Intelligent park digital twinning system and method based on artificial intelligence

Abstract

The invention discloses a smart park digital twinning system and method based on artificial intelligence, particularly relates to the technical field of digital twinning simulation optimization, and is used for solving the problems of simulation distortion and low emergency rehearsal reliability caused by insufficient entity behavior dynamic relevance modeling in the prior art. The method comprises the following steps: constructing a propagation path hierarchy analysis network by collecting entity attribute data and dynamic interaction behavior data in real time, dividing groups and individual behavior data based on path length and behavior types, identifying key disturbance nodes, and quantifying propagation path association strength of the key disturbance nodes; a simulation model adjustment strategy is generated through the dynamic change difference and the association strength, the propagation direction and density weight parameters of the key nodes are corrected, and a high-precision simulation result is generated; on the basis of a feedback mechanism of non-key node behavior deviation, a virtual-real linkage closed-loop optimization system is formed by reversely updating a behavior weight through association strength, and the reliability of digital twin simulation in a complex scene is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of digital twin simulation optimization, and more specifically, the present invention relates to an artificial intelligence-based digital twin system and method for smart campuses. Background Art

[0002] In the dynamic management of smart campuses, digital twin technology is often used to simulate the behavioral laws of complex systems such as the flow of people and vehicles to pre-enact emergency strategies. Digital twin technology usually relies on historical data and rule models to simplify and abstract the interaction behaviors of entities (such as pedestrians and vehicles) in large-scale scenarios, so as to generate predictive simulation results. However, the behaviors of entities in the real world are highly dynamic and correlated. For example, individual decisions are affected by multiple factors such as group behavior and environmental feedback, forming a complex non-linear interaction network.

[0003] Currently, during the simulation process, due to the dynamic interaction relationships between entities, local behavior perturbations may trigger global abnormal phenomena through unpredictable chain effects. For example, a path selection deviation in a certain area may be gradually amplified due to the group effect, resulting in unexpected aggregations or path congestions in the simulation results, making the prediction model unable to accurately depict the evolution law of the real scene, thereby reducing the reliability of the pre-enacted strategy. Summary of the Invention

[0004] In order to overcome the above-mentioned defects of the prior art, embodiments of the present invention provide an artificial intelligence-based digital twin system and method for smart campuses to solve the problems raised in the above background art.

[0005] To achieve the above object, the present invention provides the following technical solutions:

[0006] An artificial intelligence-based digital twin method for a smart campus, comprising the following steps:

[0007] S1. Obtain the real-time status data of entity objects in the smart campus, where the real-time status data includes entity attribute data and dynamic interaction data;

[0008] S2. Divide the dynamic interaction data into group behavior data and individual behavior data based on the propagation path length and behavior type of the dynamic interaction data;

[0009] S3. Identify key perturbation nodes that meet the preset perturbation conditions according to the hierarchical and density dynamic change conditions of the propagation path of the group behavior data;

[0010] S4. Generate a dynamic adjustment strategy by combining the dynamic change differences between the entity attribute data and historical behavior data of the key perturbation nodes and evaluating the association strength based on the propagation path levels between the key perturbation nodes and other nodes;

[0011] S5. Adjust the propagation direction and density weight of the key perturbation nodes in the dynamic simulation model based on the propagation path level and association strength to generate the simulation results;

[0012] S6. When the deviation between the behavior data of the non - key perturbation nodes and the simulation results exceeds the preset range, reverse - update the behavior weights of the non - key perturbation nodes based on the association strength of the propagation path level.

[0013] In a preferred embodiment, the entity attribute data includes the position and real - time movement speed, and the dynamic interaction data includes the avoidance behavior, following behavior, and group density distribution among entity objects.

[0014] In a preferred embodiment, based on the propagation path length and behavior type of the dynamic interaction data, the dynamic interaction data is divided into group behavior data and individual behavior data, including:

[0015] Traverse the propagation path length of the dynamic interaction data, and select the dynamic interaction data whose propagation path length exceeds the preset path length threshold;

[0016] Associate the dynamic interaction data whose propagation path length exceeds the preset path length threshold with the behavior type;

[0017] Detect the dynamic interaction data of the following behavior in the behavior type;

[0018] Divide the detected dynamic interaction data corresponding to the following behavior into group behavior data, and the remaining dynamic interaction data into individual behavior data.

[0019] In a preferred embodiment, according to the propagation path level and density dynamic change conditions of the group behavior data, identify the key perturbation nodes that meet the preset perturbation conditions, including:

[0020] Determine the propagation path level of the group behavior data, and the propagation path level is divided according to the number of nodes covered by the influence range of the group behavior data;

[0021] Detect the density change speed of the group behavior data, and the density change speed is calculated by the increase or decrease amplitude of the group density distribution value per unit time;

[0022] Compare the propagation path level with the preset propagation path level range, and compare the density change speed with the preset density change speed range;

[0023] When the propagation path level is within the preset propagation path level range and the density change speed exceeds the preset density change speed range, determine the corresponding node as the key perturbation node.

[0024] In a preferred embodiment, according to the dynamic change difference between the entity attribute data and the historical behavior data of the key perturbation node, and combining the propagation path level evaluation of the association strength between the key perturbation node and other nodes, a dynamic adjustment strategy is generated, including:

[0025] According to the position offset and speed change rate in the entity attribute data of the key perturbation node, calculate the dynamic change difference between the current moment and the historical same-period data, where the position offset is the Euclidean distance difference between the real-time position coordinate and the historical position coordinate, and the speed change rate is the ratio of the real-time speed to the historical speed;

[0026] Determine the propagation path level coverage ratio between the key perturbation node and other nodes, and the propagation path level coverage ratio is the quotient of the propagation path level number of the key perturbation node and the propagation path level number of other nodes;

[0027] Generate a dynamic adjustment strategy according to the product of the dynamic change difference and the propagation path level coverage ratio, and the dynamic adjustment strategy includes the propagation direction offset angle and the density weight allocation ratio;

[0028] When the dynamic change difference exceeds the change threshold of the historical same-period data, adjust the priority of the propagation direction offset angle based on the propagation path level coverage ratio.

[0029] In a preferred embodiment, based on the propagation path level and the association strength, adjust the propagation direction and density weight of the key perturbation node in the dynamic simulation model to generate a simulation result, including:

[0030] Based on the propagation direction offset angle, adjust the moving direction of the key perturbation node in the dynamic simulation model, and the moving direction is calculated according to the vector synthesis of the offset angle and the current propagation direction;

[0031] According to the density weight allocation ratio, adjust the density weight of the key perturbation node in the dynamic simulation model, and the density weight is updated by multiplying the current density value by the allocation ratio;

[0032] Input the adjusted moving direction and density weight into the dynamic simulation model, and perform simulation calculations to generate a simulation result;

[0033] When the density weight of the key perturbation node is updated, synchronously update the density weights of adjacent nodes, and the synchronous update is proportionally allocated according to the propagation path level association strength between the adjacent nodes and the key perturbation node.

[0034] In a preferred embodiment, when the deviation between the behavior data of the non-key perturbation node and the simulation result exceeds the preset range, reversely update the behavior weight of the non-key perturbation node based on the association strength of the propagation path level, including:

[0035] Detect the dynamic deviation rate between the behavior data of non-critical disturbance nodes and the simulation results. The dynamic deviation rate is the ratio of the absolute value of the difference between the real-time behavior data and the simulation behavior data to the average deviation in the same historical period.

[0036] Obtain the set of association strengths of the propagation path levels between non-critical disturbance nodes and all critical disturbance nodes. Each element in the set of association strengths is the coverage ratio of the propagation path levels between a single critical disturbance node and the corresponding non-critical disturbance node.

[0037] Calculate the behavior weight correction ratio of non-critical disturbance nodes according to the maximum value in the set of association strengths and the preset adjustment coefficient.

[0038] When the dynamic deviation rate exceeds the preset deviation threshold, reversely adjust the behavior weight of non-critical disturbance nodes based on the behavior weight correction ratio, and synchronously update the dynamic adjustment strategy priority of the corresponding critical disturbance nodes in the set of association strengths.

[0039] If the adjusted behavior weight still causes the dynamic deviation rate to exceed the preset deviation threshold, recalculate the behavior weight correction ratio according to the second-largest value in the set of association strengths for secondary correction.

[0040] On the other hand, the present invention provides an intelligent park digital twin system based on artificial intelligence, including:

[0041] Data acquisition module: Obtain the real-time status data of entity objects in the intelligent park. The real-time status data includes entity attribute data and dynamic interaction data.

[0042] Classification processing module: Divide the dynamic interaction data into group behavior data and individual behavior data based on the propagation path length and behavior type of the dynamic interaction data.

[0043] Node identification module: Identify critical disturbance nodes that meet the preset disturbance conditions according to the dynamic change conditions of the propagation path levels and density of the group behavior data.

[0044] Strategy generation module: Generate a dynamic adjustment strategy according to the dynamic change difference between the entity attribute data and historical behavior data of the critical disturbance nodes, and combining the evaluation of the association strength of the propagation path levels between the critical disturbance nodes and other nodes.

[0045] Simulation construction module: Adjust the propagation direction and density weight of critical disturbance nodes in the dynamic simulation model based on the propagation path levels and association strengths to generate simulation results.

[0046] Feedback correction module: When the deviation between the behavior data of non-critical disturbance nodes and the simulation results exceeds the preset range, reversely update the behavior weight of non-critical disturbance nodes based on the association strength of the propagation path levels.

[0047] Compared with the prior art, the present invention has the following beneficial effects:

[0048] 1. By collecting entity attribute data and dynamic interaction data in real time and establishing a hierarchical analysis mechanism for the propagation path of dynamic interaction behaviors, it can more accurately reflect the correlation evolution law between group behaviors and individual behaviors; based on the dual classification criteria of propagation path length and behavior type, it can effectively distinguish the impacts of group effects and individual decisions on the system, providing a reliable basis for the identification of local disturbance sources; through the collaborative analysis of the dynamic change differences of key disturbance nodes and the correlation intensity of propagation path levels, a targeted dynamic adjustment strategy is generated, enabling the simulation model parameters to adapt to the non-linear change characteristics of entity behaviors in real time and improving the reliability of evolution law prediction in complex scenarios;

[0049] 2. Through the dynamic deviation analysis of non-critical node behavior data and simulation results, the behavior weights are updated in reverse and the strategy priorities of associated nodes are corrected synchronously to achieve the two-way dynamic adaptation of the simulation model and entity behaviors; this mechanism enables the system to autonomously suppress the abnormal propagation caused by local disturbances, and through the optimization of weight distribution driven by the correlation intensity, ensures that the simulation results continuously approach the evolution trajectory of the real scenario, enhancing the effectiveness of emergency rehearsal strategies and the robustness of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 is a flowchart of a digital twin method for a smart park based on artificial intelligence according to the present invention;

[0051] Figure 2 is a schematic structural diagram of a digital twin system for a smart park based on artificial intelligence according to the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0052] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0053] Embodiment 1: Figure 1 A digital twin method for a smart park based on artificial intelligence according to the present invention is given, which includes the following steps:

[0054] S1. Obtain the real-time status data of entity objects in the smart park, and the real-time status data includes entity attribute data and dynamic interaction data;

[0055] S2. Divide the dynamic interaction data into group behavior data and individual behavior data based on the propagation path length and behavior type of the dynamic interaction data;

[0056] S3. Identify key perturbation nodes that meet the preset perturbation conditions according to the hierarchical level and density dynamic change conditions of the propagation path of the group behavior data;

[0057] S4. Generate a dynamic adjustment strategy by evaluating the association strength in combination with the hierarchical level of the propagation path between the key perturbation nodes and other nodes based on the dynamic change differences between the entity attribute data and the historical behavior data of the key perturbation nodes;

[0058] S5. Adjust the propagation direction and density weight of the key perturbation nodes in the dynamic simulation model based on the hierarchical level of the propagation path and the association strength to generate a simulation result;

[0059] S6. When the deviation between the behavior data of the non-key perturbation nodes and the simulation result exceeds the preset range, reversely update the behavior weight of the non-key perturbation nodes based on the association strength of the hierarchical level of the propagation path.

[0060] S1. Obtain the real-time status data of the entity objects in the smart park. The real-time status data includes entity attribute data and dynamic interaction data, including:

[0061] The entity attribute data includes the location and the real-time movement speed. The dynamic interaction data includes the avoidance behavior, following behavior, and group density distribution between entity objects.

[0062] The location in the entity attribute data is obtained through the global positioning system or Bluetooth beacon positioning device deployed in the smart park. The real-time movement speed is calculated by integrating the acceleration data collected by the inertial sensor set on the entity object. The avoidance behavior in the dynamic interaction data is determined by capturing the movement trajectory of the entity object through the camera deployed in the smart park and identifying the trajectory mutation points. The following behavior is determined by detecting the relative distance change and the movement direction synchronization between entity objects through the infrared sensor or camera. The group density distribution is calculated in real time by counting the number of active devices in the area through the Wi-Fi probe in the smart park and combining the camera image analysis.

[0063] The location and real-time movement speed in the entity attribute data are used for the analysis of the dynamic change differences of entity objects in the subsequent steps. The avoidance behavior and following behavior in the dynamic interaction data are used for the division of group behavior data and individual behavior data in the subsequent steps. The group density distribution is used for the determination of the density dynamic change conditions in the subsequent steps.

[0064] The acquisition frequency of real-time status data is dynamically adjusted according to the motion state of entity objects in the smart park. When the entity object is in a stationary or uniform motion state, a low-frequency acquisition mode is adopted. When the entity object is in an accelerating or turning state, it switches to a high-frequency acquisition mode. During the recognition process of the avoidance behavior of dynamic interaction data, when the camera captures that the motion trajectory of the entity object has an angular deviation or a sudden speed drop exceeding the threshold within a preset time, it is determined that an avoidance behavior has occurred. During the recognition process of the following behavior of dynamic interaction data, when the relative distance between two entity objects remains less than the preset safety distance and the included angle of the motion directions is less than the preset angle, it is determined that a following behavior has occurred. During the real-time calculation process of the population density distribution, the number of active devices counted by the Wi-Fi probe and the number of entity objects recognized by the camera are weighted and fused. When the difference between the two numbers exceeds the preset error range, the recognition result of the camera is preferentially used to correct the density distribution.

[0065] Among them, the method for setting the weighted fusion weights of the number of active devices counted by the Wi-Fi probe and the number of entity objects recognized by the camera is as follows: Based on the consistency of the recording results of the two types of devices in historical statistics, different credibility weights are pre-assigned to the Wi-Fi probe and the camera; Since the camera has shown a higher recognition accuracy in past data, its weight value is slightly higher than that of the Wi-Fi probe, and the sum of the two weights is constantly the complete credibility; When the difference in the number of active devices counted in real time by the two devices is within a reasonable fluctuation range, the system uses the weighted fusion method to comprehensively integrate the data of the two, and the camera data accounts for a larger proportion; The threshold of this fluctuation range is determined by analyzing the typical fluctuation amplitude of the difference between the two in historical data, for example, it can be set to the normal deviation level; If the real-time data difference exceeds this reasonable range, it is determined as an abnormal situation. At this time, the system will fully adopt the recognition result of the camera as the final density value and suspend the fusion calculation of the Wi-Fi probe data to give priority to ensuring the accuracy of the data; This dynamic adjustment mechanism not only ensures the data complementarity in daily situations but also automatically selects a more reliable data source when there are significant differences between devices.

[0066] S2. Based on the propagation path length and behavior type of the dynamic interaction data, the dynamic interaction data is divided into group behavior data and individual behavior data, including:

[0067] The propagation path length of the dynamic interaction data is obtained through the positioning devices deployed in the smart park. The positioning devices include global positioning system receivers or Bluetooth beacons. The propagation path length is calculated according to the movement trajectory of the entity object. The movement trajectory is generated by continuously recording the position data of the entity object. The recording interval of the position data is dynamically adjusted according to the motion state of the entity object. When the entity object is in a stationary or uniform motion state, a low-frequency recording mode is adopted. When the entity object is in an accelerating or turning state, it switches to a high-frequency recording mode.

[0068] The preset path length threshold is set according to the physical space sizes of different functional areas in the smart park. For example, it is set as the first path length threshold in the main road area of the park and the second path length threshold in the passage between buildings. The first path length threshold is greater than the second path length threshold. The specific value of the preset path length threshold is determined by statistically analyzing the average propagation distance of group behaviors in historical data.

[0069] The behavior types of dynamic interaction data are collected by cameras or infrared sensors deployed in the smart park. The determination rules for behavior types include: when the movement trajectory of an entity object shows a direction deviation exceeding a preset angle or a speed change exceeding a preset ratio within a preset time, it is determined as an avoidance behavior; when the relative distance between two entity objects continuously remains less than a preset safety distance and the included angle of the movement directions is less than a preset angle, it is determined as a following behavior.

[0070] The association between dynamic interaction data and behavior types is achieved through data tags. The data tags include timestamps, location coordinates, and behavior type codes. The storage format of the data tags is structured database entries, and each piece of dynamic interaction data corresponds to an independent data tag.

[0071] The detection of dynamic interaction data of following behaviors is achieved by traversing the behavior type codes in the data tags. The behavior type codes are combinations of preset characters or numbers. For example, the code corresponding to the following behavior is "F", and the code corresponding to the avoidance behavior is "A". During the detection process, the data tags with the behavior type code "F" are screened out, and the corresponding dynamic interaction data is marked as group behavior data.

[0072] The division of group behavior data and individual behavior data is completed by a data classifier. The data classifier is a rule-based conditional judgment program. The judgment logic of the data classifier includes: if the dynamic interaction data simultaneously meets the conditions that the propagation path length exceeds the preset path length threshold and the behavior type code is "F", it is classified as group behavior data; if the dynamic interaction data meets the condition that the propagation path length does not exceed the preset path length threshold or the behavior type code is "A", it is classified as individual behavior data.

[0073] The rule priority of the data classifier is set such that the determination of the propagation path length takes precedence over the determination of the behavior type. When the propagation path length of the dynamic interaction data exceeds the preset path length threshold but the behavior type code is "A", it is preferentially determined as group behavior data based on the propagation path length.

[0074] The divided group behavior data and individual behavior data are respectively stored in independent database partitions. The group behavior database partition is used for the identification of key perturbation nodes in subsequent steps, and the individual behavior database partition is used for the feedback correction of non-key perturbation nodes in subsequent steps.

[0075] The update period of the preset path length threshold is set according to the change frequency of the crowd density in the smart park. When it is detected that the change in the number of entity objects in the same area exceeds the preset density threshold, the recalculation of the preset path length threshold is triggered, and the recalculation process is based on the statistical distribution of the propagation path lengths of group behaviors in the current real-time data.

[0076] The preset safety distance in the following behavior detection process is set according to ergonomic standards. The value range of the preset safety distance is from 0.5 meters to 1.5 meters, and the specific value is determined by statistically analyzing the average following distance of entity objects in historical data. During the process of dividing dynamic interaction data, if the same entity object has both following behavior and avoidance behavior within a continuous time window, the final classification is determined according to the behavior type with a higher duration ratio within the time window. The duration ratio is obtained by calculating the sum of the time segments of different behavior types and the total duration.

[0077] The judgment logic of the data classifier is implemented through a programmable controller during the implementation process. The parameter configuration interface of the programmable controller allows the administrator to adjust the preset path length threshold and the behavior type determination rules according to the actual scenario. After the parameters are adjusted, the data classifier automatically reloads the configuration and performs data division.

[0078] The divided group behavior data and individual behavior data are displayed through a data visualization tool. The data visualization tool marks the group behavior data with the first color and the individual behavior data with the second color, and the data of different colors are superimposed and displayed on the digital twin map of the smart park.

[0079] The division results of the dynamic interaction data are recorded in a log file. The log file includes the division time, the number of data entries, and a summary of the classification basis. The storage path of the log file is associated with the database partition path, and the access rights of the log file are set according to the administrator role level.

[0080] S3. According to the hierarchical level and density dynamic change conditions of the propagation path of the group behavior data, identify the key disturbance nodes that meet the preset disturbance conditions, including:

[0081] The determination of the propagation path level is achieved by statistically analyzing the number of nodes affected by the group behavior data. The nodes are physical areas or devices with independent identifiers in the smart park. The physical areas include entrances and exits, intersection points of passages, or gathering areas, and the devices include cameras or sensors. The number of nodes covered by the influence range is calculated by accumulating the number of nodes passed by tracking the propagation path of the group behavior data.

[0082] The propagation path of group behavior data is generated by superimposing the movement trajectories of entity objects. The movement trajectory is a sequence of positions of the entity object within a preset time period, and the sampling interval of the position sequence is dynamically adjusted according to the movement speed of the entity object. The faster the movement speed of the entity object, the shorter the sampling interval.

[0083] The calculation of the density change speed is achieved by recording the group density distribution values within adjacent time windows and calculating the difference. The length of the time window is set according to the fluctuation frequency of the people flow in the smart park. When the people flow fluctuates frequently, a shorter time window is adopted; when the people flow fluctuates gently, a longer time window is adopted.

[0084] The preset propagation path level range is set according to the statistical distribution of the group behavior propagation paths in the historical data. For example, when the group behavior propagation paths in the historical data are concentrated in the first level to the third level, the preset propagation path level range is set to the first level to the third level. The first level corresponds to affecting 1 to 3 nodes, the second level corresponds to affecting 4 to 6 nodes, and the third level corresponds to affecting 7 to 9 nodes.

[0085] The preset density change speed range is determined by analyzing the mutation threshold of the group density in the historical data. For example, when the density increase rate per unit time exceeds 10% or the density decrease rate exceeds 15%, it is determined that the density change speed exceeds the preset range.

[0086] The comparison between the propagation path level and the preset propagation path level range is achieved by traversing the propagation path levels of the current group behavior data and checking whether it is between the preset minimum level and the maximum level. For example, when the propagation path level is the second level and the preset range is from the first level to the third level, it is determined to meet the conditions.

[0087] The comparison between the density change speed and the preset density change speed range is achieved by calculating the absolute value of the current density change speed and comparing it with the preset threshold. For example, when the density change speed reaches an increase of 12% per minute, it exceeds the preset threshold of an increase of 10% per minute, and it is determined to exceed the preset range.

[0088] The determination logic of the key disturbance node is that when and only when the propagation path level is within the preset level range and the density change speed exceeds the preset speed range, the corresponding node is marked as a key disturbance node. For example, when the propagation path level of a certain area is the second level and the density change speed is an increase of 12% per minute, it is determined that this area is a key disturbance node.

[0089] The update of the preset propagation path level range is achieved by periodically analyzing the propagation path level distribution of the latest group behavior data. The update period is set according to the activity type in the smart park. For example, it is updated once a day during large-scale events and once a week during regular operations.

[0090] The dynamic adjustment of the preset density change speed range is achieved by monitoring the variance of the density change in real-time data. When the variance continuously exceeds the historical variance mean, the preset density change speed range is automatically expanded. For example, increasing by 10% per minute is adjusted to increasing by 12% per minute. The marking information of the key disturbance nodes includes node identification, propagation path level, density change speed, and determination timestamp, and the marking information is stored in an independent key node database partition.

[0091] During the calculation of the propagation path level, if a certain node is simultaneously affected by multiple group behavior data, the propagation path level is calculated by weighting according to the number of group behavior data affecting the node. For example, when a certain node is affected by three group behavior data, the propagation path level is the average value of the levels of these three group behavior data.

[0092] During the calculation of the density change speed, if the group density distribution value within a certain time window is missing due to abnormal data collection, linear interpolation of the density distribution values of the previous and subsequent time windows is used for supplementation, and the weights of the linear interpolation are allocated according to the ratio of the time distance between the missing window and the front and back.

[0093] The association relationship between the preset propagation path level range and the preset density change speed range is achieved by establishing a two-dimensional decision matrix. The rows of the two-dimensional decision matrix correspond to the propagation path level, the columns correspond to the density change speed, and the matrix elements are the decision results (key nodes or non-key nodes). The parameters of the two-dimensional decision matrix are trained according to the occurrence frequency of key nodes in historical data.

[0094] The decision result of the key disturbance node is marked with a highlighted color block on the digital twin map through a visualization tool, and the color of the color block is displayed in grades according to the severity of the combination of the propagation path level and the density change speed. For example, red corresponds to nodes with a high level and a fast speed, and orange corresponds to nodes with a medium level and a medium speed.

[0095] The decision log record of the key disturbance node includes the decision time, node identification, propagation path level, density change speed, and decision basis. The log file is indexed by the timestamp and associated with the simulation result database, supporting retrieval by time range or node identification.

[0096] During the decision process of the key disturbance node, if a certain node is determined to be a key node in multiple consecutive time windows, the early warning mechanism is triggered, and the early warning level increases according to the number of consecutive decision time windows. For example, when a key node is determined in three consecutive time windows, a first-level early warning is triggered, and when five consecutive time windows are determined, a second-level early warning is triggered.

[0097] The initial values of the preset propagation path level range and the preset density change speed range are configured through the administrator interface, and the administrator interface provides a slider or a numeric input box for the user to adjust.

[0098] The determination result of the critical disturbance node is transmitted to the emergency management system of the smart park through the application programming interface, and the emergency management system automatically generates evacuation routes or resource scheduling instructions according to the location and severity of the critical nodes.

[0099] S4. Based on the dynamic change differences between the entity attribute data and historical behavior data of the critical disturbance node, and combining the propagation path level evaluation of the association strength between the critical disturbance node and other nodes, generate a dynamic adjustment strategy, including:

[0100] The calculation of the position offset in the entity attribute data of the critical disturbance node is achieved through the Euclidean distance difference between the real-time position coordinates and the historical position coordinates. The calculation formula of the Euclidean distance difference is: ; where represents the Euclidean distance difference, and represent the real-time position coordinates of the critical disturbance node at the current moment, and represent the position coordinates of the same node in the historical data of the same period. The historical data of the same period is the average position data within the past consecutive preset days in the same time period. For example, the average value of the position coordinates at the same moment in the past 7 days is selected as the historical data.

[0101] The calculation of the rate of change of speed is achieved through the ratio of the real-time speed to the historical speed. The real-time speed is the instantaneous speed value collected by the inertial sensor at the current moment, and the historical speed is the average speed value of the same node in the historical data of the same period.

[0102] The statistical period of the historical speed is the same as that of the historical position coordinates. The calculation of the coverage ratio of the propagation path level is achieved through the quotient of the number of propagation path levels of the critical disturbance node and the number of propagation path levels of other nodes. The number of propagation path levels is determined according to the propagation path level identified in step S3. For example, if the number of propagation path levels of the critical disturbance node is 5 and the number of propagation path levels of other nodes is 3, then the coverage ratio is 5 / 3.

[0103] The generation of the dynamic adjustment strategy is achieved through the product of the dynamic change difference and the coverage ratio of the propagation path level. The product calculation formula is: ; where is the coverage ratio of the propagation path level, is the strategy weight coefficient, is the rate of change of speed.

[0104] The strategy weight coefficient is used to calculate the propagation direction offset angle and the density weight distribution ratio. The propagation direction offset angle is calculated by multiplying the strategy weight coefficient by the preset reference angle. The preset reference angle is set according to the physical structure of the path layout in the smart park. For example, the reference angle is set to 30 degrees in a straight channel and 45 degrees in a curved channel.

[0105] The density weight allocation ratio is calculated through the ratio of the strategy weight coefficient to the preset benchmark density. The preset benchmark density is set according to the average group density distribution value of the same area in historical data. For example, when the strategy weight coefficient is 1.8 and the benchmark density is 50 people / square meter, the density weight allocation ratio is 1.8 / 50=0.036.

[0106] When the dynamic change difference exceeds the change threshold of the historical data for the same period, the change threshold of the historical data for the same period is determined by counting the 90% quantile of the dynamic change difference in the past preset period. For example, when the 90% quantile of the dynamic change difference in the past 30 days is 1.5, the change threshold is set to 1.5.

[0107] The rule for priority adjustment is: when the dynamic change difference exceeds the change threshold, the calculation weight of the propagation direction offset angle is increased to the preset priority coefficient. The preset priority coefficient is configured through the administrator interface. For example, if the priority coefficient is set to 2, the propagation direction offset angle is adjusted to twice the original calculated value.

[0108] During the generation of the dynamic adjustment strategy, if the propagation path level coverage ratio of the key disturbance node is less than 1, the density weight allocation ratio is limited to not exceed the preset upper limit value. The preset upper limit value is set according to the maximum value of the density weight in the historical data. For example, when the historical maximum density weight is 0.05, the upper limit value is set to 0.05.

[0109] The calculation results of the strategy weight coefficient are recorded in a log file, which includes the calculation time, key disturbance node identification, strategy weight coefficient and priority adjustment flag. The log file is indexed by timestamp and associated with the simulation result database.

[0110] The propagation direction offset angle and density weight distribution ratio are displayed on the digital twin map through visualization tools as vector arrows and color block transparency. The length of the vector arrow indicates the size of the offset angle, and the transparency of the color block indicates the density weight.

[0111] The effective period of the dynamic adjustment strategy is dynamically set according to the density change rate of the key disturbance nodes. The faster the density changes, the shorter the effective period. For example, when the density change rate exceeds 10% per minute, the effective period is shortened to 10 seconds.

[0112] The update of the preset reference angle and the preset reference density is achieved by periodically analyzing the latest data. The update period is synchronized with the update period of the preset disturbance condition in step S3, for example, it is updated once a week. The parameter configuration interface of the dynamic adjustment strategy allows the administrator to manually adjust the preset reference angle, the preset reference density, and the priority coefficient. After the parameters are adjusted, the strategy is automatically recalculated and the simulation model is updated.

[0113] S5. Adjust the propagation direction and density weight of the key disturbance nodes in the dynamic simulation model based on the propagation path level and the association strength, and generate simulation results, including:

[0114] The adjustment of the movement direction of the key disturbance nodes in the dynamic simulation model is achieved by the vector synthesis of the propagation direction offset angle and the current propagation direction. The current propagation direction is the real-time movement direction of the entity object in the dynamic simulation model, and the propagation direction offset angle is the parameter in the dynamic adjustment strategy generated in step S4. The calculation process of the vector synthesis is to superimpose the unit vector corresponding to the current propagation direction and the rotation vector corresponding to the offset angle, and the direction of the superimposed vector is used as the adjusted movement direction.

[0115] The adjustment of the density weight is achieved by multiplying the current density value by the density weight distribution ratio generated in step S4. The current density value is the real-time population density distribution value in the area where the key disturbance node is located in the dynamic simulation model, and the density weight distribution ratio is calculated based on the product of the dynamic change difference and the propagation path level coverage ratio in step S4. For example, when the current density value is 50 people per square meter and the distribution ratio is 0.036, the updated density weight is 50×0.036 = 1.8.

[0116] After the adjusted movement direction and density weight are input into the dynamic simulation model, the dynamic simulation model recalculates the movement trajectory and density distribution of the entity object according to the updated parameters, and generates simulation results. The output format of the simulation results is a data sequence including a timestamp, position coordinates, and density values.

[0117] The synchronous update of the density weights of adjacent nodes is achieved by proportional distribution according to the association strength. The association strength is the propagation path level coverage ratio calculated in step S4, and the synchronous update ratio is the product of the association strength and the preset ratio range. The preset ratio range is set according to the density fluctuation threshold of adjacent nodes in historical data. For example, when the association strength is 1.5 and the preset ratio range is 0.1, the synchronous update ratio is 1.5×0.1 = 0.15, and the density weight of the adjacent node is updated to the current density value×0.15.

[0118] During the vector synthesis process of the propagation direction offset angle and the current propagation direction, if the offset angle exceeds the preset angle upper limit, the offset angle is limited. The preset angle upper limit is set according to the maximum allowable turning angle of the path in the smart park. For example, at a right-angle turn, the upper limit is set to 90 degrees, and the excess part is calculated as 90 degrees.

[0119] During the application process of the density weight distribution ratio, if the updated density weight exceeds the maximum carrying threshold of the dynamic simulation model, the weight correction mechanism is triggered. The maximum carrying threshold is set according to the design parameters of the dynamic simulation model. For example, the maximum carrying threshold is 2.0, and when it is exceeded, it is forced to be corrected to 2.0.

[0120] The synchronous update range of adjacent nodes is determined according to the number of levels of the propagation path of the key disturbance node. The higher the number of levels of the propagation path, the more adjacent nodes are synchronously updated. For example, when the number of levels of the propagation path is 3, the 3 nodes directly adjacent to the key disturbance node are synchronously updated.

[0121] The effective timing of the adjusted moving direction in the dynamic simulation model is set according to the simulation time step. The simulation time step is synchronized with the real-time data acquisition frequency. For example, if the real-time data is collected once per second, the simulation time step is set to 1 second. The update result of the density weight is recorded through a log file. The log file includes the update time, node identifier, original density value, updated density value, and synchronous update ratio. The storage path of the log file is associated with the simulation result database.

[0122] The priority of synchronous update of adjacent nodes is sorted according to the magnitude of the association strength. Adjacent nodes with a higher association strength are updated first. For example, a node with an association strength of 2.0 is updated before a node with an association strength of 1.0.

[0123] During the recalculation process of the dynamic simulation model, if the deviation between the simulation result and the real-time data exceeds the preset fault tolerance threshold, the model parameter rollback mechanism is triggered. The preset fault tolerance threshold is set according to the statistical distribution of historical simulation errors. For example, when the deviation exceeds the 95% quantile of the historical error, it is rolled back to the parameter state of the previous time step.

[0124] During the vector synthesis calculation process of the propagation direction offset angle, if the current propagation direction is a zero vector, the direction corresponding to the offset angle is directly used as the adjusted moving direction.

[0125] The preset ratio range of the density weight distribution ratio is configured through the administrator interface. The administrator interface provides input boxes for the minimum and maximum values of the ratio range, and the input value is limited to a decimal between 0 and 1. For example, the ratio range is set to 0.01 to 0.05.

[0126] The effective condition of the synchronous update ratio is dynamically adjusted according to the density change speed of adjacent nodes. The faster the density changes, the lower the effective ratio. For example, when the density change speed exceeds 15% per minute, the effective ratio is reduced to 50% of the original ratio.

[0127] The output simulation results of the dynamic simulation model are displayed through a visualization interface. The visualization interface displays the adjusted movement direction as an arrow icon superimposed on the digital twin map. The length and direction of the arrow correspond to the adjusted movement direction parameters. The density weight is presented in a color gradient heat map, and the color depth corresponds to the density weight.

[0128] It is worth noting that the dynamic simulation model is a digital twin model, which is a simulation program based on real-time status data and dynamic adjustment strategies. It receives the propagation direction offset angle and density weight distribution ratio generated in step S4 as input parameters, and simulates the motion trajectory and group density distribution changes of the physical object by updating the moving direction vector and density weight value of the key disturbance node, and outputs a simulation result data sequence containing timestamps, position coordinates and density values.

[0129] S6. When the deviation between the behavior data of the non-critical disturbance node and the simulation result exceeds a preset range, the behavior weight of the non-critical disturbance node is reversely updated based on the correlation strength of the propagation path level, including:

[0130] The dynamic deviation rate is calculated by dividing the absolute value of the difference between the real-time behavior data and the simulated behavior data by the average deviation over the same period in history. The average deviation over the same period in history is the average absolute value of the deviation of the same node in the same time period over the past preset number of consecutive days. For example, the average deviation value from 9:00 to 10:00 in the morning of the past 30 days is used as the historical benchmark.

[0131] The acquisition of the correlation strength set of the propagation path level is achieved by traversing all the key disturbance nodes and calculating the propagation path level coverage ratio of each key disturbance node and the non-key disturbance node. The propagation path level coverage ratio is the quotient of the number of levels of a single key disturbance node generated in step S4 divided by the number of levels of non-key disturbance nodes. For example, when the key disturbance node level is 5 and the non-key disturbance node level is 2, the coverage ratio is 5 / 2=2.5.

[0132] The behavior weight correction ratio is calculated by multiplying the maximum value in the association strength set by the preset adjustment coefficient. The preset adjustment coefficient is set according to the effectiveness statistics of the correction ratio in the historical data. For example, when the correction ratio in the historical data is 0.8, the deviation suppression effect is best, so the preset adjustment coefficient is set to 0.8.

[0133] The preset deviation threshold of the dynamic deviation rate is determined by analyzing the 90% quantile of the dynamic deviation rate in historical data. For example, when the 90% quantile of the historical dynamic deviation rate is 1.2, the preset deviation threshold is set to 1.2.

[0134] The formula for reverse adjustment of behavior weights is: ;in, is the adjusted behavior weight, is the current behavior weight, The behavior weight correction ratio.

[0135] The priority of the dynamic adjustment strategy for synchronously updating the key disturbance nodes is achieved by reducing the priority coefficient of the corresponding key disturbance node. The reduction of the priority coefficient is 50% of the correction ratio. For example, when the correction ratio is 0.8, the priority coefficient is reduced by 0.4.

[0136] If the adjusted behavior weight still causes the dynamic deviation rate to exceed the preset deviation threshold, the correction ratio is recalculated based on the second largest value in the association strength set. The formula for the second correction is: ;in, is the secondary correction ratio; is the first revision ratio; is the secondary correction attenuation coefficient, which is set to 0.5, for example.

[0137] The secondary correction attenuation coefficient is set according to the success rate of the secondary correction in the historical data. For example, when the historical secondary correction success rate is 60%, the secondary correction attenuation coefficient is set to 0.6.

[0138] The elements in the association strength set are sorted from large to small. If the same coverage ratio appears during the sorting process, they are sorted according to the density change speed of the key disturbance nodes. The nodes with higher density change speed have higher priority. The secondary correction results of the behavior weight are recorded in the log file. The log file includes the correction time, node identification, first correction ratio, secondary correction ratio and final behavior weight. The log file storage path is associated with the simulation result database of step S5.

[0139] The reverse-adjusted behavior weights are displayed as numerical labels superimposed on the digital twin map through a visualization tool. The color of the numerical labels changes gradually according to the size of the correction ratio. For example, when the correction ratio is greater than 0.5, it is displayed in red, and when it is less than 0.3, it is displayed in green.

[0140] The synchronous update result of the priority of the key disturbance node is transmitted to the dynamic adjustment strategy of step S4, triggering the reloading and effectiveness of the strategy parameters. During the calculation of the dynamic deviation rate, if the historical average deviation over the same period is zero, the preset minimum deviation value is used instead. The preset minimum deviation value is set according to the minimum resolution of the dynamic simulation model, for example, the minimum deviation value is set to 0.1.

[0141] During the application process of the behavior weight correction ratio, if the correction ratio exceeds the preset upper limit, the correction ratio is limited. The preset upper limit is set according to the maximum valid value of the correction ratio in historical data. For example, the upper limit is set to 0.9. The update period of the association strength set is synchronized with the data collection period of non-critical disturbance nodes. For example, the data is collected every 5 seconds, and the association strength set is updated every 5 seconds.

[0142] The triggering condition for secondary correction also includes checking whether the second-largest value in the association strength set is greater than the preset association strength threshold. The preset association strength threshold is set according to the statistical distribution of valid association strengths in historical data. For example, secondary correction is allowed when the second-largest value is greater than 1.0. The effective timing of the behavior weight after reverse adjustment in the dynamic simulation model is synchronized with the simulation time step. For example, if the simulation time step is 1 second, the behavior weight is updated every 1 second.

[0143] Embodiment 2: Figure 2 A structural schematic diagram of a digital twin system for a smart park based on artificial intelligence according to the present invention is given. A digital twin system for a smart park based on artificial intelligence includes:

[0144] Data acquisition module: Obtain the real-time status data of entity objects in the smart park. The real-time status data includes entity attribute data and dynamic interaction data;

[0145] Classification processing module: Based on the propagation path length and behavior type of the dynamic interaction data, divide the dynamic interaction data into group behavior data and individual behavior data;

[0146] Node recognition module: Identify key disturbance nodes that meet the preset disturbance conditions according to the dynamic change conditions of the propagation path level and density of the group behavior data;

[0147] Strategy generation module: Generate a dynamic adjustment strategy by combining the dynamic change differences between the entity attribute data and historical behavior data of the key disturbance nodes, and evaluating the association strength based on the propagation path levels between the key disturbance nodes and other nodes;

[0148] Simulation construction module: Adjust the propagation direction and density weight of the key disturbance nodes in the dynamic simulation model based on the propagation path level and association strength to generate a simulation result;

[0149] Feedback correction module: When the deviation between the behavior data of non-critical disturbance nodes and the simulation result exceeds the preset range, reverse-update the behavior weights of non-critical disturbance nodes based on the association strength of the propagation path level.

[0150] The above formulas are all dimensionless and only take their numerical values for calculation. The formulas are obtained by collecting a large amount of data and performing software simulation to obtain a formula that is closest to the actual situation. The preset parameters and threshold selection in the formulas are set by those skilled in the art according to the actual situation.

[0151] It should be noted that the present invention can be deployed on the device itself to achieve embedded applications, or can also run on a PC or other terminals with a user interface, so as to meet various hardware environments and usage requirements.

[0152] The above embodiments can be implemented in whole or in part by software, hardware, firmware, or any other combination. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the processes or functions described in the embodiments of the present application are generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center by wire (such as infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium that the computer can access, or a data storage device such as a server or data center that contains one or more collections of available media. The available medium can be a magnetic medium (such as a floppy disk, hard disk, magnetic tape), an optical medium (such as a DVD), or a semiconductor medium. The semiconductor medium can be a solid-state drive.

[0153] Those skilled in the art can clearly understand that for the convenience and simplicity of description, the specific working processes of the systems, devices, and modules described above can refer to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0154] In the several embodiments provided in the present application, it should be understood that the disclosed systems, devices, and methods can be implemented in other ways. For example, the device embodiments described above are only illustrative. For example, the division of the modules is only a logical function division, and there can be other division methods in actual implementation. For example, multiple modules or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the displayed or discussed couplings or direct couplings or communication connections to each other can be through some interfaces, and the indirect couplings or communication connections of the devices or modules can be in electrical, mechanical, or other forms.

[0155] The module described as a separation component may or may not be physically separated. The component shown as a module may or may not be a physical module. It may be located in one place or distributed across multiple network modules. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0156] In addition, in each embodiment of the present application, each functional module may be integrated into one processing module, may exist separately as individual physical modules, or two or more modules may be integrated into one module.

[0157] If the above-mentioned function is implemented in the form of a software functional module and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, in essence, or the part that contributes to the prior art, or a part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in each embodiment of the present application. The aforementioned storage medium includes: various media that can store program codes, such as USB flash drives, mobile hard disks, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical discs.

[0158] As described above, this is only the specific implementation manner of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present application can easily think of changes or substitutions, which should all be covered by the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

[0159] Finally: The above are only the preferred embodiments of the present invention and are not used to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. An artificial intelligence-based digital twin method for smart campuses, characterized in that, The steps include: S1. Obtain real-time status data of entity objects in the smart park, including entity attribute data and dynamic interaction data; S2. Based on the propagation path length and behavior type of dynamic interaction data, the dynamic interaction data is divided into group behavior data and individual behavior data; S3. According to the propagation path level and density dynamic change conditions of the group behavior data, identify the key disturbance nodes that meet the preset disturbance conditions; S4. Based on the dynamic change difference between the entity attribute data and the historical behavior data of the key disturbance node, the correlation strength is evaluated by combining the propagation path level between the key disturbance node and other nodes to generate a dynamic adjustment strategy; S5. Adjust the propagation direction and density weight of key disturbance nodes in the dynamic simulation model based on the propagation path level and correlation strength to generate simulation results; S6. When the deviation between the behavior data of the non-critical disturbance node and the simulation result exceeds a preset range, the behavior weight of the non-critical disturbance node is reversely updated based on the correlation strength of the propagation path level.

2. The digital twin method for a smart park based on artificial intelligence according to claim 1, wherein Entity attribute data includes location and real-time movement speed, and dynamic interaction data includes avoidance behavior, following behavior and group density distribution between entity objects.

3. The digital twin method for an intelligent park based on artificial intelligence according to claim 2, wherein, Based on the propagation path length and behavior type of dynamic interaction data, dynamic interaction data is divided into group behavior data and individual behavior data, including: Traversing the propagation path length of the dynamic interaction data, and selecting the dynamic interaction data whose propagation path length exceeds a preset path length threshold; Associating dynamic interaction data whose propagation path length exceeds a preset path length threshold with a behavior type; Detect dynamic interaction data of follow-up behaviors in behavior types; The dynamic interaction data corresponding to the detected following behavior is divided into group behavior data, and the remaining dynamic interaction data is divided into individual behavior data.

4. The digital twin method for an intelligent park based on artificial intelligence according to claim 3, characterized in that, According to the propagation path level and density dynamic change conditions of group behavior data, identify the key disturbance nodes that meet the preset disturbance conditions, including: Determine the propagation path level of group behavior data, which is divided according to the number of nodes covered by the influence range of group behavior data; Detect the density change rate of group behavior data. The density change rate is calculated by the increase or decrease of the group density distribution value per unit time; Comparing the propagation path level with a preset propagation path level range, and comparing the density change speed with a preset density change speed range; When the propagation path level is within a preset propagation path level range and the density change speed exceeds a preset density change speed range, the corresponding node is determined to be a key disturbance node.

5. A digital twin method for a smart park based on artificial intelligence according to claim 4, characterized in that, According to the dynamic change difference between the entity attribute data and historical behavior data of the key disturbance node, combined with the propagation path level evaluation of the key disturbance node and other nodes, a dynamic adjustment strategy is generated, including: According to the position offset and speed change rate in the entity attribute data of the key disturbance node, the dynamic change difference between the current time and the historical data of the same period is calculated, where the position offset is the Euclidean distance difference between the real-time position coordinates and the historical position coordinates, and the speed change rate is the ratio of the real-time speed to the historical speed; Determine the propagation path level coverage ratio between the key disturbance node and other nodes, where the propagation path level coverage ratio is the quotient of the number of propagation path levels of the key disturbance node and the number of propagation path levels of other nodes; Generate a dynamic adjustment strategy based on the product of the dynamic change difference and the propagation path layer coverage ratio, the dynamic adjustment strategy includes the propagation direction offset angle and the density weight allocation ratio; When the dynamic change difference exceeds the change threshold of the historical data of the same period, the priority of the propagation direction offset angle is adjusted based on the propagation path layer coverage ratio.

6. The digital twin method for an intelligent park based on artificial intelligence according to claim 5, characterized in that Adjust the propagation direction and density weight of key disturbance nodes in the dynamic simulation model based on the propagation path hierarchy and correlation strength to generate simulation results, including: Based on the propagation direction offset angle, the moving direction of the key disturbance node in the dynamic simulation model is adjusted. The moving direction is calculated based on the vector synthesis of the offset angle and the current propagation direction. According to the density weight allocation ratio, the density weight of the key disturbance node in the dynamic simulation model is adjusted, and the density weight is updated by multiplying the current density value by the allocation ratio; Input the adjusted moving direction and density weight into the dynamic simulation model, perform simulation calculation and generate simulation results; When the density weight of the key disturbance node is updated, the density weights of the adjacent nodes are updated synchronously, and the synchronous update is proportionally distributed according to the hierarchical correlation strength of the propagation path between the adjacent nodes and the key disturbance node.

7. An artificial intelligence-based digital twin method for smart campuses according to claim 6, characterized in that When the deviation between the behavior data of the non-critical disturbance node and the simulation result exceeds the preset range, the behavior weight of the non-critical disturbance node is reversely updated based on the correlation strength of the propagation path level, including: Detect the dynamic deviation rate between the behavior data of non-critical disturbance nodes and the simulation results. The dynamic deviation rate is the ratio of the absolute value of the difference between the real-time behavior data and the simulation behavior data to the average deviation of the historical period. Obtain a set of correlation strengths of the propagation path levels of non-critical disturbance nodes and all critical disturbance nodes, where each element in the correlation strength set is a coverage ratio of the propagation path level of a single critical disturbance node to the corresponding non-critical disturbance node; Calculate the behavior weight correction ratio of non-critical disturbance nodes based on the maximum value in the association strength set and the preset adjustment coefficient; When the dynamic deviation rate exceeds the preset deviation threshold, the behavior weight of the non-critical disturbance node is reversely adjusted based on the behavior weight correction ratio, and the dynamic adjustment strategy priority of the corresponding critical disturbance node in the association strength set is synchronously updated; If the adjusted behavior weight still causes the dynamic deviation rate to exceed the preset deviation threshold, the behavior weight correction ratio is recalculated according to the second largest value in the association strength set for secondary correction.

8. An AI-based digital twin system for a smart park, which is used to implement an AI-based digital twin method for a smart park according to any one of claims 1-7, characterized in that, include: Data collection module: obtains real-time status data of entity objects in the smart park, including entity attribute data and dynamic interaction data; Classification processing module: Based on the propagation path length and behavior type of dynamic interaction data, dynamic interaction data is divided into group behavior data and individual behavior data; Node identification module: Identify key disturbance nodes that meet preset disturbance conditions based on the propagation path level and density dynamic change conditions of group behavior data; Policy Generation Module: Generate a dynamic adjustment policy based on the dynamic change differences between the entity attribute data and historical behavior data of the key perturbation nodes, and combining the propagation path hierarchy evaluation of the association strength between the key perturbation nodes and other nodes; Simulation Construction Module: Adjust the propagation direction and density weight of the key perturbation nodes in the dynamic simulation model based on the propagation path hierarchy and association strength to generate a simulation result; Feedback Correction Module: When the deviation between the behavior data of the non-key perturbation nodes and the simulation result exceeds the preset range, reversely update the behavior weights of the non-key perturbation nodes based on the association strength of the propagation path hierarchy.

Images