Distributed artificial weather influence operation cooperative instruction issuing system based on internet of things

Abstract

The application belongs to the technical field of software systems, and discloses a distributed artificial weather modification operation cooperative instruction issuing system based on an Internet of Things, which adopts a cloud-edge-end three-layer architecture. A cloud platform layer fuses multi-source meteorological data through a perception model, analyzes rain cloud changes through a rain cloud tracking and prediction model, and generates operation instructions based on a large language model; an edge node layer processes and distributes cached instructions, and supports local decision-making when there is no cloud instruction; and a terminal device layer completes data collection and operation execution. The application realizes spatiotemporal fusion of multi-source data, dynamic tracking of rain cloud trajectories, and intelligent generation of operation schemes, can determine the best catalytic window, accurately plans the type, position and operation parameters of equipment, and supports dynamic adjustment of redundant equipment. The effect significantly improves operation accuracy and efficiency, reduces operation timing judgment errors, shortens equipment response delay to seconds, improves rain enhancement efficiency, and provides reliable support for disaster prevention, activity support and the like.

Description

Technical Field

[0001] This invention belongs to the field of meteorological system technology, specifically relating to a distributed artificial weather modification operation collaborative command issuance system based on the Internet of Things. Background Technology

[0002] Meteorological artificial intervention technology aims to regulate local weather processes through scientific means to cope with natural disasters such as droughts, floods, and hail, or to ensure the smooth conduct of major events. Its core objective is to promote or inhibit precipitation formation by altering the microphysical structure of clouds. For example, artificial rain enhancement involves seeding cold clouds with ice-forming agents such as silver iodide and dry ice, generating a large number of ice crystals and accelerating the conversion of cloud water into precipitation; while artificial rain suppression involves excessive seeding of ice nuclei on the upwind side of the target area, preventing the formation of sufficiently large raindrops or causing the rain to stop prematurely. Common intervention methods include aerial seeding, rocket launches, and artillery bombardment, with catalysts mainly including silver iodide, dry ice, and salt powder. This technology, developed since the 1940s, has been widely used globally. Its principle is based on cloud physics: cold cloud catalysis utilizes the ice crystal effect to promote precipitation, while warm cloud catalysis uses hygroscopic agents to increase cloud droplets and achieve rainfall. Furthermore, artificial hail suppression inhibits the formation of large hailstones by increasing the number of hail embryos to compete for moisture. These technologies have played an important role in agricultural drought resistance, ecological restoration, and disaster prevention, but still face technological bottlenecks.

[0003] Current weather modification operations rely heavily on human decision-making and face the following core problems:

[0004] Data processing and decision-making are inefficient. Traditional methods rely on manual analysis of weather radar and satellite data, combined with experience to determine the timing and location of operations, lacking the ability to integrate multi-source data in real time. For example, operators need to manually identify cloud features and calculate catalyst quantities, which is time-consuming and easily affected by subjective factors.

[0005] Insufficient equipment coordination means that the scheduling and parameter settings of equipment such as rockets, aircraft, and ground-based smoke generators require manual coordination and cannot be dynamically adjusted based on real-time cloud conditions. For example, when the speed or intensity of rain clouds changes, manual instructions may be delayed, leading to poor operational results.

[0006] Evaluating the effectiveness of artificial rainmaking operations is challenging. Traditional statistical methods struggle to accurately quantify the impact of human intervention and lack real-time feedback mechanisms. For instance, the actual effects of artificial rainmaking may be masked by fluctuations in natural precipitation, necessitating complex comparative experiments or numerical simulations for verification.

[0007] In addition, insufficient technological maturity, aging equipment, and uneven professional qualifications of personnel also restrict operational efficiency. For example, some areas still use outdated anti-aircraft gun equipment, which has low catalytic efficiency and poses safety risks; insufficient training of operators may lead to operational errors or unreasonable parameter settings. Summary of the Invention

[0008] To address the problems existing in the prior art, this invention provides a distributed collaborative command issuance system for weather modification operations based on the Internet of Things (IoT). By automatically acquiring and fusing multi-source meteorological data, the system makes judgments and predictions about rain clouds, thereby automatically generating tasks according to needs and issuing commands through the IoT.

[0009] The technical solution adopted in this invention is as follows:

[0010] In a first aspect, the present invention provides a distributed weather modification operation collaborative instruction issuance system based on the Internet of Things, which uniformly issues task instructions for weather modification operations within a defined target area, including:

[0011] The cloud platform layer, serving as the data control center, includes a sensing model that acquires and fuses multi-source meteorological data for the target area through data channels, as well as a rain cloud tracking and prediction model and a task model based on a large language model.

[0012] The terminal equipment layer, serving as the underlying data collection and execution end, includes meteorological monitoring equipment, artificial weather modification equipment, and auxiliary equipment; and

[0013] The edge node layer, as the middle layer of the distributed architecture, is deployed in regional operation centers or edge computing servers. It includes data processing stations and device management gateways. The data processing stations process and cache instructions from the cloud platform layer and data from the terminal device layer within their management scope, while the device management gateway distributes instructions and performs protocol conversion of data.

[0014] In conjunction with the first aspect, the present invention provides a first embodiment of the first aspect, wherein the edge node layer includes the smallest meteorological management unit within a defined target area as a node, a node server is deployed, and data interaction is performed with the server of the cloud platform layer through a wired communication network and with the terminal device through a wireless IoT communication network.

[0015] In conjunction with the first aspect, the present invention provides a second implementation of the first aspect, wherein the edge node layer further includes a local decision module, which processes the acquired terminal device layer data by means of a local decision module equipped with a locally deployed and trained large language model to obtain local meteorological status data and upload it to the cloud platform layer.

[0016] The initial instructions issued by the cloud platform layer are processed. If no initial instructions are received from the cloud platform layer within the set rain cloud window time, the edge node layer processes the latest initial instructions based on the acquired meteorological status data to form task instructions and sends them to the corresponding terminal device layer.

[0017] In conjunction with the first aspect, the present invention provides a third embodiment of the first aspect, wherein the meteorological monitoring equipment includes a weather radar, an anemometer, a rain gauge, a thermometer, a hygrometer, a barometer, and a weather balloon;

[0018] The artificial intervention equipment includes ground-based launching equipment and aerial seeding equipment.

[0019] In conjunction with the first aspect, the present invention provides a fourth implementation of the first aspect, wherein the perception model acquires multi-source meteorological data for the target area through an online data channel, and performs data fusion with weighted allocation after processing the multi-source data to form operational reference data for the target area.

[0020] In conjunction with the fourth implementation of the first aspect, the present invention provides a fifth implementation of the first aspect, wherein the online data acquired by the perception model includes local monitoring data acquired by several meteorological detection devices through the terminal device layer, and meteorological satellite images acquired through the access network. The perception model identifies and processes the acquired meteorological satellite images through an image processing algorithm and acquires meteorological satellite data.

[0021] After cleaning and spatiotemporally unifying local monitoring data and meteorological satellite data, the perception model assigns weights to the spatiotemporally labeled data using a dynamic weighting method. It then uses a multimodal Transformer model to perform spatiotemporal correlation and generate a grid fusion feature tensor for the target area. Finally, the perception model outputs a grid dataset and a time-series dataset. The rain and cloud tracking prediction model processes the grid dataset and the time-series dataset and outputs visualization data of rain and cloud changes within the target area up to 72 hours, as well as the predicted trajectory changes of rain and clouds.

[0022] In conjunction with the fifth embodiment of the first aspect, the present invention provides a sixth embodiment of the first aspect, wherein the task model processes and obtains the optimal catalytic window T of the rain cloud based on the rain cloud change data obtained for the target area and by the rain cloud tracking and prediction model;

[0023] Then, based on the grid dataset, time series dataset, and visualization data of rain cloud predicted trajectory changes obtained from the perception model, the required locations of the artificial intervention equipment and the corresponding preliminary instructions for operation are determined. The cloud platform layer sends the instructions to the edge node layer, which processes the instructions to form the task instructions corresponding to the terminal devices within the communication range.

[0024] In conjunction with the sixth implementation of the first aspect, the present invention provides a seventh implementation of the first aspect, wherein the task model is trained based on an existing large language model, and historical meteorological data is collected during training. The historical meteorological data includes grid datasets and time-series datasets output by the perception model, as well as visualization data of rain cloud predicted trajectory changes output by the rain cloud tracking and prediction model. The collected data is cleaned, normalized, and feature extracted, and then the actual operation and effect evaluation data of the corresponding artificial weather modification operation are recorded.

[0025] The preprocessed data is divided into training set, validation set and test set according to the proportion. The training set is used for learning the model's parameters, the validation set is used to tune the model's hyperparameters, and the test set is used to evaluate the model's final performance.

[0026] A deep learning model is used to process time series data and spatial features. Historical meteorological data is taken as input, and the corresponding optimal catalytic window, type of operating equipment, location and operation instructions are taken as output. The model parameters are continuously adjusted through the backpropagation algorithm to output task instructions.

[0027] In conjunction with the sixth implementation of the first aspect, the present invention provides an eighth implementation of the first aspect, wherein when the task model generates the preliminary instruction, it first obtains the device information of the target area devices by the cloud platform through the device management gateway of the edge node layer, and the task model generates a device capability matrix based on the device information; the preliminary instruction generated by the task model includes alternative adjustment parameters, and the alternative parameters are associated with the performance data of alternative devices in the device capability matrix;

[0028] When a terminal device malfunctions during task execution, the local decision-making module of the edge node initiates adjustment logic, calls pre-stored alternative adjustment parameters, selects alternative devices from the device capability matrix, calculates the substitution coefficient to correct the parameters, keeps the total catalyst quantity change value less than the set threshold, and sends collaborative instructions to surrounding devices through the device management gateway to synchronously correct the device startup sequence.

[0029] The beneficial effects of this invention are as follows:

[0030] (1) The technology of this invention realizes full-process automation of meteorological data from collection and analysis to operation instruction generation by constructing a three-layer distributed architecture of cloud, edge and terminal and an intelligent task model, which significantly improves the accuracy and efficiency of artificial weather modification operations. The system can integrate multi-source data such as satellite, radar and ground sensors in real time, accurately identify rain cloud type, movement trajectory and optimal catalyst window, avoid the subjectivity and lag of manual analysis, and reduce the error in judging the timing of operations;

[0031] (2) The task model of this invention, based on deep learning and meteorological principles, can autonomously generate detailed plans including equipment type, operation location, time window, and catalyst parameters, replacing the traditional operation planning mode that relies on human experience. Through dynamic optimization algorithms, the system can adjust the equipment coordination strategy according to real-time cloud conditions. For example, when rain clouds arrive early or their intensity changes abruptly, redundant equipment is automatically triggered and the seeding parameters are corrected, ensuring that the matching degree between the operation instructions and the actual meteorological conditions is improved, and significantly reducing the problem of operation failure caused by delays in human decision-making;

[0032] (3) The present invention adopts a distributed architecture to support seamless collaboration of multiple types of equipment such as rockets, drones, and ground-based smoke generators. By combining local processing at edge nodes with global scheduling in the cloud, the system can achieve real-time monitoring of equipment status and rapid issuance of instructions. In remote areas or complex terrain scenarios, the system uses multi-mode transmission technology to ensure equipment connectivity, solving the problem of blind spots in traditional manual operations in environments without ground networks. This reduces equipment response latency from tens of minutes to seconds, expanding the operational coverage.

[0033] (4) This invention introduces a closed-loop mechanism of reinforcement learning and effect evaluation. By training the model with historical operation data, it continuously optimizes core parameters such as catalyst dosage calculation and seeding height selection, thereby improving the efficiency of artificial rain enhancement and the success rate of rain suppression. At the same time, the system’s built-in safety rules and equipment redundancy strategy can automatically avoid sensitive areas such as residential areas and airports, reduce the risk of human operation errors, improve the safety factor of operation, and provide reliable technical support for scenarios such as major event security, agricultural drought relief, and ecological restoration. Attached Figure Description

[0034] Figure 1 This is an architecture diagram of the instruction issuance system in an embodiment of the present invention;

[0035] Figure 2 This is a schematic diagram of the task model of the present invention. Detailed Implementation

[0036] The present invention will be further explained below with reference to the accompanying drawings and specific embodiments.

[0037] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. The components of the embodiments of this application described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.

[0038] Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely to illustrate selected embodiments of the application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

[0039] It should be noted that similar labels and letters in the following figures indicate similar items. Therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures.

[0040] In the description of this application, it should be noted that the use of terms such as "center," "upper," "lower," "left," "right," "vertical," "horizontal," "inner," and "outer" to indicate orientation or positional relationships is based on the orientation or positional relationships shown in the accompanying drawings, or the orientation or positional relationships commonly used when the product is in use. These terms are used solely for the convenience of describing this application and for simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this application. Furthermore, the use of terms such as "first" and "second" in the description of this application is only used to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0041] Furthermore, the use of terms such as "horizontal" and "vertical" in the description of this application does not imply that the component is required to be absolutely horizontal or suspended, but rather that it may be slightly tilted. For example, "horizontal" simply means that its direction is more horizontal relative to "vertical," and does not mean that the structure must be completely horizontal, but rather that it may be slightly tilted.

[0042] In the description of this application, it should also be noted that, unless otherwise expressly specified and limited, the terms "set up," "install," "connect," and "link" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection between two components. Those skilled in the art can understand the specific meaning of the above terms in this application based on the specific circumstances.

[0043] Example 1:

[0044] This embodiment discloses a distributed artificial weather modification operation collaborative instruction issuance system based on the Internet of Things, which aims to generate and issue automated task instructions for regional artificial weather modification operations in a target area. The determined time window is generally within 72 hours.

[0045] In this embodiment, weather modification involves using technological means to intervene in local atmospheric physical processes to achieve specific weather control objectives; in this embodiment, it specifically refers to the handling of rain clouds. Its core operational methods include:

[0046] Artificial rainmaking involves seeding clouds with catalysts such as silver iodide and dry ice using aircraft, rockets, or ground-based smoke generators to promote ice crystal formation or water droplet condensation, thereby increasing precipitation efficiency.

[0047] Artificial rain suppression involves pre-emptively catalyzing precipitation upstream of the target area or excessively seeding ice nuclei to inhibit raindrop formation, thus ensuring favorable weather conditions for major events.

[0048] This system automates the entire work process through distributed control and precise collaboration. The core mechanism is as follows:

[0049] Real-time monitoring and data fusion: Integrating multi-source data such as meteorological satellites, radar, and drones, a three-dimensional monitoring model of cloud water resources is constructed, while edge nodes transmit information such as equipment status and catalyst balance in real time through IoT modules to achieve dynamic monitoring.

[0050] Next comes the generation of collaborative operation instructions. Based on numerical weather prediction models and AI algorithms, combined with cloud structure and catalyst diffusion simulations, a multi-operation-point collaborative scheme is generated. Instructions are then issued, transmitted to the operation terminals via multiple links including 5G and satellite communication, supporting breakpoint resume and encrypted authentication. The terminal devices automatically execute instructions and provide status feedback through BeiDou time synchronization and local logic judgment, requiring no manual intervention.

[0051] To evaluate the effectiveness of the system's command issuance, radar, satellite remote sensing, and ground observation data were used, combined with an AI quantitative evaluation model, to analyze changes in cloud microphysics before and after the operation, thus verifying the rain enhancement efficiency. The system continuously optimizes its operational strategies through machine learning, such as adjusting catalyst type and dissemination location based on historical data.

[0052] Specifically, the system reference in this embodiment Figure 1 As shown in the architecture, it includes:

[0053] The cloud platform layer, serving as the data control center, includes a sensing model that acquires and fuses multi-source meteorological data for the target area through data channels, as well as a rain cloud tracking and prediction model and a task model based on a large language model.

[0054] The terminal equipment layer, serving as the underlying data collection and execution end, includes meteorological monitoring equipment, artificial weather modification equipment, and auxiliary equipment; and

[0055] The edge node layer, as the middle layer of the distributed architecture, is deployed in regional operation centers or edge computing servers. It includes data processing stations and device management gateways. The data processing stations process and cache instructions from the cloud platform layer and data from the terminal device layer within their management scope, while the device management gateway distributes instructions and performs protocol conversion of data.

[0056] As the middle layer of the distributed architecture, it is deployed in regional operation centers or edge computing servers, undertaking local data processing, real-time control logic execution, and collaborative interaction with the cloud platform. For the edge node layer, the administrative divisions and topographical conditions within the target area are comprehensively planned to determine the smallest meteorological management unit within the defined target area as a node. Node servers are deployed there, interacting with the cloud platform layer servers via wired communication networks and with terminal devices via wireless IoT networks.

[0057] The purpose of setting up the edge node layer includes:

[0058] Data preprocessing and aggregation: receiving raw meteorological data uploaded by terminal devices, performing noise reduction, calibration, and local storage, such as caching high-frequency data within 10 minutes, filtering invalid information, and uploading key data, such as cloud movement trajectory and real-time water content, to the cloud platform to reduce network transmission pressure.

[0059] Real-time status monitoring and local control: Monitor the operational status of all terminal devices within the jurisdiction in real time, such as the ammunition reserves of rocket launchers and the flight endurance of drones. In case of emergency such as network interruption, emergency operations can be performed based on locally preset rules to ensure system reliability.

[0060] Command distribution and collaborative scheduling: receiving work plans issued by the cloud platform, such as launching 3 rockets in area A at 9:00 tomorrow, breaking them down into specific equipment control commands, and coordinating multiple devices to execute them in a time sequence.

[0061] Edge nodes are regionalized distributed units that collaborate with the cloud platform in a model of local computing and global optimization. Edge nodes upload pre-processed regional data and device status to the cloud platform, and receive global operational plans and model parameter updates from the cloud platform. The cloud platform provides edge nodes with global meteorological data, operational model algorithm support, and coordinates operational plans across regions for each edge node.

[0062] And reference Figure 1 Architecture diagram and Figure 2 As shown in the flowchart, the edge node layer interacts with the cloud platform layer and the terminal device layer through IoT network data. The construction of the IoT network includes the following technical means:

[0063] In remote areas, edge nodes connect to the cloud platform via satellite communication, receive global operational plans, and distribute instructions to terminal devices. In areas with dense terminal devices, edge nodes communicate with terminal devices via LoRa or Mesh networks, reducing the load on satellite links. For example, multiple ground-based flues can be connected to edge nodes via LoRa, and then the edge nodes can upload data via satellite.

[0064] Many mobile terminal devices are typically operated manually, but can also be operated unmanned, using a separate satellite communication method similar to Starlink for positioning and data exchange.

[0065] Furthermore, the perception model acquires multi-source meteorological data for the target area through online data channels, processes the multi-source data, and performs weighted data fusion to form operational reference data for the target area. The online data acquired by the perception model includes local monitoring data obtained through several meteorological detection devices at the terminal device layer, as well as meteorological satellite images acquired through the access network. The perception model uses image processing algorithms to identify and process the acquired meteorological satellite images and obtain meteorological satellite data.

[0066] After cleaning and spatiotemporally unifying local monitoring data and meteorological satellite data, the perception model assigns weights to the spatiotemporally labeled data using a dynamic weighting method. It then uses a multimodal Transformer model to perform spatiotemporal correlation and generate a grid fusion feature tensor for the target area. Finally, the perception model outputs a grid dataset and a time-series dataset. The rain and cloud tracking prediction model processes the grid dataset and the time-series dataset and outputs visualization data of rain and cloud changes within the target area up to 72 hours, as well as the predicted trajectory changes of rain and clouds.

[0067] Data cleaning for the perception model mainly includes outlier detection and repair. For example, temperature values ​​>50℃ or <-50℃ are removed as obvious errors, and radar reflectivity is filtered to remove ground clutter <0dBZ. Furthermore, machine learning is employed, using a spatiotemporal interpolation algorithm combined with neighboring nodes and historical data to predict missing values. Outliers are detected through isolated forests and iteratively corrected using Kalman filtering.

[0068] The so-called spatiotemporal unification processing refers to the process where terminal devices use BeiDou / GPS time synchronization to align with the Coordinated Universal Time (UTC) of network data, and use linear interpolation to synchronize asynchronous data to a unified time grid. Meanwhile, the latitude of local device data is projected to UTM with a conversion error of less than 0.5 meters, and then resampled to the unified grid of the target area using bilinear interpolation.

[0069] Furthermore, the task model processes and obtains the optimal catalytic window T for rain clouds based on the target area and the rain cloud change data obtained by the rain cloud tracking and prediction model. Then, based on the grid dataset, time series dataset, and visualization data of rain cloud predicted trajectory changes obtained by the perception model, it determines the required location of the artificial intervention operation equipment and the corresponding preliminary instructions for operation. The cloud platform layer sends the instructions to the edge node layer, and the edge node layer processes the instructions to form the task instructions corresponding to the terminal devices within the communication range.

[0070] The task model is trained based on an existing large language model. During training, historical meteorological data is collected, including grid datasets and time-series datasets output by the perception model, as well as visualization data of rain cloud predicted trajectory changes output by the rain cloud tracking and prediction model. The collected data is cleaned, normalized, and feature extracted. Then, the actual operation and effect evaluation data of the corresponding artificial weather modification operations are recorded. The preprocessed data is divided into training, validation, and test sets according to the proportion. The training set is used for model parameter learning, the validation set is used to adjust the model's hyperparameters, and the test set is used to evaluate the model's final performance. A deep learning model is used to process time-series data and spatial features. Historical meteorological data is used as input, and the corresponding optimal catalytic window, operation equipment type, location, and operation instructions are used as outputs. The model parameters are continuously adjusted through the backpropagation algorithm to output task instructions.

[0071] Furthermore, when issuing instructions, if the original task cannot be executed due to weather changes or equipment failure, a safeguard mechanism is set up based on the system to automatically adjust the task instructions by utilizing edge nodes to provide redundant alternative solutions, as follows:

[0072] Based on the original three-layer cloud-edge-device architecture and the scenario of artificial weather modification operations, this paper addresses the need for cloud-based 72-hour task formulation based on device information and edge node local adjustment in the event of sudden failure of old equipment, and achieves the goal through logical refinement.

[0073] Before generating the 72-hour task plan, the cloud performs a full-area scan of devices in the target area through the device management gateway at the edge node layer, including:

[0074] The operational status of aging equipment is monitored by collecting data such as mechanical wear and ammunition reserves from local sensors at edge nodes, as well as by providing operational status data from status monitoring sensors installed on the equipment.

[0075] The core parameter library of the equipment, including data such as maximum range, single catalyst dosage, and collaborative response delay of new equipment, is fused by the perception model to generate an equipment capability matrix. The matrix includes equipment type, location, and substitutability weights.

[0076] When generating a cloud-based task model based on rain cloud prediction data, it must include the main task instruction and alternative adjustment parameters, where the alternative parameters are associated with the performance data of alternative devices in the device capability matrix.

[0077] (1) Fault detection and triggering conditions

[0078] The data processing station at the edge node monitors the status of aging equipment in real time through the following methods:

[0079] For older equipment that requires manual operation, operation feedback signals are collected via LoRa wireless communication, such as feedback from the transmit button trigger or the ammunition loading completion signal. If no preset signal is received for 30 consecutive seconds, it is determined that the equipment cannot operate.

[0080] By combining the lightweight model of the local decision-making module and analyzing historical fault data of the equipment, such as the sudden increase in the failure rate of a certain type of rocket rack below -5℃, a potential fault can be warned 10 minutes in advance.

[0081] Local adjustments are triggered when the following conditions are met: the faulty device is the core execution unit of the current task; the cloud communication is interrupted or the response delay exceeds 1 minute, as determined by the network quality monitoring module of the edge node; and the remaining time until the optimal catalytic window T is less than 30 minutes.

[0082] The edge node local decision module initiates adjustment logic, calls the equipment coordination strategy library generated by training historical schemes based on the original file task model, selects alternative equipment from the equipment capability matrix, such as replacing 1 faulty rocket launcher with 2 ground smoke generators, and calculates the substitution coefficient, such as the catalytic effect of 0.3 rockets for 1 hour of spraying by 1 smoke generator, to ensure that the total amount of catalyst meets the standard.

[0083] Simultaneously, parameter corrections are performed, adjusting the operating parameters of the remaining equipment, including:

[0084] Dosage: If the faulty rocket launcher was originally designed to spray 300g per launch, and is replaced with two smoke generators, the amount of sprayed per launcher per launch will be increased to 200g, and the total dosage will be 300g × 1.3 times the compensation coefficient.

[0085] Position offset: Based on the real-time trajectory of the rain cloud, the working position of the alternative equipment will be offset by 1-2km in the direction of the rain cloud's movement;

[0086] Cooperative scheduling: Send cooperative instructions to surrounding devices through the device management gateway and synchronously correct the device startup sequence.

[0087] Edge nodes collect micro-meteorological data in real time after the operation through local meteorological monitoring equipment and compare it with preset effect thresholds. For example, if the cloud top temperature drops by ≥2℃ 10 minutes after catalysis, if the target is not met, a second adjustment is initiated to further increase the amount of pesticides dispensed by nearby equipment until the core objective of the overall mission is met, that is, the rain enhancement efficiency is maintained at more than 85% of the original plan.

[0088] One specific embodiment of the present invention takes a city marathon event support scenario as an example to demonstrate how the system can generate and coordinate the execution of rain suppression operation instructions within 72 hours.

[0089] The cloud platform layer serves as the core control center, synchronously accessing multi-source meteorological data through data channels: on the one hand, it collects data from local monitoring equipment deployed at the terminal equipment layer, such as micro meteorological stations that acquire real-time temperature, humidity, wind speed, and wind direction, and millimeter-wave radars that monitor reflectivity and vertical liquid water content in real time; on the other hand, it acquires FY-4B satellite cloud images and numerical forecast grid data from the National Meteorological Administration through network interfaces.

[0090] The perception model preprocesses two types of data. First, it uses image recognition algorithms to extract key features such as cloud boundaries and TBB gradients from satellite cloud images. Then, it fuses local radar reflectivity data with satellite TBB data using dynamic weights. Radar data is assigned a weight of 0.6 due to its high spatiotemporal resolution, while satellite data is assigned a weight of 0.4 due to its wide coverage. The fused data is then processed by a multimodal Transformer model to generate a grid fusion feature tensor with a resolution of 200 meters, containing 12 meteorological parameters such as temperature, humidity, and reflectivity. This ultimately forms the grid dataset and time series dataset for subsequent models.

[0091] The rain cloud tracking and prediction model, based on the output data of the perception model, first processes the fused meteorological data using U-Net+ and a network to accurately segment the rain cloud outline. Then, combining the random forest algorithm, it identifies the cloud type as cumulonimbus based on features such as a VIL value of 12 kg / m² and a TBB value of 225 K. Finally, it uses a Kalman filter algorithm to dynamically track the cloud's movement trajectory, determining that the rain cloud is moving southeast at a speed of 15 m / s.

[0092] The model further combines localized WRF numerical simulation with LSTM neural networks to comprehensively analyze the thermodynamic and dynamic characteristics of the cloud, predicting that the rain cloud will pass over the stadium between 10:00 and 12:00 on the day of the event. Through real-time monitoring and analysis of microphysical parameters such as cloud top height and ice crystal concentration, the optimal catalytic window was determined to be 1-2 hours before the rain cloud arrives, i.e., 8:00-9:00. At this time, the cloud top height is stable at 4-5 km, and the ice crystal concentration can be effectively increased to 5 crystals / L by seeding with silver iodide, providing ideal physical conditions for rain suppression operations.

[0093] Based on the above predictions, the task model first extracts similar scenario data from historical data that includes rain dissipation operations for similar events in 2023. After preprocessing such as cleaning and normalization, this data is input into a finely tuned large language model. When generating operational plans, the model fully considers equipment capabilities and safety constraints, such as prohibiting the use of rocket launch equipment within 5km of the stadium and limiting the single flight time of drones to 60 minutes.

[0094] Based on this, preliminary operational instructions were generated to deploy three rocket launchers 50km upstream of the stadium, each equipped with 20 silver iodide rockets, with a catalyst dosage of 300g per rocket. These rockets were to be launched in batches between 8:30 and 9:00 to consume the water vapor in the cloud in advance. At the same time, five drones with a maximum payload of 10kg were dispatched to disperse salt particles at a height of 3km above the ground in the middle of the cloud according to a preset route. The salt particle size was controlled to be 60 micrometers to suppress the raindrop merging process.

[0095] Considering that the speed of rain cloud movement may fluctuate by 10%, the task model automatically deploys two drones as redundant equipment at the alternative operation points based on historical operation experience and real-time data uncertainty analysis. Dynamic triggering conditions are set so that when the main equipment fails or the speed of rain cloud movement suddenly increases, the redundant equipment will automatically start to perform supplementary seeding tasks to ensure the stability and effectiveness of rain suppression operations.

[0096] The edge node layer deploys node servers with the administrative region where the event is located as the smallest meteorological management unit. It achieves high-speed data interaction with the cloud platform through wired fiber optic networks, and connects with terminal devices with the help of 5G communication networks and LoRa wireless communication technology to build a communication network covering the entire target area.

[0097] Furthermore, based on the equipment failure adjustment mechanism, in this implementation, the cloud-based solution involves two old rocket launchers in area A each launching three rockets daily between 8:00 and 9:00 AM, in conjunction with a drone in area B for seeding. On the second day of the mission, the edge node detected no launch feedback signal from R1 and determined it to be a failure. At this time, there were 25 minutes remaining until the optimal window T, and cloud communication was interrupted due to heavy rain. The edge node activated the strategy library and calculated that one drone, R2, in area B could replace R1 and R2: instructing R2 to increase the single-shot explosive charge from 300g to 450g and the number of launches from 3 to 4; instructing the drone to shift its seeding route 1.5km towards area A to cover the original R1 operating area; and simultaneously activating a backup smoke generator in area C, starting seeding 10 minutes earlier to make up for the time difference.

[0098] Effect verification: Local radar monitoring showed that the temperature drop at the top of the clouds in the adjusted area was no more than 1.2℃ lower than the original plan, indicating that the rain suppression effect met the standard.

[0099] The node server's built-in data processing station performs localized parsing and caching of operation instructions issued from the cloud, and performs secondary verification and optimization based on local real-time meteorological data to ensure the timeliness and accuracy of the instructions. The device management gateway is responsible for protocol conversion of different types of terminal devices, enabling operation equipment such as rocket launchers and drones to accurately receive and execute instructions, while simultaneously transmitting equipment status and operation data back in real time, forming a complete closed loop from data acquisition and instruction generation to operation execution, providing reliable meteorological support.

[0100] This invention is not limited to the optional embodiments described above, and anyone can derive other various forms of products based on the inspiration of this invention. The specific embodiments described above should not be construed as limiting the scope of protection of this invention; the scope of protection of this invention should be determined by the claims, and the specification can be used to interpret the claims.

Claims

1. A distributed weather modification operation collaborative instruction issuance system based on the Internet of Things, which uniformly issues task instructions for weather modification operations within a defined target area, characterized by: include: The cloud platform layer, serving as the data control center, includes a sensing model that acquires and fuses multi-source meteorological data for the target area through data channels, as well as a rain cloud tracking and prediction model and a task model based on a large language model. The terminal equipment layer serves as the underlying data collection and execution end, including meteorological monitoring equipment, artificial intervention equipment, and auxiliary equipment; as well as The edge node layer, deployed as the middle layer of the distributed architecture in regional operation centers or edge computing servers, includes data processing stations and device management gateways. The data processing stations process and cache instructions from the cloud platform layer and data from the terminal device layer within their management scope, while the device management gateway distributes instructions and performs protocol conversion of data. The edge node layer also includes a local decision-making module, which processes the acquired data from the terminal device layer using a locally deployed and trained large language model to obtain local meteorological status data and uploads it to the cloud platform layer. The initial instructions issued by the cloud platform layer are processed. If the initial instructions from the cloud are not obtained within the set rain cloud window time, or if the cloud communication delay is greater than 1 minute, the local decision module of the edge node layer calls the pre-stored device capability matrix, calculates the device substitution coefficient through the local large language model, generates the task instruction of the total catalyst quantity deviation threshold, and corrects the startup sequence of the surrounding devices. The perception model acquires multi-source meteorological data for the target area through online data channels, and performs weighted data fusion after processing the multi-source data to form operational reference data for the target area. The online data acquired by the perception model includes local monitoring data acquired by several meteorological detection devices at the terminal device layer, as well as meteorological satellite images acquired through the access network. The perception model uses image processing algorithms to identify and process the acquired meteorological satellite images and obtain meteorological satellite data. The perception model cleanses and spatiotemporally unifies local monitoring data and meteorological satellite data. It then assigns weights to spatiotemporally labeled data using a dynamic weighting method and generates a grid fusion feature tensor for the target area through a multimodal Transformer model. Finally, the perception model outputs a grid dataset and a time-series dataset, which are then processed by the rain cloud tracking and prediction model to output rain cloud change data for the target area up to 72 hours in length. This rain cloud change data includes visualized data on cloud top height, ice crystal concentration, movement speed, and predicted rain cloud trajectory changes. Specifically, the cloud top height data accuracy is no greater than 50 meters, the movement speed calculation error is less than 0.5 m / s, and the trajectory prediction deviation is less than 1 km. The task model processes and obtains the optimal catalytic window T of the rain cloud based on the target area and rain cloud change data. The optimal catalytic window T must meet the following conditions: the cloud top height is in the range of 4-5km, the ice crystal concentration does not exceed 5 crystals / L, and the rain cloud arrives at the target area 1-2 hours before the rain cloud arrives. Then, based on the grid dataset, time series dataset, and visualization data of rain cloud predicted trajectory changes obtained by the perception model, the operation parameters and preliminary instructions of the artificial influence operation equipment are determined. The operation parameters include operation location, catalyst type and dosage, and operation timing. The cloud platform layer sends the preliminary instructions to the edge node layer, and the edge node layer processes the preliminary instructions to form the task instructions corresponding to the terminal devices within the communication range.

2. The distributed artificial weather modification operation collaborative instruction issuance system based on the Internet of Things as described in claim 1, characterized in that: The edge node layer includes the smallest meteorological management unit within a defined target area as a node, with node servers deployed. It interacts with the cloud platform layer server through a wired communication network and with terminal devices through a wireless IoT communication network.

3. The distributed artificial weather modification operation collaborative instruction issuance system based on the Internet of Things as described in claim 1, characterized in that: The meteorological monitoring equipment includes weather radar, anemometer, rain gauge, thermometer, hygrometer, barometer, and weather balloon; The artificial intervention equipment includes ground-based launching equipment and aerial seeding equipment.

4. The distributed artificial weather modification operation collaborative command issuance system based on the Internet of Things as described in claim 1, characterized in that: The task model is trained based on the existing large language model. During training, historical meteorological data is collected, including grid datasets and time-series datasets output by the perception model, as well as visualization data of rain cloud predicted trajectory changes output by the rain cloud tracking and prediction model. The collected data is cleaned, normalized, and feature extracted. Then, the actual operation and effect evaluation data of the corresponding artificial weather modification operation are recorded. The preprocessed data is divided into training set, validation set and test set. The training set is used for learning the model's parameters, the validation set is used to tune the model's hyperparameters, and the test set is used to evaluate the model's performance. A deep learning model is used to process time series data and spatial features. Historical meteorological data is taken as input, and the corresponding optimal catalytic window, type of operating equipment, location and operation instructions are taken as output. The model parameters are continuously adjusted through the backpropagation algorithm to output task instructions.

5. The distributed artificial weather modification operation collaborative command issuance system based on the Internet of Things as described in claim 1, characterized in that: When generating initial instructions, the task model first obtains device information from the cloud platform through the device management gateway at the edge node layer, which performs a full-domain scan of devices in the target area. The task model then generates a device capability matrix based on the device information. The initial instructions generated by the task model include alternative adjustment parameters, which are associated with the performance data of alternative devices in the device capability matrix. When a terminal device malfunctions during task execution, the local decision-making module of the edge node initiates adjustment logic, calls pre-stored alternative adjustment parameters, selects alternative devices from the device capability matrix, calculates the substitution coefficient to correct the parameters, keeps the total catalyst quantity change value less than the set threshold, and sends collaborative instructions to surrounding devices through the device management gateway to synchronously correct the device startup sequence.

Images