CN108196314A - A kind of northwest Pacific ring-type typhoon automatic recognition system - Google Patents
A kind of northwest Pacific ring-type typhoon automatic recognition system Download PDFInfo
- Publication number
- CN108196314A CN108196314A CN201711407247.5A CN201711407247A CN108196314A CN 108196314 A CN108196314 A CN 108196314A CN 201711407247 A CN201711407247 A CN 201711407247A CN 108196314 A CN108196314 A CN 108196314A
- Authority
- CN
- China
- Prior art keywords
- typhoon
- ring
- output
- time
- type
- 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
- 238000013528 artificial neural network Methods 0.000 claims abstract description 17
- 238000012549 training Methods 0.000 claims description 34
- 125000004122 cyclic group Chemical group 0.000 claims description 32
- 210000002569 neuron Anatomy 0.000 claims description 28
- 230000006870 function Effects 0.000 claims description 22
- 239000011159 matrix material Substances 0.000 claims description 17
- 238000004458 analytical method Methods 0.000 claims description 15
- 238000003062 neural network model Methods 0.000 claims description 12
- 230000005284 excitation Effects 0.000 claims description 8
- 238000000034 method Methods 0.000 claims description 7
- 238000004364 calculation method Methods 0.000 claims description 6
- 238000011478 gradient descent method Methods 0.000 claims description 3
- 230000004069 differentiation Effects 0.000 claims description 2
- 230000007246 mechanism Effects 0.000 claims description 2
- 210000004218 nerve net Anatomy 0.000 claims 1
- 230000001537 neural effect Effects 0.000 claims 1
- 238000010586 diagram Methods 0.000 description 6
- 238000012216 screening Methods 0.000 description 6
- 230000015572 biosynthetic process Effects 0.000 description 4
- 238000005516 engineering process Methods 0.000 description 3
- 230000007613 environmental effect Effects 0.000 description 3
- 125000002015 acyclic group Chemical group 0.000 description 2
- 238000013459 approach Methods 0.000 description 2
- 238000000205 computational method Methods 0.000 description 2
- 239000000463 material Substances 0.000 description 2
- 238000012950 reanalysis Methods 0.000 description 2
- 239000005436 troposphere Substances 0.000 description 2
- 108010074506 Transfer Factor Proteins 0.000 description 1
- 230000004913 activation Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 210000005036 nerve Anatomy 0.000 description 1
- 238000004064 recycling Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000004044 response Effects 0.000 description 1
- 230000001360 synchronised effect Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01W—METEOROLOGY
- G01W1/00—Meteorology
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
- Y02A50/00—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE in human health protection, e.g. against extreme weather
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
- Y02A90/00—Technologies having an indirect contribution to adaptation to climate change
- Y02A90/10—Information and communication technologies [ICT] supporting adaptation to climate change, e.g. for weather forecasting or climate simulation
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Environmental & Geological Engineering (AREA)
- Life Sciences & Earth Sciences (AREA)
- Theoretical Computer Science (AREA)
- Computational Linguistics (AREA)
- Evolutionary Computation (AREA)
- Ecology (AREA)
- Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Biodiversity & Conservation Biology (AREA)
- Data Mining & Analysis (AREA)
- Environmental Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Atmospheric Sciences (AREA)
- Image Analysis (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a kind of northwest Pacific ring-type typhoon automatic recognition systems, a kind of automatic recognition system for northwest Pacific ring-type typhoon is established for the first time, by the satellite data of typhoon to be judged, environment field, SST fields data input automatic recognition system, the typhoon a series of physical feature will be calculated in system, and then based on these characteristic parameters, pass through prescreening and linear discriminant, or the carry out automatic identification of BP neural network, warning is made to the typhoon for meeting northwest Pacific ring-type typhoon feature, the forecast for such typhoon provides help.
Description
Technical field
The invention belongs to Atmosphere System Forecasting recognition technology more particularly to northwest Pacific ring-type typhoon automatic identification with
Alarm system.
Background technology
Cyclic annular typhoon has the characteristics that intensity is big, it is long to hold time, forecast difficulty is big.Northwest Pacific ring-type typhoon generates
Region is concentrated and login probability is big, larger threat is continued to cause to East Asia Region, but not yet fully paid close attention at present.It is right at present
The prediction of typhoon is the miniature and environmental condition based on Eastern Pacific and Atlantic Ocean ring-type hurricane, establishes screening and identification, so as to
The objective identification algorithm of the Pacific Ocean and Atlantic Ocean ring-type hurricane is established, then on automatic identification early-warning and predicting, can not also accomplished in time
Quickly response.
Invention content
In view of the problems of the existing technology, the present invention provides a kind of northwest Pacific ring-type typhoon automatic identification systems
System establishes the automatic recognition system for northwest Pacific ring-type typhoon for the first time, and the early warning, forecast for such typhoon provide side
It helps.
In order to solve the above technical problems, present invention employs following technical schemes:A kind of northwest Pacific ring-type typhoon is certainly
Dynamic identifying system, includes the following steps:
Step 1:The when number that cyclic annular typhoon occurs in the historical time section of ocean is called in the analysis and calculating of physical quantity first
According to occur recently non-annularity typhoon when time data, and calculate the sea area ring-type typhoon characteristic, the characteristic
According to including intensity of typhoon, structure, environment field and SST fields characteristic;
Step 2:It is secondary during secondary and current year all non-annularity typhoons when calling all cyclic annular typhoons in northwest Pacific former years
As training set, and the result of calculation of the intensity of the training set typhoon, structure, environment field, SST fields characteristic is obtained, established
Automatic prescreening recognition mechanism:When the condition of typhoon meeting formula (1) to be evaluated, by prescreening, otherwise screened out,
In formula, VmaxFor THE MAXIMUM WIND SPEED OF TYPHOON, RcIt is averaged most cold radius for typhoon circumference, Δ TeyeAt the most cold radius of typhoon
Portion's circumference is averaged infrared brightness temperature difference maximum value in the inner, and SHRD is deep vertical wind shear, and u200 is 200hPa environment zonal winds,
SST be center of typhoon marine surface temperature, σcFor infrared brightness temperature standard deviation at the most cold radius of typhoon;
Step 3:When simultaneously to all cyclic annular typhoons in northwest Pacific former years during secondary and current year all non-annularity typhoons
Secondary training set carries out linear discriminant, with σc、VAR、ΔTeye, u200, SST as linear discriminant foundation, time instruction when selecting cyclic annular
Practice collection and the midpoint that is projected in the one-dimensional space of non-annularity training set as criteria for classification, to fall non-annularity typhoon side when
It is secondary to be screened out;
Step 4:For simultaneously by during the typhoon of step 2 and 3 prescreenings time, carrying out backpropagation neural network differentiation,
Input variable σc、VAR、ΔTeye, u200, SST, the weight coefficient obtained with training is multiplied and passes through excitation function, and it is non-to export
It is secondary during secondary typhoon when cyclic annular to be screened out;
Step 5:It checking whether during continuous two typhoons and time to have passed through step 1-4, being cyclic annular when there are two consecutive hourss time
When time, then judge that current typhoon belongs to cyclic annular typhoon.
Further, in step 3 linear discriminant,
First, with [the σ of when during training set ring-type times and non-annularity timesc,VAR,ΔTeye,u200,SST]TForm 5 × n1With 5
×n2Matrix X1And X2, wherein n1And n2Time number and time number during training set non-annularity during for training set ring-type, and by physics
It is secondary when measuring each to carry out being averaged to obtain matrixWith
Then, X is calculated1And X2Auto-covariance matrix S1And S2, and pass through auto-covariance matrix S1And S2It is calculated in class
Stroll matrix S,
Then, spin matrix W is chosen,Definition judgement midpoint Y0,
Rotation projection Rotation projectionSecondary physical quantity combination X during by calculating to be determined, into
And decision content Y=W secondary when obtaining to be determinedTThen X passes through during following judgement typhoon ring-type times:
As Y < Y0, and Y1< Y0< Y2When or Y > Y0, and Y1> Y0> Y2When, it is determined as secondary during ring-type;
As Y > Y0, and Y1< Y0< Y2When or Y < Y0, and Y1> Y0> Y2When, it is determined as secondary during non-annularity.
Further, during the BP neural network model foundation in step 4, with σc,VAR,ΔTeye,u200, SST conducts
Input layer input variable, setting hidden layer neuron number are 40, and the output result of output layer is exported for binary;
Establish BP neural network model, it is assumed that input layer hasA neuron, hidden layer have aNeuron, output layer have d
A neuron, symbol definition are as follows:
Input vector:X=(x1,x2,…xl)
Hidden layer output vector:H=(h1,h2,…hq)
Output layer output vector:Y=(y1,y2,…yd)
Desired output vector:
Error function:
Training sample:(xk,yk), k=1, number when 2 ... n, n are training set ring-type typhoon and total non-annularity typhoon;
Excitation function is
BP neural network model is established according to training set, with σc、VAR、ΔTeye, the input of u200, SST as input layer
Variable x=(σc,VAR,ΔTeye,u200, SST), setting hidden neuron number is 40, and the output result of output layer is defeated for binary
Go out, cyclic annular typhoon is 1, and non-annularity typhoon is 0, and learning procedure includes:
(1) hidden layer and the output valve of output layer neuron are calculated:
Hidden neuron output valve:
Output layer neuron output value:
In formula, W1 is hidden layer weights, and W2 is output layer weights;Serial numbers of the subscript i for hidden layer neuron, subscript j
Serial number for output layer neuron;
(2) error function is calculated to the partial derivative of output layer weights and transmitted backward, calculate error function and hidden layer is respectively weighed
The partial derivative of value:
Output layer:
Hidden layer:
(3) using gradient descent method and step (2) obtained by partial derivative, to output layer weights W1 'ijWith hidden layer weights
W2′ijIt is modified:
W1′ij=W1ij-η·g(W1ij), i=1,2 ... q, j=1,2 ... l (6)
W2′ij=W2ij-η·g(W2ij), i=1,2 ... d, j=1,2 ... q (7)
In above formula, η is learning rate;
(4) revised error function value is calculated, terminates to train if up to default precision, preserves hidden layer weights W1 and output
Layer weights W2, determines BP neural network model, otherwise chooses next sample and is learnt otherwise to choose next sample progress
Study.
Further, in the BP neural network model, secondary determination method is as follows when cyclic annular in step (4):
It is secondary during to typhoon to be determined, by the σ of when this timesc、VAR、ΔTeye, u200, SST input BP neural network model,
The hidden layer weights W1 and output layer weights W2 that training obtains is read, calculates obtained by excitation function respectively according to formula (2) and formula (3)
The hidden neuron output valve obtained and output layer neuron output value are determined as ring-type if output layer neuron output value is 1
When time, when being determined as non-annularity for 0 time.
Advantageous effect:Relative to the prior art, the present invention has the following advantages:The present invention establishes a kind of for west for the first time
The automatic recognition system of North Pacific's ring-type typhoon, by the satellite data of typhoon to be judged, environment field, SST fields data input certainly
Dynamic identifying system, the typhoon a series of physical feature will be calculated in system, and then carry out " prescreening+line based on these features
Property differentiate (or BP neural network) " automatic identification, warning is made to the typhoon for meeting northwest Pacific ring-type typhoon feature, thus
The forecast of class typhoon provides help.The present invention is applied in the analysis of 2000~2008 years all typhoons, the life of identifying system
Middle rate reaches 100%, and accuracy can mutually be equal to up to 91.3% with Eastern Pacific and Atlantic Ocean ring-type hurricane automatic recognition system
It is beautiful.
Description of the drawings
Fig. 1 is with cyclic annular typhoon WALT (1991.05.13.21:00) circumference for is averaged most cold radius Rc, most cold half
The standard deviation sigma of infrared brightness temperature at diametercAnd the maximum value Δ T of the infrared brightness temperature difference of being averaged at most cold radius with its internal circumferenceeyeIt calculates
Schematic diagram;
Fig. 2 is BP neural network model schematic;
Fig. 3 is the BP neural network model schematic based on the present invention;
Fig. 4 is northwest Pacific ring-type Typhoon Satellite Image case of the present invention;
Fig. 5 is Eastern Pacific and Atlantic Ocean ring-type hurricane formation zone and logs in situation schematic diagram;
Fig. 6 is northwest Pacific ring-type typhoon formation zone and logs in situation schematic diagram;
Fig. 7 is northwest Pacific ring-type typhoon automatic recognition system flow diagram of the present invention.
Specific embodiment
Below in conjunction with the accompanying drawings and with specific embodiment, the present invention is furture elucidated.It should be understood that these embodiments are only used for
It the bright present invention rather than limits the scope of the invention, after the present invention has been read, those skilled in the art are to of the invention
The modification of various equivalent forms falls within the application range as defined in the appended claims.
The present invention establishes a kind of automatic recognition system for northwest Pacific ring-type typhoon for the first time, by typhoon to be judged
Satellite data, environment field, SST fields data input automatic recognition system, system will be calculated the typhoon a series of physical spy
Sign, and then " prescreening+linear discriminant (or BP neural network) " automatic identification is carried out based on these features, to meeting northwest peace
The typhoon of foreign ring-type typhoon feature makes warning, and the forecast for such typhoon provides help.
The present invention is applied in the analysis of 2000~2008 years all typhoons, the hit rate of identifying system reaches 100%,
Accuracy can compare favourably up to 91.3% with Eastern Pacific and Atlantic Ocean ring-type hurricane automatic recognition system.
Satellite data and environment field, SST fields data the present invention is based on NORTHWESTERN PACIFIC TYPHOON calculate intensity of typhoon, knot
The physical quantitys such as structure, environment temperature field, wind field, SST fields, by the physical quantity being calculated by " prescreening+linear discriminant " (or
" prescreening+BP neural network ") recognizer, warning is made to the typhoon for meeting northwest Pacific ring-type typhoon feature.Specifically
Step is as follows:
1st, the analysis and calculating of physical quantity
First, typhoon position, maximum wind velocity (vmax) and infrared brightness temperature are directly obtained from typhoon satellite data.Infrared brightness
Warm data distribution, to it along after broadwise carries out linear interpolation twice respectively, obtains polar coordinate system using longitude and latitude as transverse and longitudinal coordinate
Under data, further calculate the average infrared brightness temperature of circumference on center of typhoon to radius each at center of typhoon 560km
The standard deviation of sequence and the sequence (VAR).Selection circumference is averaged most cold radius (RC) analysis station eye of wind size, it is red at most cold radius
Standard deviation (the σ of outer bright temperatureC) analysis eye wall uniformity coefficient, be averaged the infrared brightness temperature difference most with its internal circumference at most cold radius
Big value (Δ Teye) analysis eye wall convection intensity.Table 1-1 is physical quantity, (satellite data) is summarized in abbreviation and its computational methods
Table 1-1
Secondly, it according to environment field data, calculates typhoon periphery and is averaged the temperature on (200-800km) troposphere top (200hPa)
It spends (t200), obtains high-rise Characteristics of Temperature Field.Similarly, calculate 200-800km average convection layers on (200hPa), in
(500hPa), lower part (850hPa) warp, broadwise Characteristics of Wind Field (u200, v200, u500, v500, u850, v850), by high-rise and
The wind field of low layer, middle level and low layer does subtraction of vector, obtains 200-850hPa deep verticals wind shear (SHRD) and 500-
850hPa shallow-layers vertical wind shear (SHRS) obtains environment shear feature.
Finally, the marine surface temperature data in SST fields data are interpolated into center of typhoon using cubic spline interpolation
Position obtains the marine surface temperature (SST) at center of typhoon.
Table 1-2
Table 1-2 physical quantitys, abbreviation and its computational methods summarize (environment field and SST fields data)
2nd, the principle of recognizer and application
Secondary and all intensity in 2009 are acyclic more than 80knots during table 2-1 1990~2009 years ring-types of northwest Pacific
Physical Quantity Calculation result (satellite data) during shape, the variate-value for tilting overstriking represent that the mean value of two groups of data has significant difference
(α=0.05).
Table 2-1
Table 2-2 is with during 1999~2009 years all cyclic annular typhoons of northwest Pacific times and all non-annularity typhoons in 2009
Shi Ciwei training sets, according to step (1) calculate analysis northwest Pacific ring-type intensity of typhoon, structure, environment field, SST fields allusion quotation
Type feature (table 2-1,2-2).
Table 2-2
Result of calculation based on training set, it is as follows to establish recognizer:
Setting prescreening standard (table 3) first:vmax,RC,ΔTeyeDo not meet normal distribution, therefore by the sieve of these three amounts
Standard is gone to be set to the minimum value (MIN*0.95) for being slightly less than cyclic annular typhoon.The screening conditions of SHRD and u200 are less than all ring-types
When time maximum value, more than it is all cyclic annular when time minimum value, do not meet being weeded out for this condition.Finally, SST meets normal state
Distribution, thus outside the positive and negative three times standard deviation range of cyclic annular typhoon SST mean values when time will be weeded out.
Prescreening can almost ensure 100% hit rate, but will also result in higher empty report rate simultaneously.It is so right
By prescreening when time also need to apply once linear again and differentiate.
Linear discriminant is a kind of simple linear logic homing method, and basic thought is to project the sample of n-dimensional space
Onto the straight line of a specific direction, the one-dimensional space is formed.In this direction, the projection of sample on straight line can separate
It is best.In the present invention, incoherent σ is selectedc, VAR, Δ Teye, u200, SST select ring as the criterion for establishing linear discriminant
Time training set and time training set projects on 1 dimension space during non-annularity midpoint are fallen as the standard classified in non-annularity during shape
Typhoon side when time will be weeded out.
In practical programs, time [σ when training set is each cyclic annularc, VAR, Δ Teye, u200, SST]TForm 5 × n1Matrix X1
(wherein n1Time number during for training set ring-type), by physical quantity it is each when time average (matrix X1Each row is averaged)Similarly,
Secondary above-mentioned physical quantity forms matrix X during each non-annularity2, average value composition matrixX is calculated respectively1、X2Auto-covariance square
Battle array S1、S2, calculateChoose spin matrixJudge midpointAndIt falls in Y0Both sides.
For it is to be determined when time physical quantity combine X, calculate Y=WTX.If Y and Y1It falls and (is all higher than or small in homonymy
In Y0), then belong to secondary during ring-type, pass through linear discriminant;If Y and Y2It falls in homonymy, then belongs to secondary during non-annularity, not by linear
Differentiate.
In order to attempt to further improve screening accuracy rate, and BP neural network algorithm is used to substitute linear discriminant.
BP neural network is a kind of multilayer feedforward neural network (such as Fig. 2), which is mainly characterized by before signal to biography
It passs, Feedback error.In conducting forward, input signal is successively handled from input layer through hidden layer, until output layer.It is each
One layer of neuron state under the influence of the neuron state of layer.If output layer is not up to desired output, reversed operation, according to prediction
Error transfer factor network weight and threshold value, so as to which the prediction of BP neural network output be made constantly to approach desired output.It can be by BP nerves
Network is considered as the Nonlinear Mapping for being input to output, which can approach the continuous function of arbitrarily complicated degree with arbitrary accuracy.
BP network standards learning algorithm assumes that input layer hasA neuron, hidden layer have aNeuron, output layer have d
Neuron, symbol definition are as follows:
Input vector:
Hidden layer output vector:
Output layer output vector:Y=(y1, y2..., yd)
Desired output vector:
Error function:
Training sample:(xk, yk), k=1,2 ... n;(number when n is training set ring-type and total non-annularity typhoon)
Hidden layer excitation function is defined as f, and common excitation function form has sigmoid functionsOr tanh letters
Number f (x)=2sigmoid (2x) -1 etc..
Then the learning procedure of BP neural network is:
(1) hidden layer and the output valve of output layer neuron are calculated:
Hidden layer:
Output layer:
(2) error function is calculated to the partial derivative of output layer weights and transmitted backward, calculate error function and hidden layer is respectively weighed
The partial derivative of value:
Output layer:
Hidden layer:
(3) using partial derivative obtained by gradient descent method and (2), output layer weights and hidden layer weights are modified:
η is learning rate
(4) calculate revised error function value, terminate to train if up to default precision, otherwise choose next sample into
Row study.
The BP neural network that the present invention uses is using only there are one single hidden layer configuration of middle layer, activation primitive is
Sigmoid functions.Fig. 3 gives the schematic diagram of the BP neural network.Input layer is determines as 5 independent variables obtained by prescreening
It is fixed, i.e. input variable x=[σC, VAR, Δ Teye, u200, SST];It is 40 to set hidden layer neuron number;Output layer is two
Member output --- when i.e. cyclic annular time (1) or during non-annularity secondary (0).Finally, a BP neural network is obtained based on training set to differentiate
Model.
During automatic identification, by typhoon to be judged institute, rapid (1) the Physical Quantity Calculation result of hyposynchronization passes through " pre-sifted one by one sometimes
Choosing+linear discriminant " (or " prescreening+BP neural network ") recognizer is screened, by screening when there is continuous two time,
When secondary when being considered as ring-type, automatic recognition system will export current typhoon and belong to cyclic annular typhoon, otherwise it is assumed that belonging to acyclic
Shape typhoon, does not export.If time there are shortage of data during the typhoon, but when time being considered as ring-type when lacking former and later two
Secondary, it is cyclic annular typhoon that automatic recognition system, which will also export current typhoon,.
Fig. 4 is northwest Pacific ring-type Typhoon Satellite Image case described in the invention (Chu and Tan, 2014).
As can be seen that such typhoon, using distinct cyclic structure as main feature, generally in height axial symmetry, typhoon eye is than general platform
Wind is big, around almost uniform deep convection ring around typhoon eye, and lacks spiral rainband in deep convection ring periphery.In addition,
Knaff et al. (2003) point out that such typhoon only generates under certain environmental conditions.This is the present invention is based on satellite moneys
Material and environment field, SST fields data establish northwest Pacific ring-type typhoon automatic recognition system and provide theoretical foundation.
Fig. 5 and Fig. 6 be respectively Eastern Pacific and Atlantic Ocean ring-type hurricane and northwest Pacific ring-type typhoon formation zone and
Log in situation (Chu and Tan, 2014).Cyclic annular typhoon has the characteristics that intensity is big, it is long to hold time, forecast difficulty is big, but
Compared to Eastern Pacific and Atlantic Ocean ring-type hurricane formation zone far from continent, small, the northwest Pacific ring-type typhoon life that logs in probability
Into region close to continent, to log in probability big (whole of wherein R1 Area generations logs in), continue to cause larger prestige to East Asia Region
The side of body.
The embodiment of the present invention is as follows:
Apply the present invention in the identification of all typhoons of 2000-2008 northwest Pacifics.The typhoon satellite money of input
Material waits the observation note of plan B1 synchronous satellites (HURSAT-B1) from newest hurricane satellite record project and INSAT international satellite's thin clouds
Record.The 6 hour interval whole world of the environment field data from Environmental forecasting centre and American National Center for Atmospheric Research is again
Analysis of data (NCEP-NCAR reanalysis).Extra large temperature data then come from Reynolds marine surface temperature analysis of data again
(Reynolds’s SST reanalysis)。
Fig. 7 is northwest Pacific ring-type typhoon automatic recognition system flow diagram of the present invention, is specifically included:
1. Physical Quantity Calculation:
Time typhoon position, maximum wind velocity (vmax) and infrared brightness temperature when directly obtaining current from typhoon satellite data.It is red
Outer bright temperature data distribution is sat using longitude and latitude as transverse and longitudinal coordinate, to it along pole after broadwise carries out linear interpolation twice respectively, is obtained
The lower data of mark system, further average infrared of circumference on calculating center of typhoon to radius each at center of typhoon 560km
The standard deviation of bright temperature sequence and the sequence (VAR).Selection circumference is averaged most cold radius (Rc) analysis station eye of wind size, most cold radius
Locate the standard deviation (σ of infrared brightness temperaturec) analysis eye wall uniformity coefficient, be averaged at most cold radius the infrared brightness temperature difference with its internal circumference
Maximum value (Δ Teye) analysis eye wall convection intensity.
According to environment field data, time typhoon periphery is averaged (200-800km) troposphere top (200hPa) when calculating current
Temperature (t200), obtain high-rise Characteristics of Temperature Field.Similarly, calculate 200-800km average convection layers on (200hPa), in
(500hPa), lower part (800hPa) warp, broadwise Characteristics of Wind Field (u200, v200, u500, v500, u800, v800), by high-rise and
The wind field of low layer, middle level and low layer does subtraction of vector, obtains 200-850hPa deep verticals wind shear (SHRD) and 500-
850hPa shallow-layers vertical wind shear (SHRS) obtains environment shear feature.
Marine surface temperature data in SST fields data are interpolated into center of typhoon position using cubic spline interpolation,
Obtain the marine surface temperature (SST) at center of typhoon.
2. recognizer screens:
Using it is current when time vmax, Rc,ΔTeye, SHRD, u200, SST carries out a prescreening (predis.m), to logical
The when time recycling σ crossedc, VAR, Δ Teye, u200, SST carry out linear discriminant (LDA.m) or BP neural network identification
(predict.m)。
By screening when there is continuous two time, that is, when being considered as ring-type time, automatic recognition system will export current
Typhoon belongs to cyclic annular typhoon, otherwise it is assumed that belonging to non-annularity typhoon, does not export.
If when ring-type time is considered as time there are shortage of data during the typhoon, but when lacking former and later two time, automatically
It is cyclic annular typhoon that identifying system, which will also export current typhoon,.If the typhoon, which has, repeatedly continues through screening, automatic recognition system
Repeatedly the typhoon will be warned, be had been calculated until typhoon institute is sometimes secondary as only.
Experiments have shown that if two groups of automatic recognition systems are combined, i.e., " prescreening+linear discriminant " and " prescreening+BP is neural
Network " all thinks that current typhoon belongs to cyclic annular typhoon, is just identified as cyclic annular typhoon, can further improve accuracy.
It is final to obtain recognition result as follows (table 2):
22000~2008 years northwest Pacific ring-type typhoon automatic identification procedure recognition results of table
As can be seen that the present invention is applied in the analysis of 2000~2008 years all typhoons, the hit rate of identifying system
Reach 100%, accuracy can compare favourably up to 91.3% with Eastern Pacific and Atlantic Ocean ring-type hurricane automatic recognition system.
Claims (4)
1. a kind of northwest Pacific ring-type typhoon automatic recognition system, it is characterised in that include the following steps:
Step 1:The analysis and calculating of physical quantity, call first occur in the historical time section of ocean cyclic annular typhoon when time data and
Time data during the non-annularity typhoon occurred recently, and calculate the sea area ring-type typhoon characteristic, the characteristic packet
Include intensity of typhoon, structure, environment field and SST fields characteristic;
Step 2:Time conduct during secondary and current year all non-annularity typhoons when calling all cyclic annular typhoons in northwest Pacific former years
Training set, and the result of calculation of the intensity of the training set typhoon, structure, environment field, SST fields characteristic is obtained, it establishes automatic
Prescreening recognition mechanism:When the condition of typhoon meeting formula (1) to be evaluated, by prescreening, otherwise screened out,
In formula, VmaxFor THE MAXIMUM WIND SPEED OF TYPHOON, RcIt is averaged most cold radius for typhoon circumference, Δ TeyeIt is in for the most cold radius of typhoon
Internal circumference is averaged infrared brightness temperature difference maximum value, and SHRD is deep vertical wind shear, and u200 is 200hPa environment zonal winds, SST
For center of typhoon marine surface temperature, σcFor infrared brightness temperature standard deviation at the most cold radius of typhoon;
Step 3:When simultaneously to all cyclic annular typhoons in northwest Pacific former years time and during all non-annularity typhoons of current year time
Training set carries out linear discriminant, with σc、VAR、ΔTeye, u200, SST as linear discriminant foundation, time training set when selecting cyclic annular
With the midpoint that non-annularity training set projects in the one-dimensional space as criteria for classification, to fall non-annularity typhoon side when time into
Row screens out;
Step 4:For by during the typhoon of step 2 and 3 prescreenings times, progress backpropagation neural network differentiation inputs simultaneously
Variable σc、VAR、ΔTeye, u200, SST, the weight coefficient obtained with training is multiplied and passes through excitation function, export as non-annularity
When time typhoon when time screened out;
Step 5:It checks whether during continuous two typhoons and time to have passed through step 1-4, it is secondary when there are two consecutive hourss time and be cyclic annular,
Then judge that current typhoon belongs to cyclic annular typhoon.
2. northwest Pacific ring-type typhoon automatic recognition system according to claim 1, it is characterised in that:Step 3 is linearly sentenced
Not in,
First, with [the σ of when during training set ring-type times and non-annularity timesc,VAR,ΔTeye,u200,SST]TForm 5 × n1With 5 × n2
Matrix X1And X2, wherein n1And n2Time number and time number during training set non-annularity during for training set ring-type, and physical quantity is each
It is secondary when a to carry out being averaged to obtain matrixWith
Then, X is calculated1And X2Auto-covariance matrix S1And S2, and pass through auto-covariance matrix S1And S2It is calculated in class and takes a walk
Matrix S,
Then, spin matrix W is chosen,Definition judgement midpoint Y0,'s
Rotation projectionRotation projectionSecondary physical quantity combination X during by calculating to be determined, into
And decision content Y=W secondary when obtaining to be determinedTThen X passes through during following judgement typhoon ring-type times:
As Y < Y0, and Y1< Y0< Y2When or Y > Y0, and Y1> Y0> Y2When, it is determined as secondary during ring-type;
As Y > Y0, and Y1< Y0< Y2When or Y < Y0, and Y1> Y0> Y2When, it is determined as secondary during non-annularity.
3. northwest Pacific ring-type typhoon automatic recognition system according to claim 1, it is characterised in that:BP in step 4
During Establishment of Neural Model, with σc,VAR,ΔTeye,u200, SST is as input layer input variable, setting hidden layer god
It is 40 through first number, the output result of output layer is exported for binary;
Establish BP neural network model, it is assumed that input layer hasA neuron, hidden layer have aNeuron, output layer have d god
Through member, symbol definition is as follows:
Input vector:X=(x1,x2,…xl)
Hidden layer output vector:H=(h1,h2,…hq)
Output layer output vector:Y=(y1,y2,…yd)
Desired output vector:
Error function:
Training sample:(xk,yk), k=1, number when 2 ... n, n are training set ring-type typhoon and total non-annularity typhoon;
Excitation function is
BP neural network model is established according to training set, with σc、VAR、ΔTeye, the input variable x of u200, SST as input layer
=(σc,VAR,ΔTeye,u200, SST), setting hidden neuron number is 40, and the output result of output layer is exported for binary, ring
Shape typhoon is 1, and non-annularity typhoon is 0, and learning procedure includes:
(1) hidden layer and the output valve of output layer neuron are calculated:
Hidden neuron output valve:
Output layer neuron output value:
In formula, W1 is hidden layer weights, and W2 is output layer weights;Subscript i is the serial number of hidden layer neuron, and subscript j is defeated
Go out the serial number of layer neuron;
(2) error function is calculated to the partial derivative of output layer weights and transmit backward, calculating error function each weights to hidden layer
Partial derivative:
Output layer:
Hidden layer:
(3) using gradient descent method and step (2) obtained by partial derivative, to output layer weights W1 'ijWith hidden layer weights W2 'ijIt carries out
It corrects:
W1′ij=W1ij-η·g(W1ij), i=1,2 ... q, j=1,2 ... l (6)
W2′ij=W2ij-η·g(W2ij), i=1,2 ... d, j=1,2 ... q (7)
In above formula, η is learning rate;
(4) revised error function value is calculated, terminates to train if up to default precision, hidden layer weights W1 is preserved and is weighed with output layer
Value W2, determines BP neural network model, otherwise chooses next sample and is learnt otherwise to choose next sample and is learnt.
4. northwest Pacific ring-type typhoon automatic recognition system according to claim 3, it is characterised in that:The BP nerve nets
In network model, secondary determination method is as follows when cyclic annular in step (4):
It is secondary during to typhoon to be determined, by the σ of when this timesc、VAR、ΔTeye, u200, SST input BP neural network model, read
The hidden layer weights W1 that training obtains and output layer weights W2, calculates what is obtained by excitation function respectively according to formula (2) and formula (3)
Hidden neuron output valve and output layer neuron output value, it is secondary when ring-type is determined as if output layer neuron output value is 1,
When being determined as non-annularity for 0 time.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201711407247.5A CN108196314B (en) | 2017-12-22 | 2017-12-22 | Automatic northwest Pacific annular typhoon identification system |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201711407247.5A CN108196314B (en) | 2017-12-22 | 2017-12-22 | Automatic northwest Pacific annular typhoon identification system |
Publications (2)
Publication Number | Publication Date |
---|---|
CN108196314A true CN108196314A (en) | 2018-06-22 |
CN108196314B CN108196314B (en) | 2023-04-18 |
Family
ID=62583533
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201711407247.5A Active CN108196314B (en) | 2017-12-22 | 2017-12-22 | Automatic northwest Pacific annular typhoon identification system |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN108196314B (en) |
Cited By (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109583653A (en) * | 2018-11-30 | 2019-04-05 | 南京信息工程大学 | The extended peroid forecasting procedure of NORTHWESTERN PACIFIC TYPHOON based on statistical model |
CN109583095A (en) * | 2018-11-30 | 2019-04-05 | 南京信息工程大学 | NORTHWESTERN PACIFIC TYPHOON extended peroid forecasting procedure based on mixing statistics dynamic model |
CN109993282A (en) * | 2019-03-28 | 2019-07-09 | 北京航空航天大学 | A kind of typhoon involves the prediction technique of range |
CN110852169A (en) * | 2019-10-12 | 2020-02-28 | 中国气象局上海台风研究所 | Method for identifying typhoon maximum wind speed radius based on airborne data |
CN114740550A (en) * | 2022-06-14 | 2022-07-12 | 广东海洋大学 | Intelligent recognition early warning method and system for continuous storm events |
CN117036983A (en) * | 2023-10-08 | 2023-11-10 | 中国海洋大学 | Typhoon center positioning method based on physical reinforcement deep learning |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
RU2407044C1 (en) * | 2009-09-30 | 2010-12-20 | Павел Николаевич Манташьян | Method of lowering temperature of ocean surface water in area of possible occurrence of typhoon |
CN103955009A (en) * | 2014-04-25 | 2014-07-30 | 宁波市气象台 | Method for extracting typhoon objective forecast information from numerical forecasting product |
CN104036462A (en) * | 2014-06-06 | 2014-09-10 | 南京大学 | Method to estimate intensity of tropical cyclone across the Northwest Pacific Ocean by using geostationary meteorological satellite image |
CN107103173A (en) * | 2016-10-31 | 2017-08-29 | 陈柏宇 | A kind of Design Wave projectional technique for embodying the influence of the factor of typhoon three |
-
2017
- 2017-12-22 CN CN201711407247.5A patent/CN108196314B/en active Active
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
RU2407044C1 (en) * | 2009-09-30 | 2010-12-20 | Павел Николаевич Манташьян | Method of lowering temperature of ocean surface water in area of possible occurrence of typhoon |
CN103955009A (en) * | 2014-04-25 | 2014-07-30 | 宁波市气象台 | Method for extracting typhoon objective forecast information from numerical forecasting product |
CN104036462A (en) * | 2014-06-06 | 2014-09-10 | 南京大学 | Method to estimate intensity of tropical cyclone across the Northwest Pacific Ocean by using geostationary meteorological satellite image |
CN107103173A (en) * | 2016-10-31 | 2017-08-29 | 陈柏宇 | A kind of Design Wave projectional technique for embodying the influence of the factor of typhoon three |
Non-Patent Citations (1)
Title |
---|
聂高臻 等: "台风"云娜"(2004)的数值模拟:眼墙对流与环境风切变的关系" * |
Cited By (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109583653A (en) * | 2018-11-30 | 2019-04-05 | 南京信息工程大学 | The extended peroid forecasting procedure of NORTHWESTERN PACIFIC TYPHOON based on statistical model |
CN109583095A (en) * | 2018-11-30 | 2019-04-05 | 南京信息工程大学 | NORTHWESTERN PACIFIC TYPHOON extended peroid forecasting procedure based on mixing statistics dynamic model |
CN109583653B (en) * | 2018-11-30 | 2022-06-21 | 南京信息工程大学 | Northwest Pacific ocean typhoon extension period forecasting method based on statistical model |
CN109993282A (en) * | 2019-03-28 | 2019-07-09 | 北京航空航天大学 | A kind of typhoon involves the prediction technique of range |
CN110852169A (en) * | 2019-10-12 | 2020-02-28 | 中国气象局上海台风研究所 | Method for identifying typhoon maximum wind speed radius based on airborne data |
CN110852169B (en) * | 2019-10-12 | 2023-02-03 | 中国气象局上海台风研究所(上海市气象科学研究所) | Method for identifying typhoon maximum wind speed radius based on airborne data |
CN114740550A (en) * | 2022-06-14 | 2022-07-12 | 广东海洋大学 | Intelligent recognition early warning method and system for continuous storm events |
CN117036983A (en) * | 2023-10-08 | 2023-11-10 | 中国海洋大学 | Typhoon center positioning method based on physical reinforcement deep learning |
CN117036983B (en) * | 2023-10-08 | 2024-01-30 | 中国海洋大学 | Typhoon center positioning method based on physical reinforcement deep learning |
Also Published As
Publication number | Publication date |
---|---|
CN108196314B (en) | 2023-04-18 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN108196314A (en) | A kind of northwest Pacific ring-type typhoon automatic recognition system | |
Han et al. | Convolutional neural network for convective storm nowcasting using 3-D Doppler weather radar data | |
CN109345507B (en) | Dam image crack detection method based on transfer learning | |
CN105488528B (en) | Neural network image classification method based on improving expert inquiry method | |
Elsner | Tracking hurricanes | |
CN110197218A (en) | Thunderstorm gale grade forecast classification method based on multi-source convolutional neural networks | |
CN110163187A (en) | Remote road traffic sign detection recognition methods based on F-RCNN | |
CN105787501B (en) | Power transmission line corridor region automatically selects the vegetation classification method of feature | |
CN110134774A (en) | It is a kind of based on the image vision Question-Answering Model of attention decision, method and system | |
Xu et al. | Enhanced moth-flame optimization based on cultural learning and Gaussian mutation | |
CN109086540A (en) | A kind of method and device constructing Forecasting of Tropical Cyclone model | |
CN108960421A (en) | The unmanned surface vehicle speed of a ship or plane online forecasting method based on BP neural network of improvement | |
CN109143408A (en) | Combine short-term precipitation forecasting procedure in dynamic area based on MLP | |
CN115359366A (en) | Remote sensing image target detection method based on parameter optimization | |
Cai et al. | Cloud classification of satellite image based on convolutional neural networks | |
Zhao et al. | CNN-based large scale landsat image classification | |
CN113255239B (en) | Deep learning method for predicting icing thickness of ultrahigh voltage transmission line tower | |
CN106022293A (en) | Pedestrian re-identification method of evolutionary algorithm based on self-adaption shared microhabitat | |
CN107766858A (en) | A kind of method that ship detecting is carried out using diameter radar image | |
CN116822716A (en) | Typhoon prediction method, system, equipment and medium based on space-time attention | |
CN115661576A (en) | Method for identifying airplane group intention under sample imbalance | |
Xiong et al. | Research on quality control methods for surface temperature observations via spatial correlation analysis | |
Lu et al. | Dynamic evolution analysis of desertification images based on BP neural network | |
CN107231668A (en) | Transmission mechanism model under forest fire based on gray scale association analysis | |
CN114595770A (en) | Long time sequence prediction method for ship track |
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 |