CN106526708B - A kind of intelligent early-warning analysis method of the meteorological strong convective weather based on machine learning - Google Patents
A kind of intelligent early-warning analysis method of the meteorological strong convective weather based on machine learning Download PDFInfo
- Publication number
- CN106526708B CN106526708B CN201610844122.8A CN201610844122A CN106526708B CN 106526708 B CN106526708 B CN 106526708B CN 201610844122 A CN201610844122 A CN 201610844122A CN 106526708 B CN106526708 B CN 106526708B
- Authority
- CN
- China
- Prior art keywords
- data
- weather
- meteorological
- radar
- strong
- 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.)
- Active
Links
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01W—METEOROLOGY
- G01W1/00—Meteorology
- G01W1/10—Devices for predicting weather conditions
-
- 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
- Environmental & Geological Engineering (AREA)
- Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Atmospheric Sciences (AREA)
- Biodiversity & Conservation Biology (AREA)
- Ecology (AREA)
- Environmental Sciences (AREA)
- Radar Systems Or Details Thereof (AREA)
Abstract
The present invention discloses a kind of intelligent early-warning method of the meteorological strong convective weather based on machine learning, severe Convective Weather Warnings master data is received by server, including Lighting Position Data, weather electric field data, automatic Weather Station meteorological element live data, radar fact cardinal data etc., localization severe Convective Weather Warnings threshold values is provided according to history meteorological data to be based on regression analysis for each region；Real-time weather integrated data is calculated by the regression analysis of the big data of a variety of meteorological elements, and is compared with threshold value of warning, generates warning information.The present invention carries out backtracking to meteorological historical data using the regression analysis of machine learning and data push back, and according to the calculating of the dynamic of context threshold values and adjustment, radar extrapolating results and automatic Weather Station meteorological element live data are subjected to data comparison simultaneously, with strong convective weather being identified based on machine learning and early warning, the accurate early warning of the strong convection weather to specifying region is realized.
Description
Technical field
The present invention relates to weather forecast and early warning fields, and in particular to a kind of using meteorological big data and based on machine learning
Meteorological strong convective weather intelligent early-warning analysis method.
Background technology
Strong convective weather is one of main meteorological disaster weather, is the object of meteorological department's emphasis monitoring.Strong convection
Weather be occur suddenly, the diastrous weather that mobile rapid, weather is violent, destructive power is extremely strong, mainly have thunderstorm gale, hail,
Cyclone, local heavy showers etc..The identification forecast rate to strong convective weather is effectively and quickly improved, people's life wealth can be reduced
The loss of production and industrial and agricultural production.Different places by factors such as landform, seasons because being influenced so as to severe Convective Weather Warnings
Threshold values will be different, how to be localized severe Convective Weather Warnings analysis be a great problem.The strong convection of traditional approach
In weather warning, do not consider localization problem, only according to weather radar base data data carry out forecasting and warning, pre-alerting ability compared with
Weak, early warning precision is relatively low；In traditional severe Convective Weather Warnings, manual early warning threshold values processing is rule of thumb carried out, often can not
Rational threshold value of warning is obtained, and leads to the erroneous judgement of warning information or reduces the precision of early warning.
Invention content
To solve the above-mentioned problems, it is proposed that a kind of intelligent early-warning analysis of the meteorological strong convective weather based on machine learning
Method, this method carries out backtracking to automatic Weather Station historical data using the regression analysis of machine learning and data push back realization pair
Fixed point region meteorological disaster early warning threshold values carries out deducing identification and setting, and radar extrapolating results and automatic Weather Station meteorological element are live
Data carry out data comparison, with strong convective weather being identified based on machine learning and early warning.
The purpose of the present invention is realized at least through one of following technical solution.
A kind of intelligent early-warning method of the meteorological strong convective weather based on machine learning comprising：
The meteorological historical data progress backtracking in present analysis region and data are pushed back using based on machine learning method, really
Settled prefixed point region meteorological disaster early warning threshold values Y；
The real-time weather master data in present analysis region is received, the real-time weather master data includes lightning location number
According to, atmospheric electric field data, automatic Weather Station meteorological element live data, weather radar base data；
The real-time weather master data of reception is analyzed, the positioning number contained in real-time weather master data is passed through
According to, to strong convective weather generation area carry out positioning and subregion, so that it is determined that the severe Convective Weather Warnings information publishing region；
According to the real-time weather master data of acquisition, the real-time weather integrated data R in comprehensive descision present analysis region, and
It is compared with the threshold value of warning Y of the setting, is determined whether to carry out severe Convective Weather Warnings according to fiducial value size.Specific step
It is rapid as follows：
1.1 settings need the region analyzed；According to the analyzed area of setting, the real-time gas in the region obtained is extracted
As master data data；
The real time radar echo strength data of 1.3 analysis current regions, set radar echo intensity numerical value to X1；Currently
Time in season is set as X2；Air field strength data are analyzed, set its numerical value to X3；Lightning drop point is analyzed, data X4 is set as；
The wind-force for analyzing current region, sets its numerical value to X5；Current region precipitation data is analyzed, sets its numerical value to X6；
1.4 according to preset weight, respectively：Radar echo intensity W1, it time in season W2, air field strength W3, dodges
The data obtained in electric drop point W4, wind-force W5, precipitation W6, with above-mentioned 1.3 are weighted to obtain the reality in present analysis region
When meteorology integrated data R, formula R=X1*W1+X2*W2+X3*W3+X4*W4+X5*W5+X6*W6；
1.5, by real-time weather integrated data R obtained above, are compared with the threshold value of warning Y of setting；As R >=Y,
Warning information is sent out, warning information is not otherwise sent out；The threshold value of warning is carried out according to the historical data of each regional weather station
Comprehensive descision, the computational methods of threshold value of warning are as follows：
Subregion carries out machine learning analysis by the region difference historical data；First, according to history strong convection weather number
According to by machine learning, obtaining the early warning Initial Hurdle Y of a strong convective weather；Secondly according to the last or repeatedly strong right
Gas image data carries out artificial correction to initial early warning Initial Hurdle Y.
Further, the history strong convection weather data refer to since the regional weather station is built a station and having had the history of record
The data such as meteorological data, including thunderstorm gale, hail, cyclone, local heavy showers.
Further, during step 1.5 carries out artificial correction to the early warning Initial Hurdle Y of strong convective weather, the time
Historical data more rearward, reference value and weight are bigger；Closest to current primary or strong convective weather integrates several times
Meteorologic parameter R is more than early warning threshold values Y and shows strong convective weather, if true, there is no generations, corresponding to improve valve
Value, otherwise reduces threshold values.
The weather radar base data, needs through radar extrapolation algorithm, and region, data lattice point are carried out to the data
Processing；The radar extrapolation algorithm is a kind of or more in quantitative Rainfall estimates, radar return tracking, time extrapolation algorithm
Kind；
The quantitative Rainfall estimates are acquired through relational expression Z=a*r*b；Parameter Z is radar emission in the relational expression
Rate, parameter r are rate of rainall, and parameter a is automatic rain gauge data, and b is 2 kilometers of contour radar reflectivitys of height above sea level, with linear regression
Method real time correction；The quantitative Rainfall estimates have just started or rain gauge data deficiencies in the case of is estimated, a, b in rainfall
Parameter is using the default value for closing the local weather of symbol；
The radar return tracking uses the multiple dimensioned light stream calculus of variations, to capture the multiple dimensioned movement of rain belt；More rulers
Spending the light stream calculus of variations is：The objectively Multiple-Scale of analysis rain belt movement, relevant radar reflectivity data, with different points
Resolution is divided into 7 levels and carries out optical flow analysis, from down to the i.e. corresponding scale of high-resolution to small, solving corresponding light one by one greatly
Flow field；
The time extrapolation algorithm is mainly using semi-Lagrange advection (semi-Lagrangian advection, letter
Claim SLA) scheme, in particular to：Semi-Lagrange advection algorithm is used to enough meteorological historical datas, and is carried out in data
Insert, calculate this area the strong convective weather of set period of time probability of happening；The radar extrapolation algorithm is extrapolated in the time
In, it is contemplated that the needs of user personalization reporting services only choose the first higher forecast data of accuracy in two hours；
Base data is analyzed in the data lattice pointization processing, is obtained to linear interpolation or non-linear interpolation method
Go out the lattice point data in the region；What the base data linear interpolation method was suitable for not influenced by extraneous specific obstacle
Region obtains accurate lattice point data if any the region of high building blocking, otherwise, influences size according to the region barrier
Dynamic adjusts the parameter of non-linear interpolation.
Further, the radar return tracking further includes base data analysis, and therefore obtains radar echo intensity
DBZ values, and the influence according to the intensity of radar return and mobile trend to the region sets corresponding parameter value X1；The dBZ
Value is for estimating rainfall and snowfall intensity and predicting the possibility that hail or strong wind diastrous weather occur；The bigger drop of dBZ values
Rain or snowfall possibility are bigger, and intensity is also stronger；When dBZ values are greater than or equal to 40dBZ, occur the possibility of thunderstorm weather compared with
Greatly, when dBZ values are in 45dBZ or more, it is larger to there is heavy rain, hail, the possibility of strong wind strong convective weather.
Further, the setting method of the season time parameter is the method by machine learning, to the regional historical
Data are analyzed and determined, obtain the situation of change of strong convective weather Various Seasonal, obtain air field strength value data X2,
In, summer strong convective weather occurrence frequency is higher, then respective settings weight values answer seizure ratio is higher；
The setting method of the air field strength data weighting value is according to from atmospheric electric field detector and for examining air
The real-time data collection of the lightning monitoring equipment of electric field instrument warning information generates the available field strength data of strong convection business, finally
Obtain air field strength value data X3；
The setting method of the lightning impact parameter is according to acquisition from atmospheric electric field detector and for examining atmospheric electricity
The real time data that the thunder and lightning of field instrument warning information is settled in an area, is processed all kinds of meteorological datas in database, generates strong
The available thunder and lightning of convection current business is settled in an area data, and the numerical value X4 of lightning drop point is finally obtained；
The setting method of the wind data is the real-time wind data monitored according to the Regional automatic station, according to detection
To wind data size finally obtain the numerical value X5 of wind data；
The setting method of the precipitation data is the Real-time Precipitation data monitored according to the Regional automatic station, according to detection
The precipitation data size arrived finally obtains the numerical value X6 of precipitation data.
Compared with the existing technology, the present invention has the following advantages：
The present invention carries out backtracking using the regression analysis of machine learning to automatic Weather Station historical data, radar historical summary
And data push back, and carry out deducing identification and setting to fixed point region meteorological disaster early warning threshold values, and according to the dynamic of context threshold values
Calculating and adjustment, while radar extrapolating results and automatic Weather Station meteorological element live data are subjected to data comparison, with base
Strong convective weather is identified and early warning in machine learning, severe Convective Weather Warnings analysis model is generated, with machine learning
To early warning is identified to convection weather by force.
Description of the drawings
Fig. 1 is a kind of the specific of the intelligent early-warning analysis method of meteorological strong convective weather based on machine learning of the present invention
Flow diagram.
Fig. 2 is one of the intelligent early-warning analysis method of a kind of meteorological strong convective weather based on machine learning of the present invention
Embodiment flow chart.
Fig. 3 is a kind of the another of the intelligent early-warning analysis method of meteorological strong convective weather based on machine learning of the present invention
A embodiment flow chart.
Specific implementation mode
The specific implementation of the present invention is further illustrated below with reference to embodiment and attached drawing, but the guarantor of the present invention
It protects and implements without being limited thereto, be art technology if it is noted that following have not the especially process or parameter of detailed description
Personnel can refer to prior art realization.
Fig. 1 is a kind of the specific of the intelligent early-warning analysis method of meteorological strong convective weather based on machine learning of the present invention
Flow diagram, the present embodiment as shown in the figure include：
S101, severe Convective Weather Warnings analysis process start, and the strong convective weather includes thunderstorm gale, hail, spout
Wind, local heavy showers meteorological disaster.
S102, server receive Lighting Position Data in strong convective weather.
Specifically, the Lighting Position Data includes Ground flash and cloud dodges data, the Ground flash is divided into positive sudden strain of a muscle, negative sudden strain of a muscle,
The cloud sudden strain of a muscle is divided into interior cloud, thin clouds and data, and thus judges that lightning is settled in an area.
S103, server receive atmospheric electric field detector data in strong convective weather.
Specifically, the atmospheric electric field detector is used for measuring atmospheric electric field and its variation, sense is generated in the electric field using conductor
The principle of charge is answered to measure electric field strength.
S104, server receive automatic Weather Station meteorological element live data in strong convective weather.
Specifically, the automatic Weather Station meteorological element live data, including measure meteorological element data, including temperature, humidity,
Wind direction and wind velocity, air pressure, visibility and rainfall.
S105, intellectual analysis judge influence area, according to above-mentioned S102~S104 data, using based on machine learning
The region that is influenced by this factor of method intelligent decision.
S106, server receive strong convective weather in radar fact according to base data.
S107 is handled the radar fact received using radar extrapolation algorithm.
Specifically, the radar extrapolation algorithm includes quantitative Rainfall estimates, radar return tracking, time extrapolation algorithm；
Further, the quantitative Rainfall estimates are acquired through relational expression Z=arb；Parameter Z is thunder in the relational expression
Up to emissivity, parameter r is rate of rainall, and parameter a is automatic rain gauge data, 2 kilometers of contour radar reflectivitys of b height above sea level, with linear
Return Law real time correction；
Further, the quantitative Rainfall estimates, have just started in rainfall or in the case of rain gauge data deficiencies, a, b ginseng
Number is using the default value for closing the local weather of symbol；
Further, the radar return tracking uses the multiple dimensioned light stream calculus of variations in current operation system, to capture rain
The multiple dimensioned movement in area；
Specifically, the analysis scale of the first level is maximum, 4/13 computational domain of rounding for the multiple dimensioned light stream calculus of variations
Width (i.e. the radar scanning range of 256 kilometers of radius), the second layer is then set to 1/5th width of computational domain, and from third
Level starts, and every layer of resolution ratio is set as the double of last layer grade, until final layer 7 grade.And due to the light stream of every level
Analysis all provides first guess by the optical flow field of last layer grade, so among the analysis of final level, especially in no radar
On the region of echo, the motion vector of all large scales has been adopted as heir automatically.Through this multiple dimensioned analysis program, rain band exists
Motion vector under each setting resolution ratio or scale is able to successfully capture to obtain, and the phase of its whole group velocity and individual echoes
Bit rate also can thus reflect in final analysis field.
S108 carries out region, data lattice point to radar extrapolated data.
Specifically, radar extrapolated data is analyzed in the region, the processing of data lattice pointization, to linear interpolation or non-
Linear interpolation method obtains the lattice point data in the region；
Further, the region that the linear interpolation method is suitable for not influenced by extraneous specific obstacle, as high building hinders
Gear, obtains accurate lattice point data, otherwise, influences the ginseng that size dynamic adjusts non-linear interpolation according to the region barrier
Number；
S109 further carries out comprehensive analysis to the above-mentioned data received, judges, positioning strong convective weather region.
Specifically, according to the master data behind localization region, using the regression analysis based on machine learning to automatic
Historical data, the progress backtracking of radar historical summary and the data of standing push back, and are deduced to fixed point region meteorological disaster early warning threshold values
Identification and setting, and according to the calculating of the dynamic of early warning threshold values and adjustment.
S110, input radar is the same as intensity of wave parameter X1.
Specifically, the radar echo intensity dbz values are obtained according to base data analysis, and according to the strong of radar return
Degree and mobile trend, the influence to the region set corresponding parameter value X1.
S111 carries out backtracking according to automatic Weather Station historical data, radar historical summary and data pushes back, to the same intensity of wave of radar
Parameter X1 weighted W1, export new master data, and so on.
S112 inputs season time parameter X2.
S113, to season time parameter X2 weighteds W2.
S114, input air field strength parameter X3.
S115, to air field strength X3 parameter weightings weight W3.
S116, input lightning impact parameter X4.
S117, to lightning impact parameter X4 weighteds W4.
S118, input wind-force parameter X5.
S119, to wind-force parameter X5 weighteds W5.
S120, input precipitation parameter X6.
S121, to precipitation parameter X6 weighteds W6.
S122 carries out read group total, according to formula R=X1*W1+X2* according to all new datas for calculating output after weighting
W2+X3*W3+X4*W4+X5*W5+X6*W6 obtains the comprehensive parameters R of strong convective weather analysis model.
S123 is compared according to the comprehensive parameters R of strong convective weather analysis model and early warning threshold values Y.
It is compared according to modal analysis results and early warning threshold values Y specifically, the push of severe Convective Weather Warnings information sends request,
If R >=Y is big, early warning request is sent, early warning is not otherwise sent.
S124 sends out strong convection warning information.
S125, severe Convective Weather Warnings point analytic process terminate.
Fig. 2 is a kind of the another of the intelligent early-warning analysis method of meteorological strong convective weather based on machine learning of the present invention
Embodiment, the present embodiment as shown in the figure include：
S201, Severe Convective Weather Forecasting flow, the embodiment are that heavy rain strong convective weather judges flow.
S202, heavy rain strong convective weather judge that flow starts.
S203, server receive automatic Weather Station precipitation data in strong convective weather, and the precipitation data refers to measured by automatic Weather Station
Rainfall.
S204, intellectual analysis judge influence area, according to the above-mentioned precipitation data received, using based on machine learning
The region that method intelligent decision is influenced by this factor.
S205, server receive strong convective weather in radar fact according to base data.
S206 is handled the radar fact received using radar extrapolation algorithm.
Specifically, the radar extrapolation algorithm includes quantitative Rainfall estimates, radar return tracking, time extrapolation algorithm；
Further, the quantitative Rainfall estimates are acquired through relational expression Z=arb；Parameter Z is thunder in the relational expression
Up to emissivity, parameter r is rate of rainall, and parameter a is automatic rain gauge data, and b is 2 kilometers of contour radar reflectivitys of height above sea level, with line
Property Return Law real time correction；
Further, the quantitative Rainfall estimates, have just started in rainfall or in the case of rain gauge data deficiencies, a, b ginseng
Number is using the default value for closing the local weather of symbol；
Further, the radar return tracking uses " the multiple dimensioned light stream calculus of variations " in current operation system, to capture
The multiple dimensioned movement of rain belt；
S207 carries out area data lattice point to radar extrapolation process data.
Specifically, base data is analyzed in data lattice pointization processing, to linear interpolation or non-linear insert
Value method obtains the lattice point data in the region；
Further, the area that the base data linear interpolation method is suitable for not influenced by extraneous specific obstacle
Domain obtains accurate lattice point data if high building stops, otherwise, according to the region, barrier influence size dynamic adjusts non-
The parameter of linear interpolation；
S208 carries out comprehensive analysis to the above-mentioned data received, judges, positioning strong convective weather region.
Specifically, according to the master data behind localization region, using the regression analysis based on machine learning to automatic
History of standing precipitation data, radar historical summary carry out backtracking and data push back, and is carried out to fixed point region meteorological disaster early warning threshold values
Identification and setting are deduced, and according to the calculating of the dynamic of context threshold values and adjustment.
S211, input radar is the same as intensity of wave parameter X1.
Specifically, the radar echo intensity dbz values are obtained according to base data analysis, and according to the strong of radar return
Degree and mobile trend, the influence to the region set corresponding parameter value X1.
S212 carries out backtracking according to automatic Weather Station historical data, radar historical summary and data pushes back, to the same intensity of wave of radar
Parameter X1 weighted W1, export new master data, and so on.
S213 inputs season time parameter X2.
S214, to season time parameter X2 weighteds W2.
S215, input air field strength parameter X3.
S216, input precipitation parameter X6.
S217 exports new master data to precipitation parameter X6 weighted W6.
S218 carries out read group total, according to formula R=X1*W1+X2* according to all new datas for calculating output after weighting
W2+X3*W3 obtains the result R of analysis model.
S219, specifically, the push of strong convection rainstorm weather warning information sends request according to modal analysis results and early warning
Threshold values Y comparisons, if R>Y is big, then early warning request is sent, if conversely, R<Y does not send early warning then.
S220 sends out High current microsecond pulse warning information.
S221, strong convection rainstorm weather early warning analysis process terminate.
Fig. 3 is a kind of the another of the intelligent early-warning analysis method of meteorological strong convective weather based on machine learning of the present invention
Embodiment, the present embodiment as shown in the figure include：
S301, Severe Convective Weather Forecasting flow, the embodiment are that thunderstorm strong convective weather judges flow.
S302, thunderstorm strong convective weather judge that flow starts.
S303, server receive Lighting Position Data.
Specifically, the Lighting Position Data includes Ground flash and cloud dodges data, the Ground flash is divided into positive sudden strain of a muscle, negative sudden strain of a muscle,
The cloud sudden strain of a muscle is divided into interior cloud, thin clouds and data, and thus judges that lightning is settled in an area.
S304, server receive atmospheric electric field detector data in strong convective weather.
Specifically, the atmospheric electric field detector is used for measuring atmospheric electric field and its variation, sense is generated in the electric field using conductor
The principle of charge is answered to measure electric field strength.
S305, intellectual analysis judge influence area, according to the above-mentioned precipitation data received, using based on machine learning
The region that method intelligent decision is influenced by this factor.
S306, server receive strong convective weather in radar fact according to base data.
S307 is handled the radar fact received using radar extrapolation algorithm.
Specifically, the radar extrapolation algorithm includes quantitative Rainfall estimates, radar return tracking, time extrapolation algorithm；
Further, the quantitative Rainfall estimates are acquired through relational expression Z=a*r*b；Parameter Z is in the relational expression
Radar emission rate, parameter r are rate of rainall, and parameter a is automatic rain gauge data, and b is 2 kilometers of contour radar reflectivitys of height above sea level, with
Linear regression method real time correction；
Further, the quantitative Rainfall estimates, have just started in rainfall or in the case of rain gauge data deficiencies, a, b ginseng
Number is using the default value for closing the local weather of symbol；
Further, the radar return tracking uses " the multiple dimensioned light stream calculus of variations " in current operation system, to capture
The multiple dimensioned movement of rain belt；
S308 carries out area data lattice point to radar extrapolation process data.
Specifically, base data is analyzed in data lattice pointization processing, to linear interpolation or non-linear insert
Value method obtains the lattice point data in the region；
Further, the area that the base data linear interpolation method is suitable for not influenced by extraneous specific obstacle
Domain obtains accurate lattice point data if high building stops, otherwise, according to the region, barrier influence size dynamic adjusts non-
The parameter of linear interpolation；
S309 carries out comprehensive analysis to the above-mentioned data received, judges, positioning strong convective weather region.
Specifically, according to the master data behind localization region, using the regression analysis based on machine learning to automatic
History of standing precipitation data, radar historical summary carry out backtracking and data push back, and is carried out to fixed point region meteorological disaster early warning threshold values
Identification and setting are deduced, and according to the calculating of the dynamic of context threshold values and adjustment.
S310, input radar is the same as intensity of wave parameter X1.
Specifically, the radar echo intensity dbz values are obtained according to base data analysis, and according to the strong of radar return
Degree and mobile trend, the influence to the region set corresponding parameter value X1.
S311 carries out backtracking according to automatic Weather Station historical data, radar historical summary and data pushes back, to the same intensity of wave of radar
Parameter X1 weighted W1, export new master data, and so on.
S312 inputs season time parameter X2.
S313, to season time parameter X2 weighteds W2.
S314, input air field strength parameter X3.
S315, input lightning drop point X4.
S316, to lightning drop point X4 weighteds W4.
S317 carries out read group total, according to formula R=X1*W1+X2* according to all new datas for calculating output after weighting
W2+X3*W3+X4*W4 obtains the result R of analysis model.
S318, specifically, the push of strong convection rainstorm weather warning information sends request according to modal analysis results and early warning
Threshold values Y comparisons, if R>Y is big, then early warning request is sent, if conversely, R<Y does not send early warning then.
S219 sends out strong convection thunderstorm warning information.
S220, strong convection Thunderstorm Weather early warning analysis process terminate.
According to the disclosure and teachings of the above specification, those skilled in the art in the invention can also be to above-mentioned embodiment party
Formula is changed and is changed.Therefore, the invention is not limited in specific implementation modes disclosed and described above, to the present invention's
Some modifications and changes should also be as falling into the scope of the claims of the present invention.
Claims (6)
1. a kind of intelligent early-warning method of the meteorological strong convective weather based on machine learning, it is characterised in that including：
The meteorological historical data progress backtracking in present analysis region and data are pushed back using based on machine learning method, determination is worked as
Prefixed point region meteorological disaster early warning threshold values Y；
Receive the real-time weather master data in present analysis region, the real-time weather master data include Lighting Position Data,
Atmospheric electric field data, automatic Weather Station meteorological element live data, weather radar base data；
The real-time weather master data of reception is analyzed, by the location data contained in real-time weather master data,
Positioning and subregion are carried out to strong convective weather generation area, so that it is determined that the severe Convective Weather Warnings information publishing region；
According to the real-time weather master data of acquisition, the real-time weather integrated data R in comprehensive descision present analysis region, and with institute
It states current fixed-point region meteorological disaster early warning threshold values Y to be compared, is determined whether to carry out strong convective weather according to fiducial value size
Early warning；It is as follows：
1.1 settings need the region analyzed；According to the analyzed area of setting, the real-time weather in the region obtained is extracted
Master data data；
The real time radar echo strength data of 1.2 analysis current regions, set radar echo intensity numerical value to X1；Current season
Time is set as X2；Air field strength data are analyzed, set its numerical value to X3；Lightning drop point is analyzed, data X4 is set as；Analysis
The wind-force of current region sets its numerical value to X5；Current region precipitation data is analyzed, sets its numerical value to X6；
1.3 according to preset weight, respectively：Radar echo intensity W1, time in season W2, air field strength W3, lightning are fallen
The data obtained in point W4, wind-force W5, precipitation W6, with above-mentioned 1.3 are weighted to obtain the real-time gas in present analysis region
As integrated data R, formula is R=X1*W1+ X2*W2+ X3*W3+ X4*W4+ X5*W5+ X6*W6;
1.4, by real-time weather integrated data R obtained above, are compared with the threshold value of warning Y of setting；As R >=Y, send out
Otherwise warning information does not send out warning information；The threshold value of warning is integrated according to the historical data of each regional weather station
Judge, the computational methods of threshold value of warning are as follows：
Subregion carries out machine learning analysis by the region difference historical data；First, according to history strong convection weather data, lead to
Machine learning is crossed, the early warning Initial Hurdle Y of a strong convective weather is obtained；Secondly according to the last or multiple strong convection weather
Data carry out artificial correction to initial early warning Initial Hurdle Y.
2. the intelligent early-warning method of the meteorological strong convective weather according to claim 1 based on machine learning, feature exist
Refer to since the regional weather station is built a station and having had the history meteorological data of record, including thunder in the history strong convection weather data
The data such as rain strong wind, hail, cyclone, local heavy showers.
3. the intelligent early-warning method of the meteorological strong convective weather according to claim 1 based on machine learning, feature exist
In, during step 1.5 carries out artificial correction to the early warning Initial Hurdle Y of strong convective weather, the history number of time more rearward
According to reference value and weight are bigger；Closest to current primary or strong convective weather synthesis meteorologic parameter R is more than pre- several times
Alert threshold values Y simultaneously shows strong convective weather, if true, there is no generations, corresponding to improve threshold values, otherwise reduce threshold values.
4. the intelligent early-warning method of the meteorological strong convective weather according to claim 1 based on machine learning, feature exist
In the weather radar base data is needed through radar extrapolation algorithm, and region, the processing of data lattice pointization are carried out to the data；
The radar extrapolation algorithm is one or more of quantitative Rainfall estimates, radar return tracking, time extrapolation algorithm；
The quantitative Rainfall estimates are acquired through relational expression Z=a*r*b；Parameter Z is radar emission rate, ginseng in the relational expression
Number r is rate of rainall, and parameter a is automatic rain gauge data, and b is 2 kilometers of contour radar reflectivitys of height above sea level, real-time with linear regression method
Correction；The quantitative Rainfall estimates have just started or rain gauge data deficiencies in the case of is estimated in rainfall, and a, b parameter are adopted
With the default value for closing the local weather of symbol；
The radar return tracking uses the multiple dimensioned light stream calculus of variations, to capture the multiple dimensioned movement of rain belt；The multiple dimensioned light
Rheology point-score is：The objectively Multiple-Scale of analysis rain belt movement, relevant radar reflectivity data, with different resolutions
Rate is divided into 7 levels and carries out optical flow analysis, from down to the i.e. corresponding scale of high-resolution to small, solving corresponding light stream one by one greatly
?；
The time extrapolation algorithm mainly uses semi-Lagrange advection（Semi-Lagrangian advection, referred to as
SLA）Scheme, in particular to：Semi-Lagrange advection algorithm is used to enough meteorological historical datas, and carries out interpolation of data,
Calculate present analysis region the strong convective weather of set period of time probability of happening；The radar extrapolation algorithm is outside the time
In pushing away, it is contemplated that the needs of user personalization reporting services only choose the first higher forecast data of accuracy in two hours；
Base data is analyzed in the data lattice pointization processing, this is obtained to linear interpolation or non-linear interpolation method
The lattice point data in region；The area that the base data linear interpolation method is suitable for not influenced by extraneous specific obstacle
Domain obtains accurate lattice point data, otherwise, influences the ginseng that size dynamic adjusts non-linear interpolation according to the region barrier
Number.
5. the intelligent early-warning method of the meteorological strong convective weather based on machine learning as claimed in claim 4, which is characterized in that
The radar return tracking further includes base data analysis, and therefore obtains radar echo intensity dBZ values, and is returned according to radar
Influence of the intensity and mobile trend of wave to the region sets corresponding parameter value X1；The dBZ values be for estimate rainfall and
The possibility that snowfall intensity and prediction hail or strong wind diastrous weather occur；The bigger rainfall of dBZ values or snowfall possibility are bigger,
Intensity is also stronger；When dBZ values are greater than or equal to 40dBZ, the possibility for thunderstorm weather occur is larger, when dBZ values in 45dBZ or
When above, it is larger to there is heavy rain, hail, the possibility of strong wind strong convective weather.
6. the intelligent early-warning method of the meteorological strong convective weather according to claim 1 based on machine learning, feature exist
In the setting method of the season time parameter is the method by machine learning, and carrying out analysis to the regional historical data sentences
It is disconnected, it obtains the situation of change of strong convective weather Various Seasonal, obtains air field strength value data X2, wherein summer strong convection day
Gas occurrence frequency is higher, then respective settings weight values answer seizure ratio is higher；
The setting method of the air field strength data weighting value is according to from atmospheric electric field detector and for examining atmospheric electric field
The real-time data collection of the lightning monitoring equipment of instrument warning information generates the available field strength data of strong convection business, finally obtains
Air field strength value data X3；
The setting method of the lightning impact parameter is according to acquisition from atmospheric electric field detector and for examining atmospheric electric field detector
The real time data that the thunder and lightning of warning information is settled in an area is processed all kinds of meteorological datas in database, generates strong convection
The available thunder and lightning of business is settled in an area data, and the numerical value X4 of lightning drop point is finally obtained；
The setting method of the wind data is the real-time wind data monitored according to the Regional automatic station, according to what is detected
Wind data size finally obtains the numerical value X5 of wind data；
The setting method of the precipitation data is the Real-time Precipitation data monitored according to the Regional automatic station, according to what is detected
Precipitation data size finally obtains the numerical value X6 of precipitation data.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610844122.8A CN106526708B (en) | 2016-09-21 | 2016-09-21 | A kind of intelligent early-warning analysis method of the meteorological strong convective weather based on machine learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610844122.8A CN106526708B (en) | 2016-09-21 | 2016-09-21 | A kind of intelligent early-warning analysis method of the meteorological strong convective weather based on machine learning |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106526708A CN106526708A (en) | 2017-03-22 |
| CN106526708B true CN106526708B (en) | 2018-10-30 |
Family
ID=58344479
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610844122.8A Active CN106526708B (en) | 2016-09-21 | 2016-09-21 | A kind of intelligent early-warning analysis method of the meteorological strong convective weather based on machine learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106526708B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107301272A (en) * | 2017-05-24 | 2017-10-27 | 天津大学 | Strong convection monomer structure and architectural feature visualization space-time method for decomposing |
Families Citing this family (33)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111344710A (en) * | 2017-09-26 | 2020-06-26 | 沙特阿拉伯石油公司 | Method for cost-effective thermodynamic fluid property prediction using machine learning based models |
| CN107526083B (en) * | 2017-10-18 | 2019-05-31 | 国网新疆电力公司电力科学研究院 | A kind of strong convection wind scale prediction technique based on weather radar data |
| CN108549116A (en) * | 2018-03-27 | 2018-09-18 | 南京恩瑞特实业有限公司 | NRIET blending merge forecasting procedure |
| CN108711266B (en) * | 2018-06-01 | 2019-05-07 | 四川省气候中心 | Thunder and lightning based on atmospheric electric field is short to face local method for early warning |
| CN109239808A (en) * | 2018-08-01 | 2019-01-18 | 平安科技（深圳）有限公司 | Weather forecast method, device, computer equipment and storage medium |
| CN109061650A (en) * | 2018-08-10 | 2018-12-21 | 北京无线电测量研究所 | Networking detection method and system based on array weather radar |
| CN109738970B (en) * | 2018-12-07 | 2021-02-02 | 国网江苏省电力有限公司电力科学研究院 | Method, device and storage medium for realizing lightning early warning based on lightning data mining |
| CN110135654B (en) * | 2019-05-24 | 2021-08-27 | 北京百度网讯科技有限公司 | Method and apparatus for predicting strong convection weather |
| CN110161506B (en) * | 2019-07-01 | 2023-03-31 | 江苏省气象科学研究所 | Classification type hail landing area identification method based on multi-source meteorological observation data |
| CN110632681B (en) * | 2019-09-17 | 2022-05-27 | 国网湖南省电力有限公司 | Machine learning-based short-term and imminent early warning method and system for afternoon thunderstorm of power grid |
| CN110703255A (en) * | 2019-09-19 | 2020-01-17 | 天津大学 | Doppler radial velocity diagram simulation method of typical strong convection flow field |
| CN110967695A (en) * | 2019-10-28 | 2020-04-07 | 兰州大方电子有限责任公司 | Radar echo extrapolation short-term prediction method based on deep learning |
| CN110888186A (en) * | 2019-11-29 | 2020-03-17 | 天津大学 | Method for forecasting hail and short-time heavy rainfall based on GBDT + LR model |
| CN110927725B (en) * | 2019-11-30 | 2023-03-14 | 兰州大学 | Lightning early warning and monitoring method based on meteorological radar |
| CN110888184A (en) * | 2019-12-11 | 2020-03-17 | 安徽蓝业环境工程有限公司 | Environment microclimate monitoring method, device, system and server |
| CN111458768A (en) * | 2020-03-27 | 2020-07-28 | 山东大学 | Strong convection weather early warning method, computer equipment and storage medium |
| CN111783821B (en) * | 2020-05-19 | 2023-09-12 | 知天(珠海横琴)气象科技有限公司 | Method and system for discriminating strong convection gusts |
| CN111796343B (en) * | 2020-06-15 | 2022-07-01 | 兰州中心气象台(兰州干旱生态环境监测预测中心) | Strong convection weather short-time approaching recognition method based on artificial intelligence algorithm |
| CN111736156B (en) * | 2020-08-17 | 2020-11-20 | 南京信大气象科学技术研究院有限公司 | Upwind area identification method and device based on weather radar |
| CN112016615B (en) * | 2020-08-27 | 2024-10-29 | 西安易辑数字科技有限公司 | Strong convection weather identification algorithm based on convolutional neural network learning |
| CN111766641B (en) * | 2020-09-01 | 2020-12-01 | 南京信大气象科学技术研究院有限公司 | Strong convection weather identification method based on deep neural network |
| CN112232674B (en) * | 2020-10-16 | 2021-12-07 | 中国气象局气象探测中心 | Meteorological disaster assessment method, device and system |
| CN113159411B (en) * | 2021-04-14 | 2022-09-13 | 国网河南省电力公司电力科学研究院 | Method and system for testing power grid meteorological risk early warning model |
| CN114325874B (en) * | 2021-08-02 | 2023-11-07 | 苏州市气象局 | Method for establishing strong convection weather personal library system |
| CN114154658A (en) * | 2021-11-29 | 2022-03-08 | 北京比福特科技发展有限公司 | Wind power plant maintenance route calculation method and system based on oceanographic weather |
| CN114488155B (en) * | 2021-12-27 | 2023-06-27 | 国网山东省电力公司应急管理中心 | Convection monomer evolution trend discrimination method based on Doppler weather radar |
| CN114355482B (en) * | 2022-01-04 | 2022-08-16 | 成都信息工程大学 | Strong convection weather identification method and system based on optical flow and meteorological numerical prediction |
| CN114488351A (en) * | 2022-02-10 | 2022-05-13 | 中国民用航空华东地区空中交通管理局 | System and method for monitoring and early warning of field intensity echo |
| CN114935788B (en) * | 2022-07-21 | 2022-10-21 | 国网江苏省电力有限公司苏州供电分公司 | Multi-point position lightning combined early warning method and system for primary equipment of transformer substation |
| CN114994801B (en) * | 2022-08-05 | 2022-10-25 | 中国气象局公共气象服务中心(国家预警信息发布中心) | Lightning monitoring and early warning method and device |
| CN115598739A (en) * | 2022-11-08 | 2023-01-13 | 广东电网有限责任公司广州供电局（Cn） | Lightning potential forecasting method, device, equipment and medium |
| CN115494507B (en) * | 2022-11-21 | 2023-02-14 | 南京气象科技创新研究院 | Lightning identification method based on satellite-borne dual-frequency precipitation radar and microwave imager |
| CN118192312A (en) * | 2024-03-13 | 2024-06-14 | 山东链云科技有限公司 | Greenhouse curtain rolling machine control method and device and greenhouse curtain rolling machine |
Family Cites Families (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| AU1633902A (en) * | 2000-12-15 | 2002-06-24 | Daniel Rosenfeld | Location-based weather nowcast system and method |
| JP4266858B2 (en) * | 2004-03-15 | 2009-05-20 | Ｎｅｃエンジニアリング株式会社 | Local heavy rain monitoring system |
| CN101634721B (en) * | 2009-04-15 | 2011-08-31 | 华东师范大学第二附属中学 | Historical data based intelligent early warning system for typhoon and flood |
| CN104166171B (en) * | 2014-08-25 | 2016-09-21 | 清华大学 | The Forecasting Methodology of thundercloud area coverage based on lightning location system and system |
| CN105068149B (en) * | 2015-07-24 | 2017-04-12 | 国家电网公司 | Multi-information integration-based thunder and lightning monitoring and forecasting method for electric transmission and transformation equipment |
-
2016
- 2016-09-21 CN CN201610844122.8A patent/CN106526708B/en active Active
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107301272A (en) * | 2017-05-24 | 2017-10-27 | 天津大学 | Strong convection monomer structure and architectural feature visualization space-time method for decomposing |
| CN107301272B (en) * | 2017-05-24 | 2020-05-26 | 天津大学 | Strong convection monomer structure and structural feature visualization time-space analysis method |
Also Published As
| Publication number | Publication date |
|---|---|
| CN106526708A (en) | 2017-03-22 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106526708B (en) | A kind of intelligent early-warning analysis method of the meteorological strong convective weather based on machine learning | |
| US8818029B1 (en) | Weather forecasting systems and methods | |
| CN105334551B (en) | Power network weather prognosis early warning system based on numerical weather prediction model | |
| US11353625B1 (en) | Systems and methods for forecasting lightning and severe storms | |
| CN110579823B (en) | Method and system for forecasting short-term and temporary heavy rainfall | |
| CN109325633B (en) | Weather forecast set member selection method and system | |
| CN106950614A (en) | A kind of region automatic weather station hour rainfall data method of quality control | |
| US6894638B2 (en) | Radar signal processing unit and radar signal processing method for abnormal signal extraction | |
| CN112764129A (en) | Method, system and terminal for thunderstorm short-term forecasting | |
| JPH08122433A (en) | Thundercloud observation system | |
| CN104237890B (en) | The heavy rain identification that one is caused by " train effect " and forecasting procedure | |
| JP6689396B2 (en) | Weather forecasting device, weather forecasting method, and program | |
| CN104200082B (en) | Landed Typhoon Forecasting Methodology | |
| CN114325879B (en) | Quantitative precipitation correction method based on classification probability | |
| JP2019138737A (en) | Thunder risk determination device | |
| CN114863369B (en) | Method, device, equipment and medium for monitoring corn lodging by laser radar | |
| CN112327265A (en) | Division and treatment detection method based on semantic segmentation network | |
| Licznar et al. | Precipitation type specific radar reflectivity-rain rate relationships for Warsaw, Poland | |
| CN114019503A (en) | FOD detection system-based airport runway foreign matter detection method, device and storage medium | |
| CN116879900A (en) | Thunderstorm centroid positioning and tracking prediction method based on gray processing | |
| CN110441830B (en) | Small-area weather prediction method based on automatic weather station and storage medium | |
| CN112101639A (en) | Airport traffic capacity analysis method and system based on multi-source meteorological data | |
| CN116736409A (en) | Automobile safety early warning method, device, equipment and storage medium | |
| CN116702588A (en) | Wind power plant key weather factor forecasting method and system based on multi-source data | |
| CN114186726A (en) | Meteorological data processing method, and prompting method and device based on meteorological data |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| CB02 | Change of applicant information |
Address after: 528000 Guangdong, Foshan, Chancheng District, No. three Jiangwan Road, No., the software industry park area A, building 9 Applicant after: Guangdong Aobo Chengdu Westone Information Industry Inc Address before: 528000 Guangdong, Foshan, Chancheng District, No. three Jiangwan Road, No., the software industry park area A, building 9 Applicant before: Guangdong Aobo Information Industry Co Ltd |
|
| CB02 | Change of applicant information | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |