CN109783774A - A kind of temperature DATA PROCESSING IN ENSEMBLE PREDICTION SYSTEM method and system - Google Patents
A kind of temperature DATA PROCESSING IN ENSEMBLE PREDICTION SYSTEM method and system Download PDFInfo
- Publication number
- CN109783774A CN109783774A CN201811549198.3A CN201811549198A CN109783774A CN 109783774 A CN109783774 A CN 109783774A CN 201811549198 A CN201811549198 A CN 201811549198A CN 109783774 A CN109783774 A CN 109783774A
- Authority
- CN
- China
- Prior art keywords
- forecast
- time
- moment
- formula
- fields
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- 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
Abstract
Temperature DATA PROCESSING IN ENSEMBLE PREDICTION SYSTEM method and system provided by the invention, by temperature DATA PROCESSING IN ENSEMBLE PREDICTION SYSTEM from multiple numerical model forecast fields, according to past 15 days forecast fields and observation field is corrected to each forecast fields progress dynamic error and the value of forecasting is examined, and it makes adjustment to temperature jump situation, each pattern weight is calculated using gradient descent method, and then it is weighted and averaged and obtains optimal surface air temperature forecast fields, temperature DATA PROCESSING IN ENSEMBLE PREDICTION SYSTEM method and system provided by the invention, it improves numerical model and forecasts systematic error caused by error itself, mode error especially in the case of temperature jump (heating suddenly or cooling), the better value of forecasting can be obtained by acquiring each schema weight using gradient descent method, reliable reference is provided for weather forecast.
Description
Technical field
The present invention relates to weather forecast technical field, in particular to a kind of temperature ensemble prediction system and method.
Background technique
In the world, Krishnamurti etc. (1999) proposes the thought of multi-model ensemble earliest, will be multiple and different
Model predictions result carries out the result that statistic op- timization finally obtains DATA PROCESSING IN ENSEMBLE PREDICTION SYSTEM.A large number of experiments show that, DATA PROCESSING IN ENSEMBLE PREDICTION SYSTEM is effective
The error of seasonal climate prediction and weather forecast is reduced, the value of forecasting is much better than single mode and multi-mode ensemble average.?
In terms of precipitation forecast, Krishnamurti etc. is based on United Kingdom Meteorological Office (United Kingdom in TIGGE data
Meteorological Office, UKMO), the pre- measured center of U.S. environment (National Centers for
Environmental Prediction, NCEP), European Center for Medium Weather Forecasting (European Centre for
Medium-Range Weather Forecasts, ECMWF), Australian weather bureau (Bureau of Meteorology,
Australia, BOM), five centers of China Meteorological Administration (China Meteorological Administration, CMA) it is complete
Precipitation, bai-u rainy period precipitation and Landed Typhoon precipitation carry out superset when ball mode is to Monsoon region in China South China Sea monsoon onset
Prediction research, and Time effect forecast is expanded into 10d from 1-3d and is discussed, 4-10d forecast superset root mean square is missed
It is poor minimum.Multi-mode integrative prediction technology is pre- in the strong weather of the conventional prediction of various weather constituents and precipitation, typhoon etc. such as air themperature
The result of study for applying for the allocation of domain shows that it is efficiently modified seasonal climate prediction strategy, improves medium-short term prediction accuracy rate and simplicity
The advantages that practical, has been widely studied in the world and has applied.
Compared to international further investigation and application, multi-model ensemble technology is started late in China, on ground
Temperature etc. has carried out preliminary trial.Zhao Shengrong (2006) is based on China National Meteorological Center T213 mode, German weather bureau
Business model and the forecast of Japan Meteorological Agency business model 2m high-temperature establish more than 600, China station using BP neural network method
Temperature integrative prediction system, the results showed that the temperature forecast of set is substantially better than 3 individual forecast results of mode, forecast knot
The precision of fruit obtains certain raising.Intelligence association, which flies to wait, utilizes TIGGE data set European Center for Medium Weather Forecasting (ECMWF), day
The centralizations forecast result such as this meteorology Room (JMA), Environmental forecasting centre (NCEP) and United Kingdom Meteorological Office (UKMO),
DATA PROCESSING IN ENSEMBLE PREDICTION SYSTEM experiment is carried out to surface temperature and precipitation, the results showed that, although for different forecast element multi-mode integrated approaches
Applicability have differences, but on the whole the value of forecasting of multi-mode integrated approach be better than single center forecast.
Existing ensemble forecast technique there has been apparent development, but DATA PROCESSING IN ENSEMBLE PREDICTION SYSTEM model buildings method is also relatively simple, more
Using ensemble average or linear weighted function averaging method, numerical model forecasts that systematic error caused by error itself corrects aspect
Have much room for improvement, especially the mode error in the case of temperature jump (suddenly heating or cooling) is larger, Shenzhen or even Guangdong,Hongkong and Macao
, often there are the extreme weathers such as high temperature in area area, and often deviation is larger in this case for numerical model forecast model products, needs
Such a stable temperature ensemble prediction system is constructed, provides reliable reference for weather forecast.
Summary of the invention
Have in view of that, it is necessary in view of the defects existing in the prior art, a kind of stable temperature ensemble prediction system is provided,
A kind of temperature DATA PROCESSING IN ENSEMBLE PREDICTION SYSTEM method and system reliably referred to are provided for weather forecast.
To achieve the above object, the present invention adopts the following technical solutions:
On the one hand, the present invention provides a kind of temperature DATA PROCESSING IN ENSEMBLE PREDICTION SYSTEM method, include the following steps:
Obtain each model predictions field Current Temperatures forecast fields；
Judge each model predictions field Current Temperatures predicted value and the previous day synchronization predicted value difference whether be greater than or
Less than some threshold value；If it is not, performing the next step；If jumping in next step, and execute lower next step；
Result error is carried out to current temperature pattern forecast fields to correct；
Each model predictions field weight is calculated using gradient descent method；
Optimal surface air temperature forecast fields are obtained using weighted average according to each model predictions field weight；
Export the optimal surface air temperature forecast fields.
In some preferred embodiments, judging each model predictions field Current Temperatures predicted value and the previous day with for the moment
It carves in the step of whether predicted value difference is more than or less than some threshold value, the threshold value can be according to different regions and season statistics meter
It obtains.
In some preferred embodiments, it is carried out in the step of result error is corrected to Current Temperatures forecast fields, specifically
Include the following steps:
The first formula is defined, first formula is bk,r,f=Fk,r,f-Or+f, wherein Fk,r,fIt is k-th of set member in r
Moment shifts to an earlier date the forecast of f time, Or+fIt is the observation of moment r+f；
The second formula is defined, second formula isQ1, Q2And Q3It is 15 days in the past same
Moment forecast departure bk,r,f1st, 2,3 25% percentile of sequence rises when calling time as r for calculating in the past in 15 days
Carve the forecast Mean Deviation value for shifting to an earlier date the f time；
It defines third formula and result error revision is carried out to Current Temperatures forecast fields, the third formula isWherein,K-th of set member rcurrMoment shifts to an earlier date the forecast of f time,It is above-mentioned
K-th of set member that step calculates goes over 15 days r moment (with rcurrFor synchronization) to shift to an earlier date the forecast of f time average inclined
Difference,It is then k-th of set member rcurrMoment shifts to an earlier date the deviation forecast amendment field of f time
In some preferred embodiments, the step of calculating each model predictions field weight using gradient descent method
In, specifically include following methods:
Root-mean-square error cost function is defined, the root-mean-square error cost function is
Wherein, m is sliding training period data set number, F (xr,f) it is the forecast for training period shifting to an earlier date the f time at the r moment, Or+fIt is training period
R+f moment actual observed value,Wherein,To uniform weight, it is weighted and averaged to obtain predicted value F
(xr,f)；
The root-mean-square error cost function optimal solution is solved using gradient descent method, so that error is minimum；
According to the optimal solution, each model predictions field optimal weights are obtained.
In some preferred embodiments, optimal ground is being obtained using weighted average according to each model predictions field weight
In the step of face temperature forecast field, optimal surface air temperature forecast fields specifically are obtained using following formula:
F (x in formular,f) it is optimal surface air temperature forecast fields,It is asked for previous step
The mode k weight optimal solution of solution,Mode k currently gives the correct time time r in advancecurrThe deviation forecast amendment field of f time in advance.
On the other hand, the present invention also provides a kind of temperature ensemble prediction systems, comprising:
Current Temperatures forecast fields obtain module, for obtaining each model predictions field Current Temperatures forecast fields；
Predicted value difference judgment module, for judging each model predictions field Current Temperatures predicted value and the previous day with for the moment
Carve whether predicted value difference is more than or less than some threshold value；
Forecast result deviation corrects module, corrects for carrying out result error to current temperature pattern forecast fields；
Gradient descent method computing module, for calculating each model predictions field weight using gradient descent method；
Weighted average calculation module, for obtaining optimal ground using weighted average according to each model predictions field weight
Temperature forecast field；
Surface air temperature predicted value output module, for exporting the optimal surface air temperature forecast fields.
In some preferred embodiments, in the predicted value difference judgment module, the threshold value can be according to differently
Area and season are calculated.
In some preferred embodiments, the forecast result deviation corrects module and includes:
First construction unit, for defining the first formula, first formula is bk,r,f=Fk,r,f-Or+f, wherein Fk,r,f
It is the forecast that k-th of set member shifts to an earlier date the f time at the r moment, Or+fIt is the observation of moment r+f；
Second construction unit, is used for the second formula, and second formula isQ1, Q2And Q3It was
Remove 15 days synchronization forecast departure bk,r,f1st, 2,3 25% percentile of sequence rises for calculating in the past in 15 days
It calls time and shifts to an earlier date the forecast Mean Deviation value of f time for the r moment；
Third construction unit carries out result error revision to Current Temperatures forecast fields for defining third formula, and described the
Three formula areWherein,K-th of set member rcurrMoment shifts to an earlier date the forecast of f time,K-th of the set member calculated for above-mentioned steps goes over 15 days r moment (with rcurrFor synchronization) shift to an earlier date the f time
Forecast Mean Deviation value,It is then k-th of set member rcurrMoment shifts to an earlier date the deviation forecast amendment field of f time
In some preferred embodiments, the gradient descent method computing module includes:
First computing unit, for defining root-mean-square error cost function, the root-mean-square error cost function isWherein, m is sliding training period data set number, F (xr,f) it is to be mentioned at the r moment training period
The forecast of preceding f time, Or+fIt is r+f training period, actual observed value moment,Wherein,It is equal
One changes weight, is weighted and averaged to obtain predicted value F (xr,f)；
Gradient descent method solves unit, optimal for using gradient descent method to solve the root-mean-square error cost function
Solution, so that error is minimum；
Second computing unit, for obtaining each mode Optimal predictor field weight according to the optimal solution
In some preferred embodiments, the weighted average calculation module obtains optimal ground by using following formula
Face temperature forecast field:
F (x in formular,f) it is optimal surface air temperature forecast fields,It is asked for previous step
The mode k weight optimal solution of solution,Mode k currently gives the correct time time r in advancecurrThe deviation forecast amendment field of f time in advance.
The present invention by adopting the above technical scheme the advantages of be:
Temperature DATA PROCESSING IN ENSEMBLE PREDICTION SYSTEM method and system provided by the invention are forecast by temperature DATA PROCESSING IN ENSEMBLE PREDICTION SYSTEM from multiple numerical models
Field is set out, and according to past 15 days forecast fields and observation field is corrected to each forecast fields progress dynamic error and the value of forecasting is examined,
And make adjustment to temperature jump situation, each pattern weight is calculated using gradient descent method, so be weighted and averaged obtain it is optimal
Surface air temperature predicted value, temperature DATA PROCESSING IN ENSEMBLE PREDICTION SYSTEM method and system provided by the invention, improve numerical model forecast itself accidentally
Mode error in the case of systematic error caused by difference, especially temperature jump (heating suddenly or cooling), is declined using gradient
Method, which acquires each schema weight, can obtain the better value of forecasting, provide reliable reference for weather forecast.
Detailed description of the invention
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 technical 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 attached drawings.
Fig. 1 is the step flow chart for the temperature ensemble prediction system that the embodiment of the present invention 1 provides.
Fig. 2 is the structural schematic diagram for the temperature ensemble prediction system that the embodiment of the present invention 2 provides.
Specific embodiment
Following will be combined with the drawings in the embodiments of the present invention, and technical solution in the embodiment of the present invention carries out clear, complete
Site preparation description, it is clear that described embodiments are only a part of the embodiments of the present invention, instead of all the embodiments.It is based on
Embodiment in the present invention, it is obtained by those of ordinary skill in the art without making creative efforts all other
Embodiment shall fall within the protection scope of the present invention.
Embodiment 1
Please referring to Fig. 1 is to include the following steps: the present invention provides a kind of step flow chart of temperature DATA PROCESSING IN ENSEMBLE PREDICTION SYSTEM method
Step S110: each model predictions field Current Temperatures forecast fields are obtained.
It is appreciated that the temperature DATA PROCESSING IN ENSEMBLE PREDICTION SYSTEM that the present invention obtains is issued from multiple numerical model forecast fields, multiple moulds are collected
Formula forecast fields enable to the result of prediction more accurate reliable.
Step S120: judge that each model predictions field Current Temperatures predicted value is with the previous day synchronization predicted value difference
It is no to be more than or less than some threshold value；If it is not, executing step S130；If jumping in next step, and execute step S140.The threshold value
It can be calculated according to different regions and season.
It is appreciated that being directed to each model predictions field, current predicted value is greater than with the previous day synchronization predicted value difference
Or be less than some threshold value (temperature increases suddenly, or reduces) and abandon error revising, i.e.,
Wherein,K-th of set member rcurrMoment shifts to an earlier date the forecast of f time,It is assembled for k-th
Member rcurrMoment shifts to an earlier date the deviation forecast amendment field of f time, if temperature has occurred without error revising in equal expression at this time
Breakthrough if being adjusted according to history deviation, bigger systematic error is easily caused, if not occurring more typically in sudden temperature drop
Mutation carries out model predictions result error and corrects.
Step S130: result error is carried out to Current Temperatures forecast fields and is corrected.
In some preferred embodiments, specifically include the following steps:
Step S131: defining the first formula, and first formula is bk,r,f=Fk,r,f-Or+f, wherein Fk,r,fIt is k-th of collection
Synthesis person shifts to an earlier date the forecast of f time, O at the r momentr+fIt is the observation of moment r+f；
Step S132: defining the second formula, and second formula isQ1, Q2And Q3It is the past ten
Five days synchronization forecast departure bk,r,f1st, 2,3 25% percentile of sequence rises and gives the correct time for calculating in the past in 15 days
Between for the r moment shift to an earlier date the forecast Mean Deviation value of f time.
It is appreciated that due to that can not know current value model results and actual deviation, by dynamically passing by 15 days
The average forecast departure of (this time adjustable) is estimated.
Step S133: it defines third formula and result error revision is carried out to Current Temperatures forecast fields, the third formula isWherein,K-th of set member rcurrMoment shifts to an earlier date the forecast of f time,It is above-mentioned
K-th of set member that step calculates goes over 15 days r moment (with rcurrFor synchronization) to shift to an earlier date the forecast of f time average inclined
Difference,It is then k-th of set member rcurrMoment shifts to an earlier date the deviation forecast amendment field of f time
It is appreciated that for each numerical model, each forecast length revises past 15 days average deviationIt is applied in current integrative prediction result, to generate the forecast fields that a deviation is corrected.
S131~step S133 may be implemented to correct Current Temperatures forecast fields progress result error through the above steps.
Step S140: each model predictions field weight is calculated using gradient descent method.
In some preferred embodiments, following methods are specifically included:
Step S141: defining root-mean-square error cost function, and the root-mean-square error cost function isWherein, m is sliding training period data set number, F (xr,f) it is to be mentioned at the r moment training period
The forecast of preceding f time, Or+fIt is r+f training period, actual observed value moment,Wherein,It is equal
One changes weight, is weighted and averaged to obtain predicted value F (xr,f)；
Step S142: the root-mean-square error cost function optimal solution is solved using gradient descent method, so that error is minimum；
It is appreciated that solving the root-mean-square error cost function optimal solution i.e. to each variable (power using gradient descent method
Weight) it differentiates, specifically:
Wherein, ECMWF (European Center for Medium Weather Forecasting), JMA (Japan Meteorological Agency), GRAPES (Global Model set
Forecast) etc. be DATA PROCESSING IN ENSEMBLE PREDICTION SYSTEM member (can be adjusted according to demand), it will be understood that suitable learning rate is set, gradient decline
Weight optimal solution is solved, so that error is minimum.
Step S143: according to the optimal solution, each model predictions field optimal weights are obtained
It is appreciated that since each mode has different forecast features, so this system will assign different mode different power
Weight so that prediction ability it is strong pattern weight it is relatively larger, use each schema weight of gradient descent algorithm
Step S150: optimal surface air temperature forecast fields are obtained using weighted average according to each model predictions field weight.
Specifically optimal surface air temperature forecast fields are obtained using following formula:
F (x in formular,f) it is optimal surface air temperature forecast fields,For previous step
The mode k weight optimal solution of solution,Mode k currently gives the correct time time r in advancecurrThe deviation forecast amendment field of f time in advance.
Step S160: the output optimal surface air temperature forecast fields.
Temperature DATA PROCESSING IN ENSEMBLE PREDICTION SYSTEM method provided by the invention is gone out by temperature DATA PROCESSING IN ENSEMBLE PREDICTION SYSTEM from multiple numerical model forecast fields
Hair, according to past 15 days forecast fields and observation field is corrected to each forecast fields progress dynamic error and the value of forecasting is examined, and right
Temperature jump situation is made adjustment, and calculates each pattern weight using gradient descent method, and then be weighted and averaged and obtain optimal ground
Face temperature forecast value, temperature DATA PROCESSING IN ENSEMBLE PREDICTION SYSTEM method and system provided by the invention improve numerical model and forecast that error itself is drawn
Mode error in the case of the systematic error risen, especially temperature jump (heating suddenly or cooling), uses gradient descent method
The better value of forecasting can be obtained by acquiring each schema weight, provide reliable reference for weather forecast.
Embodiment 2
Referring to Fig. 2, for the present invention provides a kind of structural schematic diagrams of temperature ensemble prediction system, comprising:
Current Temperatures forecast fields obtain module 110: obtaining each model predictions field Current Temperatures forecast fields.
It is appreciated that the temperature DATA PROCESSING IN ENSEMBLE PREDICTION SYSTEM that the present invention obtains is issued from multiple numerical model forecast fields, multiple moulds are collected
Formula forecast fields enable to the result of prediction more accurate reliable.
Predicted value difference judgment module 120: for judging that each model predictions field Current Temperatures predicted value and the previous day are same
Whether one moment predicted value difference is more than or less than some threshold value.The threshold value can be calculated according to different regions and season
Out.
It is appreciated that being directed to each model predictions field, current predicted value is greater than with the previous day synchronization predicted value difference
Or be less than some threshold value (temperature increases suddenly, or reduces) and abandon error revising, i.e.,
Wherein,K-th of set member rcurrMoment shifts to an earlier date the forecast of f time,For k-th of set member
rcurrMoment shifts to an earlier date the deviation forecast amendment field of f time, if temperature has occurred without error revising in equal expression at this time
It breaks through, more typically in sudden temperature drop, if being adjusted according to history deviation, bigger systematic error is easily caused, if not dashing forward
Become progress model predictions result error to correct.
Forecast result deviation corrects module 130: carrying out result error to Current Temperatures forecast fields and corrects.
In some preferred embodiments, it specifically includes:
First construction unit: for defining the first formula, first formula is bk,r,f=Fk,r,f-Or+f, wherein Fk,r,f
It is the forecast that k-th of set member shifts to an earlier date the f time at the r moment, Or+fIt is the observation of moment r+f；
Second construction unit: for defining the second formula, second formula isQ2With
Q3It is 15 days synchronization forecast departure b of pastk,r,f1st, 2,3 25% percentile of sequence, for calculating the past 15
In it, rises and call time and shift to an earlier date the forecast Mean Deviation value of f time for the r moment.
It is appreciated that due to that can not know current value model results and actual deviation, by dynamically passing by 15 days
The average forecast departure of (this time adjustable) is estimated.
Third construction unit: carrying out result error revision to Current Temperatures forecast fields for defining third formula, and described the
Three formula areWherein,K-th of set member rcurrMoment shifts to an earlier date the forecast of f time,K-th of the set member calculated for above-mentioned steps goes over 15 days r moment (with rcurrFor synchronization) shift to an earlier date the f time
Forecast Mean Deviation value,It is then k-th of set member rcurrMoment shifts to an earlier date the deviation forecast amendment field of f time.
It is appreciated that for each numerical model, each forecast length revises past 15 days average deviationIt is applied in current integrative prediction result, to generate the forecast fields that a deviation is corrected.
It may be implemented to correct Current Temperatures forecast fields progress result error by above-mentioned module.
Gradient descent method computing module 140: each model predictions field weight is calculated using gradient descent method.
In some preferred embodiments, it specifically includes:
First computing unit: defining root-mean-square error cost function, and the root-mean-square error cost function isWherein, m is sliding training period data set number, F (xr,f) it is to be mentioned at the r moment training period
The forecast of preceding f time, Or+fIt is r+f training period, actual observed value moment,Wherein,It is equal
One changes weight, is weighted and averaged to obtain predicted value F (xr,f)；
Gradient descent method solves unit: solving the root-mean-square error cost function optimal solution using gradient descent method, makes
It is minimum to obtain error；
It is appreciated that solving the root-mean-square error cost function optimal solution i.e. to each variable (power using gradient descent method
Weight) it differentiates, specifically:
Wherein, ECMWF (European Center for Medium Weather Forecasting), JMA (Japan Meteorological Agency), GRAPES (Global Model set
Forecast) etc. be DATA PROCESSING IN ENSEMBLE PREDICTION SYSTEM member (can be adjusted according to demand), it will be understood that suitable learning rate is set, gradient decline
Weight optimal solution is solved, so that error is minimum.
Second computing unit: according to the optimal solution, each model predictions field weight optimal solution is obtained
It is appreciated that since each mode has different forecast features, so this system will assign different mode different power
Weight so that prediction ability it is strong pattern weight it is relatively larger, use each schema weight of gradient descent algorithm
Weighted average calculation module 150: optimal for being obtained according to each model predictions field weight using weighted average
Surface air temperature predicted value.
Specifically, the weighted average calculation module obtains optimal surface air temperature forecast fields by using following formula:
F (x in formular,f) it is optimal surface air temperature forecast fields,It is asked for previous step
The mode k weight optimal solution of solution,Mode k currently gives the correct time time r in advancecurrThe deviation forecast amendment field of f time in advance.
Surface air temperature predicted value output module 160: the output optimal surface air temperature forecast fields.
Temperature ensemble prediction system provided by the invention is gone out by temperature DATA PROCESSING IN ENSEMBLE PREDICTION SYSTEM from multiple numerical model forecast fields
Hair, according to past 15 days forecast fields and observation field is corrected to each forecast fields progress dynamic error and the value of forecasting is examined, and right
Temperature jump situation is made adjustment, and calculates each pattern weight using gradient descent method, and then be weighted and averaged and obtain optimal ground
Face temperature forecast field, temperature DATA PROCESSING IN ENSEMBLE PREDICTION SYSTEM method and system provided by the invention improve numerical model and forecast that error itself is drawn
Mode error in the case of the systematic error risen, especially temperature jump (heating suddenly or cooling), uses gradient descent method
The better value of forecasting can be obtained by acquiring each schema weight, provide reliable reference for weather forecast.
Certainly temperature DATA PROCESSING IN ENSEMBLE PREDICTION SYSTEM method and system of the invention can also have a variety of transformation and remodeling, it is not limited on
State the specific structure of embodiment.In short, protection scope of the present invention should include that those carry out those of ordinary skill in the art
Say obvious transformation or substitution and remodeling.
Claims (10)
1. a kind of temperature DATA PROCESSING IN ENSEMBLE PREDICTION SYSTEM method, which is characterized in that include the following steps:
Obtain each model predictions field Current Temperatures forecast fields；
Judge whether each model predictions field Current Temperatures predicted value is more than or less than with the previous day synchronization predicted value difference
Some threshold value；If it is not, performing the next step；If jumping in next step, and execute lower next step；
Result error is carried out to current temperature pattern forecast fields to correct；
Each model predictions field weight is calculated using gradient descent method；
Optimal surface air temperature forecast fields are obtained using weighted average according to each model predictions field weight；
Export the optimal surface air temperature forecast fields.
2. temperature DATA PROCESSING IN ENSEMBLE PREDICTION SYSTEM method as described in claim 1, which is characterized in that judging that each model predictions field is currently warm
In the step of whether degree predicted value and the previous day synchronization predicted value difference are more than or less than some threshold value, the threshold value can root
It is calculated according to different regions and season.
3. temperature DATA PROCESSING IN ENSEMBLE PREDICTION SYSTEM method as described in claim 1, which is characterized in that carrying out result to Current Temperatures forecast fields
In the step of deviation is corrected, specifically include the following steps:
The first formula is defined, first formula is bk,r,f=Fk,r,f-Or+f, wherein Fk,r,fIt is k-th of set member at the r moment
The forecast of f time in advance, Or+fIt is the observation of moment r+f；
The second formula is defined, second formula isQ1, Q2And Q3It is 15 days synchronizations of past
Forecast departure bk,r,f1st, 2,3 25% percentile of sequence acts to call time mentioning for the r moment for calculating in the past in 15 days
The forecast Mean Deviation value of preceding f time；
It defines third formula and result error revision is carried out to Current Temperatures forecast fields, the third formula isWherein,K-th of set member rcurrMoment shifts to an earlier date the forecast of f time,It is above-mentioned
K-th of set member that step calculates goes over 15 days r moment (with rcurrFor synchronization) to shift to an earlier date the forecast of f time average inclined
Difference,It is then k-th of set member rcurrMoment shifts to an earlier date the deviation forecast amendment field of f time.
4. temperature DATA PROCESSING IN ENSEMBLE PREDICTION SYSTEM method as described in claim 1, which is characterized in that each being calculated using gradient descent method
In the step of model predictions field weight, following methods are specifically included:
Root-mean-square error cost function is defined, the root-mean-square error cost function isIts
In, m is sliding training period data set number, F (xr,f) it is the forecast for training period shifting to an earlier date the f time at the r moment, Or+fIt is training period r+
F moment actual observed value,Wherein,To uniform weight, it is weighted and averaged to obtain predicted value F
(xr,f)；
The root-mean-square error cost function optimal solution is solved using gradient descent method, so that error is minimum；
According to the optimal solution, each model predictions field optimal weights are obtained.
5. temperature DATA PROCESSING IN ENSEMBLE PREDICTION SYSTEM method as described in claim 1, which is characterized in that adopted according to each model predictions field weight
In the step of obtaining optimal surface air temperature forecast fields with weighted average, optimal surface air temperature specifically is obtained using following formula
Forecast fields:
F (x in formulaR, f) it is optimal surface air temperature forecast fields,It is solved for previous step
Mode k weight optimal solution,Mode k currently gives the correct time time r in advancecurrThe deviation forecast amendment field of f time in advance.
6. a kind of temperature ensemble prediction system characterized by comprising
Current Temperatures forecast fields obtain module, for obtaining each model predictions field Current Temperatures forecast fields；
Predicted value difference judgment module, for judging that each model predictions field Current Temperatures predicted value and the previous day synchronization are pre-
Whether report value difference value is more than or less than some threshold value；
Forecast result deviation corrects module, corrects for carrying out result error to Current Temperatures forecast fields；
Gradient descent method computing module, for calculating each model predictions field weight using gradient descent method；
Weighted average calculation module, for obtaining optimal surface air temperature using weighted average according to each model predictions field weight
Forecast fields；
Surface air temperature predicted value output module, for exporting the optimal surface air temperature forecast fields.
7. temperature ensemble prediction system as claimed in claim 6, which is characterized in that in the predicted value difference judgment module
In, the threshold value can be calculated according to different regions and season.
8. temperature ensemble prediction system according to claim 6, which is characterized in that the forecast result deviation corrects module
Include:
First construction unit, for defining the first formula, first formula is bk,r,f=Fk,r,f-Or+f, wherein Fk,r,fIt is kth
A set member shifts to an earlier date the forecast of f time, O at the r momentr+fIt is the observation of moment r+f；
Second construction unit, is used for the second formula, and second formula isQ1, Q2And Q3It is the past ten
Five days synchronization forecast departure bk,r,f1st, 2,3 25% percentile of sequence rises and gives the correct time for calculating in the past in 15 days
Between for the r moment shift to an earlier date the forecast Mean Deviation value of f time；
Third construction unit carries out result error revision to Current Temperatures forecast fields for defining third formula, and the third is public
Formula isWherein,K-th of set member rcurrMoment shifts to an earlier date the forecast of f time,
K-th of the set member calculated for above-mentioned steps goes over 15 days r moment (with rcurrFor synchronization) shift to an earlier date the forecast of f time
Mean Deviation value,It is then k-th of set member rcurrMoment shifts to an earlier date the deviation forecast amendment field of f time.
9. temperature ensemble prediction system according to claim 6, which is characterized in that the gradient descent method computing module
Include:
First computing unit, for defining root-mean-square error cost function, the root-mean-square error cost function isWherein, m is sliding training period data set number, F (xr,f) it is to be mentioned at the r moment training period
The forecast of preceding f time, Or+fIt is r+f training period, actual observed value moment,Wherein,It is equal
One changes weight, is weighted and averaged to obtain predicted value F (xr,f)；
Gradient descent method solves unit to be made for using gradient descent method to solve the root-mean-square error cost function optimal solution
It is minimum to obtain error；
Second computing unit, for obtaining each mode Optimal predictor field weight according to the optimal solution
10. temperature ensemble prediction system according to claim 6, which is characterized in that the weighted average calculation module passes through
Optimal surface air temperature forecast fields are obtained using following formula:
F (x in formular,f) it is optimal surface air temperature forecast fields,It is solved for previous step
Mode k weight optimal solution,Mode k currently gives the correct time time r in advancecurrThe deviation forecast amendment field of f time in advance.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811549198.3A CN109783774B (en) | 2018-12-18 | 2018-12-18 | Temperature set forecasting method and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811549198.3A CN109783774B (en) | 2018-12-18 | 2018-12-18 | Temperature set forecasting method and system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109783774A true CN109783774A (en) | 2019-05-21 |
| CN109783774B CN109783774B (en) | 2023-05-23 |
Family
ID=66497247
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811549198.3A Active CN109783774B (en) | 2018-12-18 | 2018-12-18 | Temperature set forecasting method and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109783774B (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115827710A (en) * | 2022-12-20 | 2023-03-21 | 中国民用航空局空中交通管理局航空气象中心 | Method and system for screening meteorological element multi-mode forecast results |
| CN111538935B (en) * | 2019-12-26 | 2023-08-25 | 北京玖天气象科技有限公司 | Fine precipitation fusion method, system, electronic equipment and storage medium based on terrain features and multi-source mode products |
| CN117849908A (en) * | 2024-03-08 | 2024-04-09 | 江苏省气候中心 | Plum-entering and plum-exiting date prediction method and device in plum rainy season based on mode circular flow field |
| CN117849908B (en) * | 2024-03-08 | 2024-05-10 | 江苏省气候中心 | Plum-entering and plum-exiting date prediction method and device in plum rainy season based on mode circular flow field |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20140336934A1 (en) * | 2013-05-07 | 2014-11-13 | Atomic Energy Council - Institute Of Nuclear Energy Research | Ensemble wind power forecasting platform system and operational method thereof |
| CN105447593A (en) * | 2015-11-16 | 2016-03-30 | 南京信息工程大学 | Rapid updating mixing assimilation method based on time lag set |
| CN108399469A (en) * | 2018-01-31 | 2018-08-14 | 中国人民解放军空军研究院战场环境研究所 | A kind of weather phenomenon forecasting procedure based on deep learning and numerical weather forecast |
| CN108983324A (en) * | 2018-08-09 | 2018-12-11 | 深圳市雅码科技有限公司 | A kind of temperature forecast method and system based on Kalman filtering |
-
2018
- 2018-12-18 CN CN201811549198.3A patent/CN109783774B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20140336934A1 (en) * | 2013-05-07 | 2014-11-13 | Atomic Energy Council - Institute Of Nuclear Energy Research | Ensemble wind power forecasting platform system and operational method thereof |
| CN105447593A (en) * | 2015-11-16 | 2016-03-30 | 南京信息工程大学 | Rapid updating mixing assimilation method based on time lag set |
| CN108399469A (en) * | 2018-01-31 | 2018-08-14 | 中国人民解放军空军研究院战场环境研究所 | A kind of weather phenomenon forecasting procedure based on deep learning and numerical weather forecast |
| CN108983324A (en) * | 2018-08-09 | 2018-12-11 | 深圳市雅码科技有限公司 | A kind of temperature forecast method and system based on Kalman filtering |
Non-Patent Citations (1)
| Title |
|---|
| 林春泽等: "基于TIGGE资料的地面气温多模式超级集合预报", 《应用气象学报》 * |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111538935B (en) * | 2019-12-26 | 2023-08-25 | 北京玖天气象科技有限公司 | Fine precipitation fusion method, system, electronic equipment and storage medium based on terrain features and multi-source mode products |
| CN115827710A (en) * | 2022-12-20 | 2023-03-21 | 中国民用航空局空中交通管理局航空气象中心 | Method and system for screening meteorological element multi-mode forecast results |
| CN115827710B (en) * | 2022-12-20 | 2023-06-27 | 中国民用航空局空中交通管理局航空气象中心 | Screening method and system for meteorological element multi-mode forecasting result |
| CN117849908A (en) * | 2024-03-08 | 2024-04-09 | 江苏省气候中心 | Plum-entering and plum-exiting date prediction method and device in plum rainy season based on mode circular flow field |
| CN117849908B (en) * | 2024-03-08 | 2024-05-10 | 江苏省气候中心 | Plum-entering and plum-exiting date prediction method and device in plum rainy season based on mode circular flow field |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109783774B (en) | 2023-05-23 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Wang et al. | Climate, agricultural production and hydrological balance in the North China Plain | |
| Hart et al. | An evaluation of mesoscale-model-based model output statistics (MOS) during the 2002 Olympic and Paralympic Winter Games | |
| CN107341577A (en) | A kind of crop yield Forecasting Methodology and system | |
| CN109583653A (en) | The extended peroid forecasting procedure of NORTHWESTERN PACIFIC TYPHOON based on statistical model | |
| Kong et al. | Assessment of temperature extremes in China using RegCM4 and WRF | |
| CN108983324A (en) | A kind of temperature forecast method and system based on Kalman filtering | |
| CN105095670B (en) | Drawing method of electric transmission line icing-thickness growth rate distribution diagram based on continuous freezing rain | |
| CN109783774A (en) | A kind of temperature DATA PROCESSING IN ENSEMBLE PREDICTION SYSTEM method and system | |
| CN113016450A (en) | Greenhouse crop irrigation method and system | |
| CN107037504A (en) | For Severe Convective Weather Forecasting assimilation dodge the method for translation proxy radar return | |
| CN105389739A (en) | Method for determining protected grape low-temperature disaster weather index insurance pure premium rate | |
| Sokol et al. | Nowcasting of precipitation–advective statistical forecast model (SAM) for the Czech Republic | |
| CN116595394A (en) | Training method of wind speed correction model, wind speed prediction method, wind speed prediction equipment and medium | |
| Yin et al. | On statistical nowcasting of road surface temperature | |
| Shirvani et al. | Time series modelling of increased soil temperature anomalies during long period | |
| CN110059423B (en) | Tropical cyclone objective strength determining method based on multi-factor generalized linear model | |
| Bughici et al. | Evaluation and bias correction in WRF model forecasting of precipitation and potential evapotranspiration | |
| CN106355264A (en) | Combined prediction method of reference crop evapotranspiration | |
| KR101944420B1 (en) | Sea fishing index forecasting apparatus and the method thereof | |
| KR101870743B1 (en) | Tidal flat experience index forecasting apparatus and the method thereof | |
| CN109858678A (en) | A kind of method and system of determining sunflower Meteorological Output | |
| CN109615150A (en) | A kind of method and system of determining rice Meteorological Output | |
| CN109063684A (en) | A kind of near surface temperature remote-sensing monitoring method | |
| Yu et al. | An operational application of NWP models in a wind power forecasting demonstration experiment | |
| CN109615149A (en) | A kind of method and system of determining beet Meteorological Output |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |