Multi-target dynamic evolution-oriented intelligent decision support system and method for emergency management

Abstract

The application discloses an emergency management intelligent decision support system and method for multi-target dynamic evolution, and relates to the technical field of emergency management, wherein the method comprises the following steps: automatically collecting global multi-source heterogeneous emergency data, forming standardized basic data through cleaning and verification, quality control and periodic update management; completing entity recognition, relationship extraction and graph structure optimization according to the data, expanding knowledge links combined with field implicit association reasoning, and constructing an emergency management field knowledge graph; building an intelligent analysis engine by integrating emergency quantitative modeling, subject digital twinning, multi-agent simulation and large model reasoning technology, and outputting decision deduction results; building a layered decision module and a visual interactive cockpit to realize hierarchical scheme output, process backtracking, parameter adjustment and result visualization display. Through the integrated fusion of multi-field intelligent technology, the application realizes the cross-scene migration and reuse of the technical system, and solves the bottleneck of the limitation of single technology application and the difficulty of technology fusion landing.

Description

Technical Field

[0001] This invention relates to the field of emergency management technology, and in particular to an intelligent decision support system and method for emergency management oriented towards multi-objective dynamic evolution. Background Technology

[0002] With the continued intensification of global climate change and the deepening of urbanization in China, various emergencies such as natural disasters, accidents, public health emergencies, and social security incidents are occurring frequently. These emergencies are characterized by their suddenness, complex evolution, high correlation, and wide-ranging damage, constantly threatening the safety of people's lives and property, the stable operation of society, and the national public security system. The evolution of emergencies themselves exhibits significant nonlinearity and high uncertainty. The development of various situations, such as extreme weather, production safety accidents, and public health emergencies, is difficult to predict. This places stringent demands on the rapid response, scientific judgment, and efficient handling of the entire emergency management process. It also aligns with the operational needs of the entire chain of emergency management—prevention, resistance, rescue, treatment, and recovery—and conforms to the domestic emergency management operation principles of unified command, hierarchical responsibility, and local jurisdiction.

[0003] From the perspective of the current state of emergency management in China, the total number of various emergencies remains high, covering a wide range of disaster types including floods, typhoons, geological disasters, fires, hazardous chemical leaks, and public health emergencies, causing significant casualties and economic losses every year. Currently, China's emergency management system is in a critical phase of transition from passive, post-event response to proactive, pre-emptive prevention and precise, end-to-end response. Traditional emergency decision-making models, relying on experience-based extrapolation and linear analysis, are no longer adequate for complex emergency scenarios involving strong coupling of multiple factors and dynamic evolution. Traditional decision-making models generally suffer from decision-making lag, failing to keep pace with real-time developments and adjust response strategies accordingly, easily missing the golden window for rescue and response, thus amplifying various losses caused by disasters. Furthermore, there are blind spots in overall understanding, failing to comprehensively analyze the causes and development paths of emergencies, as well as the inherent relationships between multiple factors such as meteorological environment, resource reserves, transportation networks, and population distribution. Decision-making is based on limited information and lacks foresight, with fragmented overall emergency data and incomplete situational assessments being particularly prominent pain points in the industry.

[0004] Currently, China has established a relatively complete emergency organization system with sufficient emergency mobilization capabilities and mature on-site handling experience. However, it still faces multiple industry challenges in actual management and operation. Firstly, emergency decision-making is complex and subject to numerous constraints, covering the entire process from risk monitoring and early warning issuance to resource allocation, personnel evacuation, on-site rescue, and post-disaster recovery. Simultaneously, the decision-making process needs to balance multiple mutually constraining objectives such as rapid response, loss reduction, resource utilization, and operational safety. Deviations in a single decision can easily be amplified through systemic linkages, triggering chain risks and significantly increasing the overall difficulty of decision-making. Secondly, the evolution of emergencies is highly random, with frequent secondary disasters and rapid escalation, demanding extremely high capabilities for dynamic adjustment and real-time optimization of decisions, making it difficult to quickly achieve equilibrium among conflicting decision objectives. Finally, the precision of emergency resource allocation is insufficient. Under the dual constraints of time and resources, a crude allocation model easily leads to problems such as resource idleness, resource shortages, and misallocation, resulting in low resource utilization efficiency. How to achieve precise and efficient allocation of limited emergency resources has become a pressing practical problem that emergency management work urgently needs to solve.

[0005] Existing emergency decision support technologies have significant limitations in areas such as comprehensive intelligence integration, situational pattern analysis, and future scenario projection. Various emergency data sources are independent and fragmented, encompassing heterogeneous data from meteorological monitoring, geological data, material reserves, medical resources, traffic conditions, and online public opinion. This data is largely unstructured and fragmented, resulting in widespread information silos within the industry. Decision-makers struggle to form a complete and dynamic understanding of the overall situation, and decisions lack comprehensive and reliable data support. Furthermore, traditional analytical methods heavily rely on expert subjective experience and qualitative judgment, lacking standardized quantitative analysis systems. This makes it impossible to scientifically quantify the actual impact of resource input, early warning measures, and response plans on emergency effectiveness. Plan evaluations are highly subjective and have limited reference value, failing to support the iterative optimization of emergency management mechanisms and response strategies.

[0006] In addition, traditional emergency simulation models generally adopt linear evolutionary logic, simply treating various emergency response entities as command execution units, which differs significantly from real-world multi-entity collaborative emergency scenarios. In real-world emergency scenarios, management departments, rescue teams, medical institutions, grassroots communities, and related enterprises possess independent decision-making and dynamic response capabilities, with complex interactive and collaborative relationships among these entities, and significant conflicts between multiple decision-making objectives. Existing models cannot accurately simulate the complex evolutionary paths of emergencies, the effects of multi-entity collaboration, resource allocation bottlenecks, and secondary disaster risks. They lack systematic multi-objective equilibrium optimization capabilities, cannot adapt to complex and dynamic emergency response needs, and are insufficient to meet the current trend of emergency management moving towards intelligence, precision, and comprehensiveness. Overall, the current emergency decision support system has not yet formed a complete integrated technical solution. Significant technical shortcomings exist in information integration, causal analysis, trend prediction, and multi-objective coordination, necessitating a more comprehensive intelligent decision-making system to provide stable and reliable technical support for the entire process of emergency management. Summary of the Invention

[0007] The main objective of this invention is to provide an intelligent decision support method for emergency management that is oriented towards multi-objective dynamic evolution.

[0008] Another objective of this invention is to propose an intelligent decision support system for emergency management that is oriented towards multi-objective dynamic evolution.

[0009] The third objective of this invention is to provide an electronic device.

[0010] A fourth objective of this invention is to provide a non-transitory computer-readable storage medium.

[0011] To achieve the above objectives, a first aspect of the present invention proposes an intelligent decision support method for emergency management oriented towards multi-objective dynamic evolution, comprising:

[0012] By automatically collecting heterogeneous data related to emergency management across the entire domain through multiple sources, and completing data cleaning, verification, quality assessment, and periodic updating and governance, standardized basic emergency data resources are formed. Based on the aforementioned basic emergency data resources, entity recognition, relation extraction, and graph structure optimization are performed. Knowledge nodes and related links are expanded by combining domain implicit association reasoning to construct a knowledge graph for the emergency management domain. Based on the aforementioned emergency management knowledge graph, a comprehensive intelligent analysis engine is constructed by integrating emergency quantitative modeling, digital twins of response entities, multi-agent collaborative simulation, and large-scale model domain reasoning technology, and outputs simulation inference and decision-making results. A hierarchical emergency decision-making function module and a visual interactive decision-making cockpit are constructed, which are connected to the simulation and decision-making results to realize hierarchical solution output, process retrospective review, flexible parameter adjustment and intuitive display of analysis results, and complete the whole process of intelligent emergency decision-making.

[0013] Optionally, heterogeneous data related to emergency management across the entire domain can be automatically collected through multiple sources, and data cleaning, verification, quality assessment, and periodic updates can be performed to form standardized basic emergency data resources, including: It adopts automated collection methods such as web crawling, standardized data interfaces, and API calls to collect emergency open source intelligence data covering five dimensions: risks and hazards, emergency resources, emergencies, policies and regulations, and public opinion dynamics. Connect with official data interfaces of meteorological, geological and emergency management authorities to obtain real-time early warning and basic thematic data; collect event handling dynamic information from official platforms and news media through web crawlers; and capture public attention hotspots and emergency demands information with the help of public opinion collection tools. Establish a regular update mechanism with hourly, daily, or weekly updates, conduct real-time verification and deduplication preprocessing of collected data, and build a data quality assessment system to control data integrity and accuracy through quantitative scoring, ultimately forming standardized basic emergency data resources.

[0014] Optionally, based on the aforementioned basic emergency data resources, entity identification, relation extraction, and graph structure optimization are performed. Knowledge nodes and related links are expanded by combining domain implicit association reasoning to construct a knowledge graph for the emergency management domain, including: Based on standardized basic emergency data resources, deep learning entity linking technology is used to complete the identification of unstructured and semi-structured data entities, extract core entities such as risk hazards, emergency resources, emergencies, handling entities, and policy norms, and use attention mechanism extraction models to mine the causal, matching, collaborative, and constraint relationships between entities; Entities and their corresponding relationships are stored in a triple graph data structure. Reinforcement learning algorithms are used to optimize the graph structure. The action space is entity addition and redundant relationship deletion, and the state space is graph association density and relationship confidence. The topology structure is iteratively optimized based on the weighted reward function of accuracy and association density. By integrating rule-based reasoning and Bayesian network probabilistic reasoning techniques, implicit links between entities are discovered. New implicit relationships are determined based on preset conditional probability confidence thresholds, expanding knowledge graph nodes and related networks, and completing the overall construction of the knowledge graph.

[0015] Optional emergency quantitative modeling specifically includes: Based on the emergency management knowledge graph, natural language processing technology is used to analyze emergency policies, response plans and risk indicators, extract core elements of response standards and risk thresholds, and combine multi-objective trade-off algorithms to complete the weight division of objectives such as rapid response, minimum loss, resource efficiency and safety controllability. The weight coefficients of each decision objective are solved by the analytic hierarchy process (AHP) to construct a multi-objective comprehensive evaluation model, which transforms decision parameters into quantifiable and calculable indicators. For emergency resource allocation scenarios, a quantitative calculation formula is established by combining regression analysis and causal inference methods to calculate the relationship between rescue speed, loss reduction ratio, resource utilization rate and resource reserves, response time, and allocation cost. The quantitative model is improved based on scenario correction coefficients.

[0016] Optionally, the specific digital twin of the disposal entity includes: By combining emergency management knowledge graphs, basic emergency data from the early stages, industry reports, and expert experience, a full-dimensional digital twin model is constructed for emergency management departments, fire and rescue teams, medical institutions, and material reserve centers. Simulate the organizational structure, response process, rescue capabilities, resource reserves and scheduling mechanisms of each entity within the twin model, and drive the simulation of multi-entity collaborative response behavior by combining emergency quantitative model parameters; Feedback and corrections are made based on the real-time operational data of the disposal entities, and the internal parameters of the twin model are dynamically adjusted to ensure that the model accurately maps the actual operating status and decision-making logic of each entity.

[0017] Optional, multi-agent cooperative simulation specifically includes: The main body of the entire emergency response process is abstracted into an independent simulated intelligent agent. Each intelligent agent is equipped with a multi-objective decision-making algorithm and corresponding weight parameters. Based on environmental perception information including the situation of the emergency and the remaining resources, it autonomously outputs decision-making actions including resource allocation and route adjustment. A quantitative formula for information interaction intensity among intelligent agents is constructed. The interaction intensity is calculated based on the similarity of target weights, state differences, and the distance between the main functions, so as to realize information exchange and mutual influence of behavior among multiple intelligent agents. The system dynamically simulates the entire lifecycle of disaster response and outputs multi-dimensional simulation data, including rescue efficiency, secondary disaster impact, resource utilization effectiveness, and the degree of balance among multiple objectives.

[0018] Optionally, a hierarchical emergency decision-making function module and a visual interactive decision-making dashboard can be constructed, including: To meet the decision-making needs of the three levels of high-level strategy, middle-level management and grassroots execution, three functional modules are built respectively: emergency situation simulation, emergency resource allocation optimization and emergency response risk warning. Combined with multi-objective optimization and risk index quantification algorithm, hierarchical decision output and risk level warning are realized. The overall architecture of the visual interactive decision-making cockpit is built using component-based front-end development technology. It is equipped with global data status management, multi-dimensional data visualization, unified interface style specifications, and efficient project construction and compilation. This completes the development of a highly interactive component-based system, enabling intuitive display of analysis results, retrospective review of the decision-making process, and flexible parameter adjustment.

[0019] To achieve the above objectives, a second aspect of the present invention proposes an intelligent decision support system for emergency management oriented towards multi-objective dynamic evolution, comprising: The data acquisition unit is used to automatically collect heterogeneous data related to emergency management across the entire domain through multiple sources, and to complete data cleaning, verification, quality assessment and periodic updating and management to form standardized basic emergency data resources. The graph construction unit is used to perform entity recognition, relation extraction and graph structure optimization based on the basic emergency data resources, and to expand knowledge nodes and related links by combining domain implicit association reasoning to construct an emergency management domain knowledge graph. The intelligent analysis unit is used to build a comprehensive intelligent analysis engine based on the emergency management knowledge graph, integrating emergency quantitative modeling, digital twins of the response subjects, multi-agent collaborative simulation and large model domain reasoning technology, and output simulation inference and decision-making results. The decision application unit is used to construct a hierarchical emergency decision-making function module and a visual interactive decision-making cockpit. It connects to the simulation and decision-making results, realizes hierarchical scheme output, process retrospective review, flexible parameter adjustment and intuitive display of analysis results, and completes the entire process of intelligent emergency decision-making.

[0020] Regarding the system in the above embodiments, the specific ways in which each module performs operations have been described in detail in the embodiments related to the method, and will not be elaborated here.

[0021] To achieve the above objectives, a third aspect of this application provides an electronic device, including a processor and a memory; wherein the processor reads executable program code stored in the memory to run a program corresponding to the executable program code, so as to implement an intelligent decision support method for emergency management oriented towards multi-objective dynamic evolution as described in the first aspect embodiment.

[0022] To achieve the above objectives, the fourth aspect of this application proposes a non-transitory computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements an intelligent decision support method for emergency management oriented towards multi-objective dynamic evolution as described in the first aspect embodiment.

[0023] The embodiments of the present invention have the following beneficial effects: 1. Enhance the scientific nature of emergency decision-making strategies. By relying on the emergency sandbox to achieve full-process simulation and multi-objective trade-off optimization, potential risks such as resource allocation, secondary disasters, and goal conflicts can be identified in advance, effectively reducing the trial-and-error costs of emergency decision-making, abandoning the traditional experience-based passive decision-making mode, and achieving proactive prediction, precise disposal, and multi-objective coordination, meeting the strategic development needs of the upgrade of China's emergency management system.

[0024] 2. Optimize the application experience of emergency decision-making products. Build a new generative simulation emergency decision-making support system, create a human-machine symbiotic collaborative decision-making mode, combine AI computing power analysis with manual experience value judgment, make the reasoning process of the plan transparent and traceable, lower the usage threshold for non-professional decision-makers, improve decision-making efficiency, plan adaptability, and decision-making credibility, and promote the intelligent and user-friendly implementation of emergency decision-making.

[0025] 3. Break through the bottleneck of the integrated application of domain technologies. Achieve the deep integration of complex system theory, multi-agent simulation, and large model technology, form an integrated emergency decision-making technology paradigm with strong technical implementation. It not only empowers emergency management scenarios but can also be migrated and reused in similar complex system fields such as public health and urban safety, broadening the boundaries of the integrated application of multiple technologies.

[0026] 4. Facilitate the collaborative development of industry, academia, and research. Build an open emergency simulation research platform, taking into account academic theoretical research, algorithm optimization exploration, industrial technology verification, and the needs of achievement transformation, promoting interdisciplinary communication and cooperation, reducing the innovation test risks in the industry, accelerating the iteration of emergency technologies and the implementation of achievements, and empowering the high-quality innovative development of the entire emergency industry chain. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] The above and / or additional aspects and advantages of the present invention will become apparent and easy to understand from the following description of the embodiments in conjunction with the drawings, where: Figure 1 It is a flowchart of an intelligent decision-making support method for emergency management facing multi-objective dynamic evolution provided by an embodiment of the present invention; Figure 2 It is a flowchart of emergency data processing and the construction of a domain knowledge graph provided by an embodiment of the present invention; Figure 3 It is a hierarchical architecture diagram of the core engine of emergency intelligent decision-making provided by an embodiment of the present invention; Figure 4 It is an overall architecture diagram of the emergency decision-making visualization cockpit provided by an embodiment of the present invention; Figure 5 It is a structural diagram of an intelligent decision-making support system for emergency management facing multi-objective dynamic evolution provided by an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0028] It should be noted that, unless otherwise specified, the embodiments and features described in the present invention can be combined with each other. The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0029] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0030] The following description, with reference to the accompanying drawings, describes an intelligent decision support method and system for emergency management oriented towards multi-objective dynamic evolution, according to an embodiment of the present invention.

[0031] Example 1 This invention provides an intelligent decision support method for emergency management oriented towards multi-objective dynamic evolution. Figure 1 This is a flowchart illustrating an intelligent decision support method for emergency management oriented towards multi-objective dynamic evolution, provided by an embodiment of the present invention. Figure 1 As shown, the method includes the following steps: Step S1 involves automatically collecting heterogeneous data related to emergency management across the entire domain through multiple sources, and completing data cleaning, verification, quality assessment, and periodic updates to form standardized basic emergency data resources.

[0032] This step corresponds to the following: Figure 2 The data presentation layer preprocessing stage shown in this embodiment covers three major sources: global open-source intelligence, domestic open-source intelligence, and dedicated resource databases for meteorology, geology, and emergency response. The multi-source raw intelligence streams are uniformly integrated into the system to carry out full-link data governance.

[0033] In the specific data collection phase, this application comprehensively employs various automated collection methods, including web crawlers, standardized data interface integration, and backend API calls, to collect emergency open-source intelligence data covering five core dimensions: risks and hazards, emergency resources, emergencies, policies and regulations, and public opinion dynamics. In the actual implementation of data source integration, on the one hand, it proactively connects with the official standardized data interfaces of meteorological departments, geological departments, and emergency management authorities at all levels to obtain real-time basic data on topics such as extreme weather warnings, geological disaster hazard locations, emergency material reserves, and specific emergency response plans, aligning with the existing operational mechanism for emergency government data sharing in China. On the other hand, it utilizes web crawlers to extract publicly available intelligence information such as the progress of emergency response, on-site rescue dynamics, and follow-up reports from official emergency management platforms at all levels and mainstream news media, matching the industry standards for the public release of emergency information. Simultaneously, it relies on dedicated public opinion collection tools to capture social hotspots, public demands, and public opinion trends regarding various emergencies on social media and public platforms, aligning with the core working principle of emergency management: people-centeredness and timely response to public concerns.

[0034] After the initial data collection is completed, the following preprocessing operations are carried out in sequence: data verification, data cleaning, and data deduplication. In order to ensure the usability of data throughout the entire life cycle, this application has set up a regular data update mechanism that is divided into hourly, daily, and weekly levels. Real-time verification and validation are performed on each batch of newly collected and imported data to remove invalid dirty data with content errors and logical contradictions. Then, the data format is standardized and missing fields are filled in through the cleaning process. Finally, the full data deduplication process is completed to remove duplicate and redundant data entries.

[0035] At the same time, this application establishes a complete data quality assessment system to comprehensively evaluate and control the completeness and accuracy of each batch of data through quantitative scoring, thereby controlling the quality of data output throughout the process. Ultimately, it forms standardized basic emergency data resources with unified format, standardized content, and credible standards, laying a solid foundation for the underlying data in the subsequent knowledge graph construction process.

[0036] Step S2: Based on the basic emergency data resources, entity identification, relationship extraction and graph structure optimization are performed. Knowledge nodes and related links are expanded by combining domain implicit association reasoning to construct a knowledge graph for the emergency management domain.

[0037] This step still corresponds to... Figure 2 The data shown illustrates the entire process of knowledge graph construction within the data reality layer. After the cleaned and structured data is processed in step S1, it is all input into the graph construction module for in-depth knowledge processing.

[0038] In this embodiment, deep learning-driven entity linking technology is first used to perform precise entity recognition processing on unstructured and semi-structured text data within the existing basic emergency data, fully extracting five core entity objects: risks and hazards, emergency resources, emergencies, handling entities, and policies and regulations. Following the entity recognition stage, a deep learning relationship extraction model based on an attention mechanism is used to deeply explore the diverse relationships hidden between various entities. This covers multiple domain-specific relationships, including the causal relationship between risks and hazards and emergencies, the supply and demand matching relationship between emergency resources and on-site handling needs, the collaborative relationship between various handling entities, and the constraints and limitations imposed by policy and regulations on on-site handling behavior. All of these relationships align with the actual inherent laws of emergency management operations.

[0039] After extracting entities and basic relationships, this application stores all identified entities and relationships into a graph database using a triplet graph data structure for basic storage. Simultaneously, a reinforcement learning algorithm is introduced to perform global optimization of the graph topology. The specific optimization process is implemented through a dedicated mathematical model, where the entity set of the knowledge graph is denoted as . The set of relations is The overall knowledge graph can be uniformly represented as a graph. ,in It is the set of all knowledge triples.

[0040] In this optimization process, the action space of the reinforcement learning agent... Defined as all operations for structural adjustments to a knowledge graph, including adding entity associations and deleting redundant or invalid relationships; state space The corresponding structural features of the real-time graph include specific feature parameters such as the density of entity associations and the confidence level of each relationship; the reward function is defined as follows: ,in Perform actions for the intelligent agent The newly adjusted knowledge graph The overall accuracy of the adjusted map information is quantified by the entity association matching accuracy rate. This represents the correlation density of the adjusted graph, adapting to the complex and diverse entity correlation needs in the emergency response field. The two evaluation indicators are weighted coefficients, and both satisfy the constraints. The agent maximizes cumulative rewards. Complete continuous iterative optimization, where t is the number of reinforcement learning iterations and T is the maximum total number of iterations. To enhance the learning discount factor, the range of values ​​must satisfy... This will allow for continuous optimization of the overall topology of the knowledge graph, thereby improving the accuracy and practicality of knowledge storage.

[0041] Building upon structural optimization, this application further integrates two knowledge reasoning techniques: rule-based reasoning and Bayesian network probabilistic reasoning. This allows for in-depth mining of implicit connections between entities that have not been directly extracted, thus expanding the knowledge graph content. The probabilistic reasoning step relies on Bayesian network formulas for calculation, assuming risk-prone entities. With secondary disaster entities The conditional probability is The system traverses all existing associated triples within the graph to complete the conditional probability calculation between all entities. This application, based on emergency response experience, pre-sets a confidence threshold θ, ranging from 0.7 to 0.9. When the calculated conditional probabilities satisfy… When the implicit relationship exists between two entities, the graph is then updated with new related links and corresponding nodes.

[0042] The complete emergency management knowledge graph, constructed through the above-mentioned full-process processing, is stored in a dedicated graph database and a business database, along with the unique standardized data after deduplication. This not only completes the solidification and storage of all knowledge but also provides comprehensive domain knowledge and data support for the subsequent full-module analysis and simulation work of the core engine layer.

[0043] Step S3: Based on the emergency management knowledge graph, integrate emergency quantitative modeling, digital twins of the response entities, multi-agent collaborative simulation, and large-scale model domain reasoning technology to construct a comprehensive intelligent analysis engine and output simulation and decision-making results.

[0044] This step corresponds to the following: Figure 3 The core engine layer architecture shown has all functions. The emergency management knowledge graph output by the data reality layer and the unique basic data at the bottom layer flow into the core engine in a synchronous bidirectional manner. The large language model LLM serves as the core reasoning center of the entire system. It links the three core modules of emergency quantification, digital twin of the disposal subject, and multi-agent simulation to work together. Data is exchanged and logic is interlocked between the modules, which fully realizes intelligent analysis and multi-objective trade-off judgment throughout the process.

[0045] First, emergency quantitative modeling is carried out. In this embodiment, the system develops a dedicated emergency analysis tool. Relying on natural language processing (NLP) technology and connecting with domain knowledge within the knowledge graph, it performs structured text parsing on various emergency management policy provisions, on-site disposal plans, and comprehensive risk assessment indicators. It accurately extracts core element information such as disposal execution standards, various disaster risk thresholds, and disposal process specifications. At the same time, it deeply integrates a multi-objective trade-off algorithm to complete the weight division and priority labeling of the four core decision objectives: rapid response, minimum loss, resource efficiency, and safety controllability. It transforms abstract decision objectives into quantifiable parameters and standardized mathematical operation rules, and builds a complete and general emergency quantitative model.

[0046] This application employs the Analytic Hierarchy Process (AHP) to scientifically assign weights to various objectives, assuming the objective set for emergency multi-objective decision-making is . These correspond to the four main objectives: rapid response, minimal loss, resource efficiency, and security and controllability. The weights of each objective are denoted as follows: Strictly satisfy the weight constraints and The weight calculation formula is: ,in The eigenvector components obtained by solving the corresponding target judgment matrix. To solve the problem by performing pairwise matrix operations on each feature vector component corresponding to all decision objectives, the weight allocation is scientifically rigorous.

[0047] Based on this, a multi-objective comprehensive evaluation model is constructed, whereby the quantitative evaluation value of the i-th objective is set as follows: Where x is a vector composed of all emergency decision-making parameters, the formula for the comprehensive quantitative evaluation value is: This achieves the mathematical transformation of multi-objective decision-making needs, providing a unified quantitative benchmark for subsequent full-process simulation and evaluation. For typical application scenarios of emergency resource allocation, this application combines regression analysis and causal inference algorithms to establish specific quantitative calculation relationships between three core performance indicators—rescue speed, disaster loss reduction ratio, and resource utilization rate—and resource reserves, response time, allocation radius, and allocation cost. Furthermore, it incorporates specific correction coefficients for different disaster scenarios to refine the model. The quantitative formula for rescue speed is as follows: In the formula, V represents the emergency rescue advance speed, Q represents the total amount of emergency resources, T represents the emergency response time, and R represents the spatial allocation radius of emergency resources. The correction factor for rescue speed is set, with a value range of 0.8 to 1.2; the formula for quantifying the loss reduction ratio is as follows. In the formula, L represents the percentage reduction in losses from the emergency, and C represents the total cost of emergency resource allocation. The specific weight coefficient corresponding to the objective of minimizing loss. The loss effect correction coefficient ranges from 0.75 to 1.1; the resource utilization rate quantification formula is as follows: In the formula, U represents the comprehensive utilization efficiency of emergency resources. The specific weighting coefficients corresponding to the resource efficiency goals. The resource utilization correction coefficient ranges from 0.85 to 1.05, and all parameters conform to the operational logic of real emergency rescue operations.

[0048] Following the quantitative modeling phase, the construction of digital twins for the emergency response entities was carried out. Based on the emergency management knowledge graph, previously collected basic emergency data, publicly available industry reports, domain expert experience, and real-time feedback data from the data reality layer, this application builds a full-dimensional digital twin model for all key emergency response entities, covering all core participants such as emergency management authorities, fire and rescue teams, medical institutions at all levels, and emergency material reserve centers, fully matching the real-world emergency response organizational system in China. The twin model internally simulates all core business characteristics of each entity, including organizational structure, emergency response processes, professional rescue capabilities, self-owned resource reserves, and dispatching mechanisms. Combined with unified parameters output from the emergency quantitative model, it drives the simulation of each entity's response behavior and cross-entity collaborative response mode in the face of emergencies. Simultaneously, the system continuously collects real-time operational data from each response entity for real-time feedback, dynamically iterating and adjusting all parameters within the twin model to ensure that the model accurately maps the corresponding entity's real operational status and decision-making logic, providing a high-fidelity entity foundation for subsequent multi-agent simulations.

[0049] Following this, multi-agent collaborative simulation work was carried out. This application abstracts all decision-making and execution entities involved in the entire emergency response process into independent simulation agents, and builds a high-fidelity dynamic simulation environment that covers all emergency participants to match real disaster response scenarios. Each independent agent has a built-in corresponding multi-objective decision-making algorithm and inherits the exclusive objective weight parameters output by the emergency quantification model. By perceiving information such as the emergency situation, remaining resource reserves, and the operational status of other entities in the simulation environment in real time, the agent autonomously outputs decision-making actions such as resource allocation, rescue route adjustment, and initiation of collaborative requests.

[0050] In this embodiment, a multi-objective decision-making system adapted to the business needs of each simulated agent is constructed. Let the objective set of the k-th agent (k=1,2,...,K, where K is the total number of emergency response agents in the current simulation scenario) be denoted as... In the formula, m represents the number of decision-making objectives contained in the intelligent agent itself. The number of objectives varies depending on the type of the handling entity. The weights of each objective are assigned synchronously by the emergency quantification model mentioned above, denoted as... ,satisfy and In the formula Let be the weight coefficient for the i-th decision objective corresponding to the k-th agent. The real-time perception vector of the agent to the simulation environment is . In the formula, n represents the total number of environmental indicators that a single intelligent agent can perceive. This refers to the value of the i-th environmental indicator perceived by the k-th agent. Specifically, it encompasses various types of perceived data, including the quantification value of the emergency situation, the remaining emergency resources, and the operational status of other agents. All perceived indicators have undergone standardized preprocessing, and their values ​​satisfy interval constraints. The optimal decision action ultimately output by the agent is: Specifically, this includes actions such as adjusting rescue routes, determining the amount of emergency resources allocated, and sending cross-entity collaborative requests. The optimal action is determined by a built-in decision algorithm combined with multi-objective comprehensive trade-off logic operations. The core solution formula is:

[0051] In the formula, Let be the action space consisting of all possible actions of the k-th agent, and let 'a' be any candidate action selected by the agent from the action space. Let be the goal achievement function, representing the current environmental perception state of the k-th agent. After action a is executed, the completion level of the corresponding i-th target is determined by the function, which takes values ​​in the range of 0. 1. The higher the value, the better the effect of the action on the current decision goal.

[0052] This application also constructs a dedicated quantification formula for the intensity of information interaction between intelligent agents. The formula for calculating the interaction intensity between the k-th and l-th intelligent agents is as follows:

[0053] In the formula, Let μ be the information interaction strength between agents k and l, and μ be the agent collaboration interaction coefficient, which takes a value of 0.6 to 0.9 based on the degree of business collaboration. Let i be the weight of the i-th objective of the k-th agent. Let i be the weight of the i-th objective of the l-th agent. Let be the value of the i-th environmental indicator perceived by the k-th agent. Let be the value of the i-th environmental indicator perceived by the l-th agent. The formula for the functional distance between two intelligent agents reflects the principle that the higher the similarity of target weights, the greater the difference in agent states, and the closer the functional distance, the stronger the mutual information interaction. This perfectly adapts to the real-world business logic of multi-agent collaborative handling. The simulation process dynamically and completely reproduces the entire lifecycle evolution of a sudden event from its outbreak and development to its handling and conclusion, ultimately outputting simulation data results across all dimensions, including rescue efficiency, secondary disaster risk impact, resource utilization effectiveness, and the degree of comprehensive balance among multiple objectives.

[0054] Based on this, this application selects a mature large-scale language model as the dedicated reasoning center of the entire core engine. The model is guided by customized structured prompt words. The prompt words fully cover all the constraints such as task execution instructions, emergency domain knowledge constraints, multi-objective trade-off rules, and output format specifications. This drives the large model to connect with the existing knowledge of the knowledge graph and the full data of multi-agent simulation to carry out in-depth logical analysis, event trend judgment, and multi-objective comprehensive trade-off.

[0055] This application designs a proprietary formula to quantify the reliability of the model analysis results and the effect of multi-objective trade-offs. Firstly, it designs a comprehensive confidence score calculation formula for the large model analysis results, assuming the set of knowledge graph triples invoked by the model in this analysis is denoted as . In the formula This refers to the p-th knowledge triple being invoked; the set of simulation data samples being invoked is... In the formula For the q-th simulation data sample called; a single triplet The inherent confidence level is Single simulation data The fit confidence level is Since both values ​​are in the range of 0 ≤ c ≤ 1, the overall confidence level of the analysis results is: In the formula, p represents the total number of knowledge triples called, and q represents the total number of simulation data samples called, taking into account the reliability of both knowledge source and simulation source data.

[0056] Subsequently, the weighting system of the four major decision-making objectives from the previous stage was adopted, and a multi-objective trade-off comprehensive score formula was designed. The predicted achievement values ​​of each objective after large-scale model analysis were set as follows: value range , The overall score is calculated by weighing all factors. To correct the score, a comprehensive confidence level is introduced. This application sets a passing threshold η based on industry experience, with a value ranging from 0.75 to 0.85. Only when the score meets the passing threshold... At that time, the output scheme is determined to meet the requirements of multi-objective optimization. Meanwhile, this application uses a professional corpus dedicated to emergency management to conduct domain fine-tuning training on the large model, covering professional content such as disaster evolution laws, disposal technical specifications, multi-objective decision-making methods, and industry disposal guidelines, to optimize the model's domain adaptability, avoid professional analysis biases of general large models, and finally the engine outputs a complete situation assessment report, multi-objective decision suggestions, and a full-process simulation dataset, and pushes all data to the downstream decision application layer simultaneously.

[0057] Step S4: Construct a hierarchical emergency decision-making function module and a visual interactive decision-making cockpit, and integrate the simulation and decision-making results to achieve hierarchical solution output, process retrospective review, flexible parameter adjustment, and intuitive display of analysis results, thus completing the entire process of intelligent emergency decision-making.

[0058] This step corresponds to the following: Figure 3 The decision application layer, and such Figure 4 The emergency decision-making cockpit shown has a complete three-layer front-end architecture, which receives all simulation data and analysis results output by the core engine layer, builds hierarchical and exclusive decision-making capabilities for emergency decision-making entities at different levels, and builds a full-link visual interactive platform to realize the application of all results and human-machine collaborative decision-making.

[0059] Firstly, to address the differentiated work needs of three levels—high-level strategic decision-making, mid-level management decision-making, and grassroots execution decision-making—three dedicated core functional modules are built for each level. The first is the emergency situation simulation module, serving high-level macro-strategic decision-making. It comprehensively analyzes the full results of multi-agent simulations and knowledge from the emergency management knowledge graph domain to accurately predict the entire lifecycle evolution path of emergencies, potential secondary disaster risks, overall impact scope, and development trends. Combined with multi-objective trade-off logic, it outputs macro-strategic deployment basis to match the business needs of high-level overall decision-making. The second is the emergency resource allocation optimization module, serving mid-level dispatch and management decision-making. Based on emergency quantitative models and digital twins of the response entities, combined with multi-objective optimization algorithms, it conducts multi-dimensional benefit calculations for various resource allocation schemes.

[0060] This application establishes a dedicated multi-objective optimization mathematical system for this module, assuming the set of all possible resource allocation schemes is as follows: In the formula For the nth emergency resource allocation plan, a single plan corresponds to a complete set of parameters such as the amount of materials allocated and the amount of manpower deployed. At the same time, three core optimization objectives are defined: the objective of maximizing resource efficiency. , Maximize the effect of treatment Minimize allocation costs In the formula The actual utilization rate of emergency resources of type j Let be the weight coefficient corresponding to the j-th type of resource. The total reserve of emergency resources that can be dispatched throughout the region. Let k be the quantitative value of the effect of the kth treatment step. This represents the weighting coefficient corresponding to the k-th processing stage. For the unit allocation cost of resource type j, This defines the number of resources to be allocated for the j-th type in the solution; it also sets real business constraints such as a maximum limit on the total amount of resources and a maximum limit on a single type of resource. , In the formula Let j be the maximum number of resources of type j that can be allocated. Finally, a comprehensive optimization objective function is constructed through weighted fusion. In the formula For each optimization objective weight coefficient to satisfy The weighting coefficients for the three objectives (satisfying) , Given the maximum cost of resource allocation across all configuration schemes, solving the function for the optimal solution will output the optimal resource allocation scheme for multi-objective equilibrium.

[0061] The third module is the emergency response risk early warning module, which serves grassroots on-site decision-making and monitors all risk indicators, resource indicators, response progress indicators, and multi-objective achievement data in real time throughout the simulation process. This application designs a dedicated risk index evaluation system and sets a set of monitoring indicators. In the formula This refers to the k-th risk monitoring indicator; the real-time monitoring value of a single indicator. Safety threshold Warning threshold Corresponding risk weights satisfy The formula for calculating the risk index is: In the formula As a comprehensive risk index for emergency response, it limits the excessive influence of a single indicator on the overall risk rating by using an upper limit coefficient. At the same time, it divides the risk into three levels: safety, early warning, and emergency. Based on the index results, it outputs precise on-site response suggestions to match the needs of grassroots rapid response and risk prevention.

[0062] Based on the hierarchical functional modules, this application constructs a complete emergency decision-making dashboard, corresponding to, for example: Figure 4The three-tiered front-end architecture shown is divided into three main layers: the interactive presentation layer, the core technology support layer, and the data interaction layer. It also completes the selection of the entire front-end technology stack, component-based development, functional module construction, and cross-layer data interoperability design. At the core technology stack selection level, this application uses React 18 paired with TypeScript as the overall core UI framework. Leveraging its declarative development paradigm, efficient virtual DOM rendering mechanism, and static type safety verification capabilities, it adapts to the development needs of the cockpit, which features multiple modules, high interactivity, and large-scale component reuse. The Zustand library is used for global state management, enabling the distribution and subscription management of global asynchronous data such as inference scenario configuration, resource data caching, and AI interaction states throughout the application. Recharts is used as the dedicated data visualization engine, handling all chart drawing and multi-dimensional data visualization rendering. TailwindCSS combined with CSSModules is used to build a global visual style system, unifying the interface design style across the entire platform. Vite is used as the front-end project building tool, relying on its hot module updates and rapid startup capabilities to ensure efficient project iteration and rapid deployment.

[0063] At the hierarchical functional and module implementation level, the top layer is the interactive display layer, which is the core interface layer of the entire human-computer interaction. It has three main functional areas: a simulation process playback area, an analysis report display area, and an interactive operation area, which respectively support the two top-level business modules: the Emergency Intelligence Center and the Emergency Simulation Sandbox. The Emergency Intelligence Center module connects to the data source of the analysis report and is responsible for summarizing and displaying all emergency analysis results; the Emergency Simulation Sandbox module corresponds to simulation playback and full-process interactive control, supporting dynamic simulation interaction throughout the entire process. The core function of this layer is to complete the visualization of all decision-making information and the full-process human-computer interaction operation, implementing the system's "human-computer symbiosis" design concept. The middle layer is the core technology support layer, which is built by the Vite tool to coordinate the packaging and hot update maintenance of all modules, and manages all front-end underlying technology components downwards. It also links the global state management, visualization engine, style system, and all underlying capabilities to provide high-performance, highly reusable, and type-safe component support for the upper-layer interface.

[0064] The bottom layer is the data interaction layer, which establishes a full-link data communication channel and integrates three communication solutions: Fetch API basic data retrieval, SSE real-time report push, and WebSocket real-time simulation push. Each solution has a dedicated backend interface. The / api / analyze-emergency interface is responsible for acquiring analysis report data, while the / api / run-emergency-simulation interface is responsible for simulation data interaction and command control. This establishes a low-latency, highly stable, and bidirectional data transmission link, receiving all data requests from the front-end interface and connecting to the underlying core engine layer and data display layer, enabling data communication and collaborative linkage between all levels of the system and between internal and external systems.

[0065] Based on a complete cockpit architecture, the system achieves multi-dimensional interactive capabilities: the simulation process playback interface, through a timeline and key node annotations, fully reproduces the entire process of behavior, collaborative processes, and event situation evolution of each agent in the multi-agent simulation, supporting full-process traceability and review of decision-making; the analysis report display area breaks down the judgment report generated by the large model into an execution summary and detailed full text, accompanied by various trend charts, comparison charts, and correlation network diagrams to complete the visualization auxiliary display, helping decision-makers quickly grasp core information; the interactive operation area provides full parameter adjustment, scenario filtering, and scheme comparison capabilities, supporting users to customize resource parameters and disaster scenario configurations, simulate the effects of different schemes, and achieve flexible control and precise optimization of the decision-making process. Thus, the entire process, from data collection, knowledge construction, engine analysis to interactive applications, is a complete closed loop, comprehensively realizing multi-objective intelligent decision support for emergency management across all scenarios.

[0066] Example 2 This invention provides an intelligent decision support system for emergency management oriented towards multi-objective dynamic evolution. Figure 5 This is a flowchart illustrating an intelligent decision support system for emergency management oriented towards multi-objective dynamic evolution, provided as an embodiment of the present invention. Figure 5 As shown, the system includes: The data acquisition unit 100 is used to automatically collect heterogeneous data related to emergency management across the entire domain through multiple sources, and to complete data cleaning, verification, quality assessment and periodic updating and management to form standardized basic emergency data resources. The graph construction unit 200 is used to perform entity recognition, relation extraction and graph structure optimization based on the basic emergency data resources, and to expand knowledge nodes and related links by combining domain implicit association reasoning to construct an emergency management domain knowledge graph. The intelligent analysis unit 300 is used to build a comprehensive intelligent analysis engine based on the emergency management knowledge graph, integrating emergency quantitative modeling, digital twins of the disposal subjects, multi-agent collaborative simulation and large model domain reasoning technology, and output simulation inference and decision-making results. The decision application unit 400 is used to construct a hierarchical emergency decision-making function module and a visual interactive decision-making cockpit. It connects to the simulation and decision-making results, realizes hierarchical scheme output, process retrospective review, flexible parameter adjustment and intuitive display of analysis results, and completes the whole process of intelligent emergency decision-making.

[0067] Regarding the system in the above embodiments, the specific ways in which each module performs operations have been described in detail in the embodiments related to the method, and will not be elaborated here.

[0068] Example 3 To implement the methods of the above embodiments, the present invention also provides an electronic device, which includes a memory and a processor; wherein the processor reads executable program code stored in the memory to run a program corresponding to the executable program code, so as to implement the various steps of the methods described above.

[0069] Example 4 To implement the above embodiments, this application also proposes a non-transitory computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the method described in the foregoing embodiments.

[0070] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

[0071] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0072] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this invention, "a plurality of" means at least two, such as two, three, etc., unless otherwise explicitly specified.

Claims

1. A smart decision support method for emergency management oriented towards multi-objective dynamic evolution, characterized in that, include: By automatically collecting heterogeneous data related to emergency management across the entire domain through multiple sources, and completing data cleaning, verification, quality assessment, and periodic updating and governance, standardized basic emergency data resources are formed. Based on the aforementioned basic emergency data resources, entity recognition, relation extraction, and graph structure optimization are performed. Knowledge nodes and related links are expanded by combining domain implicit association reasoning to construct a knowledge graph for the emergency management domain. Based on the aforementioned emergency management knowledge graph, a comprehensive intelligent analysis engine is constructed by integrating emergency quantitative modeling, digital twins of response entities, multi-agent collaborative simulation, and large-scale model domain reasoning technology, and outputs simulation inference and decision-making results. A hierarchical emergency decision-making function module and a visual interactive decision-making cockpit are constructed, which are connected to the simulation and decision-making results to realize hierarchical solution output, process retrospective review, flexible parameter adjustment and intuitive display of analysis results, and complete the whole process of intelligent emergency decision-making.

2. The method according to claim 1, characterized in that, By automatically collecting heterogeneous data related to emergency management across multiple sources, and completing data cleaning, verification, quality assessment, and periodic updates, standardized basic emergency data resources are formed, including: It adopts automated collection methods such as web crawling, standardized data interfaces, and API calls to collect emergency open source intelligence data covering five dimensions: risks and hazards, emergency resources, emergencies, policies and regulations, and public opinion dynamics. Connect with official data interfaces of meteorological, geological and emergency management authorities to obtain real-time early warning and basic thematic data; collect event handling dynamic information from official platforms and news media through web crawlers; and capture public attention hotspots and emergency demands information with the help of public opinion collection tools. Establish a regular update mechanism with hourly, daily, or weekly updates, conduct real-time verification and deduplication preprocessing of collected data, and build a data quality assessment system to control data integrity and accuracy through quantitative scoring, ultimately forming standardized basic emergency data resources.

3. The method according to claim 2, characterized in that, Based on the aforementioned basic emergency data resources, entity recognition, relation extraction, and graph structure optimization are performed. Knowledge nodes and related links are expanded by combining domain implicit association reasoning, thus constructing a knowledge graph for the emergency management domain, including: Based on standardized basic emergency data resources, deep learning entity linking technology is used to complete the identification of unstructured and semi-structured data entities, extract core entities such as risk hazards, emergency resources, emergencies, handling entities, and policy norms, and use attention mechanism extraction models to mine the causal, matching, collaborative, and constraint relationships between entities; Entities and their corresponding relationships are stored in a triple graph data structure. Reinforcement learning algorithms are used to optimize the graph structure. The action space is entity addition and redundant relationship deletion, and the state space is graph association density and relationship confidence. The topology structure is iteratively optimized based on the weighted reward function of accuracy and association density. By integrating rule-based reasoning and Bayesian network probabilistic reasoning techniques, implicit links between entities are discovered. New implicit relationships are determined based on preset conditional probability confidence thresholds, expanding knowledge graph nodes and related networks, and completing the overall construction of the knowledge graph.

4. The method according to claim 3, characterized in that, Emergency quantitative modeling specifically includes: Based on the emergency management knowledge graph, natural language processing technology is used to analyze emergency policies, response plans and risk indicators, extract core elements of response standards and risk thresholds, and combine multi-objective trade-off algorithms to complete the weight division of objectives such as rapid response, minimum loss, resource efficiency and safety controllability. The weight coefficients of each decision objective are solved by the analytic hierarchy process (AHP) to construct a multi-objective comprehensive evaluation model, which transforms decision parameters into quantifiable and calculable indicators. For emergency resource allocation scenarios, a quantitative calculation formula is established by combining regression analysis and causal inference methods to calculate the relationship between rescue speed, loss reduction ratio, resource utilization rate and resource reserves, response time, and allocation cost. The quantitative model is improved based on scenario correction coefficients.

5. The method according to claim 4, characterized in that, The specific digital twins of the disposal entities include: By combining emergency management knowledge graphs, basic emergency data from the early stages, industry reports, and expert experience, a full-dimensional digital twin model is constructed for emergency management departments, fire and rescue teams, medical institutions, and material reserve centers. Simulate the organizational structure, response process, rescue capabilities, resource reserves and scheduling mechanisms of each entity within the twin model, and drive the simulation of multi-entity collaborative response behavior by combining emergency quantitative model parameters; Feedback and corrections are made based on the real-time operational data of the disposal entities, and the internal parameters of the twin model are dynamically adjusted to ensure that the model accurately maps the actual operating status and decision-making logic of each entity.

6. The method according to claim 5, characterized in that, Multi-agent cooperative simulation specifically includes: The main body of the entire emergency response process is abstracted into an independent simulated intelligent agent. Each intelligent agent is equipped with a multi-objective decision-making algorithm and corresponding weight parameters. Based on environmental perception information including the situation of the emergency and the remaining resources, it autonomously outputs decision-making actions including resource allocation and route adjustment. A quantitative formula for information interaction intensity among intelligent agents is constructed. The interaction intensity is calculated based on the similarity of target weights, state differences, and the distance between the main functions, so as to realize information exchange and mutual influence of behavior among multiple intelligent agents. The system dynamically simulates the entire lifecycle of disaster response and outputs multi-dimensional simulation data, including rescue efficiency, secondary disaster impact, resource utilization effectiveness, and the degree of balance among multiple objectives.

7. The method according to claim 6, characterized in that, Construct a hierarchical emergency decision-making functional module and a visual interactive decision-making dashboard, including: To meet the decision-making needs of the three levels of high-level strategy, middle-level management and grassroots execution, three functional modules are built respectively: emergency situation simulation, emergency resource allocation optimization and emergency response risk warning. Combined with multi-objective optimization and risk index quantification algorithm, hierarchical decision output and risk level warning are realized. The overall architecture of the visual interactive decision-making cockpit is built using component-based front-end development technology. It is equipped with global data status management, multi-dimensional data visualization, unified interface style specifications, and efficient project construction and compilation. This completes the development of a highly interactive component-based system, enabling intuitive display of analysis results, retrospective review of the decision-making process, and flexible parameter adjustment.

8. An intelligent decision support system for emergency management oriented towards multi-objective dynamic evolution, characterized in that, include: The data acquisition unit is used to automatically collect heterogeneous data related to emergency management across the entire domain through multiple sources, and to complete data cleaning, verification, quality assessment and periodic updating and management to form standardized basic emergency data resources. The graph construction unit is used to perform entity recognition, relation extraction and graph structure optimization based on the basic emergency data resources, and to expand knowledge nodes and related links by combining domain implicit association reasoning to construct an emergency management domain knowledge graph. The intelligent analysis unit is used to build a comprehensive intelligent analysis engine based on the emergency management knowledge graph, integrating emergency quantitative modeling, digital twins of the response subjects, multi-agent collaborative simulation and large model domain reasoning technology, and output simulation inference and decision-making results. The decision application unit is used to construct a hierarchical emergency decision-making function module and a visual interactive decision-making cockpit. It connects to the simulation and decision-making results, realizes hierarchical scheme output, process retrospective review, flexible parameter adjustment and intuitive display of analysis results, and completes the entire process of intelligent emergency decision-making.

9. An electronic device, characterized in that, Including processor and memory; The processor reads executable program code stored in the memory to run a program corresponding to the executable program code, so as to implement the method as described in any one of claims 1-7.

10. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the method as described in any one of claims 1-7.

Images