Meteorological element forecasting method and device based on genetic algorithm, equipment and medium

Abstract

The invention discloses a meteorological element forecasting method and device based on a genetic algorithm, equipment and a medium. The method comprises the following steps: acquiring actual meteorological data and initial meteorological forecasting data of different climate modes; obtaining target weather forecast data of each observation station according to the initial weather forecast data and the position information of each observation station; searching an optimal solution set in the plurality of initial weight candidate solution sets by using a genetic algorithm to obtain a target weight of each climate mode; and performing weighted average according to the predicted meteorological data in each climate mode and the target weight to obtain target predicted meteorological data. According to the method, the optimal weight combination is searched in a plurality of initial weight candidate solution sets by using the genetic algorithm, so that the optimal weight combination more reasonably reflects the prediction capability of different climate modes, weighted averaging is performed through the optimal weight combination and the weather forecast data, more accurate target prediction weather data is obtained, and the prediction efficiency is improved. Therefore, the accuracy of weather forecast data is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of weather forecasting, and in particular to a weather element forecasting method and device, equipment and medium based on a genetic algorithm. Background Art

[0002] In the field of meteorological forecasting, accurate and reliable weather forecasts are becoming increasingly important as global climate change intensifies and the demand for meteorological information continues to grow. Traditional meteorological forecasting methods primarily rely on a single climate model, but these models often have limitations and cannot fully capture the complex meteorological processes and various influencing factors. Multi-climate model ensemble forecasting methods have emerged as a response to this challenge. By combining the predictions of multiple climate models, they are expected to improve the accuracy and reliability of meteorological forecasts.

[0003] However, existing multi-climate model ensemble forecasting methods usually allocate weights based on expert experience or simple statistical methods, resulting in unreasonable weight distribution, which affects the accuracy of the ensemble meteorological forecast results. Summary of the Invention

[0004] Based on this, it is necessary to propose a meteorological element forecasting method and device, equipment and medium based on genetic algorithm to address the above problems, so as to reasonably distribute the weights of various meteorological modes and thus improve the accuracy of meteorological forecast results.

[0005] To achieve the above objectives, the present application provides, in a first aspect, a meteorological element forecasting method based on a genetic algorithm, the method comprising:

[0006] Obtaining initial meteorological forecast data for a preset historical time period predicted by different climate models under a target emission scenario, as well as actual meteorological data collected by all observation sites during the historical time period under the target emission scenario, wherein the target emission scenario is any one of all emission scenarios, and the emission scenario represents the trajectory of global greenhouse gas emissions;

[0007] Performing bilinear interpolation calculation based on the initial weather forecast data and the position information of each of the observation sites to obtain target weather forecast data for each of the observation sites;

[0008] Using a genetic algorithm, the target weight of each climate model is obtained based on the target weather forecast data of all the observation stations, the actual weather data, and a plurality of initial weight candidate solution sets, wherein the initial weight candidate solution sets include the weights of all climate models;

[0009] The target predicted meteorological data is obtained by performing a weighted average based on the predicted meteorological data under each of the climate modes and the target weight.

[0010] Furthermore, the genetic algorithm is used to obtain the target weights of each climate model based on the target weather forecast data of all the observation stations, the actual weather data and a plurality of initial weight candidate solution sets, specifically including:

[0011] Randomly generate several initial weight candidate solution sets;

[0012] Performing fitness calculation based on the target weather forecast data of all the observation sites, the actual weather data, and the initial weighted candidate solution set to obtain the individual fitness corresponding to each initial weighted candidate solution set;

[0013] Based on a genetic algorithm, a target weight candidate solution set is determined according to the individual fitness corresponding to each of the initial weight candidate solution sets, and the weights in the target weight candidate solution set are used as the target weights of each climate model.

[0014] Furthermore, the method of determining the target weight candidate solution set based on the genetic algorithm according to the individual fitness corresponding to each of the initial weight candidate solution sets specifically includes:

[0015] The maximum individual fitness among the individual fitnesses corresponding to all the initial weight candidate solution sets is obtained for the i-th time, where the initial value of i is 1;

[0016] Comparing the maximum individual fitness with a preset fitness threshold to determine whether to update the initial weighted candidate solution set;

[0017] If it is determined to update the initial weight candidate solution set, crossover and mutation operations are performed on the initial weight candidate solution set to generate a new initial weight candidate solution set, and i=i+1 is set, and the step of obtaining the maximum individual fitness among the individual fitnesses corresponding to all the initial weight candidate solution sets is continued for the i-th time until i is greater than the preset number of iterations, or it is determined not to update the initial weight candidate solution set;

[0018] The initial weight candidate solution set corresponding to the maximum individual fitness at the last iteration is used as the target weight candidate solution set.

[0019] Furthermore, performing crossover and mutation operations on the initial weight candidate solution set to generate a new initial weight candidate solution set specifically includes:

[0020] Based on the sum of the individual fitness of all the initial weight candidate solution sets, the individual fitness of each of the initial weight candidate solutions is relatively adjusted to obtain the target fitness of each of the initial weight candidate solution sets;

[0021] Randomly generating a corresponding random number for each of the initial weight candidate solution sets;

[0022] Determine a first weighted candidate solution set that meets a preset transfer condition from each of the initial candidate solution sets according to a comparison between the target fitness of each of the initial candidate solution sets and the corresponding random number;

[0023] Perform crossover and mutation operations on all the first weight candidate solution sets to generate several new initial weight candidate solution sets.

[0024] Furthermore, the individual fitness of each of the initial weight candidate solutions is relatively adjusted based on the sum of the individual fitness of all the initial weight candidate solution sets, specifically including:

[0025] Continuously numbering all the initial weight candidate solution sets to obtain a unique number for each of the initial weight candidate solution sets;

[0026] Calculating the sum of the individual fitness of all the initial weight candidate solution sets to obtain the total fitness;

[0027] Obtaining the relative fitness of each of the initial weighted candidate solution sets according to the ratio of the individual fitness of each of the initial weighted candidate solution sets to the total fitness;

[0028] The target fitness of each of the initial weighted candidate solution sets is obtained by summing up the individual fitness corresponding to the current number and the relative fitness corresponding to the previous number, wherein the current number is any one of all the numbers.

[0029] Furthermore, performing crossover and mutation operations on all the first weight candidate solution sets to generate several new initial weight candidate solution sets specifically includes:

[0030] Obtaining a second weighted candidate solution set and a third weighted candidate solution set, wherein the second weighted candidate solution set and the third weighted candidate solution set are weighted candidate solution sets corresponding to the two largest individual fitnesses among all individual fitnesses corresponding to the first weighted candidate solution sets;

[0031] randomly selecting an intersection point between the second weighted candidate solution set and the third weighted candidate solution set, and reorganizing the weights before and after the intersection point between the second weighted candidate solution set and the third weighted candidate solution set to generate a plurality of fourth weighted candidate solution sets;

[0032] performing a mutation operation on the weights in the first weight candidate solution set to generate a plurality of fifth weight candidate solution sets;

[0033] The fourth weighted candidate solution set and the fifth weighted candidate solution set constitute a new initial weighted candidate solution set.

[0034] Furthermore, the individual fitness corresponding to the initial weight candidate solution set is calculated by the following formula:

[0035]

[0036] Where f is the individual fitness corresponding to the initial weighted candidate solution set, E is the sum of the root mean square errors between the target meteorological forecast data and the actual meteorological data of all the observation stations under all climate models, j is the jth observation station, L is the total number of observation stations, t is the tth forecast time point, n is the total number of forecast time points in the historical time period, i is the i-th climate model, m is the total number of climate models, and W i,j,t is the weight coefficient of the ith climate model at the jth observation station at the tth forecast time point, R i,j,t is the target meteorological forecast data of the ith climate model at the jth observation station at the tth forecast time point, r i,j,t is the actual meteorological data of the t-th forecast time point at the j-th observation station of the ith climate model.

[0037] To achieve the above-mentioned purpose, the second aspect of the present application provides a meteorological element forecasting device based on a genetic algorithm, the device comprising:

[0038] a data acquisition unit, configured to acquire initial meteorological forecast data within a preset historical time period predicted by different climate models under a target emission scenario, and actual meteorological data collected by all observation stations within the historical time period under the target emission scenario, wherein the target emission scenario is any one of all emission scenarios, and the emission scenario represents the trajectory of global greenhouse gas emissions;

[0039] a data processing unit, configured to perform bilinear interpolation calculation based on the initial weather forecast data and the position information of each of the observation sites to obtain target weather forecast data for each of the observation sites;

[0040] Using a genetic algorithm, the target weight of each climate model is obtained based on the target weather forecast data of all the observation stations, the actual weather data, and a plurality of initial weight candidate solution sets, wherein the initial weight candidate solution sets include the weights of all climate models;

[0041] The data prediction unit is used to perform weighted averaging on the predicted meteorological data under each of the climate modes and the target weight to obtain target predicted meteorological data.

[0042] To achieve the above-mentioned objectives, the third aspect of the present application provides a computer-readable storage medium storing a computer program. When the computer program is executed by a processor, the processor executes the steps of the method described in the first aspect.

[0043] To achieve the above-mentioned objectives, the fourth aspect of the present application provides a computer device, comprising a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor executes the steps of the method described in the first aspect.

[0044] The embodiments of the present invention have the following beneficial effects:

[0045] An embodiment of the present invention proposes a meteorological element forecasting method based on a genetic algorithm, which includes: obtaining initial meteorological forecast data within a preset historical time period predicted by different climate models under a target emission scenario, and actual meteorological data collected by all observation stations within the historical time period under the target emission scenario, wherein the target emission scenario is any one of all emission scenarios, and the emission scenario represents the trajectory of global greenhouse gas emissions; performing bilinear interpolation calculation based on the initial meteorological forecast data and the location information of each observation station to obtain the target meteorological forecast data of each observation station; utilizing a genetic algorithm to obtain the target weight of each climate model based on the target meteorological forecast data of all observation stations, the actual meteorological data and a number of initial weight candidate solution sets, wherein the initial weight candidate solution set contains the weights of all climate models; performing weighted averaging based on the predicted meteorological data and the target weight under each climate model to obtain the target predicted meteorological data. The present invention uses a genetic algorithm to search for the optimal weight combination among a large number of initial weight candidate solutions, so that the weights of each climate model more reasonably reflect the prediction ability of the climate model under different meteorological elements and emission scenarios, and performs weighted averaging on the meteorological forecast data under each climate model through the optimal weight combination to obtain more accurate target predicted meteorological data, thereby improving the accuracy of meteorological forecast data. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. 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 any creative work.

[0047] in:

[0048] Figure 1 Schematic diagram of a flow chart of a meteorological element forecasting method based on a genetic algorithm in an embodiment of the present invention;

[0049] Figure 2 A Taylor diagram showing the predicted results of meteorological elements of different climate models and the observed values ​​of the observation stations in an embodiment of the present invention;

[0050] Figure 34 is a structural block diagram of a meteorological element forecasting device based on a genetic algorithm in an embodiment of the present invention;

[0051] Figure 4 2 is a diagram showing the internal structure of a computer device in an embodiment of the present invention. DETAILED DESCRIPTION

[0052] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0053] The values ​​of meteorological elements predicted by different climate models vary. To combine the prediction performance of different climate models, the prediction results from different climate models can be combined. Currently, most methods for combining prediction results from multiple climate models use the equal-weighted arithmetic ensemble average method. However, this method does not fully consider the performance differences between different climate models and the influence of temporal and spatial scales, resulting in large errors and insufficient accuracy in the resulting prediction data.

[0054] Based on this, an embodiment of the present invention proposes a meteorological element forecasting method based on a genetic algorithm, which can be found in Figure 1 , Figure 1 : is a flow chart of a meteorological element forecasting method based on a genetic algorithm in an embodiment of the present invention, the method comprising:

[0055] Step 100: Obtain initial meteorological forecast data within a preset historical time period predicted by different climate models under a target emission scenario, as well as actual meteorological data collected by all observation sites within the historical time period under the target emission scenario, wherein the target emission scenario is any one of all emission scenarios, and the emission scenario represents the trajectory of global greenhouse gas emissions.

[0056] In this embodiment, the target emission scenario can be any one of all emission scenarios. The emission scenario represents the trajectory of global greenhouse gas emissions, such as the Shared Socioeconomic Pathway (SSPs) scenario. The SSPs scenario is a tool created by the Intergovernmental Panel on Climate Change (IPCC) to assess future socioeconomic development paths and their interactions with climate change. It aims to provide basic data for research on climate change impacts, adaptation, and mitigation. The same climate model has different values ​​for the same meteorological element under different emission scenarios. Taking temperature as an example of the meteorological element, under high emission scenarios, the temperature value predicted by the climate model will be higher, while under low emission scenarios, the temperature value predicted by the climate model will be lower. Therefore, it is necessary to set a target emission scenario according to the actual application scenario to simulate the current emission scenario. For example, the target emission scenario can be the SSP-245 emission scenario. The SSP-245 emission scenario is an "intermediate path" that combines moderate greenhouse gas emissions and socioeconomic development paths and belongs to a medium emission scenario.

[0057] In one embodiment, the ACCMIP6 library can be used to access and download data from the CMIP6 database to obtain initial meteorological forecast data predicted by different climate models within a historical time period under the target emission scenario. The CMIP6 (Sixth Coupled Model Intercomparison Project) database contains simulation results of climate predictions under different greenhouse gas emission pathways from multiple climate models around the world. Table 1 shows the climate model information of CMIP6. The database contains historical records and future forecasts of multiple variables, such as temperature, precipitation, sea level, carbon cycle, and other climate elements. ACCMIP6 is a Python-based toolkit specifically for accessing and downloading data from the CMIP6 database.

[0058] Table 1 CMIP6 climate model information

[0059]

[0060]

[0061] Each climate model selected the same numbered simulation members (e.g., member r1i1p1f1 for each model) to ensure consistency within each model. In reality, each model's specific simulation run (realization_index), initial conditions (initialization_index), physical parameterization scheme (physics_index), and external forcing conditions (forcing_index) are independently set by the model development agency. Consequently, meteorological elements can differ significantly across models. r1i1p1f1 is a variant label used in the CMIP6 database to identify simulation members. The r, i, p, and f positions represent the simulation member number (realization), initialization method (initialization), physics configuration version (physics), and external forcing scheme (forcing), respectively. Each climate model selected the first member in its output data (i.e., the r1i1p1f1 combination, representing the first realization, first initialization, first set of physics configurations, and first set of external forcings), ensuring that the corresponding first simulation member was used for comparative analysis within each climate model whenever possible. This selection method ensures consistent selection criteria within each climate model. Even if each model selects members with the same numbering format (such as r1i1p1f1), it is to reduce the differences caused by the settings of each climate model.

[0062] When downloading data from various climate models, due to the length of the data, different time periods are typically stored in separate files. Therefore, the weather forecast data from the climate models must be processed using CDO (Climate Data Operation) software to facilitate subsequent experiments and analysis. For example, for the period 1985-2014, some climate models store the experimental data in a single file, while others store it in multiple files. Therefore, the data must be merged to facilitate subsequent processing and analysis.

[0063] Global climate model data is gridded data in the NC format. Each NC file contains three dimensions: time, and the second and third dimensions represent spatial information, namely longitude and latitude. Since the longitude and latitude information is the same for the same climate model, merging the first dimension (time) across different files can yield climate model grid data for a long time dimension (i.e., a historical time period).

[0064] In this embodiment, the preset historical time period is any time period in the past, for example, the period from 2018 to 2020. The initial weather forecast data may include climate elements such as temperature, precipitation, sea level, and carbon cycle.

[0065] An observation station is a fixed location used to collect current weather data. Equipped with various weather observation instruments and equipment, it monitors and records various elements of atmospheric conditions. Therefore, each region has multiple observation stations to collect actual weather data from different locations.

[0066] The embodiment of the present invention effectively integrates the prediction results of multiple models by obtaining initial forecast data from different climate models. For example, the CMIP6 database contains climate models developed by 11 different countries / institutions, including ACCESS-CM2 (Australia), BCC-CSM2-MR (China), and CESM2-WACCM (USA) (as shown in Table 1). This data collection strategy can fully capture the unique prediction capabilities of different climate models for meteorological elements, avoid the inherent bias of a single model, and provide the original data basis for subsequent weighted fusion. In addition, by setting specific emission scenarios, the predicted data is strongly correlated with specific socio-economic development paths and greenhouse gas emission trajectories. This scenario constraint ensures that the prediction results are mapped to actual decision-making needs, thereby improving the scenario adaptability of meteorological services.

[0067] Step 200: Perform bilinear interpolation calculation based on the initial weather forecast data and the location information of each observation station to obtain the target weather forecast data of each observation station.

[0068] Since the initial meteorological data predicted by different climate models are grid data, that is, meteorological data are contained at each grid point, their spatial resolution is different from the spatial resolution of the actual meteorological data measured at the observation site. Therefore, in order to refine the two meteorological data and make the two meteorological data at the same spatial resolution, the initial meteorological data under different climate models need to be interpolated to the observation site using the bilinear interpolation method, so that the interpolated target meteorological forecast data and the actual meteorological data are at the same spatial resolution, so as to improve the accuracy of subsequent analysis.

[0069] In one embodiment, step 200, performing bilinear interpolation calculation based on the initial weather forecast data and the location information of each observation station to obtain target weather forecast data for each observation station, specifically includes:

[0070] Step 210: Get the coordinates of the observation site.

[0071] In this embodiment, a point is preset as the coordinate origin, and the coordinate information of all observation sites relative to the coordinate origin is obtained. The coordinates of the observation sites are (x, y).

[0072] Step 220: Based on the coordinates of the observation site, search within the grid corresponding to the target climate model to obtain the minimum grid where the observation site is located, and obtain the grid point coordinates of the minimum grid, wherein the target climate model is any one of all climate models.

[0073] In this embodiment, the observation station is not necessarily located completely on the grid point. In order to obtain more accurate target meteorological data of the observation station, the grid corresponding to each climate mode can be searched separately to find the minimum grid position of each observation station in each climate mode. It can be understood that each minimum grid has four grid points. Furthermore, the grid coordinates of the four grid points relative to the coordinate origin can be obtained. The grid coordinates of the minimum grid are Q 11 (x1, y1), Q 12 (x1, y2), Q 21 (x2, y1) and Q 22 (x2, y2).

[0074] Step 230: Obtain grid point weather forecast data for the smallest grid point based on the initial weather forecast data for all grid points of the target climate model within the historical time period and the grid point coordinates of the smallest grid.

[0075] In this embodiment, after the coordinates of the grid points of the minimum grid are determined, a search is performed in the initial weather forecast data of all grid points of each climate model to obtain the grid point weather forecast data of the grid points of the minimum grid within the historical time period.

[0076] Step 240: Perform linear interpolation based on the grid point coordinates of the minimum grid, the grid point weather forecast data, and the coordinates of the observation station to obtain the target weather forecast data of the observation station.

[0077] In this embodiment, linear interpolation is first performed on two x-direction line segments to obtain the value of the midpoint:

[0078]

[0079] Among them, f 11 、f 21 、f 12 and f 22 It is the grid point weather forecast data on the grid point coordinates of the smallest grid.

[0080] Secondly, perform linear interpolation on the two intermediate values ​​obtained in the x direction to obtain the final result:

[0081]

[0082] Substituting formulas (1) and (2) into formula (3), we get formula (4). The meteorological forecast data of the observation station is calculated by formula (4):

[0083]

[0084] Where (x, y) is the coordinate of the observation station, f(x, y) is the target weather forecast data of the observation station, (x1, y1), (x1, y2), (x2, y1) and (x2, y2) are the grid coordinates of the minimum grid, and f 11 、f 21 、f 12 and f 22 It is the grid point weather forecast data on the grid point coordinates of the smallest grid.

[0085] The embodiment of the present invention uses a linear interpolation method to interpolate the grid data of different climate models to the location of the observation site, so that the weather forecast data and the actual observation data have the same spatial resolution, which helps to improve the comparability and consistency of the data and further improves the accuracy of the target predicted meteorological data.

[0086] Step 300: Using a genetic algorithm, the target weights of each climate model are obtained based on the target weather forecast data of all observation sites, the actual weather data and several initial weight candidate solution sets, wherein the initial weight candidate solution sets contain the weights of all climate models.

[0087] In this embodiment, different climate models have different prediction performances for each climate element. Figure 2 , Figure 2 is a Taylor diagram of meteorological element prediction results of different climate models and actual observation values ​​of observation stations in an embodiment of the present invention, Figure 2 This represents the spatial distribution of the climate model's prediction performance at each observation site. The lower the error, the higher the spatial correlation, and the closer the climate model points will be to the reference value (ref). Figure 2 It can be found that all climate models predict temperature and air pressure well, with small differences in simulation performance between models. However, there are significant differences in the simulation of near-surface wind speed between different climate models. Therefore, to obtain more accurate prediction results, each climate model can be weighted based on its performance in predicting different meteorological elements. A weighted calculation based on the weights of each model and the prediction results can be performed to obtain the optimal prediction results, thereby improving the accuracy of the prediction results.

[0088] A genetic algorithm (GA) is a search algorithm that simulates the natural selection and genetics mechanisms of Darwin's theory of biological evolution, searching for the optimal solution by simulating the natural evolutionary process. In a genetic algorithm, all potential solution sets are called populations, each potential solution set is called an individual (also called a chromosome), and the actual parameters or variables corresponding to each individual are called genes. In this embodiment, all weighted candidate solution sets are called populations, each weighted candidate solution set is called an individual (chromosome), and the weights contained in each individual (chromosome) are genes.

[0089] In this embodiment, several initial weight candidate solution sets are pre-set, each containing weights corresponding to each climate model. A genetic algorithm is used to determine the optimal weight candidate solution set from these initial weight candidate solution sets. The weights corresponding to each climate model in the optimal weight candidate solution set are used as the target weights for each climate model. Calculations are performed using the optimal weights for each climate model to obtain the optimal forecasted meteorological data.

[0090] Specifically, based on the genetic algorithm, calculations are performed based on each initial weight candidate solution set and the target meteorological forecast data to obtain the weighted average meteorological forecast data corresponding to each initial weight candidate solution set; then the error between the weighted average meteorological forecast data and the target meteorological forecast data is calculated, and the weight candidate solution set with the smallest error is taken as the optimal weight candidate solution set.

[0091] Step 400: Perform weighted averaging based on the predicted meteorological data under each climate model and the target weight to obtain target predicted meteorological data.

[0092] After determining the optimal weighted candidate solution set, a weighted average is performed on the target weight in the optimal weighted candidate solution set and the predicted meteorological data to obtain the revised target predicted meteorological data.

[0093] In one embodiment, the target forecasted meteorological data is calculated using the following formula:

[0094]

[0095] Among them, Ensemble(t) is the target forecast meteorological data, M i (t) is the predicted meteorological data of the i-th climate model at time t, ω i is the weight of the i-th climate model, and N is the total number of climate models.

[0096] This embodiment of the present invention integrates the forecast results of different climate models through a weighted averaging mechanism, leveraging the differentiated forecast capabilities of each model for specific meteorological elements. Using target weights generated by a genetic algorithm, instead of traditional equal weighting or empirical weighting methods, it significantly reduces systematic bias and ultimately achieves more accurate target forecast data.

[0097] In one embodiment of the present invention, step 300 utilizes a genetic algorithm to obtain target weights for each climate model based on target weather forecast data, actual weather data, and a plurality of initial weight candidate solution sets for all observation stations. Specifically, the method includes:

[0098] Step 310: Randomly generate several initial weight candidate solution sets.

[0099] In this embodiment, a large number of different initial weight candidate solution sets are randomly generated, wherein the initial weight candidate solution set contains the weight corresponding to each climate mode, and the sum of all weights in each initial weight candidate solution set is 1.

[0100] Step 320: Calculate the fitness based on the target meteorological forecast data, actual meteorological data and initial weight candidate solution sets of all observation stations to obtain the individual fitness corresponding to each initial weight candidate solution set.

[0101] In this embodiment, the individual fitness corresponding to the initial weight candidate solution set is calculated by the following formula:

[0102]

[0103] Where f is the individual fitness corresponding to the initial weighted candidate solution set, E is the sum of the root mean square error between the target meteorological forecast data and the actual meteorological data of all observation stations under all climate models, j is the jth observation station, L is the total number of observation stations, t is the tth forecast time point, n is the total number of forecast time points in the historical time period, i is the i-th climate model, m is the total number of climate models, and W is the total number of climate models. i,j,t is the weight coefficient of the ith climate model at the jth observation station at the tth forecast time point, R i,j,t is the target meteorological forecast data of the ith climate model at the jth observation station at the tth forecast time point, r i,j,t is the actual meteorological data of the t-th forecast time point at the j-th observation station of the ith climate model.

[0104] In this embodiment, the area requiring weather forecast is regarded as the area to be forecasted, so that the weather forecast data of the area to be forecasted can be corrected. There are multiple observation sites in the area to be forecasted, and the total number of observation sites in the area to be forecasted is the total number of observation sites.

[0105] Step 330: Based on the genetic algorithm, determine the target weight candidate solution set according to the individual fitness corresponding to each initial weight candidate solution set, and use the weights in the target weight candidate solution set as the target weights of each climate model.

[0106] In this embodiment, the individual fitness of each initial weight candidate solution set is determined according to the individual fitness calculation function in Step 320, and then based on the genetic algorithm, the optimal target weight candidate solution set is determined according to the individual fitness corresponding to each initial weight candidate solution set, and the weights in the target weight candidate solution set are used as the target weights of each climate model.

[0107] In one embodiment, Step 330, based on a genetic algorithm, determines a target weight candidate solution set according to the individual fitness corresponding to each initial weight candidate solution set, specifically including:

[0108] Step 331, for the i-th time, obtain the maximum individual fitness among the individual fitnesses corresponding to all the initial weight candidate solution sets, where the initial value of i is 1.

[0109] In this embodiment, the greater the individual fitness, the smaller the error between the weather forecast data calculated based on the initial weighted candidate solution set and the actual weather data. Conversely, the smaller the individual fitness, the larger the error between the weather forecast data calculated based on the initial weighted candidate solution set and the actual weather data. Therefore, to obtain the most accurate target predicted weather data, it is necessary to select the initial weighted candidate solution set with the smallest error, that is, the initial weighted candidate solution set corresponding to the maximum individual fitness.

[0110] Step 332: Compare the maximum individual fitness with the preset fitness threshold to determine whether to update the initial weighted candidate solution set.

[0111] In this embodiment, when the maximum individual fitness is greater than the preset fitness threshold, it means that the error calculated based on the weighted candidate solution set corresponding to the maximum individual fitness is very small and can be ignored. At this time, the weights in the weighted candidate solution set corresponding to the maximum individual fitness can be used as the target weights of each climate model; when the maximum individual fitness is not greater than the preset fitness threshold, it means that the error calculated based on the weighted candidate solution set corresponding to the maximum individual fitness is still large, and the obtained prediction result is inaccurate. At this time, it is necessary to update the initial weighted candidate solution set, generate new and different initial weighted candidate solution sets, and select the optimal weighted candidate solution set from the new initial weighted candidate solution sets.

[0112] Step 333: If it is determined to update the initial weight candidate solution set, perform crossover and mutation operations on the initial weight candidate solution set to generate a new initial weight candidate solution set, and let i = i + 1, and continue to execute the step of obtaining the maximum individual fitness among the individual fitnesses corresponding to all initial weight candidate solution sets for the i-th time, until i is greater than the preset number of iterations, or it is determined not to update the initial weight candidate solution set.

[0113] In this embodiment, when the maximum individual fitness is not greater than a preset fitness threshold, the initial weight candidate solution set is determined to be updated. The initial weight candidate solution set is then used as a chromosome, and the weights in the initial weight candidate solution set are used as genes. Crossover and mutation operations are performed on the weights in the initial weight candidate solution set to generate a new initial weight candidate solution set. Steps 331-333 are then repeated based on the new initial weight candidate solution set, with the count i incremented by 1 each time the count is repeated. This continues until the maximum individual fitness exceeds the preset fitness threshold, or the count i exceeds a preset number of iterations.

[0114] Crossover and mutation operations can introduce new weight combinations and explore more optimal solutions within the search space of candidate weight solutions. The crossover operation combines the advantages of two parent weight combinations to generate a child weight combination with potentially greater adaptability. Mutation operations randomly alter weight values ​​to prevent the algorithm from becoming trapped in local optima and introduce new gene combinations, further improving forecast accuracy. This dynamic optimization process effectively reduces the error between weather forecast data and actual observations, ensuring that the final target weights more accurately reflect the performance differences between climate models in predicting different meteorological elements, thereby improving the accuracy of the target forecasted weather data.

[0115] In one embodiment, in Step 333, crossover and mutation operations are performed on the initial weight candidate solution set to generate a new initial weight candidate solution set, specifically including:

[0116] Step 3331: Based on the sum of the individual fitness of all initial weight candidate solution sets, the individual fitness of each initial weight candidate solution is relatively adjusted to obtain the target fitness of each initial weight candidate solution set.

[0117] In this embodiment, the optimal chromosome, that is, the optimal initial weight candidate solution set, is selected through individual fitness evaluation and passed to the next generation, and a new weight candidate solution set with better quality is generated based on the optimal initial weight candidate solution set.

[0118] Specifically, the individual fitness of each initial weight candidate solution is relatively adjusted according to the sum of the individual fitness of all initial weight candidate solution sets, and the target fitness of each initial weight candidate solution set is obtained, so as to select the optimal initial weight candidate solution set according to the target fitness evaluation.

[0119] By making relative adjustments to the sum of individual fitnesses, we can more accurately assess the quality of each candidate solution set with initial weights, thereby more efficiently selecting the optimal solution set. This relative fitness adjustment process makes the algorithm robust to individual exceptionally high or low fitness values. It prevents excessive bias in one direction due to the occasional high fitness of a solution set, nor does it hinder the overall optimization process due to individual low fitness solutions. This stability helps the algorithm consistently produce high-quality predictions across different datasets and initial conditions.

[0120] In one embodiment, step 3331 performs relative fitness adjustment on the individual fitness of each initial weight candidate solution based on the sum of the individual fitness of all initial weight candidate solution sets, specifically including:

[0121] A1. Serially number all initial weight candidate solution sets to obtain a unique number for each initial weight candidate solution set.

[0122] In this embodiment, each initial weight candidate solution set is numbered consecutively, and the number of each initial weight candidate solution set is unique.

[0123] Assigning a unique number to each candidate solution set with initial weights enables more efficient management and manipulation of a large number of candidate solutions. During the iterative process of a genetic algorithm, numerous solution sets must be tracked, selected, crossover, and mutated. Unique numbers allow for quick and accurate identification and location of each solution set, avoiding confusion and errors and improving the algorithm's efficiency and accuracy. Numbering candidate solutions facilitates subsequent selection according to specific rules. For example, in a roulette wheel selection game, the probability of each solution set being selected can be determined based on the number. This allows for fair and efficient selection of superior solutions for subsequent crossover and mutation operations, improving the algorithm's ability to find the optimal solution.

[0124] A2. Calculate the sum of the individual fitness of all initial weight candidate solution sets to obtain the total fitness.

[0125] In this embodiment, the individual fitnesses of all the initial weight candidate solution sets are summed up to obtain the total fitness of all the initial weight candidate solution sets.

[0126] A3. According to the ratio of the individual fitness of each initial weight candidate solution set to the total fitness, the relative fitness of each initial weight candidate solution set is obtained.

[0127] In this embodiment, the relative fitness of each initial weight candidate solution set is calculated by the following formula:

[0128]

[0129] Where R(ε) represents the relative fitness of the initial weight candidate solution set numbered ε, f(ε) represents the individual fitness of the initial weight candidate solution set numbered ε, and M represents the total number of initial weight candidate solution sets.

[0130] A4. Calculate the target fitness of each initial weight candidate solution set by summing the individual fitness corresponding to the current number and the relative fitness corresponding to the previous number, where the current number is any one of all the numbers.

[0131] In this embodiment, the target fitness of each initial weight candidate solution set is calculated by the following formula:

[0132] C(ε=C(ε-1)+R(ε)

[0133] In the formula, C(ε) represents the target fitness of the initial weight candidate solution set numbered ε, C(ε-1) represents the target fitness of the initial weight candidate solution set numbered ε-1, and R(ε) represents the relative fitness of the initial weight candidate solution set numbered ε. It can be understood that C(0)=R(0).

[0134] The summation calculation method used in this embodiment of the present invention integrates information about individual fitness and relative fitness, ensuring that the target fitness reflects not only the fitness of the current solution set itself but also its relative position within the population. This helps more accurately assess the merits of each initial weighted candidate solution set, improving the accuracy of the selection operation and increasing the probability that excellent solutions will be selected for subsequent crossover and mutation operations, thereby accelerating the algorithm's convergence to the optimal solution.

[0135] Step 3332: Generate a corresponding random number for each initial weight candidate solution set.

[0136] In this embodiment, a random number is randomly generated for each initial weight candidate solution set.

[0137] Step 3333: Compare the target fitness of each initial candidate solution set with the corresponding random number, and determine the first weighted candidate solution set that meets the preset transfer condition from each initial candidate solution set.

[0138] In this embodiment, the preset transfer condition is that the random number is between the target fitness of the current initial candidate solution set and the target fitness of the previous initial candidate solution set. All initial weighted candidate solution sets are traversed to determine whether each initial weighted candidate solution set is the first weighted candidate solution set.

[0139] Specifically, if the random number corresponding to the current initial candidate solution set is between the target fitness of the current initial candidate solution set and the target fitness of the previous initial candidate solution set, the current initial candidate solution set is used as the first weighted candidate solution set. This can be expressed by the following formula:

[0140] C(ε-1)≤p≤C(ε)

[0141] Where C(ε) represents the target fitness of the initial weighted candidate solution set numbered ε, C(ε-1) represents the target fitness of the initial weighted candidate solution set numbered ε-1, and p represents the random number corresponding to the initial weighted candidate solution set numbered ε. If the random number corresponding to the initial weighted candidate solution set numbered ε satisfies the above formula, then the initial weighted candidate solution set numbered ε is the first weighted candidate solution set.

[0142] By comparing random numbers with target fitness, a random factor is introduced. This approach simulates the randomness of natural selection, making the selection process more scientific and rational, avoiding the premature convergence problem that can occur with deterministic selection, increasing the algorithm's exploration capabilities, and helping to find more optimal solutions. Preset transfer conditions ensure that only solution sets whose target fitness matches the random number are selected, thus ensuring that the solution sets selected for subsequent operations are of high quality, which helps improve the accuracy of the final target forecast weather data.

[0143] Step 3334: Perform crossover and mutation operations on all first weight candidate solution sets to generate several new initial weight candidate solution sets.

[0144] Specifically, crossover and mutation operations are performed on all first weight candidate solution sets to generate several new initial weight candidate solution sets, specifically including:

[0145] B1. Obtain a second weighted candidate solution set and a third weighted candidate solution set, wherein the second weighted candidate solution set and the third weighted candidate solution set are weighted candidate solution sets corresponding to the two largest individual fitnesses among the individual fitnesses corresponding to all first weighted candidate solution sets.

[0146] In this embodiment, the two individual fitnesses with the largest individual fitnesses in the first weighted candidate solution set are selected, and the initial weighted candidate solution sets corresponding to the two largest individual fitnesses are respectively used as the second weighted candidate solution set and the third weighted candidate solution set.

[0147] B2. Randomly select an intersection point from the second weighted candidate solution set and the third weighted candidate solution set, reorganize the weights before and after the intersection point of the second weighted candidate solution set and the third weighted candidate solution set, and generate several fourth weighted candidate solution sets.

[0148] Since the individual fitness of the second weight candidate solution set and the third weight candidate solution set is the largest, it means that among all the first weight candidate solution sets, the second weight candidate solution set and the third weight candidate solution set are closest to the optimal weight candidate solution set. Therefore, the quality of the new fourth weight candidate solution set generated by the cross operation based on the second weight candidate solution set and the third weight candidate solution set is better and closer to the optimal weight candidate solution set.

[0149] Specifically, a crossover operation involves randomly selecting crossover points between chromosomes and exchanging genes to produce offspring, hoping that the newly generated chromosomes will combine the strengths of their parents. In this embodiment, the genes in each weighted candidate solution set specifically correspond to the weight values ​​of each climate model. Methods such as single-point crossover and two-point crossover can be used to recombine the weights of the second and third weighted candidate solution sets to produce a more optimal fourth weighted candidate solution set.

[0150] For example, two intersection points are randomly selected on the second weight candidate solution set and the third weight candidate solution set, and the model weights between the two points are exchanged to obtain a new fourth weight candidate solution set.

[0151] In this embodiment, the second and third weighted candidate solution sets are the two solution sets with the highest fitness in the current population, representing the current optimal weight combination scheme. By randomly selecting intersections between them and reorganizing the weights, the advantageous features of the two high-quality parent solution sets can be recombined. For example, if the second weighted candidate solution set performs well in the weight distribution of certain climate modes, and the third weighted candidate solution set is more reasonable in the weight setting of other climate modes, the fourth weighted candidate solution set generated after the cross-recombination is likely to inherit the advantages of these two parents, forming a more optimal weight distribution scheme, thereby more accurately combining the prediction performance of different climate modes and improving the accuracy of meteorological element forecasts.

[0152] B3. Perform a mutation operation on the weights in the first weighted candidate solution set to generate several fifth weighted candidate solution sets.

[0153] Mutation involves replacing certain gene values ​​in an individual's code string with other gene values ​​based on the mutation probability, thereby forming a new individual. In this embodiment, balanced mutation is used. The gene mutation method involves traversing all weight values ​​in the first set of candidate weight solutions. For each weight value found, a random number is generated. The random number corresponding to each weight value is compared with the mutation probability. If the random number corresponding to a weight value is less than the mutation probability, a number between 0 and 1 is generated using a preset random function to replace the original corresponding weight value, thereby generating a new fifth set of candidate weight solutions.

[0154] By randomly changing weights, mutation introduces new gene combinations, breaking the genetic limitations of the existing population. This helps prevent the algorithm from becoming trapped in local optima and premature convergence. Mutation and crossover work together to promote the dynamic equilibrium of the population. Crossover primarily generates new solutions by combining the best features of existing solutions, while mutation, to a certain extent, overcomes the limitations of these combinations and prevents the population from becoming prematurely homogenized. This dynamic balance helps the algorithm maintain a good balance between exploration and exploitation, enabling it to effectively utilize existing information for optimization while maintaining a certain level of exploration capability, thereby improving its overall performance and efficiency.

[0155] B4, the fourth weighted candidate solution set and the fifth weighted candidate solution set constitute a new initial weighted candidate solution set.

[0156] In this embodiment, the updated initial weighted candidate solution set consists of the fourth weighted candidate solution set and the fifth weighted candidate solution set.

[0157] Step 334: The initial weight candidate solution set corresponding to the maximum individual fitness at the last iteration is used as the target weight candidate solution set.

[0158] In this embodiment, if the maximum individual fitness at the qth iteration is greater than the preset fitness threshold, or q is greater than the preset number of iterations, the initial weight candidate solution set corresponding to the maximum individual fitness at the qth iteration is used as the target weight candidate solution set.

[0159] In one embodiment of the present invention, the genetic algorithm parameters can be set as follows: the total number of chromosomes (the total number of the initial weighted candidate solution set) M is generally 20 to 100; the number of genetic operation termination evolution iterations is generally 100 to 500; the crossover probability is generally 0.4 to 0.9; and the mutation probability is generally 0.001 to 0.1.

[0160] One embodiment of the present invention also proposes a weather element forecasting device based on genetic algorithm, see Figure 3 , Figure 3 : is a structural block diagram of a meteorological element forecasting device based on a genetic algorithm in an embodiment of the present invention, the device includes:

[0161] The data acquisition unit 301 is used to obtain the initial meteorological forecast data within a preset historical time period predicted by different climate models under the target emission scenario, as well as the actual meteorological data collected by all observation sites within the historical time period under the target emission scenario, wherein the target emission scenario is any one of all emission scenarios, and the emission scenario represents the trajectory of global greenhouse gas emissions.

[0162] The data processing unit 302 is configured to perform bilinear interpolation calculations based on the initial weather forecast data and the location information of each observation station to obtain target weather forecast data for each observation station.

[0163] Using a genetic algorithm, the target weights of each climate model are obtained based on the target meteorological forecast data of all observation stations, actual meteorological data and several initial weight candidate solution sets, among which the initial weight candidate solution sets contain the weights of all climate models.

[0164] The data prediction unit 303 is used to perform weighted averaging based on the predicted meteorological data under each climate mode and the target weight to obtain target predicted meteorological data.

[0165] The meteorological element forecasting device based on genetic algorithm proposed in the embodiment of the present invention uses genetic algorithm to search for the optimal weight combination among many initial weight candidate solutions, so that the weight of each climate model more reasonably reflects the prediction ability of the climate model under different meteorological elements and emission scenarios, and performs weighted averaging on the meteorological forecast data under each climate model through the optimal weight combination to obtain more accurate target predicted meteorological data, thereby improving the accuracy of the meteorological forecast data.

[0166] Figure 4 FIG1 shows the internal structure of a computer device in one embodiment of the present invention. The computer device can be a terminal or a system. Figure 4 As shown, the computer device includes a processor, a memory, and a network interface connected via a system bus. The memory includes a non-volatile storage medium and an internal memory. The non-volatile storage medium of the computer device stores an operating system and may also store a computer program. When the computer program is executed by the processor, the processor can implement the various steps in the above method embodiment. The internal memory may also store a computer program. When the computer program is executed by the processor, the processor can implement the various steps in the above method embodiment. It will be understood by those skilled in the art that Figure 4 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than shown in the figure, or combine certain components, or have a different component arrangement.

[0167] In one embodiment, a computer device is provided, including a memory and a processor. The memory stores a computer program. When the computer program is executed by the processor, the processor executes each step in the above method embodiment.

[0168] In one embodiment, a computer-readable storage medium is provided, which stores a computer program. When the computer program is executed by a processor, the processor executes the steps in the above method embodiment.

[0169] Those skilled in the art will appreciate that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through a computer program. The program can be stored in a non-volatile computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0170] The technical features of the above embodiments can be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0171] The above-described embodiments merely represent several implementation methods of the present application. While the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the present invention. It should be noted that a person of ordinary skill in the art may make various modifications and improvements without departing from the spirit of the present invention, all of which fall within the scope of protection of the present invention. Therefore, the scope of protection of the present invention shall be determined by the appended claims.

Claims

1. A meteorological element forecasting method based on genetic algorithm, characterized in that: The method comprises: Obtaining initial meteorological forecast data for a preset historical time period predicted by different climate models under a target emission scenario, as well as actual meteorological data collected by all observation sites during the historical time period under the target emission scenario, wherein the target emission scenario is any one of all emission scenarios, and the emission scenario represents the trajectory of global greenhouse gas emissions; Performing bilinear interpolation calculation based on the initial weather forecast data and the position information of each of the observation sites to obtain target weather forecast data for each of the observation sites; Using a genetic algorithm, the target weight of each climate model is obtained based on the target weather forecast data of all the observation stations, the actual weather data, and a plurality of initial weight candidate solution sets, wherein the initial weight candidate solution sets include the weights of all climate models; The target predicted meteorological data is obtained by performing a weighted average based on the predicted meteorological data under each of the climate modes and the target weight.

2. The method according to claim 1, wherein The genetic algorithm is used to obtain the target weights of each climate model based on the target weather forecast data of all the observation stations, the actual weather data and a plurality of initial weight candidate solution sets, specifically including: Randomly generate several initial weight candidate solution sets; Performing fitness calculation based on the target weather forecast data of all the observation sites, the actual weather data, and the initial weighted candidate solution set to obtain the individual fitness corresponding to each of the initial weighted candidate solution sets; Based on a genetic algorithm, a target weight candidate solution set is determined according to the individual fitness corresponding to each of the initial weight candidate solution sets, and the weights in the target weight candidate solution set are used as the target weights of each climate model.

3. The method according to claim 2, wherein The method of determining the target weight candidate solution set based on the genetic algorithm according to the individual fitness corresponding to each of the initial weight candidate solution sets specifically includes: The maximum individual fitness among the individual fitnesses corresponding to all the initial weight candidate solution sets is obtained for the i-th time, where the initial value of i is 1; Comparing the maximum individual fitness with a preset fitness threshold to determine whether to update the initial weighted candidate solution set; If it is determined to update the initial weight candidate solution set, crossover and mutation operations are performed on the initial weight candidate solution set to generate a new initial weight candidate solution set, and i=i+1 is set, and the step of obtaining the maximum individual fitness among the individual fitnesses corresponding to all the initial weight candidate solution sets is continued for the i-th time until i is greater than the preset number of iterations, or it is determined not to update the initial weight candidate solution set; The initial weight candidate solution set corresponding to the maximum individual fitness at the last iteration is used as the target weight candidate solution set.

4. The method according to claim 3, wherein The performing crossover and mutation operations on the initial weight candidate solution set to generate a new initial weight candidate solution set specifically includes: Based on the sum of the individual fitness of all the initial weight candidate solution sets, the individual fitness of each of the initial weight candidate solutions is relatively adjusted to obtain the target fitness of each of the initial weight candidate solution sets; Randomly generating a corresponding random number for each of the initial weight candidate solution sets; Determine a first weighted candidate solution set that meets a preset transfer condition from each of the initial candidate solution sets according to a comparison between the target fitness of each of the initial candidate solution sets and the corresponding random number; Perform crossover and mutation operations on all the first weight candidate solution sets to generate several new initial weight candidate solution sets.

5. The method according to claim 4, wherein The relative fitness adjustment of the individual fitness of each of the initial weight candidate solutions based on the sum of the individual fitness of all the initial weight candidate solution sets specifically includes: Continuously numbering all the initial weight candidate solution sets to obtain a unique number for each of the initial weight candidate solution sets; Calculating the sum of the individual fitness of all the initial weight candidate solution sets to obtain the total fitness; Obtaining the relative fitness of each of the initial weighted candidate solution sets according to the ratio of the individual fitness of each of the initial weighted candidate solution sets to the total fitness; The target fitness of each of the initial weighted candidate solution sets is obtained by summing up the individual fitness corresponding to the current number and the relative fitness corresponding to the previous number, wherein the current number is any one of all the numbers.

6. The method according to claim 4, wherein The performing crossover and mutation operations on all the first weight candidate solution sets to generate a plurality of new initial weight candidate solution sets specifically includes: Obtaining a second weighted candidate solution set and a third weighted candidate solution set, wherein the second weighted candidate solution set and the third weighted candidate solution set are weighted candidate solution sets corresponding to the two largest individual fitnesses among all individual fitnesses corresponding to the first weighted candidate solution sets; randomly selecting an intersection point between the second weighted candidate solution set and the third weighted candidate solution set, and reorganizing the weights before and after the intersection point between the second weighted candidate solution set and the third weighted candidate solution set to generate a plurality of fourth weighted candidate solution sets; performing a mutation operation on the weights in the first weight candidate solution set to generate a plurality of fifth weight candidate solution sets; The fourth weighted candidate solution set and the fifth weighted candidate solution set constitute a new initial weighted candidate solution set.

7. The method according to claim 2, wherein The individual fitness corresponding to the initial weight candidate solution set is calculated by the following formula: Where f is the individual fitness corresponding to the initial weighted candidate solution set, E is the sum of the root mean square errors between the target meteorological forecast data and the actual meteorological data of all the observation stations under all climate models, j is the jth observation station, L is the total number of observation stations, t is the tth forecast time point, n is the total number of forecast time points in the historical time period, i is the i-th climate model, m is the total number of climate models, and W i,j,t is the weight coefficient of the ith climate model at the jth observation station at the tth forecast time point, R i,j,t is the target meteorological forecast data of the ith climate model at the jth observation station at the tth forecast time point, r i,j,t is the actual meteorological data of the t-th forecast time point at the j-th observation station of the ith climate model.

8. A meteorological element forecasting device based on genetic algorithm, characterized in that: The device comprises: a data acquisition unit, configured to acquire initial meteorological forecast data within a preset historical time period predicted by different climate models under a target emission scenario, and actual meteorological data collected by all observation stations within the historical time period under the target emission scenario, wherein the target emission scenario is any one of all emission scenarios, and the emission scenario represents the trajectory of global greenhouse gas emissions; a data processing unit, configured to perform bilinear interpolation calculation based on the initial weather forecast data and the position information of each of the observation sites to obtain target weather forecast data for each of the observation sites; Using a genetic algorithm, the target weight of each climate model is obtained based on the target weather forecast data of all the observation stations, the actual weather data, and a plurality of initial weight candidate solution sets, wherein the initial weight candidate solution sets include the weights of all climate models; The data prediction unit is used to perform weighted averaging on the predicted meteorological data under each of the climate modes and the target weight to obtain target predicted meteorological data.

9. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the processor is caused to perform the steps of the method according to any one of claims 1 to 7.

10. A computer device comprising a memory and a processor, characterized in that: The memory stores a computer program, and when the computer program is executed by the processor, the processor is caused to perform the steps of the method according to any one of claims 1 to 7.

Images