Quantitative prediction system and method for extreme rainfall

Abstract

The invention relates to the technical field of meteorology and data processing, in particular to a quantitative prediction system and method for extreme rainfall. The prediction system is composed of a sensor module, a data processing module, a deep learning model, a user interaction interface and a cloud computing platform, accumulative learning of historical data can be achieved, and it is ensured that a prediction model is gradually optimized along with time. The prediction method comprises the following steps: collecting real-time data; preprocessing the data; extreme rainfall prediction; generating early warning information; and issuing information. The invention provides a quantitative prediction system and method for extreme rainfall, and the system and method can achieve the precise collection, efficient preprocessing and intelligent prediction and analysis of real-time data, and the timely release of early warning information.

Description

Technical Field

[0001] The present invention relates to the field of meteorology and data processing technology, and in particular to a quantitative prediction system and method for extreme precipitation. Background Art

[0002] As global climate change intensifies, the frequency and intensity of extreme climate events are on the rise. Extreme precipitation events, such as heavy rain and snow, have a significant impact on human life and production activities, including but not limited to floods, landslides, urban waterlogging and other problems, posing a serious threat to human life and property safety. As global climate change intensifies, the frequency and intensity of extreme precipitation events are on the rise. Therefore, building a system that can effectively predict extreme precipitation and issue early warning information to the public in a timely manner is of great significance for reducing losses caused by disasters and improving risk management capabilities.

[0003] Traditional extreme precipitation prediction methods mainly rely on atmospheric physics and statistical analysis, and predict future precipitation conditions by studying historical meteorological data. Although this method can provide reference information to a certain extent, it ignores the nonlinearity and complexity of the weather, and the prediction accuracy is low, especially for extreme weather prediction. In addition, traditional methods often require a lot of manual operation and analysis, which is time-consuming and labor-intensive, and cannot achieve real-time prediction. Summary of the invention

[0004] The present invention provides a quantitative prediction system and method for extreme precipitation, which realizes accurate collection of real-time data, efficient preprocessing, intelligent prediction and analysis, and timely release of warning information by integrating sensor modules, data processing modules, deep learning models, user interaction interfaces and cloud computing platforms.

[0005] The technical solution adopted by the present invention is: a quantitative prediction system for extreme precipitation, which consists of a sensor module, a data processing module, a deep learning model, a user interaction interface, and a cloud computing platform. It can realize the cumulative learning of historical data and ensure that the prediction model is gradually optimized over time.

[0006] The sensor module is used to collect precipitation-related data in real time, including but not limited to atmospheric pressure, temperature, humidity, wind speed, wind direction and precipitation type information;

[0007] The data processing module is in communication with the sensor module and is used to receive, pre-process and store the above information, and improve the data quality through advanced data cleaning, outlier detection and removal algorithms;

[0008] The deep learning model is trained based on a large amount of historical precipitation data and can predict the probability and intensity of extreme precipitation within a certain time window in the future based on the data collected in real time;

[0009] The user interaction interface is used to display the prediction results, including graphical and tabular forms, and can provide early warning information;

[0010] The cloud computing platform is used to support computing-intensive tasks related to data storage, deep learning model training and reasoning in the system, ensuring the efficient operation of the prediction system.

[0011] As a further improvement of the present invention, the data processing module has the following functions: standardizing the collected data to unify the data format from different sensors for subsequent analysis; using time series analysis algorithms to perform trend analysis on historical data to identify seasonal change patterns and long-term climate change trends; and fusing multiple prediction models through an integrated learning algorithm to improve the accuracy and reliability of the prediction.

[0012] As a further improvement of the present invention, the deep learning model adopts a hybrid architecture of a convolutional neural network (CNN) combined with a long short-term memory network (LSTM) to simultaneously capture the changes in meteorological characteristics in space and the dynamic trends in time series.

[0013] As a further improvement of the present invention, the user interaction interface supports access from multiple terminal devices, including but not limited to mobile phone APP, web pages, and smart wearable devices, and the interface design complies with the latest standards for human-computer interaction, providing personalized service options, including precipitation forecasts for specific locations and warnings for specific time points subscribed by users.

[0014] As a further improvement of the present invention, the prediction system further includes a data acquisition module connected to the data processing module, and the data acquisition module includes a ground observation data acquisition unit, a satellite observation data acquisition unit and a radar observation data acquisition unit.

[0015] As a further improvement of the present invention, the ground observation data acquisition unit is used to collect observation data of the meteorological station at various locations on the ground; the satellite observation data acquisition unit is used to collect cloud images and other satellite remote sensing data; and the radar observation data acquisition unit is used to collect radar echo data.

[0016] A quantitative prediction method for extreme precipitation based on the above prediction system comprises the following steps:

[0017] Step 1: Real-time data collection: Collect current environmental parameters through sensors and data collection modules deployed in different geographical locations;

[0018] Step 2: Data preprocessing: clean, format, and perform time series analysis on the collected data to prepare the data set for model input;

[0019] Step 3: Extreme precipitation prediction: Input the preprocessed data into the trained deep learning model and output the precipitation prediction results for a period of time in the future;

[0020] Step 4: Warning information generation: Evaluate whether there is an extreme precipitation risk based on the forecast results. If an extreme precipitation event is predicted to occur, a warning information is automatically generated.

[0021] Step 5: Information release: timely push prediction results and warning information to users through the user interaction interface.

[0022] As a further improvement of the present invention, in the extreme precipitation prediction link of step three, the deep learning model not only considers the environmental parameters at the current moment, but also combines the data of the same location in the past 24 hours, as well as the historical and real-time data of adjacent locations to form a multi-dimensional, time-space comprehensive prediction model.

[0023] As a further improvement of the present invention, in the warning information generation link of step 4, when the predicted probability of extreme precipitation exceeds a preset threshold, the prediction system will automatically call a disaster assessment algorithm to assess the possible impact of the event, including but not limited to floods, landslides, and urban waterlogging, and generate corresponding warning levels according to different disaster types.

[0024] As a further improvement of the present invention, in the information release link of step five, users can choose the device to receive the warning information according to their own needs, and the system supports customized push, such as setting different warning strategies and notification frequencies for disaster-prone areas and specific user groups.

[0025] The beneficial effects of the present invention are as follows: (1) Improving prediction accuracy: By integrating advanced sensor technology, deep learning models and cloud computing platforms, the present invention enables the system to obtain more comprehensive and higher-quality real-time meteorological data, and utilizes a hybrid architecture model of convolutional neural networks (CNN) and long short-term memory networks (LSTM) to capture spatial meteorological characteristics and time series dynamic trends, thereby significantly improving the prediction accuracy of extreme precipitation events and the reliability of intensity estimation.

[0026] (2) Support for multi-terminal device access and personalized services: The user interaction interface of the present invention is designed to support a variety of terminal devices, including mobile phone APP, web pages, smart wearable devices, etc., to ensure that users can access prediction results and warning information anytime and anywhere, and can provide personalized service options that meet the latest standards of human-computer interaction. Reminders can be subscribed according to user-specific needs, thereby improving the practicality of the system and user satisfaction.

[0027] (3) Enhanced disaster risk management: After predicting extreme precipitation, the prediction system of the present invention can automatically call the disaster assessment algorithm to assess the potential impact, and generate corresponding warning levels according to different disaster types, and push information to the relevant public and government departments in a timely manner, which helps to take measures in advance to reduce disaster losses and improve the society's overall disaster response capabilities and risk management level. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] Figure 1 It is a system block diagram of a quantitative prediction system and method for extreme precipitation of the present invention;

[0029] Figure 2 The present invention is a flowchart of a quantitative prediction system and method for extreme precipitation. DETAILED DESCRIPTION

[0030] In order to make the technical problems, technical solutions and beneficial effects to be solved by the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0031] The present invention provides a quantitative prediction system for extreme precipitation. The prediction system consists of a sensor module, a data processing module, a deep learning model, a user interaction interface, and a cloud computing platform. The system can realize the cumulative learning of historical data and ensure that the prediction model is gradually optimized over time. The sensor module is used to collect precipitation-related data in real time, including but not limited to atmospheric pressure, temperature, humidity, wind speed, wind direction, and precipitation type information; the data processing module is communicated with the sensor module to receive, preprocess, and store the above information, and improve data quality through advanced data cleaning, outlier detection and removal algorithms; the deep learning model is trained based on a large amount of historical precipitation data, and can predict the probability and intensity of extreme precipitation within a certain time window in the future based on the data collected in real time; the user interaction interface is used to display the prediction results, including graphical and tabular forms, and can provide early warning information; the cloud computing platform is used to support data storage, deep learning model training and reasoning-related computing-intensive tasks in the system to ensure the efficient operation of the prediction system.

[0032] The data processing module described in the present invention has the following functions: standardizing the collected data to unify the data format from different sensors for subsequent analysis; using time series analysis algorithms to perform trend analysis on historical data to identify seasonal variation patterns and long-term climate change trends; and integrating multiple prediction models through integrated learning algorithms to improve the accuracy and reliability of predictions.

[0033] The deep learning model described in the present invention adopts a hybrid architecture of a convolutional neural network (CNN) combined with a long short-term memory network (LSTM) to simultaneously capture the changes in meteorological characteristics in space and the dynamic trends in time series.

[0034] The user interaction interface described in the present invention supports access from multiple terminal devices, including but not limited to mobile phone APP, web pages, and smart wearable devices, and the interface design complies with the latest standards for human-computer interaction, providing personalized service options, including precipitation forecasts for specific locations and warnings at specific time points subscribed by users.

[0035] The prediction system in the present invention also includes a data acquisition module connected to the data processing module, and the data acquisition module includes a ground observation data acquisition unit, a satellite observation data acquisition unit and a radar observation data acquisition unit. The ground observation data acquisition unit is used to collect observation data of the meteorological station at various locations on the ground; the satellite observation data acquisition unit is used to collect cloud maps and other satellite remote sensing data; the radar observation data acquisition unit is used to collect radar echo data.

[0036] A quantitative prediction method for extreme precipitation based on the above prediction system comprises the following steps:

[0037] Step 1: Real-time data collection: Collect current environmental parameters through sensors and data collection modules deployed in different geographical locations;

[0038] Step 2: Data preprocessing: clean, format, and perform time series analysis on the collected data to prepare the data set for model input;

[0039] Step 3: Extreme precipitation prediction: The preprocessed data is input into the trained deep learning model to output the precipitation prediction results for a period of time in the future. The deep learning model not only considers the environmental parameters at the current moment, but also combines the data of the same location in the past 24 hours, as well as the historical and real-time data of adjacent locations to form a multi-dimensional, spatiotemporal comprehensive prediction model.

[0040] Step 4: Generation of warning information: Based on the forecast results, assess whether there is an extreme precipitation risk. If an extreme precipitation event is predicted to occur, a warning message is automatically generated. When the predicted probability of extreme precipitation exceeds the preset threshold, the forecast system will automatically call the disaster assessment algorithm to assess the possible impact of the event, including but not limited to floods, landslides, and urban waterlogging, and generate corresponding warning levels according to different disaster types.

[0041] Step 5: Information release: The prediction results and warning information are pushed to users in a timely manner through the user interaction interface. Users can choose the device to receive the warning information according to their own needs, and the system supports customized push, such as setting different warning strategies and notification frequencies for disaster-prone areas and specific user groups.

[0042] Example 1 (Construction of urban flood warning system):

[0043] 1. Construction Background

[0044] Urban floods are a common type of extreme precipitation events, especially in old urban areas with imperfect drainage systems, where extreme precipitation often causes serious flood disasters. This embodiment aims to build a usable urban flood warning system to reduce the impact of such disasters on urban residents.

[0045] (II) System Construction

[0046] (1) Sensor module deployment: Various types of meteorological sensors are installed at key points in the city (such as roads prone to waterlogging and river inlets and outlets), while integrating data from existing hydrological monitoring stations to ensure that all parameters related to precipitation can be fully monitored.

[0047] (2) Optimization of data acquisition modules: The ground observation data acquisition unit collects ground environmental data through existing meteorological stations and newly added sensor equipment; the satellite observation data acquisition unit uses meteorological satellite data to obtain cloud map information and evaluate the formation and movement trends of precipitation clouds; the radar observation data acquisition unit collects radar signals for real-time monitoring of precipitation conditions.

[0048] (3) Deep learning model training: Historical precipitation data and corresponding disaster reports are used as training sets. The model adopts a CNN-LSTM structure. First, CNN is used to capture the meteorological characteristics of different geographical locations in space, and then LSTM is used to capture the time series characteristics in historical data. During the training process, the cross-validation method is used to avoid overfitting, and the model performance is optimized by adjusting the hyperparameters.

[0049] (4) Data processing and preprocessing: Perform preliminary cleaning of the collected sensor data to remove invalid or erroneous records. Apply standardization techniques to unify the data formats from different sources for subsequent analysis. Perform trend analysis on historical data through time series analysis algorithms to identify seasonal precipitation patterns and use them as input features as part of model training.

[0050] (5) Early warning algorithm development: Based on historical disaster data, determine the risk thresholds for different types of extreme precipitation events (such as rainstorms and short-term heavy precipitation). When the system predicts that the precipitation or precipitation probability in a certain place in the next 24 hours exceeds the preset threshold, the disaster assessment algorithm is automatically called to assess the specific impact that the extreme precipitation may bring, such as the scope of the flood and the duration of the impact. Based on the assessment results, different warning levels are assigned to different types of extreme precipitation events.

[0051] (6) User interface design: Design simple and intuitive mobile APP and web interfaces to support users to query precipitation forecast information for specific locations. Allow users to subscribe to weather warnings for specific locations and time periods. Warning information can be notified to users through APP push, email, SMS, etc. Mark disaster warning points on maps of disaster-prone areas and provide disaster response guidelines.

[0052] (III) Implementation process

[0053] (1) Data collection: Starting from 2022, data from various monitoring points in the city will be continuously collected, including temperature, humidity, wind speed and direction, precipitation, etc.

[0054] (2) Initial training: One year’s data is used for the initial training of the model. The model parameters are continuously adjusted until the model’s performance on the test set reaches the expected level.

[0055] (3) System online testing: System online testing will be carried out on a small scale starting in early 2023 to collect feedback and optimize the user interface and early warning algorithm.

[0056] (4) Full promotion: After successful testing, full promotion will begin in mid-2023, especially before the upcoming rainy season, to ensure that the system can provide services to as many users as possible.

[0057] (5) Continuous optimization: By collecting new data in real time and adding it to the training set, the deep learning model is continuously trained and optimized to improve prediction accuracy.

[0058] Example 2 (Construction of an agricultural disaster prevention and mitigation early warning system):

[0059] 1. Construction Background

[0060] Agriculture is one of the industries most sensitive to extreme precipitation events. Sudden heavy rains or long-term droughts can cause great damage to crops. By building an extreme precipitation early warning system specifically for the agricultural sector, measures can be taken in advance to reduce losses and protect agricultural production safety.

[0061] (II) System Construction

[0062] (1) Sensor module deployment: Temperature and humidity sensors, soil moisture sensors, rainfall sensors, etc. are installed in major crop planting areas to monitor environmental changes in agricultural areas. In addition, drones can be deployed for regular aerial inspections to collect information such as surface temperature and crop health.

[0063] (2) Data collection module optimization: In addition to the data from ground sensors, satellite observations are used to collect information such as cloud conditions and rainfall estimates over a large area, and radar observations are used to obtain more accurate precipitation distribution and intensity data.

[0064] (3) Deep learning model training: Agricultural meteorological data from the past 10 years were used as the training basis. These data cover multiple time periods of extreme weather events. The model also uses the CNN-LSTM structure to learn complex meteorological patterns in space and time, and how these patterns affect crop growth. During the training process, special attention was paid to avoid bias of the model towards specific geographical areas.

[0065] (4) Data processing and preprocessing: More complex algorithms are used to process data from different sensors to ensure their quality and compatibility. For example, Kalman filters and wavelet transforms are used to smooth sensor readings and eliminate noise interference.

[0066] (5) Early warning algorithm development: In collaboration with agricultural experts, we developed early warning thresholds suitable for different crop types, taking into account factors such as crop growth cycle, soil type, and drainage conditions. When the precipitation pattern in the next 72 hours is predicted to cause large-scale flooding or drought, the system automatically issues an early warning.

[0067] (6) User interface design: In response to the actual needs of farmers, an easy-to-use APP was designed that can display current weather conditions, future precipitation forecasts, and recommended disaster prevention measures. The APP supports multi-language switching, which is convenient for users in different regions. At the same time, in order to adapt to the usage habits of farmers of different ages, SMS and telephone alarm services are also provided.

[0068] (III) Implementation process

[0069] (1) Data collection: Starting from 2024, sensors and other data collection tools will be gradually deployed in major agricultural production areas to ensure the comprehensiveness and real-time nature of the data.

[0070] (2) Initial training: After collecting data for one year, the first large-scale model training will begin in early 2025 to preliminarily test the performance of the model.

[0071] (3) System online testing: In the second half of 2025, several typical agricultural production areas will be selected for pilot applications, and the system will be adjusted and optimized based on farmers’ feedback.

[0072] (4) Comprehensive promotion: It is planned to officially promote the technology to the majority of agricultural practitioners in early 2026, especially to strengthen publicity before spring sowing to help farmers better plan the planting season.

[0073] (5) Continuous optimization: Establish a data update mechanism, regularly add new observation data to the training set, and keep the model prediction capability updated. At the same time, continuously collect user feedback, improve user experience, and increase the popularity and service level of the system.

[0074] In summary, the quantitative prediction system and method of extreme precipitation of the present invention can be successfully applied to key areas, significantly reduce the losses caused by disasters, and improve the overall response capacity and risk management level of society. Therefore, the present invention not only improves the prediction ability of extreme precipitation events, but also builds security barriers in two important areas through a friendly user interface and service system, demonstrating the potential and value of the combination of meteorology and information science.

[0075] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features thereof may be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A quantitative prediction system for extreme precipitation, characterized in that: The prediction system consists of sensor modules, data processing modules, deep learning models, user interaction interfaces, and cloud computing platforms. It can realize the cumulative learning of historical data and ensure that the prediction model is gradually optimized over time. The sensor module is used to collect precipitation-related data in real time, including but not limited to atmospheric pressure, temperature, humidity, wind speed, wind direction and precipitation type information; The data processing module is in communication with the sensor module and is used to receive, pre-process and store the above information, and improve the data quality through advanced data cleaning, outlier detection and removal algorithms; The deep learning model is trained based on a large amount of historical precipitation data and can predict the probability and intensity of extreme precipitation within a certain time window in the future based on the data collected in real time; The user interaction interface is used to display the prediction results, including graphical and tabular forms, and can provide early warning information; The cloud computing platform is used to support computing-intensive tasks related to data storage, deep learning model training and reasoning in the system, ensuring the efficient operation of the prediction system.

2. The quantitative prediction system for extreme precipitation according to claim 1, characterized in that: The data processing module has the following functions: standardize the collected data to unify the data format from different sensors for subsequent analysis; use time series analysis algorithms to perform trend analysis on historical data to identify seasonal change patterns and long-term climate change trends; By integrating multiple prediction models through ensemble learning algorithms, the accuracy and reliability of predictions can be improved.

3. The quantitative prediction system for extreme precipitation according to claim 1, characterized in that: The deep learning model adopts a hybrid architecture of convolutional neural network (CNN) combined with long short-term memory network (LSTM) to simultaneously capture the changes in meteorological characteristics in space and the dynamic trends in time series.

4. The quantitative prediction system for extreme precipitation according to claim 1, characterized in that: The user interaction interface supports access from multiple terminal devices, including but not limited to mobile phone APP, web pages, and smart wearable devices, and the interface design complies with the latest standards for human-computer interaction, providing personalized service options, including precipitation forecasts for specific locations and warnings at specific time points subscribed by users.

5. The quantitative prediction system for extreme precipitation according to claim 1, characterized in that: The prediction system also includes a data acquisition module connected to the data processing module, and the data acquisition module includes a ground observation data acquisition unit, a satellite observation data acquisition unit and a radar observation data acquisition unit. The ground observation data acquisition unit is used to collect observation data of meteorological stations at various locations on the ground; the satellite observation data acquisition unit is used to collect cloud images and other satellite remote sensing data; and the radar observation data acquisition unit is used to collect radar echo data.

6. A quantitative prediction method for extreme precipitation based on any one of claims 1 to 5, characterized in that: The following steps are involved: Step 1: Real-time data collection: Collect current environmental parameters through sensors and data collection modules deployed in different geographical locations; Step 2: Data preprocessing: clean, format, and perform time series analysis on the collected data to prepare the data set for model input; Step 3: Extreme precipitation prediction: Input the preprocessed data into the trained deep learning model and output the precipitation prediction results for a period of time in the future; Step 4: Warning information generation: Evaluate whether there is an extreme precipitation risk based on the forecast results. If an extreme precipitation event is predicted to occur, a warning information is automatically generated. Step 5: Information release: timely push prediction results and warning information to users through the user interaction interface.

7. A quantitative prediction method for extreme precipitation according to claim 6, characterized in that: In the extreme precipitation prediction link of step three, the deep learning model not only considers the environmental parameters at the current moment, but also combines the data of the same location in the past 24 hours, as well as the historical and real-time data of adjacent locations to form a multi-dimensional, spatiotemporal comprehensive prediction model.

8. The method for quantitatively predicting extreme precipitation according to claim 6, characterized in that: In the warning information generation link of step 4, when the predicted probability of extreme precipitation exceeds the preset threshold, the prediction system will automatically call the disaster assessment algorithm to assess the possible impact of the event, including but not limited to floods, landslides, and urban waterlogging, and generate corresponding warning levels according to different disaster types.

9. The method for quantitatively predicting extreme precipitation according to claim 6, characterized in that: In the information release phase of step five, users can choose the device to receive warning information according to their own needs, and the system supports customized push, such as setting different warning strategies and notification frequencies for disaster-prone areas and specific user groups.

Images