阅读说明：本技术 一种基于活动区间的气温预报偏差订正方法和系统 (Temperature forecast deviation correction method and system based on activity interval ) 是由 张楠 杨晓君 于 2020-07-28 设计创作，主要内容包括：本发明公开了一种基于活动区间的气温预报偏差订正方法和系统,通过对数值模式气温精细化预报产品进行准对称统计期偏差滑动订正,并利用历史数据回算,根据预报性能最优原则确定试预报区间,不断更新气温预报偏差的准对称统计期,实现了对数值模式气温预报的有效订正。该订正方法在不断更新数值模式日最高或最低气温系统偏差的基础上,还对系统偏差的准对称统计期进行滑动更新,对模式的不断升级更新有较好的适应性,订正效果较好。(The invention discloses an activity interval-based air temperature forecast deviation correcting method and system, which are used for correcting deviation in a quasi-symmetric statistical period of a numerical-mode air temperature refined forecast product in a sliding manner, determining a trial forecast interval according to a forecast performance optimal principle by utilizing historical data back calculation, and continuously updating the quasi-symmetric statistical period of air temperature forecast deviation so as to realize effective correction of numerical-mode air temperature forecast. The correction method also performs sliding update on the quasi-symmetric statistical period of the system deviation on the basis of continuously updating the system deviation of the highest or lowest temperature of the numerical mode day, has better adaptability to the continuously updated update of the mode and has better correction effect.)

1. A temperature forecast deviation correction method based on an activity interval is characterized by comprising the following steps:

data arrangement: acquiring air temperature forecast products and live products of each forecast time period in a numerical mode, namely daily highest air temperature forecast products and daily lowest air temperature forecast products and live products, and interpolating the air temperature forecast products and the live products onto a grid of L x L by an interpolation technology, wherein L is an arbitrary value between 1 and 10 km;

determining the daily highest and lowest temperature system deviation: adopting a quasi-symmetric sliding statistical period mode, namely sliding every N days before the forecast date and after the forecast date of the previous year, carrying out day-by-day calculation on the relative error of the highest and lowest temperatures of the numerical mode days of the two periods of time, and acquiring the average relative error of the highest and lowest temperatures of the numerical mode days in different quasi-symmetric sliding statistical periods as the system deviation of the mode according to different N values, wherein the average relative error is shown in a formula (1);

wherein BIAS is the statistical system deviation, mo, of the highest or lowest air temperature of the forecast time in the mode_iForecasting the highest or lowest temperature, sk, for the mode reported on day i within the statistical period_iIs mo_iLive observation of the highest or lowest air temperature corresponding to time;

deviation correction: estimating the system deviation of the numerical prediction mode through sliding statistics, and using the system deviation to correct the air temperature prediction of the numerical mode, wherein the deviation correction specific formula is shown as a formula (2):

T_dz＝T_mo+BIAS (2)

wherein T is_dzCorrecting the forecast for the air temperature, T_moThe value mode forecast result is obtained, and BIAS is the system deviation of the value mode air temperature forecast obtained by quasi-symmetric sliding statistics;

and a service application step: and (3) according to the numerical mode air temperature forecast product and the live product, utilizing a timing operation program, and according to the time of arrival of the numerical weather forecast product, automatically operating the program every day at a fixed time, and outputting the corrected air temperature refined and gridded forecast product in real time to realize service operation.

2. The activity-interval-based air temperature forecast deviation correction method according to claim 1, further comprising setting a trial forecast interval and a forecast interval:

the forecast interval is the mode temperature forecast of the latest forecast starting time available at the forecast starting time, and the trial forecast interval is the mode temperature forecast which is started 1 to M days before the mode forecast starting time corresponding to the forecast interval.

3. The activity-interval-based air temperature forecast deviation correction method according to claim 2, further comprising an optimum N-value determination step of:

setting a test forecast interval M before the current forecast, taking different quasi-symmetric sliding statistical periods by utilizing the step of determining the daily highest and lowest temperature system deviations, determining the mode system deviations by utilizing a formula (1), correcting the mode daily highest and lowest temperature forecasts of the test forecast interval by utilizing a formula (2), storing the system deviations determined by taking different N values and the correction results of the mode daily highest and lowest temperature deviations, and calculating the forecast accuracy or average absolute error of the daily highest and lowest temperature correction results obtained by taking different N values in the test forecast interval;

selecting a system deviation correction mode with the best correction effect according to the principle of highest accuracy rate or minimum average absolute error, namely determining the optimal system deviation quasi-symmetric sliding statistical period of the highest daily temperature and the lowest daily temperature, namely determining the optimal N value, so that the N value is determined by rolling day by day;

and after the optimal N value is obtained, counting the system deviation of the highest and lowest temperature in the prediction interval mode by using a formula (1), and correcting the prediction result of the highest and lowest temperature in the prediction interval mode by using a formula (2).

4. A method for activity-interval-based air temperature forecast deviation correction according to claims 1-3, further comprising the step of determining an optimal trial forecast interval:

taking different trial forecasting intervals, namely taking different values for M, respectively determining the optimal N values according to the method in the optimal N value determining step, counting the system deviation of forecasting of the highest and lowest air temperature of the grid point day by using a formula (1), and correcting the deviation of the highest and lowest air temperature of the grid point numerical mode day of the forecasting interval by using a formula (2);

through historical back calculation, the system records and takes different test prediction intervals, namely when M takes different values, the deviation correction result of the highest and lowest temperature of the grid point numerical mode day of the prediction interval is taken, the prediction accuracy and the average absolute error of the correction result of the highest and lowest temperature of the grid point numerical mode day of the prediction interval are counted month by month, and the test prediction interval with the best correction effect is selected according to the principle of highest accuracy or minimum average absolute error.

5. The activity-interval-based air temperature forecast deviation correction method according to claim 3 or 4, characterized in that said forecast accuracy is calculated using the following equation (3):

where G represents the prediction accuracy, N_zqRepresenting the correct grid/number of forecasts, N_zAnd (4) predicting the total grid point/time of prediction, namely, the absolute error between the predicted value of the grid point/time and the actual situation value is less than 2 degrees.

6. An activity-interval-based air temperature forecast deviation correction method according to any of claims 3 or 4, characterized in that said mean absolute error statistic is calculated using the following formula (4):

where ME represents the mean absolute error of the air temperature correction result in the trial prediction interval, dz_jPattern correction result, sk, representing the highest or lowest daily temperature reported on the jth day of the test sample_jRepresents and dz_jThe observation live value of the highest or lowest temperature on the day corresponding to the time.

7. Method for correcting an activity-interval-based air temperature forecast deviation according to claims 1-6, characterized in that said business application step comprises in particular the following steps:

(1) selecting different quasi-symmetric sliding statistical periods according to the best statistical trial forecasting interval M, performing trial forecasting on the highest and lowest air temperatures of the trial forecasting interval, and selecting a system deviation correcting mode with the best correcting effect according to the principle of highest accuracy rate or minimum average absolute error, namely determining the optimal system deviation quasi-symmetric sliding statistical period of the highest or lowest daily air temperature, namely determining the optimal N value;

(2) carrying out statistics on the system deviation of the forecast of the highest or lowest air temperature in the forecast interval mode by using a formula (1);

(3) correcting the forecast of the highest or lowest temperature in the forecast interval mode by using a formula (2);

(4) and (4) linearly correcting to the timing air temperature forecast according to the maximum and minimum air temperature forecast deviation correction value of the numerical mode day on the grid points, thereby correcting the timing air temperature forecast.

8. A system for correcting a temperature forecast deviation based on an activity interval, comprising:

a data sorting module: acquiring air temperature forecast products and live products of each forecast time period in a numerical mode, namely daily highest air temperature forecast products and daily lowest air temperature forecast products and live products, and interpolating the air temperature forecast products and the live products onto a grid of L x L by an interpolation technology, wherein L is an arbitrary value between 1 and 10 km;

the determination module of the daily highest and lowest temperature system deviation: adopting a quasi-symmetric sliding statistical period mode, namely sliding every N days before the forecast date and after the forecast date of the previous year, carrying out day-by-day calculation on the relative error of the highest and lowest temperatures of the numerical mode days of the two periods of time, and acquiring the average relative error of the highest and lowest temperatures of the numerical mode days in different quasi-symmetric sliding statistical periods as the system deviation of the mode according to different N values, wherein the average relative error is shown in a formula (1);

wherein BIAS is the statistical system deviation, mo, of the highest or lowest air temperature of the forecast time in the mode_iForecasting the highest or lowest temperature, sk, for the mode reported on day i within the statistical period_iIs mo_iLive observation of the highest or lowest air temperature corresponding to time;

deviation correction module: estimating the system deviation of the numerical prediction mode through sliding statistics, and using the system deviation to correct the air temperature prediction of the numerical mode, wherein the deviation correction specific formula is shown as a formula (2):

T_dz＝T_mo+BIAS (2)

wherein T is_dzCorrecting the forecast for the air temperature, T_moThe value mode forecast result is obtained, and BIAS is the system deviation of the value mode air temperature forecast obtained by quasi-symmetric sliding statistics;

a service application module: and (3) according to the numerical mode air temperature forecast product and the live product, utilizing a timing operation program, and according to the time of arrival of the numerical weather forecast product, automatically operating the program every day at a fixed time, and outputting the corrected air temperature refined and gridded forecast product in real time to realize service operation.

9. The activity-interval-based air temperature forecast deviation correction system of claim 8, further comprising an optimal N-value determination module:

setting a test forecast interval M before the current forecast, taking different quasi-symmetric sliding statistical periods by using a determining module of the daily highest and lowest temperature system deviations, determining mode system deviations by using a formula (1), correcting deviations of the mode daily highest and lowest temperature forecasts of the test forecast interval by using a formula (2), storing the system deviations determined by taking different N values and the deviation correction results of the mode daily highest and lowest temperature, and calculating the forecast accuracy and average absolute error of the daily highest and lowest temperature correction results obtained by taking different N values in the test forecast interval;

and selecting a system deviation correction mode with the best correction effect according to the principle of highest accuracy or minimum average absolute error, namely determining the optimal system deviation quasi-symmetric sliding statistical period of the highest daily temperature and the lowest daily temperature, namely determining the optimal N value, so that the N value is determined by rolling day by day. And after the optimal N value is obtained, counting the system deviation of the highest and lowest temperature in the prediction interval mode by using a formula (1), and correcting the prediction result of the highest and lowest temperature in the prediction interval mode by using a formula (2).

10. The activity-interval-based air temperature forecast deviation correction system according to claim 8 or 9, further comprising an optimal trial forecast interval determination module:

taking different trial forecasting intervals, namely taking different values of M, respectively determining the optimal N values according to the method in the optimal N value determination module, counting the system deviation of forecasting of the highest and lowest air temperature of the grid point day by using a formula (1), and correcting the deviation of the highest and lowest air temperature of the grid point numerical mode day of the forecasting interval by using a formula (2);

through historical back calculation, the system records and takes different test prediction intervals, namely when M takes different values, the deviation correction result of the highest and lowest temperature of the grid point numerical mode day of the prediction interval is taken, the prediction accuracy and the average absolute error of the correction result of the highest and lowest temperature of the grid point numerical mode day of the prediction interval are counted month by month, and the test prediction interval with the best correction effect is selected according to the principle of highest accuracy or minimum average absolute error.

Technical Field

The invention relates to the technical field of weather forecast, in particular to a temperature forecast deviation correction method and system based on an activity interval.

Background

Temperature is a basic element in weather forecast, model products are more and more abundant along with the development of a numerical weather forecast technology, objective mode release methods are greatly developed, and the accuracy of temperature forecast is continuously improved, for example, a numerical forecast product of a central numerical forecast center in middle of Europe can reach more than 80% of the forecast accuracy (the absolute error is less than or equal to 2 degrees and is regarded as correct) of the highest temperature in 0-24 hours in the future, but the temperature forecast of a numerical model has systematic deviation inevitably because errors exist in the initial field of the model and various physical processes. In order to eliminate the system deviation, various statistical methods exist at present, for example, an air temperature objective prediction is carried out by adopting an air temperature prediction deviation sliding correction method (patent number: CN201810723060), the method adopts the sliding correction method, the problem of unstable deviation caused by rapid development and change of a numerical value mode is solved, the prediction accuracy is obviously improved compared with the air temperature prediction directly output by the numerical value mode, but the method still has the following two problems: (1) the average deviation of the previous N days (particularly when N is larger) cannot represent the system deviation of the forecast current day mode, such as the mode system deviation of 5 months and 31 days, and the average relative error of 5 months and 1 day to 5 months and 30 days is not more reasonable than the average relative error of 5 months and 16 days to 6 months and 15 days; (2) in the context of rapid development of numerical patterns, not only the systematic variation of the patterns will change with the updating of the patterns, but also the optimal statistical period of the systematic variation.

In view of the above problems, the present inventors have earnestly needed to devise a new technique to improve the problems thereof.

Disclosure of Invention

The invention aims to provide a temperature forecast deviation correction method and system based on an activity interval, which can more scientifically estimate the temperature forecast system deviation of a numerical mode and further effectively correct the mode temperature forecast.

In order to solve the technical problems, the technical scheme of the invention is as follows:

a temperature forecast deviation correction method based on an activity interval comprises the following steps:

data arrangement: acquiring air temperature forecast products and live products of each forecast time period in a numerical mode, namely daily highest air temperature forecast products and daily lowest air temperature forecast products and live products, and interpolating the air temperature forecast products and the live products onto a grid of L x L by an interpolation technology, wherein L is an arbitrary value between 1 and 10 km;

determining the daily highest and lowest temperature system deviation: adopting a quasi-symmetric sliding statistical period mode, namely sliding every N days before the forecast date and after the forecast date of the previous year, carrying out day-by-day calculation on the relative error of the highest and lowest temperatures of the numerical mode days of the two periods of time, and acquiring the average relative error of the highest and lowest temperatures of the numerical mode days in different quasi-symmetric sliding statistical periods as the system deviation of the mode according to different N values, wherein the average relative error is shown in a formula (1);

wherein BIAS is the statistical system deviation, mo, of the highest or lowest air temperature of the forecast time in the mode_iForecasting the highest or lowest temperature, sk, for the mode reported on day i within the statistical period_iIs mo_iObserving the highest and lowest air temperature in a live scene corresponding to time;

deviation correction: estimating the system deviation of the numerical prediction mode through sliding statistics, and using the system deviation to correct the air temperature prediction of the numerical mode, wherein the deviation correction specific formula is shown as a formula (2):

T_dz＝T_mo+BIAS (2)

wherein T is_dzCorrecting the forecast for the air temperature, T_moThe value mode forecast result is obtained, and BIAS is the system deviation of the value mode air temperature forecast obtained by quasi-symmetric sliding statistics;

determining the deviation of the timing air temperature system: correcting linearly to the timing air temperature forecast according to the maximum and minimum air temperature forecast deviation of the numerical mode day on the grid points, thereby correcting the timing air temperature forecast;

and a service application step: and (3) according to the numerical mode air temperature forecast product and the live product, utilizing a timing operation program, and according to the time of arrival of the numerical weather forecast product, automatically operating the program every day at a fixed time, and outputting the corrected air temperature refined and gridded forecast product in real time to realize service operation.

Preferably, the method further comprises the following steps of setting a trial forecasting interval and a forecasting interval:

the forecast interval is the mode temperature forecast of the latest forecast starting time available at the forecast starting time, and the trial forecast interval is the mode temperature forecast which is started 1 to M days before the mode forecast starting time corresponding to the forecast interval.

Preferably, the method further comprises the optimal N value determination step of:

setting a test forecast interval M before the current forecast, taking different quasi-symmetric sliding statistical periods by utilizing the step of determining the daily highest and lowest temperature system deviations, determining the mode system deviations by utilizing a formula (1), correcting the mode daily highest and lowest temperature forecasts of the test forecast interval by utilizing a formula (2), storing the system deviations determined by taking different N values and the correction results of the mode daily highest and lowest temperature deviations, and calculating the forecast accuracy or average absolute error of the daily highest and lowest temperature correction results obtained by taking different N values in the test forecast interval;

selecting a system deviation correction mode with the best correction effect according to the principle of highest accuracy rate or minimum average absolute error, namely determining the optimal system deviation quasi-symmetric sliding statistical period of the highest daily temperature and the lowest daily temperature, namely determining the optimal N value, so that the N value is determined by rolling day by day;

and after the optimal N value is obtained, counting the system deviation of the highest and lowest temperature in the prediction interval mode by using a formula (1), and correcting the prediction result of the highest and lowest temperature in the prediction interval mode by using a formula (2).

Preferably, the method further comprises the step of determining the optimal trial forecasting interval:

taking different trial forecasting intervals, namely taking different values for M, respectively determining the optimal N values according to the method in the optimal N value determining step, counting the system deviation of forecasting of the highest and lowest air temperature of the grid point day by using a formula (1), and correcting the deviation of the highest and lowest air temperature of the grid point numerical mode day of the forecasting interval by using a formula (2);

through historical back calculation, the system records and takes different test prediction intervals, namely when M takes different values, the deviation correction result of the highest and lowest temperature of the grid point numerical mode day of the prediction interval is taken, the prediction accuracy and the average absolute error of the correction result of the highest and lowest temperature of the grid point numerical mode day of the prediction interval are counted month by month, and the test prediction interval with the best correction effect is selected according to the principle of highest accuracy or minimum average absolute error.

Preferably, the forecast accuracy is calculated using the following equation (3):

where G represents the prediction accuracy, N_zqRepresenting the correct grid/number of forecasts, N_zThe forecasted total grid points/times.

Preferably, the average absolute error statistic is calculated using the following equation (4):

where ME represents the mean absolute error of the air temperature correction result in the trial prediction interval, dz_jPattern correction result, sk, representing the highest or lowest daily temperature reported on the jth day of the test sample_jRepresents and dz_jThe observation live value of the highest or lowest temperature on the day corresponding to the time.

Preferably, the service application step specifically includes the following steps:

(1) selecting different quasi-symmetric sliding statistical periods according to the best statistical trial forecasting interval M, performing trial forecasting on the highest and lowest air temperatures of the trial forecasting interval, and selecting a system deviation correcting mode with the best correcting effect according to the principle of highest accuracy rate or minimum average absolute error, namely determining the optimal system deviation quasi-symmetric sliding statistical period of the highest or lowest daily air temperature, namely determining the optimal N value;

(2) carrying out statistics on the system deviation of the forecast of the highest or lowest air temperature in the forecast interval mode by using a formula (1);

(3) correcting the forecast of the highest or lowest temperature in the forecast interval mode by using a formula (2);

(4) and (4) linearly correcting to the timing air temperature forecast according to the maximum and minimum air temperature forecast deviation correction value of the numerical mode day on the grid points, thereby correcting the timing air temperature forecast.

A system for correcting an activity interval-based air temperature forecast deviation, comprising:

a data sorting module: acquiring air temperature forecast products and live products of each forecast time period in a numerical mode, namely daily highest air temperature forecast products and daily lowest air temperature forecast products and live products, and interpolating the air temperature forecast products and the live products onto a grid of L x L by an interpolation technology, wherein L is an arbitrary value between 1 and 10 km;

the determination module of the daily highest and lowest temperature system deviation: adopting a quasi-symmetric sliding statistical period mode, namely sliding every N days before the forecast date and after the forecast date of the previous year, carrying out day-by-day calculation on the relative error of the highest and lowest temperatures of the numerical mode days of the two periods of time, and acquiring the average relative error of the highest and lowest temperatures of the numerical mode days in different quasi-symmetric sliding statistical periods as the system deviation of the mode according to different N values, wherein the average relative error is shown in a formula (1);

wherein BIAS is the statistical system deviation, mo, of the highest or lowest air temperature of the forecast time in the mode_iForecasting the highest or lowest temperature, sk, for the mode reported on day i within the statistical period_iIs mo_iObserving the highest and lowest air temperature in a live scene corresponding to time;

deviation correction module: estimating the system deviation of the numerical prediction mode through sliding statistics, and using the system deviation to correct the air temperature prediction of the numerical mode, wherein the deviation correction specific formula is shown as a formula (2):

T_dz＝T_mo+BIAS (2)

wherein T is_dzCorrecting the forecast for the air temperature, T_moThe value mode forecast result is obtained, and BIAS is the system deviation of the value mode air temperature forecast obtained by quasi-symmetric sliding statistics;

a module for determining a timing air temperature system deviation: correcting linearly to the timing air temperature forecast according to the maximum and minimum air temperature forecast deviation of the numerical mode day on the grid points, thereby correcting the timing air temperature forecast;

a service application module: and (3) according to the numerical mode air temperature forecast product and the live product, utilizing a timing operation program, and according to the time of arrival of the numerical weather forecast product, automatically operating the program every day at a fixed time, and outputting the corrected air temperature refined and gridded forecast product in real time to realize service operation.

Preferably, the method further comprises the best N value determination module:

setting a test forecast interval M before the current forecast, taking different quasi-symmetric sliding statistical periods by using a determining module of the daily highest and lowest temperature system deviations, determining mode system deviations by using a formula (1), correcting deviations of the mode daily highest and lowest temperature forecasts of the test forecast interval by using a formula (2), storing the system deviations determined by taking different N values and the deviation correction results of the mode daily highest and lowest temperature, and calculating the forecast accuracy and average absolute error of the daily highest and lowest temperature correction results obtained by taking different N values in the test forecast interval;

and selecting a system deviation correction mode with the best correction effect according to the principle of highest accuracy or minimum average absolute error, namely determining the optimal system deviation quasi-symmetric sliding statistical period of the highest daily temperature and the lowest daily temperature, namely determining the optimal N value, so that the N value is determined by rolling day by day. And after the optimal N value is obtained, counting the system deviation of the highest and lowest temperature in the prediction interval mode by using a formula (1), and correcting the prediction result of the highest and lowest temperature in the prediction interval mode by using a formula (2).

Preferably, the method further comprises an optimal trial forecasting interval determining module:

taking different trial forecasting intervals, namely taking different values of M, respectively determining the optimal N values according to the method in the optimal N value determination module, counting the system deviation of forecasting of the highest and lowest air temperature of the grid point day by using a formula (1), and correcting the deviation of the highest and lowest air temperature of the grid point numerical mode day of the forecasting interval by using a formula (2);

through historical back calculation, the system records and takes different test prediction intervals, namely when M takes different values, the deviation correction result of the highest and lowest temperature of the grid point numerical mode day of the prediction interval is taken, the prediction accuracy and the average absolute error of the correction result of the highest and lowest temperature of the grid point numerical mode day of the prediction interval are counted month by month, and the test prediction interval with the best correction effect is selected according to the principle of highest accuracy or minimum average absolute error.

By adopting the technical scheme, the invention at least comprises the following beneficial effects:

according to the temperature forecast deviation correcting method and system based on the activity interval, the deviation of the quasi-symmetric statistical period is corrected in a sliding mode through a numerical-mode temperature refined forecast product, the forecast interval is determined according to the forecast performance optimization principle by utilizing the back calculation of historical data, the quasi-symmetric statistical period of the temperature forecast deviation is continuously updated, and the effective correction of the numerical-mode temperature forecast is achieved. The correction method also performs sliding update on the quasi-symmetric statistical period of the system deviation on the basis of continuously updating the system deviation of the highest or lowest temperature of the numerical mode day, has better adaptability to the continuously updated update of the mode and has better correction effect.

Drawings

FIG. 1 is a flow chart of a method for correcting an activity interval-based air temperature forecast deviation according to the present invention;

fig. 2 is a schematic structural diagram of an activity-interval-based air temperature forecast deviation correction system according to the present invention.

Detailed Description

The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.