CN106251026A - Thunder and lightning based on PDBSCAN algorithm closes on trend prediction method - Google Patents
Thunder and lightning based on PDBSCAN algorithm closes on trend prediction method Download PDFInfo
- Publication number
- CN106251026A CN106251026A CN201610674402.9A CN201610674402A CN106251026A CN 106251026 A CN106251026 A CN 106251026A CN 201610674402 A CN201610674402 A CN 201610674402A CN 106251026 A CN106251026 A CN 106251026A
- Authority
- CN
- China
- Prior art keywords
- algorithm
- lightning
- thunder
- period
- ground
- 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.)
- Pending
Links
Classifications
-
- 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"
Abstract
The invention discloses a kind of thunder and lightning based on PDBSCAN algorithm and close on trend prediction method, for the tradition DBSCAN algorithm deficiency to lightning forecasting, select and memory optimization structure these two aspects from parameter assignment, VDBSCAN algorithm and IDBSCAN algorithm are combined, carry out waiting period cluster to lightning data with the PDBSCAN algorithm proposed after compound, by the space center of each maximum thunder and lightning bunch waited under the period is carried out linear fit, try to achieve the space coordinates of the thunder and lightning density maximum generating region of subsequent time.This algorithm not only allows for the thunder and lightning bunch internal density single situation of change, also overcomes DBSCAN algorithm to situation error-prone in the case of variable density, and the parameter of the perfect all kinds of spatial distribution forms being likely to occur thunder and lightning is chosen.Meanwhile, preferably optimize storage organization when algorithm runs, improve operation efficiency.
Description
Technical field
The present invention relates to a kind of thunder and lightning based on PDBSCAN algorithm and close on trend prediction method, belong to lightning forecasting field.
Background technology
Thunder and lightning is a kind of along with high light with the huge natural phenomena thundered.Along with the very fast development of science and technology, people
Understanding to thunder and lightning is constantly deepened, and also its understanding is turned to rationality by revering.The generating process that ground dodges is mainly in cloud layer band
The stepped leader descent that charged particle is formed, forms lightning channel by connection procedure, Return stroke occurs the most again.Its outside
Showing as a large amount of light, the heat discharged during electric charge generation neutralization, air is surprising by instantaneously heated degree, lightning channel
The interior high temperature up to 30000 degrees Celsius.When strong electric current passes through in atmosphere, air on the way is caused to expand suddenly, simultaneously
Pushing air around, the shock wave of formation is then why thunder and lightning sends the huge reason thundered.Thunder and lightning occur macroscopic view because of
Element mainly has: the convection current that vertical layer structure is unstable and the most vigorous, suitable air humidity, lifting trigger.For on microcosmic,
Electrification mechanism for thunder and lightning there is no final conclusion, but various guess has the scope of application and theoretical basis the most widely, such as: the temperature difference rises
The broken electrification of electricity, raindrop and triboelectrification etc..According to statistics, the whole world is annual about there are 16,000,000 thunders and lightnings, the most about occurs every day
44,000 times.The horizontal scale excursion of thunder and lightning is very big, can be from several kms to hundreds of km, and vertical thickness is mostly at 10km
Above.
Data mining technology and geographic information system technology, as two important technologies in information technology, are processing meteorology
Data aspect has extremely important status and effect.Data mining (Data Mining) refers in data base, comprehensive utilization system
Meter method, mode identification technology, artificial intelligence approach, nerual network technique scheduling theory, draw novel, believable, people
Interested and final intelligible knowledge, thus disclose rule, internal relation and the development trend lying in data.Ground
The preferable earth's surfaces of feature such as space characteristics, attribute character and the temporal characteristics that meteorological data can be had by reason information systems technology
Reveal to come, be the effective means realizing data management.Thunder and lightning closes on Forecasting Methodology have a lot, but because thunder and lightning has at random
Property, locality, dispersibility, sudden, instantaneity and these salient features three-dimensionality so that different thunder and lightning Forecasting Methodologies is all
There is the environment that oneself is most suitably used.Clustering algorithm in maintenance data excavation, in conjunction with GIS platform, for thunder and lightning own characteristic, to calculation
Method is optimized, can accomplish quick, convenient, calculate exactly, and meet close on the related request in trend prediction, at thunder
Electricity nowcasting work has the meaning of reality.
Based on tradition DBSCAN (Density-Based Spatial Clustering of Applications with
The features such as Noise) algorithm has abnormal data anti-interference preferable, algorithmic stability.But, when being applied to for Lightning data
Cluster time there is following two significantly limit to:
(1) ground for DBSCAN algorithm dodges radius neighborhood Eps and the assignment mode of minimal amount threshold value MinPts.Not
With under synoptic scale, lightning drop point region there may be the situation of variable density, i.e. when data in each thunder and lightning bunch group that cluster obtains
Between the biggest situation of distance difference.This point makes tradition DBSCAN algorithm be difficult to process, if rule of thumb assignment accuracy
It is difficult to weigh；
(2) storage organization is loaded down with trivial details, takies bigger memory space.Before carrying out ground at times and dodging cluster, need to set up
The R* tree of all data.
Summary of the invention
The technical problem to be solved is to provide a kind of thunder and lightning based on PDBSCAN algorithm and closes on trend prediction side
Method, proposes to be optimized parameter assignment and improve the PDBSCAN clustering algorithm of the speed of service, by this algorithm equity period
Lightning data clusters, and finds out the sudden strain of a muscle bunch of maximum kernel heart, by calculating its space center, in conjunction with the method for fitting of a polynomial,
The core Di Shan space center of prediction subsequent period.Whole Forecasting Methodology serves and improves clustering precision and improve the speed of service
Purpose.
The present invention solves above-mentioned technical problem by the following technical solutions:
The present invention provides thunder and lightning based on PDBSCAN algorithm to close on trend prediction method, it is characterised in that concrete steps are such as
Under:
Step A, utilizes lightning positioning system to detect and dodge data with recording, and the ground sudden strain of a muscle data of record are carried out pre-place
Reason, is divided into each ground waiting the period and dodges data set；
Step B, uses PDBSCAN clustering algorithm to try to achieve each maximum kernel heart sudden strain of a muscle bunch waiting the period and each maximum kernel heart
Space center's coordinate of sudden strain of a muscle bunch；
Step C, space center's coordinate of the maximum kernel heart being tried to achieve step B sudden strain of a muscle bunch carries out fitting of a polynomial, then root
Try to achieve space center's coordinate of the thunder and lightning maximum kernel heart sudden strain of a muscle bunch of lower first-class period according to the multinomial of matching, thus realize thunder and lightning and face
The forecast of nearly trend.
As the further prioritization scheme of the present invention, in step A, the ground sudden strain of a muscle data to record carry out pretreatment, are divided into each
Deng the ground sudden strain of a muscle data set of period, particularly as follows:
Step A-1, chooses the landing point coordinates data persistently occurring the ground of 20 minutes to dodge after shwoot is raw first, generates data
Collection；
Step A-2, the data set generated by A-1 carries out waiting Time segments division, obtains dodging data set with waiting the period accordingly.
As the further prioritization scheme of the present invention, step B uses PDBSCAN clustering algorithm try to achieve and each wait the period
The sudden strain of a muscle bunch of macronucleus heart and space center's coordinate of each maximum kernel heart sudden strain of a muscle bunch, particularly as follows:
Step B-1, according to DBSCAN algorithm, dodges data set to each ground waiting the period obtained in step A and tries to achieve accordingly
Each k-dist figure waiting the period；
Step B-2, the k-dist figure obtained according to B-1, observe the density dodging data in each ground waiting the period and divide different, if single
One density, then find out the point that the slope of curve is undergone mutation at first, dodges number with the ground that the k-dist value that this point is corresponding is such period
Neighborhood Eps according to collection；If the situation of multiple density level occurs, then try to achieve according to the k-dist figure of the ground sudden strain of a muscle data of such period
Corresponding DK figure, and determine neighborhood Eps according to DK map analysis；
Step B-3, sets MinPts, travels through each ground waiting the period and dodges data set, and the Eps searching for each ground flash-point successively is adjacent
Territory, dodges data to each ground waiting the period and carries out cluster calculation, choose optimum cluster result:
Step B-4, according to the optimum cluster result of B-3, it is thus achieved that each maximum kernel heart waiting the period is dodged bunch, and by bunch in
The latitude and longitude coordinates of all members, tries to achieve space center's point coordinates of each maximum kernel heart sudden strain of a muscle bunch.
As the further prioritization scheme of the present invention, step B also includes: with the internal memory building adjacency list in IDBSCAN
Thought replaces R* tree in DBSCAN.
As the further prioritization scheme of the present invention, in step B-1, according to DBSCAN algorithm, each to what step A obtained
Dodge data set try to achieve the corresponding each k-dist figure waiting the period on the ground of period, particularly as follows: i.e. to the day part obtained in step A
Ground dodge all data in data set, ask it to arrive the distance of k evidence, and the distance obtained arranged by order from low to high
Row, i.e. obtain corresponding k-dist figure.
As the further prioritization scheme of the present invention, step B-2 uses based on variable density and with noise application
Space clustering VDBSCAN algorithm, the k-dist figure dodging data according to the ground of this period tries to achieve corresponding DK figure, particularly as follows: extract
The data set of variable density occur, by it, on k-dist figure, data corresponding to Y-axis are successively from the beginning of Section 2, and latter is with front
One is done difference, i.e. obtains the DK figure of correspondence.
The present invention uses above technical scheme compared with prior art, has following technical effect that and the present invention is directed to tradition
The DBSCAN algorithm deficiency to lightning forecasting, selects and memory optimization structure these two aspects from parameter assignment, by VDBSCAN algorithm
It is combined with IDBSCAN algorithm, uses the PDBSCAN algorithm proposed after being combined to carry out lightning data waiting period cluster.This calculation
Method not only allows for the situation that the change of thunder and lightning bunch internal density is single, also overcomes DBSCAN algorithm to error-prone in the case of variable density
Situation, the parameter of the perfect all kinds of spatial distribution forms being likely to occur thunder and lightning is chosen.Meanwhile, algorithm is preferably optimized
Storage organization during operation, improves operation efficiency.By the space center of each maximum thunder and lightning bunch waited under the period is carried out line
Property matching, tries to achieve the space coordinates of the thunder and lightning density maximum generating region of subsequent time.
In the inspection of actual thunder and lightning synoptic process, it was predicted that value differs about with the maximum thunder and lightning bunch space center at lower a moment
3.114km, it was demonstrated that the effectiveness of this method.In terms of operational efficiency, Clustering Effect and precision of prediction three relevant to forefathers
Algorithm is contrasted, and result shows that thunder and lightning is closed on trend prediction and rises by method proposed by the invention to a certain extent in short-term
Arrive the effect optimized.
Accompanying drawing explanation
Fig. 1 is the method flow diagram of the present invention.
Fig. 2 is k-dist schematic diagram.
Fig. 3 is to generate corresponding DK figure according to k-dist figure.
Fig. 4 be dodge adjacency list.
Fig. 5 is thunder and lightning nowcasting model flow figure based on PDBSCAN algorithm.
Fig. 6 be wait period data collection k-dist figure, wherein, (a) is 13:05 to 13:10 data set, (b) be 13:10 extremely
13:15 data set, (c) is 13 15 to 13:20 data sets, and (d) is 13:20 to 13:25 data set.
Fig. 7 is the k-dist figure of variable density situation 13:25 to 13:30.
Fig. 8 is the DK figure of variable density situation 13:25 to 13:30.
Fig. 9 is the period cluster analysis figures such as 13:05 to 13:30, wherein, (a) be 13:05 to 13:10 Huaian Region dodge
Cluster, (b) be 13:10 to 13:15 Huaian Region dodge cluster, (c) be 13:15 to 13:20 Huaian Region dodge cluster, (d)
Be 13:20 to 13 25 Huaian Regions dodge cluster, (e) be 13:25 to 13:30 Huaian Region dodge cluster.
Figure 10 be dodge drop point prognostic chart.
Figure 11 is VDBSCAN Yu PDBSCAN operation time comparison diagram.
Figure 12 is 13:10 to 13:15IDBSCAN dendrogram.
Detailed description of the invention
Below in conjunction with the accompanying drawings and technical scheme is described in further detail by specific embodiment:
The present invention provides a kind of thunder and lightning based on PDBSCAN algorithm to close on trend prediction method, as it is shown in figure 1, for biography
The system DBSCAN algorithm deficiency to lightning forecasting, selects and memory optimization structure these two aspects from parameter assignment, is calculated by VDBSCAN
Method and IDBSCAN algorithm are combined, and use the PDBSCAN algorithm proposed after being combined to carry out lightning data waiting period cluster, lead to
Crossing the space center to each maximum thunder and lightning bunch waited under the period and carry out linear fit, the thunder and lightning density maximum trying to achieve subsequent time is sent out
The space coordinates in raw district.This algorithm not only allows for the situation that the change of thunder and lightning bunch internal density is single, also overcomes DBSCAN algorithm
To situation error-prone in the case of variable density, the parameter of the perfect all kinds of spatial distribution forms being likely to occur thunder and lightning is chosen.
Meanwhile, preferably optimize storage organization when algorithm runs, improve operation efficiency.
Thunder and lightning is predicted the study hotspot of always weather forecast neighborhood, domestic and international expert, scholar propose based on
The forecasting model of data with different and method is various informative.Owing to thunder and lightning is the weather phenomenon under the influence of small mesoscale system,
Its drop point Density Distribution spatially has the feature of high connectivity, and the passage that the core density district of thunder and lightning generation is in time
Change.So, the present invention, with Density Clustering thought as theoretical basis, carries out composite optimization to tradition DBSCAN algorithm, carries
Go out PDBSCAN algorithm.By to the maximal density after cluster the mobile route of space center's point that dodges bunch be fitted, thus
The core ground of forecast subsequent time dodges position.
When occurring due to thunder and lightning, the thunder and lightning in some areas occurs position to change over time.Carrying out DBSCAN
Need before cluster give two parameters be respectively dodge neighborhood Eps and threshold value MinPts, be i.e. expressed as in given neighborhood Eps
Lightning quantity no less than given threshold value MinPts.Some that relate to are given below in algorithm define:
(1) if the Eps neighborhood of object is including at least the object of minimal amount MinPts, then claim this to dodging with liking core
Point；
(2) a given object set D, if P is in the Eps--neighborhood of q, and q is a kernel object, then it is right to claim
As P from object q be direct density up to；
(3) if there is an object chain p1, p2..., Pn, P1=q, Pn=P, for pi∈D(I≤i≤n),pi+1
It is about P from EpsiDensity direct with MinPts up to, then claim object P from object q about Eps and MinPts density up to
(density-reachable), this relation is asymmetric relation；
(4) if object set D exists an object o so that object P and q is about can with MinPts density from Eps
Reaching, object P with q is symmetrical relations about be connected with MinPts density (density-connected), this relation；
(5) find density up to maximal density be connected the set of object.The object at any bunch is not considered as then to make an uproar
Sound point.
The features such as tradition DBSCAN algorithm has abnormal data anti-interference preferable, algorithmic stability.But, when being applied to
During for the cluster of Lightning data, there is following two significantly to limit to:
(1) for the assignment mode of Eps and MinPts of DBSCAN algorithm.Under different weather yardstick, lightning drop point district
Territory there may be the situation of variable density, i.e. when the situation that distance difference between data in each thunder and lightning bunch group that cluster obtains is the biggest.
This point makes tradition DBSCAN algorithm be difficult to process, if single by virtue of experience assignment accuracy is difficult to weigh.
(2) storage organization is loaded down with trivial details, takies bigger memory space.Before carrying out ground at times and dodging cluster, need to set up
The R* tree of all data.
In the present invention, firstly for the assignment problem of Eps, with VDBSCAN (based on variable density and with noise application
Space clustering, Varied Density Based Spatial Clustering of Applications with Noise)
It is optimized by the parameter method for selecting that algorithm is used.I.e., first the thunder and lightning space clustering data of day part are made accordingly
K-dist curve.Day part k-dist curve plotting be by make in lightning cluster set the K of each object closest to
Distance (in the present invention, K is with Common Parameters 4), dashes forward at first by searching slope in k-dist figure produced by each data set
The point become, the neighborhood Eps being notebook data collection with the value of this catastrophe point.As in figure 2 it is shown, A curve is not become close at cluster data
Ideal curve when spending, the value corresponding to point that in figure, the slope of curve is undergone mutation the earliest can be set to Eps.Gather when carrying out space
When the ground of class dodges set generation variable density situation, as shown in B curve in Fig. 2, density stabilized between a, c, e bunch, but bent when b, d, f
Line slope skyrockets.The most then need to generate corresponding DK figure, as shown in Figure 3 according to k-dist figure.By in k-dist figure adjacent two
Point PM、PM+1K-dist difference be defined as DK.Take front a certain proportion of DK value calculate average DK value Ave (P%), this way be for
Get rid of the point on density bend line and abnormity point curve as far as possible.Given threshold value Y, tries to achieve threshold value model centered by Ave (p%)
Enclose, and Ave (p%) ± Y, and as a critical field for DK value.Then according to the data point place beyond threshold range
Curve Property (this point be on k-dist figure easy curve or density turnover curve), find out correspondence radius,
Generate one group of reference radius value, generate corresponding cluster result the most respectively, and then by comparing selection optimum cluster result.
Secondly, for needing each dataset construction R* tree before cluster, operate complicated and time consumption, affect algorithm work efficiency
Problem, this problem is carried out by the thought using in IDBSCAN algorithm the thought according to graph theory to replace R* tree with adjacency list herein
Optimize.As shown in Figure 4, the data structure simple, intuitive of adjacency list, and committed memory is less.Base table and chained list thereof collectively constitute one
Individual adjacency list, base table dodges array pi in representing this period the most all ofly, and chained list thereafter represents that pi is at Eps model
Enclose interior dodging allly.Such as, the 1st element of the 1st row in Fig. 4 represents the 1st thunder and lightning bunch, and this ground dodges in threshold value Eps
In the range of have 4,2,1 these 3 ground dodge events.10th row then represents noise spot.
The present invention is directed to tradition DBSCAN algorithm for 2 problems of thunder and lightning space clustering, by VDBSCAN algorithm and
Advantage module in IDBSCAN algorithm is combined, it is proposed that the density-based algorithms PDBSCAN after Fu He, such as Fig. 5
Shown in, the recombination process of this algorithm, based on the essential idea that DBSCAN algorithm designs, also meets relevant data structure simultaneously
Specification, particularly as follows::
Step 1: data prediction.Set up space lightning cluster attribute data collection W, and increase cluster result Field ID for it
(integer), cluster result ID initial value is zero；Definition ground dodges the search collection S of data, and interim retrieval result is deposited in S.
Step 2: generate the k-dist figure of each data set, if this data set shown in image is single density trend, then should
In figure, the corresponding y numerical value of scope discontinuity is defined as Eps.In the event of multiple density levels, the i.e. situation of variable density, then
DK figure is generated according to k-dist figure.By the analysis of DK figure is determined Eps.In the present invention, another parameter MinPts is according to minimum
Call by value substitutes into and carries out cluster calculation, MinPts=2.
Step 3: traversal W, searches for the Eps neighborhood of each ground flash-point successively, and sets up in adjacency list storage neighborhood for this point
Dodge data point allly.
Step 4: traversal W, successively using each ground flash-point as initial seed point[74]Investigate, step specific as follows:
(1) for ground flash-point pi, if ID is zero, then its adjacency list is searched for；If adjacency list chained list dodges event number
Mesh is more than MinPts, then some pi is core ground flash-point, its ID is set to cluster, the institute simultaneously comprised by the adjacency list of pi
Ground flash memory is had to enter in S；
(2) traversal S, investigates each ground flash-point as seed points, for a qi, if ID is zero, searches for it adjacent
Connecing table, if sudden strain of a muscle number in neighborhood ground is more than MinPts in adjacency list, then qi is also a core point, and it is the closeest of a pi simultaneously
Degree accessible point, belongs to same class with pi, the ID of qi is set to cluster；Otherwise qi is boundary point, but the ID of qi is also set to
cluster.If qi is core ground flash-point, some o is present in qi adjacency list and o is not belonging to S, then be stored in S by an o.?
After will some a qi delete from S；
(3) the next ground flash-point in search S, if S non-NULL, then performs step (2).
Step 5: investigating ground and dodge the next point in data set W, and cluster is added 1, performing step 3, until having traveled through
Data set.
Step 6: delete ground and dodge search collection S.
Step 7: dodge bunch with finding out the relative maximum of day part, calculate its space center's coordinate, utilize least square multinomial
Day part space center coordinate is carried out curve fitting by formula.
Embodiment
It is raw in Huai'an, Jiangsu Province Thunderstorm Weather that the embodiment of the present invention chooses 13:05 to 13:35 distribution on July 6th, 2009
Instance data.Being 118 ° of 12 00 " E～119 ° 36 30 " E with longitude excursion on space scale, latitude excursion is 32 °
43 00 " N～34 ° 06 00 " N, there is thunder and lightning 1478, mean intensity 31.78kA in this period total.By above-mentioned data in the time
Be divided on yardstick every 5 minutes be an interval to divide data set, as shown in table 1.
Table 1 lightning statistical information at equal intervals
| ID | Time started | End time | Lightning frequency |
| 1 | 13:05 | 13:10 | 433 |
| 2 | 13:10 | 13:15 | 390 |
| 3 | 13:15 | 13:20 | 308 |
| 4 | 13:20 | 13:25 | 200 |
| 5 | 13:25 | 13:30 | 147 |
It is first according to the flow process of PDBSCAN algorithm, the data set chosen is made the k-dist figure of correspondence, observes by counting
According to the variation tendency of the curve that each point in collection generates, finding out the density level corresponding to each data set, if occurring, several are close
Degree level, then scheme to generate corresponding DK figure according to the k-dist of this data set, determine the neighborhood Eps carrying out clustering.Concrete outcome
As follows:
From (a) to (d) in Fig. 6, in the data set of inspection, front four k-dist waiting period data point to generate
A figure essentially mild curve.This explanation, front 45 minutes, each clustering cluster is single density data collection.At this point it is possible to it is straight
Connect the curve in the k-dist that comparison generates, find out the point that slope of curve abrupt change occurs at first, by the vertical coordinate respective value of this point
It is provided for the Eps of notebook data clustering.Wherein, on July 6th, 2009,13:05 divided the Eps=0.06 divided to 13:10；13:
10 points of Eps=0.04 divided to 13:15；The Eps=0.03 that 13:15 to 13:20 divides；The Eps=of 13:20 to 13: 25
0.06.This is compared with empirically assignment, the Eps under Severe thunderstorm is set to 8km and (under clustering based on geographical coordinate is i.e.
0.08), there is bigger discrepancy.
As it is shown in fig. 7, the 5th occurred at this thunderstorm 5 minutes, i.e. 13:25 divides to 13:30 and divides, thunder and lightning number now
There is variable density phenomenon according to the k-dist figure generated, curve table reveals obvious two density levels.Therefore, scheme according to k-dist
Make this period corresponding DK figure, as shown in Figure 8.
The present invention takes the DK value of front 90% and calculates average DK value Ave (90%), draws Ave=0.00787, given threshold value Y
=0.008, then put the critical field [0.00787-0.008,0.00787+0.008] of DK value.Because gained DK value is all non-negative
Value, so the critical field of the DK value on curve is actual should be [0,0.01587].Point beyond dotted portion has in fig. 8
P38、P40—P46、P121、P123—P125、P131、P136、P138—P142。
After consecutive points by the point checked and outside these critical fields, find the point around some P39, P121, P137
The most not in critical field, so these three point is not the point on smoothed curve.And the later curve of P138 is according to k-dist
The Curve Property of figure, this point belongs to last catastrophe point that slope sudden change occurs.Therefore on this curve of P136 to P142,
Only consider the situation of P136 1.
Finally, the k-dist character of each point is checked according to this, it is determined that the density bend line of k-dist figure is P38 and P46 connects
Logical curve, P121 with P125 connect the K-dist value corresponding to curve and P131 and P136 generated.Reality from cluster
Setting out, after data set is only taken arithmetic point, 1 degree of approximation of longitude and latitude (is taken 111 kilometers, then passes through calculating, 0.01 about 1 public affairs by 2
In).Show that one group of radius value that these points are corresponding is as shown in Figure 2: 0.02,0.03,0.07.Respectively by above-mentioned 3 clusters
Radius substitutes into the data of this period.
As shown in table 2, it is the actual Clustering Effects that draw of the data that above-mentioned 3 cluster radius substitute into this period, can see
Going out the cluster radius for this data set, to be entered as 0.07 the most suitable.Because under this radius, cluster number is less, Er Qiecheng
Merit cluster numbers is most.
Table 2 different radii Clustering Effect contrasts
| Eps | The number of the ground sudden strain of a muscle bunch of cluster | Number is dodged in bunch |
| 0.02 | 5 | 86 |
| 0.03 | 2 | 105 |
| 0.07 | 2 | 135 |
By (a) in Fig. 9 it can be seen that this choose for checking this method thunder and lightning weather example, 13:05 divide to
13:10 divides and defines one in the middle part of Huai'an and dodge by force band, from figure it can be clearly seen that, this moment ground dodges and is formed
3 core lightnings bunch, maximum of which core ground dodges a bunch major part and is distributed in the middle part of Jinhu County.In Fig. 9, (b) dodges quantity middlely
Reducing, but define 4 lightnings bunch, this embodies randomness and the feature of instantaneity that lightning itself is possessed.Its lightning bunch
Distributed areas there occurs displacement, figure can intuitively being observed, various places are dodged and bunch moves northeastward, maximum core ground dodges and bunch depends on
The most domestic in Jinhu County.In this 3rd stage, i.e. 13:15 to 13:20 tested, by (c) in Fig. 9 it is found that ground dodges number
Amount continues to reduce, and clustering cluster is also reduced to 2, and the sudden strain of a muscle bunch of maximum kernel heart is transferred to east substantially by the middle part of Jinhu County originally
Portion.In Fig. 9, in (d), lightning quantity continues to reduce, and ground sudden strain of a muscle bunch is 3, and the sudden strain of a muscle bunch of maximum kernel heart has the trend offset northeastward.Figure
9 (e) understand ground to dodge number of clusters amount is 2, and lightning quantity relatively this period this moment starts, existing huge minimizing, it is seen that this is used
Data in the method for inspection are the process that Strong Thunderstorm passes by or gradually withers away.After often waiting period cluster, calculate each maximum kernel
The space center of the heart bunch, and the significant datas such as the classification of the lightning bunch of day part, number, maximum bunch space center's coordinate are converged
Always, constitute core ground and dodge Time-space serial, as shown in table 3.
Table 3 respectively waits period core ground to dodge cluster Time-space serial
| ID | Core classes 1 | Core classes 2 | Core classes 3 | Core classes 4 | Maximum bunch space geometry centre coordinate (longitude, latitude) | Noise spot |
| 1 | 359 | 35 | 31 | Nothing | (119.0721,33.1357) | 8 |
| 2 | 256 | 62 | 45 | 10 | (119.1155,33.11448) | 17 |
| 3 | 206 | 95 | Nothing | Nothing | (119.1912,33.06153) | 7 |
| 4 | 129 | 49 | 20 | Nothing | (119.2459,33.04471) | 2 |
| 5 | 100 | 35 | Nothing | Nothing | (119.2724,33.01403) | 11 |
Dodge bunch according to the lightning space-time core tried to achieve of cluster, dodge bunch with finding out the relative maximum of day part, utilize this bunch
The coordinate dodged comprised obtains space center's coordinate of this bunch allly.The method construct utilizing quadratic polynomial matching is multinomial
Formula function y=F (x), dodges Time-space serial carry out curve fitting to core.Trying to achieve equation is y=-0.115635x2+
26.97546x-1539.39023, wherein x position longitude, y is latitude.Then the maximum of lower first-class period is simulated by this fit line
The motion track of core ground sudden strain of a muscle bunch, as shown in Figure 10.
According to matched curve, the core lightning of subsequent time is predicted, its prediction space coordinates be (119.312,
32.9941), the ground of 13:40 dodging data afterwards and carries out cluster analysis and can get similar result, its core ground dodges coordinate and is
(119.3311,32.9967) about 3.114km is differed only by with estimation range distance.Therefore deduce that, base proposed by the invention
Forecasting procedure in PDBSCAN algorithm carries out the short trend prediction that faces to thunder and lightning better effects.
What Figure 11 was given is operational efficiency comparison diagram, and the PDBSCAN algorithm that the present invention proposes is compared with VDBSCAN algorithm
Relatively, the storage form of PDBSCAN algorithm is adjacency list, and the storage form of VDBSCAN algorithm is R* tree.Structure is relatively simple
Adjacency list is when processing magnanimity meteorological data, and advantage is fairly obvious.By being 10 to the order of magnitude of stochastic generation3Spatial and temporal distributions
Point data carries out algorithm speed and compares, find to use the operation efficiency of PDBSCAN algorithm after optimizing bright be higher than traditional
DBSCAN algorithm.
Figure 12 chooses the ground sudden strain of a muscle number of data sets evidence that 13:10 to 13:15 divides, and contrasts with the Clustering Effect of (b) in Fig. 9.
Radius value selected by IDBSCAN is 8km, and the minimum number set is as 2.As shown in Figure 12, obtain according to IDBSCAN algorithm
Cluster result, has been polymerized to 3 ground sudden strains of a muscle bunch, compared with PDBSCAN algorithm proposed by the invention, ground sudden strain of a muscle bunch by whole data set
Number decreases.However, it is possible to find significantly, because experiential assignment, cluster radius is artificially expanded, the sky of lightning
Between distribution property the most well embodied.One, be two of Jinhu County southeast corner significantly are dodged bunch with maximally
Sudden strain of a muscle bunch is merged；Its two, southwest corner, Jinhu County several are sporadically dodged independent cluster by this value cluster, and calculation before
Method has been classified as noise spot.Core density point is either chosen correctness, or the calculating accuracy of space center by this
On all can reduce.
Thus it is found that the PDBSCAN algorithm set out based on testing data nature, for the definition of cluster radius
More meeting the characteristic distributions of data set, Clustering Effect more can reflect that data set midpoint is truly distributed feelings in locus
Condition.
In order to compare PDBSCAN Yu IDBSCAN difference on drop point precision of prediction, the present invention chooses one group of new thunder and lightning
Weather instance data, carries out space-time cluster by these data are called PDBSCAN Yu IDBSCAN algorithm respectively.Obtain each maximum
Bunch space center's coordinate, by the maximum thunder and lightning bunch space center position of period first-class under Curve Fitting Prediction.Contrast is final
Predict the outcome check method proposed by the invention whether on precision of prediction relative forefathers increase.
In order to weaken because the causality problems of data self existence, embody the prediction effect of method proposed by the invention
Verity and effectiveness.The thunder and lightning weather instance data of 14 points to when 6 39 points, this period when have chosen 1 day 6 August in 2009
Spot is dodged 318 times altogether.By step described above, the periods such as whole set of data are divided into 5 ground and dodge data set.Its
In, front 4 period data collection such as grade are the experimental data for space-time cluster, and last is the actual contrast number of future position distribution
According to.According to PDBSCAN from the IDBSCAN algorithm different assignment modes to Eps, PDBSCAN algorithm remains and goes out from data itself
Send out, take into full account whether data set exists the situation of variable density, select Eps according to k-dist figure and DK figure；In view of IDBSCAN
The space scale problem proposed, the period of selection is the Strong Thunderstorm period, therefore Eps still processes according to an assignment, will set by Eps
It is set to 8km.For another preset value MinPts, the present invention remains set to minima 2.
Table 4 and table 5 sets forth space-time cluster result, carry out the mode of an assignment according to IDBSCAN algorithm, and press
According to PDBSCAN algorithm polynary consideration packing density character the mode that carries out real-time update Eps for heterogeneity level, finally
The cluster result drawn also exists certain difference.Eps in table 5 is arranged, and is based on each data set density of described consideration above
Situation of change is calculated.Four Eps waiting the period are respectively set to 5km, 8km, 4km and 10km.
Table 4 IDBSCAN algorithm space-time clusters
| ID | Core classes 1 | Core classes 2 | Maximum bunch space geometry centre coordinate (longitude, latitude) | Noise spot |
| 1 | 56 | 8 | (120.7405,31.10360) | 4 |
| 2 | 63 | 5 | (120.7459,31.07428) | 5 |
| 3 | 76 | 12 | (120.7544,31.06796) | 7 |
| 4 | 56 | 22 | (120.7709,31.06306) | 3 |
Table 5 PDBSCAN algorithm space-time clusters
| ID | Core classes 1 | Core classes 2 | Core classes 3 | Maximum bunch space geometry centre coordinate (longitude, latitude) | Noise spot |
| 1 | 54 | 8 | (120.7204,31.10030) | 6 | |
| 2 | 63 | 5 | (120.7459,31.07428) | 5 | |
| 3 | 75 | 6 | 4 | (120.7558,31.06718) | 10 |
| 4 | 57 | 22 | (120.7719,31.06130) | 2 |
According to the above-mentioned maximum bunch space geometry centre coordinate (longitude, latitude) tried to achieve, obtained respectively by binomial fitting
Curvilinear equation to IDBSCAN: y=86.4985x2-20891.6x+1261490；The curvilinear equation of PDBSCAN: y=
104499x2-2524.33x+152478.4。
Maximum bunch space center's coordinate of lower first-class period that correspondence dopes is respectively as follows:
I1 (120.781,31.08707), P1 (120.7891,31.06054).
The actual value of lower first-class period is respectively as follows:
I2 (120.7731,31.0534), P2 (120.7708,31.052).
I1 Yu I2 is at a distance of 3.8km, and 2.0km apart between P1 and P2.At a distance of 0.3km between I2 and P2.So, can be further
Illustrating, thunder and lightning based on PDBSCAN algorithm trend prediction in short-term proposed by the invention is effectively, accurately, and to forefathers
Relevant forecasting procedure on precision of prediction, serve the effect of optimization.
Thus, it is possible to find, the parameter value that real-time update is composed, the Clustering Effect of data set can be improved well.As
Shown in table 4 and table 5, although in the cluster to 1,3,4 data sets, the noise spot number of IDBSCAN algorithm is than PDBSCAN algorithm
Lack, but be because not accounting for data set variable density that may be present itself, occur in that the too small or excessive problem of Eps.
The number that too small meeting causes bunch is too much, excessive, and in making bunch, the number of point is too much.Both shadow to method precision of prediction
Ring the most in the present invention arrived embodiment, so, PDBSCAN algorithm proposed by the invention can preferably evade problems.
The present invention is directed to the tradition DBSCAN algorithm deficiency to lightning forecasting, choose two sides with memory optimization from parameter
Face, is combined VDBSCAN algorithm and IDBSCAN algorithm, uses the PDBSCAN algorithm proposed after being combined to carry out lightning data
Cluster Deng the period.This algorithm not only allows for the situation that the change of thunder and lightning bunch internal density is single, also overcomes DBSCAN algorithm to change
Situation error-prone under density case, the parameter of the perfect all kinds of spatial distribution forms being likely to occur thunder and lightning is chosen.Meanwhile,
Optimize storage organization when algorithm runs to a certain extent, improve operation efficiency.By to each maximum waited under the period
The space center of thunder and lightning bunch carries out linear fit, tries to achieve the space coordinates of the thunder and lightning density maximum generating region of subsequent time.In reality
Under the inspection of example thunder and lightning synoptic process, it was predicted that value differs about 3.114km with the maximum thunder and lightning bunch space center at lower a moment, it was demonstrated that
The effectiveness of this model.Finally, carry out with the related algorithm of forefathers in terms of operational efficiency, Clustering Effect and precision of prediction three
Contrast, result shows that thunder and lightning is closed on trend prediction and serves the work of optimization by proposed algorithm to a certain extent in short-term
With.
The above, the only detailed description of the invention in the present invention, but protection scope of the present invention is not limited thereto, and appoints
What is familiar with the people of this technology in the technical scope that disclosed herein, it will be appreciated that the conversion expected or replacement, all should contain
Within the scope of the comprising of the present invention, therefore, protection scope of the present invention should be as the criterion with the protection domain of claims.
Claims (6)
1. thunder and lightning based on PDBSCAN algorithm closes on trend prediction method, it is characterised in that specifically comprise the following steps that
Step A, utilizes lightning positioning system to detect and dodge data with recording, and the ground sudden strain of a muscle data of record are carried out pretreatment, draws
It is divided into each ground waiting the period to dodge data set；
Step B, uses PDBSCAN clustering algorithm to try to achieve each maximum kernel heart sudden strain of a muscle bunch waiting the period and the sudden strain of a muscle bunch of each maximum kernel heart
Space center's coordinate；
Step C, space center's coordinate of the maximum kernel heart being tried to achieve step B sudden strain of a muscle bunch carries out fitting of a polynomial, further according to plan
The multinomial closed tries to achieve space center's coordinate of the thunder and lightning maximum kernel heart sudden strain of a muscle bunch of lower first-class period, thus realizes thunder and lightning and close on
The forecast of gesture.
Thunder and lightning based on PDBSCAN algorithm the most according to claim 1 closes on trend prediction method, it is characterised in that step
In rapid A, the ground sudden strain of a muscle data to record carry out pretreatment, are divided into each ground waiting the period and dodge data set, particularly as follows:
Step A-1, chooses the landing point coordinates data persistently occurring the ground of 20 minutes to dodge after shwoot is raw first, generates data set；
Step A-2, the data set generated by A-1 carries out waiting Time segments division, obtains dodging data set with waiting the period accordingly.
Thunder and lightning based on PDBSCAN algorithm the most according to claim 1 closes on trend prediction method, it is characterised in that step
Rapid B use PDBSCAN clustering algorithm try to achieve each maximum kernel heart sudden strain of a muscle bunch waiting the period and the sky of each maximum kernel heart sudden strain of a muscle bunch
Between centre coordinate, particularly as follows:
Step B-1, according to DBSCAN algorithm, dodges data set and tries to achieve the most each etc. each ground waiting the period obtained in step A
The k-dist figure of period；
Step B-2, the k-dist figure obtained according to B-1, observe the density dodging data in each ground waiting the period and divide different, if single close
Degree, then find out the point that the slope of curve is undergone mutation at first, dodges data set with the ground that the k-dist value that this point is corresponding is such period
Neighborhood Eps；If the situation of multiple density level occurs, then try to achieve correspondence according to the k-dist figure of the ground sudden strain of a muscle data of such period
DK figure, and determine neighborhood Eps according to DK map analysis；
Step B-3, sets MinPts, travels through each ground waiting the period and dodges data set, searches for the Eps neighborhood of each ground flash-point successively,
Each ground waiting the period is dodged data and carries out cluster calculation, choose optimum cluster result:
Step B-4, according to the optimum cluster result of B-3, it is thus achieved that each maximum kernel heart waiting the period is dodged bunch, and by bunch in all
The latitude and longitude coordinates of member, tries to achieve space center's point coordinates of each maximum kernel heart sudden strain of a muscle bunch.
Thunder and lightning based on PDBSCAN algorithm the most according to claim 3 closes on trend prediction method, it is characterised in that step
Rapid B also includes: replace R* tree in DBSCAN by the internal memory thought building adjacency list in IDBSCAN.
Thunder and lightning based on PDBSCAN algorithm the most according to claim 3 closes on trend prediction method, it is characterised in that step
Suddenly in B-1, according to DBSCAN algorithm, each ground sudden strain of a muscle data set waiting the period obtained is tried to achieve respectively wait the period in step A
K-dist figure, particularly as follows: the ground of the day part obtained in step A i.e. dodges all data in data set, ask it to arrive k evidence
Distance, and the distance that obtains is arranged by order from low to high, i.e. obtains corresponding k-dist figure.
Thunder and lightning based on PDBSCAN algorithm the most according to claim 1 closes on trend prediction method, it is characterised in that step
Rapid B-2 uses based on variable density and with the space clustering VDBSCAN algorithm of noise application, dodge number according to the ground of this period
According to k-dist figure try to achieve the DK figure of correspondence, particularly as follows: extract, the data set of variable density occurs, by its Y-axis on k-dist figure
Corresponding data are successively from the beginning of Section 2, and latter and previous item do difference, i.e. obtain the DK figure of correspondence.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610674402.9A CN106251026A (en) | 2016-08-16 | 2016-08-16 | Thunder and lightning based on PDBSCAN algorithm closes on trend prediction method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610674402.9A CN106251026A (en) | 2016-08-16 | 2016-08-16 | Thunder and lightning based on PDBSCAN algorithm closes on trend prediction method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN106251026A true CN106251026A (en) | 2016-12-21 |
Family
ID=57592240
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610674402.9A Pending CN106251026A (en) | 2016-08-16 | 2016-08-16 | Thunder and lightning based on PDBSCAN algorithm closes on trend prediction method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106251026A (en) |
Cited By (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108520023A (en) * | 2018-03-22 | 2018-09-11 | 合肥佳讯科技有限公司 | A kind of identification of thunderstorm core and method for tracing based on Hybrid Clustering Algorithm |
| CN109061774A (en) * | 2018-06-26 | 2018-12-21 | 合肥佳讯科技有限公司 | A kind of thunderstorm core relevance processing method |
| CN112187550A (en) * | 2020-10-16 | 2021-01-05 | 温州职业技术学院 | Log analysis method based on density peak value multi-attribute clustering |
| CN112183664A (en) * | 2020-10-27 | 2021-01-05 | 中国人民解放军陆军工程大学 | Novel density clustering method |
| CN112613530A (en) * | 2020-11-23 | 2021-04-06 | 北京思特奇信息技术股份有限公司 | Cell resident identification method and system based on adaptive density clustering algorithm |
| CN112668790A (en) * | 2020-12-30 | 2021-04-16 | 南京信息工程大学 | Thunder and lightning prediction method based on space-time sequence clustering algorithm and LSTM neural network |
| CN113255510A (en) * | 2021-05-21 | 2021-08-13 | 南京信息工程大学 | Rainbow cloud point charge moving path imaging method based on multi-time scale DBSCAN and sample entropy |
| CN113313717A (en) * | 2021-05-18 | 2021-08-27 | 哈尔滨理工大学 | Segmented vertebra CT image segmentation method based on deep learning |
| CN113947578A (en) * | 2021-10-18 | 2022-01-18 | 山东大学 | Prediction method of nucleation rate of vermicular cast iron based on DBSCAN clustering algorithm |
| WO2023061342A1 (en) * | 2021-10-13 | 2023-04-20 | 中兴通讯股份有限公司 | Data trend detection method, electronic device, and storage medium |
| CN116500703A (en) * | 2023-06-28 | 2023-07-28 | 成都信息工程大学 | Thunderstorm monomer identification method and device |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102253426A (en) * | 2011-04-29 | 2011-11-23 | 国网电力科学研究院武汉南瑞有限责任公司 | Method for predicting lightning motion by space density clustering |
| CN105139099A (en) * | 2015-06-04 | 2015-12-09 | 国家电网公司 | LLS-based regional lightning damage feature correlation analysis method |
-
2016
- 2016-08-16 CN CN201610674402.9A patent/CN106251026A/en active Pending
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102253426A (en) * | 2011-04-29 | 2011-11-23 | 国网电力科学研究院武汉南瑞有限责任公司 | Method for predicting lightning motion by space density clustering |
| CN105139099A (en) * | 2015-06-04 | 2015-12-09 | 国家电网公司 | LLS-based regional lightning damage feature correlation analysis method |
Non-Patent Citations (2)
| Title |
|---|
| 周董 等: "VDBSCAN：变密度聚类算法", 《计算机工程与应用》 * |
| 高文胜 等: "基于雷电定位系统监测数据的雷暴云趋势预测", 《电网技术》 * |
Cited By (16)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108520023B (en) * | 2018-03-22 | 2021-07-20 | 合肥佳讯科技有限公司 | Thunderstorm kernel identification and tracking method based on hybrid clustering algorithm |
| CN108520023A (en) * | 2018-03-22 | 2018-09-11 | 合肥佳讯科技有限公司 | A kind of identification of thunderstorm core and method for tracing based on Hybrid Clustering Algorithm |
| CN109061774A (en) * | 2018-06-26 | 2018-12-21 | 合肥佳讯科技有限公司 | A kind of thunderstorm core relevance processing method |
| CN109061774B (en) * | 2018-06-26 | 2020-10-16 | 合肥佳讯科技有限公司 | Thunderstorm core correlation processing method |
| CN112187550A (en) * | 2020-10-16 | 2021-01-05 | 温州职业技术学院 | Log analysis method based on density peak value multi-attribute clustering |
| CN112183664A (en) * | 2020-10-27 | 2021-01-05 | 中国人民解放军陆军工程大学 | Novel density clustering method |
| CN112613530A (en) * | 2020-11-23 | 2021-04-06 | 北京思特奇信息技术股份有限公司 | Cell resident identification method and system based on adaptive density clustering algorithm |
| CN112668790B (en) * | 2020-12-30 | 2023-07-25 | 南京信息工程大学 | Lightning prediction method based on space-time sequence clustering algorithm and LSTM neural network |
| CN112668790A (en) * | 2020-12-30 | 2021-04-16 | 南京信息工程大学 | Thunder and lightning prediction method based on space-time sequence clustering algorithm and LSTM neural network |
| CN113313717A (en) * | 2021-05-18 | 2021-08-27 | 哈尔滨理工大学 | Segmented vertebra CT image segmentation method based on deep learning |
| CN113255510A (en) * | 2021-05-21 | 2021-08-13 | 南京信息工程大学 | Rainbow cloud point charge moving path imaging method based on multi-time scale DBSCAN and sample entropy |
| CN113255510B (en) * | 2021-05-21 | 2023-07-25 | 南京信息工程大学 | Thunderstorm cloud point charge moving path imaging method based on multi-time scale DBSCAN and sample entropy |
| WO2023061342A1 (en) * | 2021-10-13 | 2023-04-20 | 中兴通讯股份有限公司 | Data trend detection method, electronic device, and storage medium |
| CN113947578A (en) * | 2021-10-18 | 2022-01-18 | 山东大学 | Prediction method of nucleation rate of vermicular cast iron based on DBSCAN clustering algorithm |
| CN116500703A (en) * | 2023-06-28 | 2023-07-28 | 成都信息工程大学 | Thunderstorm monomer identification method and device |
| CN116500703B (en) * | 2023-06-28 | 2023-09-01 | 成都信息工程大学 | Thunderstorm monomer identification method and device |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106251026A (en) | Thunder and lightning based on PDBSCAN algorithm closes on trend prediction method | |
| Zhang et al. | Probabilistic solar irradiation forecasting based on variational Bayesian inference with secure federated learning | |
| Hu et al. | Federated region-learning: An edge computing based framework for urban environment sensing | |
| CN102571954B (en) | Complex network clustering method based on key influence of nodes | |
| CN107918830B (en) | Power distribution network running state evaluation method based on big data technology | |
| CN105701596A (en) | Method for lean distribution network emergency maintenance and management system based on big data technology | |
| CN108683448B (en) | Influence node identification method and system suitable for aviation network | |
| CN105389621A (en) | Optimal charging pile distribution method for improving effect of electric vehicle charging load to voltage of distribution network system | |
| CN109242170A (en) | A kind of City Road Management System and method based on data mining technology | |
| Hu et al. | An intelligent UAV deployment scheme for load balance in small cell networks using machine learning | |
| Bao et al. | Lightning performance evaluation of transmission line based on data-driven lightning identification, tracking, and analysis | |
| Tian et al. | Wind power forecasting by the BP neural network with the support of machine learning | |
| CN103458032B (en) | The method and system of a kind of spatial data accessing rule dynamic statistics and Information Compression | |
| CN115329428A (en) | Method, device and equipment for determining fan arrangement mode and storage medium | |
| Shahbazi et al. | Density-based clustering and performance enhancement of aeronautical ad hoc networks | |
| CN112630836B (en) | Monitoring and stationing method and system based on power grid micro-terrain icing threat path extension analysis | |
| Wang | Optimization of wireless network node deployment in smart city based on adaptive particle swarm optimization | |
| CN114090621A (en) | Ground-air integrated ecological environment monitoring system | |
| Wang et al. | An improved multiple viewshed algorithm for region partitioning for filtering candidate viewpoints | |
| CN115759887B (en) | Multi-dimensional charging station site selection intelligent evaluation method and system | |
| CN113326883B (en) | Training method, device and medium for power utilization rate prediction model of charging station | |
| Zhou et al. | Modeling Thunderstorm Based on Paralleled and Improved Naive Bayes | |
| CN116668306B (en) | Three-view-angle-based network engineering planning method and system for mobile communication network | |
| Yao et al. | Simulation of Urban Expansion and Farmland Loss in China by Integrating Cellular Automata and Random Forest | |
| Xu | Multi-objective intelligent algorithm model design for housing environment optimization |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20161221 |