AI-based smart city emergency event prediction and resource scheduling method

Abstract

The application discloses an AI-based smart city emergency event prediction and resource scheduling method, and relates to the technical field of smart city emergency management.The application realizes fine and accurate resource matching for complex emergency events through the linkage mechanism of event deconstruction, task package generation, resource vectorization and combination optimization. The input complex event is intelligently disassembled into a standardized and parameter-defined task package set by using natural language processing and event graph technology. Each task package defines the required capability dimension and specific demand threshold. Various resources are parsed into multi-dimensional capability vectors containing the same capability dimension. The combination optimization algorithm searches and configures a resource combination for each task package from the integrated resource pool, so that the resource configuration changes from being event-oriented to being task-oriented, and solves the problem of mismatched resource supply and demand and single and rigid scheduling caused by the failure to deconstruct events in the prior art.

Description

Technical Field

[0001] This invention relates to the field of smart city emergency management technology, specifically to an AI-based method for predicting and scheduling emergency events and allocating resources in smart cities. Background Technology

[0002] As urbanization deepens, urban operating systems are becoming increasingly complex. Various sudden and complex emergency events, such as hazardous chemical explosions, major traffic accidents, and natural disasters, are characterized by high frequency, rapid chain reactions, and significant challenges in handling. Smart city construction takes emergency management as a core scenario, aiming to enhance urban safety resilience through information technology and intelligent means. Traditional emergency management models face severe challenges in response speed, resource allocation accuracy, and cross-departmental collaboration efficiency when dealing with complex events involving multiple coupled disasters and concurrent tasks, making it difficult to meet the urgent needs of modern cities for efficient emergency response.

[0003] In existing technologies, emergency resource dispatching largely relies on manual decision-making based on fixed plans or simple rule-matching systems, which have significant shortcomings:

[0004] The scheduling model is rigid and monotonous. Existing methods mostly follow a one-to-one mapping model between event type and preset resource list, treating composite events as a whole for extensive resource allocation. They fail to intelligently decompose them into standardized subtasks in parallel or serial manner, resulting in a disconnect between resource investment and refined task requirements, which can easily lead to resource shortages or redundancy.

[0005] Resource capabilities are poorly characterized, with resource descriptions typically remaining at the level of text labels or simple classifications. There is a lack of structured and quantifiable capability dimension modeling, making it difficult to achieve accurate matching between resources and tasks based on multi-dimensional capability vectors.

[0006] The dynamic adaptability is insufficient. Once the contingency plan and scheduling scheme are generated, it is difficult to dynamically and quickly replan and adjust them as the event situation evolves and the resource status changes, and there is a lack of real-time optimization capability.

[0007] In summary, existing emergency resource scheduling methods have significant shortcomings in addressing highly complex and dynamically evolving urban emergency scenarios, particularly in the areas of refined task decomposition, quantitative modeling of resource capabilities, and dynamic adaptability of solutions. Therefore, there is an urgent need for a new intelligent scheduling method that can achieve precise and coordinated matching of tasks and resources to fundamentally improve the overall efficiency of handling complex emergency events. Summary of the Invention

[0008] The purpose of this invention is to overcome the shortcomings of existing technologies and provide an AI-based method for predicting and scheduling emergency events in smart cities. This method achieves refined and precise resource matching for complex emergency events through a linkage mechanism of event deconstruction, task package generation, resource vectorization, and combinatorial optimization. Utilizing natural language processing and event graph technology, the input complex event is intelligently decomposed into a standardized set of task packages with clearly defined parameters. Each task package defines the required capability dimensions and specific demand thresholds. Various resources are parsed into multi-dimensional capability vectors containing the same capability dimensions. The combinatorial optimization algorithm transcends departmental boundaries, searching and assembling a resource combination for each task package from a fused resource pool. This transforms resource allocation from being geared towards fuzzy events to being geared towards precise tasks. It solves the problems of resource supply-demand mismatch and rigid, singular scheduling caused by the failure to deconstruct events in existing technologies. It achieves deep quantification and precise matching of task requirements and resource capabilities, and highlights the complementary capabilities and combinatorial effects between resources.

[0009] To address the aforementioned technical problems, this invention provides the following technical solution: an AI-based method for predicting and scheduling emergency events in smart cities, the specific steps of which are as follows:

[0010] S100. Based on natural language processing and event graph technology, the input complex emergency event description is parsed, the event constituent elements are identified, and the complex emergency event is decomposed into multiple standardized logically related task packages.

[0011] S200: Access the emergency resource database, obtain the static attributes and real-time status data of all available emergency resource units, and construct a multi-dimensional capability vector for each emergency resource unit;

[0012] S300: For each task package generated by S100, iterate through the set of all available emergency resource units, and use a combinatorial optimization algorithm to select and determine an optimal resource combination from the set of all emergency resource units, thus selecting the optimal resource combination for the task package.

[0013] S400. During the emergency response process, the execution progress of each task package and the status changes of each emergency resource unit are monitored in real time. When a change in task requirements, failure of an emergency resource unit, or the emergence of a new derivative task is detected, a rescheduling mechanism is triggered. Depending on the changes, either local rescheduling or global rescheduling is selected to recalculate and update the resource allocation scheme.

[0014] S500, the optimal resource combination calculated for all task packages in S300, and the adjustment scheme in S400, are transformed into specific scheduling instructions and collaborative operation suggestions, which are then issued to the corresponding emergency resource units for execution through the emergency command platform.

[0015] Furthermore, in step S100, the step of decomposing the complex emergency event into multiple task packages is as follows:

[0016] Natural language processing is used to analyze the description information of the complex emergency events to extract event elements, including the core event type, location, hazardous substances involved, confirmed secondary hazards, and affected personnel and facilities.

[0017] The extracted event elements are matched with the pre-built emergency response plan knowledge graph to deduce all the basic task types necessary to handle this complex emergency event, forming a preliminary task list;

[0018] Based on the specific information of the event, specific task parameters are calculated for each task in the preliminary task list. The task parameters include quantified capability requirement thresholds for each dimension, expected completion time calculated based on the severity and location of the event, and inter-task dependencies determined by analyzing the logical relationships between tasks. The parameterized tasks are then used as the task package.

[0019] Each task package defines a task type, a set of required capability dimensions, a threshold for each capability dimension, an expected completion time, and dependencies between tasks.

[0020] Furthermore, in S200, the multidimensional capability vector includes:

[0021] Functional capability dimension: represents the effectiveness of an emergency resource unit in performing a specific emergency function. The value of this functional capability dimension is a normalized interval of [0, 1].

[0022] Level attribute dimension: indicates the qualification certification, protection level or professional rating of the emergency resource unit;

[0023] Dynamic status dimension: Represents the real-time physical status of emergency resource units at the moment of dispatch, including geographical coordinates, current availability, and equipment integrity rate;

[0024] Collaborative attribute dimension: This represents the inherent attributes of an emergency resource unit when it collaborates with other emergency resource units, including the business it belongs to, the protocols and frequency bands supported by the communication equipment, and the collaborative partner identifier formed by historical collaborative records.

[0025] Furthermore, in S300, the combinatorial optimization algorithm employs a multi-objective optimization method to find a set of resource combination schemes such that for each task package, the corresponding objective function vector reaches the Pareto optimal frontier. The objective function vector F(C) is defined as: Where C represents a resource combination to be evaluated. It is the total capability matching degree function of resource combination C. It is the response time function of resource combination C. It is the internal collaboration cost function of resource combination C. Combinatorial optimization algorithms evaluate the cost of different resource combinations. The value is used to search and select the best option.

[0026] Furthermore, in S300, when processing a complex emergency event containing multiple task packages, the following constraints must be met when simultaneously allocating resource combinations to all task packages:

[0027] Resource exclusivity constraint: An emergency resource unit can only be allocated to one task package at any given time;

[0028] Task timing constraints: For task packages with sequential dependencies, the resource scheduling and action start time of subsequent task packages must be later than the expected completion time of their predecessor task packages.

[0029] Furthermore, the total capability matching degree function The calculation method is as follows: ,in, This indicates the task package currently awaiting assignment. Indicates task package For the first The threshold for the required capabilities, where C represents a combination of resources to be evaluated. Representing resource combination Emergency resource units in Indicates emergency resource unit In the Ability values ​​in various ability dimensions Representing resource combination In the The sum of abilities across all dimensions of ability It is the first The preset weights of various capability dimensions in the matching degree calculation, function This is used to ensure that the matching contribution is capped at 1 when the combined capability exceeds the demand.

[0030] Furthermore, the internal collaboration cost function The calculation comprehensively considers the difficulty of coordination between resources, and the calculation method is as follows: ,in, and It is a combination of resources There are two different emergency resource units, Cost(Type(p), Type(p)) is the preset emergency resource unit. and emergency resource units The communication and coordination difficulty coefficient, CoopHist(p, q) is an emergency resource unit calculated based on historical collaboration data. and emergency resource units The level of collaboration between them and These are the weighting coefficients of communication costs and historical collaboration costs in the total collaboration cost, respectively.

[0031] Furthermore, the specific steps of S400 are as follows:

[0032] S410. Triggered when any of the continuously monitored indicators meet the following conditions: the capacity requirement threshold of a single task package changes, the status of the emergency resource unit allocated to the task package becomes unavailable, or a new derivative emergency event occurs.

[0033] S420. When the triggering reason is that the capacity requirement threshold of a single task package changes or the status of the emergency resource unit allocated to the task package becomes unavailable, other unaffected task packages and their allocated resource allocation schemes are locked. Only the affected task packages are re-executed to find a new optimal resource combination.

[0034] S430. When the triggering reason is the occurrence of a new derivative emergency event, the new event description is merged with the original event information and returned as a new input to S100. A new event deconstruction is initiated to generate a new task package set containing all tasks, old and new. Based on the current latest emergency resource unit status, S300 is re-executed on all task packages.

[0035] Furthermore, in S500, the scheduling instructions and collaborative operation suggestions include:

[0036] The dispatch instructions issued for each task package clearly list the identifiers of all emergency resource units in the optimal resource combination allocated to that task package, their destinations, and the expected task content.

[0037] Collaborative operation recommendations generated by emergency resource units requiring coordination include suggested meeting points, joint routes, and designated unified communication channels or frequency bands.

[0038] Compared with existing technologies, this AI-based smart city emergency event prediction and resource scheduling method has the following beneficial effects:

[0039] I. This invention achieves refined and precise resource matching for complex emergency events through a linkage mechanism of event deconstruction, task package generation, resource vectorization, and combinatorial optimization. Utilizing natural language processing and event graph technology, the input complex event is intelligently decomposed into a standardized set of task packages with clearly defined parameters. Each task package defines the required capability dimensions and specific demand thresholds. Various resources are parsed into multi-dimensional capability vectors containing the same capability dimensions. The combinatorial optimization algorithm transcends departmental boundaries, searching and assembling a resource combination for each task package from the integrated resource pool. This transforms resource allocation from being geared towards fuzzy events to being geared towards precise tasks. It solves the problems of resource supply and demand mismatch and rigid scheduling caused by the failure to deconstruct events in existing technologies. It achieves deep quantification and precise matching of task requirements and resource capabilities, and highlights the complementary capabilities and combinatorial effects between resources.

[0040] Second, this invention establishes and optimizes internal collaboration costs, proactively optimizes the collaborative efficiency of resource combinations during the matching process, and generates combat formations with high collaborative potential. It quantifies the factors affecting collaborative efficiency between any two resource units in a resource combination, such as unit collaboration history, communication protocol compatibility, and differences in affiliated units, into a calculable value. When optimizing resource combinations, it proactively selects resources that not only meet the requirements and respond quickly, but also have a good history of cooperation, smooth communication, and low difficulty in cross-departmental coordination for formation, thus laying a structured foundation for efficient cross-departmental joint operations.

[0041] Other advantages, objectives and features of the invention will be set forth in part in the description which follows, and in part will be apparent to those skilled in the art from the following examination or study, or may be learned from the practice of the invention. Attached Figure Description

[0042] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without any creative effort.

[0043] Figure 1 A flowchart illustrating the steps of an AI-based smart city emergency event prediction and resource scheduling method.

[0044] Figure 2 This is a flowchart illustrating the steps of decomposing complex emergency events in an embodiment of the present invention;

[0045] Figure 3 This is a flowchart of an AI-based smart city emergency prediction and resource scheduling method. Detailed Implementation

[0046] To better understand the above technical solutions, a detailed description of the solutions will be provided below in conjunction with the accompanying drawings and specific embodiments. Obviously, the described embodiments are merely some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0047] The terminology used in the embodiments of this invention is for the purpose of describing particular embodiments only and is not intended to limit the invention. The singular forms “AI-based smart city emergency event prediction and resource scheduling method,” “the,” and “the” as used in the embodiments of this invention and the appended claims are also intended to include the plurality forms, unless the context clearly indicates otherwise; “plural” generally includes at least two.

[0048] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that an article or device that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such an article or device. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the article or device that includes said element.

[0049] To address the shortcomings of existing smart city emergency management systems, such as crude task deconstruction, rigid resource matching, and insufficient dynamic adjustment capabilities when facing complex events, this invention provides an AI-based method for smart city emergency event prediction and resource scheduling. This method aims to intelligently deconstruct complex events using natural language processing and event graph technology, combined with multi-dimensional vectorized modeling of emergency resources, and utilizes multi-objective combinatorial optimization algorithms to achieve accurate and coordinated matching of tasks and resources. Simultaneously, it establishes a real-time monitoring and dynamic rescheduling mechanism, thereby constructing a closed-loop intelligent scheduling system encompassing event perception, task analysis, resource optimization and matching, dynamic adjustment, and command issuance.

[0050] This invention is mainly applied to smart city emergency command centers to deal with complex emergency scenarios such as hazardous chemical leaks, major traffic accidents, natural disasters and their chain events. Traditional methods rely on fixed plans and manual decision-making, which makes it difficult to achieve refined and adaptive scheduling. This invention achieves cross-departmental and multi-resource collaborative emergency response through event element analysis and task package generation, resource capability vectorization modeling, multi-objective resource combination optimization based on Pareto fronts, and an event-driven efficient rescheduling mechanism, which significantly improves the efficiency of handling and resource utilization.

[0051] Specifically, such as Figure 1 As shown, an AI-based method for predicting and scheduling emergency events and allocating resources in smart cities includes the following steps:

[0052] S100: Based on natural language processing and event graph technology, the input complex emergency event description is parsed, the event constituent elements are identified, and the complex emergency event is decomposed into multiple standardized logically related task packages.

[0053] S200: Access the emergency resource database, obtain the static attributes and real-time status data of all available emergency resource units, and construct a multi-dimensional capability vector for each emergency resource unit;

[0054] S300: For each task package generated by S100, iterate through the set of all available emergency resource units, and use a combinatorial optimization algorithm to select and determine an optimal resource combination from the set of all emergency resource units, thus selecting the optimal resource combination for the task package.

[0055] S400. During the emergency response process, the execution progress of each task package and the status changes of each emergency resource unit are monitored in real time. When a change in task requirements, failure of an emergency resource unit, or the emergence of a new derivative task is detected, a rescheduling mechanism is triggered. Depending on the changes, either local rescheduling or global rescheduling is selected to recalculate and update the resource allocation scheme.

[0056] S500, the optimal resource combination calculated for all task packages in S300, and the adjustment scheme in S400, are transformed into specific scheduling instructions and collaborative operation suggestions, which are then issued to the corresponding emergency resource units for execution through the emergency command platform.

[0057] In the specific implementation process, the smart city emergency command system receives a description of a complex emergency event. This description, in unstructured text form, originates from a comprehensive report compiled from multiple alarm channels, social media sentiment monitoring, and IoT sensor alerts or reports. Figure 2 As shown, the steps to decompose a complex emergency event into multiple task packages are as follows:

[0058] Natural language processing is used to analyze the description information of the complex emergency events to extract event elements, including the core event type, location, hazardous substances involved, confirmed secondary hazards, and affected personnel and facilities.

[0059] The extracted event elements are matched with the pre-built emergency response plan knowledge graph to deduce all the basic task types necessary to handle this complex emergency event, forming a preliminary task list;

[0060] Based on the specific information about the event, calculate the specific task parameters for each task in the preliminary task list.

[0061] In this embodiment, S100 first performs natural language processing parsing on the input composite event description. Using a pre-trained BERT-based entity relation extraction model, it identifies and extracts key event elements from the text: core event type, location, hazardous substances involved, confirmed / potential secondary hazards, and affected personnel and facilities. The parsed event elements are then matched with a pre-built urban emergency response plan knowledge graph. This knowledge graph stores the association rules between different event types, hazardous substances, affected objects, and standardized response tasks in a graph structure. Through graph traversal and rule reasoning, the necessary basic task types for handling the event are automatically deduced, forming a preliminary task list. Based on the specific information of the event, including location, wind direction and speed, and the sensitivity of affected objects, specific task parameters are calculated for each task in the preliminary list, generating a parameterized and structured task package. This task package includes: task type, required capability dimension set, required threshold for each capability dimension, expected completion time, and inter-task dependencies. In this embodiment, the composite event is deconstructed into a set of task packages with clearly defined parameters and logical relationships.

[0062] The S200 also connects to the smart city emergency resource database, which dynamically aggregates information on various emergency resource units. For each available emergency resource unit, it acquires its static attributes and real-time status data, including:

[0063] Static attributes: affiliated unit, resource type, functional qualifications, equipment list, communication protocol, etc.

[0064] Real-time status data: current GPS location, status, equipment availability, personnel presence, etc.

[0065] Each resource unit constructs a unified multi-dimensional capability vector, which covers multiple dimensions. In this embodiment, the multi-dimensional capability vector includes:

[0066] Functional capability dimension: Represents the effectiveness of performing specific emergency functions. It is mapped to the interval [0, 1] by normalizing indicators such as equipment performance and personnel skills of resource units.

[0067] Level attribute dimension: Represents the qualification level or protection level of a resource unit, expressed through discrete values ​​or level codes.

[0068] Dynamic status dimension: Represents real-time physical status, including geographic location coordinates, current availability, and device availability rate.

[0069] Collaboration attribute dimension: Represents the inherent attributes of collaborative operations, including the code of the business line to which it belongs, the list of protocols and frequency bands supported by the communication equipment, and the list of collaboration partner identifiers formed based on historical collaboration records. Each resource unit is represented as a structured vector, and all available resource units constitute the available resource set.

[0070] For each task package generated by S100, an optimal resource combination is found from the resource pool constructed by S200. The problem of selecting a resource combination for a task package is modeled as a multi-objective combinatorial optimization problem, and the objective function vector F(C) is defined as follows: Where C represents a resource combination to be evaluated. It is the total capability matching function of resource combination C, which measures the degree of matching between the sum of capabilities of combination C and the capability requirement thresholds of each dimension of the task package. It is a response time function of resource combination C, estimating the maximum or weighted average time required for all resource units in combination C to maneuver from their current location to the mission location. This is the internal collaboration cost function of resource combination C, quantifying the expected collaboration difficulty between resource units within combination C due to factors such as different departments, communication protocols, and lack of historical collaboration experience. The combinatorial optimization algorithm evaluates the cost of collaboration between different resource combinations. When optimizing, resource exclusivity constraints and task timing constraints must be satisfied, including:

[0071] Resource exclusivity constraint: A resource unit can only be allocated to one task package at any given time.

[0072] Task timing constraints: For task packages with dependencies, the resource scheduling start time of subsequent tasks must be later than the expected completion time of their predecessor tasks.

[0073] A multi-objective optimization algorithm is used to solve the problem. For the current task package, under the premise of meeting the basic functional requirements, a batch of resource combinations is randomly generated as the initial population. For each resource combination C in the population, its objective function vector F(C) is calculated, where:

[0074] Overall capability matching degree For the task package required by the first For each capability dimension, the sum of the capability values ​​of all resource units in combination C along that dimension is calculated, and the matching degree contribution is... To ensure that the contribution cap is 1 when there is excess capacity, ultimately ,in, It is the preset weight of this capability dimension.

[0075] Response time : Take the maximum value of the estimated travel time from the current location to the task location for all resource units in combination C.

[0076] Internal collaboration costs Calculate and sum the collaboration costs between all pairwise resource units (p, q) in combination C. ,in, and It is a combination of resources There are two different emergency resource units, Cost(Type(p), Type(p)) is the preset emergency resource unit. and emergency resource units The communication and coordination difficulty coefficient, CoopHist(p, q) is an emergency resource unit calculated based on historical collaboration data. and emergency resource units The level of collaboration between them and These are the weighting coefficients of communication costs and historical collaboration costs in the total collaboration cost, respectively.

[0077] The population is iteratively evolved through operations such as selection, crossover, and mutation to find a set of solutions that are non-inferior to all three objectives, i.e., the Pareto optimal solution set. These solutions represent different trade-offs among matching degree, response time, and cooperation cost. In this embodiment, a non-dominated sorting genetic algorithm (NSGA-II) with an elitist strategy is used to solve the multi-objective optimization problem. For the current task package, under the premise of satisfying resource exclusivity constraints and task timing constraints, the following solution steps are performed:

[0078] Population initialization: The initial population size is set to 150, and the maximum number of iterations is 80 generations. Based on the currently available set of emergency resource units, an initial population that meets the constraints is randomly generated. The algorithm adopts an integer encoding method, where a single chromosome corresponds to a complete resource combination scheme. The gene positions of the chromosome correspond one-to-one with each capability dimension required by the task package, and the gene value is a unique number of the selected emergency resource unit under the corresponding capability dimension, ensuring that the encoding is fully adapted to the resource combination optimization scenario of this scheme.

[0079] Objective function value calculation: For each individual in the population (i.e., resource combination C), calculate its objective function vector: The calculation rules for each sub-function are as follows:

[0080] Overall capability matching degree For the task package required by the first Various capability dimensions, calculation combination The sum of the capability values ​​of all resource units in this dimension, the matching contribution is: To ensure that the contribution cap is 1 when there is excess capacity, ultimately: ,in, The threshold for the requirement of the i-th capability dimension for task package T is quantified by the analytic hierarchy process (AHP) and normalized to the interval [0, 1]. Let R be the capability value of the emergency resource unit in the i-th capability dimension. The preset weight is the weight of the i-th capability dimension, and the sum of the weights of all dimensions is 1.

[0081] Response time : Take the maximum value of the estimated travel time from the current location to the task location for all resource units in combination C. The travel time is calculated based on real-time traffic data of the urban road network using a GIS route planning algorithm.

[0082] Internal collaboration costs : Calculate all pairwise resource units in combination C The collaboration costs between them are summed, i.e.: Among them, Cost(Type) Type CoopHist defines the communication and coordination difficulty coefficients for emergency resource units p and q. The calculation method for the collaboration proficiency between emergency resource units p and q in this embodiment is as follows: Where N represents the number of times p and q have historically collaborated. The preset is 20 times. The historical collaborative processing success rate of p and q Let p and q be the historical average response delay for coordination. The preset time is 30 minutes. The value range is [0, 1], and the larger the value, the higher the collaboration proficiency; and These are the weighting coefficients for communication costs and historical collaboration costs, respectively, in this embodiment. Priority settings are based on cross-departmental collaboration in urban emergency response.

[0083] Non-dominated sorting and crowding calculation: Perform non-dominated sorting on all individuals in the population and divide them into Pareto levels. Level 1 is the optimal non-dominated solution set of the current population. At the same time, calculate the crowding of individuals in the same Pareto level to distinguish the quality of individuals in the same level and avoid premature convergence of the population.

[0084] Selection operation: The tournament selection operator is used. The tournament size is fixed at 5. Five individuals are randomly selected from the current population. The individual with the highest Pareto level is selected first. If the levels are the same, the individual with greater crowding is selected to enter the offspring population. The operation is repeated until the offspring population size reaches 75 (i.e. 50% of the initial population size).

[0085] Crossover operation: The simulated binary crossover operator (SBX) is used, with a crossover probability of 0.9 and a distribution index of 20. Individuals in the offspring population are randomly paired and crossover is performed to generate new individuals. During the crossover process, resource exclusivity constraints are checked in real time. If the same resource unit is repeatedly allocated after crossover, the conflicting gene positions are randomly corrected to ensure that the generated individuals always meet the constraints.

[0086] Mutation operation: A multinomial mutation operator is used, with the mutation probability set to 1 / 150 (i.e., 1 / initial population size) and the distribution index set to 20. A random mutation operation is performed on the individuals generated after crossover, randomly selecting a single gene locus of the individual and replacing it with the number of other available emergency resource units under the same ability dimension. After mutation, the constraints are checked again, and individuals that do not meet the requirements are corrected.

[0087] Elite Preservation and Iteration Termination: The parent population is merged with the mutated offspring population to obtain a merged population of 300. Non-dominated sorting and crowding calculation are performed on the merged population to select 150 high-quality individuals as the new generation of parent population. This process is repeated until the maximum number of iterations of 80 generations is reached, at which point the iteration stops and the final Pareto optimal solution set is output. These solutions represent different trade-offs among matching degree, response time, and collaboration cost, allowing emergency command personnel to select the final solution based on on-site decision-making preferences.

[0088] Emergency commanders select a final option from the Pareto front based on current decision preferences, which is then used as an optimal resource combination for the task package.

[0089] The optimization process is performed in parallel on all task packages generated by event deconstruction to generate a global resource allocation scheme.

[0090] After the resource scheduling instructions are issued and the response action begins, the real-time monitoring and dynamic adjustment phase begins. This phase monitors the execution progress feedback of each task package and the real-time status of each emergency resource unit. The rescheduling mechanism is triggered if any of the following conditions are met: change in task requirements, failure of a resource unit, or occurrence of a derivative event.

[0091] Rescheduling mechanisms include local rescheduling and global rescheduling, among which:

[0092] Local rescheduling: When the triggering reason is a change in the requirements of a single task or the failure of a resource unit, other unaffected task packages and their allocated resource schemes are locked, and the S300 optimization process is re-executed only for the affected task packages to find alternative or supplementary solutions from the remaining available resources.

[0093] Global Rescheduling: When the triggering cause is the occurrence of a new derivative emergency event, the new event description is merged with the original event information as a new composite event input, and S100 is returned to start a new event deconstruction, generating a new task package set containing the original tasks and the newly added derivative tasks. Based on the latest emergency resource unit status, the global optimization of S300 is re-executed on all task packages.

[0094] S400 ensures that the scheduling scheme can dynamically evolve with the event situation and resource status, and always remain adaptable.

[0095] The optimized resource allocation scheme is transformed into executable instructions. For each task package, detailed scheduling instructions are generated for its allocated resource combination, including:

[0096] A unique identifier for all emergency resource units assigned to this task package;

[0097] The target locations that each resource unit should go to;

[0098] Clearly defined expected tasks, key actions, and coordination requirements.

[0099] For task packages that require collaboration among multiple resource units, additional collaborative operation suggestions are generated, including: suggested rendezvous points, joint travel routes, and specified unified communication channels / frequency bands.

[0100] The aforementioned dispatch instructions and coordination suggestions are sent through the emergency command platform in a standardized message format to the mobile terminals or vehicle-mounted systems of the corresponding emergency resource units. After receiving the instructions, the personnel at the front line act accordingly and report the status through the terminals.

[0101] like Figure 3 As shown, the specific steps of the AI-based smart city emergency event prediction and resource scheduling method provided by this invention in performing emergency event prediction and resource scheduling are as follows:

[0102] (1) Event information reception and input

[0103] Receive initial description information of complex emergency events from multiple channels.

[0104] The natural language processing module is invoked to perform intelligent analysis on the event description text.

[0105] Automatically identify and extract the core components of an event, including: event type, location, hazardous substances involved, secondary hazards that have occurred, and threatened personnel and critical facilities.

[0106] (2) Knowledge graph matching and task inference

[0107] The extracted event elements are matched and reasoned with a pre-set urban emergency response plan knowledge graph.

[0108] Based on the rules and relationships in the graph, a series of standardized basic task types necessary for handling the event are automatically deduced, forming a preliminary task list.

[0109] (3) Task parameterization and packaging

[0110] Based on detailed information such as the specific geographical location, severity, and environmental conditions of the event, specific parameters are calculated for each task in the initial task list, including capability requirement thresholds, expected completion time limits, and dependencies.

[0111] Each parameterized task is encapsulated into a well-structured task package.

[0112] (4) Emergency resource data aggregation and modeling

[0113] It can access the city's emergency resource database in real time to obtain static files and real-time status of all available resource units.

[0114] For each resource unit, construct a unified multi-dimensional capability vector: functional capability dimension, level attribute dimension, dynamic state dimension, and collaborative attribute dimension.

[0115] (5) Intelligent resource matching based on multi-objective optimization

[0116] For each generated task package, find an optimal resource combination from the established resource pool.

[0117] The matching process employs a multi-objective combined optimization algorithm to achieve the following objectives: highest capability matching degree, shortest response time, and lowest internal collaboration cost.

[0118] During the matching process, the following constraints must be observed: a resource can only be allocated to one task at a time; and the resource scheduling times of dependent tasks must be sequential.

[0119] Ultimately, for each task package, one or more high-quality resource combination schemes that achieve a balance among multiple indicators are output for decision-makers to select and confirm the optimal option.

[0120] (6) Implementation of the plan and dynamic monitoring of the whole process

[0121] The optimized resource scheduling plan is transformed into specific instructions, which are then issued to the corresponding resource units for execution through the emergency command platform.

[0122] During the handling process, we continuously monitor the progress of each task, the real-time status of each resource unit, and new changes in the event situation detected through on-site feedback or sensors.

[0123] (7) Event-driven dynamic rescheduling

[0124] Once a specific change is detected, the rescheduling mechanism is immediately triggered. Triggering conditions include: changes in task requirements, resource unit failure, and the occurrence of new derivative events.

[0125] Based on the triggering reason, an intelligent rescheduling strategy is selected: local rescheduling or global rescheduling.

[0126] The determined resource allocation results are transformed into detailed, executable instructions and coordination suggestions, which are then issued through the command platform.

[0127] In summary, this invention addresses the challenges of inaccurate, uncoordinated, and inflexible scheduling in smart cities responding to complex emergency events through a comprehensive design encompassing intelligent event deconstruction, multi-dimensional resource modeling, global optimization matching, and dynamic adjustment. In its implementation, this method utilizes NLP and knowledge graphs to refine and standardize tasks; achieves quantitative alignment of supply and demand through resource vectorization; proactively optimizes coordination costs while meeting capacity and timeliness requirements through multi-objective optimization; and ensures the dynamic adaptability of the solution through an event-driven rescheduling mechanism. The entire method is highly intelligent and adaptable, fully utilizing various urban emergency resources to transform from passive response to intelligent prediction and proactive optimization scheduling, significantly improving the overall effectiveness and resilience of urban emergency management.

[0128] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any simple modifications, equivalent changes and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. An AI-based method for predicting and allocating resources for emergency events in smart cities, characterized in that: The specific steps of this method are as follows: S100. Based on natural language processing and event graph technology, the input complex emergency event description is parsed, the event constituent elements are identified, and the complex emergency event is decomposed into multiple standardized logically related task packages. S200: Access the emergency resource database, obtain the static attributes and real-time status data of all available emergency resource units, and construct a multi-dimensional capability vector for each emergency resource unit; S300: For each task package generated by S100, iterate through the set of all available emergency resource units, and use a combinatorial optimization algorithm to select and determine an optimal resource combination from the set of all emergency resource units, thus selecting the optimal resource combination for the task package. S400. During the emergency response process, the execution progress of each task package and the status changes of each emergency resource unit are monitored in real time. When a change in task requirements, failure of an emergency resource unit, or the emergence of a new derivative task is detected, a rescheduling mechanism is triggered. Depending on the changes, either local rescheduling or global rescheduling is selected to recalculate and update the resource allocation scheme. S500, the optimal resource combination calculated for all task packages in S300, and the adjustment scheme in S400, are transformed into specific scheduling instructions and collaborative operation suggestions, which are then issued to the corresponding emergency resource units for execution through the emergency command platform.

2. The AI-based smart city emergency event prediction and resource scheduling method according to claim 1, characterized in that, In step S100, the step of decomposing a complex emergency event into multiple task packages is as follows: Natural language processing is used to analyze the description information of the complex emergency events to extract event elements, including the core event type, location, hazardous substances involved, confirmed secondary hazards, and affected personnel and facilities. The extracted event elements are matched with the pre-built emergency response plan knowledge graph to deduce all the basic task types necessary to handle this complex emergency event, forming a preliminary task list; Based on the specific information of the event, specific task parameters are calculated for each task in the preliminary task list. The task parameters include quantified capability requirement thresholds for each dimension, expected completion time calculated based on the severity and location of the event, and inter-task dependencies determined by analyzing the logical relationships between tasks. The parameterized tasks are then used as the task package.

3. The AI-based smart city emergency event prediction and resource scheduling method according to claim 1, characterized in that, In S200, the multidimensional capability vector includes: Functional capability dimension: represents the effectiveness of an emergency resource unit in performing a specific emergency function. The value of this functional capability dimension is a normalized interval of [0, 1]. Level attribute dimension: indicates the qualification certification, protection level or professional rating of the emergency resource unit; Dynamic status dimension: Represents the real-time physical status of emergency resource units at the moment of dispatch, including geographical coordinates, current availability, and equipment integrity rate; Collaborative attribute dimension: This represents the inherent attributes of an emergency resource unit when it collaborates with other emergency resource units, including the business it belongs to, the protocols and frequency bands supported by the communication equipment, and the collaborative partner identifier formed by historical collaborative records.

4. The AI-based smart city emergency event prediction and resource scheduling method according to claim 1, characterized in that, In S300, the combinatorial optimization algorithm employs a multi-objective optimization method to find a set of resource combination schemes such that for each task package, the corresponding objective function vector reaches the Pareto optimal frontier. The objective function vector F(C) is defined as: Where C represents a resource combination to be evaluated. It is the total capability matching degree function of resource combination C. It is the response time function of resource combination C. It is the internal collaboration cost function of resource combination C. Combinatorial optimization algorithms evaluate the cost of different resource combinations. The value is used to search and select the best option.

5. The AI-based smart city emergency event prediction and resource scheduling method according to claim 1, characterized in that, In step S300, when processing a complex emergency event involving multiple task packages, the following constraints must be met when allocating resource combinations to all task packages simultaneously: Resource exclusivity constraint: An emergency resource unit can only be allocated to one task package at any given time; Task timing constraints: For task packages with sequential dependencies, the resource scheduling and action start time of subsequent task packages must be later than the expected completion time of their predecessor task packages.

6. The AI-based smart city emergency event prediction and resource scheduling method according to claim 4, characterized in that, The total capability matching degree function The calculation method is as follows: ,in, This indicates the task package currently awaiting assignment. Indicates task package For the first The threshold for the required capabilities, where C represents a combination of resources to be evaluated. Representing resource combination Emergency resource units in Indicates emergency resource unit In the Ability values ​​in various ability dimensions Representing resource combination In the The sum of abilities across all dimensions of ability It is the first The preset weights of various capability dimensions in the matching degree calculation, function This is used to ensure that the matching contribution is capped at 1 when the combined capability exceeds the demand.

7. The AI-based smart city emergency event prediction and resource scheduling method according to claim 4, characterized in that, The internal collaboration cost function The calculation comprehensively considers the difficulty of coordination between resources, and the calculation method is as follows: ,in, and It is a combination of resources There are two different emergency resource units, Cost(Type(p), Type(p)) is the preset emergency resource unit. and emergency resource units The communication and coordination difficulty coefficient, CoopHist(p, q) is an emergency resource unit calculated based on historical collaboration data. and emergency resource units The level of collaboration between them and These are the weighting coefficients of communication costs and historical collaboration costs in the total collaboration cost, respectively.

8. The AI-based smart city emergency event prediction and resource scheduling method according to claim 1, characterized in that, The specific steps of S400 are as follows: S410. Triggered when any of the continuously monitored indicators meet the following conditions: the capacity requirement threshold of a single task package changes, the status of the emergency resource unit allocated to the task package becomes unavailable, or a new derivative emergency event occurs. S420. When the triggering reason is that the capacity requirement threshold of a single task package changes or the status of the emergency resource unit allocated to the task package becomes unavailable, other unaffected task packages and their allocated resource allocation schemes are locked. Only the affected task packages are re-executed to find a new optimal resource combination. S430. When the triggering reason is the occurrence of a new derivative emergency event, the new event description is merged with the original event information and returned as a new input to S100. A new event deconstruction is initiated to generate a new task package set containing all tasks, old and new. Based on the current latest emergency resource unit status, S300 is re-executed on all task packages.

9. The AI-based smart city emergency event prediction and resource scheduling method according to claim 1, characterized in that, In S500, the scheduling instructions and collaborative operation suggestions include: The dispatch instructions issued for each task package clearly list the identifiers of all emergency resource units in the optimal resource combination allocated to that task package, their destinations, and the expected task content. Collaborative operation recommendations generated by emergency resource units requiring coordination include suggested meeting points, joint routes, and designated unified communication channels or frequency bands.

Images