A short-term precipitation prediction method and device based on dynamic weight determination

Abstract

The application provides a short-term precipitation prediction method and device based on dynamic weight setting, and solves the technical problem that each prediction source data cannot be comprehensively utilized according to the prediction performance change to realize accuracy and stability. The method comprises the following steps: forming a dynamic evaluation index of prediction accuracy of each prediction organization in the same area according to the frequency of accurate precipitation prediction in the same area in a past period at a current prediction time; arbitrating the prediction data of each prediction organization in a future period at the current prediction time by using the dynamic evaluation index to form a prediction of a precipitation event, dynamically quantifying the rainfall weight of the prediction data according to the dynamic evaluation index, and forming a prediction of rainfall according to the comprehensive prediction data of the weight. A measurement benchmark of the prediction performance change and prediction accuracy of the prediction source in the spatial scale under the same real-time time scale is established. A consensus basis for arbitrating the event trend and a dynamic weight setting of the data source for the event quantitative analysis are formed. The comprehensive prediction result is optimized.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of weather forecasting, in particular to a short-term and nowcasting precipitation forecasting method and device based on dynamic weighting. BACKGROUND

[0002] Short-term and nowcasting precipitation forecasting generally refers to the prediction of the occurrence probability and intensity of rainfall events within the next few hours, which has important practical significance for flood control, traffic management, agricultural planning, etc. In urban areas, short-term and nowcasting precipitation forecasting can help relevant departments to make preparations in advance to reduce the risk of urban waterlogging; in the agricultural field, timely precipitation forecasting can guide farmers to reasonably arrange irrigation and harvesting, and reduce losses. Therefore, improving the accuracy of short-term and nowcasting precipitation forecasting is of great significance to reduce the impact of natural disasters and protect people's life and property safety. Due to the rapid change and uncertainty of weather systems, short-term and nowcasting precipitation forecasting has always been a challenge in weather forecasting. With the increasing demand for meteorological services, it is particularly important to improve the accuracy and timeliness of short-term and nowcasting precipitation forecasting.

[0003] With the development of meteorological prediction technology, many forecasting agencies can currently provide rainfall forecasts at different time scales and spatial scales. However, the accuracy of a single forecasting agency is often affected by various factors, making it difficult to ensure accurate forecasts in all situations. Therefore, in order to overcome the limitations of a single weather forecasting model and improve the accuracy of rainfall forecasts, it is necessary to integrate the forecasting information from multiple forecasting agencies. There are differences in forecasting information between different forecasting models or services, and how to integrate these differences to provide stable and reliable forecasts is a technical challenge to optimize the accuracy and adaptability of the forecasting results. SUMMARY

[0004] In view of the above problems, the embodiments of the present application provide a short-term and nowcasting precipitation forecasting method and device based on dynamic weighting, which solves the technical problem that it is not possible to comprehensively utilize the forecasting data of each forecasting source according to changes in forecasting performance to achieve accuracy and stability.

[0005] The short-term and nowcasting precipitation forecasting method based on dynamic weighting of the embodiments of the present application comprises:

[0006] forming a dynamic evaluation index of the forecasting accuracy of each forecasting agency in the same area according to the frequency of accurate precipitation forecasting in the same area within the previous period at the current forecasting time;

[0007] arbitrating the forecasting data of each forecasting agency for the future period at the current forecasting time using the dynamic evaluation index to form the forecast of the precipitation event, dynamically quantifying the rainfall weight of the forecasting data according to the dynamic evaluation index, and forming the forecast of the rainfall according to the comprehensive forecasting data of the rainfall weight.

[0008] In an embodiment of the present application, the formation of the dynamic evaluation index comprises:

[0009] The prediction information and observation information of each prediction agency are normalized and converted into prediction data;

[0010] The TS scores of each prediction agency in the same region for each day of precipitation are formed according to the prediction data of the same region in the past period;

[0011] The value repetition frequency of the TS scores of each prediction agency in the same region in the same past period is counted;

[0012] The dynamic evaluation index of each prediction agency in the same region in the same past period is formed according to the TS score, the value repetition frequency and the value threshold.

[0013] In an embodiment of the present application, the past period is 180 days or a year.

[0014] In an embodiment of the present application, the past period also includes the same season or the same month of different years.

[0015] In an embodiment of the present application, the prediction of the precipitation event comprises:

[0016] The positive and negative parties supporting the occurrence of the precipitation event are determined according to the prediction data of each prediction agency for a specific future period;

[0017] The prediction result of the precipitation event is formed by taking the cumulative sum of the dynamic evaluation indexes of the positive and negative parties as the arbitration basis.

[0018] In an embodiment of the present application, the prediction of the rainfall amount comprises:

[0019] The rainfall amount weight of each prediction agency is formed according to the dynamic evaluation index;

[0020] The prediction result of the rainfall amount is formed by synthesizing the prediction data of the positive square prediction agency according to the rainfall amount weight.

[0021] In an embodiment of the present application, it also comprises:

[0022] The prediction error quantification and evaluation optimization are formed according to the difference between the prediction and the observation.

[0023] The short-term precipitation prediction device based on dynamic weight in the embodiment of the present application comprises:

[0024] The memory is used for storing the program code of the processing process of the short-term precipitation prediction method based on dynamic weight;

[0025] The processor is used for executing the program code.

[0026] The short-term and nowcasting precipitation prediction device based on dynamic weighting of the embodiment of the application comprises:

[0027] The dynamic evaluation setting module is used for forming a dynamic evaluation index of the prediction accuracy of each prediction institution in the same region according to the frequency of the accurate prediction of precipitation in the same region in the past period at the current prediction time;

[0028] The dynamic weighting prediction module is used for arbitrating the prediction data of the future period of each prediction institution at the current prediction time by using the dynamic evaluation index to form the prediction of the precipitation event, dynamically quantifying the rainfall weight of the prediction data according to the dynamic evaluation index, and forming the prediction of the rainfall according to the comprehensive prediction data of the rainfall weight.

[0029] In an embodiment of the application, the short-term and nowcasting precipitation prediction device based on dynamic weighting further comprises:

[0030] The error feedback optimization module is used for forming the prediction error quantization and evaluation optimization according to the difference between the prediction and the observation.

[0031] The short-term and nowcasting precipitation prediction method and device based on dynamic weighting of the embodiment of the application establish the measurement benchmark of the prediction performance change and the prediction accuracy in the spatial scale of the prediction source in the same real-time time scale by forming the dynamic evaluation index of the prediction accuracy of each prediction source in the past period. The timeliness and dynamic nature of the evaluation index are used to form the consensus basis for arbitrating the event trend and the dynamic weighting of the data source of the event quantitative analysis. The optimal weight configuration is obtained by real-time calculation of the weight of each prediction source, so that the comprehensive prediction result is optimized. The accuracy and adaptability of the prediction can be improved by using the complementarity of different prediction tools while adapting to different weather conditions and the performance change of the prediction source. The decision source optimization combination is realized, the importance and reliability of different prediction institutions are balanced, and the accuracy and stability of the prediction are improved. BRIEF DESCRIPTION OF DRAWINGS

[0032] Figure 1 The flowchart of the short-term and nowcasting precipitation prediction method based on dynamic weighting of an embodiment of the application is shown.

[0033] Figure 2 The architecture diagram of the short-term and nowcasting precipitation prediction device based on dynamic weighting of an embodiment of the application is shown. DETAILED DESCRIPTION

[0034] To make the purpose, technical scheme and advantages of the application more clear and explicit, the application is further described below in combination with the drawings and specific embodiments. Obviously, the described embodiments are only part of the embodiments of the application, not all the embodiments. Based on the embodiments in the application, all other embodiments obtained by those skilled in the art without creative labor belong to the protection scope of the application.

[0035] An embodiment of the present application is based on a dynamic weighting short-term precipitation prediction method as shown in Figure 1 In Figure 1 , the embodiment includes:

[0036] Step 100: Form a dynamic evaluation index of the prediction accuracy of each prediction institution in the same region according to the frequency of accurate prediction of precipitation in the same region within the past period at the current prediction time.

[0037] Those skilled in the art can understand that the current prediction time refers to the time point at which the current prediction result is made by using existing observation data, telemetry data, remote sensing data and experience data for a future determined period or a future determined time. The time point and the basis data used can be time-sequenced by the corresponding time stamp formed by the national time center clock. Different prediction institutions usually form independent regional division and meteorological observation station setting within the same geographical terrain region. The prediction institutions use independent prediction tools (such as information processing algorithms and data analysis models, etc.) to analyze the basis data to form independent prediction data to show the prediction result of short-term precipitation. The past period refers to a period of time before the current prediction time, which continues to the current prediction time, and the prediction data formed in the past period can be time-sequenced and processed.

[0038] Usually, the prediction institution makes short-term precipitation prediction at a unit time interval. The prediction result of the precipitation event is quantified by the field observation result, which is quantified into the following four states:

[0039]

[0040]

[0041] By statistically analyzing the accurate prediction results of the precipitation prediction of each prediction institution in the same region within the past period, an evaluation index of the prediction accuracy of each prediction institution in the same region can be formed. The evaluation index changes dynamically with the passage of the current prediction time.

[0042] Step 200: Arbitrate the prediction data of each prediction institution for the future period at the current prediction time by using the dynamic evaluation index to form the prediction of the precipitation event, dynamically quantify the rainfall weight of the prediction data according to the dynamic evaluation index, and form the prediction of the rainfall according to the comprehensive prediction data of the rainfall weight.

[0043] The future period refers to a certain time period or time point after the current prediction time. The prediction data at least includes the precipitation event confirmed by the prediction agency and the estimated rainfall in the precipitation event. The arbitration behavior can be embodied as a rule of majority rule, and the rule can take the dynamic evaluation index as a reference. The dynamic evaluation index is formed by iterative prediction data of regional and time series, and the weight dynamic correlation of the estimated rainfall of each prediction agency in the determined region is established through the dynamic evaluation index. The real-time of the weight correlation is formed by the time series dynamic of the dynamic evaluation index. The comprehensive estimation result of the rainfall is obtained through the weight correlation.

[0044] The short-term precipitation prediction method and device based on dynamic weighting of the embodiment of the application establish a measurement benchmark for the prediction performance change and prediction accuracy of the prediction source in the spatial scale under the same real-time time scale by forming a dynamic evaluation index for the prediction accuracy of each prediction source in the past period. The consensus basis for arbitrating the event trend and the data source dynamic weighting for event quantitative analysis are formed by using the timeliness and dynamic of the evaluation index. The optimal weight configuration is obtained by real-time calculation of the weight of each prediction source, so that the comprehensive prediction result is optimized. The accuracy and adaptability of the prediction can be improved by using the complementarity of different prediction tools while adapting to different weather conditions and prediction source performance changes. The decision source optimization combination is realized, the importance and reliability of different prediction agencies are balanced, and the accuracy and stability of the prediction are improved.

[0045] As shown in Figure 1 In an embodiment of the application, step 100 includes:

[0046] Step 110: Normalizing the prediction information and observation information of each prediction agency to convert the prediction data.

[0047] The data source of each prediction agency for precipitation event and rainfall prediction includes the observation information of the rain gauge of the meteorological observation station arranged by the prediction agency in the determined region. The regional division and meteorological observation station arrangement of each prediction agency are different. The prediction information and observation information of each prediction agency are often based on the general data structure to form differentiated data storage.

[0048] The normalization processing of the prediction information and observation information is beneficial to meet the consistency and comparability of the data, and to improve the data analysis capability by comprehensively utilizing the prediction data of multiple data sources. The normalization processing includes but is not limited to data processing in aspects such as time interval, regional division and format definition. The normalization processing mainly includes:

[0049] Data cleaning: removing outliers and missing data to ensure data quality.

[0050] Format definition conversion: Different forecast agencies have different product update frequencies and data element units. The data format of different data sources is unified to facilitate subsequent processing.

[0051] Period adaptation: The time sequence of forecast information and observation information is converted at a fixed time interval, such as adjusting the update frequency at an hourly time interval.

[0052] Region matching: The rainfall data observed by the rain gauge of the meteorological observation station is the rainfall at a specific regional location, while the forecast data provided by the forecast agency is mostly grid meteorological products. It is necessary to match the forecast data (including the corresponding observation data) of different data sources in the region to ensure the consistency of the regional coordinate space of the data.

[0053] The normalized forecast data of each forecast agency is stored in a relational database, and the new forecast data formed continuously is used for incremental update of the forecast data.

[0054] Step 120: Form the TS score of each forecast agency's daily precipitation accurate forecast in the same region according to the forecast data of the same region in the past period.

[0055] In an embodiment of the present application, the past period is 180 days from the current forecast time.

[0056] In an embodiment of the present application, the TS score (Threat Score) mechanism is used to evaluate the accuracy of precipitation forecast. The value of TS score ranges from 0 to 1, and the higher the value, the better the accuracy of the forecast. A completely accurate forecast will get a TS score of 1, while a completely random forecast (i.e. random forecast) will usually have a TS score close to 0.

[0057] In an embodiment of the present application, the forecast data of each forecast agency is hourly forecast data, i.e. one day includes 24 sets of forecast data.

[0058] In an embodiment of the present application, there are four kinds of forecast results, including correct forecast of rain, correct forecast of no rain, false forecast of rain (actually no rain) and missed forecast of rain (actually rain but forecast no rain). In a day, the frequency of the four kinds of forecast results is recorded as TP (True Positive), TN (True Negative), FP (False Positive) and FN (False Negative).

[0059] The value of TS score is:

[0060]

[0061] Where i is the identifier of the forecast agency, j is the identifier of the forecast region, and k is the date in the past period, i.e. TSi,j,k TS score of the forecast organization i in the forecast area j on the date k.

[0062] When TN=24, TP=FP=FN=0 (in fact, no rain and no rain is predicted), the TS score is recorded as 0.5.

[0063] Step 130: Count the number of repeated values of TS scores of each forecast organization in the same region in the same past period.

[0064] The repeated frequency of the same TS score value in the past period (180 days) is f i,j (TS i,j,k ), that is, the number of times each TS score value TS i,j,k occurs in 180 TS scores (corresponding to 180 days).

[0065] Step 140: Form a dynamic evaluation index of each forecast organization in the same region in the same past period according to the TS score, the number of repeated values, and the value threshold.

[0066] In an embodiment of the present application, the dynamic evaluation index is as follows:

[0067] S i,j =∑ TS≥.5 TS i,j,k ×f i,j (TS i,j,k ) (2)

[0068] Wherein, TS≥0.5 is the value threshold of the TS score, TS i,j,k ×f i,j (TS i,j,k ) is the product of the TS score value and the repeated frequency, and the dynamic evaluation index S i,j is the sum of the products of the TS score values greater than 0.5 of the forecast organization i in the forecast area j. The dynamic evaluation index S i,j changes with the adjustment of the current forecast time, the length of the past period, and the TS score value threshold. With the passage of time, the dynamic evaluation index is in dynamic updating.

[0069] The short-term precipitation forecast method based on dynamic weighting of the embodiment of the present application can feedback the reliability and accuracy of the past forecast data through the dynamic evaluation index. The accuracy measurement in a specific length of time is quantified by using the TS score, the measurement precision is improved by using the accumulation and amplification of the TS score product, and the real-time measurement of the improvement of the forecast performance is obtained by using the dynamic nature of the TS score with the passage of time. The dynamic evaluation index establishes the real-time basis for the forecast accuracy in the past period of the current forecast time, the improvement of the forecast performance, and the quantitative comparison between organizations.

[0070] S i,jAs the dynamic evaluation index of the institution i at the weather station j, the higher the index score is, the more accurate the prediction result of the institution at the weather station position is. When the prediction product is comprehensively predicted, the weight of the institution is greater.

[0071] After the prediction data of each prediction institution is fully accumulated and enriched, the selection of the past period can be further optimized as an adjustment parameter. In an embodiment of the present application, the past period can be extended to the annual scale, and the same season, the same month or the same date of different years can be additionally selected, so that the data of the same period in previous years is included in the calculation of the comprehensive score, so as to consider the influence of seasonal changes on the prediction accuracy.

[0072] As shown in the formula (1), in an embodiment of the present application, step 200 includes: Figure 1

[0073] Step 210: determining the positive and negative parties supporting that the precipitation event will occur according to the prediction data of each prediction institution for a specific future period.

[0074] As understood by those skilled in the art, predicting whether a precipitation event will occur is to predict an on-off state, and there is a clear conclusion in the prediction data of each prediction institution. According to the clear conclusion, the positive and negative parties of the prediction institution supporting that the precipitation event will occur can be determined.

[0075] Step 220: taking the cumulative sum of the dynamic evaluation indexes of the positive and negative parties as the arbitration basis to form the prediction result of the precipitation event.

[0076] The dynamic evaluation index reflects the prediction reliability of the prediction institution in the determined area. The arbitration basis is formed by the following cumulative sum difference, and the prediction of the precipitation event is made:

[0077] ∑ 有雨 S i,j -∑ 无雨 S i,j ≥ 0, precipitation;

[0078] ∑ 有雨 S i,j -∑ 无雨 S i,j < 0, no precipitation.

[0079] Step 230: forming the rainfall weight of each prediction institution according to the dynamic evaluation index.

[0080] When the positive party prediction institution has an arbitration advantage, the calculation of the rainfall weight is performed.

[0081] The dynamic evaluation index of the participating prediction institution is used to form the rainfall weight W of each positive party prediction institution according to the following formula: i,j

[0082] ​​

[0083] i.e., S i,j represents the dynamic evaluation index of the regional organization i, and the denominator is the sum of the dynamic evaluation indexes of all participating organizations. In this way, the sum of the weights of all organizations is ensured to be 1, thereby reflecting the relative importance of each organization's forecast in the comprehensive rainfall prediction.

[0084] Step 240: Form the prediction result of the rainfall according to the rainfall weight to integrate the prediction data of the square forecast organization.

[0085] By the rainfall weight W i,j The predicted rainfall from different forecast organizations is integrated to form the weighted rainfall PZ i,j,m :

[0086]

[0087] wherein P i,j,m represents the predicted rainfall of m hours in the future of the organization i at the location of the weather station j, the weighted rainfall PZ i,j,m integrates the predictions of different organizations to obtain a more accurate rainfall prediction.

[0088] The short-term precipitation prediction method based on dynamic weight setting of the embodiment of the application utilizes the dynamic nature of the dynamic evaluation index to form dynamic weight setting. The influence of historical results on the current weight is reduced, and the accuracy of the weight evaluation is improved. By continuously adjusting the weight, the difference between the prediction result and the actual observation data is minimized. In short-term precipitation prediction, the weights of different data sources and prediction methods can be optimized through a dynamic process, so that the comprehensive prediction result is more accurate.

[0089] As Figure 1 shown, in an embodiment of the application, it further includes:

[0090] Step 300: Form the prediction error quantification and evaluation optimization according to the difference between the rainfall prediction and observation.

[0091] The difference between the predicted rainfall and the observed rainfall is used as the error measurement data to measure the error of the short-term rainfall prediction process of the embodiment, to calculate the system error or system loss function. The commonly used error measurement methods of the loss function include root mean square error (RMSE) and standard deviation (STD), which can be used to iteratively quantify and narrow the difference between the prediction output of the complete prediction process and the actual rainfall data. The error measurement forms a continuous optimization process of the key parameter formation process in the short-term rainfall prediction process of the embodiment. Further, dynamic optimization and regular evaluation of the prediction model are formed. This helps to discover performance bottlenecks in a timely manner, and adjust the optimization strategy according to the evaluation results. The evaluation of the model performance can be carried out through various indicators, including but not limited to the accuracy, stability and reliability of the prediction. Through the quantification of the error in the prediction process, iterative optimization of the technical parameters and state parameters in the prediction process is formed until the system error or system loss function is stabilized within an acceptable threshold range.

[0092] An embodiment of the short-term rainfall prediction device based on dynamic weighting of the present application comprises:

[0093] A memory for storing the program code of the short-term rainfall prediction method based on dynamic weighting of the above-mentioned embodiments;

[0094] A processor for executing the program code of the short-term rainfall prediction method based on dynamic weighting of the above-mentioned embodiments.

[0095] The processor can be a DSP (Digital Signal Processor) digital signal processor, a FPGA (Field-Programmable Gate Array) field programmable gate array, a MCU (Microcontroller Unit) system board, a SoC (system on a chip) system board or a PLC (Programmable Logic Controller) minimum system including I / O.

[0096] An embodiment of the short-term rainfall prediction device based on dynamic weighting of the present application is shown in Figure 2 In Figure 2 , the embodiment comprises:

[0097] A dynamic evaluation setting module 10 for forming a dynamic evaluation index of the prediction accuracy of each prediction institution in the same area according to the frequency of accurate rainfall prediction in the same area within the past period at the current prediction time;

[0098] The dynamic weighted forecast module 20 is used to arbitrate the forecast data of various forecasting agencies for future periods at the current forecast time using dynamic evaluation indicators to form a forecast of precipitation events. It dynamically quantifies the rainfall weight of the forecast data according to the dynamic evaluation indicators, and forms a rainfall forecast by integrating the forecast data based on the rainfall weight.

[0099] like Figure 2 As shown, in one embodiment of the present invention, the dynamic evaluation setting module 10 includes:

[0100] The data preprocessing unit 11 is used to normalize the forecast information and observation information from various forecasting agencies and convert them into forecast data;

[0101] The basic TS scoring unit 12 is used to generate TS scores for the accurate daily precipitation forecasts of various forecasting agencies in the same region based on forecast data from previous periods for the same region.

[0102] TS score statistics unit 13 is used to count the frequency of numerical repetition of TS scores from different forecasting agencies in the same region during the same historical period.

[0103] Evaluation index formation unit 14 is used to form dynamic evaluation indicators for each forecasting agency in the same region during the same historical period based on TS score, numerical repetition frequency and numerical threshold.

[0104] like Figure 2 As shown, in one embodiment of the present invention, the dynamic weighting prediction module 20 includes:

[0105] Precipitation event classification unit 21 is used to determine the positive and negative sides supporting the occurrence of precipitation events based on forecast data from various forecasting agencies for a specific future period;

[0106] The precipitation event arbitration unit 22 is used to form the precipitation event forecast result based on the cumulative sum of the dynamic evaluation indicators of both sides.

[0107] The dynamic weighting unit 23 is used to form the rainfall weights of each forecasting agency based on dynamic evaluation indicators.

[0108] The weighted rainfall forecast unit 24 is used to generate a rainfall forecast result by integrating the forecast data of the positive forecasting agency based on the rainfall weight.

[0109] like Figure 2 As shown, in one embodiment of the present invention, it further includes:

[0110] Error feedback optimization module 30 is used to quantify and evaluate forecast errors based on the differences between forecasts and observations of rainfall.

[0111] The above merely describes preferred specific embodiments of the present application, but the protection scope of the present application is not limited thereto, any changes or replacements within the technical scope disclosed by the present application, which can be easily thought by those skilled in the art, should be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A short-term precipitation forecasting method based on dynamic weighting, characterized in that, include: Based on the frequency of accurate precipitation forecasts made in the same area during past periods at the current forecast time, a dynamic evaluation index is formed to assess the forecast accuracy of each forecasting agency in the same area. The forecast of precipitation events is formed by arbitrating the forecast data of various forecasting agencies for future periods at the current forecast time using dynamic evaluation indicators. The forecast data is dynamically quantified by weighting the rainfall based on the dynamic evaluation indicators, and the rainfall forecast is formed by integrating the forecast data based on the rainfall weights. The method of arbitrating precipitation events by using dynamic evaluation indicators to arbitrate the forecast data of various forecasting agencies for future periods at the current forecast time includes: Based on forecast data from various forecasting agencies for a specific future period, determine the two opposing viewpoints supporting the occurrence of precipitation events; The forecast results for precipitation events are formed by summing up the dynamic evaluation indicators of both sides.

2. The short-term precipitation forecasting method based on dynamic weighting as described in claim 1, characterized in that, The formation of the dynamic evaluation indicators includes: The forecast and observation information from various forecasting agencies are normalized and converted into forecast data; Based on forecast data of the same region over previous periods, a TS score is generated for the daily precipitation accuracy forecast of each forecasting agency in the same region. The frequency of repeated TS scores from different forecasting agencies in the same region during the same historical time period was statistically analyzed. Dynamic evaluation indicators for forecasting agencies in the same region during the same historical period are formed based on TS scores, numerical repetition frequency, and numerical thresholds.

3. The short-term precipitation forecasting method based on dynamic weighting as described in claim 2, characterized in that, The aforementioned period is 180 days or a year.

4. The short-term precipitation forecasting method based on dynamic weighting as described in claim 3, characterized in that, The aforementioned past periods also include the same season or the same month in different years.

5. The short-term precipitation forecasting method based on dynamic weighting as described in claim 1, characterized in that, The forecasts that will result in rainfall include: The rainfall weights for each forecasting agency are determined based on dynamic evaluation indicators; The rainfall forecast is generated by combining the forecast data from positive forecasting agencies based on the weighted rainfall amount.

6. The short-term precipitation forecasting method based on dynamic weighting as described in claim 1, characterized in that, Also includes: The forecast error is quantified and evaluated based on the difference between the forecast and the observed rainfall.

7. A short-term precipitation forecasting device based on dynamic weighting, characterized in that, include: A memory for storing program code for the processing of the dynamic weighting-based short-term precipitation forecasting method as described in any one of claims 1 to 6; A processor for executing the program code.

8. A short-term precipitation forecasting device based on dynamic weighting, characterized in that, include: The dynamic evaluation setting module is used to generate dynamic evaluation indicators for the accuracy of forecasts in the same area by each forecasting agency based on the frequency of accurate precipitation forecasts in the same area during the previous time period at the current forecast time. The dynamic weighted forecast module is used to arbitrate the forecast data of various forecasting agencies for future periods at the current forecast time using dynamic evaluation indicators to form a forecast of precipitation events. It dynamically quantifies the rainfall weight of the forecast data according to the dynamic evaluation indicators, and forms a rainfall forecast by integrating the forecast data based on the rainfall weight. The method of arbitrating precipitation events by using dynamic evaluation indicators to arbitrate the forecast data of various forecasting agencies for future periods at the current forecast time includes: Based on forecast data from various forecasting agencies for a specific future period, determine the two opposing viewpoints supporting the occurrence of precipitation events; The forecast results for precipitation events are formed by summing up the dynamic evaluation indicators of both sides.

9. The short-term precipitation forecasting device based on dynamic weighting as described in claim 8, characterized in that, Also includes: The error feedback optimization module is used to quantify and evaluate forecast errors based on the differences between forecasts and observations of rainfall.

Images