Rainfall climate prediction system based on multiple machine learning algorithms

Abstract

The invention provides a rainfall climate prediction system based on multiple machine learning algorithms, which belongs to the technical field of climate prediction, and comprises a data acquisition unit, a statistical analysis unit, a data decoding module, a key factor screening module, a key factor local cache module and a prediction unit, the output end of the statistical analysis unit and the output end of the data decoding module are connected with the key factor screening module, the key factor screening module is connected with the key factor local cache module, and the key factor local cache module is connected with the prediction unit. According to the method, rainstorm climate prediction analysis is carried out, rainstorm day number prediction based on a numerical mode, Chinese rose scale rainstorm concentration degree climate prediction and season-year scale rainstorm concentration period climate prediction are provided, and reference is provided for rainstorm disaster climate prediction business service.

Description

Technical Field

[0001] This invention relates to the field of climate prediction technology, and in particular to a precipitation climate prediction system based on multiple machine learning algorithms. Background Technology

[0002] In recent years, climate prediction research has received unprecedented attention both domestically and internationally. Climate prediction research not only has significant social and economic implications but also substantial scientific value. It is currently a major focus of international scientific research and has become one of the priority scientific and technological fields for development by various countries at the end of this century and the beginning of the next. Short-term climate prediction has a solid scientific foundation, and current short-term climate prediction systems built upon this foundation have demonstrated certain predictive skills. However, current short-term climate prediction systems, whether relying on empirical statistical methods, climate model predictions, or based on dynamics or statistical downscaling, suffer from problems such as weak regional specificity, short prediction timeframes, low accuracy, slow computation speed, and, in particular, limited predictive targets and unstable prediction results. Therefore, it is necessary to design a precipitation and climate prediction system based on multiple machine learning algorithms. Summary of the Invention

[0003] The purpose of this invention is to provide a precipitation and climate prediction system based on multiple machine learning algorithms, which solves the technical problems of existing climate prediction systems having a single target and unstable prediction results.

[0004] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0005] A precipitation and climate prediction system based on multiple machine learning algorithms includes a data acquisition unit, a statistical analysis unit, a data decoding module, a key factor screening module, a key factor local caching module, and a prediction unit. The output of the data acquisition unit is connected to the data decoding module, the outputs of the statistical analysis unit and the data decoding module are connected to the key factor screening module, the key factor screening module is connected to the key factor local caching module, and the key factor local caching module is connected to the prediction unit.

[0006] The data acquisition unit is used to collect climate-related data, the statistical analysis unit is used to statistically analyze rainstorm data, the data decoding module decodes the collected data, the key factor screening module is used to extract key factors based on the collected data and rainstorm data, the key factor local caching module is used to store key factor data, and the prediction unit is used to predict the climate based on particle swarm neural network and random forest, and then the prediction structure is verified and displayed as a chart.

[0007] Furthermore, the data acquisition unit includes a CIMISS precipitation data acquisition module, a CRA40 atmospheric reanalysis data acquisition module, an ERSST sea surface temperature field reanalysis data acquisition module, and a BCC_CSM and CFSv2.0 forecast data acquisition module. The ERSST sea surface temperature field reanalysis data in the ERSST sea surface temperature field reanalysis data acquisition module comes from the Internet. The CRA40 atmospheric reanalysis data and precipitation data in the CRA40 atmospheric reanalysis data acquisition module both come from the National Meteorological Administration's intranet. The CIMISS precipitation data acquisition module obtains data from the CIMISS statistical meteorological basic data interface and stores it in a MySQL database. The model data download in the data acquisition unit is implemented using a C / S program and stored in the data storage server. To ensure the stability of the download process and to prevent timeouts, a breakpoint resume method is used for downloading.

[0008] Furthermore, the statistical analysis unit includes a rainstorm days statistical analysis module, a rainstorm concentration statistical analysis module, and a rainstorm concentration period statistical analysis module. The rainstorm days statistical analysis module is used to statistically analyze the number of rainstorm days and output the analysis results. The rainstorm concentration statistical analysis module is used to perform statistical analysis on the concentration of rainstorms, and the rainstorm concentration period statistical analysis module is used to perform statistical analysis on the concentration of rainstorms.

[0009] Furthermore, the prediction unit includes a numerical model intelligent climate prediction module for the number of rainstorm days, a numerical model intelligent climate prediction module for concentration, a numerical model intelligent climate prediction module for concentration of atmospheric and oceanic indices, a numerical model intelligent climate prediction module for the concentration period, and a numerical model intelligent climate prediction module for the concentration period. The numerical model intelligent climate prediction module for the number of rainstorm days is used to predict the climate of the number of rainstorm days. The numerical model intelligent climate prediction module for concentration predicts the climate of concentration based on the data model. The numerical model intelligent climate prediction module for concentration predicts the climate of concentration based on the atmospheric and oceanic index data. The numerical model intelligent climate prediction module for the concentration period predicts the climate of the concentration period based on the numerical model. The numerical model intelligent climate prediction module for the concentration period predicts the climate of the concentration period based on the atmospheric and oceanic index data.

[0010] Furthermore, the prediction unit is equipped with both a particle swarm neural network and a random forest network to predict the climate simultaneously. The predicted structures of the two networks are then stored, and the ACC and AS tests of the predicted structures are performed.

[0011] Furthermore, the intelligent climate prediction module for heavy rainfall days based on the CFS numerical model uses data from the BCC-CSM and CFSv2.0 numerical models, as well as NCEP reanalysis data, to calculate the influence areas of key factors for heavy rainfall days, obtain forecast factors, and establish a monthly-seasonal scale heavy rainfall day climate prediction model based on machine learning methods such as particle swarm neural networks and random forests to predict the number of heavy rainfall days for any time period.

[0012] Furthermore, the concentration-based intelligent climate prediction module of the numerical model is based on the concentration-based intelligent climate prediction of the numerical model. It statistically analyzes the concentration of rainstorms over the years, uses BCC-CSM and CFSv2.0 model prediction data and NCEP / NCAR reanalysis data to calculate and statistically analyze key impact areas, screen out prediction factors, establish a rainstorm concentration climate prediction model based on particle swarm neural network and random forest methods, perform rainstorm concentration climate prediction for any time period, and display the prediction results.

[0013] Furthermore, the intelligent climate prediction module based on the concentration of atmospheric and oceanic indices calculates the concentration of rainstorms over the years, uses data from 130 atmospheric circulation and oceanic temperature indices from the National Climate Center, calculates and statistically analyzes highly correlated indices, selects predictive factors, establishes a rainstorm concentration climate prediction model based on particle swarm neural network and random forest method, performs rainstorm concentration climate prediction for any time period, and displays the prediction results.

[0014] Furthermore, the numerical model's concentrated period intelligent climate prediction module is based on the numerical model's concentrated period climate prediction. It statistically analyzes the concentrated periods of heavy rainfall over the years, uses BCC-CSM and CFSv2.0 numerical model data, NCEP / NCAR real-time reanalysis data, statistically analyzes key impact areas, obtains prediction factors, establishes a concentrated period climate prediction model based on particle swarm neural network and random forest method, performs prediction calculations, and displays the prediction results.

[0015] Furthermore, the intelligent climate prediction module for concentrated periods of atmospheric and oceanic indices predicts the climate of concentrated periods of heavy rainfall based on atmospheric and oceanic temperature indices. It statistically analyzes the concentrated periods of heavy rainfall over the years, uses 130 atmospheric circulation and oceanic temperature index data from the National Climate Center, calculates and statistically analyzes highly correlated indices, establishes a climate prediction model for concentrated periods of heavy rainfall based on particle swarm neural network and random forest method, performs prediction calculations, and displays the prediction results.

[0016] The present invention, by adopting the above-described technical solution, has the following beneficial effects:

[0017] This invention provides forecasting and analysis of rainstorm climate, offering numerical model-based forecasts of rainstorm days, monthly / seasonal-scale rainstorm concentration, and seasonal / annual-scale rainstorm concentration periods. It serves as a reference for operational services related to rainstorm disaster climate forecasting. Utilizing CFSv2.0 forecast data from the U.S. National Oceanic and Atmospheric Prediction Center (NOAA) and CSM climate model data from the National Climate Center (NCC), it establishes forecasting models for monthly / seasonal rainstorm days, rainstorm concentration, and rainstorm concentration periods based on various machine learning algorithms such as particle swarm optimization (PSO) and random forest. The forecast results are output and displayed, resulting in more accurate climate predictions. Attached Figure Description

[0018] Figure 1 This is a block diagram of the system structure modules of the present invention;

[0019] Figure 2 This is a functional schematic diagram of the system of the present invention;

[0020] Figure 3 This is the main interface diagram of the climate prediction of heavy rain days based on the CFS numerical model of this invention;

[0021] Figure 4 This is the main interface diagram of the concentration-based intelligent climate prediction based on numerical models of this invention;

[0022] Figure 5 This is a graph showing the prediction results of the particle swarm neural network method for reporting the concentration anomaly of rainstorms from January to December 2022, based on the present invention.

[0023] Figure 6 This is the main interface diagram of the climate prediction for concentrated rainstorm periods based on numerical models in this invention;

[0024] Figure 7 This is the main interface diagram of the climate prediction for concentrated periods of heavy rainfall based on atmospheric and ocean temperature indices in this invention. Detailed Implementation

[0025] To make the objectives, technical solutions, and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings and preferred embodiments. However, it should be noted that many details listed in the specification are merely to provide the reader with a thorough understanding of one or more aspects of the present invention, and these aspects of the invention can be implemented even without these specific details.

[0026] like Figure 1-2As shown, a climate prediction system based on multiple machine learning algorithms includes a data acquisition unit, a statistical analysis unit, a data decoding module, a key factor screening module, a key factor local caching module, and a prediction unit. The output of the data acquisition unit is connected to the data decoding module, the outputs of the statistical analysis unit and the data decoding module are connected to the key factor screening module, the key factor screening module is connected to the key factor local caching module, and the key factor local caching module is connected to the prediction unit.

[0027] The data acquisition unit is used to collect climate-related data, the statistical analysis unit is used to statistically analyze rainstorm data, the data decoding module decodes the collected data, the key factor screening module is used to extract key factors based on the collected data and rainstorm data, the key factor local caching module is used to store key factor data, and the prediction unit is used to predict the climate based on particle swarm neural network and random forest, and then the prediction structure is verified and displayed as a chart.

[0028] Based on the network and server environment of the Guangxi Climate Center, a detailed technical architecture design was carried out. The system adopts cross-platform programming, seamlessly connecting with software systems and equipment, possessing excellent interoperability, manageability, ease of use, and maintainability. Its performance indicators and processing capabilities fully meet the operational needs of the Guangxi Meteorological Research Institute. The system architecture adopts a B / S architecture mode, while data download uses a C / S mode, ensuring breakpoints and no timeouts during the download process, ultimately guaranteeing data integrity. Data decoding is primarily implemented using Grads, a commonly used meteorological tool, through a hybrid programming approach combining C# and Grads scripts. The system's plotting primarily utilizes the Surfer 11.0 COM component, which offers high interpolation efficiency, especially for multi-map plotting, providing fast speed, high flexibility, and aesthetically pleasing images. The database uses a 64-bit version of MySQL, with Redis used as a cache database to provide direct data support to various systems. The default cache duration is one month; after expiration, cached data is automatically destroyed, reducing server memory pressure. The development languages ​​primarily include C# 7.0, JavaScript, and HTML.

[0029] In this embodiment of the invention, the data acquisition unit includes a CIMISS precipitation data acquisition module, a CRA40 atmospheric reanalysis data acquisition module, an ERSST sea surface temperature field reanalysis data acquisition module, and a BCC_CSM and CFSv2.0 forecast data acquisition module. The ERSST sea surface temperature field reanalysis data in the ERSST reanalysis data acquisition module comes from the internet. The CRA40 atmospheric reanalysis data and precipitation data in the CRA40 atmospheric reanalysis data acquisition module both come from the National Meteorological Administration's intranet. The CIMISS precipitation data acquisition module obtains data from the CIMISS data statistical meteorological basic data interface and stores it in a MySQL database. The model data download in the data acquisition unit is implemented using a C / S program and stored on a data storage server. To ensure a stable download process without timeouts, a breakpoint resume method is used. Intelligent forecasting requires key area analysis based on the accessed data: precipitation, CRA40, and ERSST. CRA40 and SST data are decoded, and after calculation and statistics, the data is input into the particle swarm optimization-neural network algorithm to obtain and display the forecast results. Data decoding was performed using Opengrads 2.0, and the particle swarm optimization (PSO) neural network algorithm was called using Matlab 2017.

[0030] In this embodiment of the invention, the statistical analysis unit includes a rainstorm days statistical analysis module, a rainstorm concentration statistical analysis module, and a rainstorm concentration period statistical analysis module. The rainstorm days statistical analysis module is used to statistically analyze the number of rainstorm days and output the analysis results. The rainstorm concentration statistical analysis module is used to perform rainstorm concentration statistical analysis, and the rainstorm concentration period statistical analysis module is used to perform rainstorm concentration period statistical analysis.

[0031] In this embodiment of the invention, the prediction unit includes a numerical model intelligent climate prediction module for the number of rainstorm days, a numerical model intelligent climate prediction module for concentration, an atmospheric and oceanic index intelligent climate prediction module for concentration, a numerical model intelligent climate prediction module for the concentration period, and an atmospheric and oceanic index intelligent climate prediction module for the concentration period. The numerical model intelligent climate prediction module for the number of rainstorm days is used to predict the rainstorm day climate. The numerical model intelligent climate prediction module for concentration predicts the concentration climate based on the data model. The atmospheric and oceanic index intelligent climate prediction module for concentration predicts the concentration climate based on the atmospheric and oceanic index data. The numerical model intelligent climate prediction module for the concentration period predicts the concentration period climate based on the numerical model. The atmospheric and oceanic index intelligent climate prediction module for the concentration period predicts the concentration period climate based on the atmospheric and oceanic index data.

[0032] The images used in the display were created using Surfer software. Surfer is a professional 2D and 3D geographic mapping software developed by Golden Software in the United States. It allows users to customize line styles and colors, edit text, fill graphic colors, generate visualized geographic images from data, design grid representations, and create accurate geographic maps. Surfer is mainly used for modeling terrain and depth maps, visualizing landscapes, analyzing surfaces, and drawing 3D surface maps.

[0033] Surfer provides input / output interfaces for various popular graphic image file formats as well as major GIS software file formats, including GSI, SRF, HDF, CSV, DXF, ECW, ASC, ERS, and TAB, greatly facilitating the exchange of files and data.

[0034] Surfer's powerful interpolation and mapping capabilities make it the preferred software for processing XYZ data and an essential professional mapping tool for geologists. It can easily create base maps, data point maps, categorical data maps, contour maps, wireframe maps, topographic maps, trend maps, vector maps, and 3D surface maps. It offers 11 data gridding methods, covering almost all popular data statistical calculation methods. The new scripting engine greatly enhances automation capabilities.

[0035] Surfer is used in scientific research by geologists, hydrologists, archaeologists, biologists, oceanographers, climatologists, medical researchers, and geophysicists. It supports large datasets and grid formats, giving you complete control over the appearance of your data maps. If you work in the scientific field and need a map visualization tool for XYZ data, Surfer is an excellent choice.

[0036] All bar charts, line charts, and pie charts in the system are implemented using Baidu's open-source Echarts.js.

[0037] ECharts is an open-source visualization library implemented in JavaScript. It can run smoothly on PCs and mobile devices and is compatible with most current browsers (IE8 / 9 / 10 / 11, Chrome, Firefox, Safari, etc.). It relies on the lightweight vector graphics library ZRender and provides intuitive, interactive, and highly customizable data visualization charts.

[0038] ECharts offers a variety of chart types, including line charts, bar charts, scatter plots, pie charts, and candlestick charts; box plots for statistics; maps, heatmaps, and line charts for geographic data visualization; relationship diagrams and sunburst charts for relational data visualization; parallel coordinates for multidimensional data visualization; funnel charts and dashboards for business intelligence; and it supports mixing and matching different chart types.

[0039] ECharts effectively solves compatibility issues when displaying charts in various browsers, completely replacing the problem of Flash not being supported.

[0040] In this embodiment of the invention, the prediction unit is equipped with a particle swarm neural network and a random forest network to predict the climate simultaneously. Then, the predicted structures of the two networks are stored, and the ACC test and AS test of the predicted structures are performed.

[0041] In this embodiment of the invention, the intelligent climate prediction module for heavy rainfall days based on the CFS numerical model uses daily precipitation data from 1991 to 2020 to count the number of days with precipitation ≥50 mm. It then uses BCC-CSM, CFSv2.0 numerical model prediction data, and NCEP reanalysis data to calculate the influence areas of key factors for heavy rainfall days, obtain forecast factors, and establish a monthly / seasonal scale heavy rainfall day prediction model based on particle swarm neural network and random forest machine learning methods to predict the number of heavy rainfall days for any given time period. Figure 3 As shown.

[0042] Basic operations:

[0043] (1) Heavy Rain Days Forecast: Select the reporting month, forecast month, forecast algorithm, model data, and modeling method, then click query to display the predicted heavy rain days under the current conditions, presented in the form of color-coded maps and tables, such as... Figure 3 As shown.

[0044] (2) Image display: Click "Image" to display the predicted number of rainstorm days.

[0045] (3) Data: Click “Data” to display the table of predicted number of rainstorm days.

[0046] (4) Re-predict: Clicking “Re-predict” will ignore the previously predicted result and re-predict the result using the current conditions.

[0047] (5) Prediction score: Compare the historical prediction results with the actual data to verify the prediction results. Click "Historical score" to check the effect of historical predictions.

[0048] (6) Redraw: Select Yes to redraw the color patch map based on the query results; select No to use the previously cached image. Especially since re-prediction is performed, not selecting Redraw may result in incorrect images.

[0049] In this embodiment of the invention, the concentration-based intelligent climate prediction module of the numerical model uses daily precipitation data from 1991 to 2020 to statistically analyze the concentration of rainstorms over the years. It uses BCC-CSM, CFSv2.0 model prediction data and NCEP / NCAR reanalysis data to calculate and statistically analyze key impact areas, screen predictive factors, establish a rainstorm concentration climate prediction model based on particle swarm neural network and random forest methods, perform rainstorm concentration climate prediction for any time period, and display the prediction results.

[0050] (1) Query: Select the reporting start month, forecast period, prediction algorithm, model data, and modeling method, then click Query to display the concentration intelligent climate prediction results under the current conditions, presented in a color patch map ( Figure 4 Displayed in both ) and tabular formats.

[0051] (2) Image display: Click "Image" to display the concentration of intelligent climate prediction results.

[0052] (3) Data: Click “Data” to display the table of concentrated intelligent climate prediction results.

[0053] (4) Re-predict: Click “Re-predict” to re-predict using the current conditions.

[0054] In this embodiment of the invention, the intelligent climate prediction module based on the concentration of atmospheric and oceanic indices uses daily precipitation data from 1991 to 2020. It calculates the concentration of heavy rainfall over the years using a concentration calculation formula, calculates and statistically analyzes 130 atmospheric circulation and oceanic temperature indices from the National Climate Center, selects predictive factors, and establishes a heavy rainfall concentration climate prediction model based on particle swarm neural networks and random forest methods. This model predicts the heavy rainfall concentration for any given time period and displays the prediction results. Figure 5 As shown.

[0055] The basic operation is as follows: (1) Query: Select the reporting month, forecast period, and prediction algorithm, and then click query to display the concentration of intelligent climate prediction results under the current conditions, which are presented in the form of color patch map and table.

[0056] (2) Image display: Click "Image" to display the concentration of intelligent climate prediction results.

[0057] (3) Data: Click “Data” to display the table of concentrated intelligent climate prediction results.

[0058] (4) Re-predict: Click “Re-predict” to re-predict using the current conditions.

[0059] (5) Historical Scoring: Compare the predicted results with the actual data to verify and evaluate the predicted results. Click "Historical Scoring" to view the prediction performance. Figure 6 As shown.

[0060] (6) Redraw: Select Yes to redraw the color patch map based on the query results; select No to use the previously cached image. Especially since re-prediction is performed, not selecting Redraw may result in incorrect images.

[0061] In this embodiment of the invention, the numerical model's concentrated period intelligent climate prediction module is based on the numerical model's concentrated period climate prediction of heavy rainfall. It uses daily precipitation data from 1991 to 2023 to statistically analyze the concentrated periods of heavy rainfall over the years. It utilizes BCC-CSM and CFSv2.0 numerical model data, NCEP / NCAR real-time reanalysis data, statistically analyzes key impact areas, obtains prediction factors, establishes a concentrated period climate prediction model of heavy rainfall based on particle swarm neural network and random forest method, performs prediction calculations, and displays the prediction results.

[0062] Basic Operations

[0063] (1) Query: Select the reporting month, forecast period, prediction algorithm, model data, and modeling method, and then click query to display the prediction results of the concentrated period of heavy rain under the current conditions, in the form of color patch map and table.

[0064] (2) Re-predict: Clicking “Re-predict” will ignore the previously predicted result and re-predict the result using the current conditions.

[0065] (3) Historical scoring: Compare the prediction results with the actual data to verify and evaluate the prediction results. Click "Historical scoring" to check the prediction effect.

[0066] (4) Redraw: Select Yes to redraw the color patch map based on the query results; select No to use the previously cached image. Especially since re-prediction is performed, not selecting Redraw may result in incorrect images.

[0067] In this embodiment of the invention, the intelligent climate prediction module for concentrated periods of atmospheric and oceanic indices is based on the climate prediction of concentrated periods of heavy rainfall using atmospheric and oceanic temperature indices. It uses daily precipitation data from 1991 to 2023 to statistically analyze the concentrated periods of heavy rainfall over the years. It uses 130 atmospheric circulation and oceanic temperature index data from the National Climate Center to calculate and statistically analyze highly correlated indices. It establishes a climate prediction model for concentrated periods of heavy rainfall based on particle swarm neural network and random forest method, performs prediction calculations, and displays the prediction results.

[0068] Basic operations:

[0069] Query: Select the start month, forecast period, and prediction algorithm, then click Query to display the prediction results for the concentrated period of heavy rainfall under the current conditions, presented in a color-coded map. Figure 7 Displayed in both ) and tabular formats.

[0070] (2) Re-predict: Clicking “Re-predict” will ignore the previously predicted result and re-predict the result using the current conditions.

[0071] (3) Historical scoring: Compare the prediction results with the actual data to verify and evaluate the prediction results. Click "Historical scoring" to check the prediction effect.

[0072] (4) Redraw: Select Yes to redraw the color patch map based on the query results; select No to use the previously cached image. Especially since re-prediction is performed, not selecting Redraw may result in incorrect images.

[0073] Matters not covered in this invention are common knowledge.

[0074] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A precipitation and climate prediction system based on machine learning algorithms, characterized in that: It includes a data acquisition unit, a statistical analysis unit, a data decoding module, a key factor screening module, a key factor local caching module, and a prediction unit. The output of the data acquisition unit is connected to the data decoding module, the output of the statistical analysis unit and the data decoding module are connected to the key factor screening module, the key factor screening module is connected to the key factor local caching module, and the key factor local caching module is connected to the prediction unit. The data acquisition unit is used to collect climate-related data, the statistical analysis unit is used to statistically analyze rainstorm data, the data decoding module decodes the collected data, the key factor screening module is used to extract key factors based on the collected data and rainstorm data, the key factor local caching module is used to store key factor data, and the prediction unit is used to predict the climate based on particle swarm neural network and random forest, and then the prediction structure is verified and displayed as a chart.

2. The precipitation and climate prediction system based on machine learning algorithms according to claim 1, characterized in that: The data acquisition unit includes a CIMISS precipitation data acquisition module, a CRA40 atmospheric reanalysis data acquisition module, an ERSST sea surface temperature field reanalysis data acquisition module, and a BCC_CSM and CFSv2.0 forecast data acquisition module. The ERSST sea surface temperature field reanalysis data in the ERSST sea surface temperature field reanalysis data acquisition module comes from the Internet. The CRA40 atmospheric reanalysis data and precipitation data in the CRA40 atmospheric reanalysis data acquisition module come from the National Meteorological Administration's intranet. The CIMISS precipitation data acquisition module obtains data from the CIMISS statistical meteorological basic data interface and stores it in a MySQL database. The model data download in the data acquisition unit is implemented using a C / S program and stored in the data storage server. To ensure the stability of the download process and to prevent timeouts, a breakpoint resume method is used for downloading.

3. The precipitation and climate prediction system based on machine learning algorithm according to claim 1, characterized in that: The statistical analysis unit includes a heavy rain day statistics analysis module, a heavy rain concentration statistics analysis module, and a heavy rain concentration period statistics analysis module. The heavy rain day statistics analysis module is used to count and analyze the number of heavy rain days and output the analysis results. The heavy rain concentration statistics analysis module is used to perform statistical analysis on the concentration of heavy rain. The heavy rain concentration period statistics analysis module is used to perform statistical analysis on the period of heavy rain.

4. The precipitation and climate prediction system based on machine learning algorithm according to claim 1, characterized in that: The prediction unit includes a numerical model-based intelligent climate prediction module for the number of rainstorm days, a numerical model-based intelligent climate prediction module for concentration, a numerical model-based intelligent climate prediction module for concentration of atmospheric and oceanic indices, a numerical model-based intelligent climate prediction module for the concentration period, and a numerical model-based intelligent climate prediction module for the concentration period. The numerical model-based intelligent climate prediction module for the number of rainstorm days is used to predict the climate of the number of rainstorm days. The numerical model-based intelligent climate prediction module for concentration predicts the climate of concentration based on the data model. The numerical model-based intelligent climate prediction module for concentration predicts the climate of concentration based on atmospheric and oceanic index data. The numerical model-based intelligent climate prediction module for the concentration period predicts the climate of the concentration period based on the numerical model. The numerical model-based intelligent climate prediction module for the concentration period predicts the climate of the concentration period based on atmospheric and oceanic index data.

5. A precipitation and climate prediction system based on machine learning algorithms according to claim 1, characterized in that: The prediction unit is equipped with both a particle swarm neural network and a random forest network to predict climate simultaneously. The predicted structures of the two networks are then stored, and the ACC and AS tests of the predicted structures are performed.

6. The precipitation and climate prediction system based on machine learning algorithm according to claim 4, characterized in that: The intelligent climate prediction module for heavy rainfall days based on the CFS numerical model uses data from the BCC-CSM and CFSv2.0 numerical models, as well as NCEP reanalysis data, to calculate the influence areas of key factors for heavy rainfall days, obtain forecast factors, and establish a monthly-seasonal scale heavy rainfall day prediction model based on particle swarm neural network and random forest machine learning methods to predict the number of heavy rainfall days for any time period.

7. A precipitation and climate prediction system based on machine learning algorithms according to claim 4, characterized in that: The numerical model concentration-based intelligent climate prediction module is based on the numerical model concentration-based intelligent climate prediction. It statistically analyzes the concentration of rainstorms over the years, uses BCC-CSM and CFSv2.0 model prediction data and NCEP / NCAR reanalysis data to calculate and statistically analyze key impact areas, selects prediction factors, establishes a rainstorm concentration climate prediction model based on particle swarm neural network and random forest methods, performs rainstorm concentration climate prediction for any time period, and displays the prediction results.

8. A precipitation and climate prediction system based on machine learning algorithms according to claim 4, characterized in that: The intelligent climate prediction module based on the concentration of atmospheric and oceanic indices predicts the concentration of heavy rainfall over the years. It uses data from 130 atmospheric circulation and oceanic temperature indices from the National Climate Center to calculate and statistically analyze highly correlated indices, selects predictive factors, and establishes a heavy rainfall concentration climate prediction model based on particle swarm neural network and random forest method. The model then performs heavy rainfall concentration climate prediction for any time period and displays the prediction results.

9. A precipitation and climate prediction system based on machine learning algorithms according to claim 4, characterized in that: The numerical model's concentrated period intelligent climate prediction module is based on the concentrated period climate prediction of heavy rainfall in numerical models. It statistically analyzes the concentrated periods of heavy rainfall over the years, uses data from the BCC-CSM and CFSv2.0 numerical models, and NCEP / NCAR real-time reanalysis data to statistically analyze key impact areas, obtains prediction factors, establishes a concentrated period climate prediction model based on particle swarm neural network and random forest method, performs prediction calculations, and displays the prediction results.

10. A precipitation and climate prediction system based on machine learning algorithms according to claim 4, characterized in that: The intelligent climate prediction module for concentrated periods of atmospheric and oceanic indices predicts the climate of concentrated periods of heavy rainfall based on atmospheric and oceanic temperature indices. It statistically analyzes the concentrated periods of heavy rainfall over the years, uses 130 atmospheric circulation and oceanic temperature index data from the National Climate Center, calculates and statistically analyzes highly correlated indices, establishes a climate prediction model for concentrated periods of heavy rainfall based on particle swarm neural network and random forest method, performs prediction calculations, and displays the prediction results.

Images