Simulation risk simulation deduction early warning method and system

Abstract

The application provides a simulation risk simulation deduction early warning method and system, which comprises the following steps: obtaining multi-dimensional situation data corresponding to emergency materials, and performing standardization preprocessing on the multi-dimensional situation data to obtain material base data; inputting the material base data and disaster monitoring data into a large model to obtain a virtual simulation scene of disaster evolution and material demand coupling; based on the virtual simulation scene, deploying a multi-agent cluster to perform autonomous interaction deduction to obtain material collaborative deduction data; according to the material collaborative deduction data, performing parallel simulation iteration under a preset differentiated boundary condition, and based on the full-link material management and control index data in the simulation iteration process, constructing a simulation evaluation data set; quantitatively evaluating the simulation evaluation data set to generate a material guarantee strategy, and outputting risk warning information and a pre-positioned guarantee scheme according to the material guarantee strategy. The application realizes proactive defense before a disaster for emergency material guarantee, and guarantees accurate scheduling of emergency materials.

Description

Technical Field

[0001] This invention relates to the field of disaster early warning technology, and in particular to a simulation risk simulation and early warning method and system. Background Technology

[0002] Current emergency supplies support simulations remain at the level of static data statistics and fixed templates, failing to achieve a deep integration of large-scale intelligent cognition and dynamic simulation across all scenarios.

[0003] Traditional models rely on established historical disaster templates to formulate fixed material deployment plans, which can only conduct fixed-point drills for single conventional disasters and cannot simulate complex pressure scenarios such as multi-disaster coupling, multi-point concurrency, road network interruption, and warehouse over-limitation. At the same time, existing plans cannot dynamically link real-time warehouse capacity, on-the-way transportation capacity and other dynamic variables, making it difficult to adapt to the dynamic needs of emergency material full-chain support under complex disasters.

[0004] Therefore, there is an urgent need for a simulation-based risk simulation and early warning method and system to solve the above problems. Summary of the Invention

[0005] To address the problems existing in the prior art, this invention provides a simulation risk simulation and early warning method and system.

[0006] This invention provides a simulation risk modeling and early warning method, comprising: Acquire multi-dimensional situational data corresponding to emergency supplies, and perform standardized preprocessing on the multi-dimensional situational data to obtain basic data on the quantity of supplies; The material baseline data and disaster monitoring data are input into the large model to obtain a virtual simulation scenario that couples the disaster evolution and material demand, output by the large model. Based on the virtual simulation scenario, a multi-agent cluster is deployed to perform autonomous interactive simulations to obtain material collaborative simulation data. Based on the material collaborative simulation data, parallel simulation iterations are performed under preset differentiated boundary conditions, and a simulation evaluation dataset is constructed based on the full-link material control index data during the simulation iteration process. The simulation evaluation dataset is quantitatively evaluated to generate a material support strategy, and based on the material support strategy, risk warning information and pre-emptive support plan are output.

[0007] According to the simulation risk simulation and early warning method provided by the present invention, the step of acquiring multi-dimensional situational data corresponding to emergency supplies includes: Based on emergency material reserve warehouses, RFID inventory terminals, transportation capacity dispatching platforms, GPS positioning terminals, and meteorological monitoring stations, information on the types of materials in storage, available storage capacity, batches of materials, transportation capacity, road network capacity, and disaster monitoring indicators can be obtained. The standardization preprocessing of the multidimensional situational data includes: The multidimensional situational data is cleaned, anomaly removed, standardized, coded, and time-series aligned to obtain the material baseline data.

[0008] According to the present invention, a simulation risk modeling and early warning method is provided, wherein the basic material data and disaster monitoring data are input into a large model to obtain a virtual simulation scenario coupled with disaster evolution and material demand output by the large model, including: Based on the large model, the spatiotemporal coupling relationship between disaster evolution and material demand is analyzed. Based on the spatiotemporal coupling relationship, scenarios such as conventional material supply guarantee, multi-disaster superposition pressure scenario, road network blockage supply scenario, and warehouse overload diversion scenario are generated. The peak material demand, warehouse capacity crowding nodes, and transportation congestion nodes under different scenarios are determined.

[0009] According to the present invention, a simulation risk simulation and early warning method is provided, wherein the multi-agent cluster is a multi-department, multi-management role material control intelligent agent, and the multi-department, multi-management role material control intelligent agent is embedded with a material business-specific cognitive reasoning engine and reinforcement learning algorithm. Based on the virtual simulation scenario, a multi-agent cluster is deployed to perform autonomous interactive simulations to obtain material collaborative simulation data, including: The multi-department, multi-management role intelligent agent for material control is deployed in the virtual simulation scenario. Based on historical scheduling cases and warehousing specifications, it executes adaptive behaviors corresponding to the autonomous interactive simulation to obtain the material collaborative simulation data. The adaptive behaviors include at least manual reconciliation, warehouse capacity allocation, transportation capacity anchoring, and cross-warehouse linkage behaviors.

[0010] According to a simulation risk simulation and early warning method provided by the present invention, the method involves performing parallel simulation iterations under preset differentiated boundary conditions based on the material collaborative simulation data, and constructing a simulation evaluation dataset based on the end-to-end material control indicator data during the simulation iteration process, including: The preset differentiated boundary conditions are determined based on the differentiated disaster intensity, the timing of material shortages, and road network constraints. Based on the preset differentiated boundary conditions, multiple rounds of material closed-loop simulation are executed in parallel. During the multi-round closed-loop simulation of materials, the simulation evaluation dataset is obtained based on the collected data on material arrival timeliness, warehouse capacity utilization rate, transportation capacity empty rate, and end-point shortage duration.

[0011] According to a simulation risk simulation and early warning method provided by the present invention, the step of quantitatively evaluating the simulation evaluation dataset, generating a material support strategy, and outputting risk early warning information and pre-emptive support plan based on the material support strategy includes: Based on the large model, a multidimensional weighted quantitative evaluation is performed on the simulation evaluation dataset to output the material support strategy. The multidimensional weighted quantitative evaluation includes comparing the adaptability, anti-disturbance capability, timeliness capability, and cost control capability of different material pre-positioning, transportation capacity grouping and batch replenishment schemes. Based on the supply and demand pressure situation corresponding to the material support strategy, the system generates tiered early warning information for material supply and demand imbalance, transportation capacity gap, and supply disruption across the entire region, and outputs a support plan including a warehouse capacity optimization list, a pre-positioned material quota table, and a transportation capacity grouping plan.

[0012] The present invention also provides a simulation risk simulation and early warning system, comprising: The data acquisition and processing module is used to acquire multi-dimensional situational data corresponding to emergency supplies, and to perform standardized preprocessing on the multi-dimensional situational data to obtain the basic data of the supplies. The coupled dynamic scene generation module is used to input the material base data and disaster monitoring data into the large model to obtain a virtual simulation scene that couples the disaster evolution and material demand output by the large model. The autonomous interactive reasoning module is used to deploy a multi-agent cluster to perform autonomous interactive reasoning based on the virtual simulation scenario, and obtain material collaborative reasoning data; The simulation iteration module is used to perform parallel simulation iteration under preset differentiated boundary conditions based on the material collaborative simulation data, and to construct a simulation evaluation dataset based on the full-link material control index data during the simulation iteration process. The simulation and early warning result output module is used to quantitatively evaluate the simulation evaluation dataset, generate material support strategies, and output risk warning information and pre-emptive support plans based on the material support strategies.

[0013] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the simulation risk simulation and early warning method as described above.

[0014] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the simulation risk simulation and early warning method as described above.

[0015] The present invention also provides a computer program product, including a computer program that, when executed by a processor, implements the simulation risk simulation and early warning method as described above.

[0016] The simulation risk simulation and early warning method and system provided by this invention dynamically generates virtual simulation scenarios that couple disaster evolution and material demand through a large model, and replaces manual collaborative reconciliation and route trial calculation based on multi-agent cluster autonomous interactive simulation, greatly improving simulation efficiency and depth; furthermore, through parallel simulation iteration and multi-dimensional weighted quantitative evaluation under differentiated boundary conditions, it outputs the optimal material support strategy, realizes proactive defense of emergency material support before disasters, and ensures the accurate scheduling of emergency materials. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0018] Figure 1 A flowchart illustrating the simulation risk simulation and early warning method provided by this invention; Figure 2 This is a schematic diagram of the structure of the simulation risk simulation and early warning system provided by the present invention; Figure 3 This is a schematic diagram of the structure of the electronic device provided by the present invention. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0020] Emergency supplies are the core resources for disaster response and on-site rescue. Their adequacy, rational storage, efficient allocation, and smooth coordination directly determine the outcome of emergency response to sudden events.

[0021] Currently, most mainstream emergency material support drills and risk assessments in the industry adopt a model of "offline, real-world drills, fixed paper-based plans, and outdated material dispatch case studies" to conduct comprehensive material support verification. In practical applications, both traditional models and existing auxiliary systems focus on core aspects such as emergency material reserves, dispatch, and transportation capacity allocation. However, they consider limited variable factors and fail to consider extreme cases such as global variables. For example, existing systems typically only perform static analysis for a single disaster type, a single reserve, and fixed transportation routes, failing to incorporate complex factors such as weather evolution, road network dynamics, social transportation capacity fluctuations, and multi-reservoir collaborative operations into a unified analysis framework.

[0022] Meanwhile, existing technologies have not achieved a deep integration of large-scale intelligent cognition and full-scenario dynamic simulation. Traditional emergency material management systems lack the core capabilities of real-time perception of the overall material situation, adaptive game theory simulation, and iterative optimization. They remain at the level of static data statistics and cannot adapt to the dynamic needs of full-chain emergency material support under complex disaster situations.

[0023] Based on actual emergency supplies management scenarios, existing technologies have the following shortcomings, which seriously affect the effectiveness of emergency supplies support: The scenarios are insufficiently covered and statically fixed. Traditional emergency material support simulations rely on pre-defined historical disaster templates to formulate fixed material distribution plans. They can only conduct targeted material distribution drills for single conventional disasters and cannot simulate complex material support pressure scenarios caused by multiple disasters coupled together, multiple points of concurrency, road network disruptions, and warehouse overcapacity. At the same time, existing plans cannot dynamically link with dynamic variables such as real-time warehouse capacity, on-transit transportation capacity, and the amount of available materials in the local area. The scenario update cycle is long and lagging, making it difficult to adapt to the real-time fluctuations in material supply and demand. When extreme disasters occur, the actual material demand often far exceeds the plan settings, resulting in delayed discovery of reserve gaps and slow response to allocation orders.

[0024] The simulations are inefficient and lack depth. Current material support simulations rely heavily on manual labor, requiring collaboration from multiple departments and personnel to manually assess reserve gaps, plan transportation routes, and verify coordination points. This not only results in high labor costs and lengthy simulation times but also fails to deeply simulate the complex real-world material operations under sudden disasters, such as material losses, warehouse capacity constraints, transportation congestion, and dynamic shortage replenishment. Manual simulations are limited by the experience level and collaboration efficiency of participants, making it difficult to complete large-scale, multi-scenario, and high-concurrency material support stress tests within a limited timeframe, and thus unable to reflect the overall pressure limits of material support.

[0025] The output solutions lack adaptability and scientific guidance. Traditional models output static, fixed templates for material support, failing to consider massive uncertainties and disturbances such as extreme weather disrupting road networks, differences in the shelf life of batches of materials, and temporary shortages in social transportation capacity. There is a lack of quantitative simulation and evaluation of the completeness, dynamic adaptability, cost controllability, and timeliness of the entire material supply chain solution, making it impossible to scientifically guide pre-positioning of emergency material storage locations, pre-positioning of transportation capacity, and cross-warehouse coordination. Managers often make decisions based on experience, leading to unreasonable layout of pre-positioning points, unbalanced reserve structures, and wasted transportation resources.

[0026] A fully closed-loop management mechanism has not been established. Existing technologies have not achieved closed-loop management encompassing "risk prediction, material simulation, strategy optimization, and precise pre-deployment." Simulation results are disconnected from actual material preparation work, making it impossible to translate optimized strategies into pre-disaster deployment actions and hindering the construction of a proactive material risk prediction mechanism. Traditional models remain at the stage of passive allocation and replenishment after disasters, failing to complete the simulation, optimization, and pre-deployment of material support plans before disasters strike, thus missing the golden preparation period for emergency material management.

[0027] To address the aforementioned issues, this invention establishes an integrated system of large-scale models and digital simulations, integrating multi-dimensional data on material storage, transportation capacity, supply and demand, and geographic and meteorological conditions to dynamically generate virtual scenarios that couple disasters and materials. Simultaneously, based on large-scale model technology, it constructs a multi-agent cluster for material management with autonomous cognitive and decision-making capabilities, replicating all business behaviors and conducting autonomous interactive simulations. Through high-frequency simulation iterations, it quantifies and optimizes material support strategies for output.

[0028] Figure 1 This is a flowchart illustrating the simulation risk simulation and early warning method provided by the present invention, as shown below. Figure 1 As shown, the present invention provides a simulation risk simulation and early warning method, comprising: Step 101: Obtain multi-dimensional situational data corresponding to emergency supplies, and perform standardized preprocessing on the multi-dimensional situational data to obtain the basic data of the supplies.

[0029] This invention can be specifically connected to emergency material reserve depots, RFID inventory terminals, transportation dispatch platforms, GPS positioning terminals, and meteorological monitoring stations to collect real-time data on the types of materials in storage, available storage capacity, material batches, transportation capacity information, road network capacity, and disaster monitoring indicators. This acquired data covers multiple dimensions, including material warehousing, transportation dispatch, supply and demand matching, and geographic and meteorological dimensions, forming a comprehensive, multi-dimensional situational data database of materials.

[0030] Then, the acquired multidimensional situational data is subjected to standardized preprocessing, including data cleaning, anomaly removal, standardized coding, and time sequence alignment, to build a dynamic material database that integrates materials, space, and disaster situation, thereby ensuring that the simulation data is real-time, accurate, and traceable.

[0031] Step 102: Input the material baseline data and disaster monitoring data into the large model to obtain a virtual simulation scenario that couples the disaster evolution and material demand, output by the large model.

[0032] In this invention, a large-scale material support model with special fine-tuning can be used to deeply link pre-processed material baseline data with disaster monitoring data to analyze the spatiotemporal coupling relationship between disaster evolution and material demand.

[0033] In this invention, the large model adaptively generates diverse stress test scenarios based on the analysis results, including routine material supply scenarios, multi-disaster pressure scenarios, road network disruption supply scenarios, and warehouse overload diversion scenarios. It dynamically recreates the real-world situations such as peak material demand, warehouse capacity saturation points, and transportation congestion points under different conditions. For example, in a typhoon disaster scenario, the large model combines typhoon path prediction data, wind force evolution timelines, and rainfall intensity distribution information to generate spatiotemporal distribution maps of material demand corresponding to different typhoon levels, and simultaneously simulates real-world conditions such as road blockages in mountainous areas, limited warehouse capacity at township forward deployment points, and insufficient transportation capacity.

[0034] Step 103: Based on the virtual simulation scenario, deploy a multi-agent cluster to perform autonomous interactive simulation and obtain material collaborative simulation data.

[0035] In this invention, a multi-department, multi-management intelligent agent for material control is deployed in a virtual simulation scenario, including a material warehouse management intelligent agent, a transportation scheduling intelligent agent, and a cross-regional joint guarantee intelligent agent. Each intelligent agent is embedded with a material business-specific cognitive reasoning engine and reinforcement learning algorithm to learn from historical scheduling cases, warehousing standards, emergency plans, etc., and replicate adaptive behaviors such as manual reconciliation, warehouse capacity allocation, transportation capacity anchoring, and cross-warehouse linkage. In the virtual simulation scenario, each intelligent agent autonomously conducts collaborative operations throughout the entire process, including data verification, warehouse capacity allocation, emergency transportation, material diversion, and gap replenishment, replicating the collaborative logic of manual material management and achieving complex material collaborative simulation without human intervention.

[0036] Step 104: Based on the material collaborative simulation data, perform parallel simulation iteration under preset differentiated boundary conditions, and construct a simulation evaluation dataset based on the full-link material control index data during the simulation iteration process.

[0037] In this invention, preset differentiated boundary conditions are determined based on factors such as differentiated disaster intensity, material shortage timing, and road network constraints. Then, based on the preset differentiated boundary conditions, multiple rounds of material closed-loop simulation are executed in parallel, such as setting different parameters for typhoon intensity, material shortage periods, road network blockage range, and linkage response time in batches, and conducting multiple differentiated simulations.

[0038] During the simulation process, core control indicators such as the time it takes for materials to arrive, peak warehouse capacity occupancy, empty transport capacity rate, and time of stockout at the end are automatically recorded throughout the process, forming a simulation material big data set, namely the simulation evaluation dataset.

[0039] Step 105: Quantitatively evaluate the simulation evaluation dataset, generate a material support strategy, and output risk warning information and pre-emptive support plan based on the material support strategy.

[0040] In this invention, a multi-dimensional weighted quantitative evaluation can be performed on the simulation evaluation dataset based on a large model. The adaptability, anti-disturbance capability, timeliness capability and cost control capability of different material pre-layout, transportation capacity grouping and batch replenishment schemes can be compared, and the optimal practical strategy, i.e. material support strategy, can be iteratively output.

[0041] Furthermore, based on the supply and demand pressure situation corresponding to the material support strategy, the system generates tiered early warning information such as material supply and demand imbalance, transportation capacity gap, and supply disruption across the entire region. It also outputs pre-implementable support solutions such as warehouse capacity optimization lists, pre-positioned material quota tables, and transportation capacity grouping plans, and pushes them to the material management dashboard, warehouse control terminal, and transportation capacity scheduling platform with one click, so as to achieve early risk warning and early material deployment.

[0042] The simulation risk simulation and early warning method provided by this invention dynamically generates a virtual simulation scenario that couples disaster evolution with material demand through a large model. Based on the autonomous interactive simulation of a multi-agent cluster, it replaces manual collaborative reconciliation and route trial calculation, greatly improving the efficiency and depth of the simulation. Furthermore, through parallel simulation iteration and multi-dimensional weighted quantitative evaluation under differentiated boundary conditions, it outputs the optimal material support strategy, realizes proactive defense of emergency material support before disasters, and ensures the accurate scheduling of emergency materials.

[0043] Based on the above embodiments, the acquisition of multi-dimensional situational data corresponding to emergency supplies includes: Based on emergency material reserve warehouses, RFID inventory terminals, transportation capacity dispatching platforms, GPS positioning terminals, and meteorological monitoring stations, information on the types of materials in storage, available storage capacity, batches of materials, transportation capacity, road network capacity, and disaster monitoring indicators can be obtained. The standardization preprocessing of the multidimensional situational data includes: The multidimensional situational data is cleaned, anomaly removed, standardized, coded, and time-series aligned to obtain the material baseline data.

[0044] This invention enables targeted integration with multi-source heterogeneous data systems, achieving real-time aggregation and standardized management of emergency supplies data across the entire supply chain. Targeted integration with various data sources can be established through standardized API interfaces, IoT protocols, and direct database connections.

[0045] Specifically, it connects to the management information systems of emergency material reserve depots at all levels to obtain real-time information on the types of materials in storage, including detailed information such as the quantity, specifications, quality grade, storage location, entry time, and expiration date of emergency living supplies (e.g., drinking water, convenience food, and blankets), medical supplies (e.g., first-aid medicines, medical equipment, and protective gear), and engineering rescue supplies (e.g., sand, gravel, water pumps, and generators). Simultaneously, it obtains available storage capacity information for each reserve depot, including total storage capacity, used storage capacity, reserved storage capacity, refrigerated storage capacity, and ambient temperature storage capacity.

[0046] By deploying RFID electronic tags on carriers such as packaging, shelves, and pallets, and combining them with handheld or fixed RFID reader terminals, automated inventory and real-time tracking of materials can be achieved. The system acquires batch information reported by the RFID inventory terminal, including batch number, production date, shelf life, supplier information, and quality inspection report number, ensuring traceability throughout the entire lifecycle of the materials.

[0047] It connects to transportation capacity dispatching platforms, contracted transportation company management systems, and vehicle GPS monitoring platforms to obtain real-time transportation capacity information, including the number of available vehicles, vehicle type distribution, load capacity, current location, operating status, driver information, and estimated arrival time. Simultaneously, it obtains road network capacity information, including real-time traffic status of each road segment, historical average traffic speed, road grade, bridge and tunnel weight limits, and the distribution of construction sections.

[0048] By connecting with vehicle GPS positioning terminals, personnel positioning terminals, and material positioning tags, it can obtain real-time spatial location information such as the geographical coordinates, movement trajectory, and dwell time of transportation vehicles, operators, and key materials, providing spatial data support for subsequent route planning, capacity anchoring, and material tracking.

[0049] By connecting with data sources such as meteorological departments, local meteorological bureaus, and self-built meteorological monitoring stations, disaster monitoring indicators can be obtained in real time, including meteorological and hydrological disaster monitoring data such as typhoon path, wind force level, central pressure, movement speed, rainfall intensity, temperature change, geological disaster risk level, and flood warning level.

[0050] After acquiring the aforementioned multi-dimensional situational data, a standardized preprocessing workflow is executed. First, during the data cleaning phase, a pre-defined rule engine is used to perform quality checks on the raw data, removing abnormal data such as duplicate records, format errors, and logical inconsistencies. Then, statistical analysis methods are used to identify and remove outliers, while business rules are combined to remove illogical data combinations.

[0051] Furthermore, a unified encoding and conversion process is implemented for multi-source heterogeneous data, establishing a standardized data dictionary and encoding system. For example, vehicle descriptions from different transportation companies are uniformly converted into standard vehicle codes (such as mini, light, medium, heavy, and container vehicles), and meteorological element units from different meteorological data sources are uniformly converted into international standard units. Simultaneously, based on a unified time reference, time-series alignment processing is performed on data from different data sources with different sampling frequencies to ensure that data from all dimensions can be correlated, analyzed, and extrapolated within the same time coordinate system.

[0052] After the above-mentioned data cleaning, anomaly removal, standardization coding and time sequence alignment processes, this invention constructs basic material data, providing high-quality, standardized and real-time updated data for subsequent large-scale model scene generation and simulation.

[0053] Based on the above embodiments, the step of inputting the material baseline data and disaster monitoring data into a large model to obtain a virtual simulation scenario output by the large model that couples disaster evolution with material demand includes: Based on the large model, the spatiotemporal coupling relationship between disaster evolution and material demand is analyzed. Based on the spatiotemporal coupling relationship, scenarios such as conventional material supply guarantee, multi-disaster superposition pressure scenario, road network blockage supply scenario, and warehouse overload diversion scenario are generated. The peak material demand, warehouse capacity crowding nodes, and transportation congestion nodes under different scenarios are determined.

[0054] In this invention, a specialized large-scale model for material support, based on specific fine-tuning, achieves deep coupling modeling and dynamic scenario generation of disaster evolution and material demand. This large-scale model is based on the Transformer architecture and, through domain-specific fine-tuning on a general-purpose large-scale language model, incorporates professional knowledge from the field of emergency material management, historical disaster cases, material dispatch rules, and warehousing management standards, forming a specialized large-scale model with cognitive reasoning capabilities in the field of material support.

[0055] Furthermore, the pre-processed basic material data (including the types, quantities, batches, and storage capacity of materials in each reserve warehouse, the material configuration at each forward location, and the distribution of contracted transportation resources, etc.) and disaster monitoring data (including typhoon path, wind force level, rainfall intensity, geological disaster risk, and flood warnings, etc.) are input into the large model. The large model first performs disaster evolution analysis, predicting the evolution trend of the disaster in the next few hours to days based on meteorological disaster dynamics models, historical disaster evolution patterns, and real-time monitoring data, including the spread path of the disaster's impact range, the timing of intensity changes, and the probability of secondary disasters being triggered.

[0056] Then, the large-scale model performs spatiotemporal coupling analysis of material demand. Based on the disaster evolution prediction results, combined with population distribution data, economic density data, infrastructure vulnerability data, and historical material demand response data, the large-scale model establishes a nonlinear mapping relationship between disaster parameters and material demand. For example, based on the historical correlation model between typhoon wind force level and the number of households without power, the large-scale model calculates the demand for generators and emergency lighting equipment under different wind force levels; based on the coupling relationship between rainfall intensity and geological disaster risk level, it calculates the spatiotemporal distribution demand for disaster relief materials such as life jackets, inflatable boats, and sand and gravel.

[0057] Next, based on the spatiotemporal coupling relationship between the disaster evolution and material demand, the large model adaptively generates four typical stress test scenarios, as follows: The first category is the routine material supply scenario, which corresponds to general disasters or routine emergency needs. The intensity of the disaster is within a controllable range, the road network is basically normal, the storage capacity of each reserve is sufficient, and transportation resources are abundant. In this scenario, the large model generates a standard material allocation process, simulating the entire process of material outbound, transportation, distribution, and release under routine needs, serving as a benchmark scenario.

[0058] The second category is the multi-hazard superposition pressure scenario, which simulates a complex pressure scenario formed by the simultaneous occurrence or chain-triggered occurrence of multiple disasters. For example, the passage of a typhoon triggers multiple disasters such as strong winds, torrential rains, floods, and geological disasters, resulting in multiple regions simultaneously raising material demands, and some reserve warehouses facing multi-directional allocation pressure at the same time. The large-scale model generates complex situations such as peak material demand, concurrent allocation conflicts between multiple warehouses, and cross-regional material coordination pressure under the multi-hazard superposition scenario based on the spatiotemporal overlap of each disaster, the intersection of their impact range, and the superposition effect of material demand.

[0059] The third category is the road network disruption and supply scenario. This scenario simulates extreme situations where critical transportation networks are disrupted due to extreme disasters, blocking material transport channels. The large-scale model combines wind speed, rainfall, and geological disaster risk levels from disaster monitoring data to identify potentially damaged road sections, bridges, and tunnels. It simulates complex constraints such as detour transportation of materials after road network disruption, selection of alternative routes, and re-anchoring of transport capacity. For example, when a mountain highway is blocked by a landslide, the large-scale model automatically generates an emergency replenishment plan for materials at forward-positioned reserve depots near that road section, as well as alternative routes for detour distribution through reserve depots in adjacent counties.

[0060] The fourth category is the scenario of diverting excess storage. This scenario simulates the extreme situation where the demand for supplies surges during the peak of a disaster, leading to some reserve warehouses or forward locations exceeding their capacity limits, resulting in stockpiling of supplies and saturation of receiving and dispatching capabilities. Based on the peak demand prediction results, the large model identifies nodes where storage capacity may be squeezed and generates diversion and receiving plans for surrounding reserve warehouses, temporary storage point deployment plans, and material category adjustment plans.

[0061] While generating the four scenarios mentioned above, the large model further determines key node parameters for different scenarios, including peak material demand (maximum material demand in each region and time period), warehouse capacity overcrowding nodes (reservoirs and time points where warehouse capacity utilization reaches or exceeds the warning threshold), and transportation capacity congestion nodes (time periods and regions where transportation demand exceeds available transportation capacity supply). These key node parameters serve as boundary constraints for subsequent multi-agent inference and simulation iteration, ensuring that the virtual simulation scenario realistically reflects the pressure situation of material supply under complex disaster conditions.

[0062] Based on the above embodiments, the multi-agent cluster is a multi-department, multi-management role material control intelligent agent, and the multi-department, multi-management role material control intelligent agent is embedded with a material business-specific cognitive reasoning engine and reinforcement learning algorithm; Based on the virtual simulation scenario, a multi-agent cluster is deployed to perform autonomous interactive simulations to obtain material collaborative simulation data, including: The multi-department, multi-management role intelligent agent for material control is deployed in the virtual simulation scenario. Based on historical scheduling cases and warehousing specifications, it executes adaptive behaviors corresponding to the autonomous interactive simulation to obtain the material collaborative simulation data. The adaptive behaviors include at least manual reconciliation, warehouse capacity allocation, transportation capacity anchoring, and cross-warehouse linkage behaviors.

[0063] In this invention, a multi-department, multi-management intelligent agent for material control is deployed in a virtual simulation scenario to achieve complex collaborative simulation of materials without human intervention. The multi-agent cluster includes, but is not limited to, the following types of intelligent agents: a material warehouse management intelligent agent, responsible for warehouse management operations such as material inbound management, outbound verification, inventory monitoring, batch tracking, and warehouse capacity optimization; a transportation scheduling intelligent agent, responsible for transportation management operations such as vehicle dispatching, route planning, on-the-spot monitoring, capacity matching, and empty-load optimization; a cross-regional joint guarantee intelligent agent, responsible for cross-regional material coordination, collaboration between adjacent warehouses, material allocation, and gap replenishment; a demand acceptance intelligent agent, responsible for front-end operations such as receiving material demand applications from disaster-stricken areas, demand verification, priority ranking, and demand aggregation; and a quality monitoring intelligent agent, responsible for quality control operations such as material quality sampling inspection, shelf-life warning, handling of non-conforming products, and quality traceability.

[0064] In this invention, each intelligent agent is embedded with a cognitive reasoning engine and reinforcement learning algorithm specifically for material operations. The cognitive reasoning engine leverages the semantic understanding and logical reasoning capabilities of a large language model, combined with a knowledge graph in the field of emergency material management (including material classification systems, warehousing management standards, transportation scheduling rules, emergency response procedures, and departmental collaboration mechanisms), to achieve understanding, reasoning, and decision-making within business scenarios. For example, when the material warehouse management intelligent agent receives an emergency material allocation request from a forward location, the cognitive reasoning engine first parses the material category, quantity, and time limit requirements in the request. Then, based on the knowledge graph, it retrieves the corresponding outbound approval process, inventory availability verification rules, and emergency allocation green channel mechanism, ultimately generating an outbound instruction or a shortage warning.

[0065] In this invention, reinforcement learning algorithms can employ frameworks such as Deep Q-Network (DQN) or Proximal Policy Optimization (PPO). Each agent learns the optimal behavioral policy through continuous interaction with the virtual simulation environment. During the training phase, the agent learns offline based on real-world cases in a historical scheduling case library, absorbing expert decision-making experience in scenarios such as material scheduling, warehouse management, and transportation capacity coordination. During the extrapolation phase, the agent autonomously selects actions based on state observations of the current virtual simulation scenario and continuously adjusts its policy parameters according to reward signals from the environment (such as improved material arrival timeliness, optimized warehouse capacity utilization, and reduced transportation capacity idle rate), achieving adaptive behavior optimization.

[0066] During the autonomous interactive simulation, each intelligent agent executes various adaptive behaviors based on historical scheduling cases and warehousing specifications. Specifically, the material warehouse management intelligent agent periodically or under triggered conditions performs inventory reconciliation operations to verify the consistency between the system inventory records and the physical inventory. When an inventory discrepancy is detected, the intelligent agent initiates a discrepancy analysis process. Based on multi-source information such as historical inbound and outbound records, RFID inventory data, and transportation data in transit, it locates the cause of the discrepancy (such as missed scans, incorrect scans, and losses) and generates discrepancy handling suggestions.

[0067] When the capacity utilization rate of a certain reserve warehouse or forward location reaches a warning threshold (e.g., 85%), the intelligent material management agent automatically initiates the capacity allocation process. Based on information such as the current inventory structure, material turnover rate, shelf-life distribution, and warehousing plan, the agent generates a capacity optimization plan, including: suggesting the transfer of some low-frequency materials to nearby reserve warehouses with ample capacity; suggesting accelerating the turnover and release of materials nearing their expiration date; and suggesting the use of temporary storage space or leased warehouses. When both capacity shortages and a surge in material demand occur simultaneously, multiple intelligent material management agents reach a capacity-sharing agreement through a negotiation mechanism, achieving cross-regional capacity coordination and allocation.

[0068] The intelligent transportation scheduling agent can perform transportation matching and anchoring operations based on current material allocation needs and available transportation resources. This agent calculates the optimal vehicle scheduling plan based on multiple dimensions of factors, including material type, quantity, urgency, destination location, road conditions, vehicle load capacity, and current location. For urgent material needs, the agent prioritizes anchoring to the nearest vehicle with matching load capacity and available status, and generates the optimal transportation route. For large-scale material allocation, the agent adopts a vehicle platooning mode, uniformly scheduling multiple vehicles for collaborative transportation to reduce empty load rates and transportation costs. In road network disruption scenarios, the intelligent transportation scheduling agent receives real-time road condition updates, dynamically adjusts transportation routes, and re-anchors alternative transportation resources.

[0069] When the inventory of a single reserve warehouse cannot meet the material needs within its jurisdiction, the cross-regional joint guarantee intelligent agent activates the cross-warehouse linkage mechanism. This intelligent agent can retrieve the available inventory and storage capacity status of reserve warehouses in adjacent areas based on comprehensive material inventory data, and initiate material transfer requests. During the cross-warehouse linkage process, multiple intelligent agents, including those responsible for material warehouse management (transferring out and in), transportation capacity scheduling, and demand verification, engage in autonomous negotiation. When consensus is reached on factors such as the type and quantity of transferred materials, handover time, transportation method, and cost settlement, a cross-warehouse linkage execution plan is generated.

[0070] This invention obtains material coordination simulation data through the autonomous interaction simulation of the above-mentioned multi-agent system, including the behavior sequence, decision logic, interaction records and resource allocation results of each agent.

[0071] Based on the above embodiments, the step of performing parallel simulation iterations under preset differentiated boundary conditions according to the material collaborative simulation data, and constructing a simulation evaluation dataset based on the end-to-end material control indicator data during the simulation iteration process, includes: The preset differentiated boundary conditions are determined based on the differentiated disaster intensity, the timing of material shortages, and road network constraints. Based on the preset differentiated boundary conditions, multiple rounds of material closed-loop simulation are executed in parallel. During the multi-round closed-loop simulation of materials, the simulation evaluation dataset is obtained based on the collected data on material arrival timeliness, warehouse capacity utilization rate, transportation capacity empty rate, and end-point shortage duration.

[0072] In this invention, based on the material collaborative simulation data generated by multi-agent inference, a large-scale parallel simulation iteration is performed under preset differentiated boundary conditions to construct a simulation evaluation dataset covering the entire material control indicators.

[0073] Specifically, a multi-dimensional parameterized boundary condition space is constructed based on three dimensions: differentiated disaster intensity, timing of material shortages, and road network constraints. In the disaster intensity dimension, multiple levels of disaster intensity parameters can be set. For example, for typhoon disasters, three levels are set: Level 3 typhoon (wind speed 10.8-17.1 m / s), Level 2 typhoon (wind speed 17.2-24.4 m / s), and Level 1 super typhoon (wind speed ≥ 24.5 m / s). Each level corresponds to different sub-parameters such as the scope of impact, degree of damage, and duration.

[0074] In terms of the timing of material shortages, different material demand surge curves can be set, including linear growth (demand grows at a constant rate over time), exponential growth (demand grows explosively in a short period of time), and pulse (demand peaks at a specific point in time), and different peak arrival times can be set (such as reaching the peak demand 6 hours, 12 hours, or 24 hours after the disaster).

[0075] In terms of road network constraints, different road network blocking schemes can be set, including single-point blocking (interruption of a single critical road segment), multi-point blocking (interruption of multiple road segments at the same time), and regional isolation (complete disconnection of a certain area from external traffic). The duration of the blocking (such as 2 hours, 6 hours, 12 hours, 24 hours) and the recovery sequence (such as simultaneous recovery, segment-by-segment recovery, priority recovery of main lines, etc.) can be set.

[0076] In this invention, orthogonal experimental design and other methods can be used to extract representative combinations of boundary conditions from the aforementioned multidimensional parameter space to form a preset set of differentiated boundary conditions. For example, a set of simulation tasks containing 100 different boundary conditions can be generated, with each set of boundary conditions corresponding to a unique combination scenario of disaster, resources, and road network.

[0077] Next, based on a pre-defined set of differentiated boundary conditions, multiple rounds of material closed-loop simulation are launched in parallel using a distributed computing framework. Each simulation round constructs an independent virtual simulation environment instance, loads the corresponding boundary condition parameters, deploys a pre-trained multi-agent cluster, and executes the complete material closed-loop process from demand acceptance, inventory verification, outbound approval, transportation scheduling, in-transit transportation, and last-mile delivery to demand fulfillment. Since each simulation instance does not interfere with each other and runs independently, the parallel processing capability of the computing cluster is fully utilized to achieve a simulation iteration scale of tens of thousands or more.

[0078] During parallel simulation, the running status of each simulation instance is monitored in real time, and issues such as abnormal interruptions, insufficient resources, and data overflows are automatically handled to ensure the integrity and reliability of the simulation task. In the multi-round closed-loop material simulation, automated indicator collection is performed for each simulation instance, recording key performance indicators across the entire material support chain.

[0079] Specifically, the time interval from the submission of a material demand request to the actual delivery of the materials to their destination is recorded, including the breakdown of timeframes such as demand processing time, warehousing preparation time, loading time, transit time, and unloading and delivery time. Statistical indicators such as average arrival time, maximum arrival time, and timeliness achievement rate (e.g., 24-hour arrival rate, 48-hour arrival rate) are compiled for different regions, different material categories, and different disaster levels.

[0080] Record the temporal changes in storage capacity utilization rates of each reserve warehouse and forward location during the simulation process, including peak utilization rate, average utilization rate, utilization rate fluctuation amplitude, and over-limit duration. By analyzing the utilization rate distribution characteristics under different storage capacity allocation strategies, identify storage capacity bottleneck nodes and idle resource nodes.

[0081] Record the load changes of each transport vehicle during the simulation process, and statistically analyze indicators such as empty mileage, empty time, and empty load rate (the proportion of empty mileage to total mileage). By analyzing the impact of different capacity grouping schemes and route planning schemes on the empty load rate, identify the waste of transport capacity resources.

[0082] Record the duration of material shortages in each disaster-stricken area and each forward location during the simulation process, including the start time of the shortage, the end time of the shortage, the type of material in short supply, the quantity of material in short supply, and the population affected by the shortage; and statistically analyze indicators such as the total duration of shortages, the maximum duration of a single shortage, and the shortage coverage rate under different material support plans.

[0083] This invention summarizes, cleans, and structures the data of the above-mentioned end-to-end material control indicators collected from all simulation instances to form a simulation evaluation dataset. This dataset uses simulation instance numbers, boundary condition parameters, and strategy configuration parameters as indexes, linking corresponding indicator observations to form a high-dimensional structured data table, providing a data foundation for large-scale model quantitative evaluation and strategy optimization. For example, a record in the simulation evaluation dataset can be represented as: {Simulation ID: 001, Typhoon Level: Level 2, Peak Demand Time: 12h, Road Network Blocking Scheme: Multi-point Blocking, Pre-positioning Scheme: Scheme A, Transport Capacity Grouping Scheme: Scheme X, Average Arrival Time: 18.5h, Peak Warehouse Capacity Utilization Rate: 92%, Average Empty Load Rate: 15%, Last-mile Shortage Duration: 4.2h}.

[0084] Based on the above embodiments, the step of quantitatively evaluating the simulation evaluation dataset, generating a material support strategy, and outputting risk warning information and pre-emptive support plans according to the material support strategy includes: Based on the large model, a multidimensional weighted quantitative evaluation is performed on the simulation evaluation dataset to output the material support strategy. The multidimensional weighted quantitative evaluation includes comparing the adaptability, anti-disturbance capability, timeliness capability, and cost control capability of different material pre-positioning, transportation capacity grouping and batch replenishment schemes. Based on the supply and demand pressure situation corresponding to the material support strategy, the system generates tiered early warning information for material supply and demand imbalance, transportation capacity gap, and supply disruption across the entire region, and outputs a support plan including a warehouse capacity optimization list, a pre-positioned material quota table, and a transportation capacity grouping plan.

[0085] In this invention, the large model first reads the simulation evaluation dataset, performs feature engineering on each record in the dataset, and extracts structured feature vectors such as boundary condition features (e.g., disaster level, demand time series pattern and road network blockage degree), strategy configuration features (e.g., pre-layout scheme coding, capacity grouping scheme coding and batch replenishment scheme coding) and performance index features (e.g., arrival timeliness, warehouse capacity utilization rate, empty load rate and stockout duration).

[0086] Furthermore, the large model performs a multi-dimensional weighted quantitative evaluation. In this invention, a multi-dimensional evaluation weight system is preset, including adaptability weight, disturbance resistance weight, timeliness weight, and cost control weight. The weights of each dimension can be dynamically adjusted based on historical best-case scenarios. For example, in the emergency preparation phase before a typhoon, the timeliness weight can be set to the highest (e.g., 40%), followed by the disturbance resistance weight (e.g., 30%), then the adaptability weight (e.g., 20%), and the cost control weight is relatively low (e.g., 10%). In the normalized reserve optimization phase, the cost control weight can be increased to the highest (e.g., 35%), followed by the adaptability weight (e.g., 25%), with disturbance resistance and timeliness each weighing 20%.

[0087] In terms of adaptability assessment, the large model analyzes the degree of matching between different pre-positioning plans for supplies and various disaster scenarios. For example, for typhoon disasters, it assesses whether the pre-positioning point layout covers all potentially affected areas, whether it considers the impact of terrain on transportation, and whether it takes into account the different needs of urban and rural areas. The higher the adaptability score, the more effectively the plan can respond to supply needs in more scenarios.

[0088] In terms of disturbance resilience assessment, large-scale model analysis examines the stability of different schemes when faced with boundary condition disturbances. This invention uses sensitivity analysis to calculate the changes in scheme performance indicators when key parameters such as disaster intensity, demand timing, and road network status fluctuate to a certain extent. A higher disturbance resilience score indicates that the scheme can maintain stable protection effectiveness even with increased uncertainty. For example, if one scheme only increases delivery time by 10% when the typhoon level increases from level two to level one, while another scheme increases it by 50%, then the former's disturbance resilience is significantly better than the latter.

[0089] In terms of timeliness assessment, the large model comprehensively evaluates indicators such as the timeliness of material delivery, response speed, and end-point coverage speed. In this invention, not only is the average timeliness considered, but also the fairness of timeliness distribution, that is, ensuring that all regions and all groups can obtain material support within a reasonable time limit, avoiding the imbalance where some areas have excellent timeliness while others suffer from long-term shortages.

[0090] In terms of cost control capability assessment, the large-scale model comprehensively evaluates all cost elements across the entire cost chain, including material reserve costs, warehousing operation costs, transportation costs, labor costs, and opportunity costs. This invention can identify optimizable aspects of the cost structure through cost-benefit analysis, such as reducing emergency procurement costs through optimized pre-positioning, lowering empty-load fuel costs through optimized transport capacity grouping, and reducing inventory backlog losses through optimized batch replenishment.

[0091] Based on the aforementioned multi-dimensional weighted quantitative evaluation results, the large-scale model comprehensively ranks all schemes in the simulation evaluation dataset and selects the scheme with the highest comprehensive score as the optimal material support strategy. Along with the optimal strategy, the large-scale model also outputs a set of suboptimal strategies and strategy switching conditions, providing alternative solutions and decision-making references. For example, when the optimal strategy performs well in terms of timeliness but has a slightly higher cost, the suboptimal strategy is recommended as an alternative in cost-sensitive scenarios.

[0092] After generating a material supply strategy, this invention automatically generates tiered early warning information based on the supply and demand pressure situation corresponding to the strategy and a preset risk threshold system. The risk threshold system includes: a material supply and demand imbalance early warning threshold (triggered when the predicted demand for a certain type of material in a certain region exceeds a certain percentage of available inventory), a transportation capacity gap early warning threshold (triggered when the transportation capacity demand in a certain region exceeds a certain percentage of available transportation capacity during a certain period), and a region-wide supply disruption early warning threshold (triggered when multiple regions experience severe material shortages simultaneously and inter-regional allocation cannot compensate for them).

[0093] In one embodiment, the early warning information can be divided into three levels: blue warning (general risk, indicating that there may be a slight shortage of materials in some areas, and it is recommended to pay attention and prepare to start allocation), yellow warning (significant risk, indicating that there will be a significant shortage of materials or a shortage of transportation capacity in some areas, and it is recommended to start emergency preparation and optimize the forward deployment), orange warning (major risk, indicating that a large area will face a serious imbalance between supply and demand of materials or a shortage of transportation capacity, and it is recommended to immediately start cross-regional linkage and emergency procurement), and red warning (particularly serious risk, indicating that there may be a complete supply disruption or a systemic failure of material security, and it is recommended to start the highest level of emergency response and request support from higher authorities).

[0094] While outputting early warning information, the system can also output proactive support plans, including: a warehouse capacity optimization list, which details suggested inventory adjustment plans for each reserve warehouse and forward location, including suggested categories and quantities of materials to be added, suggested destinations for material transfers, and a list of near-expiry materials to be cleared; a forward material quota table, which details the quota information such as the categories, quantities, batch requirements, and arrival deadlines of materials to be pre-positioned at each forward location; and a transportation capacity formation plan, which details the vehicle configuration, driver arrangements, departure time, estimated arrival time, and return trip plan for each transportation route. In this invention, the generated plans are simultaneously output in structured data formats (such as JSON and XML) and visual report forms, supporting one-click push to business terminals such as material management dashboards, warehouse control terminals, and transportation capacity scheduling platforms.

[0095] In one embodiment, taking the provision of supplies before a typhoon as an example, the specific implementation process is as follows: Step 1: Comprehensive Material Situation Data Collection and Preprocessing. The system automatically retrieves data from all reserve warehouses and forward deployment points within the jurisdiction, including the quantity of materials in stock, batch expiration dates, storage capacity, contracted transportation capacity, road network access status, and typhoon path predictions. Abnormal data is automatically removed, missing fields are filled in, time-series data is aligned, and the integrated data base for comprehensive material resources is updated.

[0096] Step 2: Generation of Disaster and Material Coupling Scenarios. The large model, combined with typhoon path prediction data, automatically generates three types of coupled material pressure scenarios for Category 3 typhoons, Category 2 typhoons, and Category 1 super typhoons. Simultaneously, it simulates real-world conditions such as road blockages in some mountainous areas, limited storage capacity at some township forward points, and insufficient transportation capacity, fully recreating the complex constraints on material supply during typhoon passage.

[0097] Step 3: Multi-agent interactive simulation. After the intelligent agents responsible for material warehousing, transportation scheduling, and cross-regional joint guarantee have completed their fine-tuning and training, they enter the virtual scenario and autonomously carry out collaborative operations throughout the entire process, including data verification, warehouse capacity allocation, emergency transportation, material diversion, and replenishment of shortages, replicating the collaborative logic of manual material management.

[0098] Step 4, Simulation and Indicator Collection. Batch set different boundary conditions such as typhoon intensity, period of material shortage, road network blockage range, and linkage response time, and conduct multiple differentiated simulations in parallel. The core control indicators such as material arrival time, peak warehouse capacity occupancy, transportation capacity empty rate, and end-point shortage duration are automatically recorded throughout the process to form a simulation material big data set.

[0099] Step 5, Strategy Iteration and Optimization. The large model performs multi-dimensional weighted quantitative evaluation on all simulated material operation data, compares the anti-disturbance capability, timeliness capability, and cost control capability of different material pre-positioning, transport capacity grouping, and batch replenishment schemes, and iteratively outputs the optimal practical strategy.

[0100] Step 6: Early Warning Issuance and Plan Implementation. Based on the simulated and predicted pressure on material supply and demand, an orange alert for material transportation capacity shortage is issued. Simultaneously, the optimized material layout, transportation grouping, and deployment plan are pushed to the material management dashboard, warehouse control terminal, and transportation dispatch platform with one click. Management personnel confirm and verify the information and carry out preliminary preparations.

[0101] The simulation risk simulation and early warning method provided by this invention can generate extreme material pressure scenarios such as multiple disasters overlapping, road network blockage, and warehouse over-limitation. It comprehensively tests the extreme supply guarantee capabilities of reserve warehouses, forward deployment points, and contracted transportation capacity, solving the shortcomings of traditional exercises that can only cover conventional scenarios and cannot test extreme shortages. At the same time, multiple intelligent agents autonomously complete the full-domain material simulation without the need for manual collaborative reconciliation and route trial calculation, improving simulation efficiency by several times and avoiding material scheduling errors caused by manual ledger errors, thus significantly reducing the labor costs of material management. Furthermore, through big data simulation and quantitative comparison, it determines the forward deployment points, reserve volume, outbound batches, and transportation capacity deployment of materials, improving material utilization, significantly reducing the risk of stockouts, and lowering the overall guarantee cost.

[0102] The simulation risk simulation and early warning system provided by the present invention is described below. The simulation risk simulation and early warning system described below can be referred to in correspondence with the simulation risk simulation and early warning method described above.

[0103] Figure 2 This is a schematic diagram of the structure of the simulation risk simulation and early warning system provided by the present invention, as shown below. Figure 2 As shown, this invention provides a simulation risk simulation and early warning system, including a data acquisition and processing module 201, a coupled dynamic scenario generation module 202, an autonomous interactive reasoning module 203, a simulation iteration module 204, and a simulation and early warning result output module 205. The data acquisition and processing module 201 acquires multi-dimensional situational data corresponding to emergency supplies and performs standardized preprocessing on the multi-dimensional situational data to obtain basic supply data. The coupled dynamic scenario generation module 202 inputs the basic supply data and disaster monitoring data into a large model to obtain a coupled disaster evolution and supply demand output by the large model. The system includes a virtual simulation scenario; an autonomous interactive reasoning module 203, which deploys a multi-agent cluster to perform autonomous interactive inference based on the virtual simulation scenario, and obtains material collaborative inference data; a simulation iteration module 204, which performs parallel simulation iteration under preset differentiated boundary conditions based on the material collaborative inference data, and constructs a simulation evaluation dataset based on the full-link material control index data during the simulation iteration process; and an inference early warning result output module 205, which performs quantitative evaluation on the simulation evaluation dataset, generates a material support strategy, and outputs risk early warning information and pre-emptive support plan based on the material support strategy.

[0104] The simulation risk simulation and early warning system provided by this invention dynamically generates virtual simulation scenarios that couple disaster evolution and material demand through a large model. Based on the autonomous interactive simulation of a multi-agent cluster, it replaces manual collaborative reconciliation and route trial calculation, greatly improving the efficiency and depth of the simulation. Furthermore, through parallel simulation iteration and multi-dimensional weighted quantitative evaluation under differentiated boundary conditions, it outputs the optimal material support strategy, realizes proactive pre-disaster defense for emergency material support, and ensures the accurate scheduling of emergency materials.

[0105] The system provided in this embodiment of the invention is used to execute the above-described method embodiments. For specific processes and details, please refer to the above embodiments, which will not be repeated here.

[0106] Figure 3 This is a schematic diagram of the structure of the electronic device provided by the present invention, such as... Figure 3As shown, the electronic device may include: a processor 301, a communications interface 302, a memory 303, and a communication bus 304, wherein the processor 301, the communications interface 302, and the memory 303 communicate with each other through the communication bus 304. The processor 301 can call logic instructions in the memory 303 to execute a simulation risk simulation and early warning method. This method includes: acquiring multi-dimensional situational data corresponding to emergency supplies, and performing standardized preprocessing on the multi-dimensional situational data to obtain basic supply data; inputting the basic supply data and disaster monitoring data into a large model to obtain a virtual simulation scenario output by the large model that couples disaster evolution with supply demand; deploying a multi-agent cluster to perform autonomous interactive simulation based on the virtual simulation scenario to obtain collaborative supply simulation data; performing parallel simulation iteration under preset differentiated boundary conditions based on the collaborative supply simulation data, and constructing a simulation evaluation dataset based on the full-link supply control indicator data during the simulation iteration process; performing a quantitative evaluation on the simulation evaluation dataset to generate a supply support strategy, and outputting risk warning information and pre-emptive support plans based on the supply support strategy.

[0107] Furthermore, the logical instructions in the aforementioned memory 303 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, essentially, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0108] On the other hand, the present invention also provides a computer program product, which includes a computer program stored on a non-transitory computer-readable storage medium. The computer program includes program instructions, and when the program instructions are executed by the computer, the computer can execute the simulation risk simulation and early warning method provided by the above methods. The method includes: acquiring multi-dimensional situational data corresponding to emergency supplies, and performing standardized preprocessing on the multi-dimensional situational data to obtain basic material data; inputting the basic material data and disaster monitoring data into a large model to obtain a virtual simulation scenario output by the large model that couples disaster evolution with material demand; deploying a multi-agent cluster to perform autonomous interactive simulation based on the virtual simulation scenario to obtain material collaborative simulation data; performing parallel simulation iteration under preset differentiated boundary conditions based on the material collaborative simulation data, and constructing a simulation evaluation dataset based on the full-link material control indicator data during the simulation iteration process; performing quantitative evaluation on the simulation evaluation dataset to generate a material support strategy, and outputting risk early warning information and pre-emptive support plan based on the material support strategy.

[0109] In another aspect, the present invention also provides a non-transitory computer-readable storage medium storing a computer program thereon. When executed by a processor, the computer program implements the simulation risk simulation and early warning method provided in the above embodiments. The method includes: acquiring multi-dimensional situational data corresponding to emergency supplies, and performing standardized preprocessing on the multi-dimensional situational data to obtain basic material data; inputting the basic material data and disaster monitoring data into a large model to obtain a virtual simulation scenario output by the large model that couples disaster evolution with material demand; deploying a multi-agent cluster to perform autonomous interactive simulation based on the virtual simulation scenario to obtain material collaborative simulation data; performing parallel simulation iteration under preset differentiated boundary conditions based on the material collaborative simulation data, and constructing a simulation evaluation dataset based on the full-link material control indicator data during the simulation iteration process; performing quantitative evaluation on the simulation evaluation dataset to generate a material support strategy, and outputting risk early warning information and pre-emptive support plan based on the material support strategy.

[0110] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0111] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0112] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A simulation-based risk prediction and early warning method, characterized in that, include: Acquire multi-dimensional situational data corresponding to emergency supplies, and perform standardized preprocessing on the multi-dimensional situational data to obtain basic data on the quantity of supplies; The material baseline data and disaster monitoring data are input into the large model to obtain a virtual simulation scenario that couples the disaster evolution and material demand, output by the large model. Based on the virtual simulation scenario, a multi-agent cluster is deployed to perform autonomous interactive simulations to obtain material collaborative simulation data. Based on the material collaborative simulation data, parallel simulation iterations are performed under preset differentiated boundary conditions, and a simulation evaluation dataset is constructed based on the full-link material control index data during the simulation iteration process. The simulation evaluation dataset is quantitatively evaluated to generate a material support strategy, and based on the material support strategy, risk warning information and pre-emptive support plan are output.

2. The simulation risk simulation and early warning method according to claim 1, characterized in that, The acquisition of multi-dimensional situational data corresponding to emergency supplies includes: Based on emergency material reserve warehouses, RFID inventory terminals, transportation capacity dispatching platforms, GPS positioning terminals, and meteorological monitoring stations, information on the types of materials in storage, available storage capacity, batches of materials, transportation capacity, road network capacity, and disaster monitoring indicators can be obtained. The standardization preprocessing of the multidimensional situational data includes: The multidimensional situational data is cleaned, anomaly removed, standardized, coded, and time-series aligned to obtain the material baseline data.

3. The simulation risk simulation and early warning method according to claim 1, characterized in that, The process of inputting the basic material data and disaster monitoring data into a large model to obtain a virtual simulation scenario output by the large model that couples disaster evolution with material demand includes: Based on the large model, the spatiotemporal coupling relationship between disaster evolution and material demand is analyzed. Based on the spatiotemporal coupling relationship, scenarios such as conventional material supply guarantee, multi-disaster superposition pressure scenario, road network blockage supply scenario, and warehouse overload diversion scenario are generated. The peak material demand, warehouse capacity crowding nodes, and transportation congestion nodes under different scenarios are determined.

4. The simulation risk simulation and early warning method according to claim 1, characterized in that, The multi-agent cluster is a multi-department, multi-management role material control intelligent agent, and the multi-department, multi-management role material control intelligent agent is embedded with a material business-specific cognitive reasoning engine and reinforcement learning algorithm. Based on the virtual simulation scenario, a multi-agent cluster is deployed to perform autonomous interactive simulations to obtain material collaborative simulation data, including: The multi-department, multi-management role intelligent agent for material control is deployed in the virtual simulation scenario. Based on historical scheduling cases and warehousing specifications, it executes adaptive behaviors corresponding to the autonomous interactive simulation to obtain the material collaborative simulation data. The adaptive behaviors include at least manual reconciliation, warehouse capacity allocation, transportation capacity anchoring, and cross-warehouse linkage behaviors.

5. The simulation risk simulation and early warning method according to claim 1, characterized in that, The process involves performing parallel simulation iterations based on the material collaborative simulation data, under preset differentiated boundary conditions, and constructing a simulation evaluation dataset based on the end-to-end material control indicator data during the simulation iteration process, including: The preset differentiated boundary conditions are determined based on the differentiated disaster intensity, the timing of material shortages, and road network constraints. Based on the preset differentiated boundary conditions, multiple rounds of material closed-loop simulation are executed in parallel. During the multi-round closed-loop simulation of materials, the simulation evaluation dataset is obtained based on the collected data on material arrival timeliness, warehouse capacity utilization rate, transportation capacity empty rate, and end-point shortage duration.

6. The simulation risk simulation and early warning method according to claim 1, characterized in that, The process of quantitatively evaluating the simulation evaluation dataset, generating a material support strategy, and outputting risk warning information and pre-emptive support plans based on the material support strategy includes: Based on the large model, a multidimensional weighted quantitative evaluation is performed on the simulation evaluation dataset to output the material support strategy. The multidimensional weighted quantitative evaluation includes comparing the adaptability, anti-disturbance capability, timeliness capability, and cost control capability of different material pre-positioning, transportation capacity grouping and batch replenishment schemes. Based on the supply and demand pressure situation corresponding to the material support strategy, the system generates tiered early warning information for material supply and demand imbalance, transportation capacity gap, and supply disruption across the entire region, and outputs a support plan including a warehouse capacity optimization list, a pre-positioned material quota table, and a transportation capacity grouping plan.

7. A simulation risk deduction and early warning system, characterized in that, include: The data acquisition and processing module is used to acquire multi-dimensional situational data corresponding to emergency supplies, and to perform standardized preprocessing on the multi-dimensional situational data to obtain the basic data of the supplies. The coupled dynamic scene generation module is used to input the material base data and disaster monitoring data into the large model to obtain a virtual simulation scene that couples the disaster evolution and material demand output by the large model. The autonomous interactive reasoning module is used to deploy a multi-agent cluster to perform autonomous interactive reasoning based on the virtual simulation scenario, and obtain material collaborative reasoning data; The simulation iteration module is used to perform parallel simulation iteration under preset differentiated boundary conditions based on the material collaborative simulation data, and to construct a simulation evaluation dataset based on the full-link material control index data during the simulation iteration process. The simulation and early warning result output module is used to quantitatively evaluate the simulation evaluation dataset, generate material support strategies, and output risk warning information and pre-emptive support plans based on the material support strategies.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the simulation risk simulation and early warning method as described in any one of claims 1 to 6.

9. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the simulation risk simulation and early warning method as described in any one of claims 1 to 6.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the simulation risk simulation and early warning method as described in any one of claims 1 to 6.

Images