Emergency decision-making auxiliary system and method based on knowledge graph, electronic equipment and storage medium

Abstract

The invention provides an emergency decision-making auxiliary system based on a knowledge graph. Comprising a data acquisition and processing module, a knowledge graph construction and reasoning module, an information retrieval and generation module, a multi-level reasoning and thinking chain prompting module, a multi-agent cooperation and game optimization module, an incremental learning and dynamic knowledge updating module and a block chain trust guarantee module. The emergency decision-making efficiency, accuracy, cooperation capability and safety are broken through, dynamic and complex emergency scenes can be effectively dealt with, and efficient, reliable and intelligent technical support is provided for decision makers.

Description

Technical Field

[0001] The present invention relates to the field of knowledge graph technology, and in particular to an emergency decision support system, method and storage medium based on a knowledge graph. Background Art

[0002] With the frequent occurrence of emergencies such as natural disasters, accidents, and public health incidents, rapid and accurate emergency decision-making has become a core requirement for reducing losses and ensuring life safety. However, emergency scenarios often involve multi-source heterogeneous data (such as sensor monitoring data, social media information, expert experience, etc.), multiple participants (such as government departments, rescue teams, medical institutions, etc.), and dynamically changing environmental factors, which poses severe challenges to traditional emergency decision-making technologies. Currently, the research and application of related technologies at home and abroad still have the following limitations:

[0003] 1. The rigidity of traditional decision support systems. Existing emergency response systems are mostly based on preset rules and static models, capable of processing only structured data and single scenarios. For example, disaster warning systems rely on historical data for linear predictions, making it difficult to integrate unstructured data such as real-time video and images. Resource scheduling models are limited to fixed priority algorithms and cannot adapt to dynamic adjustments in emergencies. When faced with complex and changing emergency situations, such systems often suffer from delayed or even erroneous decisions due to inefficient data integration and poor model generalization.

[0004] 2. Limitations of big data and artificial intelligence technologies. In recent years, some studies have attempted to introduce big data analysis and machine learning technologies to enhance emergency response capabilities. For example, the disaster impact prediction model based on deep learning can analyze the affected area through satellite imagery, but it relies on offline training data and is difficult to update in real time; natural language processing (NLP) technology can extract disaster information from social media, but lacks correlation analysis with geographic information and resource status. In addition, existing technologies focus more on optimizing a single link (such as early warning or resource scheduling) and have not formed an end-to-end collaborative decision-making system, resulting in a prominent "information island" problem and limited global decision-making efficiency.

[0005] 3. Insufficient application of knowledge graphs. Knowledge graph technology has shown potential in the field of emergency response due to its powerful knowledge representation and reasoning capabilities. Existing research has linked entities such as events, locations, and resources by constructing disaster knowledge graphs, but the construction process relies on manual labeling, making it difficult to automatically integrate multimodal data (such as text, images, and sensor streams); the graph update mechanism mostly uses regular batch updates, which cannot support incremental learning in real-time dynamic environments. For example, in earthquake rescue, disaster data evolves rapidly over time, and traditional graphs are difficult to integrate new casualty information or road damage status in real time, resulting in outdated decision-making basis.

[0006] 4. Inefficiency of multi-agent collaboration. Multi-agent systems (MAS) can enhance emergency response flexibility through distributed decision-making, but existing technologies have not effectively addressed conflicts of interest and resource competition among agents. For example, rescue robots, drones, and command centers need to coordinate tasks, but traditional collaboration models are often based on fixed rules or simple game-playing, making it difficult to dynamically optimize strategies when resources are limited. Furthermore, the lack of a reliable data sharing mechanism leads to delays and the risk of tampering in information transmission between agents.

[0007] 5. Lack of data security and trust mechanisms. Emergency data often involves sensitive information (such as personnel locations and medical records). Existing systems often use centralized databases for storage, making them vulnerable to single points of failure or malicious attacks. Although blockchain technology has been proposed for data security, its application in emergency scenarios is still limited to simple evidence storage and lacks deep integration with decision-making processes, making it impossible to achieve dynamic data verification and granular permission control.

[0008] In summary, existing technologies have significant shortcomings in real-time multi-source data fusion, dynamic knowledge reasoning, multi-agent game optimization, and data security and reliability, making them difficult to meet the needs of efficient, accurate, and secure decision-making in complex emergency scenarios. Therefore, an innovative technical solution integrating knowledge graphs, incremental learning, multi-agent collaboration, and blockchain is urgently needed to overcome existing bottlenecks and comprehensively enhance the comprehensive capabilities of emergency decision-making systems. Summary of the Invention

[0009] This invention proposes an emergency decision-making support system, method, and device based on a knowledge graph, which solves the problems of low efficiency in data standardization and governance and the inability to fully cover the entire process from data governance to intelligent analysis in existing technologies. The technical solution of this invention is achieved as follows:

[0010] The emergency decision support system based on knowledge graph includes:

[0011] Data acquisition and processing module, used to collect multi-source heterogeneous emergency data and clean, denoise, standardize and convert the data into different formats;

[0012] The knowledge graph construction and reasoning module is used to build a dynamic knowledge graph through multimodal data fusion technology. The nodes of the knowledge graph represent entities in the emergency event, and the edges represent the relationships between entities. The graph convolutional network (GCN) is used to fuse and reason about multimodal data.

[0013] The information retrieval and generation module builds a semantic space model based on a deep neural network, matches user queries with knowledge graph nodes, and generates emergency decision-making solutions through a generative adversarial network (GAN);

[0014] The multi-level reasoning and thought chain prompt module uses recurrent neural networks (RNNs) to achieve hierarchical decomposition of tasks and combines this with a reasoning engine to provide multi-level decision support.

[0015] Multi-agent collaboration and game optimization module, which optimizes resource allocation and collaboration strategies among multiple agents through game theory models and dynamically adjusts resource scheduling;

[0016] The incremental learning and dynamic knowledge update module uses incremental learning algorithms to update the knowledge graph in real time to ensure real-time decision support;

[0017] The blockchain trust assurance module records key decisions and data transactions through blockchain technology, uses smart contracts to verify the authenticity and integrity of data, and controls data access rights.

[0018] As a further technical solution, the knowledge graph construction and reasoning module specifically includes:

[0019] The data preprocessing unit uses convolutional neural networks (CNN) and recurrent neural networks (RNN) to clean and extract features from multimodal data;

[0020] Heterogeneous data fusion unit, mapping text, image, video and sensor data into a unified knowledge graph;

[0021] The dynamic update unit updates the knowledge graph in real time based on the incremental learning algorithm. The update formula is:

[0022] G(t)=G(t-1)∪ΔG(t)

[0023] Among them, ΔG(t) is the newly added knowledge part, and G(t) is the updated knowledge graph.

[0024] As a further technical solution, in the information retrieval and generation module,

[0025] Semantic retrieval calculates the matching degree between user query and knowledge graph node through cosine similarity. The formula is:

[0026]

[0027] Among them, x is the query input, z is the node in the knowledge graph, and f embed (x) represents the input embedding vector, and cosine similarity is used to measure their semantic similarity;

[0028] The Generative Adversarial Network (GAN) generates a decision plan based on the retrieval results and real-time environmental data. The generation formula is:

[0029] y=GenerateAnswer(x,I real-time ,θ)

[0030] Among them, y is the generated emergency decision plan, z is the retrieved relevant knowledge, I real-time is the real-time environment data, and θ is the parameter of the generated network.

[0031] As a further technical solution, in the multi-agent collaboration and game optimization module:

[0032] The game model realizes resource allocation by optimizing the objective function, which is:

[0033] max∑ i R i X i subject∑ i C i X i ≤S max ,

[0034] Among them, R i is the reward of agent i, X i To allocate resources, C i is resource consumption, S max is the total amount of resources;

[0035] The collaboration weight is dynamically adjusted through the game theory formula, which is:

[0036]

[0037] Among them, a i and a j For the intelligent agent, a ij Collaboration weight, Collaboration(a i ,a j ) represents the degree of collaboration between agents.

[0038] As a further technical solution, the blockchain trust assurance module specifically includes:

[0039] The decentralized storage unit records the key data hash value to the blockchain. The verification formula is:

[0040] H block =Hash(Data n ,Signature user ,Timestamp)

[0041] Among them, H block is the hash value of the data in the blockchain, Hash is the cryptographic hash function, Data n For transaction data, Signature user The user signature and Timestamp are timestamps.

[0042] Smart contract unit controls data access rights, and the execution condition formula is:

[0043] Access_allowed=Verify(SmartContract,UserPermission),

[0044] Among them, Access_allowed is the access permission verification result, SmartContract is the executed smart contract, and UserPermission is the user permission information.

[0045] The emergency decision support system based on knowledge graph is characterized in that the incremental learning and dynamic knowledge update module uses support vector machine (SVM) or deep learning network for online training.

[0046] As a further technical solution, the agents in the multi-agent collaboration and game optimization module include decision makers, executors, expert groups and emergency response teams, which adjust collaboration strategies through real-time communication.

[0047] An emergency decision-making method of the system, characterized by comprising the following steps:

[0048] Step S1: Collect multi-source heterogeneous data and perform preprocessing;

[0049] Step S2: Build a dynamic knowledge graph and perform reasoning through a graph convolutional network (GCN);

[0050] Step S3: Generate emergency plans based on semantic retrieval and generative adversarial network (GAN);

[0051] Step S4: Optimize resource allocation through multi-agent game;

[0052] Step S5: Using blockchain technology to ensure data security and transparent sharing;

[0053] Step S6: Update the knowledge graph in real time through incremental learning.

[0054] A computer-readable storage medium stores a computer program, which implements the steps of an emergency decision-making method when executed by a processor.

[0055] An electronic device includes a storage medium, a processor and a computer program stored in the memory. When the processor executes the program stored in the storage medium, the steps of the emergency decision-making method are implemented.

[0056] Compared with the existing technology, this solution has the following beneficial effects:

[0057] (1) Significantly improve emergency decision-making efficiency. Through multimodal data fusion (text, images, videos, etc.) and deep learning technology, multi-source heterogeneous information can be quickly integrated, shortening data preprocessing and knowledge extraction time. Dynamic knowledge graphs combined with incremental learning algorithms can update event status and resource distribution in real time, avoiding decision lags caused by data delays in traditional systems.

[0058] (2) Enhance decision-making accuracy and adaptability. The knowledge graph's graph convolutional network (GCN) reasoning capabilities support complex entity relationship analysis, ensuring that decision-making solutions meet the needs of actual scenarios. The incremental learning module continuously optimizes model parameters and dynamically adjusts decision logic based on real-time environmental data, improving adaptability to sudden changes.

[0059] (3) Optimize multi-party collaboration and resource allocation. The multi-agent collaboration and game optimization module balances the interests of multiple parties through game theory models, resolves resource conflicts, and improves task execution efficiency. The priority scheduling algorithm dynamically adjusts the order of tasks based on real-time feedback (such as disaster level and resource consumption) to ensure the coordination of global decision-making and local execution.

[0060] (4) Ensure data security and system credibility. Blockchain technology enables decentralized storage of key data and smart contract verification, ensuring that data cannot be tampered with and is fully traceable. The smart contract mechanism based on permission control limits access to sensitive information and prevents unauthorized data leakage or abuse.

[0061] (5) Supporting intelligent decision-making in complex scenarios. Generative adversarial networks (GANs) combined with reinforcement learning generate diverse emergency response plans and optimize execution paths, breaking through the limitations of traditional rule bases. The multi-level reasoning module decomposes the task hierarchy through recurrent neural networks (RNNs), balancing the coherence of strategic planning and tactical execution. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0063] Figure 1 This is a structural block diagram of the emergency decision support system based on knowledge graph of the present invention. DETAILED DESCRIPTION

[0064] The following will clearly and completely describe the technical solutions of the present invention in conjunction with the embodiments of the present invention. Obviously, the embodiments described are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making any creative efforts shall fall within the scope of protection of the present invention.

[0065] Reference Figure 1 This invention proposes an emergency decision support system based on knowledge graph, which combines multiple advanced technologies (such as knowledge graph, deep learning, natural language processing (NLP), multi-agent collaboration (MAS), blockchain, incremental learning, etc.) to provide comprehensive decision support for emergency decision makers through real-time data fusion and dynamic reasoning. The system is applicable to various emergency events, including natural disasters, accident emergencies, public health events, etc. The core innovation lies in comprehensively improving the efficiency and accuracy of emergency decision-making by combining multiple intelligent technologies.

[0066] 1. Overall system architecture and implementation process

[0067] The system's overall architecture is based on a modular design and comprises six key modules: knowledge graph construction and reasoning, information retrieval and generation, multi-level reasoning and thought chain prompting, multi-agent collaboration and game optimization, incremental learning and dynamic knowledge updating, and blockchain trust assurance and data security. The system's implementation process follows the following sequence: data acquisition and preprocessing → knowledge graph construction → information retrieval and generation → multi-level reasoning and decision making → multi-agent collaboration and execution → incremental learning and dynamic updating → data logging and blockchain security. The collaboration between these modules ensures the efficiency, accuracy, and real-time performance of the emergency response system.

[0068] 2. Knowledge Graph Construction and Reasoning Module

[0069] 2.1 Data Preprocessing and Standardization

[0070] The data preprocessing stage is the foundation of the system. Its main tasks include data cleaning, denoising, standardization, and format conversion. We use deep learning models based on convolutional neural networks (CNNs) and recurrent neural networks (RNNs) to automatically extract key information from multimodal data, such as scene features from videos and images, and speech information from audio data. Deep learning is then used to automatically annotate and associate data. The formalized processing is:

[0071] D clean =f clean (D raw ) where D raw is the original data, D clean For cleaned data

[0072] The goal of this stage is to remove redundant parts of the data and ensure the standardized processing of information to facilitate the subsequent construction of the knowledge graph.

[0073] 2.2 Graph Construction and Knowledge Fusion

[0074] This system uses multimodal data fusion technology to build a dynamic knowledge graph. The nodes in the knowledge graph represent key entities in various emergency events (such as disaster type, location, participants, etc.), and the edges represent the relationships between these entities (such as command relationships, resource allocation relationships, etc.). The formula is expressed as:

[0075] G=(V,E) where V={v1,v2,...,v n}E={e1,e2,...,e m}

[0076] In this paper, V represents an entity node and E represents an edge between entities. Node attributes are automatically extracted from the literature using natural language processing (NLP) technology, and edge relationships are analyzed using a graph convolutional network (GCN).

[0077] In addition, we introduced a heterogeneous data fusion model to map data of different modalities (such as text, images, videos, etc.) into a unified knowledge graph, so that each type of data can provide supplementary information for emergency decision-making.

[0078] G fused =G text ∪G image ∪G sensor

[0079] Here, G fused Represents the fused knowledge graph, which contains knowledge from text, images, and sensor data.

[0080] 2.3 Dynamic Reasoning and Knowledge Update

[0081] The dynamic update mechanism of the knowledge graph is based on incremental learning and deep graph reasoning models. Whenever an emergency occurs, the system automatically updates the graph and dynamically adds new information to the existing knowledge structure. The incremental reasoning formula is expressed as:

[0082] G(t)=G(t-1)∪ΔG(t)

[0083] Among them, ΔG(t) is the newly added knowledge part. The system continuously corrects the relationships and entities in the knowledge graph through the incremental learning model to ensure that the information used in emergency decision-making is the latest.

[0084] 3. Information retrieval and generation module

[0085] 3.1 Semantic Retrieval and Multi-Dimensional Matching

[0086] In this module, the system first builds a semantic space model based on a deep neural network (DNN), converts the user's query into a vector form, and matches it with the nodes in the knowledge graph. The similarity between the query and the graph node is calculated using the following formula based on cosine similarity:

[0087]

[0088] Among them, x is the query input, z is the node in the knowledge graph, and f embed represents the embedding vector of the input, and cosine similarity is used to measure their semantic similarity.

[0089] 3.2 Information Generation and Reasoning Optimization

[0090] Based on information retrieval, this system further introduces a generative adversarial network (GAN) to generate emergency decision-making solutions that match the input query. The GAN can generate reasonable response plans based on the current emergency scenario and optimize the generated solutions through reinforcement learning. The generation formula is as follows:

[0091] y=GenerateAnswer(x,I real-time ,θ)

[0092] Among them, y is the generated emergency decision plan, z is the retrieved relevant knowledge, I real-time is the real-time environment data, and θ is the parameter of the generated network.

[0093] 4. Multi-level reasoning and thinking chain prompt module

[0094] 4.1 Task Decomposition and Hierarchical Reasoning

[0095] In this module, the system uses recurrent neural networks (RNNs) to implement hierarchical task decomposition and provides decision support at each level through an inference engine. The system sequentially processes high-level strategic decisions and low-level tactical execution through multiple inference levels to ensure the consistency of global decisions and the efficiency of local execution. The reasoning at each level is described by the following formula:

[0096] Prompt i =Prompt i-1 ∪ΔTask i

[0097] Among them, Prompt i is the decision hint for level i, Task i New tasks added to level i.

[0098] 4.2 Real-time task priority adjustment

[0099] The system uses a priority scheduling algorithm to dynamically adjust the execution order of each task based on real-time data (such as disaster level, resource status, etc.). The priority scheduling algorithm is based on the following formula:

[0100] P task =αT+βR+γD

[0101] Among them, P task is the priority of the task, T is the urgency of the task, R is the required resources, D is the real-time data feedback, and α, β and γ are adjustment factors.

[0102] 5. Multi-agent collaboration and game optimization module

[0103] 5.1 Agent Collaboration and Game Model

[0104] This module uses game theory to optimize collaboration and resource allocation among agents. Agents in the system include decision makers, executors, expert groups, emergency response teams, etc. Each agent optimizes its strategy through game theory to achieve maximum benefits. The optimization formula of the game model is:

[0105]

[0106] Among them, a i and a j For the intelligent agent, a ij Collaboration weight, Collaboration(a i ,a j ) represents the degree of collaboration between agents.

[0107] 5.2 Game Optimization and Resource Scheduling

[0108] Through the game optimization algorithm, the system dynamically adjusts resource allocation according to the current resource status, task priority and the status of each agent. The objective function in the optimization process is:

[0109]

[0110] Among them, R i is the reward of agent i, X i is the resource allocated to agent i, C i is resource consumption, S max The total amount of resources.

[0111] 6. Incremental learning and dynamic knowledge update module

[0112] 6.1 Incremental Learning Algorithm

[0113] The incremental learning algorithm continuously improves the knowledge graph by updating the real-time data stream. This module uses online learning models (such as support vector machines (SVMs) and deep learning networks) for incremental training to cope with the rapid changes in knowledge in a dynamic environment. The incremental update formula is:

[0114] G(t)=G(t-1)+ΔGnew(t)

[0115] Among them, ΔGnew(t) is the newly added knowledge at the current moment, and G(t) is the updated knowledge graph at time t.

[0116] 7. Blockchain trust assurance and data security module

[0117] 7.1 Decentralized Data Storage and Verification

[0118] To ensure data security and transparency in emergency decision-making, all key decisions and data transactions are recorded on the blockchain, and smart contracts are used to automatically verify the authenticity and integrity of the data. The blockchain security verification process is as follows:

[0119] H block =Hash(Data n ,Signature user ,Timestamp)

[0120] Among them, H block is the hash value of the data in the blockchain, Hash is the cryptographic hash function, Data n For transaction data, Signature user The user signature and Timestamp are timestamps.

[0121] 7.2 Trusted Data Sharing Mechanism

[0122] Blockchain controls data permissions and access through smart contracts, ensuring that different organizations and users can only access authorized data. The conditional formula for smart contract execution is:

[0123] Access_allowed=Verify(SmartContract,UserPermission)

[0124] Among them, Access_allowed is the access permission verification result, SmartContract is the executed smart contract, and UserPermission is the user permission information.

[0125] Compared with the existing technology, this solution has the following beneficial effects:

[0126] (1) Efficient decision support: By combining multimodal data, deep learning, and reasoning technology, the present invention can quickly integrate information from multiple parties, provide immediate emergency decision support, and significantly improve response efficiency.

[0127] (2) Accurate decision-making capabilities: Knowledge graphs and incremental learning algorithms ensure that the decision-making process can be updated in real time, providing accurate decision-making solutions based on the latest environmental data and changes.

[0128] (3) Intelligence and adaptability: The multi-agent collaboration and game optimization module can adjust strategies based on real-time feedback and environmental changes, thereby improving the system's adaptability.

[0129] (4) Data security and transparency: Blockchain technology ensures data security, transparency, and immutability in the emergency decision-making process, greatly enhancing the credibility of the system.

[0130] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. The emergency decision support system based on knowledge graph is characterized by: include: Data acquisition and processing module, used to collect multi-source heterogeneous emergency data and clean, denoise, standardize and convert the data into different formats; The knowledge graph construction and reasoning module is used to build a dynamic knowledge graph through multimodal data fusion technology. The nodes of the knowledge graph represent entities in the emergency event, and the edges represent the relationships between entities. The graph convolutional network (GCN) is used to fuse and reason about multimodal data. The information retrieval and generation module builds a semantic space model based on a deep neural network, matches user queries with knowledge graph nodes, and generates emergency decision-making solutions through a generative adversarial network (GAN); The multi-level reasoning and thought chain prompt module uses recurrent neural networks (RNNs) to achieve hierarchical decomposition of tasks and combines this with a reasoning engine to provide multi-level decision support. Multi-agent collaboration and game optimization module, which optimizes resource allocation and collaboration strategies among multiple agents through game theory models and dynamically adjusts resource scheduling; The incremental learning and dynamic knowledge update module uses incremental learning algorithms to update the knowledge graph in real time to ensure real-time decision support; The blockchain trust assurance module records key decisions and data transactions through blockchain technology, uses smart contracts to verify the authenticity and integrity of data, and controls data access rights.

2. The knowledge graph-based emergency decision support system according to claim 1, characterized in that: The knowledge graph construction and reasoning module specifically includes: The data preprocessing unit uses convolutional neural networks (CNN) and recurrent neural networks (RNN) to clean and extract features from multimodal data; Heterogeneous data fusion unit, mapping text, image, video and sensor data into a unified knowledge graph; The dynamic update unit updates the knowledge graph in real time based on the incremental learning algorithm. The update formula is: G(t)=G(t-1)∪ΔG(t) Among them, ΔG(t) is the newly added knowledge part, and G(t) is the updated knowledge graph.

3. The knowledge graph-based emergency decision support system according to claim 1, characterized in that: In the information retrieval and generation module, Semantic retrieval calculates the matching degree between user query and knowledge graph node through cosine similarity. The formula is: Among them, x is the query input, z is the node in the knowledge graph, and f embed represents the embedding vector of the input, Cosine similarity is used to measure their semantic similarity; The Generative Adversarial Network (GAN) generates a decision plan based on the retrieval results and real-time environmental data. The generation formula is: y=GenerateAnswer(x,I real-time ,i) Among them, y is the generated emergency decision plan, z is the retrieved relevant knowledge, I real-time is the real-time environment data, and θ is the parameter of the generated network.

4. The knowledge graph-based emergency decision support system according to claim 1, characterized in that: In the multi-agent collaboration and game optimization module: The game model realizes resource allocation by optimizing the objective function, which is: Among them, R i is the reward of agent i, X i To allocate resources, C i is resource consumption, S max Total amount of resources; The collaboration weight is dynamically adjusted through the game theory formula, which is: Among them, a i and a j For the intelligent agent, a ij Collaboration weight, Collaboration(a i ,a j ) represents the degree of collaboration between agents.

5. The knowledge graph-based emergency decision support system according to claim 1, characterized in that: The blockchain trust assurance module specifically includes: The decentralized storage unit records the key data hash value to the blockchain. The verification formula is: H block =Hash(Data n ,Signature user ,Timestamp) Among them, H block is the hash value of the data in the blockchain, Hash is the cryptographic hash function, Data n For transaction data, Signature user The user signature and Timestamp are timestamps. Smart contract unit controls data access rights, and the execution condition formula is: Access_allowed=Verify(SmartContract,UserPermission) Among them, Access_allowed is the access permission verification result, SmartContract is the executed smart contract, and UserPermission is the user permission information.

6. The knowledge graph-based emergency decision support system according to any one of claims 1 to 5, characterized in that: The incremental learning and dynamic knowledge updating module uses a support vector machine (SVM) or a deep learning network for online training.

7. The knowledge graph-based emergency decision support system according to claim 1, characterized in that: The agents in the multi-agent collaboration and game optimization module include decision makers, executors, expert groups and emergency response teams, which adjust collaboration strategies through real-time communication.

8. An emergency decision-making method based on the system of claim 1, characterized in that: The following steps are involved: Step S1: Collect multi-source heterogeneous data and perform preprocessing; Step S2: Build a dynamic knowledge graph and perform reasoning through a graph convolutional network (GCN); Step S3: Generate emergency plans based on semantic retrieval and generative adversarial network (GAN); Step S4: Optimize resource allocation through multi-agent game; Step S5: Using blockchain technology to ensure data security and transparent sharing; Step S6: Update the knowledge graph in real time through incremental learning.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps of the emergency decision-making method according to claim 8 are implemented.

10. An electronic device comprising a storage medium, a processor, and a computer program stored in the memory, wherein: When the processor executes the program stored in the storage medium according to claim 9, the steps of the method according to claim 8 are implemented.

Images