Complex disastrous weather characteristic parameter extraction system

Abstract

The invention provides a complex disastrous weather characteristic parameter extraction system, which relates to the technical field of meteorological information processing, and can obtain a meteorological data source through a data acquisition module, a data preprocessing module, a disaster type identification module, a characteristic parameter extraction module and an early warning report module. The method comprises the steps of collecting original meteorological data of a meteorological data source in real time, performing preprocessing operation, generating corresponding target meteorological data, performing identification and judgment on the target meteorological data based on a convolutional neural network, obtaining a corresponding disastrous weather type, calling a preset feature parameter library, constructing a feature parameter extraction model, and extracting the disastrous weather according to the feature parameter extraction model. And determining key feature parameters corresponding to the disastrous weather type, generating corresponding disastrous weather forecast and early warning information according to the key feature parameters, and sending and displaying the disastrous weather forecast and early warning information to the management end. According to the method, corresponding characteristic parameters can be efficiently, accurately and automatically identified and extracted for different types of complex disastrous weather, and a scientific basis is provided for meteorological early warning and disaster prevention and reduction.

Description

Technical Field

[0001] The present invention relates to the technical field of meteorological information processing, and in particular to a system for extracting characteristic parameters of complex disastrous weather. Background Art

[0002] In recent years, global climate change has led to increasingly frequent and complex disastrous weather phenomena, such as super typhoons, continuous heavy rains, long-term droughts, severe thunderstorms, and large-scale sandstorms. Extreme weather events have gradually had a serious impact on human society and economic activities. These disastrous weather not only threaten people's lives and property, but also have a profound impact on many key areas such as agriculture, transportation, energy, and water resources management. Each type of disastrous weather has unique physical and chemical characteristics, and the corresponding characteristic parameters can reflect the essential attributes of disastrous weather. For example, the intensity and path of typhoons are affected by meteorological parameters such as atmospheric circulation, sea temperature distribution, and water vapor transport; heavy rains are closely related to parameters such as convective intensity, cloud structure, and precipitation efficiency; drought phenomena are often associated with parameters such as long-term low precipitation, decreased soil moisture, and increased surface temperature. By monitoring and analyzing the changes in these meteorological characteristic parameters, the occurrence and development trends of disastrous weather can be more accurately predicted, providing a scientific basis for disaster prevention and mitigation.

[0003] In the prior art, traditional meteorological observation and analysis methods generally rely on manual intuitive judgment and long-term experience accumulation. This method has significant limitations when dealing with complex and changeable meteorological characteristic parameters, and it is difficult to comprehensively collect and analyze meteorological data in real time. Due to the subjectivity of manual judgment and the non-standardization of experience accumulation, meteorological characteristic parameters are often not fully and accurately captured, which greatly limits the accuracy and timeliness of disastrous weather forecasts. Especially in the face of extreme weather events, this traditional method that relies on manual labor is often difficult to meet the urgent needs of modern society for high-precision weather forecasts and rapid response.

[0004] Therefore, it is necessary to provide a complex disastrous weather characteristic parameter extraction system to solve the above technical problems. Summary of the invention

[0005] In order to solve the above technical problems, the present invention provides a complex disastrous weather characteristic parameter extraction system to solve the problem that traditional meteorological observation and analysis methods often rely on manual judgment and experience accumulation, and it is difficult to capture complex and changeable meteorological characteristic parameters comprehensively and in real time, which greatly limits the accuracy and timeliness of disastrous weather forecasts.

[0006] The present invention provides a complex disastrous weather characteristic parameter extraction system, comprising:

[0007] A data acquisition module is used to obtain a meteorological data source and collect raw meteorological data from the meteorological data source in real time;

[0008] A data preprocessing module, used for performing preprocessing operations on the original meteorological data to generate corresponding target meteorological data;

[0009] A disaster type identification module is used to identify and judge the target meteorological data based on a convolutional neural network to obtain the corresponding disastrous weather type;

[0010] A feature parameter extraction module is used to retrieve a preset feature parameter library, construct a feature parameter extraction model, and determine key feature parameters corresponding to the disastrous weather type;

[0011] The warning report module is used to generate corresponding disastrous weather forecast and warning information according to the key characteristic parameters, and send and display it to the management end.

[0012] Preferably, the raw meteorological data includes temperature, humidity, wind speed, wind direction, air pressure, precipitation, cloud cover, visibility and lightning activity data.

[0013] Preferably, the data preprocessing module specifically includes:

[0014] Removing invalid data from the original meteorological data to generate first meteorological data;

[0015] Smoothing the first meteorological data based on a smoothing filtering algorithm to remove noise in the first meteorological data and generate second meteorological data;

[0016] The missing values ​​in the second meteorological data are filled based on a linear interpolation algorithm to generate the target meteorological data.

[0017] Preferably, the disaster type identification module is used to identify and judge the target meteorological data based on a convolutional neural network to obtain the corresponding disastrous weather type, specifically including:

[0018] The target meteorological data is input into the convolution layer in the convolutional neural network, the meteorological feature pattern in the target meteorological data is extracted, and the corresponding initial meteorological feature is generated. The expression of the initial meteorological feature is:

[0019] CSTZ=W0*TZMS+b0

[0020] Where CSTZ represents the initial meteorological characteristics; W0 represents the weight matrix corresponding to the convolution layer; TZMS represents the meteorological characteristic pattern in the target meteorological data; b0 represents the bias vector corresponding to the convolution layer;

[0021] The initial meteorological features are input into the activation layer of the convolutional neural network to generate corresponding intermediate meteorological features. The expression of the intermediate meteorological features is:

[0022] ZJTZ=max(0,CSTZ)

[0023] In the formula, ZJTZ represents the intermediate meteorological characteristics; max() represents the maximum value operation; CSTZ represents the initial meteorological characteristics;

[0024] The intermediate meteorological features are input into the pooling layer in the convolutional neural network to reduce the dimension of the intermediate meteorological features, the dimension including height and width, and the intermediate meteorological features are subjected to maximum pooling processing to generate corresponding key meteorological features, the expression of the key meteorological features is:

[0025]

[0026] In the formula, GJTZ ij Represents the key meteorological characteristics at location index (i, j); ZJTZ mn Represents the intermediate meteorological characteristics at the location index (m, n); max() represents the maximum value operation; R ij Represents the area of ​​the pooling window corresponding to the position index (i, j) in the key meteorological feature on the intermediate meteorological feature.

[0027] Preferably, the key meteorological features are input into the fully connected layer in the convolutional neural network to generate the corresponding predicted disaster type, and the predicted disaster type is processed based on the activation function to obtain the corresponding predicted disaster occurrence probability, specifically including:

[0028] The expression for predicting the disaster type is:

[0029] ZHLX=W1*GJTZ+b1

[0030] Where ZHLX represents the predicted disaster type; W1 represents the weight matrix corresponding to the fully connected layer; GJTZ represents the key meteorological features; b1 represents the bias vector corresponding to the fully connected layer;

[0031] The expression for predicting the probability of disaster occurrence is:

[0032] ZHGL=σ(ZHLX)

[0033] In the formula, ZHGL represents the predicted probability of disaster occurrence; σ represents the activation function; ZHLX represents the predicted disaster type;

[0034] According to the magnitude of the predicted disaster occurrence probability, the corresponding predicted disaster types are sorted from large to small, the preset number of disaster types is obtained, and the predicted disaster type whose serial number is less than or equal to the preset number of disaster types is selected as the disastrous weather type.

[0035] Preferably, the characteristic parameter extraction module is used to retrieve a preset characteristic parameter library, construct a characteristic parameter extraction model, and determine the key characteristic parameters corresponding to the disastrous weather type, specifically including:

[0036] The preset characteristic parameter library is retrieved, the characteristic parameter extraction model is constructed, and the correlation between the disastrous weather type and each characteristic parameter in the preset characteristic parameter library is determined based on the Pearson correlation coefficient calculation formula. The expression of the correlation degree is:

[0037]

[0038] Wherein, D(ZHTQ,TZCS) represents the correlation between the type of disastrous weather and each characteristic parameter in the preset characteristic parameter library; ZHTQ represents the type of disastrous weather; TZCS represents each characteristic parameter in the preset characteristic parameter library; n represents the number of characteristic parameters in the preset characteristic parameter library; Σ represents the summation symbol;

[0039] According to the magnitude of the correlation, the corresponding characteristic parameters are sorted from large to small, a preset number of characteristic parameters is obtained, and the characteristic parameters whose sequence numbers are less than or equal to the preset number of characteristic parameters are selected as the key characteristic parameters.

[0040] Preferably, the warning information includes the disaster level, possible affected areas and disaster response measures corresponding to the disastrous weather type.

[0041] Compared with the related art, the complex disastrous weather characteristic parameter extraction system provided by the present invention has the following beneficial effects:

[0042] The present invention can obtain meteorological data sources through data acquisition modules, data preprocessing modules, disaster type identification modules, characteristic parameter extraction modules and early warning reporting modules, and collect the original meteorological data of meteorological data sources in real time, perform preprocessing operations on the original meteorological data, generate corresponding target meteorological data, identify and judge the target meteorological data based on convolutional neural networks, obtain the corresponding disastrous weather types, call the preset characteristic parameter library, build a characteristic parameter extraction model, determine the key characteristic parameters corresponding to the disastrous weather types, generate corresponding disastrous weather forecasts and early warning information according to the key characteristic parameters, and send and display them to the management end. The present invention can extract characteristic parameters more accurately for different complex disastrous weather through automatic identification algorithms and dynamic adjustment strategies, thereby improving the accuracy of disastrous weather forecasts; and the system of the present invention can process and analyze meteorological data in real time, quickly generate early warning information, and shorten the response time of disaster warnings; at the same time, the present invention can provide comprehensive and timely disastrous weather information for meteorological departments, emergency management departments and the public, which is helpful to formulate effective disaster prevention and mitigation strategies; and can promote the innovation and development of meteorological information processing technology, and provide reference and reference for meteorological applications in other fields. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 A system block diagram of a complex disastrous weather characteristic parameter extraction system of the present invention;

[0044] Figure 2 A schematic diagram of a disastrous weather forecast and warning information display of a complex disastrous weather characteristic parameter extraction system of the present invention;

[0045] Figure 3 A flowchart of data preprocessing operations of a complex disastrous weather characteristic parameter extraction system of the present invention;

[0046] Figure 4 It is a schematic diagram of the classification of warning information of a complex disastrous weather characteristic parameter extraction system of the present invention. DETAILED DESCRIPTION

[0047] The present invention will be further described below in conjunction with the accompanying drawings and implementation modes.

[0048] Embodiment 1

[0049] like Figure 1 As shown, a complex disastrous weather characteristic parameter extraction system includes:

[0050] A data acquisition module is used to obtain a meteorological data source and collect raw meteorological data from the meteorological data source in real time;

[0051] The data acquisition module integrates sensor interface technology and network communication protocols to access multiple data sources such as ground observation stations, satellite remote sensing, radar detection, and numerical weather forecast models in real time and stably, thereby ensuring the comprehensiveness and timeliness of the collected meteorological information. The collected meteorological data includes multiple meteorological elements such as temperature, humidity, air pressure, wind speed, wind direction, and precipitation, which facilitates the subsequent analysis and processing of meteorological data.

[0052] A data preprocessing module, used for performing preprocessing operations on the original meteorological data to generate corresponding target meteorological data;

[0053] It can be understood that the data preprocessing module can effectively reduce the noise interference in the original meteorological data, enhance the comparability and consistency of meteorological data, and obtain target meteorological data with standardized format and reliable quality through operations such as data cleaning, data format conversion, and data normalization or standardization.

[0054] Specifically, data cleaning operations include removing outliers, filling missing values, etc., and data format conversion operations include unified data encoding, timestamp alignment, etc.

[0055] A disaster type identification module is used to identify and judge the target meteorological data based on a convolutional neural network to obtain the corresponding disastrous weather type;

[0056] It should be noted that based on the convolutional neural network (CNN) in deep learning technology, the disaster type recognition module can perform in-depth analysis and processing on the pre-processed target meteorological data through its feature extraction and pattern recognition capabilities. By training a large amount of historical disaster weather case data, the disaster type recognition module can automatically identify and classify various types of disastrous weather such as rainstorms, typhoons, droughts and cold waves, and can ensure the accuracy of the recognition results, improve the generalization ability of the disaster recognition module, and provide a scientific basis for subsequent disaster warnings.

[0057] A feature parameter extraction module is used to retrieve a preset feature parameter library, construct a feature parameter extraction model, and determine key feature parameters corresponding to the disastrous weather type;

[0058] Among them, based on the preset feature parameter library, by constructing a feature parameter extraction model, the feature parameter extraction module can automatically match and screen out the key feature parameters corresponding to the current disastrous weather type, so that the disaster intensity and disaster development trend corresponding to the disastrous weather type can be evaluated based on these key feature parameters.

[0059] Specifically, the preset characteristic parameter library refers to the parameter library to which various characteristic parameters preset in advance by the staff belong, and the preset characteristic parameter library includes the key physical quantities and their threshold ranges corresponding to various types of disastrous weather, such as the accumulated precipitation of heavy rain, the wind speed and path information of typhoons, etc.

[0060] The warning report module is used to generate corresponding disastrous weather forecast and warning information according to the key characteristic parameters, and send and display it to the management end.

[0061] like Figure 2 As shown in the figure, the early warning report module can automatically generate detailed disastrous weather forecasts and warning information based on the extracted key feature parameters using statistical methods, and these warning information include disastrous weather forecasts, warning levels, impact ranges, and recommended measures, etc. The early warning report module ensures that the warning information can be sent quickly and accurately to management terminals at all levels through the efficient data transmission protocol and secure data exchange mechanism of the terminal equipment, and is intuitively displayed through a visual interface, providing certain information guarantees for decision support and emergency response.

[0062] In the specific implementation process, the original meteorological data includes temperature, humidity, wind speed, wind direction, air pressure, precipitation, cloud cover, visibility and lightning activity data.

[0063] In practical applications, the characteristic parameters of complex disastrous weather include the following four aspects:

[0064] The first is the characteristic parameters related to precipitation, including:

[0065] Heavy rain refers to daily rainfall reaching or exceeding 50 mm. Specifically, when daily rainfall reaches or exceeds 100 mm, it is called heavy rain; when daily rainfall reaches or exceeds 200 mm, it is called extremely heavy rain.

[0066] The precipitation deviation percentage refers to an index for assessing whether the precipitation is too little or too much, and the severity of drought or flood. The specific calculation method is the ratio of the precipitation in a certain place and time period to the average precipitation in the same period of the year. When the ratio is greater than or equal to 20%, it indicates flood; when the ratio is less than a certain value, it indicates drought.

[0067] The second is the characteristic parameters related to wind, including:

[0068] A gale is defined as an instantaneous maximum wind speed reaching or exceeding 16 m / s, i.e., above level 7. A strong wind is defined as an average wind speed reaching or exceeding 8 m / s, i.e., above level 6. A squall is defined as a strong wind that is actually strong and is often accompanied by a change in wind direction, an increase in wind force, and a sharp drop in temperature. It is often accompanied by thunderstorms, hail, and other weather conditions. A tornado is defined as a violent weather condition such as rotating winds caused by the drooping funnel cloud at the bottom of a cumulonimbus cloud touching the ground.

[0069] The third is the characteristic parameters related to temperature, including:

[0070] A cold wave refers to a strong cold front that causes the average daily temperature to drop by more than 8°C in one day, or more than 10°C in two days, and the lowest temperature drops below 5°C. Low temperature refers to an extreme minimum temperature of 5°C or below. Frost refers to a short-term low temperature weather in the warm season, when the average temperature is above 0°C, and the temperature of the soil surface, plant surface or air layer close to the ground drops to a level high enough to cause damage or death to plants. High temperature refers to a summer maximum temperature of 36°C or above.

[0071] The fourth is other related characteristic parameters, including:

[0072] Dry hot wind refers to a dry and hot wind that often occurs in the warm season of late spring and early summer. Generally, the maximum temperature is greater than or equal to 25℃, the minimum relative humidity is less than or equal to 30%, and the wind speed is greater than or equal to 4-5 meters per second as the standard of dry hot wind. Hail refers to the solid precipitation phenomenon produced by strong storm systems such as cumulonimbus clouds that develop strongly and have a long life span. Fog refers to the weather in which the air is saturated with humidity, water vapor condenses near the ground, and visibility is reduced to less than 1 km. It often occurs in humid and windless conditions.

[0073] like Figure 3 As shown, the data preprocessing module specifically includes:

[0074] Removing invalid data from the original meteorological data to generate first meteorological data;

[0075] Smoothing the first meteorological data based on a smoothing filtering algorithm to remove noise in the first meteorological data and generate second meteorological data;

[0076] The missing values ​​in the second meteorological data are filled based on a linear interpolation algorithm to generate the target meteorological data.

[0077] The data preprocessing module is used to process the original meteorological data to ensure that the data quality meets the needs of subsequent analysis and application. Through the data cleaning operation, invalid data in the original meteorological data can be removed. These invalid data may include abnormal values ​​or null values ​​caused by sensor failure, transmission errors or incomplete data records, and then the corresponding first meteorological data can be obtained.

[0078] Smoothing filtering is a mathematical method commonly used in signal processing and time series analysis. It can effectively remove high-frequency noise components in data and retain useful low-frequency signals.

[0079] In order to further improve the data quality, a smoothing filtering algorithm can be applied to perform corresponding smoothing processing on the first meteorological data, and then the corresponding second meteorological data can be obtained, thereby retaining the original information while ensuring that the processed meteorological data has higher smoothness and accuracy.

[0080] Linear interpolation is a method of estimating unknown data points based on the linear relationship between known data points. It can effectively estimate and fill in the blank areas in the time series, thereby generating complete and coherent target meteorological data.

[0081] In view of the problem of missing values ​​that may exist in the second meteorological data, the present invention adopts a linear interpolation algorithm to fill the missing values, thereby improving the integrity of the meteorological data and providing a reliable data basis for subsequent meteorological analysis, prediction and decision support.

[0082] Through the above methods, the original meteorological data can be effectively cleaned, smoothed and missing values ​​filled to generate high-quality target meteorological data, ensuring the accuracy and comprehensiveness of the meteorological data.

[0083] The disaster type identification module is used to identify and judge the target meteorological data based on a convolutional neural network to obtain the corresponding disastrous weather type, specifically including:

[0084] The target meteorological data is input into the convolution layer in the convolutional neural network, the meteorological feature pattern in the target meteorological data is extracted, and the corresponding initial meteorological feature is generated. The expression of the initial meteorological feature is:

[0085] CSTZ=W0*TZMS+b0

[0086] Where CSTZ represents the initial meteorological characteristics; W0 represents the weight matrix corresponding to the convolution layer; TZMS represents the meteorological characteristic pattern in the target meteorological data; b0 represents the bias vector corresponding to the convolution layer;

[0087] The initial meteorological features are input into the activation layer of the convolutional neural network to generate corresponding intermediate meteorological features. The expression of the intermediate meteorological features is:

[0088] ZJTZ=max(0,CSTZ)

[0089] In the formula, ZJTZ represents the intermediate meteorological characteristics; max() represents the maximum value operation; CSTZ represents the initial meteorological characteristics;

[0090] The intermediate meteorological features are input into the pooling layer in the convolutional neural network to reduce the dimension of the intermediate meteorological features, the dimension including height and width, and the intermediate meteorological features are subjected to maximum pooling processing to generate corresponding key meteorological features, the expression of the key meteorological features is:

[0091]

[0092] In the formula, GJTZ ij Represents the key meteorological characteristics at location index (i, j); ZJTZ mn Represents the intermediate meteorological characteristics at the location index (m, n); max() represents the maximum value operation; R ij Represents the area of ​​the pooling window corresponding to the position index (i, j) in the key meteorological feature on the intermediate meteorological feature.

[0093] In this embodiment, a target meteorological data set may include observation values ​​of multiple meteorological elements within a period of time, and these meteorological elements may be temperature, humidity, wind speed, air pressure, etc. Based on the convolutional neural network (CNN), the target meteorological data set can be identified and judged to obtain the corresponding disastrous weather types, such as heavy rain, typhoon, and drought.

[0094] Specifically, the target meteorological data set can first be input into the convolution layer of the convolutional neural network, and a convolution operation can be performed on the input target meteorological data through a set of weight matrices, i.e., convolution kernels. Assume that there is a weight matrix W0, which represents the meteorological feature pattern to be detected, such as a specific temperature distribution pattern. Through the convolution operation, i.e., the dot product operation between the weight matrix W0 and the local area in the target meteorological data, the corresponding meteorological feature pattern can be extracted and the corresponding initial meteorological feature CSTZ can be generated.

[0095] Furthermore, the initial meteorological feature CSTZ can be input into the activation layer of the convolutional neural network. The activation layer can transform the input initial meteorological feature CSTZ through a nonlinear function to introduce nonlinear factors and enhance the expression ability of the network. In the nonlinear function, each initial meteorological feature CSTZ will be processed separately. Specifically, if the eigenvalue is greater than 0, it remains unchanged; if the eigenvalue is less than or equal to 0, the output is 0. In the above manner, the corresponding intermediate meteorological feature ZJTZ can be generated, which retains important information while being easier to process in the subsequent pooling layer.

[0096] Finally, the intermediate meteorological features ZJTZ can be input into the pooling layer of the convolutional neural network. The pooling layer can reduce the dimension of the features, i.e., height and width, and retain the most important information. In the maximum pooling operation, the convolutional neural network can select the largest feature value in each pooling window as the output, and then generate the corresponding key meteorological features. These key meteorological features can reduce the computational complexity and overfitting risk of the pooling process while retaining the most important information.

[0097] Through the above method, the convolutional neural network can extract key meteorological features from the target meteorological data, and identify and judge the type of disastrous weather based on these key meteorological features to obtain the corresponding type of disastrous weather. Then, corresponding meteorological warnings and response measures can be taken to reduce the impact of disastrous weather on human society and the natural environment.

[0098] The key meteorological features are input into the fully connected layer in the convolutional neural network to generate the corresponding predicted disaster type, and the predicted disaster type is processed based on the activation function to obtain the corresponding predicted disaster occurrence probability, specifically including:

[0099] The expression for predicting the disaster type is:

[0100] ZHLX=W1*GJTZ+b1

[0101] Where ZHLX represents the predicted disaster type; W1 represents the weight matrix corresponding to the fully connected layer; GJTZ represents the key meteorological features; b1 represents the bias vector corresponding to the fully connected layer;

[0102] The expression for predicting the probability of disaster occurrence is:

[0103] ZHGL=σ(ZHLX)

[0104] In the formula, ZHGL represents the predicted probability of disaster occurrence; σ represents the activation function; ZHLX represents the predicted disaster type;

[0105] According to the magnitude of the predicted disaster occurrence probability, the corresponding predicted disaster types are sorted from large to small, the preset number of disaster types is obtained, and the predicted disaster type whose serial number is less than or equal to the preset number of disaster types is selected as the disastrous weather type.

[0106] Among them, the key meteorological features GJTZ can be firstly input into the fully connected layer in the convolutional neural network (CNN). Through the feature extraction capability of the fully connected layer in the convolutional neural network (CNN), the key meteorological features GJTZ can be nonlinearly combined to generate the corresponding predicted disaster type ZHLX.

[0107] Specifically, the expression for predicting the disaster type ZHLX is: ZHLX=W1*GJTZ+b1, where ZHLX represents the vector corresponding to the predicted disaster type; W1 represents the weight matrix corresponding to the fully connected layer, which can be continuously optimized through the learning process to map the input features to the disaster type space in the best way; GJTZ represents the vector corresponding to the key meteorological features; b1 represents the bias vector corresponding to the fully connected layer, which is used to adjust the reference position of the output result.

[0108] Furthermore, the predicted disaster type ZHLX can be processed by the activation function to obtain the corresponding predicted disaster probability ZHGL. The expression of the predicted disaster probability ZHGL is: ZHGL = σ(ZHLX), where ZHGL represents the vector corresponding to the predicted disaster probability; σ represents the activation function, which can map any real value to the interval [0,1], thereby reflecting the possibility of disaster occurrence; ZHLX represents the vector corresponding to the predicted disaster type.

[0109] Finally, according to the predicted probability of disaster occurrence, the corresponding predicted disaster types can be sorted in descending order. According to the pre-set number of disaster types, the predicted disaster types with serial numbers less than or equal to this number can be selected as the final disastrous weather type output, so that the most urgent meteorological disaster risks can be quickly identified and responded to in practical applications.

[0110] The characteristic parameter extraction module is used to retrieve a preset characteristic parameter library, construct a characteristic parameter extraction model, and determine the key characteristic parameters corresponding to the disastrous weather type, specifically including:

[0111] The preset characteristic parameter library is retrieved, the characteristic parameter extraction model is constructed, and the correlation between the disastrous weather type and each characteristic parameter in the preset characteristic parameter library is determined based on the Pearson correlation coefficient calculation formula. The expression of the correlation degree is:

[0112]

[0113] Wherein, D(ZHTQ,TZCS) represents the correlation between the type of disastrous weather and each characteristic parameter in the preset characteristic parameter library; ZHTQ represents the type of disastrous weather; TZCS represents each characteristic parameter in the preset characteristic parameter library; n represents the number of characteristic parameters in the preset characteristic parameter library; Σ represents the summation symbol;

[0114] According to the magnitude of the correlation, the corresponding characteristic parameters are sorted from large to small, a preset number of characteristic parameters is obtained, and the characteristic parameters whose sequence numbers are less than or equal to the preset number of characteristic parameters are selected as the key characteristic parameters.

[0115] It should be noted that, firstly, the preset characteristic parameter library and characteristic parameter extraction model that have been constructed can be called up to accurately identify the key characteristic parameters corresponding to a specific type of disastrous weather. The preset characteristic parameter library includes physical quantities and statistical indicators related to disastrous weather, such as wind speed, precipitation, temperature, humidity, and air pressure change rate.

[0116] It can be understood that based on the Pearson correlation coefficient calculation formula, the linear correlation between the type of disastrous weather and each characteristic parameter in the preset characteristic parameter library can be quantified. The value range of the Pearson correlation coefficient is between -1 and 1, and the closer its absolute value is to 1, the stronger the linear relationship between the two variables.

[0117] The calculation formula for the correlation between the corresponding disastrous weather type and each characteristic parameter in the preset characteristic parameter library is: Among them, D(ZHTQ, TZCS) reflects the correlation between the disastrous weather type and the characteristic parameters in the preset characteristic parameter library, ZHTQ represents the disastrous weather type; TZCS represents the characteristic parameters in the preset characteristic parameter library; n represents the number of characteristic parameters in the preset characteristic parameter library.

[0118] After obtaining the degree of correlation between the disastrous weather type and each characteristic parameter in the preset characteristic parameter library, the characteristic parameters can be sorted in descending order according to the degree of correlation, so as to screen out the key characteristic parameters with a high degree of correlation with the disastrous weather type. Specifically, a threshold value of the number of characteristic parameters can be preset according to actual needs, and characteristic parameters with serial numbers less than or equal to the threshold can be selected as key characteristic parameters. These key characteristic parameters can reflect the essential characteristics of disastrous weather to the greatest extent, and provide a scientific basis for subsequent weather warnings, risk assessments and emergency responses.

[0119] like Figure 4 As shown, the warning information includes the disaster level corresponding to the disastrous weather type, the possible affected area and disaster response measures.

[0120] In actual applications, early warning information includes disaster level assessments corresponding to disastrous weather types, analysis of potential impact areas, and planning of disaster response measures. Specifically, disaster level refers to a quantitative indicator of disaster intensity derived through scientific algorithms based on meteorological principles and historical disaster data, which is used to intuitively reflect the degree of damage and scope of impact that may be caused by disasters. The possible affected area refers to the geographical scope that may be affected by disasters, which is accurately delineated through geographic information system (GIS) technology combined with meteorological model predictions. It can provide accurate disaster risk assessments for relevant departments. Disaster response measures refer to comprehensive disaster prevention and mitigation strategies, including emergency evacuation plans, emergency material allocation, infrastructure reinforcement, and post-disaster recovery and reconstruction plans, which can minimize the losses caused by complex disastrous weather and ensure the safety of people's lives and property.

[0121] The present invention can obtain a meteorological data source through a complex disastrous weather characteristic parameter extraction system, a data acquisition module, a data preprocessing module, a disaster type identification module, a characteristic parameter extraction module and an early warning report module, and collect the original meteorological data of the meteorological data source in real time, perform preprocessing operations on the original meteorological data, generate corresponding target meteorological data, identify and judge the target meteorological data based on a convolutional neural network, obtain the corresponding disastrous weather type, call a preset characteristic parameter library, build a characteristic parameter extraction model, determine the key characteristic parameters corresponding to the disastrous weather type, generate corresponding disastrous weather forecast and early warning information according to the key characteristic parameters, and send and display them to a management terminal. The present invention can extract characteristic parameters more accurately for different disastrous weather through automatic identification algorithms and dynamic adjustment strategies, thereby improving the accuracy of disastrous weather forecasts; and the system of the present invention can process and analyze meteorological data in real time, quickly generate warning information, and shorten the disaster warning response time; at the same time, the present invention can provide comprehensive and timely disastrous weather information for meteorological departments, emergency management departments, and the general public, which is helpful to formulate effective disaster prevention and mitigation strategies; and can promote the innovation and development of meteorological information processing technology, and provide reference and reference for meteorological applications in other fields.

[0122] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0123] Those skilled in the art can understand that all or part of the steps in the various methods of the above embodiments can be completed by instructing related hardware through a program, and the program can be stored in a computer-readable storage medium, and the storage medium includes a read-only memory (ROM), a random access memory (RAM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), a one-time programmable read-only memory (OTPROM), an electronically erasable rewritable read-only memory (EEPROM), a compact disc (CD-ROM) or other optical disc storage, magnetic disk storage, magnetic tape storage, or any other computer-readable medium that can be used to carry or store data.

[0124] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

Claims

1. A complex disastrous weather characteristic parameter extraction system, characterized in that: include: A data acquisition module is used to obtain a meteorological data source and collect raw meteorological data from the meteorological data source in real time; A data preprocessing module, used for performing preprocessing operations on the original meteorological data to generate corresponding target meteorological data; A disaster type identification module is used to identify and judge the target meteorological data based on a convolutional neural network to obtain the corresponding disastrous weather type; A feature parameter extraction module is used to retrieve a preset feature parameter library, construct a feature parameter extraction model, and determine key feature parameters corresponding to the disastrous weather type; The warning report module is used to generate corresponding disastrous weather forecast and warning information according to the key characteristic parameters, and send and display it to the management end.

2. A complex disastrous weather characteristic parameter extraction system according to claim 1, characterized in that: The raw meteorological data include temperature, humidity, wind speed, wind direction, air pressure, precipitation, cloud cover, visibility and lightning activity data.

3. A complex disastrous weather characteristic parameter extraction system according to claim 1, characterized in that: The data preprocessing module specifically includes: Removing invalid data from the original meteorological data to generate first meteorological data; Smoothing the first meteorological data based on a smoothing filtering algorithm to remove noise in the first meteorological data and generate second meteorological data; The missing values ​​in the second meteorological data are filled based on a linear interpolation algorithm to generate the target meteorological data.

4. A complex disastrous weather characteristic parameter extraction system according to claim 1, characterized in that: The disaster type identification module is used to identify and judge the target meteorological data based on a convolutional neural network to obtain the corresponding disastrous weather type, specifically including: The target meteorological data is input into the convolution layer in the convolutional neural network, the meteorological feature pattern in the target meteorological data is extracted, and the corresponding initial meteorological feature is generated. The expression of the initial meteorological feature is: CSTZ=W0*TZMS+b0 Where CSTZ represents the initial meteorological characteristics; W0 represents the weight matrix corresponding to the convolution layer; TZMS represents the meteorological characteristic pattern in the target meteorological data; b0 represents the bias vector corresponding to the convolution layer; The initial meteorological features are input into the activation layer of the convolutional neural network to generate corresponding intermediate meteorological features, and the expression of the intermediate meteorological features is: ZJTZ=max(0,CSTZ) In the formula, ZJTZ represents the intermediate meteorological characteristics; max() represents the maximum value operation; CSTZ represents the initial meteorological characteristics; The intermediate meteorological features are input into the pooling layer in the convolutional neural network to reduce the dimension of the intermediate meteorological features, the dimension including height and width, and the intermediate meteorological features are subjected to maximum pooling processing to generate corresponding key meteorological features, the expression of the key meteorological features is: In the formula, GJTZ ij Represents the key meteorological characteristics at location index (i, j); ZJTZ mn Represents the intermediate meteorological characteristics at the location index (m,n); max() represents the maximum value operation; R ij Represents the area of ​​the pooling window corresponding to the position index (i, j) in the key meteorological feature on the intermediate meteorological feature.

5. A complex disastrous weather characteristic parameter extraction system according to claim 4, characterized in that: The key meteorological features are input into the fully connected layer in the convolutional neural network to generate the corresponding predicted disaster type, and the predicted disaster type is processed based on the activation function to obtain the corresponding predicted disaster occurrence probability, specifically including: The expression for predicting disaster type is: ZHLX=W1*GJTZ+b1 Where ZHLX represents the predicted disaster type; W1 represents the weight matrix corresponding to the fully connected layer; GJTZ represents the key meteorological features; b1 represents the bias vector corresponding to the fully connected layer; The expression for predicting the probability of disaster occurrence is: ZHGL=σ(ZHLX) In the formula, ZHGL represents the predicted probability of disaster occurrence; σ represents the activation function; ZHLX represents the predicted disaster type; According to the magnitude of the predicted disaster occurrence probability, the corresponding predicted disaster types are sorted from large to small, the preset number of disaster types is obtained, and the predicted disaster type whose serial number is less than or equal to the preset number of disaster types is selected as the disastrous weather type.

6. A complex disastrous weather characteristic parameter extraction system according to claim 1, characterized in that: The characteristic parameter extraction module is used to retrieve a preset characteristic parameter library, construct a characteristic parameter extraction model, and determine the key characteristic parameters corresponding to the disastrous weather type, specifically including: The preset characteristic parameter library is retrieved, the characteristic parameter extraction model is constructed, and the correlation between the disastrous weather type and each characteristic parameter in the preset characteristic parameter library is determined based on the Pearson correlation coefficient calculation formula. The expression of the correlation degree is: Wherein, D(ZHTQ,TZCS) represents the correlation between the type of disastrous weather and each characteristic parameter in the preset characteristic parameter library; ZHTQ represents the type of disastrous weather; TZCS represents each characteristic parameter in the preset characteristic parameter library; n represents the number of characteristic parameters in the preset characteristic parameter library; Σ represents the summation symbol; According to the magnitude of the correlation, the corresponding characteristic parameters are sorted from large to small, a preset number of characteristic parameters is obtained, and the characteristic parameters whose sequence numbers are less than or equal to the preset number of characteristic parameters are selected as the key characteristic parameters.

7. A complex disastrous weather characteristic parameter extraction system according to claim 1, characterized in that: The warning information includes the disaster level corresponding to the disastrous weather type, the possible affected area and disaster response measures.

Images