Objective forecast and prediction product statistical release method, device and equipment and medium

Abstract

The invention relates to the technical field of forecasting and forecasting, and discloses an objective forecasting and forecasting product statistical release method, device and equipment and a medium, and the method comprises the steps: determining a characteristic quantity type, a release model, a weight function model, a distance model and space-time neighborhood control parameters required by forecasting and forecasting of a target object; obtaining observation data and historical characteristic quantity required by forecasting of the target object, and constructing a sample set; obtaining a characteristic quantity required by forecasting and forecasting of the target object in a forecasting and forecasting period, and interpolating the characteristic quantity into a required discrete grid; forecasting and predicting the target object by adopting the plurality of groups of parameter values, the sample set and the characteristic quantity required for forecasting and predicting the target object in the forecasting and predicting period to obtain a plurality of forecasting and predicting results; and processing the plurality of forecast and prediction results to obtain a final forecast and prediction result. Therefore, the potential of the numerical model and the large model forecasting product in the geoscience field can be fully excavated, and the statistical release level of the numerical model and the large model forecasting product in the geoscience field is further improved.

Description

Technical Field

[0001] The present invention relates to the field of forecasting and prediction technology, and in particular to a method, device, equipment and computer-readable storage medium for statistical interpretation of objective forecasting and prediction products. Background Art

[0002] Accurate weather forecasts and climate predictions are helpful for people to arrange various production and life activities and prevent and reduce disasters in advance. With the development of earth observation technology, information and communication technology, supercomputer technology, data assimilation technology, etc., numerical model forecasts have become the basis of weather forecasts and climate forecasts. However, due to the dynamic framework of numerical models, the physical process is only an approximation of the real atmosphere, the initial values ​​of the model obtained by data assimilation also inevitably have errors, and the chaotic characteristics of atmospheric motion, there are always uncertainties in the forecast products of numerical models. Therefore, statistical interpretation of numerical model forecast products is an important way to improve the level of meteorological forecasts.

[0003] At present, common methods for statistical interpretation of numerical model forecast products include bias correction, linear regression, Kalman filtering, neural network, LightGBM, deep learning, logistic regression, optimal subset regression, frequency matching, simple ensemble average, bias-eliminated ensemble average, pattern weighted average, Bayesian pattern average, multivariate Gaussian ensemble kernel fitting method, and large model-based methods. However, the existing statistical interpretation methods usually use designated training samples to establish "an optimized model" for the target object and a selected interpretation model, and then interpret the corresponding forecast products. This cannot fully consider the uncertainty of numerical model forecast products, nor can it fully tap the potential of numerical model forecast products. The above problems exist not only in the statistical interpretation of weather and climate numerical model forecast products, but also in the statistical interpretation of numerical model forecast products in the ocean and other fields. Similarly, the forecast products of various large models such as meteorology and ocean are also faced with the problem of how to interpret them. Summary of the invention

[0004] The purpose of the present invention is to provide a method for statistical interpretation of objective forecast products, which can fully tap the potential of numerical models and large-model forecast products in the field of geology, and further improve the level of statistical interpretation of numerical models and large-model forecast products in the field of geology; another purpose of the present invention is to provide an objective forecast product statistical interpretation device, equipment and computer-readable storage medium.

[0005] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0006] A statistical interpretation method for objective forecasting products, comprising:

[0007] Determine the target object, and determine the type of characteristic quantity, interpretation model, weight function model, distance model, and spatiotemporal neighborhood control parameters required for forecasting the target object;

[0008] Obtaining the observation data and historical feature quantities required for the forecast of the target object, and constructing a sample set; preprocessing and quality controlling the constructed sample set, and storing it for future use;

[0009] Obtaining the feature quantity required for forecasting the target object in the forecast period, and interpolating it to the required discrete grid;

[0010] Setting multiple sets of parameter values ​​for parameter groups consisting of interpretation model parameters, weight function model parameters, distance model parameters, and spatiotemporal neighborhood control parameters;

[0011] Using the set multiple groups of parameter values, the stored sample sets and the acquired feature quantities required for forecasting the target object in the forecast period, forecasting the target object respectively, and obtaining multiple forecasting results corresponding to the multiple groups of parameter values;

[0012] The multiple forecast prediction results corresponding to the multiple groups of parameter values ​​are processed to obtain a final forecast prediction result.

[0013] In the first aspect, in the above-mentioned objective forecast product statistical interpretation method provided by the present invention, multiple groups of set parameter values, stored sample sets and acquired feature quantities required for forecasting the target object in the forecast period are used to forecast the target object respectively, and multiple forecast results corresponding to the multiple groups of parameter values ​​are obtained, including:

[0014] According to a set of parameter values, for any discrete grid point, the relationship between the observed data and the characteristic quantity is constructed by using the samples in its spatiotemporal neighborhood and the corresponding interpretation model, weight function model, and distance model;

[0015] The target object at the discrete grid point is predicted and forecasted by utilizing the constructed relationship between the observation data and the characteristic quantity and the characteristic quantity of the discrete grid point to obtain a prediction result of the target object at the discrete grid point.

[0016] On the other hand, in the above objective forecast product statistical interpretation method provided by the present invention, in the process of forecasting the target object, it includes:

[0017] The distance model defined by the following formula is used to calculate the distance between the samples in the sample set and the discrete grid points:

[0018] ;

[0019] in, For sample To discrete grid The distance For sample The space-time coordinates of Discrete grid points The space-time coordinates of , ,……, For sample The characteristic quantity of , ,……, Discrete grid points The characteristic quantity, q, , , ,……, is the distance model parameter;

[0020] Or, , , , , ,……, Input into the distance model to obtain the samples in the sample set To discrete grid Distance ;

[0021] Wherein, the distance model is a neural network model, For sample To discrete grid The distance For sample The space-time coordinates of Discrete grid points The space-time coordinates of , ,……, For sample The characteristic quantity of , ,……, Discrete grid points The characteristic quantity of .

[0022] On the other hand, in the above objective forecast product statistical interpretation method provided by the present invention, in the process of forecasting the target object, it includes:

[0023] Calculating the weight corresponding to each sample according to the distance from each sample to the discrete grid point using the weight function model;

[0024] When the weight function model is a distance threshold model, the following formula is used to calculate the weight corresponding to each sample according to the distance from each sample to the discrete grid point:

[0025] ;

[0026] When the weight function model is a Gaussian model, the weight corresponding to each sample is calculated according to the distance from each sample to the discrete grid point using the following formula:

[0027] ;

[0028] When the weight function model is an exponential model, the weight corresponding to each sample is calculated according to the distance from each sample to the discrete grid point using the following formula:

[0029] ;

[0030] When the weight function model is a bisquare model, the weight corresponding to each sample is calculated according to the distance from each sample to the discrete grid point using the following formula:

[0031] ;

[0032] in, For sample To discrete grid The distance is the weight function model parameter, For sample The weight of

[0033] Alternatively, the weight function model is a function of the feature quantity.

[0034] On the other hand, in the above-mentioned objective forecast product statistical interpretation method provided by the present invention, the multiple sets of parameter values ​​include an optimized set of parameter values;

[0035] Or, the plurality of groups of parameter values ​​include a plurality of groups of parameter values ​​whose complementary interpretation effects meet a preset standard;

[0036] Alternatively, the plurality of sets of parameter values ​​include a set of parameter values ​​that is optimal for a particular weather process.

[0037] On the other hand, in the above objective forecast product statistical interpretation method provided by the present invention, multiple forecast results corresponding to the multiple groups of parameter values ​​are processed to obtain the final forecast result, including:

[0038] The test index is used to test the forecast results of each group of parameter values ​​hour by hour, and the test results of each hour are obtained;

[0039] The forecast results corresponding to a set of parameter values ​​with better test results in the most recent time are taken as the final forecast results;

[0040] The test index is one or a combination of mean square error, mean absolute error, accuracy, CRPS score, and TS score.

[0041] On the other hand, in the above objective forecast product statistical interpretation method provided by the present invention, multiple forecast results corresponding to the multiple groups of parameter values ​​are processed to obtain the final forecast result, including:

[0042] When the target object is a continuous object, the multiple forecast results of any discrete grid points corresponding to the multiple sets of parameter values ​​are processed by taking any one of the means of average, maximum, minimum, mode, median and quantile, or the multiple forecast results of any discrete grid points corresponding to the multiple sets of parameter values ​​are taken as a whole to evaluate the uncertainty of the target object at the discrete grid points;

[0043] When the target object is a classified object, a voting method is used to process multiple forecast prediction results of any discrete grid points corresponding to the multiple groups of parameter values, or multiple forecast prediction results of any discrete grid points corresponding to the multiple groups of parameter values ​​are taken as a whole to evaluate the uncertainty of the target object at the discrete grid points.

[0044] In order to solve the above technical problems, the present invention also provides an objective forecast product statistical interpretation device, comprising:

[0045] The model and parameter determination module is used to determine the target object and determine the type of characteristic quantity, interpretation model, weight function model, distance model, and spatiotemporal neighborhood control parameters required for the forecast of the target object;

[0046] A sample set construction module is used to obtain the observation data and historical feature quantities required for the forecast of the target object, and to construct a sample set; to preprocess and quality control the constructed sample set, and to store it for future use;

[0047] A feature quantity acquisition module is used to acquire the feature quantity required for the forecast of the target object in the forecast period and interpolate it to the required discrete grid;

[0048] A parameter value setting module, used to set multiple groups of parameter values ​​for a parameter group consisting of an interpretation model parameter, a weight function model parameter, a distance model parameter, and a spatiotemporal neighborhood control parameter;

[0049] A forecast result acquisition module is used to use the set multiple groups of parameter values, the stored sample set and the acquired feature quantities required for forecasting the target object in the forecast period to forecast the target object respectively and obtain multiple forecast results corresponding to the multiple groups of parameter values;

[0050] The final forecast prediction result obtaining module is used to process the multiple forecast prediction results corresponding to the multiple groups of parameter values ​​to obtain the final forecast prediction result.

[0051] In order to solve the above technical problems, the present invention also provides an objective forecast product statistical interpretation device, comprising:

[0052] Memory for storing computer programs;

[0053] A processor is used to implement the steps of the statistical interpretation method of objective forecast products as described above when executing the computer program.

[0054] A computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of the statistical interpretation method for objective forecast products as described above.

[0055] It can be seen from the above technical scheme that the statistical interpretation method of an objective forecast product provided by the present invention includes: determining a target object, and determining the type of characteristic quantity, interpretation model, weight function model, distance model, and spatiotemporal neighborhood control parameters required for forecasting the target object; obtaining the observation data and historical characteristic quantities required for forecasting the target object, and constructing a sample set; preprocessing and quality controlling the constructed sample set, and storing it for standby use; obtaining the characteristic quantities required for forecasting the target object in the forecast period, and interpolating them to the required discrete grid; setting multiple groups of parameter values ​​for the parameter group consisting of the interpretation model parameters, weight function model parameters, distance model parameters, and spatiotemporal neighborhood control parameters; using the set multiple groups of parameter values, the stored sample set, and the obtained characteristic quantities required for forecasting the target object in the forecast period, respectively forecasting and predicting the target object, and obtaining multiple forecast prediction results corresponding to the multiple groups of parameter values; processing the multiple forecast prediction results corresponding to the multiple groups of parameter values ​​to obtain the final forecast prediction result.

[0056] The beneficial effect of the present invention is that the statistical interpretation method of the objective forecast product provided by the present invention sets multiple sets of parameter values ​​for the parameter group consisting of the interpretation model parameters, weight function model parameters, distance model parameters and spatiotemporal neighborhood control parameters, uses the set multiple sets of parameter values, the stored sample set and the acquired feature quantities required for the forecast of the target object in the forecast period, respectively forecasts the target object, obtains multiple forecast results corresponding to the multiple sets of parameter values, processes the multiple forecast results corresponding to the multiple sets of parameter values, and obtains the final forecast result. Thus, the potential of numerical models and large model forecast products in the field of geoscience can be fully explored, and the statistical interpretation level of numerical models and large model forecast products in the field of geoscience can be further improved.

[0057] In addition, the present invention also provides objective forecast and prediction product statistical interpretation devices, equipment and computer-readable storage media corresponding to the above-mentioned objective forecast and prediction product statistical interpretation method, which have the above-mentioned technical effects and will not be repeated here. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or related technologies, the drawings required for use in the embodiments or related technical descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0059] Figure 1 A flow chart of a method for statistical interpretation of objective forecast products provided by an embodiment of the present invention;

[0060] Figure 2 A schematic diagram of the structure of a device for statistical interpretation of objective forecast and prediction products provided by an embodiment of the present invention;

[0061] Figure 3 A schematic diagram of the structure of a device for statistical interpretation of objective forecast and prediction products provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0062] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0063] The core of the present invention is to provide a method, device, equipment and medium for statistical interpretation of objective forecast products to solve the technical problem of not fully considering the uncertainty of numerical model forecast products and not fully tapping the potential of numerical model forecast products.

[0064] In order to enable those skilled in the art to better understand the scheme of the present invention, the present invention is further described in detail below in conjunction with the accompanying drawings and specific implementation methods. Figure 1 A flowchart of a statistical interpretation method for objective forecast products provided by an embodiment of the present invention, such as Figure 1 As shown, the method includes:

[0065] S101, determining the target object, and determining the type of characteristic quantity, interpretation model, weight function model, distance model, and spatiotemporal neighborhood control parameters required for forecasting the target object;

[0066] When it is necessary to forecast the target object in the forecast period, the target object is determined, and the type of characteristic quantity, interpretation model, weight function model, distance model, and spatiotemporal neighborhood control parameters required for the forecast of the target object are determined.

[0067] The space-time neighborhood control parameters control the size and shape of the space-time neighborhood; the space-time neighborhood control parameters may include the maximum number of samples S of the space-time neighborhood, the edge length of the cube space-time neighborhood, the sliding training period T, etc.

[0068] S102, obtaining the observation data and historical characteristic quantities required for the forecast of the target object, and constructing a sample set; preprocessing and quality control the constructed sample set, and storing it for later use;

[0069] After determining the target object and determining the type of feature quantity, interpretation model, weight function model, distance model, and spatiotemporal neighborhood control parameters required for forecasting the target object, obtain the observation data and historical feature quantities required for forecasting the target object, and construct a sample set. For example, a sample set in the form of "observation data, feature quantity 1, feature quantity 2, ..., feature quantity n" can be constructed, and the constructed sample set is preprocessed and quality controlled, and stored for later use.

[0070] The observation data can be ground station observation data, buoy station observation data, radiosonde observation data, multi-source data fusion real-time data, and reanalysis data. The preprocessing and quality control of the sample set include but are not limited to removing redundant samples and abnormal samples, and converting samples as needed.

[0071] The target object can be one or a combination of a variety of ground layer variables (such as 2-meter air temperature, 10-meter U wind component, 10-meter V wind component, wind speed, sea level pressure, precipitation), atmospheric variables at different altitudes (such as geopotential height, air temperature, U component of wind, V component of wind, relative humidity, air pressure), ocean variables at different depths (such as ocean temperature, salinity, density, current velocity, latitudinal and longitudinal components of ocean currents), and soil variables at different depths (such as soil temperature, soil moisture).

[0072] S103, obtaining the feature quantity required for forecasting the target object in the forecast period, and interpolating it to the required discrete grid;

[0073] After determining the target object and determining the type of feature quantity, interpretation model, weight function model, distance model, and spatiotemporal neighborhood control parameters required for forecasting the target object, obtain the feature quantity required for forecasting the target object in the forecast period and interpolate it to the required discrete grid.

[0074] The characteristic quantities required for the forecast of the target object may include one or a combination of multiple forecast products of multiple weather and climate numerical models, multiple earth system numerical models, multiple ocean numerical models, multiple meteorological large models and multiple ocean large models. The forecast products may be the temperature, air pressure, wind speed, relative humidity, visibility, soil temperature, soil moisture, precipitation, geopotential height at different levels forecasted by weather and climate numerical models, earth system numerical models and meteorological large models, and the seawater temperature, salinity, density, velocity and latitudinal and longitudinal components of ocean currents at different depths forecasted by ocean numerical models and ocean large models.

[0075] The characteristic quantities required for the forecast of the target object may also include altitude, error in the start time of reporting, and various derived quantities of the forecast product. For example, for the daily maximum, minimum and hourly temperature forecast, the derived quantity may be one or more of the temperature at 2 meters, the daily maximum and minimum temperature, the sea level pressure, the relative humidity, the difference between the wind field and the previous day, the wind field, cloud cover, height field, the 12-hour average of the sea level pressure field, and the 12-hour cumulative value of precipitation.

[0076] S104, setting multiple groups of parameter values ​​for the parameter group consisting of the interpretation model parameters, the weight function model parameters, the distance model parameters and the spatiotemporal neighborhood control parameters;

[0077] After obtaining the characteristic quantities required for forecasting the target object in the forecast period, multiple groups of parameter values ​​are set for the parameter group consisting of the interpretation model parameters, weight function model parameters, distance model parameters and spatiotemporal neighborhood control parameters. The interpretation model parameters, weight function model parameters, distance model parameters and spatiotemporal neighborhood control parameters can specifically include multiple parameters. The interpretation model parameters can be interpretation model hyperparameters or other parameters, which are not limited in the embodiments of the present invention. For example, the parameter group can include interpretation model hyperparameter 1, interpretation model hyperparameter 2, ..., interpretation model hyperparameter M, weight function model parameter 1, weight function model parameter 2, ..., weight function model parameter W, distance model parameter 1, distance model parameter 2, ..., distance model parameter L, spatiotemporal neighborhood control parameter 1, spatiotemporal neighborhood control parameter 2, ..., spatiotemporal neighborhood control parameter N.

[0078] Depending on the specific implementation, some parameters in the parameter group may be discarded. For a weighted support vector machine, the model hyperparameters may include a penalty coefficient, a kernel function parameter, and the like.

[0079] S105, using the set multiple groups of parameter values, the stored sample sets and the acquired feature quantities required for forecasting the target object in the forecast period, forecasting the target object respectively, and obtaining multiple forecasting results corresponding to the multiple groups of parameter values;

[0080] After setting multiple groups of parameter values ​​for the parameter group consisting of the interpretation model parameters, weight function model parameters, distance model parameters and spatiotemporal neighborhood control parameters, the target objects are forecasted and predicted respectively using the set multiple groups of parameter values, the stored sample set and the acquired characteristic quantities required for forecasting the target object in the forecast period, and multiple forecast results corresponding to the multiple groups of parameter values ​​are obtained, thereby realizing the forecasting of the target object using the multiple groups of parameter values.

[0081] S106: Process multiple forecast results corresponding to multiple groups of parameter values ​​to obtain a final forecast result.

[0082] After obtaining multiple forecast results corresponding to multiple sets of parameter values, the multiple forecast results corresponding to the multiple sets of parameter values ​​are processed to obtain the final forecast results. By using multiple sets of parameter values ​​to forecast the target object, the potential of numerical models and large model forecast products in the field of geoscience can be fully explored, and the statistical interpretation level of numerical models and large model forecast products in the field of geoscience can be further improved.

[0083] In the above-mentioned objective forecast product statistical interpretation method provided by the present invention, multiple groups of parameter values ​​are set for the parameter group consisting of the interpretation model parameters, weight function model parameters, distance model parameters and spatiotemporal neighborhood control parameters, and the set multiple groups of parameter values, the stored sample set and the acquired characteristic quantities required for the forecast of the target object in the forecast period are used to forecast the target object respectively, and multiple forecast results corresponding to the multiple groups of parameter values ​​are obtained, and the multiple forecast results corresponding to the multiple groups of parameter values ​​are processed to obtain the final forecast results. In this way, the potential of numerical models and large model forecast products in the field of geoscience can be fully explored, and the statistical interpretation level of numerical models and large model forecast products in the field of geoscience can be further improved.

[0084] It should be noted that, based on the above embodiment, the embodiment of the present invention also provides corresponding improved solutions. In the subsequent embodiments, the same steps or corresponding steps as those in the above embodiment can be referenced to each other, and the corresponding beneficial effects can also be referenced to each other, which will not be repeated one by one in the following improved embodiments.

[0085] In a specific implementation of the present invention, step S105 may include the following steps:

[0086] Step 1: According to a set of parameter values, for any discrete grid point, the relationship between the observed data and the characteristic quantity is constructed by using the samples in its spatiotemporal neighborhood and the corresponding interpretation model, weight function model, and distance model;

[0087] Step 2: Use the constructed relationship between the observed data and the characteristic quantity and the characteristic quantity of the discrete grid points to forecast the target object of the discrete grid points and obtain a forecast result of the target object of the discrete grid points.

[0088] For the convenience of description, the above two steps can be combined for explanation.

[0089] When using multiple sets of set parameter values ​​to forecast the target object, according to the set set of parameter values, for any discrete grid point, the relationship between the observed data and the characteristic quantity is constructed using the samples in its spatiotemporal neighborhood and the corresponding interpretation model, weight function model, and distance model. The target object at the discrete grid point is forecasted using the constructed relationship between the observed data and the characteristic quantity and the characteristic quantity of the discrete grid point to obtain a forecast result for the discrete grid point target object.

[0090] The applied models can be weighted linear regression, weighted partial least squares regression, weighted logistic regression, weighted lasso regression, weighted neural network, weighted random forest, weighted XGBoost, weighted LightGBM, weighted support vector machine and related multi-scale models.

[0091] In a specific implementation of the present invention, the process of forecasting the target object may include the following steps:

[0092] The distance model defined by the following formula is used to calculate the distance from the sample in the sample set to the discrete grid point:

[0093] ;

[0094] in, For sample To discrete grid The distance For sample The space-time coordinates of Discrete grid points The space-time coordinates of , ,……, For sample The characteristic quantity of , ,……, Discrete grid points The characteristic quantity, q, , , ,……, is the distance model parameter;

[0095] Or, , , , , ,……, Input into the distance model to obtain samples in the sample set To discrete grid Distance ;

[0096] Among them, the distance model is a neural network model. For sample To discrete grid The distance For sample The space-time coordinates of Discrete grid points The space-time coordinates of , ,……, For sample The characteristic quantity of , ,……, Discrete grid points The characteristic quantity of .

[0097] In the process of forecasting the target object, the sample is calculated by using the above calculation formula or neural network model. To discrete grid Distance , realizing the sample With discrete grid Distance Fast and accurate calculation, simple and easy to operate.

[0098] In a specific implementation of the present invention, the process of forecasting the target object may include the following steps:

[0099] The weight function model is used to calculate the weight corresponding to each sample according to the distance from each sample to the discrete grid point;

[0100] When the weight function model is a distance threshold model, the following formula is used to calculate the weight corresponding to each sample according to the distance from each sample to the discrete grid point:

[0101] ;

[0102] When the weight function model is a Gaussian model, the following formula is used to calculate the weight corresponding to each sample according to the distance from each sample to the discrete grid point:

[0103] ;

[0104] When the weight function model is an exponential model, the following formula is used to calculate the weight corresponding to each sample according to the distance from each sample to the discrete grid point:

[0105] ;

[0106] When the weight function model is a bisquare model, the following formula is used to calculate the weight corresponding to each sample according to the distance from each sample to the discrete grid point:

[0107] ;

[0108] in, For sample To discrete grid The distance is the weight function model parameter, For sample The weight of

[0109] Alternatively, the weight function model is a function of the feature quantity.

[0110] In the process of forecasting the target object, the weight corresponding to each sample is calculated by the above formula or the function of the characteristic quantity. For example, each sample can be weighted in such a way that the smaller the distance, the greater the weight, thereby realizing the accurate and rapid determination of the weight of each sample. The observation data with a large weight plays a greater role in the forecast result, making the forecast result closer to the actual situation.

[0111] The weight function model can be a function of the characteristic quantity. For example, the weight function model can be a function P of the precipitation P predicted by the numerical model or the large meteorological model forecast. α , where α is the weight function model parameter.

[0112] In one embodiment of the present invention, the plurality of sets of parameter values ​​include an optimized set of parameter values;

[0113] Or, the plurality of sets of parameter values ​​include a plurality of sets of parameter values ​​whose complementary interpretation effects meet a preset standard;

[0114] Alternatively, the plurality of sets of parameter values ​​include a set of parameter values ​​that is optimal for a particular weather process.

[0115] When setting multiple sets of parameter values, the multiple sets of parameter values ​​include the most optimized set of parameter values, or the multiple sets of parameter values ​​include several sets of parameter values ​​whose complementary interpretation effects meet preset standards, or the multiple sets of parameter values ​​include the optimal set of parameter values ​​for a specific weather process. This can fully tap the potential of numerical models and large-scale model forecast products in the field of geosciences, and further improve the statistical interpretation level of numerical models and large-scale model forecast products in the field of geosciences.

[0116] In a specific implementation of the present invention, step S106 may include the following steps:

[0117] Step 1: Use the test index to test the forecast results of each group of parameter values ​​hour by hour to obtain the test results of each hour;

[0118] Step 2: Take the forecast results corresponding to a set of parameter values ​​with better test results in the most recent time as the final forecast results;

[0119] The test index is one or more combinations of mean square error, mean absolute error, accuracy, CRPS score, and TS score.

[0120] For the convenience of description, the above two steps can be combined for explanation.

[0121] When processing multiple forecast results corresponding to multiple groups of parameter values, the forecast results of each group of parameter values ​​are tested hour by hour using one or more of the test indicators of mean square error, mean absolute error, accuracy, CRPS score, and TS score, and the test results of each time are obtained. The forecast results corresponding to a group of parameter values ​​with better test results in the most recent time are used as the final forecast results. By setting the test indicators to screen multiple forecast results, the forecast results corresponding to a group of parameter values ​​with better test results in the most recent time are used as the final forecast results, thereby improving the accuracy of the final forecast results.

[0122] In a specific implementation of the present invention, step S106 may include the following steps:

[0123] Step 1: When the target object is a continuous object, use any one of the following methods: average, maximum, minimum, mode, median, and quantile to process multiple forecast results of any discrete grid point corresponding to multiple sets of parameter values, or take multiple forecast results of any discrete grid point corresponding to multiple sets of parameter values ​​as a whole to evaluate the uncertainty of the discrete grid point target object;

[0124] Step 2: When the target object is a categorized object, the voting method is used to process multiple forecast results of any discrete grid point corresponding to multiple sets of parameter values, or the multiple forecast results of any discrete grid point corresponding to multiple sets of parameter values ​​are taken as a whole to evaluate the uncertainty of the discrete grid target object.

[0125] When the target object is a continuous object, the multiple forecast results of any discrete grid point corresponding to multiple sets of parameter values ​​are processed by taking any of the following methods: average, maximum, minimum, mode, median, or quantile, or the multiple forecast results of any discrete grid point corresponding to multiple sets of parameter values ​​are taken as a whole to evaluate the uncertainty of the discrete grid target object. When the target object is a categorical object, the voting method is used to process the multiple forecast results of any discrete grid point corresponding to multiple sets of parameter values, or the multiple forecast results of any discrete grid point corresponding to multiple sets of parameter values ​​are taken as a whole to evaluate the uncertainty of the discrete grid target object. By processing multiple forecast results corresponding to multiple sets of parameter values, the scientific nature of the final forecast results is further improved.

[0126] In a specific implementation application, for example, the target object is determined to be the 2-meter daily minimum temperature in Hainan Island in the next 24 hours, and the feature quantity required for the target object forecast is determined to be the 2-meter daily minimum temperature in the next 24 hours predicted by the European Center for Medium-Range Weather Forecasts (ECMWF). The observation data is the daily minimum temperature observed by the automatic weather station in Hainan Island and a sample set is constructed. The interpretation model is weighted linear regression, and the weight function model is a distance threshold model. The formula is as follows:

[0127] ;

[0128] in, For sample To discrete grid The distance is the weight function model parameter (bandwidth), r takes a sufficiently large value, For sample The weight of .

[0129] The distance model is as follows:

[0130] ;

[0131] in, , q=2, and the other distance model parameters are all taken as 0. The spatiotemporal neighborhood control parameter 1 is set as the sliding training period T, and the spatiotemporal neighborhood control parameter 2 is set as the maximum number of spatiotemporal neighborhood samples S. Two sets of parameter values ​​with strong complementarity are used (spatiotemporal neighborhood control parameter 1=10 days, spatiotemporal neighborhood control parameter 2=10) and (spatiotemporal neighborhood control parameter 1=50 days, spatiotemporal neighborhood control parameter 2=50). The 2-meter daily minimum temperature in the next 24 hours predicted by the European Center for Medium-Range Weather Forecasts (ECMWF) is statistically interpreted, and the forecast results of the two sets of parameters are tested hour by hour using the mean absolute error. According to the latest test results, the forecast result corresponding to the set of parameters with the smallest mean absolute error is taken as the forecast result of the 2-meter daily minimum temperature in Hainan Island in the next 24 hours.

[0132] Corresponding to the above method embodiment, the present invention further provides an objective forecast and prediction product statistical interpretation device. The objective forecast and prediction product statistical interpretation device described below and the objective forecast and prediction product statistical interpretation method described above can refer to each other.

[0133] See also Figure 2 , Figure 2 This is a structural block diagram of a device for statistical interpretation of objective forecast products in an embodiment of the present invention. The device may include:

[0134] The model and parameter determination module 10 is used to determine the target object and determine the type of characteristic quantity, interpretation model, weight function model, distance model, and spatiotemporal neighborhood control parameters required for the forecast of the target object;

[0135] The sample set construction module 11 is used to obtain the observation data and historical feature quantities required for the forecast of the target object, and to construct a sample set; to preprocess and quality control the constructed sample set, and to store it for future use;

[0136] The feature quantity acquisition module 12 is used to acquire the feature quantity required for the forecast of the target object in the forecast period and interpolate it to the required discrete grid;

[0137] The parameter value setting module 13 is used to set multiple groups of parameter values ​​for the parameter group consisting of the interpretation model parameters, the weight function model parameters, the distance model parameters and the spatiotemporal neighborhood control parameters;

[0138] The forecast result acquisition module 14 is used to use the set multiple groups of parameter values, the stored sample set and the acquired feature quantities required for forecasting the target object in the forecast period to forecast the target object respectively and obtain multiple forecast results corresponding to the multiple groups of parameter values;

[0139] The final forecast prediction result obtaining module 15 is used to process multiple forecast prediction results corresponding to multiple groups of parameter values ​​to obtain a final forecast prediction result.

[0140] In the above-mentioned objective forecast product statistical interpretation method provided by the present invention, multiple groups of parameter values ​​are set for the parameter group consisting of the interpretation model parameters, weight function model parameters, distance model parameters and spatiotemporal neighborhood control parameters, and the set multiple groups of parameter values, the stored sample set and the acquired characteristic quantities required for the forecast of the target object in the forecast period are used to forecast the target object respectively, and multiple forecast results corresponding to the multiple groups of parameter values ​​are obtained, and the multiple forecast results corresponding to the multiple groups of parameter values ​​are processed to obtain the final forecast results. In this way, the potential of numerical models and large model forecast products in the field of geoscience can be fully explored, and the statistical interpretation level of numerical models and large model forecast products in the field of geoscience can be further improved.

[0141] Since the embodiments of the device part correspond to the embodiments of the method part, the embodiments of the device part refer to the description of the embodiments of the method part, and are not described here. And it has the same beneficial effects as the above-mentioned objective forecast product statistical interpretation method.

[0142] Furthermore, in a specific implementation, in the above-mentioned objective forecast product statistical interpretation device provided in an embodiment of the present invention, the forecast result acquisition module 14 includes:

[0143] The relationship building submodule is used to build the relationship between the observed data and the characteristic quantity for any discrete grid point based on a set of parameter values, using the samples in its spatiotemporal neighborhood and the corresponding interpretation model, weight function model, and distance model;

[0144] The forecast result obtaining submodule is used to forecast the target object of the discrete grid point by using the constructed relationship between the observation data and the characteristic quantity and the characteristic quantity of the discrete grid point to obtain a forecast result of the target object of the discrete grid point.

[0145] Furthermore, in the specific implementation, in the above-mentioned objective forecast product statistical interpretation device provided in the embodiment of the present invention, the forecast result acquisition module 14 includes a distance calculation submodule, which is specifically used to calculate the distance from the sample in the sample set to the discrete grid point in the process of forecasting the target object using the distance model defined by the following formula:

[0146] ;

[0147] in, For sample To discrete grid The distance For sample The space-time coordinates of Discrete grid points The space-time coordinates of , ,……, For sample The characteristic quantity of , ,……, Discrete grid points The characteristic quantity, q, , , ,……, is the distance model parameter;

[0148] Or, , , , , ,……, Input into the distance model to obtain samples in the sample set To discrete grid Distance ;

[0149] Among them, the distance model is a neural network model. For sample To discrete grid The distance For sample The space-time coordinates of Discrete grid points The space-time coordinates of , ,……, For sample The characteristic quantity of , ,……, Discrete grid points The characteristic quantity of .

[0150] Furthermore, in the specific implementation, in the above-mentioned objective forecast product statistical interpretation device provided by the embodiment of the present invention, the forecast result acquisition module 14 includes a weight calculation submodule, and the weight calculation submodule is specifically used to calculate the weight corresponding to each sample according to the distance from each sample to the discrete grid point by using the weight function model in the process of forecasting the target object;

[0151] When the weight function model is a distance threshold model, the following formula is used to calculate the weight corresponding to each sample according to the distance from each sample to the discrete grid point:

[0152] ;

[0153] When the weight function model is a Gaussian model, the following formula is used to calculate the weight corresponding to each sample according to the distance from each sample to the discrete grid point:

[0154] ;

[0155] When the weight function model is an exponential model, the following formula is used to calculate the weight corresponding to each sample according to the distance from each sample to the discrete grid point:

[0156] ;

[0157] When the weight function model is a bisquare model, the following formula is used to calculate the weight corresponding to each sample according to the distance from each sample to the discrete grid point:

[0158] ;

[0159] in, For sample To discrete grid The distance is the weight function model parameter, For sample The weight of

[0160] Alternatively, the weight function model is a function of the feature quantity.

[0161] Furthermore, in a specific implementation, in the above-mentioned objective forecast product statistical interpretation device provided in an embodiment of the present invention, the final forecast result acquisition module 15 includes:

[0162] The submodule for obtaining the test results at each time is used to test the forecast results of each group of parameter values ​​at each time by using the test index to obtain the test results at each time;

[0163] The final forecast result obtaining submodule is used to take the forecast result corresponding to a set of parameter values ​​with better test results in the most recent time as the final forecast result;

[0164] The test index is one or more combinations of mean square error, mean absolute error, accuracy, CRPS score, and TS score.

[0165] Furthermore, in a specific implementation, in the above-mentioned objective forecast product statistical interpretation device provided in an embodiment of the present invention, the final forecast result acquisition module 15 includes:

[0166] The first forecast result processing submodule is used to process multiple forecast results of any discrete grid point corresponding to multiple sets of parameter values ​​by taking any one of the following methods: average value, maximum value, minimum value, mode, median, and quantile when the target object is a continuous object, or to take multiple forecast results of any discrete grid point corresponding to multiple sets of parameter values ​​as a whole to evaluate the uncertainty of the discrete grid point target object;

[0167] The second forecast result processing submodule is used to process multiple forecast results of any discrete grid point corresponding to multiple groups of parameter values ​​by voting method when the target object is a classified object, or to take multiple forecast results of any discrete grid point corresponding to multiple groups of parameter values ​​as a whole to evaluate the uncertainty of the discrete grid point target object.

[0168] Figure 3 This is a schematic diagram of the structure of the objective forecast product statistical interpretation device provided by the embodiment of the present invention. This embodiment is based on the hardware perspective, such as Figure 3 As shown in the figure, the statistical interpretation equipment for objective forecast products includes:

[0169] A memory 20, used for storing computer programs;

[0170] The processor 21 is used to implement the steps of the objective forecast product statistical interpretation method mentioned in the above embodiment when executing the computer program.

[0171] Among them, the processor 21 may include one or more processing cores, such as a 4-core processor, an 8-core processor, etc. The processor 21 can be implemented in at least one hardware form of a digital signal processor (DSP), a field-programmable gate array (FPGA), and a programmable logic array (PLA). The processor 21 may also include a main processor and a coprocessor. The main processor is a processor for processing data in the awake state, also known as a CPU; the coprocessor is a low-power processor for processing data in the standby state. In some embodiments, the processor 21 may be integrated with a graphics processing unit (GPU), which is responsible for rendering and drawing the content to be displayed on the display screen. In some embodiments, the processor 21 may also include an artificial intelligence (AI) processor, which is used to process computing operations related to machine learning.

[0172] The memory 20 may include one or more computer-readable storage media, which may be non-transitory. The memory 20 may also include a high-speed random access memory, and a non-volatile memory, such as one or more disk storage devices, flash memory storage devices. In this embodiment, the memory 20 is at least used to store the following computer program 201, wherein, after the computer program is loaded and executed by the processor 21, it can implement the relevant steps of the objective forecast product statistical interpretation method disclosed in any of the aforementioned embodiments. In addition, the resources stored in the memory 20 may also include an operating system 202 and data 203, etc., and the storage method may be temporary storage or permanent storage. Among them, the operating system 202 may include Windows, Unix, Linux, etc. The data 203 may include, but is not limited to, the data involved in the above-mentioned objective forecast product statistical interpretation method, etc.

[0173] In some embodiments, the objective forecast product statistical interpretation device may further include a display screen 22, an input and output interface 23, a communication interface 24, a power supply 25 and a communication bus 26. Those skilled in the art will appreciate that Figure 3The structure shown in does not constitute a limitation on the statistical interpretation and utilization device of objective forecast and prediction products, and may include more or fewer components than those shown in the figure. The statistical interpretation and utilization device of objective forecast and prediction products provided by the embodiment of the present invention includes a memory and a processor. When the processor executes the program stored in the memory, it can implement the following method: the statistical interpretation and utilization method of objective forecast and prediction products, the effect is the same as above.

[0174] Finally, the present invention also provides an embodiment corresponding to a computer-readable storage medium. The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps recorded in the above method embodiment are implemented.

[0175] It is understandable that if the method in the above embodiment is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium and executes all or part of the steps of the methods of each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk and other media that can store program codes. The computer-readable storage medium provided by the present invention can realize the above-mentioned objective forecast product statistical interpretation method, and the effect is the same as above.

[0176] Finally, the present invention also provides an embodiment corresponding to a computer program product. The computer program product includes a computer program / instruction, and when the computer program / instruction is executed by a processor, the steps recorded in the above-mentioned objective forecast product statistical interpretation method embodiment are implemented. The computer program product provided by the present invention can implement the above-mentioned objective forecast product statistical interpretation method, and the effect is the same as above.

[0177] It should also be noted that, in this specification, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "comprise", "include" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, an element defined by the statement "comprises a ..." does not exclude the presence of other identical elements in the process, method, article or device including the element.

[0178] The objective forecast product statistical interpretation method, device, equipment and medium provided by the present invention are introduced in detail above. The various embodiments in the specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same and similar parts between the various embodiments can be referenced to each other. For the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part description. It should be pointed out that for ordinary technicians in this technical field, without departing from the principle of the present invention, several improvements and modifications can be made to the present invention, and these improvements and modifications also fall within the scope of protection of the present invention.

Claims

1. A statistical interpretation method for objective forecast products, characterized in that: include: Determine the target object, and determine the type of characteristic quantity, interpretation model, weight function model, distance model, and spatiotemporal neighborhood control parameters required for forecasting the target object; Obtaining the observation data and historical feature quantities required for the forecast of the target object, and constructing a sample set; preprocessing and quality controlling the constructed sample set, and storing it for future use; Obtaining the feature quantity required for forecasting the target object in the forecast period, and interpolating it to the required discrete grid; Setting multiple sets of parameter values ​​for parameter groups consisting of interpretation model parameters, weight function model parameters, distance model parameters, and spatiotemporal neighborhood control parameters; Using the set multiple groups of parameter values, the stored sample sets and the acquired feature quantities required for forecasting the target object in the forecast period, forecasting the target object respectively, and obtaining multiple forecasting results corresponding to the multiple groups of parameter values; The multiple forecast prediction results corresponding to the multiple groups of parameter values ​​are processed to obtain a final forecast prediction result.

2. The objective forecast product statistical interpretation method according to claim 1, characterized in that: Using the set multiple groups of parameter values, the stored sample sets and the acquired feature quantities required for forecasting the target object in the forecast period, forecasting the target object is performed respectively to obtain multiple forecasting results corresponding to the multiple groups of parameter values, including: According to a set of parameter values, for any discrete grid point, the relationship between the observed data and the characteristic quantity is constructed by using the samples in its spatiotemporal neighborhood and the corresponding interpretation model, weight function model, and distance model; The target object at the discrete grid point is predicted and forecasted by utilizing the constructed relationship between the observation data and the characteristic quantity and the characteristic quantity of the discrete grid point to obtain a prediction result of the target object at the discrete grid point.

3. The statistical interpretation method of objective forecast products according to claim 2, characterized in that: The process of forecasting the target object includes: The distance model defined by the following formula is used to calculate the distance between the samples in the sample set and the discrete grid points: ； in, For sample To discrete grid The distance For sample The space-time coordinates of Discrete grid points The space-time coordinates of , ,……, For sample The characteristic quantity of , ,……, Discrete grid points The characteristic quantity, q, , , ,……, is the distance model parameter; Or, , , , , ,……, Input into the distance model to obtain the samples in the sample set To discrete grid Distance ; Wherein, the distance model is a neural network model, For sample To discrete grid The distance For sample The space-time coordinates of Discrete grid points The space-time coordinates of , ,……, For sample The characteristic quantity of , ,……, Discrete grid points The characteristic quantity of .

4. The statistical interpretation method of objective forecast products according to claim 2, characterized in that: The process of forecasting the target object includes: Calculating the weight corresponding to each sample according to the distance from each sample to the discrete grid point using the weight function model; When the weight function model is a distance threshold model, the following formula is used to calculate the weight corresponding to each sample according to the distance from each sample to the discrete grid point: ； When the weight function model is a Gaussian model, the weight corresponding to each sample is calculated according to the distance from each sample to the discrete grid point using the following formula: ； When the weight function model is an exponential model, the weight corresponding to each sample is calculated according to the distance from each sample to the discrete grid point using the following formula: ； When the weight function model is a bisquare model, the weight corresponding to each sample is calculated according to the distance from each sample to the discrete grid point using the following formula: ； in, For sample To discrete grid The distance is the weight function model parameter, For sample The weight of Alternatively, the weight function model is a function of the feature quantity.

5. The objective forecast product statistical interpretation method according to claim 1, characterized in that: The multiple sets of parameter values ​​include an optimized set of parameter values; Or, the plurality of groups of parameter values ​​include a plurality of groups of parameter values ​​whose complementary interpretation effects meet a preset standard; Alternatively, the plurality of sets of parameter values ​​include a set of parameter values ​​that is optimal for a particular weather process.

6. The objective forecast product statistical interpretation method according to claim 1, characterized in that: Processing the multiple forecast results corresponding to the multiple groups of parameter values ​​to obtain a final forecast result includes: The test index is used to test the forecast results of each group of parameter values ​​hour by hour, and the test results of each hour are obtained; The forecast results corresponding to a set of parameter values ​​with better test results in the most recent time are taken as the final forecast results; The test index is one or a combination of mean square error, mean absolute error, accuracy, CRPS score, and TS score.

7. The objective forecast product statistical interpretation method according to claim 1, characterized in that: Processing the multiple forecast results corresponding to the multiple groups of parameter values ​​to obtain a final forecast result includes: When the target object is a continuous object, the multiple forecast results of any discrete grid points corresponding to the multiple sets of parameter values ​​are processed by taking any one of the means of average, maximum, minimum, mode, median and quantile, or the multiple forecast results of any discrete grid points corresponding to the multiple sets of parameter values ​​are taken as a whole to evaluate the uncertainty of the target object at the discrete grid points; When the target object is a classified object, a voting method is used to process multiple forecast prediction results of any discrete grid points corresponding to the multiple groups of parameter values, or multiple forecast prediction results of any discrete grid points corresponding to the multiple groups of parameter values ​​are taken as a whole to evaluate the uncertainty of the target object at the discrete grid points.

8. An objective forecast product statistical interpretation device, characterized in that: include: The model and parameter determination module is used to determine the target object and determine the type of characteristic quantity, interpretation model, weight function model, distance model, and spatiotemporal neighborhood control parameters required for the forecast of the target object; A sample set construction module is used to obtain the observation data and historical feature quantities required for the forecast of the target object, and to construct a sample set; to preprocess and quality control the constructed sample set, and to store it for future use; A feature quantity acquisition module is used to acquire the feature quantity required for the forecast of the target object in the forecast period and interpolate it to the required discrete grid; A parameter value setting module, used to set multiple groups of parameter values ​​for a parameter group consisting of an interpretation model parameter, a weight function model parameter, a distance model parameter, and a spatiotemporal neighborhood control parameter; A forecast result acquisition module is used to use the set multiple groups of parameter values, the stored sample set and the acquired feature quantities required for forecasting the target object in the forecast period to forecast the target object respectively and obtain multiple forecast results corresponding to the multiple groups of parameter values; The final forecast prediction result obtaining module is used to process the multiple forecast prediction results corresponding to the multiple groups of parameter values ​​to obtain the final forecast prediction result.

9. An objective forecast product statistical interpretation device, characterized in that: include: Memory for storing computer programs; A processor, used to implement the steps of the statistical interpretation method of objective forecast products as described in any one of claims 1 to 7 when executing the computer program.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of the method for statistical interpretation of objective forecast products as claimed in any one of claims 1 to 7.

Images