj,kIndicate pixel pixel in gray scale wind speed characteristic patternj,kGray scale, rj,k、gj,k、bj,kRespectively indicate picture
Vegetarian refreshments pixelj,kR component, G component and B component before gray processing processing, r 'j,k、g'j,k、b′j,kRespectively indicate pixel
pixelj,kIn gray processing treated R component, G component and B component, wr、wg、wbRespectively indicate R component, G component and B component
Gray processing weight.
Further, before step 4 extracts the wind speed indicatrix in each region, place first is filtered to each region
Reason: obtaining the mean value of all pixels point gray value in region, sets first threshold factor sigma1, by the mean value and first threshold system
Number σ1Product as first threshold；The pixel that gray value is less than first threshold is non-feature pixel, and rest of pixels point is
The gray value of feature pixel non-in region is set 0 by feature pixel.
This programme is by being filtered each region in wind speed characteristic pattern, the resolution ratio of feature pixel in raising region,
To improve the accuracy of cluster.
Further, the Num1A elevational position refers to 0.5 meter of elevational position as starting, with 0.5 meter for step-length, with
Elevational position Num1/ 2 meters are the one group of arithmetic progression terminated.
Beneficial effect
This programme is by calculating the wind speed incidence coefficient of target area, when selecting highest 3 samples of wind speed incidence coefficient
Carve section, using 3 sample moment sections in each elevational position of air measuring station wind speed sample sequence as RGB component, build
Vertical fusion strong wind is special in the wind speed of the elevational position Spatial Dimension represented and the changing rule of the time dimension of wind speed timing representative
Sign figure, is equivalent to and the flow path of strong wind and changing rule is fully recorded in wind speed characteristic pattern；Further to wind
Fast characteristic pattern carries out gray processing processing and extraction indicatrix obtains the wind speed feature histogram of air measuring station, to realize similar wind speed
The cluster of air measuring station group is realized in the matching of characteristic pattern.Therefore, this programme is changed carries out currently with single historical wind speed sequence
The method of air measuring station cluster is excavated the association syncretic relation between wind speed big data sufficiently to cluster to each air measuring station, is subtracted
The mottled degree of a small number of evidences improves the science and practicability of air measuring station clustering method.
In addition, being clustered by this programme to air measuring station group, high degree is improved between same category air measuring station
Overall relevance reduces the mottled degree of wind speed big data, and the high air measuring station of matching degree uniformly establishes forecasting wind speed model, drops
Low calculating cost, improves the generalization ability and precision of prediction of prediction model.
Detailed description of the invention
Fig. 1 is the flow diagram of the method for the invention.
Specific embodiment
Elaborate below to the embodiment of the present invention, the present embodiment with the technical scheme is that according to development,
The detailed implementation method and specific operation process are given, is further explained explanation to technical solution of the present invention.
As shown in Figure 1, a kind of air measuring station group space clustering method based on images match provided by the invention, including it is following
Step:
Step 1, for air measuring station each in target area, acquisition historical wind speed data is to construct wind speed sample；
There is Num in target area1A air measuring station, each air measuring station can acquire the wind speed number of Different Altitude position
According to.
With identical sample frequency, successively acquire in identical historical time section, each air measuring station is in Num1The wind of a elevational position
Fast data obtain the historical wind speed sequence that each air measuring station corresponds to Different Altitude position.Wherein, sample frequency is to sample 1 time for 5 seconds,
The time span of historical time section is not less than Num1/ 20 hours；Num1A elevational position refers to 0.5 meter Wei Qi of elevational position
Begin, with 0.5 meter for step-length, with elevational position Num1/ 2 meters are the one group of arithmetic progression terminated.
Successively by all historical wind speed sequences of above-mentioned each air measuring station, using the wind speed maximum value in time interval T as sample
This moment wind speed obtains the wind speed sample sequence of corresponding Different Altitude position, forms the wind speed sample of corresponding air measuring station.Wherein, T
Value be 1 minute.
Therefore, each air measuring station has a wind speed sample sequence, Num in each elevational position respectively1A wind speed sample sequence
Column constitute wind speed sample corresponding with air measuring station.
In the present embodiment, Num1Value be 100.
Step 2, for each air measuring station, all in accordance with wind speed sample acquisition wind speed characteristic pattern；
Step 2.1, different sample moment sections is obtained from historical time section, when each sample moment section is by history
Between continuous Num in section1A sample moment is constituted, and therefore, two adjacent sample moment sections have the Num of overlapping1- 1 sample
This moment.
Step 2.2, the wind speed incidence coefficient of the target area in each sample moment section is calculated, selects wind speed incidence coefficient most
3 high sample moment sections, and 3 sample moment sections are followed successively by c by wind speed incidence coefficient from high to low1、c2、c3。
Wherein, the detailed process of the wind speed incidence coefficient of the target area in each sample moment section is calculated are as follows:
Step A1 successively calculates each survey using the wind speed sample sequence of the coefficient of variation and each elevational position of all air measuring stations
The most stable state elevational position at wind station；Wherein, most stable state elevational position refers to corresponding to the smallest wind speed sample sequence of the coefficient of variation
Elevational position, the calculation formula of coefficient of variation CV are as follows:
In formula, xi, i=1,2,3 ..., n indicates the air speed value of i-th of sample moment in wind speed sample sequence；N indicates wind
The number of sample moment in fast sample sequence；Indicate the average value of all sample moment air speed values in wind speed sample sequence；
The most stable state elevational position ascending order of each air measuring station is arranged, selects intermediate value as target area most stable state by step A2
Elevational position is denoted as pos1Position；
Step A3, using each air measuring station in current sample in the wind speed sample sequence of target area most stable state elevational position
Air speed data in moment section is calculated the wind speed Pearson came relative coefficient between any two air measuring station, will own
The wind speed Pearson came relative coefficient summation of any two air measuring station is averaged in air measuring station, as the sample moment currently calculated
The wind speed incidence coefficient of the target area in section；
Wind speed Pearson came relative coefficient calculation formula between any two air measuring station is as follows:
In formula, r (ws1, ws2) indicates the wind speed Pearson came relative coefficient between air measuring station ws1 and ws2, ws1i, ws2i,
I=1,2,3 ..., Num1, respectively indicate the air speed data of air measuring station i-th of sample moment of ws1 and ws2；Point
Not Biao Shi all sample moment air speed datas of air measuring station ws1 and ws2 average value.
Step 2.3, the wind speed characteristic pattern of each air measuring station is successively constructed, each wind speed characteristic pattern includes Num1*Num1
A pixel；
In the wind speed characteristic pattern of air measuring station, the true color of pixel is indicated with RGB component, note pixel is pixelj,k,
J=1,2,3 ..., Num1, k=1,2,3 ..., Num1, pixelj,kRepresent the picture of j-th of elevational position, k-th of sample moment
Vegetarian refreshments, pixelj,kR component be j-th of elevational position wind speed sample sequence in c1K-th of sample moment section sample moment
Air speed value, G component be j-th of elevational position wind speed sample sequence in c2The wind of k-th of sample moment section sample moment
Speed value, B component are c in the wind speed sample sequence of j-th of elevational position3The air speed value of k-th of sample moment section sample moment.
Step 3, for the wind speed characteristic pattern of each air measuring station, progress gray processing processing obtains gray scale wind speed characteristic pattern；
The method of the gray processing processing of wind speed characteristic pattern is specific as follows, remembers pixel pixelj,kR points before gray processing processing
Amount, G component and B component are followed successively by rj,k、gj,k、bj,k, remember pixel pixelj,kGray processing treated R component, G component and B
Component is followed successively by r 'j,k、g'j,k、b′j,k, gray processing processing is carried out as follows:
r′j,k=g'j,k=b'j,k=rj,k*wr+gj,k*wg+bj,k*wb,
wr+wg+wb=1,
In formula, wr、wg、wbSuccessively indicate the gray processing weight of R component, G component and B component；
After gray processing processing, gray scale wind speed characteristic pattern is obtained, remembers pixel pixelj,kGray scale be hj,k, there is hj,k=
r′j,k=g'j,k=b 'j,k。
Step 4, for the gray scale wind speed characteristic pattern of each air measuring station, it is divided into num2A mutually disjoint region, mentions
The wind speed indicatrix for taking each region constitutes the wind speed indicatrix of air measuring station；
Step 4.1: being Num with size2*Num2Rectangular window, gray scale wind speed characteristic pattern is divided into several mutually not phases
The region of friendship remembers that the quantity in region is num2；
Step 4.2: successively each region being filtered；
The mean value of all pixels point gray value in region is obtained, first threshold factor sigma is set1, by the mean value and first
Threshold coefficient σ1Product as first threshold；The pixel that gray value is less than first threshold is non-feature pixel, afterimage
Vegetarian refreshments is characterized pixel, and the gray value of feature pixel non-in region is set 0；
Step 4.3: successively extracting the wind speed indicatrix in each region, constitute air measuring station wind speed indicatrix, specific method
Are as follows:
Step B1 randomly chooses arbitrary characteristics pixel, the starting point as a wind speed indicatrix；
Step B2 judges in 8 pixels adjacent with the feature pixel of wind speed indicatrix least significant end, if contain
It is not belonging to the feature pixel of current wind speed indicatrix, if so, the selection highest feature pixel of gray value, as current wind
Next point of fast indicatrix, enters step B2, if it is not, obtaining a wind speed indicatrix；
Whether step B3 judges containing the feature pixel for being not belonging to any wind speed indicatrix in region, if so, selection
Wherein starting point of the arbitrary characteristics pixel as a wind speed indicatrix repeats step B2 and step B3, if it is not, entering step
Rapid B4；
Second threshold σ is arranged in step B42, reject of length no more than second threshold σ2Wind speed indicatrix；
Third threshold value σ is arranged in step B53, successively select any 2 wind speed features containing same characteristic features pixel bent
The length of line, 2 wind speed indicatrixes is denoted as length respectively1And length2, and length1≤length2, same characteristic features picture
The quantity of vegetarian refreshments is denoted as Num3If meeting Num3≥length1*σ3, rejecting length is length1Wind speed indicatrix；
The length of the wind speed indicatrix is the quantity for the feature pixel that wind speed indicatrix includes.
In the present embodiment, first threshold factor sigma1Value be 120%, second threshold σ2Value be 5, third threshold value
σ3Value be 70%.
Step 5, using the wind speed indicatrix of each air measuring station, air measuring station cluster is realized；
Step 5.1, for the wind speed indicatrix of each air measuring station, the wind speed feature histogram of air measuring station is obtained；
Wind speed indicatrix is divided into 4 classes by direction: 0 °, 45 °, 90 °, 135 ° are divided into 3 classes by length: low length, middle length
Degree, Gao Changdu, so that building obtains 12 wind speed feature classes, respectively feam, m=1,2,3 ... 12；Then by air measuring station
Wind speed indicatrix is distributed to corresponding wind speed feature class, and records the number for the wind speed indicatrix that each wind speed feature class includes
Amount, constitutes the wind speed feature histogram of air measuring station.
Wherein, the direction of the wind speed indicatrix refers to: in wind speed indicatrix, between all adjacent feature pixels
Direction mode；Mode refers to the most number of frequency of occurrence, is herein, all adjacent feature pictures in wind speed indicatrix
In direction between vegetarian refreshments, using the most direction of frequency of occurrence as the direction of wind speed indicatrix.The wind speed indicatrix
In, the direction between the adjacent feature pixel of any two is, straight line first quartiles where two feature pixels or second as
Limit the angle between direction and horizontal line right direction；The low length refers to indicatrix length in section [6,10], middle length
Refer to indicatrix length in section [11,15], middle length refers to indicatrix length at section [16 ,+∞].
Step 5.2, using each air measuring station as element to be clustered, using wind speed feature histogram as the generalized coordinates of element,
Each air measuring station is clustered, method particularly includes:
Step C1, using each air measuring station as element to be clustered, using wind speed feature histogram as the generalized coordinates of element, with
Machine selects Num4For a element as cluster centre, each cluster centre represents 1 clustering cluster；
Step C2 randomly selects a certain element, calculates its Pasteur's distance with each cluster centre, and by the Elemental partition
To Pasteur's clustering cluster representated by the nearest cluster centre between the element, immediately by the poly- of the increased clustering cluster of element
Class center is updated to the average coordinates of all elements in clustering cluster, randomly selects next element, repeats step C2, until completing
The distribution of all elements；
Wherein, calculation formula of the air measuring station with Pasteur's distance of cluster centre are as follows:
In formula, Pasteur's distance between Dist (coor1, coor2) indicates coordinate coor1 and coordinate coor2, pcoor1
(feam) and pcoor2(feam) successively indicate wind speed feature class feamIt is general in the element that coordinate coor1 and coordinate coor2 is represented
Rate, and
In addition, the average coordinates of several elements are, mean value successively is taken to the wind speed feature class in wind speed feature histogram.
Step C3, if clustering cluster meets: element is no more than element with poly- with the maximum value of cluster centre distance in clustering cluster
The 30% of the intermediate value of class centre distance, then the clustering cluster is referred to as stable state clustering cluster；If the Num that step C2 is obtained4A clustering cluster is equal
For stable state clustering cluster, cluster is completed, and otherwise, enables Num4=Num4+ 1, repeat step C1- step C3.
Above embodiments are preferred embodiment of the present application, those skilled in the art can also on this basis into
The various transformation of row or improvement these transformation or improve this Shen all should belong under the premise of not departing from the application total design
Within the scope of please being claimed.
Claims (10)
1. a kind of air measuring station group space clustering method based on images match, which comprises the following steps:
Step 1, for air measuring station each in target area, historical wind speed data is obtained to construct wind speed sample；
Obtain the Num of each air measuring station in target area1The wind of each sample moment of a elevational position in identical historical time section
Fast data construct each air measuring station in the wind speed sample sequence of each elevational position；The wind of each air measuring station whole elevational position
Fast sample sequence constitutes wind speed sample corresponding with air measuring station；
Step 2, for each air measuring station, all in accordance with wind speed sample acquisition wind speed characteristic pattern；
Step 2.1, different sample moment sections is obtained from historical time section, each sample moment section is by historical time section
Interior continuous Num1A sample moment is constituted；
Step 2.2, the wind speed incidence coefficient of the target area in each sample moment section is calculated, selects wind speed incidence coefficient highest
3 sample moment section c1、c2、c3；
Step 2.3, by j-th of elevational position of air measuring station in first sample moment section c1K-th of sample moment air speed data
As the pixel pixel in the air measuring station wind speed characteristic patternj,kR component, by j-th of elevational position of air measuring station in the second sample
This moment section c2K-th of sample moment air speed data as the pixel pixel in the air measuring station wind speed characteristic patternj,k
G component, by j-th of elevational position of air measuring station in third sample moment section c3K-th of sample moment air speed data make
For the pixel pixel in the air measuring station wind speed characteristic patternj,kB component；
Step 3, for the wind speed characteristic pattern of each air measuring station, progress gray processing processing obtains gray scale wind speed characteristic pattern；
Step 4, for the gray scale wind speed characteristic pattern of each air measuring station, it is divided into num2A mutually disjoint region is extracted each
The wind speed indicatrix in region constitutes the wind speed indicatrix of air measuring station；
Step 5, using the wind speed indicatrix of each air measuring station, air measuring station cluster is realized；
Step 5.1, for the wind speed indicatrix of each air measuring station, the wind speed feature histogram of air measuring station is obtained；
Wind speed indicatrix is divided into 4 classes by direction, is divided into 3 classes by length, so that building obtains 12 wind speed feature classes, respectively
For feam, m=1,2,3 ... 12；Then the wind speed indicatrix of air measuring station is distributed to corresponding wind speed feature class, and recorded
The quantity for the wind speed indicatrix that each wind speed feature class includes constitutes the wind speed feature histogram of air measuring station；
Step 5.2, using each air measuring station as element to be clustered, using wind speed feature histogram as the generalized coordinates of element, simultaneously
By the quantity and cluster centre of preset rules setting clustering cluster；For each element to be clustered, element to be clustered and each is calculated
The distance between cluster centre of clustering cluster carries out each air measuring station according to the distance between element to be clustered and each cluster centre
Cluster.
2. the method according to claim 1, wherein calculating the target in this moment of various kinds section in step 2.2
The detailed process of the wind speed incidence coefficient in region are as follows:
Step A1 successively calculates each air measuring station using the wind speed sample sequence of the coefficient of variation and each elevational position of all air measuring stations
Most stable state elevational position；Wherein, most stable state elevational position refers to sea corresponding to the smallest wind speed sample sequence of the coefficient of variation
Pull out position, the calculation formula of coefficient of variation CV are as follows:
In formula, xi, i=1,2,3 ..., n indicates the air speed value of i-th of sample moment in wind speed sample sequence；N indicates wind speed sample
The number of sample moment in this sequence；Indicate the average value of all sample moment air speed values in wind speed sample sequence；
The most stable state elevational position ascending order of each air measuring station is arranged, selects intermediate value as target area most stable state height above sea level by step A2
Position is denoted as pos1Position；
Step A3, using each air measuring station in current sample moment in the wind speed sample sequence of target area most stable state elevational position
The wind speed Pearson came relative coefficient between any two air measuring station is calculated in air speed data in section, by all survey wind
The wind speed Pearson came relative coefficient summation of any two air measuring station is averaged in standing, as the sample moment section currently calculated
Target area wind speed incidence coefficient.
3. the method according to claim 1, wherein extracting the wind in each region in gray feature figure in step 4
The method of fast indicatrix specifically:
Step B1 randomly chooses arbitrary characteristics pixel, the starting point as a wind speed indicatrix；
Step B2 judges in 8 pixels adjacent with the feature pixel of wind speed indicatrix least significant end, if containing not belonging to
In the feature pixel of current wind speed indicatrix, if so, the selection highest feature pixel of gray value, special as current wind speed
The next point for levying curve, enters step B2, if it is not, obtaining a wind speed indicatrix；
Whether step B3 judges containing the feature pixel for being not belonging to any wind speed indicatrix in region, if so, selection is wherein
Starting point of the arbitrary characteristics pixel as a wind speed indicatrix repeats step B2 and step B3, if it is not, entering step
B4；
Second threshold σ is arranged in step B42, reject of length no more than second threshold σ2Wind speed indicatrix；
Third threshold value σ is arranged in step B53, successively select any 2 wind speed indicatrixes containing same characteristic features pixel, 2
The length of wind speed indicatrix is denoted as length respectively1And length2, and length1≤length2, same characteristic features pixel
Quantity is denoted as Num3If meeting Num3≥length1*σ3, rejecting length is length1Wind speed indicatrix；
The length of the wind speed indicatrix is the quantity for the feature pixel that wind speed indicatrix includes.
4. the method according to claim 1, wherein in the step 5.2, using Pasteur's distance and k-means
Means clustering algorithm clusters each air measuring station, method particularly includes:
Step C1, it is random to select using wind speed feature histogram as the generalized coordinates of element using each air measuring station as element to be clustered
Select Num4For a element as cluster centre, each cluster centre represents 1 clustering cluster；
Step C2 randomly selects a certain element, calculates its Pasteur's distance with each cluster centre, and by the Elemental partition to
Pasteur's clustering cluster representated by the nearest cluster centre between the element, immediately will be in the cluster of the increased clustering cluster of element
The heart is updated to the average coordinates of all elements in clustering cluster, randomly selects next element, repeats step C2, all until completing
The distribution of element.
5. according to the method described in claim 4, it is characterized in that, in the quantity Num that step C1 is cluster centre4Setting is initial
Value 3 also sets up step C3 after executing step C2:
Judge current Num4Whether a clustering cluster is otherwise stable state clustering cluster if then clustering completion enables Num4=Num4+
1, and repeat C1- step C3；Wherein, stable state clustering cluster refers to, Pasteur's distance in clustering cluster between each element and cluster centre
Maximum value be no more than Pasteur's distance between each element and cluster centre intermediate value 30%.
6. according to the method described in claim 4, it is characterized in that, air measuring station with Pasteur's distance of cluster centre calculation formula
Are as follows:
In formula, Pasteur's distance between Dist (coor1, coor2) indicates coordinate coor1 and coordinate coor2, pcoor1(feam) and
pcoor2(feam) successively indicate wind speed feature class feamProbability in the element that coordinate coor1 and coordinate coor2 is represented, and
7. the method according to claim 1, wherein 4 class directions of wind speed indicatrix are divided in the step 5.1
Not are as follows: 0 °, 45 °, 90 °, 135 °；The direction of the wind speed indicatrix refers to: in wind speed indicatrix, all adjacent feature pictures
The mode in the direction between vegetarian refreshments；In the wind speed indicatrix, the direction between the adjacent feature pixel of any two is two
Angle where a feature pixel between straight line first quartile or the second quadrantal heading and horizontal line right direction；
3 class length of wind speed indicatrix are respectively low length, middle length, Gao Changdu in the step 5.1, and the low length is
Refer to indicatrix length in section [6,10], middle length refers to indicatrix length in section [11,15], and middle length refers to feature
Length of curve is at section [16 ,+∞].
8. the method according to claim 1, wherein the step 3 carries out gray processing processing as follows:
hj,k=r 'j,k=g 'j,k=b 'j,k=rj,k*wr+gj,k*wg+bj,k*wb,
wr+wg+wb=1,
In formula, hj,kIndicate pixel pixel in gray scale wind speed characteristic patternj,kGray scale, rj,k、gj,k、bj,kRespectively indicate pixel
pixelj,kR component, G component and B component before gray processing processing, r 'j,k、g′j,k、b′j,kRespectively indicate pixel pixelj,k
In gray processing treated R component, G component and B component, wr、wg、wbRespectively indicate the gray processing of R component, G component and B component
Weight.
9. the method according to claim 1, wherein before step 4 extracts the wind speed indicatrix in each region,
First each region is filtered: obtaining the mean value of all pixels point gray value in region, sets first threshold factor sigma1,
By the mean value and first threshold factor sigma1Product as first threshold；The pixel that gray value is less than first threshold is non-spy
Pixel is levied, rest of pixels point is characterized pixel, the gray value of feature pixel non-in region is set 0.
10. the method according to claim 1, wherein the Num1A elevational position refers to 0.5 meter of elevational position
To originate, with 0.5 meter for step-length, with elevational position Num1/ 2 meters are the one group of arithmetic progression terminated.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910266398.6A CN110009035B (en) | 2019-04-03 | 2019-04-03 | Wind measuring station cluster space clustering method based on image matching |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910266398.6A CN110009035B (en) | 2019-04-03 | 2019-04-03 | Wind measuring station cluster space clustering method based on image matching |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110009035A true CN110009035A (en) | 2019-07-12 |
| CN110009035B CN110009035B (en) | 2020-10-27 |
Family
ID=67169679
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910266398.6A Active CN110009035B (en) | 2019-04-03 | 2019-04-03 | Wind measuring station cluster space clustering method based on image matching |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110009035B (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2022016884A1 (en) * | 2020-07-22 | 2022-01-27 | 江苏科技大学 | Method for extracting sea surface wind speed on basis of k-means clustering algorithm |
| CN115408301A (en) * | 2022-10-31 | 2022-11-29 | 北京千尧新能源科技开发有限公司 | Test set construction method and system for fan simulation |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2003058883A (en) * | 2001-08-10 | 2003-02-28 | Soichi Oka | Feature extracting system for hyper spectrum image |
| CN102163224A (en) * | 2011-04-06 | 2011-08-24 | 中南大学 | Adaptive spatial clustering method |
| CN106355197A (en) * | 2016-08-24 | 2017-01-25 | 广东宝乐机器人股份有限公司 | Navigation image matching filtering method based on K-means clustering algorithm |
| CN108267614A (en) * | 2017-12-29 | 2018-07-10 | 华润电力风能（阳江）有限公司 | A kind of wind detection method and wind measuring system |
| WO2018204524A1 (en) * | 2017-05-03 | 2018-11-08 | Uptake Technologies, Inc. | Computer system & method for predicting an abnormal event at a wind turbine in a cluster |
| CN109034476A (en) * | 2018-07-27 | 2018-12-18 | 中南大学 | A kind of line of high-speed railway extreme wind speeds big data Forecast method |
| CN109102565A (en) * | 2018-07-05 | 2018-12-28 | 内江市下代互联网数据处理技术研究所 | A method of automatically generating Virtual Terrain |
| CN109214581A (en) * | 2018-09-20 | 2019-01-15 | 中南大学 | A kind of Along Railway wind speed forecasting method considering wind direction and confidence interval |
-
2019
- 2019-04-03 CN CN201910266398.6A patent/CN110009035B/en active Active
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2003058883A (en) * | 2001-08-10 | 2003-02-28 | Soichi Oka | Feature extracting system for hyper spectrum image |
| CN102163224A (en) * | 2011-04-06 | 2011-08-24 | 中南大学 | Adaptive spatial clustering method |
| CN106355197A (en) * | 2016-08-24 | 2017-01-25 | 广东宝乐机器人股份有限公司 | Navigation image matching filtering method based on K-means clustering algorithm |
| WO2018204524A1 (en) * | 2017-05-03 | 2018-11-08 | Uptake Technologies, Inc. | Computer system & method for predicting an abnormal event at a wind turbine in a cluster |
| CN108267614A (en) * | 2017-12-29 | 2018-07-10 | 华润电力风能（阳江）有限公司 | A kind of wind detection method and wind measuring system |
| CN109102565A (en) * | 2018-07-05 | 2018-12-28 | 内江市下代互联网数据处理技术研究所 | A method of automatically generating Virtual Terrain |
| CN109034476A (en) * | 2018-07-27 | 2018-12-18 | 中南大学 | A kind of line of high-speed railway extreme wind speeds big data Forecast method |
| CN109214581A (en) * | 2018-09-20 | 2019-01-15 | 中南大学 | A kind of Along Railway wind speed forecasting method considering wind direction and confidence interval |
Non-Patent Citations (4)
| Title |
|---|
| HAGKWEN KIM 等: "Three dimensional clustering in wind farms with storage for reliability analysis", 《2013 IEEE GRENOBLE CONFERENCE》 * |
| 刘辉 等: "基于小波分析法与滚动式时间序列法的风电场风速短期预测优化算法", 《中南大学学报（自然科学版）》 * |
| 史静 等: "聚类分析数据挖掘技术在气象站点布局管理中的应用", 《第32届中国气象学会年会S19气象信息化--机遇与挑战》 * |
| 李松鹤: "洮儿河流域察尔森水库汛期调度方案分析", 《中国优秀硕士学位论文全文数据库 工程科技II辑》 * |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2022016884A1 (en) * | 2020-07-22 | 2022-01-27 | 江苏科技大学 | Method for extracting sea surface wind speed on basis of k-means clustering algorithm |
| CN115408301A (en) * | 2022-10-31 | 2022-11-29 | 北京千尧新能源科技开发有限公司 | Test set construction method and system for fan simulation |
| CN115408301B (en) * | 2022-10-31 | 2023-01-31 | 北京千尧新能源科技开发有限公司 | Test set construction method and system for fan simulation |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110009035B (en) | 2020-10-27 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112766199B (en) | Hyperspectral image classification method based on self-adaptive multi-scale feature extraction model | |
| CN110222769B (en) | Improved target detection method based on YOLOV3-tiny | |
| CN103984946B (en) | High resolution remote sensing map road extraction method based on K-means | |
| CN110929602A (en) | Foundation cloud picture cloud shape identification method based on convolutional neural network | |
| CN103413142B (en) | Remote sensing image land utilization scene classification method based on two-dimension wavelet decomposition and visual sense bag-of-word model | |
| CN109447008A (en) | Population analysis method based on attention mechanism and deformable convolutional neural networks | |
| CN109523013A (en) | A kind of air particle pollution level estimation method based on shallow-layer convolutional neural networks | |
| CN106056591A (en) | Method for estimating urban density through fusion of optical spectrum image and laser radar data | |
| CN108876797A (en) | A kind of image segmentation system and method based on Spiking-SOM neural network clustering | |
| CN111310756A (en) | Damaged corn particle detection and classification method based on deep learning | |
| CN108681718A (en) | A kind of accurate detection recognition method of unmanned plane low target | |
| CN104112062B (en) | The acquisition methods of wind-resources distribution based on interpolation method | |
| CN109409438B (en) | Remote sensing image classification method based on IFCM clustering and variational inference | |
| CN105279769A (en) | Hierarchical particle filtering tracking method combined with multiple features | |
| CN103984963B (en) | Method for classifying high-resolution remote sensing image scenes | |
| CN110009035A (en) | A kind of air measuring station group space clustering method based on images match | |
| CN113887517B (en) | Crop remote sensing image semantic segmentation method based on parallel attention mechanism | |
| CN107679644A (en) | A kind of website Rainfall data interpolating method based on rain types feature | |
| CN102737381B (en) | Image partitioning method based on mixed bipartite graph clustering integration | |
| CN112801332B (en) | Short-term wind speed prediction method based on gray level co-occurrence matrix | |
| CN109635726A (en) | A kind of landslide identification method based on the symmetrical multiple dimensioned pond of depth network integration | |
| CN104751111A (en) | Method and system for recognizing human action in video | |
| CN110866472A (en) | Unmanned aerial vehicle ground moving target identification and image enhancement system and method | |
| CN110334584A (en) | A kind of gesture identification method based on the full convolutional network in region | |
| CN114330516A (en) | Small sample logo image classification based on multi-graph guided neural network model |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |