Adaptive generation method and system for emergency strategy of highway tunnel emergencies

Abstract

The invention discloses an emergency strategy self-adaptive generation method and system for highway tunnel emergencies, and relates to the technical field of intelligent traffic control, and the method comprises the steps: collecting multi-modal data; an event basic feature, a tunnel structure feature and an environment dynamic feature are extracted layer by layer through a dynamic feature hierarchical extraction module, and a multi-dimensional feature vector is generated; performing event type primary screening, severity secondary screening and environment adaptation verification in sequence through a control item dynamic screening module, and screening out candidate control items from the total control item set; and performing multi-objective optimization on the candidate control items, generating a Pareto optimal strategy set, and outputting a final emergency strategy in combination with an interactive decision support interface. According to the method, the event type identification accuracy and the environmental risk prediction timeliness can be effectively improved, the candidate control item screening precision is higher through three-stage verification of the control item dynamic screening module, and strategy conflicts and resource waste are effectively avoided.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent traffic control, and more specifically, it relates to a method and system for adaptively generating emergency strategies for highway tunnel emergencies. Background Art

[0002] As an important part of traffic infrastructure, the enclosed space structure and complex operating environment of highway tunnels make emergencies such as fires and traffic accidents extremely harmful. Emergencies often exhibit characteristics such as multi-factor coupling, rapid dynamic evolution, and high risk of secondary disasters. Traditional emergency decision-making relies on manual experience and static pre-plan libraries, and there are problems such as lagging response, poor environmental adaptability, and difficulty in coordinating multi-objective conflicts.

[0003] With the popularization of Internet of Things sensing technology and intelligent algorithms, building a data-driven adaptive emergency strategy generation system has become a key direction for improving the safety operation level of tunnels. The prior art records an emergency decision-making framework based on fuzzy logic, but it uses a single sensor data fusion method, with a single feature extraction dimension, and cannot effectively process the spatio-temporal correlation of multi-modal data, such as capturing the synergistic effects of key parameters such as the smoke diffusion gradient and visibility attenuation rate; some prior art also records using genetic algorithms for strategy optimization, but no feature hierarchical mechanism is established, resulting in insufficient modeling of the coupling relationship between dynamic environmental parameters and structural constraints; in addition, the emergency strategy generation methods in the above prior art are mainly for emergencies on highway roads, and when applied to highway tunnels, the control item screening process lacks a verification mechanism, which is likely to lead to situations such as the ventilation strategy not matching the fluid mechanics characteristics and the evacuation path conflicting with the crowd dynamics model.

[0004] Therefore, how to research and design a method and system for adaptively generating emergency strategies for highway tunnel emergencies that can overcome the above defects is an urgent problem for us to solve currently. Summary of the Invention

[0005] To solve the deficiencies in the prior art, the purpose of the present invention is to provide a method and system for adaptively generating emergency strategies for highway tunnel emergencies. By constructing a multi-dimensional feature vector through a dynamic feature hierarchical extraction module and combining a three-level feature fusion mechanism of an event basic feature layer, a tunnel structure feature layer, and an environmental dynamic feature layer, compared with traditional single-layer feature extraction methods, it can effectively improve the accuracy of event type recognition and the timeliness of environmental risk prediction. Through the three-stage verification of the control item dynamic screening module, the accuracy of candidate control item screening is higher, effectively avoiding strategy conflicts and resource waste.

[0006] The above technical purpose of the present invention is achieved through the following technical solutions: In the first aspect, a method for adaptively generating emergency strategies for highway tunnel emergencies is provided, including the following steps: Collect sensor data, video surveillance data and environmental parameters in the target tunnel to obtain multi-modal data; Based on the multi-modal data, layer by layer extract event basic features, tunnel structure features and environmental dynamic features through a dynamic feature hierarchical extraction module to generate a multi-dimensional feature vector; Based on the multi-dimensional feature vector, sequentially perform initial screening of event types, secondary screening of severity, and environmental adaptation verification through a control item dynamic screening module, and screen out candidate control items from the total control item set; Use the NSGA-Ⅲ algorithm or the improved NSGA-Ⅲ algorithm to perform multi-objective optimization on the candidate control items, generate a Pareto optimal strategy set, and output the final emergency strategy in combination with an interactive decision support interface.

[0007] Furthermore, the dynamic feature hierarchical extraction module includes: An event basic feature layer for extracting event basic features including temperature, smoke concentration, CO concentration, vehicle density and accident point location data; A tunnel structure feature layer for extracting tunnel structure features including tunnel curvature, ventilation shaft distribution, escape passage topology and fire fighting facility layout; An environmental dynamic feature layer for extracting environmental dynamic features including wind speed vector field, temperature and humidity gradient matrix and visibility attenuation rate.

[0008] Furthermore, the initial screening of event types is implemented using a multi-modal fusion classification model; Among them, the modal fusion classification model takes the event basic features and the tunnel structure features as inputs, performs a feature splicing operation on the event basic features and the tunnel structure features, and then applies a multi-class activation function to output a probability distribution vector of event types.

[0009] Furthermore, the secondary screening of severity is achieved through risk index calculation, specifically including: Solve the time reciprocal of the environmental dynamic features to obtain the environmental change risk; Perform normalization processing on the event basic features and then calculate the L2 norm to obtain the global event risk; Determine the risk index based on the sum of the environmental change risk and the environmental change risk.

[0010] Furthermore, the environmental adaptation verification includes: Perform a fluid dynamics simulation verification on the ventilation system control item; Perform a crowd dynamics model verification on the evacuation path planning item; And / or, perform an accessibility topology analysis on the fire rescue deployment item.

[0011] Furthermore, the objective function of the multi-objective optimization takes the activation state of the candidate control item as the decision variable; The objective function includes: Minimize the time objective, including the coupling effect of the execution time of the control item and the dynamic environment gradient; Minimize the cost objective, introduce the opportunity cost, and quantify the loss caused by giving up other measures due to enabling the control item; Minimize the risk objective, and use the product form to characterize the complexity risk amplification effect brought by the superposition of multiple control items; Constraint conditions, including mutually exclusive control item constraints, dependent control item constraints, and tunnel physical constraints.

[0012] Furthermore, the improvement of the improved NSGA-Ⅲ algorithm includes: Guide the search direction through the dynamic reference point, and adaptively divide the hyperplane according to the dimension of the candidate control item; In addition, adopt the constraint violation degree weighting strategy to punish the solutions that do not meet the tunnel structure constraints.

[0013] Furthermore, the interactive decision support interface outputs the final emergency strategy, including: Calculate the solution with the minimum Manhattan distance in the Pareto optimal strategy set as the recommended strategy, and output the final emergency strategy.

[0014] Furthermore, the method further includes: Dynamically update the feature extraction rule base and the weight parameters of the multi-objective optimization according to the execution results of the final emergency strategy; Among them, the dynamic update of the feature extraction rule base adopts an incremental reinforcement learning mechanism.

[0015] In the second aspect, a highway tunnel emergency strategy adaptive generation system is provided. This system is used to implement a highway tunnel emergency strategy adaptive generation method as described in any one of the first aspects, including: The data acquisition module is used to collect sensor data, video monitoring data, and environmental parameters in the target tunnel to obtain multi-modal data; The feature extraction module is used to layer by layer extract the event basic features, tunnel structure features, and environmental dynamic features based on the multi-modal data through the dynamic feature hierarchical extraction module, and generate a multi-dimensional feature vector; The control screening module is used to screen out candidate control items from the total control item set by sequentially performing event type primary screening, severity secondary screening, and environmental adaptability verification through the control item dynamic screening module based on the multi-dimensional feature vector; The target optimization module is used to perform multi-objective optimization on the candidate control items by using the NSGA-Ⅲ algorithm or the improved NSGA-Ⅲ algorithm, generate a Pareto optimal strategy set, and output the final emergency strategy in combination with the interactive decision support interface.

[0016] Compared with the prior art, the present invention has the following beneficial effects: 1. For the method for adaptively generating an emergency strategy for highway tunnel emergencies provided by the present invention, a multi-dimensional feature vector is constructed through the dynamic feature hierarchical extraction module, and a three-level feature fusion mechanism combining the event basic feature layer, the tunnel structure feature layer, and the environmental dynamic feature layer is adopted. Compared with the traditional single-layer feature extraction method, it can effectively improve the accuracy of event type recognition and the timeliness of environmental risk prediction. Through the three-stage verification of the control item dynamic screening module, the screening accuracy of the candidate control items is higher, effectively avoiding strategy conflicts and resource waste; 2. The present invention adopts the improved NSGA-Ⅲ algorithm to achieve multi-objective optimization. Through the dynamic reference point to guide the search direction and the constraint violation degree weighting strategy, in the multi-dimensional control item scenario, the convergence speed of the Pareto front can be accelerated, the coverage rate of the strategy set can be improved, and combined with the minimum Manhattan distance solution recommendation mechanism, the decision-making efficiency is higher than that of the traditional Euclidean distance method; 3. The present invention introduces the joint verification of fluid mechanics simulation and crowd dynamics model in the environmental adaptation verification link, which can improve the smoke dispersion efficiency of the ventilation strategy and shorten the residence time of personnel in the evacuation path planning; and optimize the fire rescue path through reachability topological analysis, which can reduce the response time of emergency resource deployment. Description of the Drawings

[0017] The drawings described herein are used to provide a further understanding of the embodiments of the present invention, form a part of this application, and do not constitute a limitation to the embodiments of the present invention. In the drawings: Figure 1 is the flowchart in Embodiment 1 of the present invention; Figure 2 is the system block diagram in Embodiment 2 of the present invention. Detailed Embodiments

[0018] To make the objectives, technical solutions, and advantages of the present invention clearer and more understandable, the present invention will be further described in detail below in conjunction with the embodiments and the drawings. The illustrative embodiments and descriptions thereof of the present invention are only used to explain the present invention and do not constitute a limitation to the present invention.

[0019] Embodiment 1: A method for adaptively generating an emergency strategy for highway tunnel emergencies, as Figure 1 shown, includes the following steps: S1: Collect sensor data, video surveillance data, and environmental parameters in the target tunnel to obtain multi-modal data; S2: Based on multimodal data, the event basic features, tunnel structure features and environmental dynamic features are extracted layer by layer through the dynamic feature hierarchical extraction module to generate a multidimensional feature vector; S3: Based on the multi-dimensional feature vector, the control item dynamic screening module performs event type primary screening, severity secondary screening and environmental adaptation verification in sequence to select candidate control items from the total control item set; S4: Use NSGA-Ⅲ algorithm or improved NSGA-Ⅲ algorithm to perform multi-objective optimization on candidate control items, generate Pareto optimal strategy set, and output the final emergency strategy in combination with interactive decision support interface. Among them, NSGA-Ⅲ algorithm is the third generation non-dominated sorting genetic algorithm.

[0020] In step S1, the sensor data is the core data for real-time monitoring of the tunnel environment status, which is collected through distributed IoT devices, including but not limited to temperature, smoke concentration, carbon monoxide concentration, carbon dioxide concentration, oxygen concentration, harmful gases, vibration frequency of tunnel lining, crack width, water seepage pressure, vehicle density, vehicle speed and accident point location information.

[0021] Video surveillance data is a key input for visual perception and event recognition. It extracts information through multi-source cameras and intelligent analysis algorithms, including but not limited to vehicle accident types, number of stranded personnel, visible light recognition of flames / smoke, distribution of thermal radiation in high-temperature areas, detection of hidden fire sources, three-dimensional outlines of obstacles, spatial positions of vehicles / personnel, detection of running / falling personnel, and judgment of vehicle reversing / illegal parking.

[0022] Environmental parameters are auxiliary data for assessing dynamic risks and strategy feasibility, combined with real-time meteorological and equipment status collection, including but not limited to wind speed / direction, humidity, air pressure, light intensity, fan speed, fire water pressure, remaining power of emergency power supply, wireless signal strength and other information.

[0023] In step S2, the dynamic feature hierarchical extraction module includes an event basic feature layer, a tunnel structure feature layer, and an environmental dynamic feature layer. The present invention adopts a three-level hierarchical extraction mechanism for the heterogeneity of multimodal data and the dynamic risk evolution characteristics of highway tunnel emergency scenarios; among them, the event basic feature layer mainly extracts features that directly reflect the essence of the event from the original data; the tunnel structure feature layer mainly loads the static topology information of the tunnel to ensure that the strategy complies with physical constraints; and the environmental dynamic feature layer predicts the environmental evolution trend and realizes forward-looking strategy optimization.

[0024] Specifically, the event-based feature layer is used to extract event-based features such as temperature, smoke concentration, CO concentration, vehicle density, and accident point location data; the tunnel structure feature layer is used to extract tunnel structure features such as tunnel curvature, ventilation shaft distribution, escape passage topology, and fire facility layout; the environmental dynamic feature layer is used to extract environmental dynamic features such as wind speed vector field, temperature and humidity gradient matrix, and visibility attenuation rate.

[0025] For example, the data input of the event-based feature layer is: temperature (95°C), smoke concentration (800 μg / m³), CO concentration (1200 ppm), vehicle density (12 vehicles / 100 m). Calculate the mean / extreme values (such as the highest temperature, maximum smoke concentration) within a 50 m range around the accident point. Through Kalman filtering to fuse multi-sensor data, eliminate noise and generate a continuous spatio-temporal distribution map, the obtained event-based features .

[0026] The data input of the tunnel structure feature layer is: BIM model (the accident point is located at the K23+500 bend section), ventilation shaft position (the upstream well distance is 350 m, the downstream well distance is 200 m), the available number of fire hydrants (2, water pressure 0.8 MPa). Calculate the shortest path between the accident point and key facilities (such as the distance of the escape passage), define the maximum allowable fire truck width in the bend area (3.5 m), the obtained tunnel structure features .

[0027] The data input of the environmental dynamic feature layer is: wind speed vector (2.8 m / s, direction S→N), temperature and humidity gradient (temperature rise rate 8°C / min, humidity drop rate 15% / min), visibility attenuation rate (-15% / min). Based on fluid dynamics simulation to predict the smoke diffusion path, use long short-term memory network to estimate the environmental state in the next 5 minutes, the obtained environmental dynamic features .

[0028] After obtaining the event-based features, tunnel structure features, and environmental dynamic features through layering, the three-layer features are integrated into a unified vector through feature-level fusion, which is the multi-dimensional feature vector .

[0029] In the highway tunnel fire scenario, the dynamic feature extraction through multi-source data fusion and hierarchical processing transforms the original data into feature vectors with clear physical meanings: the event-based features locate the core parameters of the accident and determine the response level; the tunnel structure features constrain the feasibility of the strategy and optimize resource deployment; the environmental dynamic features predict the development trend of the situation and achieve forward-looking control. The present invention enables the emergency strategy generation to quickly complete feature extraction in a complex environment, provides a high signal-to-noise ratio input for subsequent multi-objective optimization, and significantly improves the efficiency and reliability of emergency response.

[0030] In step S3, the initial screening of event types filters out irrelevant control items according to event types (such as fires, traffic accidents). For example: if it is a fire, smoke exhaust and lane closure may be required, while for a traffic accident, diversion and lighting control may be needed. It is necessary to combine the event basic features in the multi-dimensional feature vector, such as temperature and smoke concentration, to determine the event type.

[0031] In some examples, the initial screening of event types is implemented using a multi-modal fusion classification model; among them, the modal fusion classification model takes the event basic features and tunnel structure features as inputs. After performing a feature splicing operation on the event basic features and tunnel structure features, a multi-class activation function is applied to output the probability distribution vector of the event type.

[0032] Specifically, the expression of the modal fusion classification model is: ; Among them, represents the probability distribution vector of the event type, such as the prediction probabilities of fire, traffic accident, and hazardous material leakage; represents the multi-class activation function; represents the trainable weight matrix, with a dimension equal to the product of the number of event types and the total feature dimension; represents the event basic features; represents the tunnel structure features; represents the bias term; represents the feature splicing operation.

[0033] The secondary screening of severity further narrows down the range of control items according to the severity of the event. For example, a minor fire may only require partial lane closure, while a severe fire requires full closure. It may be necessary to use a risk index formula and combine dynamic features such as the smoke concentration gradient and temperature change rate to evaluate the severity.

[0034] In some examples, the secondary screening of severity is achieved through risk index calculation, specifically including: solving the time reciprocal of the environmental dynamic features to obtain the environmental change risk; calculating the L2 norm after normalizing the event basic features to obtain the global event risk; determining the risk index based on the sum of the environmental change risk and the environmental change risk.

[0035] The specific expression of the risk index is: ; Among them, represents the risk index; represents the environmental dynamic sensitivity coefficient, which is used to amplify the impact of environmental mutations; It represents the event-based sensitivity coefficient, which is used to measure the harm degree of the event itself. In the present invention, a major response is triggered when the risk index exceeds the set threshold. For example, in case of a fire, key items such as activating the smoke exhaust system to operate at full power and two-way traffic control can be activated.

[0036] The environmental adaptation verification is to ensure that the selected control items are compatible with the current environmental dynamics and tunnel structure. For example, use CFD simulation to verify whether the ventilation strategy is feasible, or check whether the evacuation path is unavailable due to structural problems. It is necessary to combine the tunnel structure characteristics (such as the location of ventilation shafts) and environmental dynamic characteristics (such as wind speed, visibility) for verification.

[0037] In some examples, the environmental adaptation verification can adopt one or more of the following methods: perform a fluid mechanics simulation verification on the ventilation system control items; perform a crowd dynamics model verification on the evacuation path planning items; perform an accessibility topological analysis on the fire fighting and rescue deployment items.

[0038] In some examples, performing a fluid mechanics simulation verification on the ventilation system control items is mainly to verify whether the ventilation strategies such as smoke exhaust and air supply conform to the law of smoke diffusion, so as to prevent the strategies from failing or aggravating the risk.

[0039] Based on computational fluid dynamics (CFD), a three-dimensional tunnel model is established, and real-time parameters are loaded: structural parameters and dynamic parameters. The structural parameters include but are not limited to the tunnel cross-section size, the location of ventilation shafts, and the fan power. The dynamic parameters include but are not limited to the real-time wind speed and the smoke concentration gradient. By setting the boundary conditions and simulating the smoke diffusion under different ventilation strategies, key indicators such as the visibility recovery time and the reduction rate of the high-temperature area can be obtained.

[0040] The expression of the boundary condition is: ; Among them, represents the velocity field; represents time; represents the convective acceleration; represents the air density; represents the pressure gradient; represents the kinematic viscosity of air; represents the Laplacian of the velocity field; represents the smoke buoyancy.

[0041] In some examples, performing a crowd dynamics model verification on the evacuation path planning items is mainly to ensure that the planned escape path meets the crowd behavior characteristics and dynamic environmental constraints, and prevent trampling or congestion.

[0042] For example, the path planning solution: Guide to the cross tunnel H3 (width 3.2m); Real-time data: Personnel density (2 persons / m²), visibility (15m); Predict the passage time of the crowd through the cross tunnel H3 within 5 minutes through dynamic simulation and detect bottleneck areas. The bottleneck areas such as the curve where the speed drops by 30% due to limited visibility.

[0043] In some examples, performing reachability topology analysis on fire rescue deployment items is mainly to verify the physical reachability and timeliness of the fire resource dispatching path, ensuring that the rescue force arrives quickly.

[0044] For example, build a tunnel space topology map based on the BIM model. The nodes in the tunnel space topology map: accident point (K23+500), fire hydrant (HYD1 / 2), cross tunnel (H3 / H4); Edge weight: passage distance × dynamic risk coefficient. For example, when the CO concentration > 1000 ppm, the weight +50%; Then update the obstacles in real time: collapse area (coordinates K23+480 → K23+520); Solve the optimal path through the algorithm: The fire truck detours from the entrance → cross tunnel H4.

[0045] In step S4, the objective function of multi-objective optimization uses the activation state of candidate control items as the decision variable.

[0046] The objective function includes: minimizing the time objective, including the coupled influence of the execution time of the control item and the dynamic environment gradient; minimizing the cost objective, introducing the opportunity cost to quantify the loss caused by giving up other measures due to enabling the control item; minimizing the risk objective, using the product form to characterize the complexity risk amplification effect brought by the superposition of multiple control items.

[0047] The constraint conditions include mutually exclusive control item constraints, dependent control item constraints, and tunnel physical constraints.

[0048] Specifically, the expression of the objective function is: ; Among them, represents minimizing the time objective; represents the set of binary decision variables; represents the th candidate control item is activated or not. The value of 1 means activated, and the value of 0 means not activated; represents the number of variables in the set of binary decision variables; represents the th candidate control item from startup to take effect; represents the environmental dynamic change penalty weight, adjusting the influence weight of the environmental parameter change rate (such as the smoke diffusion speed) on timeliness; Represents the gradient norm of environmental characteristics, integrating the severity of dynamic parameters such as smoke concentration gradient and temperature change rate; Represents the cost minimization objective; Represents the direct resource input cost caused by the th candidate control item; Represents the opportunity cost weight, adjusting the priority of indirect economic losses in the total economic objective; Represents the indirect economic loss caused by the th candidate control item; Represents the risk minimization objective; Represents the global risk adjustment coefficient, amplifying or suppressing the influence weight of the event risk level on the strategy security; Represents the event risk level, a risk level divided according to the event type and severity, such as low risk of level I and high risk of level III; Represents the complex risk coefficient of the th candidate control item, the larger the value, the higher the risk; Represents the th set of mutually exclusive control item sets, and the control items within the same set cannot be activated simultaneously; Represents the mutually exclusive control item group; Represents the candidate control item is the candidate control item precondition of, and to activate the candidate control item the candidate control item must be activated first; Represents the dependent control item pair; Represents the constraint coefficient matrix, describing the relationship between the control item and the tunnel structure parameters, such as lane width and ventilation shaft spacing; Represents the constraint threshold vector, characterizing the physical limit values of the tunnel structure, such as the maximum load and the minimum net height.

[0049] The improvements of the improved NSGA-III algorithm in the present invention include: guiding the search direction through a dynamic reference point and adaptively dividing the hyperplane according to the dimension of the candidate control item; and, adopting a constraint violation degree weighting strategy to punish the solutions that do not meet the tunnel structure constraints.

[0050] In some examples, the expression for guiding the search direction through a dynamic reference point is: ; wherein, represents the dynamic reference point; represents the minimum value in the time minimization objective; represents the maximum value in the time minimization objective; represents the real-time environment offset of the time minimization objective, such as the fire spread stage Increased to bias towards time optimization; Represents the minimum value in the cost minimization objective; Represents the maximum value in the cost minimization objective; Represents the real-time environment offset of the cost minimization objective; Represents the minimum value in the risk minimization objective; Represents the maximum value in the risk minimization objective; Represents the real-time environment offset of the risk minimization objective.

[0051] The present invention divides the target space hyperplane through dynamic reference points, avoiding the solution set aggregation caused by fixed reference points in traditional algorithms.

[0052] In some examples, the constraint violation degree weighting strategy is adopted, and the expression for punishing the solutions that do not meet the tunnel structure constraints is: ; Wherein, Represents the total penalty value of the constraint violation degree, the weighted penalty imposed on the solutions that violate the tunnel structure constraints, and the larger the value, the more infeasible the solution; Represents the dynamic adjustment coefficient, which adjusts the penalty intensity according to the event urgency or safety level, and the larger the value, the stricter the penalty; Represents the number of tunnel structure constraints; Represents the th structural constraint function, quantifying the degree of violation of the solution against the physical limitations of the tunnel, and a positive value indicates violation.

[0053] In some examples, the interactive decision support interface outputs the final emergency strategy, which mainly calculates the solution with the minimum Manhattan distance in the Pareto optimal strategy set as the recommended strategy and outputs the final emergency strategy. The present invention adopts the Manhattan distance to emphasize the balance of each objective more, avoiding the extreme optimization of a single objective, and can meet the trade-off requirements of emergency decision-making compared with the Euclidean distance.

[0054] In some examples, the present invention can also dynamically update the feature extraction rule library and the weight parameters of multi-objective optimization according to the execution results of the final emergency strategy; among them, the dynamic update of the feature extraction rule library adopts an incremental reinforcement learning mechanism.

[0055] Embodiment 2: A highway tunnel emergency strategy adaptive generation system, which is used to implement a highway tunnel emergency strategy adaptive generation method as recorded in Embodiment 1, as Figure 2 shown, including a data acquisition module, a feature extraction module, a control and screening module, and a target optimization module.

[0056] Among them, the data acquisition module is used to collect sensor data, video surveillance data, and environmental parameters in the target tunnel to obtain multi-modal data; the feature extraction module is used to layer and extract event basic features, tunnel structure features, and environmental dynamic features from the multi-modal data through the dynamic feature hierarchical extraction module to generate a multi-dimensional feature vector; the control screening module is used to sequentially perform initial screening of event types, secondary screening of severity, and environmental adaptation verification through the control item dynamic screening module based on the multi-dimensional feature vector to screen out candidate control items from the total control item set; the target optimization module is used to perform multi-objective optimization on the candidate control items using the NSGA-Ⅲ algorithm or the improved NSGA-Ⅲ algorithm to generate a Pareto optimal strategy set, and output the final emergency strategy in combination with the interactive decision support interface.

[0057] Working principle: The present invention constructs a multi-dimensional feature vector through the dynamic feature hierarchical extraction module, and combines the three-level feature fusion mechanism of the event basic feature layer, the tunnel structure feature layer, and the environmental dynamic feature layer. Compared with the traditional single-layer feature extraction method, it can effectively improve the accuracy of event type recognition and the timeliness of environmental risk prediction. Through the three-stage verification of the control item dynamic screening module, the screening accuracy of candidate control items is higher, effectively avoiding strategy conflicts and resource waste.

[0058] Those skilled in the art should understand that the embodiments of the present application can be provided as a method, a system, or a computer program product. Therefore, the present application can adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0059] The present application is described with reference to the flowcharts and / or block diagrams of methods, apparatuses (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or block in the flowchart and / or block diagram can be implemented by computer program instructions, and the combination of the processes and / or blocks in the flowchart and / or block diagram can also be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing devices generate means for implementing the functions specified in Figure 1 one process or multiple processes and / or blocks Figure 1 one block or multiple blocks.

[0060] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to operate in a particular manner, such that the instructions stored in the computer-readable memory produce a manufacture including an instruction means that implements the functions specified in one or more of the processes Figure 1 a process or processes and / or blocks Figure 1 specified in a block or blocks.

[0061] These computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process, whereby the instructions executed on the computer or other programmable apparatus provide steps for implementing the functions specified in one or more of the processes Figure 1 a process or processes and / or blocks Figure 1 specified in a block or blocks.

[0062] The specific embodiments described above further elaborate on the objectives, technical solutions, and beneficial effects of the present invention. It should be understood that the above description is only for the specific embodiments of the present invention and is not intended to limit the protection scope of the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present invention shall be included within the protection scope of the present invention.

Claims

1. A method for adaptively generating an emergency response strategy for highway tunnel emergencies, characterized in that, It includes the following steps: Collect sensor data, video surveillance data, and environmental parameters in the target tunnel to obtain multi-modal data; Based on the multi-modal data, layer-by-layer extract event basic features, tunnel structure features, and environmental dynamic features through a dynamic feature hierarchical extraction module to generate a multi-dimensional feature vector; Based on the multi-dimensional feature vector, sequentially perform initial screening of event types, secondary screening of severity, and environmental adaptation verification through a control item dynamic screening module to screen out candidate control items from the total control item set; Use the NSGA-Ⅲ algorithm or the improved NSGA-Ⅲ algorithm to perform multi-objective optimization on the candidate control items, generate a Pareto optimal strategy set, and output the final emergency strategy in combination with an interactive decision support interface.

2. The adaptive generation method of an emergency response strategy for highway tunnel emergencies according to claim 1, wherein, The dynamic feature hierarchical extraction module includes: An event basic feature layer for extracting event basic features including temperature, smoke concentration, CO concentration, vehicle density, and accident point location data; A tunnel structure feature layer for extracting tunnel structure features including tunnel curvature, ventilation shaft distribution, escape route topology, and fire protection facility layout; An environmental dynamic feature layer for extracting environmental dynamic features including wind speed vector field, temperature and humidity gradient matrix, and visibility attenuation rate.

3. A method for adaptively generating an emergency response strategy for highway tunnel emergencies according to claim 1, characterized in that, The initial screening of event types is implemented using a multi-modal fusion classification model; Among them, the modal fusion classification model takes the event basic features and the tunnel structure features as inputs, performs a feature splicing operation on the event basic features and the tunnel structure features, and then applies a multi-classification activation function to output a probability distribution vector of event types.

4. A method for adaptively generating an emergency response strategy for highway tunnel emergencies according to claim 1, characterized in that, The secondary screening of severity is achieved through risk index calculation, specifically including: Solve the reciprocal of time for the environmental dynamic features to obtain the environmental change risk; Normalize the event basic features and then calculate the L2 norm to obtain the global event risk; Determine the risk index based on the sum of the environmental change risk and the environmental change risk.

5. The emergency strategy adaptive generation method for highway tunnel emergencies according to claim 1, characterized in that The environmental adaptation verification includes: Perform fluid mechanics simulation verification on the ventilation system control items; Perform crowd dynamics model verification on the evacuation path planning items; And / or, perform reachability topological analysis on the fire rescue deployment items.

6. A method for adaptively generating an emergency response strategy for highway tunnel emergencies according to claim 1, characterized in that, The objective function of the multi-objective optimization takes the activation state of the candidate control items as the decision variable; The objective function includes: Minimize the time objective, including the coupling effect of the execution time of the control item and the dynamic environmental gradient; Minimize the cost objective, introduce the opportunity cost, and quantify the loss caused by giving up other measures due to enabling the control item; Minimize the risk objective, and use a product form to represent the complexity risk amplification effect brought by the superposition of multiple control items; Constraint conditions, including mutually exclusive control item constraints, dependent control item constraints, and tunnel physical constraints.

7. A method for adaptively generating an emergency response strategy for highway tunnel emergencies according to claim 1, characterized in that, The improvement of the improved NSGA-Ⅲ algorithm includes: Guide the search direction through a dynamic reference point and adaptively divide the hyperplane according to the dimension of the candidate control items; And, adopt a constraint violation degree weighting strategy to punish the solutions that do not meet the tunnel structure constraints.

8. A method for adaptively generating an emergency response strategy for highway tunnel emergencies according to claim 1, characterized in that, The interactive decision support interface outputs the final emergency strategy, including: Calculate the solution with the minimum Manhattan distance in the Pareto optimal strategy set as the recommended strategy and output the final emergency strategy.

9. The adaptive generation method for an emergency response strategy for highway tunnel emergencies according to claim 1, wherein, This method also includes: Dynamically update the feature extraction rule base and the weight parameters of the multi-objective optimization according to the execution result of the final emergency strategy; Among them, the dynamic update of the feature extraction rule base adopts an incremental reinforcement learning mechanism.

10. An emergency strategy adaptive generation system for highway tunnel emergencies, characterized in that, This system is used to implement an adaptive generation method for highway tunnel emergency strategies described in any one of claims 1-9, including: A data acquisition module for collecting sensor data, video surveillance data, and environmental parameters in the target tunnel to obtain multi-modal data; A feature extraction module for layer-by-layer extracting event basic features, tunnel structure features, and environmental dynamic features based on the multi-modal data through a dynamic feature hierarchical extraction module to generate a multi-dimensional feature vector; A control screening module for sequentially performing initial screening of event types, secondary screening of severity, and environmental adaptation verification based on the multi-dimensional feature vector through a control item dynamic screening module to screen out candidate control items from the total control item set; A target optimization module for performing multi-objective optimization on the candidate control items using the NSGA-Ⅲ algorithm or an improved NSGA-Ⅲ algorithm to generate a Pareto optimal strategy set and output a final emergency strategy in combination with an interactive decision support interface.

Images