A Case-Based Reasoning-Based Method and System for Fire Emergency Response Plan Simulation and Verification

Abstract

This invention provides a method and system for fire emergency response plan simulation and verification based on case-based reasoning. The method includes establishing a structured historical fire case database, calculating the mixed similarity between the current scenario and historical cases based on rough set theory and cloud models, and selecting similar source cases; adapting the handling strategies of the source cases to the model by constraining the model to generate an initial plan to be verified; establishing a generalized stochastic Petri net model representing the evolution of the disaster-affected state and the interaction of rescue resources, mapping the initial plan to be verified to the transition rate and initial identifier in the database within the model, performing concurrent conflict detection and temporal logic deduction, and evaluating transient performance indicators; if the target rescue success indicator does not reach a preset threshold, iteratively correcting the resource scheduling parameters of the plan based on the sensitivity analysis results of key transitions, and repeating the deduction steps until the termination condition is met, outputting the target fire emergency response plan.

Description

Technical Field

[0001] This application belongs to the field of verification, and in particular relates to a method and system for verifying fire emergency response plans based on case reasoning. Background Technology

[0002] Case-based reasoning involves retrieving historical cases most similar to the current fire scenario and reusing or modifying the response plan as the current contingency plan. However, fire scene data contains fuzzy and random information, and has numerous characteristic attributes. The method for calculating feature weights, based on expert experience, struggles to objectively reflect the importance of attributes in different fire scenarios. Furthermore, similarity calculations, based on quantitative Euclidean distance or qualitative matching, struggle to handle the conversion and mapping between quantitative data and qualitative concepts. Historical case response strategies were formulated based on the resource allocation at the time; when transferred to the current scenario, they ignore the differences in constraints of existing rescue resources, leading to resource conflicts or scheduling infeasibility in actual implementation. Fire rescue is a concurrent and uncertain process involving the evolution of the affected body's state and the collaborative interaction of multiple rescue forces. Existing technologies struggle to model and simulate the temporal logic of rescue operations, resource competition, and the synergy of parallel tasks before plan implementation. Due to the lack of a deduction and verification mechanism, the generated plan cannot anticipate potential logical deadlocks, temporal conflicts, or delays at critical nodes during actual rescue operations. Furthermore, existing technologies are mostly open-loop decision-making processes, which cannot perform sensitivity analysis and iterative optimization of resource scheduling parameters in the plan based on the expected rescue effect. Therefore, there is an urgent need for a fire emergency plan generation method that can integrate multi-dimensional feature retrieval, adapt to differences in resource constraints, and perform deductive verification and closed-loop optimization of the plan. Summary of the Invention

[0003] To address the problems that existing technologies struggle to integrate multi-dimensional feature retrieval, adapt to varying resource constraints, and perform simulation, verification, and optimization of contingency plans.

[0004] In the first aspect, the present invention proposes a case-based reasoning-based method for simulating and verifying fire emergency response plans, comprising:

[0005] Construct a historical fire case library containing structured environmental attributes and fire evolution characteristics, obtain static feature parameters and evolution trends of the current fire scene; calculate the dependency weight of feature attributes based on rough set conditional entropy, use a cloud model inverse generator to transform historical quantitative features into qualitative concept clouds, and combine weights to calculate the mixed similarity between the current scene and historical cases.

[0006] Extract a set of source cases whose similarity meets the threshold, and use a resource difference correction model based on constraint satisfaction to adapt and transfer the source case handling strategy to generate an initial test plan containing resource allocation and action sequence.

[0007] A generalized stochastic Petri net model representing the evolution of the disaster-affected state and the interaction of rescue resources is established. The initial contingency plan to be tested is mapped to the transition rate and initial identifier of the storage location in the model, and concurrent conflict detection and temporal logic deduction are performed.

[0008] The transient performance index of the simulation model is calculated. If the target rescue success index does not reach the preset threshold, the resource scheduling parameters of the plan are iteratively corrected based on the sensitivity analysis results of key changes, and the simulation steps are repeated until the termination condition is met, and the target fire emergency plan is output.

[0009] In another aspect, the present invention also proposes a case-based reasoning-based fire emergency plan simulation and verification system, comprising the following modules:

[0010] The calculation module is used to build a historical fire case library containing structured environmental attributes and fire evolution characteristics, obtain static feature parameters and evolution trends of the current fire scene; calculate the dependency weight of feature attributes based on rough set conditional entropy, use the cloud model inverse generator to transform historical quantitative features into qualitative concept cloud, and combine the weights to calculate the mixed similarity between the current scene and historical cases.

[0011] The generation module is used to extract a set of source cases whose similarity meets the threshold, and to adapt and transfer the source case handling strategy using a resource difference correction model based on constraint satisfaction, thereby generating an initial test plan that includes resource allocation and action sequence.

[0012] The mapping module is used to establish a generalized stochastic Petri net model representing the evolution of the disaster-affected state and the interaction of rescue resources. It maps the initial contingency plan to be tested into the transition rate and initial identifier of the storage location in the model, and performs concurrent conflict detection and temporal logic deduction.

[0013] The output module is used to calculate the transient performance index of the simulation model. If the target rescue success index does not reach the preset threshold, the resource scheduling parameters of the plan are iteratively corrected based on the sensitivity analysis results of key changes, and the simulation steps are repeated until the termination condition is met, and the target fire emergency plan is output.

[0014] This invention utilizes rough set conditional entropy and cloud model theory to achieve objective weighting and qualitative-quantitative transformation of fire scene characteristics. By employing a constraint-based correction model, it can differentiate historical strategies based on current resource constraints, ensuring the feasibility of the generated contingency plan. Using a generalized stochastic Petri net model, the contingency plan execution process is mapped to a mathematical network structure. Through analysis of transition rates and location identifiers, it can not only pre-identify potential resource conflicts and temporal logic flaws in the plan but also pinpoint factors affecting rescue effectiveness through sensitivity analysis. Through this closed-loop deduction and iterative correction mechanism, the contingency plan can undergo thorough logical verification and parameter optimization before actual action, enhancing the scientific rigor and precision of the fire emergency plan. Attached Figure Description

[0015] Figure 1 A flowchart of the first embodiment; Figure 2 This is a schematic diagram of a qualitative concept cloud. Figure 3 This is a schematic diagram of the initial test plan. Detailed Implementation

[0016] The features and exemplary embodiments of various aspects of this application will be described in detail below. To make the objectives, technical solutions, and advantages of this application clearer, the application will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are only intended to explain this application and not to limit it. For those skilled in the art, this application can be implemented without some of these specific details. The following description of the embodiments is merely to provide a better understanding of this application by illustrating examples.

[0017] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising..." does not exclude the presence of additional identical elements in the process, method, article, or apparatus that includes said element.

[0018] Figure 1 The flowchart of the first embodiment of the present invention is as follows: Figure 1The method for simulating and verifying fire emergency response plans based on case-based reasoning, as shown, includes:

[0019] S1. Construct a historical fire case library containing structured environmental attributes and fire evolution characteristics, obtain static feature parameters and evolution trends of the current fire scene; calculate the dependency weight of feature attributes based on rough set conditional entropy, use cloud model inverse generator to transform historical quantitative features into qualitative concept cloud, and combine weights to calculate the mixed similarity between the current scene and historical cases.

[0020] Specifically, past fire accident reports and on-site investigation data are collected to establish a non-relational database to store historical cases. Case attributes are divided into static environmental attributes and evolution attributes. Static environmental attributes include building type, building area, fire resistance rating, internal combustible load, and layout of fire protection facilities. Evolution attributes include time-series data on ignition point location, fire spread rate, smoke diffusion rate, and temperature rise gradient. In the current fire scenario, on-site data is collected in real time by heat detectors, smoke detectors, and thermal imaging cameras deployed on-site. The current temperature value, smoke concentration value, and flame coverage area are extracted as static feature parameters. At the same time, continuous monitoring data within the previous time window are selected, and the least squares method is used for linear fitting. The slope of the fitted line is calculated as the evolution trend parameters of temperature rise trend and smoke diffusion trend.

[0021] The continuous feature attribute data in the historical case database is discretized to construct a decision information table. The conditional entropy of the decision attribute relative to the conditional attribute is calculated. The importance of each feature attribute is calculated based on the conditional entropy, and the dependence weight of each feature attribute is obtained after normalization. For quantitative features such as temperature and area in historical cases, the sample mean is calculated as the expectation of the cloud model. The sample variance is calculated, and the entropy and hyper-entropy of the cloud model are derived based on the sample variance and expectation, thus generating qualitative concept cloud features composed of expectation, entropy, and hyper-entropy. When calculating similarity, Hamming distance is used to calculate the matching degree for qualitative text attributes. For quantitative numerical attributes, the attribute value of the current scene is input into the corresponding qualitative concept cloud, and the certainty of the value under the cloud model is calculated using a Gaussian membership function as the similarity value. Figure 2 The similarity of each attribute is weighted and summed using the dependency weights calculated above to obtain the global mixed similarity between the current scene and historical cases.

[0022] In an optional embodiment, calculating the dependency weights of feature attributes based on rough set conditional entropy includes:

[0023] Establish a decision information system for historical fire cases, wherein the decision information system consists of a domain, a set of conditional attributes, and a set of decision attributes;

[0024] Calculate the conditional entropy of each attribute in the conditional attribute set relative to the decision attribute set;

[0025] Calculate the overall conditional entropy of the conditional attribute set;

[0026] The importance metric for an attribute is the difference between the conditional entropy of the conditional attribute set after removing the attribute and the overall conditional entropy of the original conditional attribute set.

[0027] Normalize the importance metric to obtain the dependency weights of the feature attributes.

[0028] Construct a form of S=<U,C,D> The decision table is defined as follows: U is the domain containing, for example, 500 historical fire cases; C is the set of conditional attributes, such as ignition source, wind speed, ambient temperature, and building type; and D is the set of decision attributes, such as rescue success or failure, and response level. For each conditional attribute... The conditional entropy of the attribute under decision attribute D is calculated using the Shannon entropy formula. The overall conditional entropy H(D|C) under the full attribute set is calculated. In one embodiment, all samples are divided into different feature subsets according to the combination of values ​​of all conditional attributes. The information entropy of the decision results in each subset is calculated separately. The information entropy is then weighted and summed according to the sample proportion of each subset to obtain the overall conditional entropy. The formula is used to calculate this. Calculate attribute importance, the value of which reflects the importance of removing attributes. The degree of increase in uncertainty; for example, if the entropy increase is greatest after removing ignition material, then the attribute is critical. All attributes... The value is normalized, that is This allows us to obtain the dependency weight of each feature attribute in the similarity calculation.

[0029] To address the ambiguity and randomness in fire data, in an optional embodiment, the process of transforming historical quantitative features into a qualitative conceptual cloud using a cloud model inverse generator includes:

[0030] Collect quantitative feature data of the same attribute from the historical case database to form a sample set;

[0031] Calculate the sample mean and first-order sample absolute central moments of the sample set;

[0032] Set the sample mean as the expectation of the cloud model;

[0033] The entropy of the cloud model is calculated using the first-order sample absolute central moments;

[0034] Calculate the sample variance, and take the arithmetic square root of the absolute value of the difference between the sample variance and the square of the entropy as the hyperentropy;

[0035] The output includes cloud digital features containing expectation, entropy, and hyperentropy as historical qualitative conceptual cloud attributes.

[0036] Taking wind speed as an example, a sample set is constructed by extracting all wind speed values ​​for a certain type of fire scenario from the historical case library. For example, n=100 data points. Calculate the sample mean. The expectation Ex of the cloud model reflects the central value of the concept; the first-order sample absolute central moments are calculated. And the entropy is derived using the first-order absolute central moment formula. This parameter represents the range of ambiguity in the concept of wind speed; calculate the sample variance. and using the formula The hyperentropy He is calculated to represent the uncertainty of entropy and the discreteness of cloud droplets. The output is a qualitative concept cloud consisting of triples (Ex, En, He), such as (3.5, 0.8, 0.1), realizing the transformation from discrete quantitative historical data to qualitative linguistic concepts with soft partitioning capabilities.

[0037] In an optional embodiment, the step of calculating the mixed similarity between the current scene and historical cases by combining weights includes:

[0038] For symbolic features, the local similarity is recorded as 1 when the feature value of the current scene is completely consistent with the feature value of historical cases, and as 0 when they are inconsistent.

[0039] For numerical features, the distance between the current scene feature value and the historical case feature value is calculated using the normalized Euclidean distance formula, and the distance is converted into a local similarity in the range of 0 to 1.

[0040] For cloud model features, the quantitative feature values ​​of the current scene are substituted into the Gaussian membership function of the qualitative concept cloud corresponding to the historical case, and the calculated membership value is used as the local similarity.

[0041] By using the dependency weights of feature attributes to perform a weighted sum of all local similarities, the global similarity between the current scene and historical cases is obtained.

[0042] For symbolic features such as building structure, the Crisp logic is used. If the current scenario is a reinforced concrete structure and the historical cases are also reinforced concrete structures, then the similarity is... Otherwise, it is 0. For the deterministic numerical characteristics of floor height, a value range [min, max] is set, and the calculation is performed. Cloud model features with fuzzy properties regarding wind speed; if the current wind speed is... If the concept cloud corresponding to the historical cases is (Ex, En, He), then a random normal number with En as the expectation and He as the standard deviation is generated. And using the Gaussian membership formula Calculate similarity To eliminate randomness, the calculation is usually repeated multiple times and the average value is taken. The dependency weights obtained from the above calculations are then combined. Using the weighted summation formula The global similarity is calculated, which represents the degree of matching between the current scenario and historical cases.

[0043] S2, extract the set of source cases whose similarity meets the threshold, and use the resource difference correction model based on constraint satisfaction to adapt and transfer the source case handling strategy to generate an initial test plan containing resource allocation and action sequence;

[0044] Specifically, a similarity threshold of 0.85 is set. The calculation results are iterated, and historical cases with a global mixed similarity greater than this threshold are selected to form the source case set. The resource allocation adaptation problem is defined as a constraint satisfaction problem. Variables include fire truck type, number of firefighters, type of extinguishing agent, and length of water supply pipeline. The value range of the variables is the list of resources that the current rescue team can actually call upon. Constraints include the required water supply intensity being less than or equal to the actual water supply capacity and the rescue arrival time being less than or equal to the critical time of fire out of control. The ratio of the current scene's burning area to the burning area of ​​the source case is calculated as a correction coefficient. The resource input in the source case is multiplied by this correction coefficient to obtain the demand. If the demand is within the current resource value range and meets the constraints, the quantity is reused and corrected. If the constraints are not met, a backtracking search algorithm is used to find a resource combination that meets all constraints within the value range. The output is an initial test plan containing the specific task allocation of each combat unit, vehicle parking positions, water hose laying routes, and attack and retreat times.

[0045] In an optional embodiment, the step of extracting a set of source cases whose similarity meets a threshold, adapting the source case handling strategy using a resource difference correction model based on constraint satisfaction, and generating an initial contingency plan including resource allocation and action sequence includes:

[0046] The quantity of rescue resources that need to be adjusted is defined as the decision variable, which includes the number of fire trucks, the amount of fire extinguishing agent used, and the number of firefighters.

[0047] Define the maximum range of the actual number of resources that can be dispatched at the current rescue station as the range constraint of the variable;

[0048] Define the minimum fire extinguishing requirement and arrival time limit calculated based on the current fire intensity as satisfying the constraint;

[0049] Construct an objective function to calculate the cost of the deviation between the assigned values ​​of the decision variables and the resource allocation of the source case;

[0050] A constraint satisfaction solution algorithm is used to find variable assignment schemes that satisfy all constraints and optimize the objective function within the range constraints;

[0051] The corresponding resource configuration in the source case is replaced by the variable assignment scheme obtained from the solution, and an initial test plan is generated.

[0052] The problem of revising the contingency plan is modeled as a constraint satisfaction problem (CSP) to address the discrepancy between the resource allocation in the source case and the current actual situation. A set of decision variables is defined. Define constraint C: hard constraints that must be satisfied, such as... ,in The maximum number of available vehicles at fire stations in the current jurisdiction, and the constraints to be met, such as... ,in For injection efficiency, The current heat release rate at the fire site, Let the cooling coefficient be used. Construct the objective function. ,in , These are preset weighting coefficients used to balance the relative importance of different resources. The number of fire trucks configured in the source case. In the case of fire extinguishing agent dosage, to prevent discrepancies in dimensions, adjustments can be made. , Eliminating the differences caused by the unit of measurement is obvious to those skilled in the art. The goal is to find the solution that minimizes the configuration differences from the source case, thus preserving experience and strictly conforming to the current constraints. A backtracking search algorithm is used to search the variable space. For example, if the source case suggests 3 vehicles, but the current fire calculation requires at least 4, and there are currently 6 vehicles in the area, the output will be... If there are only 3 vehicles in the jurisdiction, output 3 vehicles and mark them with a risk warning. Fill the optimized resource parameters into the contingency plan template to generate an initial contingency plan for simulation, such as... Figure 3 .

[0053] S3. Establish a generalized stochastic Petri net model representing the evolution of the disaster-affected state and the interaction of rescue resources. Map the initial contingency plan to be verified to the transition rate and initial identifier of the storage location in the model, and perform concurrent conflict detection and temporal logic deduction.

[0054] Specifically, the set of locations in a generalized random Petri net represents the state of the fire scene and the state of rescue resources, including uncontrolled fire, fire spread, resources in place, and water supply for firefighting. The set of transitions represents fire development events and rescue action events, including flame ignition of adjacent areas, fire truck movement, laying of hoses, and spraying of extinguishing agents. The resource quantities in the initial test plan are configured as the initial token quantities of the corresponding locations. Directed arcs connecting locations and transitions are formed according to the action logic set in the plan. The implementation rate of fire spread transitions is determined using the probability distribution function obtained from historical data statistical analysis. The implementation rate of rescue action transitions is determined based on the performance parameters of the rescue equipment and the action time specified in the plan.

[0055] The model's reachability graph is constructed, and all state nodes in the reachability graph are traversed to check for deadlock states, such as the absence of a token (also known as an identifier) ​​in the resource repository but the need to trigger a rescue transition, i.e., no fire truck is available but water is required to extinguish the fire. The model also checks for state loops that contradict the actual logic, such as the fire being extinguished but the extinguishing agent is still being consumed, thus completing the concurrent conflict detection. Through temporal logic deduction, the flow process of tokens in the network is simulated, such as the change from the flow of tokens from the fire garage to the laying of water hoses → then to the flow of water to extinguish the fire, corresponding to the timing of resource investment in rescue actions. The entire state sequence from the initial state to the fire extinguishing state is recorded, such as fire ignition → spread → controlled → extinguished, and the time consumption of each stage.

[0056] In an optional embodiment, establishing a generalized stochastic Petri net model representing the evolution of the disaster-affected state and the interaction of relief resources includes:

[0057] Define a set of storage locations, which includes state storage locations representing different stages of fire development and resource storage locations representing fire trucks, water cannons, and extinguishing agents;

[0058] Define a set of transitions, which includes random temporal transitions of fire spread, deterministic temporal transitions of rescue preparation, and instantaneous transitions of rescue decisions;

[0059] Define a set of directed arcs, establish the input arc pointing from the resource repository to the rescue transition, and the output arc pointing from the rescue transition to the next stage fire state repository;

[0060] Define an association function and set a weight on each arc, representing the amount of resources consumed in the rescue operation or the triggering condition for the fire state transition.

[0061] Constructing the GSPN model six-tuple The set of places P is divided into a subset of states. ,like Fire spread, Controlled and resource subsets ,like Water tanker, Task force. The transition set T contains: an exponentially distributed random transition representing the natural development of the fire. The rate parameter is related to the properties of the combustible material; it represents the deterministic changes in the actions of laying water hoses and breaching rescue structures. The time delay is a fixed value, such as 300 seconds; and it represents the instantaneous changes in command and decision-making. A directed arc F defines the logical flow direction, such as from... point to The input arc, with a weight W of 2, indicates that triggering the general offensive transition requires consuming or occupying 2 assault group tokens. Through this topology, it is possible to simulate the intervention process of resource input (i.e., token flow) on the fire state evolution path, i.e., the changes marked in the state database.

[0062] To instantiate static plans into model parameters and solve the engineering mapping problem, in an optional embodiment, mapping the initial plan to be tested to the transition rate and initial location identifier in the model includes:

[0063] Extract the number of rescue vehicles and firefighters from the initial contingency plan to be tested, and use the values ​​as the initial number of tokens in the resource library;

[0064] Extract the estimated deployment time from the initial contingency plan to be tested, and assign the reciprocal of the estimated deployment time to the implementation rate of the rescue preparation change;

[0065] Based on the ratio of the extinguishing agent's rated flow rate to the fire area, a fire suppression rate function is constructed.

[0066] The calculated fire suppression rate is assigned to the implementation rate of the corresponding fire suppression transition in the generalized stochastic Petri net model.

[0067] For the initial identifier If the contingency plan includes 4 main battle fire trucks and 20 firefighters, then a resource depot should be set up. , Regarding the rate of change If the planned water supply pipeline laying time is 5 minutes, then the corresponding deterministic transition rate is set to... For fire suppression transitions, i.e., random transitions, a function is constructed. ,in The total outflow rate mobilized in the contingency plan, Let k be the current fire area, and k be an empirical fire extinguishing efficiency coefficient, for example, 0.05, used to standardize the model's rate dimensions and fit the actual fire extinguishing efficiency. For example, when the total flow rate is 100 L / s and the fire area is 200... At that time, the calculated The value will assign changes to fire extinguishing in the model, thereby controlling the simulation process to the resource intensity configured in the plan.

[0068] S4. Calculate the transient performance index of the simulation model. If the target rescue success index does not reach the preset threshold, the resource scheduling parameters of the plan are iteratively corrected based on the sensitivity analysis results of key changes, and the simulation steps are repeated until the termination condition is met, and the target fire emergency plan is output.

[0069] Specifically, based on the constructed generalized stochastic Petri net model, a continuous-time Markov chain is generated isomorphically. By solving the state transition equation of the Markov chain, the probability of reaching the fire extinguishing state within the preset rescue time limit is calculated as the target rescue success index. If this probability is lower than the preset success rate threshold of 95%, a sensitivity analysis is performed on the transition rate representing the rescue action in the model. The partial derivatives of the target index with respect to each transition rate are calculated, and the transition with the largest absolute value of the partial derivative is selected as the key transition. For the rescue action corresponding to the key transition, the resource input is increased or the action start time is adjusted within the resource constraint range. The corresponding parameters in the initial test plan are updated and remapped to the transition rate or initial identifier of the Petri net model. The deduction and calculation are performed again until the target rescue success index is higher than the threshold. The verified parameter combination is then used to generate a textual target fire emergency plan and output.

[0070] In an optional embodiment, the iterative correction of the resource scheduling parameters of the contingency plan based on the sensitivity analysis results of key changes includes:

[0071] Select an instantaneous performance index function R that represents the probability of fire suppression or the time of rescue;

[0072] Calculate the rate of change of the index function R for each rescue-related transition in the model. partial derivatives ;

[0073] The transition with the largest absolute value of the partial derivative is selected as the key transition;

[0074] Identify the types of rescue resources associated with key changes;

[0075] Increase the number of rescue resources to be dispatched within the resource constraints;

[0076] Update the initial token quantity of the corresponding resource repository and the change rate dependent on the resource quantity based on the increased resource quantity, and proceed to the next round of simulation.

[0077] After the Petri net simulation, the system performance index R is defined as the probability that the fire state Kussocken is 0 within 30 minutes. The sensitivity vector is calculated using numerical analysis methods. ,in Corresponding to the rates of different rescue stages, such as the rate of internal attack and the rate of water supply. Compare the magnitudes of each component; if found... The largest absolute value indicates that increasing the water supply rate is effective in improving the success rate of the rescue, thus identifying the change in water supply as the key change. The associated resource is identified as a high-powered water tanker; therefore, without exceeding the upper limit for the jurisdiction, the number of water tankers N in the plan is increased to N+1. The calculation is then performed using the method described above. and the updated inhibition rate This triggers a new round of GSPN simulations until the indicator R meets the preset success threshold, such as 95%, thereby outputting the optimal fire emergency plan.

[0078] In the second embodiment, the present invention also proposes a case-based reasoning-based fire emergency plan simulation and verification system, comprising the following modules:

[0079] The calculation module is used to build a historical fire case library containing structured environmental attributes and fire evolution characteristics, obtain static feature parameters and evolution trends of the current fire scene; calculate the dependency weight of feature attributes based on rough set conditional entropy, use the cloud model inverse generator to transform historical quantitative features into qualitative concept cloud, and combine the weights to calculate the mixed similarity between the current scene and historical cases.

[0080] The generation module is used to extract a set of source cases whose similarity meets the threshold, and to adapt and transfer the source case handling strategy using a resource difference correction model based on constraint satisfaction, thereby generating an initial test plan that includes resource allocation and action sequence.

[0081] The mapping module is used to establish a generalized stochastic Petri net model representing the evolution of the disaster-affected state and the interaction of rescue resources. It maps the initial contingency plan to be tested into the transition rate and initial identifier of the storage location in the model, and performs concurrent conflict detection and temporal logic deduction.

[0082] The output module is used to calculate the transient performance index of the simulation model. If the target rescue success index does not reach the preset threshold, the resource scheduling parameters of the plan are iteratively corrected based on the sensitivity analysis results of key changes, and the simulation steps are repeated until the termination condition is met, and the target fire emergency plan is output.

[0083] In an optional embodiment, calculating the dependency weights of feature attributes based on rough set conditional entropy includes:

[0084] Establish a decision information system for historical fire cases, wherein the decision information system consists of a domain, a set of conditional attributes, and a set of decision attributes;

[0085] Calculate the conditional entropy of each attribute in the conditional attribute set relative to the decision attribute set;

[0086] Calculate the overall conditional entropy of the conditional attribute set;

[0087] The importance metric for an attribute is the difference between the conditional entropy of the conditional attribute set after removing the attribute and the overall conditional entropy of the original conditional attribute set.

[0088] Normalize the importance metric to obtain the dependency weights of the feature attributes.

[0089] In an optional embodiment, the step of using a cloud model inverse generator to transform historical quantitative features into qualitative conceptual clouds includes:

[0090] Collect quantitative feature data of the same attribute from the historical case database to form a sample set;

[0091] Calculate the sample mean and first-order sample absolute central moments of the sample set;

[0092] Set the sample mean as the expectation of the cloud model;

[0093] The entropy of the cloud model is calculated using the first-order sample absolute central moments;

[0094] Calculate the sample variance, and take the arithmetic square root of the absolute value of the difference between the sample variance and the square of the entropy as the hyperentropy;

[0095] The output includes cloud digital features containing expectation, entropy, and hyperentropy as historical qualitative conceptual cloud attributes.

[0096] In an optional embodiment, the step of calculating the mixed similarity between the current scene and historical cases by combining weights includes:

[0097] For symbolic features, the local similarity is recorded as 1 when the feature value of the current scene is completely consistent with the feature value of historical cases, and as 0 when they are inconsistent.

[0098] For numerical features, the distance between the current scene feature value and the historical case feature value is calculated using the normalized Euclidean distance formula, and the distance is converted into a local similarity in the range of 0 to 1.

[0099] For cloud model features, the quantitative feature values ​​of the current scene are substituted into the Gaussian membership function of the qualitative concept cloud corresponding to the historical case, and the calculated membership value is used as the local similarity.

[0100] By using the dependency weights of feature attributes to perform a weighted sum of all local similarities, the global similarity between the current scene and historical cases is obtained.

[0101] In an optional embodiment, the step of extracting a set of source cases whose similarity meets a threshold, adapting the source case handling strategy using a resource difference correction model based on constraint satisfaction, and generating an initial contingency plan including resource allocation and action sequence includes:

[0102] The quantity of rescue resources that need to be adjusted is defined as the decision variable, which includes the number of fire trucks, the amount of fire extinguishing agent used, and the number of firefighters.

[0103] Define the maximum range of the actual number of resources that can be dispatched at the current rescue station as the range constraint of the variable;

[0104] Define the minimum fire extinguishing requirement and arrival time limit calculated based on the current fire intensity as satisfying the constraint;

[0105] Construct an objective function to calculate the cost of the deviation between the assigned values ​​of the decision variables and the resource allocation of the source case;

[0106] A constraint satisfaction solution algorithm is used to find variable assignment schemes that satisfy all constraints and optimize the objective function within the range constraints;

[0107] The corresponding resource configuration in the source case is replaced by the variable assignment scheme obtained from the solution, and an initial test plan is generated.

[0108] In an optional embodiment, establishing a generalized stochastic Petri net model representing the evolution of the disaster-affected state and the interaction of relief resources includes:

[0109] Define a set of storage locations, which includes state storage locations representing different stages of fire development and resource storage locations representing fire trucks, water cannons, and extinguishing agents;

[0110] Define a set of transitions, which includes random temporal transitions of fire spread, deterministic temporal transitions of rescue preparation, and instantaneous transitions of rescue decisions;

[0111] Define a set of directed arcs, establish the input arc pointing from the resource repository to the rescue transition, and the output arc pointing from the rescue transition to the next stage fire state repository;

[0112] Define an association function and set a weight on each arc, representing the amount of resources consumed in the rescue operation or the triggering condition for the fire state transition.

[0113] In an optional embodiment, mapping the initial proposed scenario to the transition rate and initial location identifier in the model includes:

[0114] Extract the number of rescue vehicles and firefighters from the initial contingency plan to be tested, and use the values ​​as the initial number of tokens in the resource library;

[0115] Extract the estimated deployment time from the initial contingency plan to be tested, and assign the reciprocal of the estimated deployment time to the implementation rate of the rescue preparation change;

[0116] Based on the ratio of the extinguishing agent's rated flow rate to the fire area, a fire suppression rate function is constructed.

[0117] The calculated fire suppression rate is assigned to the implementation rate of the corresponding fire suppression transition in the generalized stochastic Petri net model.

[0118] In an optional embodiment, the iterative correction of the resource scheduling parameters of the contingency plan based on the sensitivity analysis results of key changes includes:

[0119] Select an instantaneous performance index function R that represents the probability of fire suppression or the time of rescue;

[0120] Calculate the rate of change of the index function R for each rescue-related transition in the model. partial derivatives ;

[0121] The transition with the largest absolute value of the partial derivative is selected as the key transition;

[0122] Identify the types of rescue resources associated with key changes;

[0123] Increase the number of rescue resources to be dispatched within the resource constraints;

[0124] Update the initial token quantity of the corresponding resource repository and the change rate dependent on the resource quantity based on the increased resource quantity, and proceed to the next round of simulation.

[0125] In the above embodiments, several specific steps are described and illustrated as examples. However, the methods and specific values ​​of this application are not limited to the specific steps and specific values ​​described and illustrated. Those skilled in the art can make various changes, modifications and additions, or change the order of steps after understanding the spirit of this application.

[0126] The above description is merely a specific implementation of this application. Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, modules, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here. It should be understood that the protection scope of this application is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in this application, and these modifications or substitutions should all be covered within the protection scope of this application.

Claims

1. A method for simulating and verifying fire emergency response plans based on case-based reasoning, characterized in that, include: Construct a historical fire case library containing structured environmental attributes and fire evolution characteristics, and obtain the static characteristic parameters and evolution trends of the current fire scenario; The dependency weights of feature attributes are calculated based on the conditional entropy of rough sets. The cloud model inverse generator is used to transform historical quantitative features into qualitative concept clouds. The weights are then combined to calculate the mixed similarity between the current scene and historical cases. Extract a set of source cases whose similarity meets the threshold, and use a resource difference correction model based on constraint satisfaction to adapt and transfer the source case handling strategy to generate an initial test plan containing resource allocation and action sequence. A generalized stochastic Petri net model representing the evolution of the disaster-affected state and the interaction of rescue resources is established. The initial contingency plan to be tested is mapped to the transition rate and initial identifier of the storage location in the model, and concurrent conflict detection and temporal logic deduction are performed. The transient performance index of the simulation model is calculated. If the target rescue success index does not reach the preset threshold, the resource scheduling parameters of the plan are iteratively corrected based on the sensitivity analysis results of key changes, and the simulation steps are repeated until the termination condition is met, and the target fire emergency plan is output.

2. The method according to claim 1, characterized in that, The calculation of the dependency weights of feature attributes based on rough set conditional entropy includes: Establish a decision information system for historical fire cases, wherein the decision information system consists of a domain of discourse, a set of conditional attributes, and a set of decision attributes; Calculate the conditional entropy of each attribute in the conditional attribute set relative to the decision attribute set; Calculate the overall conditional entropy of the conditional attribute set; The importance metric for an attribute is the difference between the conditional entropy of the conditional attribute set after removing the attribute and the overall conditional entropy of the original conditional attribute set. Normalize the importance metric to obtain the dependency weights of the feature attributes.

3. The method according to claim 2, characterized in that, The process of using a cloud model inverse generator to transform historical quantitative features into qualitative conceptual clouds includes: Collect quantitative feature data of the same attribute from the historical case database to form a sample set; Calculate the sample mean and first-order sample absolute central moments of the sample set; Set the sample mean as the expectation of the cloud model; The entropy of the cloud model is calculated using the first-order sample absolute central moments; Calculate the sample variance, and take the arithmetic square root of the absolute value of the difference between the sample variance and the square of the entropy as the hyperentropy; The output includes cloud digital features containing expectation, entropy, and hyperentropy as historical qualitative conceptual cloud attributes.

4. The method according to claim 1, characterized in that, The weighted calculation of the mixed similarity between the current scene and historical cases includes: For symbolic features, the local similarity is recorded as 1 when the feature value of the current scene is completely consistent with the feature value of historical cases, and as 0 when they are inconsistent. For numerical features, the distance between the current scene feature value and the historical case feature value is calculated using the normalized Euclidean distance formula, and the distance is converted into a local similarity in the range of 0 to 1. For cloud model features, the quantitative feature values ​​of the current scene are substituted into the Gaussian membership function of the qualitative concept cloud corresponding to the historical case, and the calculated membership value is used as the local similarity. By using the dependency weights of feature attributes to perform a weighted sum of all local similarities, the global similarity between the current scene and historical cases is obtained.

5. The method according to claim 1, characterized in that, The process involves extracting a set of source cases whose similarity meets a threshold, then adapting the source case handling strategy using a resource difference correction model based on constraint satisfaction, generating an initial contingency plan that includes resource allocation and action timing, including: The quantity of rescue resources that need to be adjusted is defined as the decision variable, which includes the number of fire trucks, the amount of fire extinguishing agent used, and the number of firefighters. Define the maximum range of the actual number of resources that can be dispatched at the current rescue station as the range constraint of the variable; Define the minimum fire extinguishing requirement and arrival time limit calculated based on the current fire intensity as satisfying the constraint; Construct an objective function to calculate the cost of the deviation between the assigned values ​​of the decision variables and the resource allocation of the source case; A constraint satisfaction solution algorithm is used to find variable assignment schemes that satisfy all constraints and optimize the objective function within the range constraints; The corresponding resource configuration in the source case is replaced by the variable assignment scheme obtained from the solution, and an initial test plan is generated.

6. The method according to claim 4, characterized in that, The establishment of a generalized stochastic Petri net model representing the evolution of the disaster-affected state and the interaction of relief resources includes: Define a set of storage locations, which includes state storage locations representing different stages of fire development and resource storage locations representing fire trucks, water cannons, and extinguishing agents; Define a set of transitions, which includes random temporal transitions of fire spread, deterministic temporal transitions of rescue preparation, and instantaneous transitions of rescue decisions; Define a set of directed arcs, establish the input arc pointing from the resource repository to the rescue transition, and the output arc pointing from the rescue transition to the next stage fire state repository; Define an association function and set a weight on each arc, representing the amount of resources consumed in the rescue operation or the triggering condition for the fire state transition.

7. The method according to claim 1, characterized in that, The mapping of the initial contingency plan to be tested to the transition rate and initial location identifier in the model includes: Extract the number of rescue vehicles and firefighters from the initial contingency plan to be tested, and use the values ​​as the initial number of tokens in the resource library; Extract the estimated deployment time from the initial contingency plan to be tested, and assign the reciprocal of the estimated deployment time to the implementation rate of the rescue preparation change; Based on the ratio of the extinguishing agent's rated flow rate to the fire area, a fire suppression rate function is constructed. The calculated fire suppression rate is assigned to the implementation rate of the corresponding fire suppression transition in the generalized stochastic Petri net model.

8. The method according to claim 1, characterized in that, The iterative correction of the resource scheduling parameters of the contingency plan based on the sensitivity analysis results of key changes includes: Select an instantaneous performance index function R that represents the probability of fire suppression or the time of rescue; Calculate the rate of change of the index function R for each rescue-related transition in the model. The partial derivatives; The transition with the largest absolute value of the partial derivative is selected as the key transition; Identify the types of rescue resources associated with key changes; Increase the number of rescue resources to be dispatched within the resource constraints; Update the initial token quantity of the corresponding resource repository and the change rate dependent on the resource quantity based on the increased resource quantity, and proceed to the next round of simulation.

9. A case-based reasoning-based fire emergency response plan simulation and verification system, characterized in that, Includes the following modules: The calculation module is used to build a historical fire case library containing structured environmental attributes and fire evolution characteristics, obtain static feature parameters and evolution trends of the current fire scene; calculate the dependency weight of feature attributes based on rough set conditional entropy, use the cloud model inverse generator to transform historical quantitative features into qualitative concept cloud, and combine the weights to calculate the mixed similarity between the current scene and historical cases. The generation module is used to extract a set of source cases whose similarity meets the threshold, and to adapt and transfer the source case handling strategy using a resource difference correction model based on constraint satisfaction, thereby generating an initial test plan that includes resource allocation and action sequence. The mapping module is used to establish a generalized stochastic Petri net model representing the evolution of the disaster-affected state and the interaction of rescue resources. It maps the initial contingency plan to be tested into the transition rate and initial identifier of the storage location in the model, and performs concurrent conflict detection and temporal logic deduction. The output module is used to calculate the transient performance index of the simulation model. If the target rescue success index does not reach the preset threshold, the resource scheduling parameters of the plan are iteratively corrected based on the sensitivity analysis results of key changes, and the simulation steps are repeated until the termination condition is met, and the target fire emergency plan is output.

10. The system according to claim 9, characterized in that, The calculation of the dependency weights of feature attributes based on rough set conditional entropy includes: Establish a decision information system for historical fire cases, wherein the decision information system consists of a domain of discourse, a set of conditional attributes, and a set of decision attributes; Calculate the conditional entropy of each attribute in the conditional attribute set relative to the decision attribute set; Calculate the overall conditional entropy of the conditional attribute set; The importance metric for an attribute is the difference between the conditional entropy of the conditional attribute set after removing the attribute and the overall conditional entropy of the original conditional attribute set. Normalize the importance metric to obtain the dependency weights of the feature attributes.

Images