A stadium intelligent management and control method and system based on digital twinning

Abstract

This invention relates to the field of intelligent management and control technology for sports venues, and discloses a method and system for intelligent management and control of sports venues based on digital twins. The method includes: iteratively predicting the smoke concentration and temperature of each passage in multiple future time windows through a smoke diffusion calculation engine to generate a time-varying traffic capacity function; dividing the network into sub-networks according to physical floors and functional zones, and compressing the time-series prediction sequence of traffic safety levels into boundary potential energy values; each edge computing node solving evacuation paths based on a local time extension network; a central coordinator coordinating boundary flow quotas; spectator mobile terminals collecting displacement speeds to generate congestion perception event reports; edge computing nodes asynchronously decaying traffic capacity and triggering local replanning; and the central coordinator continuously correcting the smoke diffusion calculation engine and writing the results back to the digital twin model.

Description

Technical Field

[0001] This invention relates to the field of intelligent management and control technology for sports venues, specifically to an intelligent management and control method and system for sports venues based on digital twins. Background Technology

[0002] In the event of a fire in a large, multi-story sports stadium, smoke spreads much faster along vertical atriums, stairwells, and ventilation ducts than along horizontal passageways. Existing venue management methods based on digital twins rely solely on statically marking passageways based on current sensor alarm status, lacking prediction of smoke spread trends within future time windows. This can easily lead people to passageways soon to be covered by smoke, causing crowds to return and resulting in secondary congestion.

[0003] After introducing flue gas diffusion time-series prediction, each time slice in the time-extended network is a complete copy of the venue network, with waiting arc connections and capacity transitive dependencies between adjacent slices. Global centralized solutions face the problem of excessive computational latency in venues accommodating tens of thousands of people, and cannot eliminate the temporal dimension coupling constraints through simple spatial decomposition.

[0004] Furthermore, the actual congestion status in video blind spots such as bends and intersections within the venue cannot be perceived by the central control system. The update frequency of crowdsourced perception data does not match the flue gas simulation calculation cycle. The lack of a way for the two types of asynchronous information sources to feed back and integrate into the digital twin model results in evacuation guidance instructions failing to reflect the actual channel status in a timely manner. Summary of the Invention

[0005] This invention provides a smart management and control method and system for sports venues based on digital twins, which solves the technical problems in related technologies such as the inability to coordinate smoke diffusion prediction and evacuation path planning in fire emergency evacuation scenarios, and the lack of dynamic coordination in traffic distribution among distributed computing nodes.

[0006] This invention discloses a smart management and control method for sports venues based on digital twins, which includes at least steps A and B. Step A includes: after receiving a fire alarm signal, iteratively calculating the predicted values ​​of smoke concentration in each channel within multiple future time windows through a smoke diffusion calculation engine, and converting the attributes of each channel edge into a time-varying capacity function. The venue space topology network is divided into multiple sub-network areas according to physical floors and functional zones. The safety time window summary information of each boundary connection edge is encoded into the boundary potential energy value to generate a zone control data package. Step B includes: The partition control data packet is distributed to each edge computing node. Each edge computing node expands a finite-time slice to generate a local time extended network and solves the local minimum cost flow problem to obtain the direction recommendation probability distribution. Each edge computing node reports the actual traffic used by the boundary connection edge to the central coordinator. The central coordinator performs traffic coordination between adjacent sub-networks and issues traffic quotas to update the direction recommendation probability distribution.

[0007] Furthermore, the step of converting the attributes of each channel edge into a time-varying traffic capacity function includes: When the predicted flue gas concentration at the edge of the channel is lower than the safety threshold in the target time window, the passage capacity of the channel edge shall be the physical maximum passage capacity. When the predicted value of the flue gas concentration is between the safety threshold and the danger threshold, the passage capacity of the channel edge is equal to the physical maximum passage capacity multiplied by the linear attenuation factor, where the linear attenuation factor is the ratio obtained by dividing the difference between the danger threshold and the predicted value of the flue gas concentration by the difference between the danger threshold and the safety threshold. When the predicted flue gas concentration exceeds the danger threshold, the passage capacity of the passage side is zero.

[0008] Furthermore, the generation of the security time window summary information includes: For each boundary connection edge between sub-network regions, the access security level of that boundary connection edge in each time window is extracted from the access security level time series prediction sequence; Starting from the current moment, traverse forward along the time axis and count the number of consecutive time windows in which the boundary connecting edge remains passable. Record the duration corresponding to the number of consecutive time windows as the remaining safe duration. Calculate the average traffic capacity within the continuous time window, and record this average as the average available capacity within the safe period; the remaining safe period and the average available capacity together constitute the safe time window summary information of the boundary connecting edge.

[0009] Furthermore, encoding the safety time window summary information of each boundary connection edge into the boundary potential energy value includes: The boundary potential energy value of each boundary connecting edge is equal to the weighted sum of the path distance cost component and the time safety margin cost component; Wherein, the path distance cost component is the dimensionless value of the normalized path distance cost from the boundary connection edge to the nearest exit, and the time safety margin cost component is the dimensionless value of the normalized reciprocal of the remaining safety duration of the boundary connection edge. The two components are multiplied by their respective weighting coefficients and then summed; the larger the boundary potential energy value, the higher the evacuation cost of connecting the edges through that boundary.

[0010] Furthermore, the direction recommendation probability distribution obtained by solving the local minimum cost flow problem includes: Each edge computing node models the evacuation problem in its region as a local minimum cost flow problem with finite time steps. In the local time extension network, the capacity of each channel edge in each time slice is taken from the time-varying traffic capacity function, and each boundary connection edge serves as a sink node with the sinking cost taken from the corresponding boundary potential value. The minimum cost flow algorithm is used to solve the flow value of each edge in each time slice; the flow value allocated to each node on each outgoing edge in each time step is normalized to obtain the direction recommendation probability distribution of each node in each time step. Each component in the direction recommendation probability distribution corresponds to the direction of each adjacent channel of the node.

[0011] Furthermore, the central coordinator performs traffic coordination between adjacent sub-networks, including: The central coordinator compares the sum of the time-series traffic demands of adjacent sub-networks sharing the same boundary connection edge with the physical capacity on that boundary connection edge. When the total demand exceeds the physical capacity, the traffic quota obtained by each sub-network on the boundary connection edge is equal to the physical capacity limit multiplied by the ratio of the composite weight of the sub-network to the sum of the composite weights of all shared sub-networks. The composite weight of each sub-network is the product of the number of people to be evacuated in that sub-network and the urgency level. The urgency level is the reciprocal of the remaining safe time of the passage closest to the fire source within that sub-network. Each edge computing node re-solves the local minimum cost flow problem with the allocated traffic quota as a constraint and updates the direction recommendation probability distribution. The traffic coordination and local re-solution are repeated until the traffic allocation of all boundary connection edges converges.

[0012] Furthermore, it also includes: Each audience member's mobile terminal obtains its current location coordinates through indoor positioning within the venue and requests navigation data from the nearest edge computing node. The edge computing node matches the nearest network node based on the audience member's mobile terminal location and returns the corresponding direction recommendation probability distribution. The audience member's mobile terminal determines the individual evacuation direction based on the direction recommendation probability distribution and renders navigation arrows.

[0013] Furthermore, it also includes: Each audience mobile terminal collects the displacement speed of surrounding audience mobile terminals at a preset frequency. When the number of audience mobile terminals that meet the low-speed condition in the same channel segment reaches a preset lower limit, a congestion perception event report is generated and uploaded to the corresponding edge computing node. The edge computing node obtains the density of audience mobile terminals reporting congestion sensing events within the channel segment marked as congested, calculates the ratio of this density to the designed traffic density of the channel segment, and multiplies the attenuation factor obtained by subtracting this ratio by the current time-varying traffic capacity of the channel segment to obtain the attenuated traffic capacity value. The attenuated traffic capacity value is injected into the capacity parameter of the local cost flow in this region, and the direction recommendation probability distribution is recalculated with the updated capacity parameter in the next solution cycle.

[0014] Furthermore, it also includes: The central coordinator acquires measured data from smoke and temperature sensors in each channel at a preset monitoring cycle, calculates the deviation between the measured smoke concentration and the predicted smoke concentration output by the smoke diffusion calculation engine, and when the deviation exceeds the correction threshold, corrects the boundary condition parameters of the smoke diffusion calculation engine with the measured data, recalculates the traffic safety level time series prediction sequence for subsequent time windows, and regenerates the safety time window summary information of the affected boundary connection edges based on the updated traffic safety level time series prediction sequence, and sends it incrementally to the affected edge computing nodes to trigger local replanning.

[0015] This invention provides a smart management and control system for sports venues based on digital twins, comprising: The smoke prediction and capacity generation module is used to receive fire alarm signals and then iteratively calculate the predicted smoke concentration values ​​of each channel within multiple future time windows through the smoke diffusion calculation engine, and convert the attributes of each channel edge into a time-varying passage capacity function. The subnetwork partitioning and boundary coding module is used to divide the venue space topology into multiple subnetwork areas according to physical floors and functional zones, encode the safety time window summary information of each boundary connection edge into the boundary potential value, and generate a partition control data packet. The distributed pathfinding module, deployed on each edge computing node, is used to receive the partition control data packet, expand finite time slices to generate a local time extended network and solve the local minimum cost flow problem to obtain the direction recommendation probability distribution; The boundary traffic coordination module, deployed in the central coordinator, is used to receive the actual traffic used by the boundary connection edges reported by each edge computing node, perform traffic coordination between adjacent sub-networks, and issue traffic quotas to update the recommended probability distribution of the direction.

[0016] This invention uses a flue gas diffusion calculation engine to iteratively predict the flue gas concentration and temperature of each channel in multiple future time windows, generates a time-series prediction sequence of traffic safety level, and converts it into a time-varying traffic capacity function. This enables evacuation route planning to perceive changes in the safety status of the channel in future time periods, overcoming the problem of insufficient foresight of the channel status caused by relying solely on static marking of the current alarm status of sensors.

[0017] This invention compresses the temporal prediction sequence of access security levels into a summary of the security time windows of each boundary connection edge, and encodes this summary into the boundary potential value. This allows each sub-network region to obtain a local solution with time security guarantees without retaining the complete coupling constraints between adjacent time slices during local solution processing; it only requires expanding the finite time slices within the current security time window. This compression transforms the temporal coupling constraints into spatial boundary cost parameters, making spatial decomposition feasible under the condition of including time security constraints, thereby reducing the computational latency of centralized solutions for large-scale venues.

[0018] This invention collects the displacement speed of surrounding audience mobile terminals at a preset frequency using audience mobile terminals and generates congestion perception event reports. Edge computing nodes attenuate the time-varying traffic capacity of congested channels according to the ratio of congestion perception event report density to channel design density and then inject it into the capacity parameters of the local cost flow. This allows crowdsourced perception data to directly update local solution parameters independently of the flue gas simulation calculation cycle, compensating for the deficiency that the actual congestion status in video coverage blind spots cannot be perceived by the central control system. At the same time, it avoids information delays caused by the mismatch in update frequencies between the two types of asynchronous information sources.

[0019] This invention continuously compares the sensor's measured data with the predicted flue gas concentration through a central coordinator. When the deviation exceeds the correction threshold, the boundary condition parameters of the flue gas diffusion calculation engine are corrected with the measured data, and the updated safety time window summary information is incrementally sent out. At the same time, the distributed calculation results and measured correction data are written back to the digital twin model, ensuring the real-time performance and global consistency of the global control situation mirror. Attached Figure Description

[0020] Figure 1 This is a flowchart of a smart sports venue management and control method based on digital twins provided in an embodiment of the present invention; Figure 2 This is a schematic diagram illustrating the change of the time-varying throughput capacity of channel E-W3 with a time window, provided in an embodiment of the present invention. Figure 3 This is a schematic diagram comparing the boundary potential values ​​of each boundary connection edge provided in an embodiment of the present invention; Figure 4 This is a schematic diagram comparing the remaining safe duration of each boundary connection edge provided in an embodiment of the present invention; Figure 5 This is a schematic diagram of the probability distribution of directional recommendations for sub-network SW3 nodes provided in an embodiment of the present invention; Figure 6 This is a schematic diagram of the flow quota allocation for the boundary connection edge B-SE3-SE4 provided in an embodiment of the present invention; Figure 7This is a schematic diagram of congestion sensing and capacity attenuation in channel segment N-W3-C provided in an embodiment of the present invention; Figure 8 This is a schematic diagram illustrating the deviation between the predicted and measured values ​​of flue gas concentration at node W3-09 provided in this embodiment of the invention. Figure 9 This is a schematic diagram comparing the remaining safe time of the boundary connection edge B-SW3-NW3 before and after correction, provided in an embodiment of the present invention. Detailed Implementation Example 1

[0021] In large, multi-story sports stadiums, smoke spreads much faster along vertical atriums, stairwells, and ventilation ducts than along horizontal passageways during a fire. Existing digital twin-based stadium management methods rely solely on static labeling of passageways based on current sensor alarm states, lacking prediction of smoke spread trends within future time windows. This can easily lead people to passageways soon to be covered by smoke, causing crowd backflow and secondary congestion. Furthermore, when introducing smoke spread time-series prediction, each time slice in the time-extended network is a complete copy of the stadium network, with waiting arc connections and capacity transfer dependencies between adjacent slices. Global centralized solutions face excessive computational latency in stadiums accommodating tens of thousands of people and cannot eliminate temporal coupling constraints through simple spatial decomposition. Moreover, the actual congestion status in video coverage blind spots such as bends and intersections within the stadium cannot be perceived by the central control system, the update frequency of crowdsourced perception data does not match the smoke simulation calculation cycle, and there is a lack of ways to feed back and integrate these two asynchronous information sources into the digital twin model.

[0022] This embodiment provides a smart management and control method for sports venues based on digital twins. The method operates in a distributed computing environment comprising a central coordinator and multiple edge computing nodes. The central coordinator communicates with the server hosting the digital twin model. Each edge computing node is deployed in a physical zone of the venue and communicates with the central coordinator. Each spectator's mobile terminal within the venue supports indoor positioning and communicates with the nearest edge computing node.

[0023] According to an embodiment of the present invention, a smart management and control method for sports venues based on digital twins, such as... Figure 1 As shown, it includes the following steps: Step 1: Predict flue gas diffusion and generate time-varying traffic capacity; Once a fire alarm signal is received, the system acquires fire source location data, venue ventilation system operating parameters, and geometric cross-sectional parameters of each passage from the digital twin model. This data is then input into the smoke diffusion calculation engine, which iteratively calculates the predicted smoke concentration and temperature values ​​for each passage node within multiple future time windows using a preset time step. This generates a time-series prediction sequence of the passage safety level for each passage within each time window.

[0024] It should be noted that the iterative calculation of the aforementioned flue gas diffusion calculation engine refers to determining the initial heat release rate distribution using fire source location data, determining the airflow transmission characteristics of each channel segment using ventilation system operating parameters and geometric cross-sectional parameters of each channel, and progressively solving for the flue gas concentration and temperature of each channel node in each time window according to a preset time step. The preset time step can be a value in the range of 30 seconds to 120 seconds, and the total span of future time windows can be a value in the range of 10 minutes to 30 minutes.

[0025] Based on the time-series prediction sequence of traffic safety levels, the attributes of each channel edge are converted into a time-varying traffic capacity function. The time-varying traffic capacity function is defined as follows: ; in, For the channel edge In the time window Traffic capacity; For the channel edge The physical maximum throughput capacity; For the channel edge In the time window The predicted value of flue gas concentration; The safety threshold represents the upper limit at which the smoke concentration does not affect the passage of people; The danger threshold represents the lower limit at which the flue gas concentration renders the passage completely impassable. . , and All are expressed in the same unit of flue gas concentration, with consistent dimensions for the numerator and denominator, and the ratio is dimensionless. The dimensions of the result after multiplication are the same as same.

[0026] When the predicted flue gas concentration is below the safe threshold, the passage capacity is the maximum physical passage capacity. When the predicted flue gas concentration is within the warning range between the safe and dangerous thresholds, the passage capacity decreases linearly according to the concentration. When the predicted flue gas concentration exceeds the dangerous threshold, the passage capacity is set to zero.

[0027] Simultaneously, the time-sharing reuse capacity limit is marked for each stairwell node. The time-varying passage capacity function and the time-sharing reuse capacity limit are written into the channel attribute layer of the digital twin model.

[0028] Step 2: Subnetwork partitioning and boundary safety time window compression; Based on the venue space topology network in the digital twin model, the venue space topology network is divided into multiple sub-network areas according to physical floors and functional zones.

[0029] Step 2 may specifically include: Step 201: Obtain the floor identifier and functional zone identifier of each node in the venue space topology network, group the nodes in the venue space topology network according to the floor identifier and functional zone identifier, generate multiple sub-network regions, and mark the boundary connection edges between each sub-network region.

[0030] Step 202: For each boundary connection edge between sub-network regions, extract the passage security level of that boundary connection edge in each time window from the passage security level time series prediction sequence. Starting from the current moment, traverse forward along the time axis, count the number of consecutive time windows in which the boundary connection edge remains passable, and calculate the average passage capacity within each consecutive time window. Record the duration corresponding to the number of consecutive time windows as the remaining safe duration, and record the average value as the average available capacity within the safe duration. The remaining safe duration and the average available capacity together constitute the summary information of the safe time window of that boundary connection edge.

[0031] It should be noted that the aforementioned security time window summary information is the process of compressing all time slice information related to the boundary connection edges in the passage security level time series prediction sequence into two scalar values. This compression allows each sub-network region to perceive the time dimension security constraints without retaining a complete temporally extended network copy during local solution, relying solely on the boundary potential value.

[0032] Step 203: Encode the safety time window summary information into the boundary potential values ​​of each sub-network. Boundary potential values The calculation formula is: ; in, Connect edges to the boundary The boundary potential energy value; Connect edges to the boundary The dimensionless value of the path distance cost to the nearest export after mean normalization based on the range; Connect edges to the boundary Remaining safe time The dimensionless value of the reciprocal of the mean normalized based on the range, representing the remaining safe time. Taken from the remaining safe time calculated in step 202; and These are the weighting coefficients for the path distance cost component and the time safety margin cost component, respectively. , Both components are normalized to The interval is then substituted into the formula for calculation to eliminate the influence of dimensional differences on the weighted summation result. The larger the value, the higher the evacuation cost via that boundary connection edge. The local topology of each sub-network region, the time-varying capacity function of each channel edge within the region, and the boundary potential value of each boundary connection edge together constitute the partition control data packet carrying time security constraints.

[0033] Step 3, Solving for distributed local evacuation paths; The partition control data packets carrying time-safety constraints are distributed to the edge computing nodes corresponding to each sub-network region. Each edge computing node performs local evacuation path solving based on the local topology of its sub-network and the boundary potential value carrying the time-safety constraints.

[0034] Step 3 may specifically include: Step 301: Each edge computing node receives a partition control data packet carrying time security constraints, and obtains the time-varying passage capacity function of each channel edge and the boundary potential value of each boundary connection edge within its region.

[0035] Step 302: Each edge computing node models the evacuation problem in its region as a local minimum cost flow problem with finite time steps. Only finite time slices within the current safe time window are expanded to generate a local time-extended network. In the local time-extended network, the capacity of each channel edge in each time slice is taken from the time-varying passage capacity function, and each boundary connection edge serves as a sink node, with its sinking cost taken from the corresponding boundary potential value.

[0036] Step 303: On the local time-extended network, the minimum cost flow algorithm is used to solve for the optimal path and flow allocation from each node to each boundary connection edge. The minimum cost flow algorithm takes as input the set of nodes, the set of edges, the capacity and cost of each edge in the local time-extended network; the output is the flow value of each edge at each time slice. Based on the output flow value, the flow value allocated to each outgoing edge of each node in the local region at each time step is normalized to obtain the direction recommendation probability distribution of each node at each time step. In the direction recommendation probability distribution, each component corresponds to the direction of each adjacent channel of the node, and the larger the component value, the higher the recommendation priority of that direction.

[0037] Step 4: Boundary flow coordination and evacuation guidance output; Each edge computing node reports the actual traffic usage and time step allocation of each boundary connection edge to the central coordinator. The central coordinator performs traffic coordination between adjacent sub-networks.

[0038] Step 4 may specifically include: Step 401: The central coordinator receives the actual traffic and time step allocation data of the boundary connection edges reported by each edge computing node.

[0039] Step 402: The central coordinator compares the total time-series traffic demand of adjacent sub-networks sharing the same boundary connection edge with the physical capacity on that boundary connection edge. When the total demand exceeds the physical capacity, the traffic quota of that boundary connection edge is reallocated based on the number of people to be evacuated and the urgency of each sub-network, and the traffic quota is then distributed to the corresponding edge computing nodes.

[0040] It should be noted that in the above weighted allocation, the sub-network Traffic quotas obtained on the boundary connection edge for: ; in, For sub-networks The traffic quota obtained on the boundary connection edge; This represents the upper limit of the physical capacity of the edge connecting this boundary. For sub-networks The number of people to be evacuated is taken from the number of people currently present in this area reported by the edge computing nodes, in person; For sub-networks The urgency level is calculated as the reciprocal of the remaining safe time along the passage closest to the fire source within the area, expressed as the reciprocal of time. The total number of sub-networks sharing the same boundary connection edge; For sub-networks The number of people to be evacuated; For sub-networks The urgency; The subnetwork index is used for summation and traversal, and its value range is... to . and Multiplying them yields a composite weight, which appears in both the numerator and denominator. The resulting ratio is dimensionless. Multiplying by... Dimensional and Consistent. The shorter the remaining safe time, the better. The larger the value, the greater the weight the corresponding subnetwork receives in traffic quota allocation.

[0041] Step 403: Each edge computing node receives the updated boundary traffic quota, uses the boundary traffic quota as a constraint to resolve the local minimum cost flow problem, updates the direction recommendation probability distribution, and reports the actual traffic usage of the boundary connection edges to the central coordinator again. Steps 402 and 403 are repeated until the traffic allocation of all boundary connection edges converges.

[0042] Step 404: After convergence, each edge computing node stores the final direction recommendation probability distribution in its local cache for use by terminal devices.

[0043] In this embodiment of the application, in order for the evacuation guidance instructions to reach each individual audience member, the following steps are also included: Step 405: Each audience member's mobile terminal obtains its current location coordinates through indoor positioning within the venue and requests navigation data from the nearest edge computing node. The edge computing node matches the audience member's mobile terminal location to the nearest network node and returns the corresponding direction recommendation probability distribution. The audience member's mobile terminal determines the individual evacuation direction based on the direction recommendation probability distribution and renders navigation arrows displayed on the audience member's mobile terminal screen.

[0044] In this embodiment of the application, in order to perceive the actual congestion status within the video coverage blind spot, the following steps are included in addition to step 405: Step 406: Each audience mobile terminal collects the displacement speed of surrounding audience mobile terminals at a preset frequency. When multiple audience mobile terminals within the same channel segment are detected to have moving speeds below a low-speed threshold, a congestion awareness event report is generated. The congestion awareness event report includes the channel segment identifier, the number of audience mobile terminals, and the average moving speed. The congestion awareness event report is then uploaded to the corresponding edge computing node.

[0045] It should be noted that the preset frequency can be once every 5 to 15 seconds. The low-speed threshold can be a preset proportion of normal walking speed. A congestion perception event report is only triggered when the number of audience mobile terminals meeting the low-speed condition within the same passage segment reaches a preset lower limit, in order to filter out false alarms caused by occasional individual pauses.

[0046] In this embodiment of the application, in order to integrate crowdsourced congestion data into the local solution process to correct the traffic capacity parameters, the following steps are also included in step 406: Step 407: Edge computing nodes aggregate congestion-aware event reports within their respective regions. For a passage segment marked as congested, the density of spectator mobile terminals reporting congestion-aware events within that passage segment is obtained. The ratio of this density to the designed traffic density of the passage segment is then used to attenuate the current time-varying traffic capacity of that passage segment, yielding the attenuated traffic capacity value. : ; in, For channel segment In the current time window Traffic capacity after congestion reduction; For channel segment In the current time window Time-varying traffic capacity; For channel segment The density of audience mobile terminals currently reporting congestion sensing events, i.e. the number of audience mobile terminals triggering low-speed conditions within a unit length channel, in units per meter; For channel segment The design traffic density refers to the maximum number of mobile terminals per unit length of the passageway under full-load traffic conditions, expressed in units per meter. . and The units are the same, the ratio is dimensionless, the result in parentheses is dimensionless, and the result is the same as... Dimensions after multiplication Consistent. The range of values ​​is constrained as follows: To ensure the attenuation coefficient Stay Within the interval, thus ensuring .

[0047] Step 408 involves injecting the attenuated capacity value as an asynchronous update message into the capacity parameters of the local cost flow in this region. In the next solution cycle, the edge computing nodes recalculate the direction recommendation probability distribution with the updated capacity parameters, enabling subsequent evacuation guidance to bypass congested passages. Simultaneously, the location and severity of congestion are reported to the central coordinator, triggering a readjustment of traffic quotas in adjacent regions.

[0048] It should be noted that the timing of the asynchronous update message injection is independent of the computation cycle of the flue gas diffusion calculation engine. Edge computing nodes can perform capacity decay and local replanning immediately upon receiving a congestion awareness event report, without waiting for the next round of global computation by the flue gas diffusion calculation engine. This achieves asynchronous integration between the flue gas simulation cycle and the crowdsourced awareness update frequency.

[0049] In this embodiment of the application, in order to ensure the continuous accuracy of flue gas diffusion prediction and the global consistency of the digital twin model, the following steps are also included: Step 409: The central coordinator acquires measured data from the smoke and temperature sensors of each channel at a preset monitoring cycle. The deviation between the measured smoke concentration and the predicted smoke concentration output by the smoke diffusion calculation engine is calculated.

[0050] Step 410: When the deviation exceeds the correction threshold, the boundary condition parameters of the flue gas diffusion calculation engine are corrected using measured data, and the time series prediction sequence of the passage safety level for subsequent time windows is recalculated. Based on the updated time series prediction sequence of the passage safety level, the safety time window summary information of the affected boundary connection edges is regenerated. The updated safety time window summary information is incrementally distributed to the affected edge computing nodes, triggering local replanning of the corresponding sub-network regions.

[0051] Step 411: The evacuation progress data, congestion labeling data and corrected flue gas concentration prediction values ​​collected from each edge computing node are synchronously written into the digital twin model, and a real-time updated global control situation image is output.

[0052] This implementation introduces a flue gas diffusion calculation engine into the digital twin model to iteratively predict the flue gas concentration and temperature of each channel within multiple future time windows, generating a time-series prediction sequence of passage safety levels and converting it into a time-varying passage capacity function. This enables evacuation route planning to perceive changes in the safety status of channels in future periods, avoiding guiding personnel to channels that are about to be covered by flue gas. This overcomes the lack of foresight in channel status caused by relying solely on static snapshot marking of the current alarm status of sensors.

[0053] This implementation compresses the temporal prediction sequence of access security levels into a summary of the security time windows of each boundary connection edge, and encodes this summary into the boundary potential value. This allows each sub-network region to obtain a locally optimal solution with time security guarantees by expanding only the finite time slices within the current security time window, without needing to retain the complete coupling constraints between adjacent time slices during local solution. This compression transforms the coupling constraints in the time dimension into boundary cost parameters in the spatial dimension, making spatial decomposition feasible under the condition of including time security constraints, thereby reducing the computational latency of centralized solutions for large-scale venues.

[0054] This implementation method collects the displacement velocity of surrounding audience mobile terminals at a preset frequency using each audience mobile terminal and generates congestion perception event reports. Edge computing nodes then attenuate the time-varying capacity of the congested passage according to the ratio of the congestion perception event report density to the passage's design density and inject it into the capacity parameters of the local cost flow. This allows crowdsourced perception data to directly update the local solution parameters of the edge computing nodes independently of the flue gas simulation calculation cycle. This asynchronous fusion method compensates for the inability of the central control system to perceive actual congestion in video coverage blind spots, while avoiding information delays caused by mismatched update frequencies between the two asynchronous information sources.

[0055] This implementation method continuously compares the sensor measured data with the predicted flue gas concentration through a central coordinator. When the deviation exceeds the correction threshold, the boundary condition parameters of the flue gas diffusion calculation engine are corrected with the measured data and the updated safety time window summary information is incrementally sent out. At the same time, the distributed calculation results and measured correction data are written back to the digital twin model, ensuring the real-time performance and global consistency of the global control situation mirror.

[0056] The following is an example of an application of the present invention, such as Figure 2-9 As shown, the implementation process is as follows: A large, multi-story sports stadium (hereinafter referred to as "Venue A") has five floors and is designed to accommodate 48,000 spectators. Venue A hosted the finals on the evening of a certain event day in 20XX, with approximately 41,000 spectators present. At 21:17 on the event day in 20XX, a fire broke out in the equipment room on the west side of the third floor of Venue A. The automatic fire alarm system was triggered, and emergency evacuation and control measures were immediately activated. Venue A is equipped with 32 smoke detectors and 28 temperature sensors. The stadium's spatial topology network is divided into 12 sub-network areas, corresponding to 12 edge computing nodes. The mobile terminal access rate for the entire audience was approximately 78%, meaning approximately 31,980 terminals were online.

[0057] Implementation example of step 1 After a fire alarm is triggered, the central coordinator reads the fire source location data (equipment room on the west side of the third floor, coordinate node W3-07), the current operating parameters of the ventilation system (fan speed, damper opening), and the geometric cross-sectional parameters of each channel from the digital twin model. This data is then input into the smoke diffusion calculation engine, which iteratively calculates the predicted smoke concentration and temperature values ​​for each channel node within the next 20 minutes (20 time windows) using a preset time step of 60 seconds.

[0058] Taking the E-W3 core passage on the west side of the third floor as an example, its maximum physical throughput capacity The safety threshold is 180 people per minute. The threshold value is 0.10 (volume fraction, the same below). The value is 0.25. The flue gas diffusion calculation engine outputs the predicted flue gas concentration for this channel in each time window, based on which the time-varying passage capacity is calculated. .

[0059] With time window (For example, 3 minutes after the fire starts) It is within the warning zone. Substitute it into the formula: ; Table 1. Calculation results of time-varying traffic capacity for channel E-W3.

[0060] from Starting from the 4th minute, the time-varying capacity of channel E-W3 was set to zero, and the channel became unavailable for evacuation route planning.

[0061] Implementation example of step 2 Based on the spatial topology of Venue A, the entire venue is divided into 12 sub-network areas according to floor and functional zoning labels. The western area on the third floor is numbered as sub-network SW3, and the eastern area on the third floor is numbered as sub-network SE3. There is a boundary connecting the two areas, B-SW3-SE3.

[0062] In step 202, the time-window-by-time access security level of the boundary connection edge B-SW3-SE3 is extracted from the time-series prediction sequence of access security levels. Starting from the current moment, a forward traversal is performed, and this boundary connection edge... to Maintain continuous passability. It has been converted to an impassable route. Therefore, its remaining safe time is... Minutes, average available capacity within the safe period is People per minute.

[0063] In step 203, the path distance cost for all boundary connecting edges of venue A is calculated. The values ​​are then normalized to the mean based on the range. Taking the boundary connection edge B-SW3-SE3 as an example, the normalized path distance cost is 0.43. The normalized value is 0.81, and the weighting coefficient is taken as... , ,but: ; in, The boundary potential energy value. , These are the normalized values ​​of path distance cost and Weighting coefficients for normalized values.

[0064] Table 2 shows the calculated boundary potential energy values ​​for some boundary connection edges.

[0065] A higher boundary potential energy value indicates a higher evacuation cost via the boundary connection. The boundary potential energy value of B-SW3-SE3 (0.66) is much higher than that of B-NW3-NW2 (0.21), which is consistent with its actual situation of only 3 minutes of remaining safety time and the urgent threat of flue gas. The partition control data packet carrying the above boundary potential energy value is then sent to each edge computing node.

[0066] Implementation example of step 3 After receiving the partition control data packet, each edge computing node uses the boundary potential value of each boundary connection edge as the inflow fee, and only expands a limited time slice within the current safe time window (sub-network SW3 expands 3 time slices, and sub-network SE3 expands 12 time slices) to generate a local time extension network.

[0067] Taking the edge computing nodes of subnetwork SW3 as an example, the local temporal spread network includes all nodes within SW3. to The time-slice replicas are used, and the capacity of each channel edge is taken from the time-varying traffic capacity function. B-SW3-SE3 and B-SW3-SW2 are used as sink nodes, with sinking costs of 0.66 and 0.44, respectively. After solving the minimum cost flow algorithm on this local network, the outgoing edge traffic of each node in SW3 at each time step is normalized to obtain the direction recommendation probability distribution.

[0068] Table 3. Distribution of orientation recommendation probabilities for some nodes within subnetwork SW3.

[0069] The recommended probability for the downward direction (via B-SW3-SW2) is much higher than that for the eastward direction (via B-SW3-SE3), which is consistent with the calculation results that the boundary potential energy value of B-SW3-SW2 (0.44) is lower than that of B-SW3-SE3 (0.66).

[0070] Implementation example of step 4 In steps 401 to 403, the central coordinator receives the actual traffic usage of the boundary connection edges reported by each edge computing node. Taking the boundary connection edge B-SW3-SW2 as an example, the traffic demand of sub-network SW3 is 128 people per minute, and the traffic demand of sub-network SW2 is 47 people per minute, with a total demand of 175 people / minute. This does not exceed the physical capacity limit of 180 people / minute for this boundary connection edge, and therefore no reallocation is required.

[0071] Taking another boundary connecting edge B-SE3-SE4 as an example, the total demand on both sides exceeds the physical capacity limit (220 people per minute, physical capacity limit 200 people / minute). The number of people awaiting evacuation in subnetwork SE3. People, urgency minute Number of people awaiting evacuation in sub-network SE4 People, urgency minute Substitute into the traffic quota formula: ; ; Each edge computing node receives the updated traffic quota and re-solves the local minimum cost flow problem. After two rounds of iteration, the traffic allocation of all boundary connection edges converges.

[0072] Implementation examples of steps 405 to 408 In step 405, a viewer's mobile terminal (terminal ID: TM-8831) in sub-network SW3 determines the node corresponding to its current location W3-12 through indoor positioning, requests navigation data from the edge computing node of SW3, obtains the direction recommendation probability distribution (downward direction probability 0.72), and renders a navigation arrow pointing down the stairs on the mobile terminal screen.

[0073] In step 406, at 21:23, within the N-W3-C section of the north corridor on the third floor (video coverage blind spot), 11 mobile terminals were detected with an average moving speed of 0.31 m / s, which is lower than the low speed threshold (35% of the normal walking speed of 1.2 m / s, i.e., 0.42 m / s), triggering a congestion perception event report, which was uploaded to the NW3 edge computing node.

[0074] In step 407, the NW3 edge computing node aggregates the report to obtain the audience mobile terminal density of channel segment N-W3-C. Number per meter, designed traffic density Units per meter. Calculate the capacity after attenuation: ; Table 4. Congestion perception data and capacity decay results

[0075] In step 408, the NW3 edge computing node injects the attenuated capacity value of 83 people / minute as an asynchronous update message into the local cost flow capacity parameter. In the next solution cycle, it updates the direction recommendation probability distribution, ensuring that subsequent guidance bypasses channel segment N-W3-C. This update is independent of the 60-second calculation cycle of the flue gas diffusion calculation engine and is executed immediately after the congestion perception event report is uploaded.

[0076] Implementation examples of steps 409 to 411 In steps 409 to 410, at 21:31, the central coordinator reads the measured smoke concentration from the smoke sensor on the west side of the third layer and finds that the measured value at node W3-09 is 0.19, while the predicted value of the smoke diffusion calculation engine in the corresponding time window is 0.13, with a deviation of 0.06, exceeding the correction threshold of 0.04. The central coordinator corrects the boundary condition parameters with the measured value and recalculates the time series prediction sequence of the passage safety level for subsequent time windows. The remaining safety duration of the affected boundary connection edge B-SW3-NW3 is corrected from the original prediction of 7 minutes to 4 minutes. The safety time window summary information is sent to the SW3 and NW3 edge computing nodes incrementally, triggering local replanning.

[0077] In step 411, the evacuation progress data, congestion labeling data (channel segment N-W3-C), and corrected flue gas concentration prediction values ​​of the SW3 and NW3 edge computing nodes are synchronously written into the digital twin model, and the global control situation mirror is updated in real time at 21:31.

[0078] Table 5 Summary of Data Input for Digital Twin Model

[0079] The data flow throughout the entire implementation process embodies a complete closed-loop logic. In Step 1, the time-varying flow capacity function output by the flue gas diffusion calculation engine is compressed into boundary potential energy values ​​in Step 2 and then transmitted to each edge computing node. In Step 3, the edge computing nodes solve for the local minimum cost flow based on the boundary potential energy values ​​and output the direction recommendation probability distribution. In Step 4, the central coordinator coordinates the boundary flow and feeds back the updated flow quotas to the edge computing nodes, forming an iterative convergence between distributed and centralized approaches. In Steps 406 to 408, crowdsourced sensing data directly corrects the capacity parameters of the edge computing nodes asynchronously, decoupling them from the flue gas simulation cycle. In Steps 409 to 411, sensor measurement data corrects flue gas prediction deviations, and all distributed computing results are written back to the digital twin model to ensure the consistency of the global control situation mirror.

[0080] The embodiments of the present invention have been described above. However, the embodiments are not limited to the specific implementation methods described above. The specific implementation methods described above are merely illustrative and not restrictive. Those skilled in the art can make more equivalent embodiments under the guidance of the present embodiments, and all of them are within the protection scope of the present embodiments.

Claims

1. A method for intelligent management and control of sports venues based on digital twins, characterized in that, It includes at least steps A and B, wherein step A includes: After receiving the fire alarm signal, the smoke concentration prediction value of each channel in multiple future time windows is calculated iteratively through the smoke diffusion calculation engine, and the attributes of each channel edge are converted into a time-varying passage capacity function. The venue space topology network is divided into multiple sub-network areas according to physical floors and functional zones. The safety time window summary information of each boundary connection edge is encoded into the boundary potential energy value to generate a zone control data package. Step B includes: The partition control data packet is distributed to each edge computing node. Each edge computing node expands a finite-time slice to generate a local time extended network and solves the local minimum cost flow problem to obtain the direction recommendation probability distribution. Each edge computing node reports the actual traffic used by the boundary connection edge to the central coordinator. The central coordinator performs traffic coordination between adjacent sub-networks and issues traffic quotas to update the direction recommendation probability distribution.

2. The intelligent management and control method for sports venues based on digital twins according to claim 1, wherein, The step of converting the attributes of each channel edge into a time-varying traffic capacity function includes: When the predicted flue gas concentration at the edge of the channel is lower than the safety threshold in the target time window, the passage capacity of the channel edge shall be the physical maximum passage capacity. When the predicted value of the flue gas concentration is between the safety threshold and the danger threshold, the passage capacity of the channel edge is equal to the physical maximum passage capacity multiplied by the linear attenuation factor, where the linear attenuation factor is the ratio obtained by dividing the difference between the danger threshold and the predicted value of the flue gas concentration by the difference between the danger threshold and the safety threshold. When the predicted flue gas concentration exceeds the danger threshold, the passage capacity of the passage side is zero.

3. The intelligent management and control method for sports venues based on digital twins according to claim 1, wherein, The generation of the security time window summary information includes: For each boundary connection edge between sub-network regions, the access security level of that boundary connection edge in each time window is extracted from the access security level time series prediction sequence; Starting from the current moment, traverse forward along the time axis and count the number of consecutive time windows in which the boundary connecting edge remains passable. Record the duration corresponding to the number of consecutive time windows as the remaining safe duration. Calculate the average traffic capacity within the continuous time window, and record this average as the average available capacity within the safe period; the remaining safe period and the average available capacity together constitute the safe time window summary information of the boundary connecting edge.

4. The intelligent management and control method for sports venues based on digital twins according to claim 3, wherein, Encoding the safe time window summary information of each boundary connection edge into the boundary potential energy value includes: The boundary potential energy value of each boundary connecting edge is equal to the weighted sum of the path distance cost component and the time safety margin cost component; Wherein, the path distance cost component is the dimensionless value of the normalized path distance cost from the boundary connection edge to the nearest exit, and the time safety margin cost component is the dimensionless value of the normalized reciprocal of the remaining safety duration of the boundary connection edge. The two components are multiplied by their respective weighting coefficients and then summed; the larger the boundary potential energy value, the higher the evacuation cost of connecting the edges through that boundary.

5. The intelligent management and control method for sports venues based on digital twins according to claim 1, wherein, The probability distribution for direction recommendation obtained by solving the local minimum cost flow problem includes: Each edge computing node models the evacuation problem in its region as a local minimum cost flow problem with finite time steps. In the local time extension network, the capacity of each channel edge in each time slice is taken from the time-varying traffic capacity function, and each boundary connection edge serves as a sink node with the sinking cost taken from the corresponding boundary potential value. The minimum cost flow algorithm is used to solve the flow value of each edge in each time slice; the flow value allocated to each node on each outgoing edge in each time step is normalized to obtain the direction recommendation probability distribution of each node in each time step. Each component in the direction recommendation probability distribution corresponds to the direction of each adjacent channel of the node.

6. The intelligent management and control method for sports venues based on digital twins according to claim 1, wherein, The central coordinator performs traffic coordination between adjacent sub-networks, including: The central coordinator compares the sum of the time-series traffic demands of adjacent sub-networks sharing the same boundary connection edge with the physical capacity on that boundary connection edge. When the total demand exceeds the physical capacity, the traffic quota obtained by each sub-network on the boundary connection edge is equal to the physical capacity limit multiplied by the ratio of the composite weight of the sub-network to the sum of the composite weights of all shared sub-networks. The composite weight of each sub-network is the product of the number of people to be evacuated in that sub-network and the urgency level. The urgency level is the reciprocal of the remaining safe time of the passage closest to the fire source within that sub-network. Each edge computing node re-solves the local minimum cost flow problem with the allocated traffic quota as a constraint and updates the direction recommendation probability distribution. The traffic coordination and local re-solution are repeated until the traffic allocation of all boundary connection edges converges.

7. The intelligent management and control method for sports venues based on digital twins according to any one of claims 1 to 6, wherein, Also includes: Each audience member's mobile device obtains its current location coordinates through indoor positioning within the venue and requests navigation data from the nearest edge computing node; The edge computing node is matched to the nearest network node based on the location of the viewer's mobile terminal and returns the corresponding direction recommendation probability distribution; The audience's mobile terminal determines the individual evacuation direction based on the direction recommendation probability distribution and renders navigation arrows.

8. The intelligent management and control method for sports venues based on digital twins according to claim 7, wherein, Also includes: Each audience mobile terminal collects the displacement speed of surrounding audience mobile terminals at a preset frequency. When the number of audience mobile terminals that meet the low-speed condition in the same channel segment reaches a preset lower limit, a congestion perception event report is generated and uploaded to the corresponding edge computing node. The edge computing node obtains the density of audience mobile terminals reporting congestion sensing events within the channel segment marked as congested, calculates the ratio of this density to the designed traffic density of the channel segment, and multiplies the attenuation factor obtained by subtracting this ratio by the current time-varying traffic capacity of the channel segment to obtain the attenuated traffic capacity value. The attenuated traffic capacity value is injected into the capacity parameter of the local cost flow in this region, and the direction recommendation probability distribution is recalculated with the updated capacity parameter in the next solution cycle.

9. The intelligent management and control method for sports venues based on digital twins according to any one of claims 1 to 6, wherein, Also includes: The central coordinator acquires measured data from smoke and temperature sensors in each channel at a preset monitoring cycle, calculates the deviation between the measured smoke concentration and the predicted smoke concentration output by the smoke diffusion calculation engine, and when the deviation exceeds the correction threshold, corrects the boundary condition parameters of the smoke diffusion calculation engine with the measured data, recalculates the traffic safety level time series prediction sequence for subsequent time windows, and regenerates the safety time window summary information of the affected boundary connection edges based on the updated traffic safety level time series prediction sequence, and sends it incrementally to the affected edge computing nodes to trigger local replanning.

10. A smart sports venue management and control system based on digital twins, used to execute the smart sports venue management and control method based on digital twins as described in any one of claims 1 to 9, characterized in that, include: The smoke prediction and capacity generation module is used to receive fire alarm signals and then iteratively calculate the predicted smoke concentration values ​​of each channel within multiple future time windows through the smoke diffusion calculation engine, and convert the attributes of each channel edge into a time-varying passage capacity function. The subnetwork partitioning and boundary coding module is used to divide the venue space topology into multiple subnetwork areas according to physical floors and functional zones, encode the safety time window summary information of each boundary connection edge into the boundary potential value, and generate a partition control data packet. The distributed pathfinding module, deployed on each edge computing node, is used to receive the partition control data packet, expand finite time slices to generate a local time extended network and solve the local minimum cost flow problem to obtain the direction recommendation probability distribution; The boundary traffic coordination module, deployed in the central coordinator, is used to receive the actual traffic used by the boundary connection edges reported by each edge computing node, perform traffic coordination between adjacent sub-networks, and issue traffic quotas to update the recommended probability distribution of the direction.

Images