Rainstorm flood disaster emergency decision system and method based on domain knowledge ontology

Abstract

The application discloses a rainstorm and flood disaster emergency decision system and method based on a domain knowledge ontology, and relates to the technical field of emergency management of sudden disasters and artificial intelligence. The system comprises a domain knowledge ontology module, a multi-agent decision module, a data interaction module and a visual output module. The domain knowledge ontology module is connected with the multi-agent decision module through a knowledge interface. Each agent in the multi-agent decision module transmits information through a communication protocol. The data interaction module is connected with an external data source and a multi-department business system, and the data interaction module is connected with the multi-agent decision module through a data transmission channel. The application is suitable for the whole process of monitoring analysis, early warning and prediction, response disposal and evaluation optimization of rainstorm and flood disasters. Through structured knowledge reasoning and multi-agent cooperation, the application provides a predictive and intelligent decision scheme for disaster prevention and reduction and emergency management departments.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of emergency management of sudden disasters and artificial intelligence, in particular to a rainstorm and flood disaster emergency decision system and method based on domain knowledge ontology. BACKGROUND

[0002] With the influence of climate change and strong human activities, unconventional sudden water disasters are increasingly frequent and severe worldwide. Rainstorm and flood disasters have the characteristics of strong suddenness, wide influence, and multiple chain and secondary disasters. The emergency environment presents high uncertainty and complexity, and the decision response time is urgent. The timeliness and scientificity of emergency decision-making directly affect the degree of disaster loss.

[0003] The existing rainstorm and flood disaster emergency decision system mainly has the following technical defects, which is difficult to meet the needs of complex disaster scenes:

[0004] (1) Fragmented knowledge: Emergency decision systems involve various structured and unstructured knowledge, such as geographic information, environmental monitoring, disaster cases, and emergency plans, which are stored in different databases. Lack of structured association and integration leads to "data islands", making it difficult to quickly match decision-making needs;

[0005] (2) Decision response lag: In the face of complex flood disaster scenes (such as mountain torrents, embankment breaches and dam breaches), decision-making subjects involve multiple departments at different levels. Decision-making relies on expert experience and consultation mechanisms, and the response time usually exceeds 30 minutes, missing the best emergency disposal window;

[0006] (3) Lack of multi-agent collaboration: Emergency management involves meteorology, water conservancy, transportation, civil affairs and other departments. The existing decision system lacks a collaborative interaction mechanism. The decision efficiency of multi-organizational participation in the emergency environment is low. Disaster monitoring, forecasting, risk, early warning and emergency response and action subjects cannot interact in real time and efficiently. Emergency decision information is not smooth;

[0007] (4) Poor decision dynamics: Emergency systems often only execute pre-set processes and cannot dynamically adjust plans according to real-time changes (such as rapid river water level rise, sudden rainfall increase, and embankment piping). Flood disaster emergency decision-making lacks foresight and weak adaptability. SUMMARY

[0008] To solve the above problems, the present application provides a rainstorm and flood disaster emergency decision system and method based on domain knowledge ontology. The technical solution is as follows:

[0009] On the one hand, the present application provides a rainstorm and flood disaster emergency decision system based on domain knowledge ontology, which comprises:

[0010] The domain knowledge ontology module, the multi-agent decision module, the data interaction module and the visual output module;

[0011] The domain knowledge ontology module is connected with the multi-agent decision module through a knowledge interface, and provides knowledge support for the collaborative decision of multiple agents in the multi-agent decision module;

[0012] The agents in the multi-agent decision module communicate information through a communication protocol;

[0013] The data interaction module is connected with external data sources and multi-department business systems respectively, and the data interaction module is connected with the multi-agent decision module through a data transmission channel, and provides real-time data and department resource state data for the multi-agent decision module;

[0014] The visual output module is used for receiving disaster situation, decision scheme and effect evaluation result from the multi-agent decision module, and receiving knowledge ontology structure data from the domain knowledge ontology module, and realizing data visual display.

[0015] As a further scheme of the application, the domain knowledge ontology module adopts a three-level architecture of a mode layer, a task layer and a data layer, is constructed based on an OWL language, and is stored in an ontology database composed of Protégé+MySQL;

[0016] The mode layer is used for defining core concepts and relationships in the rainstorm flood disaster emergency management field, and the relationships between concepts are associated through Ad-hoc binary relationships;

[0017] The task layer is used for defining emergency decision tasks and standardized processes;

[0018] The data layer establishes a mapping relationship between mode layer concepts and data based on integrated rainstorm flood disaster emergency management knowledge, adopts a ternary group storage of 'concept-attribute-data field', supports real-time data updating, and automatically updates attribute values of corresponding concepts when new data is pushed by external data sources.

[0019] As a further scheme of the application, the multi-agent decision module includes a collaborative scheduling agent, a real-time sensing agent, a diagnosis and evaluation agent, a decision response agent and a learning optimization agent;

[0020] The collaborative scheduling agent manages the interaction, task allocation and conflict resolution among other agents, triggers other agents according to the emergency situation, interfaces with multi-department business systems to obtain department resource states, combines disaster data and preliminary analysis of the real-time sensing agent to trigger the diagnosis and evaluation agent and the decision response agent, receives initial emergency decision and response action scheme and pushes it to the learning optimization agent, and updates it to the visual output module at the same time;

[0021] The real-time sensing agent collects flood-related data from sensors, satellite images, meteorological data, and social media according to the schedule set by the collaborative scheduling agent, performs data preprocessing, and makes a preliminary assessment of the current disaster situation in combination with historical data. When a potential disaster event is detected, it sends information to the collaborative scheduling agent, pushes standardized data to the diagnostic assessment agent, and stores it in the system's local database.

[0022] After detecting a potential disaster scenario, the diagnostic assessment agent triggers a simulation and prediction of the physical processes of the water disaster system. It utilizes data provided by the real-time sensing agent to drive a hydrological and hydrodynamic model to simulate and extrapolate flood processes, supporting the risk assessment model in analyzing the hazard level of the disaster-affected area, the vulnerability of different disaster-bearing entities, and the overall regional flood risk. The comprehensive flood risk index formula is adopted:

[0023]

[0024] Calculate the comprehensive flood risk index, identify and predict the flood disaster risk level in the affected area, receive emergency plans from the decision response agent, and assess the evolution of disaster risk and emergency situation;

[0025] in, Represents a comprehensive flood risk index. This represents the total number of grid cells within the disaster-stricken area covered by the risk assessment model. The identifier representing a grid cell at a specific spatial location. Represents grid cell The weight of the disaster-causing factors Represents grid cell The intensity values ​​of the comprehensive disaster-causing factors derived from the hydrological and hydrodynamic model are as follows. This represents a preset risk threshold for disaster-causing factors. This represents the maximum reference value for the combined intensity of potential disaster-causing factors within the study area. Represents grid cell The vulnerability weight of disaster-bearing bodies This represents the total number of asset types included in the vulnerability assessment of the disaster-bearing body. Identifiers representing specific asset types Representing the The importance weight of asset classes Represents grid cell Inner Vulnerability scores for asset classes Represents grid cell Risk evolution adjustment coefficient;

[0026] The decision response agent receives current flood state data, introduces preset rule-based reasoning (RBR) and case-based reasoning (CBR) methods, matches disaster scenes based on the "data-concept" mapping of the domain knowledge ontology, issues flood disaster early warning and pushes contacts, field experts, department heads and mass media, generates an initial decision scheme in combination with the emergency task flow of the task layer, starts the agent cooperation mechanism to request other agents to assist in formulating a suggestion scheme for reducing flood disaster risk, generates an emergency response and action scheme by using the iterative analysis strategy of "proposing-criticizing-improving", and sends it to the visualization output module.

[0027] The learning optimization agent evaluates the current scheme effect based on preset evaluation indicators combined with the effect benchmark value of the historical case library, analyzes the reasons when the scheme effect is not up to standard, adjusts the scheme based on the domain knowledge ontology and pushes it to the decision response agent to update the scheme, and stores the scene, scheme and execution effect of this disaster to the historical case library.

[0028] As a further scheme of the application, the data interaction module comprises a data interface layer and an authority control layer.

[0029] The external data interface of the data interface layer supports interaction with the monitoring data of the meteorological department, the water conservancy department and the traffic department, the department business system interface is connected with the reservoir gate pump scheduling system, the road traffic monitoring system and the shelter management system, and supports bidirectional data interaction.

[0030] As a further scheme of the application, the visualization output module is developed based on WebGIS and deployed on the browser end, supporting PC and mobile end access.

[0031] The visualization output function includes dynamic risk map display, disaster situation display, decision scheme display and execution effect display, and the decision scheme display and execution effect display include evacuation routes, resource allocation paths and detailed attribute information thereof.

[0032] The visualization output module supports interactive functions.

[0033] As a further scheme of the application, the active learning process of the learning optimization agent comprises data collection, model training and model updating.

[0034] Data collection collects scene, scheme and execution effect data of this disaster, model training trains the built-in model of the system using the collected data, and model updating updates the trained model parameters to the system to enhance the adaptability and prediction ability of the system and improve the accuracy of future emergency decision-making.

[0035] As a further scheme of the application, the data preprocessing process of the real-time perception agent comprises data cleaning, data conversion and data fusion.

[0036] Data cleaning removes outliers and missing values, data conversion uniformly converts data of different formats into a preset format, and data fusion integrates multi-source data to eliminate data redundancy and conflicts.

[0037] As a further scheme of the present application, the hydrology and hydrodynamic model in the diagnosis evaluation agent adopts a MIKE11 or HEC-HMS model.

[0038] The risk assessment model is constructed based on an analytic hierarchy process or a fuzzy comprehensive evaluation method, and ensures the accuracy of the flood process simulation and deduction and the reliability of the disaster risk assessment.

[0039] As a further scheme of the present application, the RBR in the decision response agent constructs a rule base based on laws and regulations, technical standards and emergency plans in the field of rainstorm flood disaster emergency management, and the CBR retrieves similar cases from a historical case base through a similarity algorithm, to provide a reference for the generation of an initial decision scheme.

[0040] In another aspect, the present application also provides a rainstorm flood disaster emergency decision method based on a domain knowledge ontology, which adopts the above system, and the method comprises the following steps:

[0041] S1, domain knowledge ontology construction and update: the domain knowledge ontology module adopts an OWL language to construct a rainstorm flood disaster emergency management domain knowledge ontology with a three-level architecture of a mode layer, a task layer and a data layer, the data layer establishes a mapping relationship between the mode layer concepts and data based on integrated emergency management knowledge, and when new data is pushed by an external data source, the attribute values of the corresponding concepts are automatically updated;

[0042] S2, multi-source data acquisition and interaction: the data interaction module acquires real-time data from sensors, satellite images, meteorological departments, water conservancy departments and traffic departments through external data interfaces, acquires department resource state data through department business system interfaces to department business systems, and after role permission verification based on a permission control layer, transmits the data to the multi-agent decision module;

[0043] S3, multi-agent collaborative decision: the collaborative scheduling agent triggers the real-time sensing agent according to the data transmitted by the data interaction module, the real-time sensing agent pushes the preprocessed collected data to the diagnosis evaluation agent after preprocessing the collected data; the diagnosis evaluation agent drives the hydrology and hydrodynamic model to deduce the flood process, generates a disaster risk assessment result in combination with the risk assessment model, and feeds back to the collaborative scheduling agent; the collaborative scheduling agent triggers the decision response agent, the decision response agent generates an emergency response and action scheme based on the data-concept mapping of the domain knowledge ontology in combination with the RBR and CBR methods;

[0044] S4, scheme evaluation and optimization: the learning optimization agent evaluates the scheme effect generated by the decision response agent based on the preset evaluation index and the benchmark value of the historical case library, and when the scheme effect is not up to standard, adjusts the scheme based on the domain knowledge ontology and pushes it to the decision response agent for updating, and stores the disaster scene, scheme and execution effect of this time to the historical case library;

[0045] S5, visual display and interaction: the visual output module receives the disaster situation, decision scheme, effect evaluation result of the multi-agent decision module and the knowledge ontology structure data of the domain knowledge ontology module, realizes dynamic display based on WebGIS, and supports emergency management personnel to modify the scheme parameters and synchronously generates a new scheme to the decision response agent.

[0046] Compared with the prior art, the technical scheme provided by the application has at least the following beneficial effects:

[0047] Solve the problem of knowledge fragmentation and realize the correlation and reuse of multiple information: by constructing the domain knowledge ontology of the three-level architecture of "mode layer-task layer-data layer", the dispersed geographic information, environmental monitoring data, disaster cases, emergency schemes and other knowledge are structured and integrated, the Ad-hoc binary relationship between concepts is defined based on OWL language, the association chain of "data-concept-task" is formed, the "data island" is broken, the knowledge is quickly matched and reused, and the decision efficiency is improved.

[0048] Improve the decision response timeliness and grasp the best disposal window: a multi-agent collaborative decision framework is adopted, each agent has clear division of labor and real-time interaction, manual intervention links are reduced, real-time automatic data collection and preprocessing of the agent, rapid deduction of the diagnosis and evaluation agent, and rule-based scheme generation of the decision response agent, so that the whole decision response process is short in time, compared with the existing system, the response time is greatly shortened, and the disaster loss is reduced.

[0049] Strengthen the multi-department collaboration capability and improve the decision efficiency: the data interaction module realizes two-way data interaction with meteorological, water conservancy, traffic, civil affairs and other business systems through standardized interfaces, the permission control layer ensures that each department only accesses the data within the permission range, which guarantees data security and realizes information sharing; the collaborative scheduling agent of the multi-agent decision module uniformly coordinates the agents related to each department, realizes multi-department collaboration in the whole process of disaster monitoring, early warning, disposal and evaluation, when formulating a personnel evacuation scheme, the system can simultaneously obtain flood prediction data of the water conservancy department and road traffic data of the traffic department, ensure that the scheme takes into account the flood risk and traffic feasibility, and avoid the low efficiency problem of the existing system due to the lack of information between departments.

[0050] Enhance the dynamic adaptability of decision-making and improve the scientificity of the scheme: the learning optimization agent evaluates the scheme execution effect in real time based on the preset evaluation index and the historical case benchmark value, and dynamically adjusts, and at the same time optimizes the model parameters through active learning, so that the system can adapt to the real-time changes of rainstorm flood disasters, avoid the defects that the existing system can only execute the preset process and cannot dynamically respond to unexpected situations, and ensure the scientificity and adaptability of the emergency decision-making scheme. BRIEF DESCRIPTION OF DRAWINGS

[0051] In order to more clearly illustrate the technical solutions in the specific embodiments or prior art of the present application, the drawings needed in the specific embodiment or prior art description will be briefly introduced as follows. Obviously, the drawings in the following description are some embodiments of the present application, and other drawings can also be obtained by those skilled in the art without creative labor.

[0052] Fig. 1 The overall architecture diagram of the rainstorm flood disaster emergency decision-making system based on the domain knowledge ontology provided by the embodiments of the present application is shown in the figure.

[0053] Fig. 2 The structure diagram of the domain knowledge ontology module provided by the embodiments of the present application is shown in the figure.

[0054] Fig. 3 The flow chart of the multi-agent decision-making module cooperatively supporting the emergency decision-making process provided by the embodiments of the present application is shown in the figure. DETAILED DESCRIPTION

[0055] The technical solutions of the present application will be described in detail below with reference to the drawings. Obviously, the described embodiments are part of the embodiments of the present application, not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0056] The present application will be further explained and described below in conjunction with specific embodiments.

[0057] Please refer to Figs. 1-3 The embodiments of the present application provide a rainstorm flood disaster emergency decision-making system based on domain knowledge ontology. The system realizes the intelligentization of the whole process of rainstorm flood disaster emergency decision-making through the double-layer architecture of "domain knowledge ontology + multi-agent cooperation" in flood disaster emergency management, combined with data interaction and visual output. The system comprises:

[0058] The domain knowledge ontology module, the multi-agent decision-making module, the data interaction module and the visual output module.

[0059] Among them, the domain knowledge ontology module is connected with the multi-agent decision module through the "knowledge interface" to provide knowledge support for multi-agent collaborative decision-making, and the multi-agent module communicates information among the agents through the communication protocol. The data interaction module is connected with the external data source and the multi-department business system, and realizes the permission control of data access with the help of permission control. The data interaction module is connected with the multi-agent decision module through the "data transmission channel" to provide real-time data and department resource state data. The visual output module receives the disaster situation, decision scheme, effect evaluation and other results from the multi-agent module and the knowledge ontology structure data from the domain knowledge ontology module, and realizes the visual display of data.

[0060] The domain knowledge ontology module adopts a three-level architecture of "pattern layer-task layer-data layer", is constructed based on OWL (Web Ontology Language) language, and is stored in the ontology database (implemented by Protégé+MySQL). The pattern layer defines the core concepts and relationships in the field of rainstorm flood disaster emergency management, such as disaster element class, department responsibility class, geographic space class, secondary disaster class, etc. The relationship between concepts is associated through language-defined Ad-hoc binary relationship. The task layer defines the key emergency decision-making tasks and standardized processes, which are stored in the form of "trigger condition-execution step-output result-restriction condition" quadruple, such as disaster warning task, personnel evacuation task, rescue force deployment task, etc. Based on the integrated rainstorm flood disaster emergency management knowledge (such as rescue technology guide, disaster case, emergency plan and disposal experience, etc.), the data layer establishes the mapping relationship between the pattern layer concepts and data, which is stored in the form of "concept-attribute-data field" triple, supports real-time data update, and automatically updates the attribute value of the corresponding concept when new data is pushed by the external data source.

[0061] The multi-agent decision module: This module contains 5 types of agents, namely collaborative scheduling agent, real-time sensing agent, diagnosis and evaluation agent, decision response agent and learning optimization agent. Multiple agent systems evolve collaboratively to support emergency decision-making, simulating the self-organizing behavior of complex emergency decision-making systems.

[0062] ①The collaborative scheduling agent is the administrator of the multi-agent module, which manages the interaction, task allocation and conflict resolution among other agents to ensure the overall consistency of the system. According to the emergency situation, it timely triggers other agents, and is responsible for the start and task allocation of other agents. Through the data interaction module, it queries and obtains the department resource state from the multi-department business system, and based on the domain knowledge ontology, combined with the disaster data and preliminary analysis of the real-time sensing agent, it timely triggers the diagnosis and evaluation agent and the decision response agent, receives the initial scheme of emergency decision and response action, and pushes it to the learning optimization agent, and updates it to the visual output module.

[0063] ②Real-time sensing agent collects flood-related data (rainfall, weather forecast, river water level, ground water, reservoir operation and geographic information, etc.) in real time from sensors, satellite images, weather data and social media according to the schedule set by the collaborative scheduling agent, pre-processes the data, and preliminarily assesses the current disaster situation in combination with historical data. When a potential disaster event is detected, the collaborative scheduling agent is informed, and standardized data (such as JSON format) is pushed to the diagnostic assessment agent, and stored in the local database of the system.

[0064] ③After detecting a potential disaster scenario, the water disaster system triggers physical process simulation prediction behavior, uses the data provided by the real-time sensing agent to drive the hydrological and hydrodynamic model to carry out flood process simulation and deduction, supports the risk assessment model to analyze the hazard level of the disaster-affected area and the vulnerability degree of different disaster-bearing bodies, and the regional comprehensive flood risk. The formula of comprehensive flood risk index is adopted:

[0065]

[0066] The comprehensive flood risk index is calculated to identify and estimate the flood disaster risk level of the disaster-affected area, receive the emergency plan of the decision response agent, and evaluate the disaster risk evolution and emergency situation;

[0067] Wherein, represents the comprehensive flood risk index, represents the total number of grid cells in the disaster-affected area covered by the risk assessment model, represents the identification of the grid cell at a specific spatial location, represents the disaster-causing factor weight of the grid cell , represents the comprehensive disaster-causing factor intensity value derived by the hydrological and hydrodynamic model in the grid cell , represents the preset disaster-causing factor risk threshold, represents the maximum reference value of the comprehensive disaster-causing factor intensity that may occur in the study area, represents the disaster-bearing body vulnerability weight of the grid cell , represents the total number of asset types included in the disaster-bearing body vulnerability assessment, represents the identification of a specific asset type, represents the importance weight of the th asset type, represents the vulnerability score of the th asset type in the grid cell , represents the risk evolution adjustment coefficient of the grid cell .

[0068] After the diagnosis evaluation agent detects a potential disaster scenario, for example, it monitors the upstream basin and finds that the cumulative rainfall in 6 hours reaches 150 mm, triggering the water disaster system physical process simulation prediction behavior, which first uses the data provided by the real-time perception agent to drive the hydrological and hydrodynamic model, specifically, the model calls the river water level sensor data deployed at three locations A, B, and C in the region, with values of 5.2 m, 5.5 m, and 5.1 m, respectively, and the flow rate sensor data with values of 1.5 m / s, 1.8 m / s, and 1.4 m / s, respectively, while combining the data of 50 mm of rainfall in the next 3 hours obtained from the meteorological system, to carry out flood process simulation deduction. The research area is divided into 1kmx1km grid cells, and a total of 3 cells are taken as an example (( ), by numerically solving the two-dimensional shallow water equation, the water depth and flow rate of each grid cell at different time steps in the future are iteratively calculated, for example, it is calculated that after 1 hour, the comprehensive disaster-causing factor intensity values of grid cells 1, 2, and 3 are 4.5, 5.8, and 3.9, respectively, which are obtained by weighted summation of the simulated water depth (unit: m) and flow rate (unit: m / s) of the cell. The risk assessment model uses the comprehensive flood risk index formula:

[0069] ;

[0070] To calculate, first determine the parameters in the formula, the disaster-causing factor risk threshold is set in reference to historical flood event data, and the average disaster-causing factor intensity of 5 flood events in the past 10 years that triggered mild disasters (such as a small amount of farmland being flooded, no casualties) is selected as the benchmark, and is calculated, while the maximum reference value of the comprehensive disaster-causing factor intensity that may occur in the research area is based on the simulation results of a 100-year flood scenario, and is set to The disaster-causing factor weight and the vulnerability weight of each grid cell are determined by expert scoring combined with the analytic hierarchy process, ensuring that the sum of the two is 1. For grid cell 1 (mainly industrial area), set , For grid cell 2 (mainly residential area), set , For grid cell 3 (mainly suburban farmland), set , The total number of asset types included in the vulnerability assessment of the disaster-affected body , are the importance weights of building facilities (j = 1) and transportation network (j = 2) Based on their roles in the socio-economic system, the mean values are calculated by statistical analysis of the questionnaire data from 20 experts in the field, , , the vulnerability scores of each asset in each grid cell According to their specific characteristics, the quantification standards are shown in Table 1:

[0071] Table 1 Quantification table of asset vulnerability score

[0072]

[0073] As shown in Table 1, the assets in each grid cell are evaluated. In grid cell 1, the main assets are reinforced concrete structure workshops and main roads, with an asset vulnerability score of , In grid cell 2, the main assets are brick and concrete structure residential buildings and secondary roads, with an asset vulnerability score of , In grid cell 3, the main assets are simple structure greenhouses and branch roads, with an asset vulnerability score of , The setting of the risk evolution adjustment coefficient is based on the assessment of the emergency response capacity of each grid cell, including the coverage rate of early warning information, the perfect rate of emergency evacuation channels, and the amount of emergency supplies. After standardization and weighted summation, the result is obtained. For example, grid cell 1 has a complete early warning system and material reserves, with an Grid cell 2 has a blind area in early warning coverage, with an Grid cell 3 is short of emergency resources, with an Then, the actual values of all the above parameters are substituted into the formula for calculation. For grid cell 1 ( ):

[0074] ;

[0075] ;

[0076] For grid cell 2 ( ):

[0077] ;

[0078] For grid cell 3 ( ):

[0079] ;

[0080] The final comprehensive flood risk index for:

[0081] ;

[0082] The calculation logic of this formula lies in first performing absolute value operations. Obtain the intensity of the disaster-causing factor Deviation from risk threshold The degree, divided by the maximum reference value. Normalization is performed, and the square root operation is used to adjust the sensitivity to intensity differences, which is then compared with the hazard factor weights. Multiplying yields the disaster risk component, while weighted summation is performed. Calculate the weighted vulnerability score for each asset type and divide by the total number of asset types. The average vulnerability is obtained, and then it is compared with the vulnerability weight of the disaster-bearing entity. Multiplying these two components yields the vulnerability risk component. Adding these two risk components together and multiplying by a risk evolution adjustment coefficient based on emergency response capabilities... Finally, by summing and taking the average The comprehensive risk index for the entire region was obtained, and the comprehensive flood risk index was calculated to be 19.41. This index is used to identify and predict the flood disaster risk level in the affected area. If the risk levels are divided into low risk (0-10), medium risk (10-20), and high risk (above 20), the current calculation result of 19.41 indicates that the region as a whole is at a medium risk level. Among them, grid cells 2 and 3 have higher risk contribution values, which provides a quantitative basis for the risk assessment model. This allows the model to receive emergency plans from the decision-making response agent and assess the evolution of disaster risk and emergency situation. The advantage of the formula lies in the introduction of a risk evolution adjustment coefficient. This method couples static disaster-causing factors and disaster-bearing body assessments with dynamic emergency response capabilities, and uses absolute values ​​and square root operations to process the intensity of disaster-causing factors, thus refining the expression of risk differences; among them, Represents a comprehensive flood risk index. This represents the total number of grid cells within the disaster-stricken area covered by the risk assessment model. The identifier representing a grid cell at a specific spatial location. Represents grid cell The weight of the disaster-causing factors Represents grid cell The intensity values ​​of the comprehensive disaster-causing factors derived from the hydrological and hydrodynamic model are as follows. This represents a preset risk threshold for disaster-causing factors. This represents the maximum reference value for the combined intensity of potential disaster-causing factors within the study area. Represents grid cell The vulnerability weight of disaster-bearing bodies This represents the total number of asset types included in the vulnerability assessment of the disaster-bearing body. Identifiers representing specific asset types Representing the The importance weight of asset classes Represents grid cell Inner Vulnerability scores for asset classes Represents grid cell Risk evolution adjustment coefficient.

[0083] ④ The decision-response agent receives current flood status data, introduces pre-defined rule-based reasoning (RBR) and case-based reasoning (CBR) methods, matches disaster scenarios based on the "data-concept" mapping of domain knowledge ontology, issues flood disaster warnings, and pushes information to contacts, domain experts, department heads, and mass media according to pre-defined rules. Based on the emergency task process at the task layer, it generates an initial decision plan in conjunction with the scenario, and simultaneously activates the agent cooperation mechanism to request other agents to assist in formulating suggested plans to reduce flood disaster risks, such as the coordinated operation of reservoir gates and pumps and protective actions (such as warnings to people at disaster sites, personnel evacuation procedures, and pre-positioning of rescue forces). The generated suggested plans adopt an iterative analysis method, following a "propose-critique-improve" strategy, and finally generate a satisfactory emergency response and action plan, which is then sent to the visualization output module.

[0084] ⑤ The learning and optimization agent evaluates the effectiveness of the current plan based on preset evaluation indicators (such as evacuation accuracy, evacuation completion rate, resource arrival rate, and secondary disaster incidence rate) and the benchmark values ​​in the historical case library. If the effectiveness is not up to standard, the reasons are analyzed, the plan is adjusted based on the domain knowledge ontology, and the updated plan is pushed to the decision-response agent. The scenario, plan, and execution effect of this disaster are stored in the historical case library. In this way, the system's adaptability and predictive ability are enhanced through active learning, continuously learning from disaster events, optimizing model parameters, and improving the accuracy of future responses.

[0085] Data Interaction Module: This module includes a data interface layer and an access control layer. For the data interface layer, external data interfaces support data exchange with monitoring data from meteorological, water resources, and transportation departments; departmental business system interfaces connect to business systems such as reservoir gate and pump scheduling systems, road traffic monitoring systems, and shelter management systems, and support bidirectional data exchange. For the access control layer, different roles are defined, including system administrators (who can access all data), department operators (who can only access data within their own department), and viewing users (who can only view public data such as early warning information). A "role-data permission" mapping relationship is stored in the database. When a user / system accesses data, the interface automatically verifies permissions and rejects unauthorized access.

[0086] Visualization Output Module: This module is developed based on WebGIS and deployed on the browser side, supporting access from PC and mobile devices. The visualization output functions include dynamic risk maps and disaster situation display, decision-making schemes and execution effects display (such as evacuation routes and resource allocation paths and their detailed attribute information), and support interactive functions. For example, emergency managers can modify the scheme parameters by clicking the mouse, and the modifications are automatically synchronized to the response action agent to generate a new scheme.

[0087] Furthermore, this embodiment of the invention also provides an emergency decision-making method for rainstorm and flood disasters based on domain knowledge ontology. Using the above-described system, the method includes the following steps:

[0088] S1. Domain Knowledge Ontology Construction and Update: The domain knowledge ontology module uses the OWL language to construct a three-level architecture of rainstorm and flood disaster emergency management domain knowledge ontology, consisting of a schema layer, a task layer, and a data layer. The data layer establishes a mapping relationship between schema layer concepts and data based on integrated emergency management knowledge. When new data is pushed from external data sources, the attribute values ​​of the corresponding concepts are automatically updated.

[0089] S2. Multi-source data acquisition and interaction: The data interaction module collects real-time data from sensors, satellite images, meteorological departments, water conservancy departments, and transportation departments through external data interfaces. It also connects to multiple departmental business systems through departmental business system interfaces to obtain departmental resource status data. After role and permission verification based on the permission control layer, the data is transmitted to the multi-agent decision-making module.

[0090] S3. Multi-agent collaborative decision-making: The collaborative scheduling agent triggers the real-time sensing agent based on the data transmitted by the data interaction module. The real-time sensing agent preprocesses the collected data and pushes it to the diagnostic assessment agent. The diagnostic assessment agent drives the hydrological and hydrodynamic model to deduce the flood process, combines the risk assessment model to generate disaster risk assessment results and feeds them back to the collaborative scheduling agent. The collaborative scheduling agent triggers the decision response agent, which generates emergency response and action plans based on the data-concept mapping of the domain knowledge ontology and combines RBR and CBR methods.

[0091] S4. Solution Evaluation and Optimization: The learning optimization agent evaluates the effectiveness of the solutions generated by the decision response agent based on preset evaluation indicators and historical case library benchmarks. When the solution effectiveness does not meet the standards, the solution is adjusted based on the domain knowledge ontology and pushed to the decision response agent for update. At the same time, the disaster scenario, solution and execution effect are stored in the historical case library.

[0092] S5, visual display and interaction: the visual output module receives the disaster situation, decision scheme, effect evaluation result of the multi-agent decision module and the knowledge ontology structure data of the domain knowledge ontology module, realizes dynamic display based on WebGIS, supports emergency management personnel to modify scheme parameters and synchronizes to the decision response agent to generate a new scheme.

[0093] Compared with the prior art, the present application has the following advantages:

[0094] Solve the problem of knowledge fragmentation and realize the correlation and reuse of multiple information: by constructing the domain knowledge ontology of the three-level architecture of "mode layer-task layer-data layer", the dispersed geographic information, environmental monitoring data, disaster cases, emergency schemes and other knowledge are structured and integrated, the Ad-hoc binary relationship between concepts is defined based on OWL language, the correlation chain of "data-concept-task" is formed, the "data island" is broken, the rapid matching and reuse of knowledge are realized, and the decision efficiency is improved.

[0095] Improve the decision response timeliness and grasp the best disposal window: a multi-agent collaborative decision framework is adopted, each agent has clear division of labor and real-time interaction, manual intervention links are reduced, automatic data collection and preprocessing of the agent, rapid deduction of the diagnosis evaluation agent model, and rule-based scheme generation of the decision response agent, so that the whole decision response process is short in time, compared with the existing system, the response time is greatly shortened, and the disaster loss is reduced.

[0096] Strengthen the multi-department collaboration ability and improve the decision efficiency: the data interaction module realizes two-way data interaction with meteorological, water conservancy, traffic, civil affairs and other business systems through standardized interface, the permission control layer ensures that each department only accesses the data within the permission range, which guarantees data security and realizes information sharing; the collaborative scheduling agent of the multi-agent decision module uniformly coordinates the related agents of each department, realizes the multi-department collaboration of the whole process of disaster monitoring, early warning, disposal and evaluation, when formulating the personnel evacuation scheme, the system can simultaneously obtain the flood prediction data of the water conservancy department and the road traffic data of the traffic department, ensure that the scheme takes into account the flood risk and traffic feasibility, and avoid the low efficiency problem of the existing system due to the lack of information between departments.

[0097] Enhance the dynamic adaptability of decision and improve the scientific nature of scheme: the learning optimization agent evaluates the scheme execution effect in real time based on the preset evaluation index and historical case benchmark value and dynamically adjusts, and simultaneously optimizes the model parameters through active learning, so that the system can adapt to the real-time changes of rainstorm flood disaster, avoid the defects of the existing system that can only execute the preset process and cannot dynamically respond to sudden situations, and guarantee the scientific nature and adaptability of the emergency decision scheme.

[0098] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, and are not intended to limit the present application; although the present application has been described in detail with reference to the above embodiments, those skilled in the art should understand that the technical solutions recorded in the above embodiments can be modified, or some or all of the technical features can be replaced by equivalents; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the present application.

Claims

1. A rainstorm and flood disaster emergency decision system based on domain knowledge ontology, characterized in that, The system includes: a domain knowledge ontology module, a multi-agent decision-making module, a data interaction module, and a visualization output module; The domain knowledge ontology module is connected to the multi-agent decision-making module through a knowledge interface, providing knowledge support for collaborative decision-making among multiple agents in the multi-agent decision-making module. In the multi-agent decision-making module, each agent transmits information through a communication protocol. The data interaction module is connected to external data sources and multi-department business systems respectively, and the data interaction module is connected to the multi-agent decision-making module through a data transmission channel to provide the multi-agent decision-making module with real-time data and departmental resource status data; The visualization output module is used to receive disaster situation, decision-making scheme and effect evaluation results from the multi-agent decision-making module, as well as knowledge ontology structure data from the domain knowledge ontology module, to realize data visualization display; The multi-agent decision-making module includes a collaborative scheduling agent, a real-time perception agent, a diagnostic evaluation agent, a decision response agent, and a learning optimization agent. The collaborative scheduling agent manages the interaction, task allocation and conflict resolution among other agents, triggers other agents according to the emergency situation, connects with multi-department business systems to obtain departmental resource status, and triggers diagnostic assessment agents and decision response agents by combining disaster data from real-time sensing agents and preliminary analysis. It receives initial emergency decisions and response action plans and pushes them to the learning and optimization agent, while updating them to the visualization output module. The real-time sensing agent collects flood-related data from sensors, satellite images, meteorological agencies, and social media according to the schedule set by the collaborative scheduling agent, performs data preprocessing, and makes a preliminary assessment of the current disaster situation by combining historical data. When a potential disaster event is detected, it sends information to the collaborative scheduling agent and pushes standardized data to the diagnostic assessment agent and stores it in the system's local database. After detecting a potential disaster scenario, the diagnostic assessment agent triggers a simulation and prediction of the physical processes of the water disaster system. It utilizes data provided by the real-time sensing agent to drive a hydrological and hydrodynamic model to simulate and extrapolate flood processes, supporting the risk assessment model in analyzing the hazard level of the disaster-affected area, the vulnerability of different disaster-bearing entities, and the overall regional flood risk. The comprehensive flood risk index formula is used: ； Calculate the comprehensive flood risk index, identify and predict the flood disaster risk level in the affected area, receive emergency plans from the decision response agent, and assess the evolution of disaster risk and emergency situation; in, Represents a comprehensive flood risk index. This represents the total number of grid cells within the disaster-stricken area covered by the risk assessment model. The identifier representing a grid cell at a specific spatial location. Represents grid cell The weight of the disaster-causing factors Represents grid cell The intensity values ​​of the comprehensive disaster-causing factors derived from the hydrological and hydrodynamic model are as follows. This represents a preset risk threshold for disaster-causing factors. This represents the maximum reference value for the combined intensity of potential disaster-causing factors within the study area. Represents grid cell The vulnerability weight of disaster-bearing bodies This represents the total number of asset types included in the vulnerability assessment of the disaster-bearing body. Identifiers representing specific asset types Representing the The importance weight of asset classes Represents grid cell Inner Vulnerability scores for asset classes Represents grid cell Risk evolution adjustment coefficient; The decision-response agent receives current flood status data, introduces preset rule reasoning and case reasoning methods, matches disaster scenarios based on the "data-concept" mapping of domain knowledge ontology, issues flood disaster warnings and pushes them to contacts, domain experts, department heads and mass media, generates an initial decision plan in combination with the emergency task process of the task layer, initiates the agent cooperation mechanism to request other agents to assist in formulating suggested plans to reduce flood disaster risks, and generates emergency response and action plans by adopting an iterative analysis strategy of "propose-critique-improve", which is then sent to the visualization output module. The learning optimization agent evaluates the current scheme effect based on the preset evaluation index and the effect benchmark value of the historical case library, analyzes the reasons when the scheme effect is not up to standard, adjusts the scheme based on the domain knowledge ontology, and pushes the updated scheme to the decision response agent, stores the scene, scheme and execution effect of the disaster to the historical case library.

2. The system of claim 1, wherein, The domain knowledge ontology module adopts a three-level architecture of pattern layer, task layer and data layer, is constructed based on OWL language, and is stored in an ontology database composed of Protégé+MySQL; The pattern layer is used to define the core concepts and relationships in the field of rainstorm flood disaster emergency management, and the relationships between concepts are associated through Ad-hoc binary relationship; The task layer is used to define the emergency decision tasks and standardized processes; The data layer establishes the mapping relationship between the pattern layer concepts and data based on the integrated rainstorm flood disaster emergency management knowledge, stores in the form of "concept-attribute-data field" triple, supports real-time data update, and automatically updates the attribute values of the corresponding concepts when new data is pushed by external data sources.

3. The system of claim 1, wherein, The data interaction module includes a data interface layer and a permission control layer; The external data interface of the data interface layer supports interaction with monitoring data of meteorological departments, water conservancy departments and traffic departments, and the department business system interface is connected with the reservoir gate pump dispatching system, the road traffic monitoring system and the shelter management system, and supports bidirectional data interaction.

4. The system of claim 1, wherein, The visualization output module is developed based on WebGIS and deployed on the browser side, supporting PC and mobile access; The visualization output function includes dynamic risk map display, disaster situation display, decision scheme display and execution effect display, and the decision scheme display and execution effect display include evacuation routes, resource allocation paths and detailed attribute information; The visualization output module supports interactive functions.

5. The system of claim 2, wherein, The active learning process of the learning optimization agent includes data collection, model training and model updating; Data collection collects scene, scheme and execution effect data of the disaster, model training trains the built-in model of the system using the collected data, and model updating updates the trained model parameters to the system to enhance the adaptability and prediction ability of the system and improve the accuracy of future emergency decision-making.

6. The system of claim 2, wherein, The data preprocessing process of the real-time perception agent includes data cleaning, data conversion and data fusion; Data cleaning removes outliers and missing values, data conversion converts different formats of data into a preset format, and data fusion integrates multi-source data to eliminate data redundancy and conflicts.

7. The system of claim 2, wherein, The hydrology and hydrodynamics model in the diagnosis and evaluation agent adopts MIKE11 or HEC-HMS model; The risk assessment model is constructed based on the analytic hierarchy process or fuzzy comprehensive evaluation method, ensuring the accuracy of the flood process simulation and the reliability of the disaster risk assessment.

8. The system of claim 2, wherein, The RBR in the decision response agent constructs a rule base based on laws, regulations, technical standards and emergency plans in the field of rainstorm flood disaster emergency management, and the CBR retrieves similar cases from the historical case library through a similarity algorithm to provide a reference for the generation of the initial decision scheme.

9. A storm flood disaster emergency decision-making method based on a domain knowledge ontology, adopting the system of any one of claims 1-8, characterized in that, The method comprises the following steps: S1, domain knowledge ontology construction and update: through the domain knowledge ontology module, the rainstorm flood disaster emergency management domain knowledge ontology with three-level architecture of mode layer, task layer and data layer is constructed by using OWL language. The mapping relationship between the concepts of mode layer and data is established based on the integrated emergency management knowledge. When new data is pushed by external data source, the attribute value of corresponding concept is automatically updated; S2, multi-source data collection and interaction: the data interaction module collects real-time data from sensors, satellite images, meteorological departments, water conservancy departments and transportation departments through external data interface. The department resource state data is obtained by connecting multi-department business systems through department business system interface. After the role permission verification based on the permission control layer, the data is transmitted to the multi-agent decision module; S3, multi-agent collaborative decision: the real-time sensing agent is triggered by the collaborative scheduling agent according to the data transmitted by the data interaction module. The real-time sensing agent pushes the preprocessed collected data to the diagnosis and evaluation agent. The diagnosis and evaluation agent drives the hydrological and hydrodynamic model to deduce the flood process, generates disaster risk assessment results combined with risk assessment model and feeds back to the collaborative scheduling agent. The decision response agent is triggered by the collaborative scheduling agent. The decision response agent generates emergency response and action plan based on the data-concept mapping of domain knowledge ontology combined with RBR and CBR methods; S4, scheme evaluation and optimization: the learning and optimization agent evaluates the scheme effect generated by the decision response agent based on the preset evaluation index and historical case library benchmark value. When the scheme effect is not up to standard, the scheme is adjusted based on the domain knowledge ontology and pushed to the decision response agent for update. At the same time, the disaster scene, scheme and execution effect of this time are stored in the historical case library; S5, visualization display and interaction: the visualization output module receives the disaster situation, decision scheme, effect evaluation result of multi-agent decision module and knowledge ontology structure data of domain knowledge ontology module, realizes dynamic display based on WebGIS, supports emergency management personnel to modify scheme parameters and synchronizes to decision response agent to generate new scheme.

Images