Multi-mode mountain torrent forecast integration method based on network data

Abstract

The invention relates to the technical field of mountain torrent forecasting, in particular to a multi-mode mountain torrent forecasting integration method based on network data, which comprises the following steps of: crawling weather forecasting data from a weather station weather forecasting network, and synchronously updating the weather forecasting data to a weather driving data set every 3 hours; building a digital watershed based on an SWAT model, and coupling a random forest and an LSTM algorithm to carry out rainfall-runoff simulation and forecasting; and finally, constructing an integrated system based on an MFC framework. A network data crawler technology, a digital watershed technology and a deep learning technology are adopted to realize early warning and effective management of flood disasters in a mountainous area, and on the basis of result output and display on a data watershed, runoff forecast data recognized by various modes in the future 7 days at any point of a river channel are provided, the disaster prevention and reduction efficiency of the mountainous area is improved, and the flood disaster prevention and reduction efficiency is improved. And a scientific basis is provided for flood control decision. The problems that an existing runoff forecasting system is poor in practicability and insufficient in timeliness in a mountainous area drainage basin are solved.

Description

Technical Field

[0001] The invention relates to the technical field of flash flood forecasting, and in particular to a multi-mode flash flood forecasting integration method based on network data. Background Art

[0002] Affected by the special geographical environment and climatic conditions, flood disasters frequently occur in various river basins in mountainous areas, especially in the southwestern mountainous areas. Mountain torrents are the main type of flood disasters causing casualties, and the disaster prevention situation is severe.

[0003] Runoff forecasting is the basic advance and important foundation for flood disaster prevention and control. The runoff forecasting system refers to a computer software system that includes functions such as data reception, analysis and processing, model selection, parameter calibration, operation forecasting and output of results. From the perspective of development history, it has mainly experienced three stages: "hydrological automatic measurement and reporting", "online real-time forecasting" and "interactive forecasting". The initial "hydrological automatic measurement and reporting" realized the transmission of hydrological information (such as rainfall, river water level, etc.) based on telemetry technology; "online real-time forecasting" further introduced hydrological and hydrodynamic models, and simulated and forecasted downstream runoff through hydrological and meteorological information monitored upstream; "interactive forecasting" is based on "online real-time forecasting" and adds an interactive forecasting program, realizing the use of graphic interactive processing technology and network big data technology to manually intervene in the calculation structure of the flood forecasting mathematical model, thereby obtaining the results of publishing continuous river forecasts and ensuring the continuity of flood forecasting operations.

[0004] The current runoff forecasting system has been widely and well used in cities and major rivers. However, due to the relatively slow economic development and lack of infrastructure in mountainous areas, it is difficult to accurately and timely obtain the various driving data required by the above-mentioned flood forecasting system, so these forecasting systems are less practical in mountainous watersheds. In addition, the formation of mountain floods is sudden, and coupled with the rapid convergence of mountainous terrain, there are higher timeliness requirements for the forecast of mountain runoff. At present, mountain floods and the secondary disasters they cause have become the most important work content in mountainous disaster prevention and mitigation work. The suddenness and susceptibility of mountain floods have put great pressure on mountain flood control. Summary of the invention

[0005] The purpose of the present invention is to provide a multi-mode flash flood forecasting integrated method based on network data, which solves the problems of poor practicability and insufficient timeliness of the existing runoff forecasting system in mountainous watersheds.

[0006] To achieve the above object, the present invention provides a multi-mode flash flood forecast integration method based on network data, comprising the following steps:

[0007] Obtain weather forecast data for the next 7 days from the weather station weather forecast website and synchronize it to the weather driving dataset every 3 hours;

[0008] Based on the SWAT model, a digital watershed is constructed. Taking into account the spatial heterogeneity of the water cycle in the watershed, a distributed precipitation-runoff process simulation is carried out. The precipitation-runoff relationship in the water cycle of the watershed is identified by coupling the random forest and LSTM algorithms. Furthermore, a data reading interface between the pattern recognition algorithm and the driving data set is built, and the runoff forecast of the river is carried out based on the real-time updated meteorological forecast data.

[0009] Based on the MFC framework, in accordance with the basic process of data acquisition, data processing and updating, forecast model selection, and forecast operation output, object-oriented integration technology is used to build a multi-mode flash flood forecast integration system based on network data, which automatically completes the entire process from weather forecast to flash flood forecast and provides flash flood warning information.

[0010] The weather forecast data for the next 7 days is obtained from the weather forecast website of the weather station, and is synchronously updated to the weather driving data set every 3 hours. The steps also include:

[0011] Use the requests library to send HTTP requests to the specified weather station weather forecast network API interface or web page;

[0012] If the data returned is in HTML format, use the BeautifulSoup library to parse the HTML and extract the required meteorological data;

[0013] If structured data such as JSON or XML is returned, parse the data directly;

[0014] Extract the weather forecast data for the next 7 days from the parsed data, clean it, remove invalid characters, and convert data types;

[0015] Define a unified data format, including field names and data types, according to the requirements of the meteorological driven dataset;

[0016] Convert the extracted and cleaned data into a defined data format;

[0017] Establish database connections to weather-driven datasets;

[0018] Insert the standard edited data into the corresponding table or collection in the database;

[0019] Set up scheduled tasks to execute crawler scripts regularly to update weather data in real time.

[0020] The weather forecast data for the next 7 days is obtained from the weather forecast website of the weather station, and is synchronously updated to the weather driving data set every 3 hours. The steps also include:

[0021] The weather forecast data includes precipitation, temperature, wind speed and relative humidity.

[0022] Among them, based on the SWAT model, a digital watershed is constructed, and on the basis of considering the spatial heterogeneity of the water cycle in the watershed, a distributed precipitation-runoff process simulation is carried out, and the precipitation-runoff relationship in the water cycle of the watershed is identified by coupling the random forest and LSTM algorithms. Furthermore, a data reading interface between the pattern recognition algorithm and the driving data set is built, and based on the real-time updated meteorological forecast data, the runoff forecast of the river is carried out. The steps also include:

[0023] Create a new project in the SWAT model, input the processed basic data, including land use, soil properties or topography, set the start and end time of the simulation, and select the set time step;

[0024] The watershed was divided into several sub-watersheds based on topographic data, and each sub-watershed was further subdivided into HRUs to reflect the changes in vegetation cover and soil type;

[0025] Set key parameters in the model, including evapotranspiration method, infiltration model and runoff calculation method;

[0026] Load real-time driving data, start the SWAT model for simulation, and obtain simulation results for different paths;

[0027] The forest regression model is trained using historical runoff data and real-time driving data output by the SWAT model;

[0028] Use LSTM regression model to learn from real-time updated meteorological driving data to predict future runoff changes;

[0029] The simulation results of the SWAT model, the prediction results of the forest regression model and the LSTM regression model are integrated to perform multi-model integration, and the forecast results are regularly updated based on the real-time updated meteorological data;

[0030] Based on the predicted runoff data and combined with the flood threshold of the basin, flood identification is carried out to identify the possibility, time, location and scale of flood occurrence, and output multiple modes of flood identification results, including the probability of flood occurrence and the scope of impact.

[0031] Among them, based on the SWAT model, a digital watershed is constructed, and on the basis of considering the spatial heterogeneity of the water cycle in the watershed, a distributed precipitation-runoff process simulation is carried out, and the precipitation-runoff relationship in the water cycle of the watershed is identified by coupling the random forest and LSTM algorithms. Furthermore, a data reading interface between the pattern recognition algorithm and the driving data set is built, and based on the real-time updated meteorological forecast data, the runoff forecast of the river is carried out. The steps also include:

[0032] Runoff forecasting is to set up computing nodes at random on the river channel to forecast the runoff conditions at any potential flood disaster point.

[0033] Among them, based on the MFC framework, according to the basic process of data acquisition, data processing and updating, forecast model selection, and forecast operation output, an object-oriented integration technology is adopted to build a multi-mode flash flood forecast integration system based on network data, automatically completing the whole process from weather forecast to flash flood forecast, and providing flash flood warning information. The steps also include:

[0034] Use Winform to create the software system interface, modularize the functions, separate the functions with Button controls, design different interfaces with customized user controls, and then project them to the Panel for display. Finally, integrate and develop independent functional modules.

[0035] The invention discloses a multi-mode flash flood forecasting integration method based on network data, which crawls the weather forecast data from the weather forecast website of the meteorological station, and updates it synchronously to the meteorological driving data set every 3 hours; constructs a digital watershed based on the SWAT model, couples the random forest and LSTM algorithms to simulate and forecast precipitation-runoff; and finally builds an integrated system based on the MFC framework. The network data crawler technology, the digital watershed technology and the deep learning technology are used to realize the early warning and effective management of flood disasters in mountainous areas, so as to protect the life and property safety of residents in mountainous areas. The MFC framework is used to integrate the functions of data acquisition, data processing and updating, runoff forecasting operation, result output display and the like under the support of various technologies, and construct a multi-mode integrated runoff forecasting operation under the support of network data. The data collection and processing uses the network crawler technology to crawl the weather forecast data for the next 7 days from the weather forecast website of the Meteorological Bureau, including rainfall, temperature, wind speed and humidity, etc., to overcome the problem of data scarcity in mountainous areas and update the driving data set in real time. The runoff forecasting operation is based on the open source SWAT model, constructs a distributed hydrological model of the basin, and combines random forest and LSTM learning algorithms to carry out cross-validation, providing multi-mode selection and comparison, improving forecast accuracy and the referenceability of the results. The output and display of the results are based on the data basin, providing runoff forecast data for the next 7 days at any point in the river channel, and multiple modes of recognition, providing disaster prevention and mitigation efficiency in mountainous areas, and providing a scientific basis for flood control decisions. It solves the problem that the existing runoff forecasting system is poorly practical and lacks timeliness in mountainous basins. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art are briefly introduced below.

[0037] Figure 1 It is a flow chart of the steps of the multi-mode flash flood forecast integration method based on network data of the present invention. DETAILED DESCRIPTION

[0038] Embodiments of the present invention are described in detail below. Examples of the embodiments are shown in the accompanying drawings. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present invention, but should not be construed as limiting the present invention.

[0039] The first embodiment of the present application is:

[0040] See also Figure 1 ,in, Figure 1 The present invention provides a multi-mode flash flood forecast integration method based on network data, which includes the following steps:

[0041] S101: Obtain weather forecast data for the next 7 days from the weather station weather forecast website, and synchronize and update it to the weather driving data set every 3 hours;

[0042] Specifically, the weather forecast data of precipitation, temperature, wind speed, relative humidity, etc. for the next 7 days are crawled from the weather forecast website of the weather station in real time through web crawler technology. After the data is edited in a standard format, it is automatically updated to the system's meteorological driven data set every 3 hours to overcome the problem of flood forecast data completeness caused by weak infrastructure in mountainous watersheds. Among them, the requests library is used to send HTTP requests to the specified API interface or web page. If the data returned is in HTML format, the HTML is parsed using libraries such as BeautifulSoup to extract the required meteorological data (such as date, weather conditions, temperature, wind direction, etc.). If the data returned is structured data such as JSON or XML, the data is parsed directly. The weather forecast data for the next 7 days is extracted from the parsed data, and necessary cleaning is performed, such as removing invalid characters and converting data types. According to the requirements of the meteorological driven data set, a unified data format is defined, including field names, data types, etc. The extracted and cleaned data is converted to the defined data format. The converted data is verified to ensure the accuracy and completeness of the data. A database connection with the meteorological driven data set is established. The standard edited data is inserted into the corresponding table or collection in the database. Set up a scheduled task (such as using a cron job in Linux system) to execute the crawler script regularly to update the weather data in real time.

[0043] S102: Based on the SWAT model, a digital watershed is constructed. Based on the spatial heterogeneity of the water cycle in the watershed, a distributed precipitation-runoff process simulation is carried out. The precipitation-runoff relationship in the water cycle of the watershed is identified by coupling the random forest and LSTM algorithms. Furthermore, a data reading interface between the pattern recognition algorithm and the driving data set is built, and the runoff forecast of the river is carried out based on the real-time updated meteorological forecast data.

[0044] Specifically, the SWAT model is used to simulate the physical hydrological process based on land use, soil properties, topography, etc., combined with real-time updated meteorological driving data such as precipitation, daily maximum and minimum temperatures, wind speed, relative humidity, etc., to obtain forecast results for different paths. Furthermore, by coupling the random forest and LSTM algorithms, the "precipitation-runoff" relationship in the water cycle of the basin is identified, runoff forecasts are carried out, and flood identification results of various modes are provided to improve the accuracy and reference value of the forecast. Among them, the hydrological simulation method of the SWAT digital basin: SWAT is a semi-distributed hydrological model based on physical processes. The SWAT model is driven by land use, soil properties, precipitation, daily maximum and minimum temperatures, wind speed, relative humidity, etc. The simulation of the hydrological process mainly includes precipitation, vegetation interception, surface runoff, soil flow, groundwater, snowmelt and evapotranspiration data. The SWAT model divides the watershed into sub-basins connected by river networks. In each sub-basin, the uneven distribution of underlying surface factors is considered, and different hydrological response units (HRUs) are determined by three attribute values: land use, soil type, and slope. HRU is the smallest calculation unit of SWAT. The spatial position relationship between HRUs is not considered. Each HRU is independently calculated based on the water balance equation. In the river channel, the variable water storage coefficient model or Muskingum method is used to calculate the river confluence, and the flow of the basin outlet section is obtained by deducting water surface evaporation, confluence loss, and water use in the basin. Deep learning "precipitation-runoff" relationship identification method: including random forest regression and LSTM regression algorithm, is an algorithm based on ensemble learning. It integrates the prediction results of multiple decision trees to establish a regression relationship, and has no special requirements for data distribution and linear relationships between factors. In RF regression, each decision tree is trained independently and randomly selected subsamples, so it has good generalization ability and can effectively reduce the risk of overfitting. It is suitable for data with complex nonlinear relationships such as precipitation and explanatory variables. LSTM is a special recurrent neural network architecture for processing sequence data. It controls the flow of information through gating units, including forgetting gates, input gates, and output gates. This enables LSTM to capture long-term dependencies more effectively and is suitable for tasks such as language models and time series prediction. Runoff forecasting is to set up computing nodes arbitrarily on the river channel to forecast the runoff conditions of any potential flood disaster point.

[0045] S103: Based on the MFC framework, in accordance with the basic process of data acquisition, data processing and updating, forecast model selection, and forecast operation output, object-oriented integration technology is used to build a multi-mode flash flood forecast integration system based on network data, automatically completing the entire process from weather forecast to flash flood forecast, and providing flash flood warning information.

[0046] Specifically, according to the functional requirements of the flash flood forecasting system, based on the MFC framework, in accordance with the basic process of data acquisition, data processing and updating, forecast model selection, and forecast operation output, object-oriented integration technology is adopted, and Winform is used to create the software system interface. Each functional module is distinguished by Button controls, and different interfaces are designed using custom user controls, and then projected to the Panel for display. Finally, independent functional modules are integrated and developed to build a multi-mode flash flood forecasting integrated system based on network data. System development method for integrating multiple functional modules: Using Visual Studio as the development environment and C# as the development language, Winform is used to create the software system interface. Each function is modularized, and the functions are distinguished by Button controls. Different interfaces are designed using custom user controls, and then projected to the Panel for display. Finally, independent functional modules are integrated and developed.

[0047] The method of the present invention adopts network data crawler technology, digital watershed technology and deep learning technology to achieve early warning and effective management of flood disasters in mountainous areas to ensure the safety of life and property of residents in mountainous areas. The MFC framework is adopted to integrate the functions of data acquisition, data processing and updating, runoff forecasting operation, result output and display under various technical support, and build a multi-mode integrated runoff forecasting operation under network data support. Data collection and processing use network crawler technology to capture meteorological forecast data for the next 7 days from the weather forecast network of the Meteorological Bureau, including rainfall, temperature, wind speed and humidity, etc., to overcome the problem of lack of data in mountainous areas and update the driving data set in real time. The runoff forecasting operation is based on the open source SWAT model to build a distributed hydrological model of the basin, and combines random forest and LSTM learning algorithms to carry out cross-validation, provide multi-mode selection and comparison, and improve the forecast accuracy and referenceability of the results. The result output and display provide runoff forecast data for the next 7 days at any point in the river channel, and multiple modes are identified on the basis of the data basin, providing the efficiency of disaster prevention and mitigation in mountainous areas, and providing a scientific basis for flood control decision-making. It solves the problem that the existing runoff forecasting system has poor practicality and insufficient timeliness in mountainous watersheds.

[0048] What is disclosed above is only one or more preferred embodiments of the present application, and cannot be used to limit the scope of rights of the present application. Ordinary technicians in this field can understand that all or part of the processes of implementing the above embodiments and equivalent changes made according to the claims of the present application are still within the scope covered by the present application.

Claims

1. A multi-mode flash flood forecasting integration method based on network data, characterized in that: The following steps are involved: Obtain weather forecast data for the next 7 days from the weather station weather forecast website and synchronize it to the weather driving dataset every 3 hours; Based on the SWAT model, a digital watershed is constructed. Taking into account the spatial heterogeneity of the water cycle in the watershed, a distributed precipitation-runoff process simulation is carried out. The precipitation-runoff relationship in the water cycle of the watershed is identified by coupling the random forest and LSTM algorithms. Furthermore, a data reading interface between the pattern recognition algorithm and the driving data set is built, and the runoff forecast of the river is carried out based on the real-time updated meteorological forecast data. Based on the MFC framework, in accordance with the basic process of data acquisition, data processing and updating, forecast model selection, and forecast operation output, object-oriented integration technology is used to build a multi-mode flash flood forecast integration system based on network data, which automatically completes the entire process from weather forecast to flash flood forecast and provides flash flood warning information.

2. The multi-mode flash flood forecasting integration method based on network data as claimed in claim 1, characterized in that: Obtain weather forecast data for the next 7 days from the weather station weather forecast website, and update it to the weather driving data set every 3 hours. The steps also include: Use the requests library to send HTTP requests to the specified weather station weather forecast network API interface or web page; If the data returned is in HTML format, use the BeautifulSoup library to parse the HTML and extract the required meteorological data; If structured data such as JSON or XML is returned, parse the data directly; Extract the weather forecast data for the next 7 days from the parsed data, clean it, remove invalid characters, and convert data types; Define a unified data format, including field names and data types, according to the requirements of the meteorological driven dataset; Convert the extracted and cleaned data into a defined data format; Establish database connections to weather-driven datasets; Insert the standard edited data into the corresponding table or collection in the database; Set up scheduled tasks to execute crawler scripts regularly to update weather data in real time.

3. The multi-mode flash flood forecasting integration method based on network data as claimed in claim 2, characterized in that: Obtain weather forecast data for the next 7 days from the weather station weather forecast website, and update it to the weather driving data set every 3 hours. The steps also include: The weather forecast data includes precipitation, temperature, wind speed and relative humidity.

4. The multi-mode flash flood forecasting integration method based on network data as claimed in claim 1, characterized in that: Based on the SWAT model, a digital watershed is constructed, and a distributed precipitation-runoff process simulation is carried out on the basis of considering the spatial heterogeneity of the water cycle in the watershed. The precipitation-runoff relationship in the watershed water cycle is identified by coupling the random forest and LSTM algorithms. Furthermore, a data reading interface between the pattern recognition algorithm and the driving data set is built, and the runoff forecast of the river is carried out based on the real-time updated meteorological forecast data. The steps also include: Create a new project in the SWAT model, input the processed basic data, including land use, soil properties or topography, set the start and end time of the simulation, and select the set time step; The watershed was divided into several sub-watersheds based on topographic data, and each sub-watershed was further subdivided into HRUs to reflect the changes in vegetation cover and soil type; Set key parameters in the model, including evapotranspiration method, infiltration model and runoff calculation method; Load real-time driving data, start the SWAT model for simulation, and obtain simulation results for different paths; The forest regression model is trained using historical runoff data and real-time driving data output by the SWAT model; Use LSTM regression model to learn from real-time updated meteorological driving data to predict future runoff changes; The simulation results of the SWAT model, the prediction results of the forest regression model and the LSTM regression model are integrated to perform multi-model integration, and the forecast results are regularly updated based on the real-time updated meteorological data; Based on the predicted runoff data and combined with the flood threshold of the basin, flood identification is carried out to identify the possibility, time, location and scale of flood occurrence, and output multiple modes of flood identification results, including the probability of flood occurrence and the scope of impact.

5. The multi-mode flash flood forecasting integration method based on network data as claimed in claim 4, characterized in that: Based on the SWAT model, a digital watershed is constructed, and a distributed precipitation-runoff process simulation is carried out on the basis of considering the spatial heterogeneity of the water cycle in the watershed. The precipitation-runoff relationship in the watershed water cycle is identified by coupling the random forest and LSTM algorithms. Furthermore, a data reading interface between the pattern recognition algorithm and the driving data set is built, and the runoff forecast of the river is carried out based on the real-time updated meteorological forecast data. The steps also include: Runoff forecasting is to set up computing nodes at random on the river channel to forecast the runoff conditions at any potential flood disaster point.

6. The multi-mode flash flood forecasting integration method based on network data as claimed in claim 5, characterized in that: Based on the MFC framework, in accordance with the basic process of data acquisition, data processing and updating, forecast model selection, and forecast operation output, an object-oriented integration technology is used to build a multi-mode flash flood forecast integration system based on network data, which automatically completes the entire process from weather forecast to flash flood forecast and provides flash flood warning information. The steps also include: Use Winform to create the software system interface, modularize the functions, separate the functions with Button controls, design different interfaces with customized user controls, and then project them to the Panel for display. Finally, integrate and develop independent functional modules.

Images