CN108491877A - A kind of classification convection weather probability forecast method and device - Google Patents
A kind of classification convection weather probability forecast method and device Download PDFInfo
- Publication number
- CN108491877A CN108491877A CN201810230385.9A CN201810230385A CN108491877A CN 108491877 A CN108491877 A CN 108491877A CN 201810230385 A CN201810230385 A CN 201810230385A CN 108491877 A CN108491877 A CN 108491877A
- Authority
- CN
- China
- Prior art keywords
- forecast
- weather
- convection
- field
- convection weather
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Business, Economics & Management (AREA)
- Strategic Management (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Human Resources & Organizations (AREA)
- Economics (AREA)
- General Physics & Mathematics (AREA)
- Development Economics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Evolutionary Biology (AREA)
- Game Theory and Decision Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Engineering & Computer Science (AREA)
- Entrepreneurship & Innovation (AREA)
- Marketing (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Tourism & Hospitality (AREA)
- General Business, Economics & Management (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The embodiment of the invention discloses a kind of classification convection weather probability forecast method and device, method includes:It obtains the null field of the NWP in forecast area or analyzes field again, choose observation data;Field is analyzed according to the observation data markers null field of convection weather or again, builds the sample set of convection weather;Predetermined depth learning model is trained according to the sample set of convection weather, obtains the forecasting model of classification convection weather probability;Forecast field data is inputted forecasting model by the forecast field data for obtaining the NWP in forecast area, obtains the probability of happening of convection weather, and the convection weather reporting services of corresponding period are carried out according to probability of happening and numerical weather prediction model.By observing the null field of the numerical weather forecast in data markers forecast area or analyzing field again, build the sample set of convection weather, training deep learning model obtains the forecasting model of classification convection weather probability and carries out weather forecast, disclosure satisfy that forecast requirement of the different user to convection weather.
Description
Technical field
The present embodiments relate to technical field of data processing, and in particular to a kind of classification convection weather probability forecast method
And device.
Background technology
Strong convective weather refer to this refer to convective system occur lightning, Convective strong wind (>=17.2m/s),
The diastrous weathers such as hail, short-time strong rainfall (>=20mm/h), spout are with high one of the diastrous weather influenced.China
All kinds of strong convective weathers are multiple, cause larger casualties and economic loss.However, Convective Weather System scale is small, office
Ground is strong, the raw development that disappears is quick, therefore has larger forecast difficulty.
In recent years, national weather forecast business department attempts to carry out Severe Convective Weather Forecasting business.Strong convection subjectivity and
What objective forecast used is mainly " dosing method ".Dosing method determines the basic constituent element (i.e. " dispensing ") of forecast, inscape
Usually relatively independent meteorological variables provide the clearly thinking of weather forecast to forecaster.Object based on different weather
Manage model and " dosing method ", the convection environment that weatherman devises different type strong convective weather occurrence and development is (steam, dynamic
Power, energy etc.) physical parameter combinations, give class probability forecast model products.Strong convective weather " dosing method " etc. is subjective pre-
Report Product evaluation shows forecaster's subjectivity extraction convection current feature and has by the strong convection forecast result that " dispensing combination " obtains
Preferable validity and prediction ability, but necessarily can also have some limitations.
First, Chinese Regional is vast, topography is complicated, different regions orographic condition, weather conditions significant difference, strong convection day
Gas occurrence and development condition will necessarily have differences with characteristic threshold value range, it is difficult to use a set of unified feature physical quantity sets of threshold values
It closes to realize the forecast of whole nation classification strong convective weather;Secondly, it needs to use mass data during weather forecast, by forecast
The subjective extraction feature physical quantity of member and threshold range, are easy to ignore a large amount of valuable detailed information hidden in data, especially
It is some Small and Medium Sized information;In addition, strong convective weather is complicated and changeable, if forecaster recognizes strong convection regularity of occurrence and development
Know not enough deeply, comprehensively, will necessarily also ignore some developing validity features of strong convective weather.
The Heuristics and the awareness to strong convective weather that the subjective extraction of physics-mechanics character is limited to forecaster, profit
Above-mentioned limitation will effectively be broken through by carrying out Automatic Feature Extraction with machine learning method.Existing machine learning algorithm is in weather forecast
Also there were many trials in field, but its value of forecasting still cannot be satisfied forecast requirement of the different user to convection weather.
Invention content
Since existing method is there are the above problem, the embodiment of the present invention proposes a kind of classification convection weather probability forecast method
And device.
In a first aspect, the embodiment of the present invention proposes a kind of classification convection weather probability forecast method, including:
It obtains the null field of the numerical weather forecast NWP in forecast area or analyzes field again, choose the observation number of convection weather
According to;
Null field described in observation data markers according to the convection weather analyzes field again, builds the sample of convection weather
Collection;
Predetermined deep learning model is trained according to the sample set of the convection weather, it is general to obtain classification convection weather
The forecasting model of rate;
The forecast field data is inputted the forecasting model by the forecast field data for obtaining the NWP in the forecast area,
The probability of happening of convection weather is obtained, and carries out the convection current of corresponding period according to the probability of happening and numerical weather prediction model
Weather forecast service.
Optionally, null field described in the observation data markers according to the convection weather or field is analyzed again, build convection current
The sample set of weather, specifically includes:
The null field is marked using over-sampling or down-sampling mode according to the observation data of the convection weather or is analyzed again
, build the sample set of convection weather.
Optionally, the forecast field data is inputted institute by the forecast field data for obtaining the NWP in the forecast area
Forecasting model is stated, obtains the probability of happening of convection weather, and it is pre- to carry out according to the probability of happening convection weather of corresponding period
Report service, specifically includes:
The forecast field data is normalized in the forecast field data for obtaining the NWP in the forecast area, will
Forecast field data after normalized inputs the forecasting model, obtains the probability of happening of convection weather, and according to the hair
Raw probability carries out the convection weather probability forecast of corresponding period.
Optionally, the feature of the sample set include the temperature of numerical weather prediction model output, it is geopotential unit, wet
Degree and the basic meteorological element feature of wind field and longitude, latitude, height above sea level geographic factor, and/or common strong convection index.
Optionally, it includes pair that the predetermined deep learning model, which uses Softmax graders, the Softmax graders,
Flow the probability of happening of weather 0-1.
Second aspect, the embodiment of the present invention also propose a kind of classification convection weather probability forecast device, including:
Data decimation module, null field for obtaining the numerical weather forecast NWP in forecast area or analyzes field again, chooses
The observation data of convection weather;
Sample set builds module, analyzes for null field described in the observation data markers according to the convection weather or again field,
Build the sample set of convection weather;
Model training module, for being trained to predetermined deep learning model according to the sample set of the convection weather,
Obtain the forecasting model of classification convection weather probability;
Weather forecast module, the forecast field data for obtaining the NWP in the forecast area, by the forecast field data
The forecasting model is inputted, obtains the probability of happening of convection weather, and according to the probability of happening and numerical weather prediction model
Carry out the convection weather reporting services of corresponding period.
Optionally, the sample set structure module is specifically used for using over-sampling according to the observation data of the convection weather
Or down-sampling mode marks the null field or analyzes field again, builds the sample set of convection weather.
Optionally, the weather forecast module is specifically used for obtaining the forecast field data of the NWP in the forecast area,
The forecast field data is normalized, the forecast field data after normalized is inputted into the forecasting model, is obtained
To the probability of happening of convection weather, and carry out according to the probability of happening convection weather reporting services of corresponding period.
The third aspect, the embodiment of the present invention also propose a kind of electronic equipment, including:
At least one processor;And
At least one processor being connect with the processor communication, wherein:
The memory is stored with the program instruction that can be executed by the processor, and the processor calls described program to refer to
Order is able to carry out the above method.
Fourth aspect, the embodiment of the present invention also propose a kind of non-transient computer readable storage medium, the non-transient meter
Calculation machine readable storage medium storing program for executing stores computer program, and the computer program makes the computer execute the above method.
As shown from the above technical solution, the embodiment of the present invention is pre- by observing the Numerical Weather in data markers forecast area
The null field of report analyzes field again, builds the sample set of convection weather, and training deep learning model obtains classification convection weather probability
Forecasting model and carry out weather forecast, disclosure satisfy that forecast requirement of the different user to convection weather.
Description of the drawings
In order to more clearly explain the embodiment of the invention or the technical proposal in the existing technology, to embodiment or will show below
There is attached drawing needed in technology description to be briefly described, it should be apparent that, the accompanying drawings in the following description is only this
Some embodiments of invention for those of ordinary skill in the art without creative efforts, can be with
Other attached drawings are obtained according to these figures.
Fig. 1 is a kind of flow diagram for classification convection weather probability forecast method that one embodiment of the invention provides;
Fig. 2 is a kind of flow diagram for classification convection weather probability forecast method that another embodiment of the present invention provides;
Fig. 3 is a kind of structural schematic diagram for classification convection weather probability forecast device that one embodiment of the invention provides;
Fig. 4 is the logic diagram for the electronic equipment that one embodiment of the invention provides.
Specific implementation mode
Below in conjunction with the accompanying drawings, the specific implementation mode of the present invention is further described.Following embodiment is only used for more
Technical scheme of the present invention is clearly demonstrated, and not intended to limit the protection scope of the present invention.
Fig. 1 shows a kind of flow diagram of classification convection weather probability forecast method provided in this embodiment, including:
S101, numerical weather forecast (Numerical Weather Prediction, NWP) in forecast area is obtained
Null field analyzes field again, chooses the observation data of convection weather.
Specifically, obtain forecast area in NWP null field or analyze field again, choose temperature, air pressure, humidity, wind field and
The observation data of common strong convection index and convection weather (thunderstorm, short-time strong rainfall, hail, thunderstorm gale etc.).
Null field described in S102, the observation data markers according to the convection weather analyzes field again, builds convection weather
Sample set.
S103, predetermined deep learning model is trained according to the sample set of the convection weather, obtains classification convection current
The forecasting model of weather probability.
Wherein, the predetermined deep learning model uses Softmax graders, and the Softmax graders include convection current
The probability of happening of weather 0-1.
The feature of the sample set includes temperature, geopotential unit, humidity and the wind of the numerical weather prediction model output
The geographic factor of the basic meteorological element feature in field and longitude, latitude, height above sea level, and/or common strong convection index.Wherein, institute
The feature for stating sample set must include the bases such as temperature, geopotential unit, humidity and the wind field of numerical weather prediction model output
The geographic factors such as this meteorological element feature and longitude, latitude, height above sea level;May include common strong convection index, it can not also
Including common strong convection index.
Specifically, predetermined deep learning model is trained it using convection weather sample set, automatic to learn convection current day
Gas occurrence condition, obtains forecasting model.
Deep learning (Deep Learning) is the artificial neural network with multilayer hidden layer.With shallow-layer neural network, branch
It holds the conventional machines learning algorithm such as vector machine to compare, deep neural network can provide modeling for Complex Nonlinear System, can be with
Compacter succinct mode expresses the function set more much bigger than shallow-layer network, for model provides higher abstraction hierarchy,
To improve the ability in feature extraction of model, has compared to conventional method in fields such as image recognition, speech processes and significantly carry
It rises.Deep learning network application to strong convection is forecast field by the present embodiment, can be obtained better than artificial forecast and traditional objective
The value of forecasting of forecasting procedure.
S104, the forecast field data for obtaining NWP in the forecast area, by the forecast field data input forecast
Model obtains the probability of happening of convection weather, and carries out the corresponding period according to the probability of happening and numerical weather prediction model
Convection weather reporting services.
The present embodiment also provides a kind of probability forecast system of classification convection weather, which is based on GIS-Geographic Information System
The Visualization Platform of GIS, including server and client side, as shown in Fig. 2, the server is used for:Obtain classification convection current observation
Data and NWP null fields analyze field data again, analyze using observation data markers numerical weather forecast null field or again field data, obtain
To training and test sample collection;Including for supporting concurrent operation and with video card (GPU, the Graphics of multioperation core
Processing Unit), using training sample, training deep learning model obtains classification convection current probability forecast model;It obtains
The forecast field data of numerical weather prediction model in forecast area, and be normalized;Including parallel for supporting
Normalized forecast data is inputted deep learning forecasting model, calculated by operation and the video card (GPU) with multioperation core
To convection weather probability of happening, the lattice point probability weather forecast field in forecast range is obtained.
In the present embodiment, convection weather occurrence and development are automatically extracted from history big-sample data using deep learning model
The features such as steam, instability condition, energy judge strong convection probability of happening, and the subjectivity such as " dosing method " is based in advance to substitute with this
Reporting method and conventional machines learn Objective forecasting method, improve the prediction performance of convection weather forecast, improve convection weather
Forecast work efficiency;The ability of convection current feature is automatically extracted using deep learning network, history big-sample data collection is built, to depth
Degree learning network is trained, and obtains forecasting model, can realize the reporting services of classification convection weather automatically.
The present embodiment is built by observing the null field of the numerical weather forecast in data markers forecast area or analyzing field again
The sample set of convection weather, further training pattern obtain the forecasting model of classification convection weather probability and carry out weather forecast,
It disclosure satisfy that forecast requirement of the different user to convection weather.
Further, on the basis of above method embodiment, S102 is specifically included:
The null field is marked using over-sampling or down-sampling mode according to the observation data of the convection weather or is analyzed again
, build the sample set of convection weather.
Specifically, it analyzes using the observation data logarithm weather forecast null field of convection weather or again characteristic and carries out
Label;During marker samples, since considerably beyond strong convection sample does not occur for the generation sample of strong convection event, use
The methods of over-sampling or down-sampling make positive negative sample balance.
Further, on the basis of above method embodiment, S104 is specifically included:
The forecast field data is normalized in the forecast field data for obtaining the NWP in the forecast area, will
Forecast field data after normalized inputs the forecasting model, obtains the probability of happening of convection weather, and according to the hair
Raw probability carries out the convection weather reporting services of corresponding period.
The embodiment of the present invention provides a kind of automatic probability forecast method and system of classification convection weather, it is intended to solve current
In convection weather forecasting process, main, Objective forecasting method and conventional machines study forecasting procedure forecast based on " dosing method " etc.
The poor problem of effect.
In order to illustrate technical solutions according to the invention, illustrated below by specific embodiment.Fig. 3 illustrates this
The implementation process for the convection weather forecasting procedure that inventive embodiments provide, details are as follows:
As shown in S301, obtains the numerical weather prediction model null field in forecast area or analyze field and convection current day again
After gas observes data, structure training and test sample collection.Such as 1 ° × 1 °F NL from NCEP points of Forecast characteristic (Feature)
Analysis data.NCEP FNL analysiss of data one day four times (02:00,08:00,14:00 and 20:00BJT), it is capable of providing global range
Analysis field.Forecast characteristic not only has chosen the basic meteorological element field such as temperature, geopotential unit, humidity, wind field, while also choosing
The common strong convection physical quantity of the conditions such as reflection steam, power, energy, such as BCAPE (optimal convective available potential energy), PWAT
The strong convections such as (flood precipitable water), K indexes.In view of the geographical environment difference that strong convection occurs, height above sea level, warp are also had chosen
The features such as degree, latitude amount to feature 144.
Thunderstorm and strong convective weather live data are supervised using the strong convection business of the strong weather forecast center of National Meteorological Center
Measured data collection.Thunderstorm observes the data then Lighting Position Data from National Lightning Detecting Network.Precipitation measurement data are from the whole nation
The rainfall gauge of face weather station observes data, and wherein short-time strong rainfall data refer to observation of hour rainfall not less than 20mm and remember
Record.Hail and thunderstorm gale then sources such as the significant weather report from national weather station, hazardous weather report.
Thunderstorm and strong convective weather can be seen as the classification problem of one 2 classification, and 0 is does not occur, and 1 is generation.Therefore,
With strong convection fact come tag value pattern analysis field, structure training and test sample collection.Since numerical model forecast is lattice point
, and fact is observed scatterplot data, therefore observation data lattice pointization is handled, if there is such day within the scope of lattice point radius R
Gas, then it is assumed that weather (can mark and be) occurs in the lattice point, otherwise not occur strong convection lattice point (label 0).In view of pair
Stream weather is generated by small mesoscale system, is finally taken R by experiment, if R is too small, is susceptible to strong convection and fails to report, instead
Be then susceptible to empty report.
Using the above method, about 430,000, thunderstorm sample is had chosen, short-time strong rainfall sample about 110,000, hail sample is about
3.6 ten thousand, 110,000, thunderstorm gale sample.
Such as S302, deep learning model is built.This example constructs depth convolution disaggregated model to carry out convection weather
Forecast.The structure of depth convolutional network is by convolutional layer (Convolution), pond layer (Pooling), full articulamentum (Fully
Connect), the compositions such as Softmax graders, activation primitive choose ReLU functions.
Such as S303, trains deep learning model using above-mentioned training and test sample collection, in training process, be related to various
The adjustment of hyper parameter is obtained with highest classification accuracy, the strongest forecasting model of generalization ability.
Such as S304, the NCEP comprising 144 Forecast characteristics is forecast into field data, is normalized, and feature is turned
The 4 dimension groups (M is forecast sample size) for turning to M × 12 × 12 × 1, as forecast data.
As S305 is passed through based on the trained deep learning Convective Forecasting model and the forecast data
Forecasting model operation obtains the probability forecast of convection weather.As a result of Softmax graders, therefore forecast result is lattice
Point probability forecast product.
Operand is huge during depth convolutional network model training, compared with a small amount of logical unit of CPU, GPU
It is exactly a huge calculating matrix, GPU has thousands of calculating cores, supports vital to deep learning parallel
Computing capability, can be quicker than conventional processors, greatly accelerates training process.Training is used with forecasting process
The library NVidia CUDA (Compute Unified Device Architecture), hardware use NVidia
Gerforece1080Ti video cards.
Test shows that 1 ° × 1 ° within the scope of -72 hours China's Mainlands of full 0 of lattice point forecast can be completed in 10 minutes, complete
The full demand that disclosure satisfy that service application.
It should be noted that the size of the serial number of each step is not meant that the order of the execution order in above-described embodiment,
The execution sequence of each process should be determined by its function and internal logic, and the implementation process without coping with the embodiment of the present invention, which is constituted, appoints
What is limited.The foregoing is merely illustrative of the preferred embodiments of the present invention, is not intended to limit the invention, all spirit in the present invention
With within principle made by all any modification, equivalent and improvement etc., should all be included in the protection scope of the present invention.
Compare the performance of deep learning model and conventional machines learning algorithm on same short-time strong rainfall data set, hair
It is existing:
Traditional machine learning algorithm has Severe Convective Weather Forecasting certain validity, such as logistic regression side
The method training time is short, and classification accuracy reaches 90.8%;The decision Tree algorithms training time is relatively long, obtains 91.36% standard
True rate, multi-layer perception (MLP) can then reach 92.2% accuracy rate.
In contrast, for deep learning model due to having deeper network structure, model parameter is more, therefore the training time
Longer, training performance tool improves, such as 1800 seconds Belief Network training times of Deep, training accuracy rate
93.3%;And depth convolutional network, it is trained using GPU, by training in 1200 seconds, 93.5% accuracy can be reached.
It can be seen that relative to traditional machine learning algorithm, deep learning algorithm is in Severe Convective Weather Forecasting data set
It is upper that there is preferably performance.
Show the Subjective forecast relative to forecaster, deep learning by the forecast experiment of 4-9 month synoptic process in 2015
Algorithm all has improving for the prediction performance of all kinds of strong convections.
The TS (critical success index) of short-time strong rainfall has reached 0.32, and 0.25 relative to forecaster improves 28%.
Rate of failing to report has certain improvement relative to forecaster with empty report rate.It is noted that the rate of failing to report (0.66) of forecaster is big
In empty report rate (0.52), it is meant that there is a large amount of precipitation to fail to report.On the contrary, the rate of failing to report (0.48) of deep learning algorithm is less than sky
Report rate (0.55), lower rate of failing to report can provide better directive function, and the guidance to forecaster can be better achieved and make
With.
The forecast TS scorings of thunderstorm have reached 0.42, have been more than the 0.37 of forecaster, have improved 13.5%. deep learnings
Model has higher hit rate, lower sky report rate, but rate of failing to report is higher with respect to forecaster.
The raising being presented with by a relatively large margin of hail and thunderstorm gale relative to forecaster, TS are respectively reached
0.065 and 0.063, corresponding Subjective forecast 0.025 and 0.057 improves 160% and 10.5% respectively.Deep learning model exists
Hit rate, empty report rate, rate of failing to report etc. have better performance.
Corresponding to the probability Objective forecasting method of the convection weather described in foregoing embodiments, Fig. 2 shows implementations of the present invention
The structure diagram of the nowcasting system for the convection weather that example provides illustrates only related to the present embodiment for convenience of description
Part.The system is the visualization nowcasting platform based on GIS-Geographic Information System, is based on B/S and C/S hybrid architectures
It realizes, C/S structure programs (background program) run on server end, and multi-thread concurrent executes, obtaining NWP null fields and analyzing again
After field data, training sample set is built, obtains deep learning training pattern, and normalized forecast field data is inputted into forecast mould
Type obtains class probability Convective Forecasting.Client is that display platform is used to show institute as B/S structure programs (foreground program)
State the probability forecast result of server.Design storage is reasonable and possesses the database purchase system of good data interface, design
Mainly including relationship and storing process between table structure (data field, data type, remarks), table and application layer (such as index)
Design and structure.Background program system is built, server end is run on, is responsible for multi-thread concurrent execution.In view of data when
Effect property, background program multithreading executes, and is uniformly included in background program system, and the state of background program operation can monitor can be preceding
Platform is shown.
Whole system is based on Python design and realizes that program is run in Windows PC servers, and server is extremely
Include one piece of NVidia GPU, the forecast operation for accelerating deep learning forecasting model less.Test result shows that system is connecing
1 ° × 1 ° in 0-72 hours the whole country of 4 kinds of convection weathers forecast can be completed after receiving NWP data, in 10 minutes.Therefore,
System has stronger business practicability.
It is described Fig. 2 shows a kind of structural schematic diagram of classification convection weather probability forecast device provided in this embodiment
Device includes:Data decimation module 201, sample set structure module 202, model training module 203 and weather forecast module 204,
Wherein:
The data decimation module 201 is used to obtain the null field of the numerical weather forecast NWP in forecast area or analyzes again
, choose the observation data of convection weather;
The sample set structure module 202 is for null field described in the observation data markers according to the convection weather or divides again
Field is analysed, the sample set of convection weather is built;
The model training module 203 is used to carry out predetermined deep learning model according to the sample set of the convection weather
Training obtains the forecasting model of classification convection weather probability;
The weather forecast module 204 is used to obtain the forecast field data of the NWP in the forecast area, by the forecast
Field data inputs the forecasting model, obtains the probability of happening of convection weather, and pre- according to the probability of happening and Numerical Weather
Report pattern carries out the convection weather reporting services of corresponding period.
Specifically, the data decimation module 201 obtains the null field or again of the numerical weather forecast NWP in forecast area
Field is analyzed, the observation data of convection weather are chosen;The sample set builds observation data of the module 202 according to the convection weather
It marks the null field or analyzes field again, build the sample set of convection weather;The model training module 203 is according to the convection current day
The sample set of gas is trained predetermined deep learning model, obtains the forecasting model of classification convection weather probability;The weather
Forecast module 204 obtains the forecast field data of the NWP in the forecast area, and the forecast field data is inputted the forecast mould
Type obtains the probability of happening of convection weather, and carries out the corresponding period according to the probability of happening and numerical weather prediction model
Convection weather reporting services.
In the server, the server is connect with client for the classification convection weather probability forecast device deployment.Institute
Stating server further includes:GPU scientific algorithm video cards, training and forecast for accelerating deep learning model.
Wherein, the model training module 203 and the weather forecast module 204 include at least one piece of video card (GPU,
Graphics Processing Unit);Video card supports concurrent operation and has multioperation core, for accelerating deep learning pre-
The operation of model is reported to calculate.
The present embodiment is built by observing the null field of the numerical weather forecast in data markers forecast area or analyzing field again
The sample set of convection weather, training deep learning model obtains the forecasting model of classification convection weather probability and to carry out weather pre-
Report, disclosure satisfy that forecast requirement of the different user to convection weather.
Further, on the basis of above-mentioned apparatus embodiment, the sample set structure module is specifically used for according to
The observation data of convection weather are marked the null field or are analyzed field again using over-sampling or down-sampling mode, build convection weather
Sample set
Further, on the basis of above-mentioned apparatus embodiment, the weather forecast module is specifically used for obtaining described pre-
The forecast field data for reporting the NWP in region, is normalized the forecast field data, by the forecast after normalized
Field data inputs the forecasting model, obtains the probability of happening of convection weather, and carry out the corresponding period according to the probability of happening
Convection weather probability forecast.
Classification convection weather probability forecast device described in the present embodiment can be used for executing above method embodiment, former
Reason is similar with technique effect, and details are not described herein again.
Reference Fig. 4, the electronic equipment, including:Processor (processor) 401, memory (memory) 402 and total
Line 403;
Wherein,
The processor 401 and memory 402 complete mutual communication by the bus 403;
The processor 401 is used to call the program instruction in the memory 402, to execute above-mentioned each method embodiment
The method provided.
The present embodiment discloses a kind of computer program product, and the computer program product includes being stored in non-transient calculating
Computer program on machine readable storage medium storing program for executing, the computer program include program instruction, when described program instruction is calculated
When machine executes, computer is able to carry out the method that above-mentioned each method embodiment is provided.
The present embodiment provides a kind of non-transient computer readable storage medium, the non-transient computer readable storage medium
Computer instruction is stored, the computer instruction makes the computer execute the method that above-mentioned each method embodiment is provided.
The apparatus embodiments described above are merely exemplary, wherein the unit illustrated as separating component can
It is physically separated with being or may not be, the component shown as unit may or may not be physics list
Member, you can be located at a place, or may be distributed over multiple network units.It can be selected according to the actual needs
In some or all of module achieve the purpose of the solution of this embodiment.Those of ordinary skill in the art are not paying creativeness
Labour in the case of, you can to understand and implement.
Through the above description of the embodiments, those skilled in the art can be understood that each embodiment can
It is realized by the mode of software plus required general hardware platform, naturally it is also possible to pass through hardware.Based on this understanding, on
Stating technical solution, substantially the part that contributes to existing technology can be expressed in the form of software products in other words, should
Computer software product can store in a computer-readable storage medium, such as ROM/RAM, magnetic disc, CD, including several fingers
It enables and using so that a computer equipment (can be personal computer, server or the network equipment etc.) executes each implementation
Method described in certain parts of example or embodiment.
It should be noted that:The above embodiments are merely illustrative of the technical solutions of the present invention, rather than its limitations;Although reference
Invention is explained in detail for previous embodiment, it will be understood by those of ordinary skill in the art that:It still can be right
Technical solution recorded in foregoing embodiments is modified or equivalent replacement of some of the technical features;And this
A little modification or replacements, the spirit and model of various embodiments of the present invention technical solution that it does not separate the essence of the corresponding technical solution
It encloses.
Claims (10)
1. a kind of classification convection weather probability forecast method, which is characterized in that including:
It obtains the null field of the numerical weather forecast NWP in forecast area or analyzes field again, choose the observation data of convection weather;
Null field described in observation data markers according to the convection weather analyzes field again, builds the sample set of convection weather;
Predetermined deep learning model is trained according to the sample set of the convection weather, obtains classification convection weather probability
Forecasting model;
The forecast field data is inputted the forecasting model, obtained by the forecast field data for obtaining the NWP in the forecast area
The probability of happening of convection weather, and carry out according to the probability of happening and numerical weather prediction model the convection weather of corresponding period
Reporting services.
2. according to the method described in claim 1, it is characterized in that, the observation data markers institute according to the convection weather
It states null field or analyzes field again, build the sample set of convection weather, specifically include:
The null field is marked using over-sampling or down-sampling mode according to the observation data of the convection weather or analyzes field, structure again
Build the sample set of convection weather.
3. according to the method described in claim 1, it is characterized in that, the forecast fields for obtaining the NWP in the forecast area
The forecast field data is inputted the forecasting model by data, obtains the probability of happening of convection weather, and occur generally according to described
Rate carries out the convection weather reporting services of corresponding period, specifically includes:
The forecast field data for obtaining the NWP in the forecast area is normalized the forecast field data, by normalizing
Changing treated, forecast field data inputs the forecasting model, obtains the probability of happening of convection weather, and according to it is described occur it is general
Rate carries out the convection weather reporting services of corresponding period.
4. according to the method described in claim 1, it is characterized in that, the feature of the sample set includes the numerical weather forecast
Temperature, geopotential unit, humidity and the basic meteorological element feature of wind field and longitude, latitude, height above sea level geographic factor of pattern output,
And/or common strong convection index.
5. according to the method described in claim 1, it is characterized in that, the predetermined deep learning model is classified using Softmax
Device, the Softmax graders include the probability of happening of convection weather 0-1.
6. a kind of classification convection weather probability forecast device, which is characterized in that including:
Data decimation module, null field for obtaining the numerical weather forecast NWP in forecast area or analyzes field again, chooses convection current
The observation data of weather;
Sample set builds module, analyzes for null field described in the observation data markers according to the convection weather or again field, builds
The sample set of convection weather;
Model training module is obtained for being trained to predetermined deep learning model according to the sample set of the convection weather
The forecasting model of classification convection weather probability;
Weather forecast module, the forecast field data for obtaining the NWP in the forecast area input the forecast field data
The forecasting model obtains the probability of happening of convection weather, and is carried out according to the probability of happening and numerical weather prediction model
The convection weather reporting services of corresponding period.
7. device according to claim 6, which is characterized in that the sample set structure module is specifically used for according to described right
The observation data of stream weather are marked the null field or are analyzed field again using over-sampling or down-sampling mode, build the sample of convection weather
This collection.
8. device according to claim 6, which is characterized in that the weather forecast module is specifically used for obtaining the forecast
The forecast field data of NWP in region is normalized the forecast field data, by the forecast fields after normalized
Data input the forecasting model, obtain the probability of happening of convection weather, and carry out the corresponding period according to the probability of happening
Convection weather reporting services.
9. a kind of electronic equipment, which is characterized in that including:
At least one processor;And
At least one processor being connect with the processor communication, wherein:
The memory is stored with the program instruction that can be executed by the processor, and the processor calls described program to instruct energy
Enough methods executed as described in claim 1 to 5 is any.
10. a kind of non-transient computer readable storage medium, which is characterized in that the non-transient computer readable storage medium is deposited
Computer program is stored up, the computer program makes the computer execute the method as described in claim 1 to 5 is any.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810230385.9A CN108491877A (en) | 2018-03-20 | 2018-03-20 | A kind of classification convection weather probability forecast method and device |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810230385.9A CN108491877A (en) | 2018-03-20 | 2018-03-20 | A kind of classification convection weather probability forecast method and device |
Publications (1)
Publication Number | Publication Date |
---|---|
CN108491877A true CN108491877A (en) | 2018-09-04 |
Family
ID=63318776
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201810230385.9A Pending CN108491877A (en) | 2018-03-20 | 2018-03-20 | A kind of classification convection weather probability forecast method and device |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN108491877A (en) |
Cited By (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109188567A (en) * | 2018-09-30 | 2019-01-11 | 康美中药材数据信息服务有限公司 | A kind of production of crude drugs weather warning method, electronic equipment and storage medium |
CN110059082A (en) * | 2019-04-17 | 2019-07-26 | 东南大学 | A kind of weather prediction method based on 1D-CNN and Bi-LSTM |
CN111650671A (en) * | 2020-05-13 | 2020-09-11 | 苏翔 | Rainfall forecast space inspection method based on multi-object optimal matching technology |
CN111783821A (en) * | 2020-05-19 | 2020-10-16 | 知天(珠海横琴)气象科技有限公司 | Method and system for distinguishing strong convection gust |
CN111966962A (en) * | 2020-08-27 | 2020-11-20 | 山东省气象台(山东省海洋气象台) | Method and device for judging triggering condition of convection weather |
CN111985727A (en) * | 2020-09-03 | 2020-11-24 | 南京信息工程大学 | Weather prediction method and system based on circulation parting model |
CN112464158A (en) * | 2020-11-05 | 2021-03-09 | 重庆市气象台 | Method and system for automatically recommending numerical weather forecast product |
CN113537336A (en) * | 2021-03-10 | 2021-10-22 | 沈阳工业大学 | XGboost-based short-term thunderstorm and strong wind forecasting method |
CN113962465A (en) * | 2021-10-26 | 2022-01-21 | 国家气象中心(中央气象台) | Precipitation forecasting method, equipment, device and storage medium |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102956023A (en) * | 2012-08-30 | 2013-03-06 | 南京信息工程大学 | Bayes classification-based method for fusing traditional meteorological data with perception data |
CN107728234A (en) * | 2017-09-17 | 2018-02-23 | 北京工业大学 | A kind of intensity of lightning value Forecasting Methodology based on atmospheric electric field data |
-
2018
- 2018-03-20 CN CN201810230385.9A patent/CN108491877A/en active Pending
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102956023A (en) * | 2012-08-30 | 2013-03-06 | 南京信息工程大学 | Bayes classification-based method for fusing traditional meteorological data with perception data |
CN107728234A (en) * | 2017-09-17 | 2018-02-23 | 北京工业大学 | A kind of intensity of lightning value Forecasting Methodology based on atmospheric electric field data |
Non-Patent Citations (6)
Title |
---|
修媛媛等: "《基于机器学习方法的强对流天气识别研究》", 《电子设计工程》 * |
周后福: "《安徽省雷电检测预报及风险评估》", 30 April 2010, 中国科学技术大学出版社 * |
尹姗: "《2017年9-11月T639、ECMWF及日本模式中期预报性能检验》", 《气象》 * |
扬﹒克莱斯尔 著,李法民 等译: "《太阳能资源预报与评估》", 31 January 2018, 中国三峡出版社 * |
楼振飞: "《大数据技术与应用》", 31 March 2016, 上海科学技术出版社 * |
河南省人民政府办公厅主管: "《河南年鉴》", 31 December 2016 * |
Cited By (13)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109188567A (en) * | 2018-09-30 | 2019-01-11 | 康美中药材数据信息服务有限公司 | A kind of production of crude drugs weather warning method, electronic equipment and storage medium |
CN110059082A (en) * | 2019-04-17 | 2019-07-26 | 东南大学 | A kind of weather prediction method based on 1D-CNN and Bi-LSTM |
CN111650671A (en) * | 2020-05-13 | 2020-09-11 | 苏翔 | Rainfall forecast space inspection method based on multi-object optimal matching technology |
CN111650671B (en) * | 2020-05-13 | 2021-10-22 | 苏翔 | Rainfall forecast space inspection method based on multi-object optimal matching technology |
CN111783821A (en) * | 2020-05-19 | 2020-10-16 | 知天(珠海横琴)气象科技有限公司 | Method and system for distinguishing strong convection gust |
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 |
CN111966962A (en) * | 2020-08-27 | 2020-11-20 | 山东省气象台(山东省海洋气象台) | Method and device for judging triggering condition of convection weather |
CN111966962B (en) * | 2020-08-27 | 2023-11-24 | 山东省气象台(山东省海洋气象台) | Method and device for judging triggering conditions of convection weather |
CN111985727B (en) * | 2020-09-03 | 2023-07-28 | 南京信息工程大学 | Method and system for predicting weather based on loop parting model |
CN111985727A (en) * | 2020-09-03 | 2020-11-24 | 南京信息工程大学 | Weather prediction method and system based on circulation parting model |
CN112464158A (en) * | 2020-11-05 | 2021-03-09 | 重庆市气象台 | Method and system for automatically recommending numerical weather forecast product |
CN113537336A (en) * | 2021-03-10 | 2021-10-22 | 沈阳工业大学 | XGboost-based short-term thunderstorm and strong wind forecasting method |
CN113962465A (en) * | 2021-10-26 | 2022-01-21 | 国家气象中心(中央气象台) | Precipitation forecasting method, equipment, device and storage medium |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN108491877A (en) | A kind of classification convection weather probability forecast method and device | |
CN111104981B (en) | Hydrological prediction precision evaluation method and system based on machine learning | |
CN108375808A (en) | Dense fog forecasting procedures of the NRIET based on machine learning | |
CN109614997A (en) | A kind of stealing Risk Forecast Method and device based on deep learning | |
CN108537368A (en) | A kind of lightning method for early warning, apparatus and system | |
Marzban et al. | Ceiling and visibility forecasts via neural networks | |
CN111401419A (en) | Improved RetinaNet-based employee dressing specification detection method | |
CN110378786A (en) | Model training method, promise breaking conduction Risk Identification Method, device and storage medium | |
CN110471131B (en) | High-spatial-resolution automatic prediction method and system for refined atmospheric horizontal visibility | |
Zhang et al. | Weather visibility prediction based on multimodal fusion | |
CN112285807A (en) | Meteorological information prediction method and device | |
CN113901977A (en) | Deep learning-based power consumer electricity stealing identification method and system | |
Biard et al. | Automated detection of weather fronts using a deep learning neural network | |
CN115994325B (en) | Fan icing power generation data enhancement method based on TimeGAN deep learning method | |
CN111178585A (en) | Fault reporting amount prediction method based on multi-algorithm model fusion | |
CN114493052B (en) | Multi-model fusion self-adaptive new energy power prediction method and system | |
CN109003091A (en) | A kind of risk prevention system processing method, device and equipment | |
Larraondo et al. | Automating weather forecasts based on convolutional networks | |
CN110992307A (en) | Insulator positioning and identifying method and device based on YOLO | |
Otache et al. | ARMA modelling of Benue River flow dynamics: comparative study of PAR model | |
CN111784022A (en) | Short-time adjacent fog prediction method based on combination of Wrapper method and SVM method | |
CN109033322A (en) | A kind of test method and device of multidimensional data | |
KR100920529B1 (en) | Method of estimating salt contamination | |
Salman Aal-shamkhi et al. | Extraction and modeling of urban sprawl development in Karbala City using VHR satellite imagery | |
CN114548595A (en) | Strong convection weather physical characteristic quantity prediction method and system based on attention mechanism |
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 | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20180904 |
|
RJ01 | Rejection of invention patent application after publication |