CN-121995540-A - Numerical forecasting optimization method and system based on error analysis

Abstract

The invention discloses a numerical forecasting optimization method and a numerical forecasting optimization system based on error analysis, wherein the method comprises the steps of calculating a hysteresis forecasting error variance to form a sample set; the method comprises the steps of defining a numerical forecasting core parameter based on physical characteristics of an atmospheric chaotic system, establishing a function relation between a perception error variance estimation value and the core parameter, constructing a cost function, carrying out minimum solution to obtain a global optimal estimation value of the core parameter, obtaining a real analysis error variance, a real forecasting error variance and an analysis-forecasting error correlation coefficient model according to the global optimal estimation value of the core parameter so as to describe an evolution rule of an error statistical characteristic along with forecasting time, and carrying out constraint or adjustment on an error statistical structure in a data assimilation system to realize error correction and assimilation optimization of the numerical forecasting system. The invention can be independent of a data assimilation system, and can carry out unbiased statistical analysis of space continuation estimation on analysis and short-term prediction error variance, thereby improving the overall performance of the numerical prediction system.

Inventors

• WANG JINYU
• FENG JIE
• LI MIAOMIAO
• WANG CHAO
• YAO DEGUI
• LIANG YUN
• LI ZHE
• LU MING
• LAN GUANGYU
• WANG XIAOHUI
• MA YUCHENG
• KE JIAYING

Assignees

• 国网河南省电力公司电力科学研究院
• 复旦大学

Dates

Publication Date

20260508

Application Date

20251209

Claims (12)

1. 1. The numerical forecasting optimization method based on error analysis is characterized by comprising the following steps of: acquiring forecast field data of different forecast timelines of a numerical weather forecast system at the same analysis time, and calculating a hysteresis forecast error variance based on the forecast field data to serve as an observation perception error variance to form a sample set; defining a numerical forecasting core parameter based on physical characteristics of an atmospheric chaotic system and establishing a functional relation between a perception error variance estimation value and the core parameter; constructing a cost function based on the sample set and the function relation, and carrying out minimization solution to obtain a global optimal estimated value of the core parameter; Obtaining a real analysis error variance, a real prediction error variance and an analysis-prediction error correlation coefficient model according to the global optimal estimation value of the core parameter so as to describe the evolution rule of the error statistical characteristic along with the prediction time; And utilizing the evolution rule of the error statistical characteristics along with the forecast aging to restrict or adjust the error statistical structure in the data assimilation system, so as to realize the error correction and assimilation optimization of the numerical forecast system.
2. 2. The method for optimizing numerical forecast based on error analysis according to claim 1, wherein: the calculation formula of the hysteresis prediction error variance is as follows: (1) Wherein, the The hysteresis forecast error variance is calculated for a forecast field F i 、F i+l based on the forecast aging of the ith and the (i+l) th, wherein T is the real state of an atmospheric system; And Is a true forecast error.
3. 3. The method for optimizing numerical forecast based on error analysis according to claim 1, wherein: based on physical characteristics of the atmospheric chaotic system, defining numerical forecasting core parameters including true analysis error standard deviation Error growth rate Analysis-prediction error correlation coefficient 。
4. 4. A numerical forecast optimization method based on error analysis according to claim 3, characterized in that: the establishing a functional relationship between the perceptual error variance estimation value and the core parameter comprises: establishing an evolution model of a real prediction error variance, wherein the model is an exponential growth model or a logic growth model, and the exponential growth model is as follows: (4) in the formula, For the first forecast time interval real prediction error variance of step length; the forecasting time length corresponding to the interval step length of the first forecasting time is the forecasting time length; the logical growth model is: (5) In the parameters of ; Is the error saturation level; and (3) establishing an attenuation model for analyzing and forecasting error correlation coefficients: (6) in the formula, A correlation coefficient between the analysis error and the forecast error of the first forecast time interval step length is obtained; A correlation coefficient between the analysis error and the forecast error of the 1 st forecast time interval step length is obtained; Based on the evolution model and the attenuation model, a perception error variance estimation value is established And core parameters 、 And Functional relationship between: (7) Wherein, the For the first forecast time interval step is a true prediction error standard deviation of (c).
5. 5. The method for optimizing numerical forecast based on error analysis according to claim 1, wherein: constructing a cost function based on the sample set and the functional relation, wherein the cost function is specifically: (10) Wherein, the As a cost function; the observed perception error variance is calculated for the forecast field data based on the 0 th and the l th forecast interval step; a perceived error variance estimation value obtained based on the functional relation; the observed perception error variance is calculated for the forecast field data based on the ith and the (i+2) th forecast; Error variance estimation value is forecast for hysteresis; And Is the weight.
6. 6. The method for optimizing numerical forecast based on error analysis according to claim 5, wherein: hysteresis forecast error variance estimate The method comprises the following steps: ,(9) Wherein, the The observed perception error variance calculated for the forecasting fields based on the 2 nd and 4 th forecasting timelines; the forecast duration of the i-2 th forecast time.
7. 7. The method for optimizing numerical forecast based on error analysis according to claim 5, wherein: Weighting of The calculation mode of (a) is as follows: From autocorrelation coefficients of samples delayed by 1 prediction time Calculating a sample autocorrelation adjustment factor f, and combining the sample size N and the sample standard deviation of the observed perception error variance Calculating to obtain the sample mean standard error corresponding to the i < th > and the i < th > forecast timeliness Calculating The ratio to the sum of standard errors of all sample means, obtain the weight 。
8. 8. The method for optimizing numerical forecast based on error analysis according to claim 1, wherein: obtaining a real analysis error variance, a real prediction error variance and an analysis-prediction error correlation coefficient model according to the global optimal estimation value of the core parameter so as to describe the evolution rule of the error statistical characteristic along with the prediction aging, wherein the method specifically comprises the following steps: Substituting the global optimal estimated value of the core parameter into an evolution model of the real prediction error variance and an attenuation model of the analysis-prediction error correlation coefficient to obtain the real analysis error variance, the real prediction error variance and the analysis-prediction error correlation coefficient model, wherein the model is used for describing the evolution rule of the error statistical characteristic along with the prediction aging.
9. 9. The method for optimizing numerical forecast based on error analysis according to claim 1, wherein: And utilizing the evolution rule of the error statistical characteristic along with the forecast aging to restrict or adjust the error statistical structure in the data assimilation system so as to realize the error correction and assimilation optimization of the numerical forecast system, and specifically comprising the following steps: Taking the spatial distribution of the forecast error as a reference, comprehensively evaluating the initial set disturbance generated by the unified forecast moment set Kalman filtering, comparing the consistency of the set disturbance among the spatial structure, the local amplitude and the forecast error, identifying deviation, underestimation or structural defects of the set disturbance in the aspect of error sampling capability, and analyzing the reasons for generating the defects; The method comprises the steps of quantitatively evaluating the description capability of each set disturbance to an error field by calculating the projection coefficient of the set disturbance on the corresponding forecast error field, and judging the problem of the set disturbance in the aspect of error sampling according to a projection result; Based on the evaluation of the original set disturbance sampling performance, the set Kalman filtering initial set disturbance is adjusted based on the analysis error variance; And (3) evaluating the sampling performance of the set disturbance after adjustment, comparing the changes of the set disturbance before and after adjustment in the aspects of spatial structure and amplitude characteristics, and selecting a disturbance adjustment scheme with better sampling performance.
10. 10. A numerical forecast optimization system based on error analysis, running the method of any of claims 1-9, characterized in that the system comprises: the data acquisition and calculation module is used for acquiring forecast field data of different forecast timeliness of the numerical weather forecast system at the same analysis time and calculating a hysteresis forecast error variance based on the forecast field data to serve as an observation perception error variance to form a sample set; The function construction and solving module is used for defining a numerical forecast core parameter based on the physical characteristics of the atmospheric chaotic system, establishing a function relation between a perception error variance estimation value and the core parameter, constructing a cost function based on the sample set and the function relation, and carrying out minimum solving to obtain a global optimal estimation value of the core parameter; The model construction module is used for obtaining a real analysis error variance, a real prediction error variance and an analysis-prediction error correlation coefficient model according to the global optimal estimated value of the core parameter so as to describe the evolution rule of the error statistical characteristic along with the prediction time; And the numerical forecasting optimization module is used for restraining or adjusting the error statistical structure in the data assimilation system by utilizing the evolution rule of the error statistical characteristics along with forecasting aging so as to realize error correction and assimilation optimization of the numerical forecasting system.
11. 11. A terminal comprises a processor and a storage medium, and is characterized in that: The storage medium is used for storing instructions; The processor being operative according to the instructions to perform the steps of the method according to any one of claims 1-9.
12. 12. Computer readable storage medium, on which a computer program is stored, characterized in that the program, when being executed by a processor, implements the steps of the method according to any one of claims 1-9.

Description

Numerical forecasting optimization method and system based on error analysis Technical Field The invention belongs to the technical field of numerical weather forecast, and relates to a numerical forecast optimization method and system based on error analysis. Background The key challenges faced by the existing numerical weather forecast system are the chaos, initial value error and mode error of the atmospheric system, and even if the numerical mode is continuously optimized, the analysis field and short-term forecast still cannot be completely accurate. In order to better evaluate the performance of the assimilation and forecast system and optimize the data assimilation and set forecast strategy, it is particularly critical to accurately evaluate the error variance between the numerical pattern analysis field and the forecast field and the real atmospheric conditions. The current error variance estimation method mainly comprises two types, namely a method based on the integrated Kalman filtering and the like of the internal output of a data assimilation system, wherein the method can provide error estimation, but has the defects of high calculation cost, sensitivity to initial disturbance assumption and the like, the estimation result depends on the prior assumption of the assimilation system and lacks independence, and the method based on the historical forecast data is a national weather center (NMC) method for estimating background error covariance by calculating the difference between hysteresis forecast fields, but the estimation accuracy seriously depends on the selection of hysteresis time and is difficult to accurately reflect the error variance of actual analysis and short-term forecast. The above methods have obvious disadvantages in estimating spatially distributed fine error structures. In particular, in the aspects of analysis time and short-term (0-48 hours) forecast error estimation, the conventional method has systematic deviation, can not meet the precision requirements of the current high-resolution set forecast and variation data assimilation system, and influences the overall performance of the numerical forecast system. Disclosure of Invention In order to solve the defects in the prior art, the invention provides a numerical forecasting optimization method and a numerical forecasting optimization system based on error analysis, which can be independent of a data assimilation system, carry out unbiased analysis of space continuation estimation on analysis and short-term forecasting error variance, and improve the overall performance of the numerical forecasting system by objectively and quantitatively estimating the analysis error and the forecasting error variance. The invention adopts the following technical scheme. The first aspect of the present invention provides a numerical forecast optimization method based on error analysis, including: acquiring forecast field data of different forecast timelines of a numerical weather forecast system at the same analysis time, and calculating a hysteresis forecast error variance based on the forecast field data to serve as an observation perception error variance to form a sample set; defining a numerical forecasting core parameter based on physical characteristics of an atmospheric chaotic system and establishing a functional relation between a perception error variance estimation value and the core parameter; constructing a cost function based on the sample set and the function relation, and carrying out minimization solution to obtain a global optimal estimated value of the core parameter; Obtaining a real analysis error variance, a real prediction error variance and an analysis-prediction error correlation coefficient model according to the global optimal estimation value of the core parameter so as to describe the evolution rule of the error statistical characteristic along with the prediction time; And utilizing the evolution rule of the error statistical characteristics along with the forecast aging to restrict or adjust the error statistical structure in the data assimilation system, so as to realize the error correction and assimilation optimization of the numerical forecast system. Preferably, the calculation formula of the hysteresis prediction error variance is: (1) Wherein, the The hysteresis forecast error variance is calculated for a forecast field F i、F i+l based on the forecast aging of the ith and the (i+l) th, wherein T is the real state of an atmospheric system; And Is a true forecast error. Preferably, the numerical prediction core parameters are defined based on the physical characteristics of the atmospheric chaotic system, including the standard deviation of true analysis errorsError growth rateAnalysis-prediction error correlation coefficient。 Preferably, said establishing a functional relationship between the perceptual error variance estimate and the core parameter comprises: establishing an evolution model of a rea