City region sand and dust weather forecasting method based on environment monitoring stations and sensitive cities

Abstract

The invention provides a method for forecasting sand and dust weather in a customs region based on an environment monitoring station and a sensitive city. According to the method, environmental data are collected by setting and maintaining environmental monitoring stations spread all over cities around a sand source and historical sand and dust transmission paths. A sensitive city is set on a sand and dust transmission path, and a corresponding environment monitoring station is set in the city to serve as a front forecasting city for sand and dust in the Guanzhong region. A mathematical model between the sand and dust weather and sensitive city environment monitoring data is established by combining historical experience data of the sand and dust weather. Through real-time monitoring and data analysis, when sand and dust weather occurs in a sensitive city, whether the sand and dust weather affects a central region or not is predicted according to the mathematical model, and the transmission time, the concentration and the continuous condition of the sand and dust weather are predicted. The method can improve timeliness and accuracy of sand and dust weather forecast, and has important value for environmental protection and public life.

Description

Technical Field

[0001] The invention belongs to the field of sandstorm forecasting, and more specifically relates to a sandstorm weather forecasting method in Guanzhong area based on environmental monitoring stations and sensitive cities. Background Art

[0002] Dust, as a natural disaster, has a serious impact on human production and life. Traditional dust weather forecasting is a very complex and difficult subject. Existing dust weather forecasting methods are mainly based on simulation and prediction of meteorological satellite data. This method often requires cross-departmental and cross-regional collaboration, involving many links such as data acquisition, data integration, model building and forecast result transmission. Due to the timeliness and accuracy of the data and the complexity of the model itself, the accuracy of this forecasting method is often difficult to achieve as expected. In addition, this method relies on a large number of external resources and tools, and cannot meet the independent, rapid and accurate forecasting needs of a certain region for dust weather.

[0003] First, existing forecasting methods mostly rely on meteorological satellite data. Although these data have the characteristics of wide coverage and long time range, they are limited by application conditions such as satellite orbits and collection timing, and their real-time and local forecasting capabilities for sandstorms are insufficient. For non-satellite-professional meteorological forecasting departments, obtaining and processing these satellite data requires cross-departmental and cross-regional collaboration, which is a complex process and has limited timeliness in forecasting.

[0004] Secondly, the existing sandstorm weather prediction models are mostly based on satellite remote sensing data and meteorological models, which are often ineffective for accurate prediction and real-time warning in local areas. At the same time, the processing of satellite data and the application of models require professional meteorological and remote sensing knowledge, which places high demands on the technical level of forecasting department personnel and has high learning and training costs. Summary of the invention

[0005] This paper proposes a sandstorm weather forecasting method for the Guanzhong region based on environmental monitoring stations and sensitive cities, which uses the existing rich ground observation network, combined with mathematical models and artificial intelligence algorithms, to conduct real-time and efficient sandstorm weather forecasting. Compared with the forecasting method that relies on satellite data, this method can better meet the timeliness requirements, reduce the complexity of forecasting work, and improve the accuracy and feasibility of forecasting.

[0006] In order to achieve the above object, the present invention is implemented by adopting the following technical solutions: The method includes: Set up and maintain environmental monitoring stations in cities around the dust source areas and along the historical dust transmission paths to collect environmental data; Set up sensitive cities along the dust transmission path and set up corresponding environmental monitoring stations in the cities. These sensitive cities will serve as the advance forecast cities for dust in the Guanzhong region. Study the historical experience data of sandstorm weather and establish a mathematical model between sandstorm weather and sensitive urban environmental monitoring data; Real-time monitoring and data analysis of collected environmental data; when sandstorms occur in sensitive cities, the mathematical model between sandstorms and environmental monitoring data of sensitive cities is used to predict whether the sandstorms will affect the Guanzhong region; this is used in scenarios where weather information needs to be obtained quickly in emergency situations; Provide continuous forecasts of sandstorm transmission time, sandstorm concentration and sandstorm weather.

[0007] In one embodiment, the environmental monitoring station collects PM2.5 and PM10 concentration data.

[0008] In one scenario, the sensitive city is a potential sensitive city, which is geographically located at a key hub for sand and dust transmission and can effectively reflect the movement dynamics of sand and dust from the source to the Guanzhong region.

[0009] In one scheme, the mathematical model between sandstorm weather and sensitive urban environmental monitoring data adopts a sandstorm weather prediction model based on a graph convolutional long short-term memory network, collects data related to sandstorm weather from multiple data sources, and pre-processes the data to ensure data quality and consistency.

[0010] In one embodiment, the real-time monitoring and data analysis of the collected environmental data includes: (1) Real-time data collection and integration; (2) Data preprocessing and feature engineering; (3) Application of real-time prediction models; (4) Determination of the impact of sandstorms and setting of thresholds; (5) Real-time alerts and decision support.

[0011] In one solution, the continuous prediction adopts a hybrid model of centralized heterogeneous networks and deep neural networks: this model takes into account diverse factors, including terrain factors and climate differences; by incorporating such factors into the training data, the effect of the training model is improved, which is suitable for scenarios of detailed analysis and long-term planning.

[0012] In one embodiment, the hybrid model of the centralized heterogeneous network and the deep neural network includes: (1) Data integration and feature fusion; (2) Building a centralized heterogeneous network; (3) Deep neural network model design; (4) Model training and optimization.

[0013] Beneficial effects of the present invention: 1. The method of the present invention can obtain and analyze environmental parameters related to sandstorm weather in real time by using environmental monitoring stations distributed around sand sources and on sandstorm transmission paths, as well as environmental monitoring data of sensitive cities, thereby improving the timeliness of sandstorm weather forecasts.

[0014] 2. The present invention adopts artificial intelligence algorithms and mathematical models, integrates environmental monitoring data and environmental conditions of sensitive cities, and can more accurately predict and judge parameters such as the transmission path, impact range, and duration of sandstorm weather, thereby improving the accuracy of the forecast.

[0015] 3. Compared with the forecasting method relying on satellite data, the present invention saves the complicated links such as multi-department coordination and data processing, simplifies the forecasting process, improves work efficiency and reduces the forecasting cost.

[0016] 4. The present invention directly uses the equipment that can be managed by the unit to make forecasts, making sandstorm weather forecasts more convenient and efficient, and more convenient to implement.

[0017] In general, the sandstorm weather forecast method for the Guanzhong area based on environmental monitoring stations and sensitive cities proposed in the present invention has the characteristics of strong real-time performance, high accuracy, and easy operation. It can provide timely and accurate sandstorm weather forecasts for government decision-making departments and the public, which is helpful to do a good job in the prevention and control of sandstorm weather, protect the environment, and reduce the impact of sandstorm weather on people's lives. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 is a flow chart of the method of the present invention; Figure 2 To establish a mathematical model flow chart between sandstorm weather and sensitive urban environmental monitoring data; Figure 3 A flow chart for real-time monitoring and data analysis of collected environmental data; Figure 4 Flowchart for prediction of a hybrid model combining heterogeneous networks and deep neural networks. DETAILED DESCRIPTION

[0019] In order to facilitate the understanding of the present invention, the present invention will be described more fully below with reference to the relevant drawings. Typical embodiments of the present invention are given in the drawings. However, the present invention can be implemented in many different forms and is not limited to the embodiments described in the present invention. On the contrary, the purpose of providing these embodiments is to make the disclosure of the present invention more thorough and comprehensive.

[0020] Unless otherwise defined, all technical and scientific terms used in the present invention have the same meaning as those understood by a person skilled in the art of the present invention. The terms used in the present invention in the specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention. In order to facilitate the understanding of the present invention, the present invention will be described more fully below with reference to the relevant drawings. Typical embodiments of the present invention are shown in the drawings. However, the present invention can be implemented in many different forms and is not limited to the embodiments described in the present invention. On the contrary, the purpose of providing these embodiments is to make the disclosure of the present invention more thorough and comprehensive.

[0021] The general concept of the present invention is: a method for forecasting sandstorm weather in Guanzhong region based on environmental monitoring stations and sensitive cities. This method uses environmental monitoring stations in cities around the sand source and environmental monitoring stations on the historical transmission path of sandstorms to read the concentrations of PM2.5 and PM10 monitored by environmental monitoring stations to judge sandstorm weather. Then, a sensitive city is set on the sandstorm transmission path, and the sensitive city serves as a pre-forecast city for sandstorms in Guanzhong region. When the environmental monitoring station in the sensitive city monitors sandstorm pollution, historical experience data is used, and then analyzed and processed to judge whether sandstorm weather will enter Guanzhong region and how much impact it will have on Guanzhong region. The selection of sensitive cities is based on historical data, that is, the sandstorms in Guanzhong region are all transmitted to Guanzhong region through sandstorms in the sand source through sensitive cities, that is, if there is sandstorm weather in Guanzhong region, there will definitely be sandstorm weather in sensitive cities.

[0022] Through the above method, combined with artificial intelligence algorithm analysis, sandstorm weather in Guanzhong area can be forecasted. The forecast content includes: sandstorm transmission time, sandstorm concentration, sandstorm duration and other information.

[0023] The difference between the method of the present invention and the prior art is that the sandstorm forecast of the prior art mostly uses satellite data to forecast sandstorm weather. The present invention uses existing environmental monitoring stations throughout the region and selects sensitive cities to make forecasts. Because the use of satellite data for forecasting requires coordination in many aspects, the technical solution of the present application can directly use the equipment that the unit can manage for forecasting, which improves the timeliness of the forecast and reduces the intermediate work process.

[0024] like Figure 1 As shown, step one: set up and maintain environmental monitoring stations throughout the cities surrounding the sand source area and along the historical transmission paths of sand and dust to collect relevant environmental data such as the concentration of PM2.5 and PM10.

[0025] Conduct a comprehensive geographical and climatic analysis of the sand source area and its surrounding areas to determine the main sources of sand and dust and their potential transmission paths. Based on historical sand and dust weather data and transmission patterns, select key cities as the basis for the layout of monitoring stations to ensure that these stations can effectively cover the main paths where sand and dust may be transmitted. Transmit monitoring data to the central data processing center in real time via wireless or wired networks. Through the implementation and maintenance of the above steps, a comprehensive and stable environmental monitoring network can be established to provide reliable data support for subsequent sand and dust weather forecasts.

[0026] Step 2: Set up sensitive cities on the dust transmission path and set up corresponding environmental monitoring stations in the cities. These sensitive cities will serve as advance forecast cities for dust in the Guanzhong region.

[0027] Implementation process of step 2 First, we conducted an in-depth analysis of the historical data of dust transmission in the Guanzhong region to identify the main transmission routes and key node cities of dust. These key node cities are potential sensitive cities. Because of their geographical location at the key hub of dust transmission, they can effectively reflect the movement dynamics of dust from the source to the Guanzhong region. Next, based on the distribution and transmission patterns of historical dust events, we selected several sensitive cities with representativeness and warning value to ensure that these cities cover the main dust transmission routes and have high data collection capabilities.

[0028] After the sensitive cities are identified, environmental monitoring stations in each sensitive city should be used. These monitoring stations should be equipped with advanced monitoring equipment, which can accurately and real-time collect concentration data of key particulate matter such as PM2.5 and PM10, and have stable data transmission capabilities to transmit monitoring data to the central data processing system in real time.

[0029] In order to realize the function of sensitive cities as advance forecasting points, it is also necessary to establish a close cooperation mechanism with local governments and relevant departments to ensure that monitoring data can be shared and applied to forecast models in a timely manner. By setting up efficient environmental monitoring stations in sensitive cities, the occurrence and transmission dynamics of sandstorms can be captured in advance, providing important advance information for sandstorm forecasts in the Guanzhong region, thereby improving the accuracy and response speed of the overall forecast system.

[0030] Finally, combined with geographic information system (GIS) technology, the monitoring data of sensitive cities are spatially analyzed and visualized to help decision makers more intuitively understand the path and impact range of dust transmission. Through the systematic implementation of the above steps, an efficient and reliable sensitive city monitoring network can be established on the dust transmission path, providing solid data support and technical guarantee for early warning and response to sandstorm weather in Guanzhong area.

[0031] Step 3: Study the historical experience data of sandstorms and establish a mathematical model between sandstorms and sensitive urban environmental monitoring data. Construct a sandstorm forecasting model based on graph convolutional long short-term memory network, collect sandstorm-related data from multiple data sources, and preprocess these data to ensure data quality and consistency. These data sources include meteorological observation data (such as wind speed, humidity, temperature), historical sandstorm records, etc.

[0032] like Figure 2 As shown in the figure, after completing steps one and two, that is, identifying sensitive cities and setting up environmental monitoring stations there, the goal of step three is to use these monitoring data and historical sandstorm weather events to establish an innovative mathematical model to achieve accurate prediction of sandstorm weather. To this end, this step will use the Graph Convolutional Long Short-Term Memory (GCLSTM), a deep learning model that combines a graph convolutional network (GCN) and a long short-term memory network (LSTM), which is specifically designed to process complex data with spatial and temporal dependencies.

[0033] S301. Data collection and preprocessing First, historical data from sensitive urban environmental monitoring stations were collected, including multiple meteorological indicators such as PM2.5, PM10, wind speed, wind direction, humidity, and air pressure. At the same time, records of sandstorm weather events in the corresponding time period were collected, such as the time of occurrence, intensity, and duration. To ensure the integrity of the data, missing values ​​were filled by interpolation, and the data were normalized using standardization methods (such as Z-score standardization) to eliminate the dimensional differences between different indicators.

[0034] S302, Constructing a spatial relationship diagram Since the transmission of sandstorm weather has significant spatial dependence, it is necessary to construct a graph structure that reflects the spatial relationship between sensitive cities. Suppose there are N sensitive cities, each of which is a node in the graph. The edge weights between nodes are It can be defined based on factors such as the geographical distance between cities, the direction of airflow, and the frequency of historical dust transport, for example:

[0035] in, represents the distance between city i and city j, It represents the angle between the airflow direction and the line connecting the cities. To adjust the parameters.

[0036] S303. Constructing the Graph Convolutional Network (GCN) layer Graph convolutional networks are used to capture the spatial dependencies between cities. Suppose the input feature matrix is , where F is the number of features for each city (such as PM2.5, PM10, etc.). The graph convolution operation can be expressed as:

[0037] in, To add the adjacency matrix with self-connection, for The degree matrix of is the trainable weight matrix, is the activation function (such as ReLU).

[0038] S304. Build a long short-term memory network (LSTM) layer In order to capture the dynamic features in the time series, LSTM is used to process the spatial features extracted by GCN. Suppose the graph convolution output is the node feature at time step t, and the update equation of LSTM is:

[0039] in, , , They are forget gate, input gate and output gate respectively. and are cell state and hidden state respectively, Represents element-wise multiplication.

[0040] S305, Model training and optimization Combine GCN with LSTM to form a GCLSTM model, and train the model parameters through historical data. The loss function can use mean square error (MSE) for regression tasks, or cross entropy loss for classification tasks (such as whether a dust event occurs). Use the back propagation algorithm and optimizer (such as Adam) to update parameters, and apply regularization techniques (such as Dropout) to prevent overfitting.

[0041] Through the above steps, the GCLSTM model can effectively integrate spatial and temporal information, capture the complex dynamics of sandstorm weather, and provide an innovative and high-precision mathematical tool for sandstorm weather forecasting in Guanzhong area.

[0042] Step 4: Real-time monitoring and data analysis of the collected environmental data. When sandstorms occur in sensitive cities, the mathematical model established in step 3 is used to predict whether the sandstorms will affect the Guanzhong region. Step 4 relies on the multi-source data and preprocessing results provided by step 3 to conduct real-time monitoring and preliminary predictions. These preliminary prediction results are passed to step 5 as input for further in-depth analysis and advanced predictions to form more sophisticated and comprehensive prediction results. It is suitable for scenarios where weather information needs to be obtained quickly in emergency situations.

[0043] like Figure 3 As shown in the figure, after completing the first three steps, namely, determining sensitive cities and establishing a sandstorm weather prediction model based on a graph convolutional long short-term memory network (GCLSTM), the goal of step four is to achieve real-time monitoring and analysis of the collected environmental data, and when sandstorm weather occurs in sensitive cities, use the mathematical model established in step three to predict whether the sandstorm weather will affect the Guanzhong region. The specific implementation process is as follows: S401, Real-time data collection and integration First, an efficient real-time data collection system needs to be established to continuously obtain key environmental indicator data from environmental monitoring stations in sensitive cities. These data include but are not limited to PM2.5, PM10, wind speed, wind direction, humidity, air pressure, etc. The data collection system should have high reliability and low latency to ensure that data can be transmitted to the central processing platform in real time.

[0044] Real-time data can be represented as a time series vector:

[0045] in, It represents the environmental index data of the i-th monitoring station at time t, and N is the number of monitoring stations.

[0046] S402. Data preprocessing and feature engineering After data collection, real-time data needs to be preprocessed, including missing value filling, outlier detection and processing, and data standardization. The commonly used standardization method is Z-score standardization:

[0047] in, and are the mean and standard deviation of the historical data of the ith monitoring station respectively.

[0048] In addition, feature engineering is required to extract key features in the time series, such as moving average, trend changes, etc., to enhance the model's predictive ability. For example, calculate the moving average of the past k time steps:

[0049] S403. Application of real-time prediction model The preprocessed real-time data will be used as input and dynamically passed to the GCLSTM model trained in step 3 to generate dust weather forecast results. The model prediction process can be expressed as:

[0050] in, Represents the time step The input data sequence to t, is the adjacency matrix between sensitive cities, is a predicted value (such as PM2.5 concentration or the probability of occurrence of sandstorm weather events).

[0051] S404. Determination of sandstorm weather impact and threshold setting In order to convert the model output into a specific impact judgment, a threshold needs to be set. For example, when the predicted PM2.5 concentration exceeds a certain set threshold When the sandstorm weather is judged to affect the Guanzhong area:

[0052] Setting Thresholds It can be based on historical data analysis and environmental standards, such as the World Health Organization (WHO) recommended standards for air quality.

[0053] S405, Real-time alarm and decision support When the judgment result is "yes", the system should automatically trigger the early warning mechanism and send sandstorm weather warning information to relevant departments and the public through multiple channels such as SMS, email, mobile applications, etc. In addition, it can be combined with the Geographic Information System (GIS) to achieve a visual display of the impact range of sandstorm weather and assist decision makers in formulating targeted protective measures, such as traffic control and restrictions on outdoor activities.

[0054] Through the above systematic steps, step four realizes real-time monitoring and accurate prediction of sandstorm weather, which can timely warn the Guanzhong area of ​​the possible impact of sandstorm weather, provide scientific basis and technical support, and help relevant departments to take effective protection and response measures.

[0055] Step 5: Input the prediction results of step 4 into the artificial intelligence algorithm to continuously predict the transmission time of sandstorms, sandstorm concentration and sandstorms. The artificial intelligence algorithm adopts: a hybrid model of centralized heterogeneous networks and deep neural networks (Heterogeneous Network Fusion): This type of model can take into account diverse factors, including terrain factors, climate differences, and other environmental factors. By taking such factors into account in the training data, the effect of the training model can be improved. It is suitable for scenarios that require detailed analysis and long-term planning.

[0056] Step 5 further analyzes and optimizes the preliminary prediction results of step 4 by introducing a hybrid model of centralized heterogeneous networks and deep neural networks (Heterogeneous Network Fusion, HNF), comprehensively considers multiple influencing factors, and accurately predicts the transmission time, concentration and duration of sandstorm weather.

[0057] like Figure 4 As shown in the figure, after completing step 4, i.e. real-time monitoring and preliminary prediction of the impact of sandstorm weather on Guanzhong region, step 5 aims to further improve the accuracy and comprehensiveness of the prediction. This step further predicts the transmission time, dust concentration and duration of sandstorm weather by inputting the prediction results of step 4 into the hybrid model of centralized heterogeneous network and deep neural network (Heterogeneous Network Fusion, HNF), taking into account terrain factors, climate differences and other environmental factors. The following is a detailed implementation process and the mathematical formulas involved: S501. Data integration and feature fusion First, collect multi-source heterogeneous data, including but not limited to the real-time prediction results in step 4, terrain data, climate data, vegetation coverage, land use type, historical dust weather records, etc. These data sources have different structures and characteristics, so they need to be effectively integrated and fused.

[0058] Assume that there are M different types of data sources, and the characteristics of each data source are expressed as , where N is the number of samples, is the feature dimension of the mth data source. The feature fusion process can be expressed as:

[0059] in, Indicates that various features are spliced ​​in the feature dimension to obtain the fused feature matrix ,in .

[0060] S502. Building a centralized heterogeneous network The centralized heterogeneous network aims to capture the complex relationships between different types of data. There are K different types of nodes, each with specific semantics and attributes. By constructing a graph structure with multiple relationships, the associations between multi-source data can be fully represented.

[0061] For example, in sandstorm weather forecasting, node types can include monitoring stations, meteorological elements, terrain features, etc. It can be defined based on spatial distance, semantic association or other related indicators:

[0062] in, It is a function that combines distance and semantic relevance, which can be designed according to actual conditions, such as:

[0063] in, represents the distance between node i and node j, Indicates the semantic relevance, , , is the weight parameter.

[0064] S503. Deep neural network model design Based on the heterogeneous network, a deep neural network architecture is designed to fully mine the complex patterns in multi-source data. The hybrid model includes the following key components: Graph Convolutional Network (GCN) layer: used to capture the spatial relationship and structural information between nodes in the network.

[0065] The graph convolution operation can be expressed as:

[0066] in, is the normalized adjacency matrix, is the node feature of the lth layer, is the trainable weight matrix, is the activation function (e.g. ReLU).

[0067] Multi-Layer Perceptron (MLP) layer: used to process fused features and extract high-level abstract information.

[0068] The calculation process of MLP is:

[0069] in, and are the weight matrix and bias vector respectively.

[0070] Fusion Layer: Fusion the spatial information extracted by the GCN layer with the feature information extracted by the MLP layer to form a unified representation.

[0071]

[0072] Output layer: Design the corresponding output layer according to different prediction tasks. For example, for regression tasks (such as predicting dust concentration), the output layer can use a linear activation function; for classification tasks (such as whether a dust event occurs or not), the Softmax activation function is used.

[0073]

[0074] S504, Model training and optimization Using historical data and the prediction results from step 4, train the HNF hybrid model. The choice of loss function depends on the specific prediction task: Regression tasks (such as prediction of dust concentration, transmission time, and duration) use mean square error (MSE):

[0075] Classification tasks (such as whether a dust event occurs or not) use cross entropy loss:

[0076] Back propagation algorithm and optimizer (such as Adam optimizer) are used to update parameters. At the same time, regularization techniques (such as L2 regularization and Dropout) are applied to prevent model overfitting and improve the generalization ability of the model.

[0077] Combining the above steps, the mathematical formula of the HNF hybrid model is as follows:

[0078] S505, Prediction and Application After the model training is completed and its performance is verified, it is applied to the actual dust weather prediction task. The specific process is as follows: 1. Input data preparation: Collect the latest real-time data and other relevant heterogeneous data for preprocessing and feature fusion.

[0079] 2. Feature extraction and fusion: GCN is used to extract spatial relationship features, and MLP is used to process the fused multi-source features.

[0080] 3. Information fusion and prediction: The outputs of GCN and MLP are fused to generate prediction results of dust weather transmission time, concentration and duration through the output layer.

[0081] 4. Application of results: The prediction results will be used in the decision support system to assist relevant departments in formulating protective measures, such as issuing warnings, adjusting traffic flow, arranging outdoor activities, etc.

[0082] The mathematical representation of the prediction process is:

[0083] in, represents a hybrid model of centralized heterogeneous networks and deep neural networks, is the fused multi-source feature, is the adjacency matrix of the heterogeneous network, is the predicted output (including transit time, concentration and duration).

[0084] Through the above steps, step five realizes the accurate prediction of the transmission time, concentration and duration of sandstorm weather, greatly improving the intelligence level and predictive ability of the entire sandstorm weather warning system, and providing strong technical support for sandstorm weather protection in the Guanzhong area.

[0085] Step 4 focuses on real-time and rapid response, and is suitable for scenarios where weather information needs to be obtained quickly in emergency situations. The real-time data and preliminary forecast results generated in step 4 can be used as inputs to the advanced model in step 5 to improve the quality and accuracy of advanced forecasts. Emergency departments and the public mainly rely on the immediate warnings and short-term forecasts in step 4 to take emergency protective measures.

[0086] Step 5 focuses on the depth and accuracy of predictions and is suitable for scenarios that require detailed analysis and long-term planning. The detailed prediction information provided by step 5 can also be fed back to step 4 to optimize the model and strategy of real-time prediction. Scientific research institutions, policy makers and long-term planners rely more on the in-depth analysis and long-term prediction results of step 5 to guide strategic decision-making.

[0087] Combining steps 4 and 5, we can build a comprehensive sandstorm weather forecasting system that has both rapid response capabilities and high-precision forecasting capabilities. The two work together to ensure that appropriate forecasting services can be provided in different scenarios, improving the adaptability and reliability of the overall system.

[0088] The innovation of this implementation step lies mainly in: introducing data from sensitive cities and their environmental monitoring stations as the basis for prediction; combining artificial intelligence algorithms with historical data to make the prediction results more accurate; and improving the timeliness of sandstorm weather warnings through real-time monitoring data analysis.

[0089] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through a computer program, and the program can be stored in a computer-readable storage medium, and when the program is executed, it can include the processes of the embodiments of the above-mentioned methods. The storage medium can be a disk, an optical disk, a read-only memory (ROM) or a random access memory (RAM).

[0090] It should be understood that the detailed description of the technical solutions of the present invention by means of the preferred embodiments is illustrative rather than restrictive. A person skilled in the art may modify the technical solutions described in the embodiments, or replace some of the technical features by equivalents, based on reading the specification of the present invention; and 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 sandstorm weather forecasting method for Guanzhong region based on environmental monitoring stations and sensitive cities, characterized by: The method includes: Set up and maintain environmental monitoring stations in cities around the dust source areas and along the historical dust transmission paths to collect environmental data; Set up sensitive cities along the dust transmission path and set up corresponding environmental monitoring stations in the cities. These sensitive cities will serve as the advance forecast cities for dust in the Guanzhong region. Study the historical experience data of sandstorm weather and establish a mathematical model between sandstorm weather and sensitive urban environmental monitoring data; Real-time monitoring and data analysis of collected environmental data; when sandstorms occur in sensitive cities, the mathematical model between sandstorms and environmental monitoring data of sensitive cities is used to predict whether the sandstorms will affect the Guanzhong region; this is used in scenarios where weather information needs to be obtained quickly in emergency situations; Provide continuous forecasts of sandstorm transmission time, sandstorm concentration and sandstorm weather.

2. A sandstorm weather forecasting method for Guanzhong region based on environmental monitoring stations and sensitive cities according to claim 1, characterized in that: The environmental monitoring site collects PM2.5 and PM10 concentration data.

3. The method for forecasting sandstorm weather in Guanzhong region based on environmental monitoring stations and sensitive cities according to claim 1, characterized in that: The sensitive cities mentioned above are potential sensitive cities, which are located at key hubs for sand and dust transmission and can effectively reflect the movement dynamics of sand and dust from the source to the Guanzhong region.

4. The method for forecasting sandstorm weather in Guanzhong region based on environmental monitoring stations and sensitive cities according to claim 1, characterized in that: The mathematical model between the sandstorm weather and the sensitive urban environment monitoring data adopts a sandstorm weather prediction model based on a graph convolutional long short-term memory network, collects data related to sandstorm weather from multiple data sources, and pre-processes the data to ensure data quality and consistency.

5. The method for forecasting sandstorm weather in Guanzhong region based on environmental monitoring stations and sensitive cities according to claim 1, characterized in that: The real-time monitoring and data analysis of the collected environmental data includes: (1) Real-time data collection and integration; (2) Data preprocessing and feature engineering; (3) Application of real-time prediction models; (4) Determination of the impact of sandstorms and setting of thresholds; (5) Real-time alerts and decision support.

6. The method for forecasting sandstorm weather in Guanzhong region based on environmental monitoring stations and sensitive cities according to claim 1, characterized in that: The continuous prediction described above adopts a hybrid model of centralized heterogeneous networks and deep neural networks: this model takes into account diverse factors, including terrain factors and climate differences; by incorporating such factors into the training data, the effect of the training model is improved, which is suitable for scenarios of detailed analysis and long-term planning.

7. The method for forecasting sandstorm weather in Guanzhong region based on environmental monitoring stations and sensitive cities according to claim 6, characterized in that: The hybrid model of centralized heterogeneous network and deep neural network includes: (1) Data integration and feature fusion; (2) Building a centralized heterogeneous network; (3) Deep neural network model design; (4) Model training and optimization.

Images