A park dynamic risk heat map-based inspection route planning method and system

Abstract

The application discloses a kind of based on park dynamic risk heat map and system of inspection route planning method, the described method includes following steps: step 1: collection park basic data and real-time sensing data, and pre-processing is carried out;Step 2: construct park global space continuous dynamic risk heat map;Step 3: with park dynamic risk heat map as core driving, construct multi-objective optimization model and hard constraint condition for park inspection scene;Step 4: based on multi-objective optimization model and hard constraint condition, using the improved genetic-ant colony hybrid algorithm for park inspection scene is solved global optimal inspection route;Step 5: preset dynamic re-planning trigger condition, constructs the dynamic closed-loop management and control mechanism of inspection whole process.

Description

Technical Field

[0001] This invention relates to the fields of smart park security management and intelligent inspection route planning, and in particular to an inspection route planning method and system based on a dynamic risk heat map of the park. Background Technology

[0002] With the comprehensive advancement of smart park construction, the requirements for refined and real-time park security management are constantly increasing. As a core link in park security control, the rationality of the route planning of inspections directly determines the effectiveness of risk control and the efficiency of inspection execution.

[0003] The currently available technical solutions for park inspection can be mainly divided into two categories: one is static inspection route planning based on fixed locations and fixed cycles, which can only achieve the shortest path optimization and cannot adapt to the dynamic changes in park risks; the other is inspection route optimization based on risk quantification, which introduces risk assessment dimensions, but both have core technical defects and cannot meet the actual management and control needs of the park.

[0004] The existing technologies that are closest to this solution are as follows: Patent publication number CN116563968A, titled "A Method, Device, Electronic Equipment and Storage Medium for Generating Inspection Paths": This solution discloses an inspection path generation method based on risk heat maps. However, its core logic is based solely on static risk heat maps. Dynamic heat maps can only be updated in batches after a single inspection, based on the results of that inspection, without real-time multi-source perception data-driven capabilities. Path generation is merely a simple connection of high-risk points, without constructing a multi-objective fusion optimization model or considering the specific control constraints of the park scenario. Furthermore, it can only achieve single route planning before inspection, without a dynamic replanning mechanism during the inspection execution process.

[0005] Patent publication number CN121638827A, titled "Operation and Maintenance Safety Management System Adapted to Smart Parks": This solution is a system-type solution related to smart park inspections, which is fundamentally different from the method-type object for which this invention is claimed; its risk quantification only performs single-point risk confidence scoring for discrete inspection points, without constructing a continuous dynamic risk heat map of the entire park space, and thus cannot achieve accurate quantification of risks across the entire area; the route planning adopts a dual-route two-choice mode that completely separates risk and efficiency, lacks the ability to solve for the global optimal solution through multi-objective fusion, and also lacks a real-time dynamic replanning closed-loop design during the inspection process.

[0006] Currently available technical solutions cannot simultaneously solve the core problems of accurate risk quantification across the entire park area, multi-objective global optimal planning, dynamic risk real-time adaptation, and deep compatibility with park scenarios in park inspections. The specific core shortcomings are as follows: 1. The ability to quantify risks is severely insufficient, making it impossible to achieve precise prevention and control across the entire area: Existing technologies either only score the risk confidence of discrete inspection points, failing to generate a continuous risk distribution across the entire park and unable to avoid high-risk areas outside the inspection points; or they rely solely on static risk heat maps, only able to perform post-inspection batch updates based on the inspection results after a single inspection, lacking real-time multi-source sensing data-driven capabilities and the ability to predict the spread of risks within the park. Risk management is always lagging behind, and it is impossible to achieve early warning.

[0007] 2. The route planning logic has inherent flaws and cannot balance safety and efficiency: Existing technologies either only achieve single-objective optimization (shortest path priority / high-risk point priority), which cannot balance multi-dimensional management and control needs; or they adopt a dual-route two-choice mode that completely separates risk and efficiency, which cannot achieve a global optimal solution that integrates multiple objectives. This results in inspection routes that either expose too much risk or have extremely low inspection efficiency, making them unsuitable for complex management and control scenarios in the park.

[0008] 3. Lack of dynamic adaptability, unable to cope with sudden risks in the park: Existing technology can only complete static route planning before the inspection task is started. The route is completely fixed during a single inspection. It can only achieve minor adjustments to the path for local obstacle avoidance. It cannot complete dynamic replanning of the entire route based on real-time risk changes in the park. There is no complete real-time closed-loop control mechanism. Its adaptability to sudden risks such as fire alarms, perimeter intrusion, and equipment abnormalities is extremely poor.

[0009] 4. Poor scenario adaptability and extremely limited application: Existing technologies are either for general inspection scenarios without differentiation, without considering the park's specific control rules and hard constraints; or they are only customized for specific scenarios such as chemical parks and power plants, and cannot adapt to the control needs of general smart parks such as industrial parks, industrial parks, and logistics parks. They require large-scale customized transformation for different parks, resulting in high implementation costs and poor compatibility.

[0010] In summary, none of the currently available technical solutions can simultaneously solve the core problems of accurate quantification of risks across the entire park, optimal global planning for multiple objectives, real-time dynamic risk adaptation, and deep compatibility with park scenarios in park inspections.

[0011] The above content is only used to help understand the technical solution of the present invention and does not represent an admission that the above content is prior art. Summary of the Invention

[0012] The purpose of this invention is to address the shortcomings of existing technologies by proposing a method and system for planning inspection routes based on dynamic risk heat maps of industrial parks.

[0013] To achieve the above objectives, the present invention adopts the following technical solution: A method for planning inspection routes based on a dynamic risk heat map of a park includes the following steps: Step 1: Collect basic data and real-time sensing data of the park, and perform preprocessing; Step 2: Construct a continuous dynamic risk heat map of the entire park area; Step 3: Using the dynamic risk heat map of the park as the core driver, the real-time risk weight of the grid of the dynamic risk heat map is used as the core cost function for path planning, and a multi-objective optimization model and hard constraints are constructed for the park inspection scenario. Step 4: Based on the multi-objective optimization model and hard constraints, the genetic-ant colony hybrid algorithm improved for the park inspection scenario is used to solve the global optimal inspection route; Step 5: Preset dynamic replanning trigger conditions to build a dynamic closed-loop control mechanism for the entire inspection process.

[0014] Furthermore, in step 1, the collected basic data and real-time sensing data of the park include: basic geographic information of the park, park inspection and control rules, and multi-source real-time sensing data; The basic geographic information of the park includes: the spatial coordinates and attribute information of the park's high-precision GIS map, building and pipeline topology, inspection point distribution, access control points, fire lanes, restricted areas, and restricted routes; The park's inspection and control rules include: key control area classification standards, mandatory inspection point requirements, inspection time window constraints, inspection personnel permissions, day / night / holiday inspection mode switching rules, and load balancing distribution rules. Multi-source real-time sensing data includes: real-time data from the park's fire protection system, alarm data from the perimeter security system, real-time data from equipment operation and maintenance, video AI recognition results, and park weather warning data; In step 1, data preprocessing includes: performing data cleaning, outlier removal, spatiotemporal alignment, and normalization on the collected multi-source data to unify the data format and spatial coordinate system, ensuring the timeliness and consistency of the data.

[0015] Furthermore, step 2 includes the following steps: Step 2.1: Divide the park into spatial grids: Based on the high-precision GIS map of the park, divide the entire park area into several continuous spatial grids of equal size, with each grid corresponding to unique spatial coordinates and attribute labels; Step 2.2 Calculate multi-dimensional risk weights: For each spatial grid, based on the preprocessed multi-source real-time sensing data and combined with the park's key area classification standards, calculate the grid's real-time comprehensive risk weight; specifically, calculate the risk weight by weighting and summing the real-time alarm level, equipment anomaly probability, area control level, and historical risk occurrence frequency. Step 2.3: Construct a risk diffusion prediction model: Based on the topology of buildings, pipelines, and circuits in the park, construct a risk diffusion prediction model for the park. Combine the current real-time risk weights to simulate the diffusion path and speed of high-risk areas and generate a park risk prediction heat map for the next 5-30 minutes. Step 2.4: Real-time update of heat map: According to the preset frequency, the real-time risk heat map and the predicted heat map of the park are updated synchronously, and the real-time risk value and the predicted risk value of each spatial grid are output. Through weighted fusion, a continuous dynamic risk heat map of the entire park space is formed, which provides basic data support for subsequent inspection route planning.

[0016] Furthermore, step 3 includes the following steps: Step 3.1: Construct decision variables: Divide the entire park area into M continuous spatial grids, each grid corresponding to a unique spatial coordinate (x, y). i ,y i ) and real-time risk weight r i ; Obtain the guaranteed inspection points for a single inspection task, with the corresponding coordinate set being P={p1,p2,...,p N}, where N is the number of mandatory inspection points; the inspection route is a closed or open path passing through all mandatory inspection points, represented by an ordered sequence of points as R={p s1 ,p s2 ,...,p sN}, where s1, s2, ..., s N It represents the permutation of 1 to N, corresponding to the access order of the inspection points; Step 3.2: Construct the core cost function: The core cost function for path planning is defined using the real-time risk weights of the grids in the dynamic risk heatmap. The unit path cost of the inspection route passing through a certain grid is defined as follows: ; in, Let be the unit path cost of the i-th grid; α is the risk weight coefficient, with a value range of [0.6, 0.9]. The real-time risk weight for the i-th grid is derived from the dynamic risk heatmap and has a value range of [0, 100]. The base distance cost per unit path is fixed at 1. Step 3.3: Constructing the multi-objective optimization function: Integrating five core inspection sub-objectives, including: minimizing the total risk exposure of the inspection route, minimizing the total length of the inspection route, achieving 100% coverage of guaranteed inspection points, maximizing the satisfaction of inspection time window constraints, and maximizing the load balancing of multiple inspection entities, a normalized multi-objective overall optimization function is constructed; the overall optimization function is constructed through linear normalization and linear weighted summation. Step 3.4: Normalization and Fusion of Multi-Objective Functions: To address the dimensional differences among the various sub-objectives, a linear normalization method is used to standardize each sub-objective, resulting in normalized sub-objectives. The overall optimization function is constructed using a linear weighted summation method. ; in, Let ω1 be the weight coefficient of each sub-objective, and let ω2 be the sum of ω3 and ω4, where ω2 > ω5. Step 3.5: Embed hard constraints of the park: Construct spatial constraints, permission constraints, and compliance constraints as mandatory constraints for model solving.

[0017] Furthermore, in step 3.3, the sub-objective function that minimizes the total risk exposure of the inspection route is as follows: ; Where L is the total number of grid cells traversed by the route. Let be the path length of the route within the i-th grid, and the optimization objective is . Minimum; The sub-objective function for minimizing the total length of the inspection route is as follows: ; in, The Euclidean distance between two adjacent inspection points is the optimization objective. Minimum; The sub-objective function for achieving 100% coverage of inspection points is as follows: ; in, The optimization objective is to determine the number of guaranteed reachable points covered by route R. =0; The sub-objective function that achieves the highest satisfaction of the inspection time window constraint is as follows: ; in,[ , For each guaranteed access point pi, there is a corresponding allowed inspection time window. The optimization objective is to... =0; The sub-objective function that yields the highest load balancing across multiple inspection entities is as follows: ; Among them, w k Given the workload of the k-th inspection entity among K inspection entities, the optimization objective is: Minimum, meaning the difference in workload among the various inspection entities is minimized.

[0018] Furthermore, in step 3.5, the spatial constraint is that the route must not enter the restricted area, as shown in the following formula: ; in, This is a set of coordinates for restricted areas within the park; routes that must pass through fire lanes must meet certain requirements. ∈ ,in This is a set of coordinates for fire escape routes. Access control requires that the route taken by the patrol entity must match its access control permissions, i.e., the set of access control points the route passes through. And it must belong to the set of permissions of the inspection entity. ,Right now ; The compliance constraints are that the key controlled areas traversed by the route must meet the requirements of the park's safety production standards regarding the length of stay and the order of inspections.

[0019] Furthermore, the genetic-ant colony hybrid algorithm in step 4 includes the following improvements: Improvement 1: The grid risk weights of the dynamic risk heatmap are used as the core factor of the heuristic function of the ant colony algorithm; Improvement 2: The hard constraints of the park are embedded in the tabu table of the algorithm, and solutions that do not meet the constraints are automatically eliminated during the iteration process; Improvement 3: The crossover and mutation operation of the genetic algorithm is used to optimize the initial pheromone distribution of the ant colony algorithm, and the local search capability of the genetic algorithm is optimized through the positive feedback mechanism of the ant colony algorithm.

[0020] Furthermore, in the genetic-ant colony hybrid algorithm in step 4, the number of ants, m, is 1.5 ± 0.5 times the number of guaranteed inspection points N; the pheromone importance factor, γ = 1.2 ± 0.2, characterizes the influence of pheromones on path selection; the risk heuristic factor, δ = 2.5 ± 0.5, characterizes the influence of risk cost on path selection; the pheromone evaporation coefficient, ρ = 0.1 ± 0.02, ranges from [0,1]; the maximum number of iterations, MaxIter = 200 ± 10; and the genetic algorithm crossover probability, P... c =0.7±0.1, Probability of mutation: P m =0.05±0.01.

[0021] Furthermore, in step 4, the algorithm is solved iteratively through the following steps: Step 4.1: Based on the full permutation of the inspection points, generate an initial path population of size m. At the same time, embed the park's hard constraints into the population generation rules to ensure that all paths in the initial population meet the mandatory constraints. Step 4.2: Using the overall optimization function F(R) from Step 3 as the fitness function, perform selection, crossover, and mutation operations on the initial population, iterating 50 times to obtain the optimized initial path population; based on the optimized initial path population, generate the initial pheromone distribution for the ant colony algorithm, as shown in the following formula: ; in, This is the initial pheromone base value. Let pheromone intensity be constant. The fitness value is the result of optimization by the genetic algorithm for the optimal path. Step 4.3: Using the roulette wheel method and introducing a risk heuristic factor, calculate the state transition probability of ant k choosing the next visiting point j at point i. The formula is: ; in, The improved risk heuristic function is defined as follows: =1 / ,in The total risk cost of the path from point i to point j is derived from the dynamic risk heatmap; This is the set of points that ant k has not visited. Points that do not meet the hard constraints of the park are removed from this set, i.e., the tabu list is embedded. Step 4.4: After completing one iteration, an improved pheromone rule combining global and local methods is used for updating to avoid the algorithm getting trapped in local optima. Specifically, for local pheromone updates, a local update is performed immediately after each ant traverses a path (i,j), as shown in the following formula: ; The global pheromone update, after traversing all ants' paths, only performs a global update on the optimal path of the current iteration, as shown in the following formula: ; in, This is the fitness value of the optimal path in this iteration; Step 4.5: Repeat steps 4.3 and 4.4 until the maximum number of iterations MaxIter is reached, or the fitness value of the optimal path remains unchanged for 30 consecutive iterations, at which point the iteration terminates; output the path with the smallest fitness value during the iteration process as the globally optimal inspection route, and simultaneously send it to the inspection execution end.

[0022] Furthermore, step 5 includes the following steps: Step 5.1: During the inspection process, monitor the dynamic risk heat map of the park generated in Step 2 in real time, and track the inspection progress and location information simultaneously; Step 5.2: Preset dynamic replanning trigger conditions. Replanning will be automatically triggered when any condition is met. Specifically, the dynamic replanning trigger conditions include: the risk value change in the core control area of ​​the park exceeds the preset threshold; a new level 1 / 2 high-risk alarm is added in the park; the inspection execution progress deviates from the preset plan by more than the preset threshold; and abnormal situations are reported by the inspection execution terminal. Among them, the preset threshold for the risk value change in the core control area of ​​the park is 15%-30%, and the preset threshold for the inspection execution progress deviating from the preset plan is 20-40 minutes. Step 5.3: When replanning is triggered, steps 2 to 4 are executed again. Based on the latest dynamic risk heat map and inspection execution status, the updated global optimal inspection route is re-solved and generated. Step 5.4: Send the updated inspection route to the inspection execution terminal in real time, and update the inspection task synchronously to complete the real-time closed loop of the entire process of "data collection → heat map update → route optimization → execution feedback".

[0023] To achieve the above objectives, the present invention adopts the following technical solution: An inspection route planning system based on a dynamic risk heat map of a park includes: Data Input Layer: Acquires raw data to provide basic data support for subsequent full-process calculations; specifically includes three types of data sources: basic static data of the park, real-time dynamic sensing data, and park inspection and control rules; Data Preprocessing Layer: The input end connects to the data input layer, and the output end connects to the dynamic risk heat map construction module of the core computing layer. This module receives all the raw data from the data input layer and performs data cleaning, outlier removal, spatiotemporal alignment, and normalization processing in sequence to unify the format and spatial coordinate system of all data and output standardized, highly consistent compliant data to ensure the timeliness and accuracy of subsequent calculations. Core Computing Layer: The three core functional modules are sequentially connected according to the data flow to realize the entire process of calculation from risk quantification to optimal route solution; the three core functional modules include: dynamic risk heat map construction module, multi-objective optimization model construction module, and global optimal inspection route solution module; Control and execution layer: Enables the issuance, execution, feedback, and dynamic adjustment of inspection tasks; includes: dynamic replanning closed-loop control module and inspection execution terminal; Visualized Control Layer: Corresponding to the park's intelligent control visualization platform, the input end synchronously connects to all modules of the core computing layer and the control execution layer, and synchronizes all data in real time, including dynamic risk heat maps, multi-objective optimization model calculation process, inspection route solution results, inspection task execution progress, replanning trigger records, etc., to realize the visualization display, task scheduling, anomaly alarm and data statistical analysis of the entire park inspection process.

[0024] Compared with the prior art, the beneficial effects of this invention are as follows: 1. Steps 1-2 provide a dynamic risk heatmap generation mechanism that does not rely on static heatmaps, is driven by multi-source real-time data from the park, is spatially continuous across the entire area, and has the ability to predict risk diffusion, thus solving the problem of insufficient risk quantification capabilities in existing technologies. Steps 2-4 provide a globally optimal inspection route planning method that uses the risk weights of the entire heatmap as the core cost function, integrates multi-dimensional control objectives, and embeds park-specific hard constraints, thus solving the problem of logical defects in route planning in existing technologies. Step 5 constructs a real-time closed loop for dynamic replanning of the entire route during the inspection execution process, enabling rapid response to sudden risks in the park and adaptive route adjustment, thus solving the problem of lack of dynamic adaptability in existing technologies. The general and specific control rules of the park are embedded into the entire process of the technical solution, adapting to various general smart park scenarios, greatly improving the feasibility and compatibility of the solution, and solving the problem of poor scenario adaptability in existing technologies.

[0025] 2. Step 3 is a multi-objective fusion optimization model customized for park inspection scenarios. It abandons the conventional approach of single-objective optimization and dual-route choice in existing technologies. It uses the real-time risk weight of the dynamic risk heat map as the core cost function, and integrates five park-specific control objectives. It achieves the global optimal solution through normalized weighted fusion, taking into account both inspection safety and efficiency.

[0026] 3. Step 4 is an improved genetic-ant colony hybrid algorithm for park inspection scenarios. It uses dynamic risk weights as the core heuristic factor, embeds the park's hard constraints into the algorithm's taboo table, and optimizes the initial pheromone distribution and pheromone update rules. This not only solves the problems of premature convergence of conventional genetic algorithms and the lack of initial pheromone in conventional ant colony algorithms, but also makes the algorithm naturally adaptable to the risk prevention and control requirements and management rules of park inspection. Attached Figure Description

[0027] Figure 1 A flowchart of an inspection route planning method based on a dynamic risk heat map of a park; Figure 2 This is an architecture diagram of an inspection route planning system based on a dynamic risk heat map of a park. Detailed Implementation

[0028] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0029] Example 1: A method for planning inspection routes based on dynamic risk heat maps of the park, such as Figure 1 As shown, it includes the following steps: Step 1: Collect basic data and real-time sensing data of the park, and perform preprocessing.

[0030] In this embodiment, in step 1, three types of core basic data of the target park are acquired and standardized preprocessing is completed to provide data support for subsequent heat map generation and route planning.

[0031] In this embodiment, the basic data and real-time sensing data collected in step 1 include: basic geographic information of the park, park inspection and control rules, and multi-source real-time sensing data.

[0032] The basic geographic information of the park includes: the spatial coordinates and attribute information of the park's high-precision GIS map, building and pipeline topology, inspection point distribution, access control points, fire lanes, restricted areas, and restricted routes.

[0033] The park's inspection and control rules include: key control area classification standards, mandatory inspection point requirements, inspection time window constraints, inspection personnel permissions, day / night / holiday inspection mode switching rules, and load balancing distribution rules.

[0034] The multi-source real-time sensing data includes: real-time data of the park's fire protection system (water pressure, smoke, and temperature), alarm data of the perimeter security system, real-time data of equipment operation and maintenance (power distribution cabinets, elevators, and temporary storage equipment for hazardous chemicals), video AI recognition results (open flames, illegal gatherings, road occupancy, and foreign object intrusion), and park weather early warning data.

[0035] In this embodiment, step 1, data preprocessing includes: performing data cleaning, outlier removal, spatiotemporal alignment, and normalization on the collected multi-source data to unify the data format and spatial coordinate system, ensuring the timeliness and consistency of the data.

[0036] Step 2: Construct a dynamic risk heat map that provides a continuous view of the entire park area.

[0037] In this embodiment, step 2 does not rely on the static risk heat map at all, and directly generates a dynamic heat map of the entire domain based on the real-time data preprocessed in step 1.

[0038] In this embodiment, step 2 includes the following steps: Step 2.1: Divide the park into spatial grids: Based on the high-precision GIS map of the park, divide the entire park area into several continuous spatial grids of equal size, with each grid corresponding to unique spatial coordinates and attribute labels.

[0039] In this embodiment, step 2.1 divides the entire park area into several continuous spatial grids of equal size, which can be adjusted according to the park area, with 5m×5m being recommended.

[0040] Step 2.2 Calculate multi-dimensional risk weights: For each spatial grid, based on the preprocessed multi-source real-time sensing data and combined with the park's key area classification standards, calculate the grid's real-time comprehensive risk weight; specifically, calculate the risk weight by weighting and summing the real-time alarm level, equipment anomaly probability, area control level, and historical risk occurrence frequency.

[0041] In this embodiment, the weighting coefficients of real-time alarm level, equipment anomaly probability, area control level, and historical risk occurrence frequency in step 2.2 can be set according to the park control rules.

[0042] Step 2.3: Construct a risk diffusion prediction model: Based on the topology of buildings, pipelines, and circuits in the park, construct a risk diffusion prediction model for the park. Combine the current real-time risk weights to simulate the diffusion path and speed of high-risk areas and generate a heat map of park risk prediction for the next 5-30 minutes.

[0043] In this embodiment, step 2.3 includes the following steps: Step 2.3.1, Topology Modeling: Divide the park into spatially continuous grid units. Based on physical boundaries such as building walls, fire compartments, pipeline routes, and circuit loops, construct a connectivity topology map of the park and define the passable paths and obstruction coefficients for risk diffusion.

[0044] Step 2.3.2, Risk Diffusion Rule Customization: For typical risk types in the park: fire, gas leak, power failure, and personnel intrusion, different diffusion models are constructed respectively. (1) Fire risk: A simplified diffusion model based on fluid dynamics is used, combined with wind speed, building openings and sprinkler system response time, to calculate the speed of fire spread and the scope of impact; (2) Gas leakage: A Gaussian plume model was used, combined with wind direction, wind speed and leakage point pressure, to simulate the diffusion distribution of gas concentration over time; (3) Power faults: Based on the fault propagation model of circuit topology, calculate the influence range and diffusion path of the fault current; (4) Personnel intrusion: Based on the topological relationship of security points in the park, simulate the movement path and area coverage of intruders.

[0045] Step 2.3.3, Diffusion Rate Correction: Introduce dynamic parameters of the park environment, including temperature, humidity, personnel density, and equipment operating status, to correct the diffusion rate and avoid prediction bias of the general model in the park scenario.

[0046] Step 2.3.4, Predictive Heatmap Generation: Combining the basic weights of the current real-time risk heatmap, the diffusion prediction results of various types of risks are weighted and integrated to generate a multi-time node risk prediction heatmap with 1-minute intervals for the next 5-30 minutes.

[0047] Step 2.4: Real-time update of heat map: According to the preset frequency (e.g., once every minute), the real-time risk heat map and the predicted heat map of the park are updated synchronously, and the real-time risk value and the predicted risk value of each spatial grid are output. Through weighted fusion, a continuous dynamic risk heat map of the entire park space is formed, which provides basic data support for subsequent inspection route planning.

[0048] In this embodiment, the preset frequency in step 2.4 can be set according to the park management and control requirements, such as 1 minute / time.

[0049] Step 3: Using the dynamic risk heat map of the park as the core driver, the real-time risk weight of the grid of the dynamic risk heat map is used as the core cost function for path planning, and a multi-objective optimization model and hard constraints are constructed for the park inspection scenario.

[0050] In this embodiment, step 3 includes the following steps: Step 3.1: Construct decision variables: Divide the entire park area into M continuous spatial grids, each grid corresponding to a unique spatial coordinate (x, y). i ,y i ) and real-time risk weight r i ; Obtain the guaranteed inspection points for a single inspection task, with the corresponding coordinate set being P={p1,p2,...,p N}, where N is the number of mandatory inspection points; the inspection route is a closed or open path passing through all mandatory inspection points, represented by an ordered sequence of points as R={p s1 ,p s2 ,...,p sN}, where s1, s2, ..., s N It represents all permutations from 1 to N, corresponding to the order in which the inspection points are visited.

[0051] In this embodiment, the real-time risk weight r in step 3.1 i The result is obtained through step 2.

[0052] Step 3.2: Construct the core cost function: The core cost function for path planning is defined using the real-time risk weights of the grids in the dynamic risk heatmap. The unit path cost of the inspection route passing through a certain grid is defined as follows: ; in, α is the unit path cost of the i-th grid; α is the risk weight coefficient, with a value range of [0.6, 0.9], which can be dynamically adjusted according to the park's control level (α=0.9 under Level 1 control mode, α=0.6 under normal mode). The real-time risk weight for the i-th grid is derived from the dynamic risk heatmap and has a value range of [0, 100]. The base distance cost per unit path is 1, which is a fixed value.

[0053] Step 3.3: Constructing the multi-objective optimization function: Integrating five core inspection sub-objectives, including: minimizing the total risk exposure of the inspection route, minimizing the total length of the inspection route, achieving 100% coverage of guaranteed inspection points, maximizing the satisfaction of inspection time window constraints, and maximizing the load balancing of multiple inspection entities, a normalized multi-objective overall optimization function is constructed; the overall optimization function is constructed through linear normalization and linear weighted summation.

[0054] In this embodiment, in step 3.3, the sub-objective function that minimizes the total risk exposure of the inspection route is as follows: ; Where L is the total number of grid cells traversed by the route. Let be the path length of the route within the i-th grid, and the optimization objective is . Minimum.

[0055] In this embodiment, the total risk exposure of the inspection route in step 3.3 is the sum of the risk costs of all grids traversed by the route.

[0056] In this embodiment, in step 3.3, the sub-objective function for minimizing the total length of the inspection route is as follows: ; in, The Euclidean distance between two adjacent inspection points is the optimization objective. Minimum.

[0057] In this embodiment, the sub-objective function for achieving 100% coverage of inspection points in step 3.3 is as follows: ; in, The optimization objective is to determine the number of guaranteed reachable points covered by route R. =0.

[0058] In this embodiment, in step 3.3, the sub-objective function with the highest satisfaction of the inspection time window constraint is as follows: ; in,[ , For each guaranteed access point pi, there is a corresponding allowed inspection time window. The optimization objective is to... =0.

[0059] In this embodiment, in step 3.3, the sub-objective function that achieves the highest load balancing degree among the multiple inspection subjects is as follows: ; Among them, w k The task load of the k-th inspection entity among K inspection entities (a weighted sum of the number of inspection points and the total route length) is optimized as follows: Minimum, meaning the difference in workload among the various inspection entities is minimized.

[0060] Step 3.4: Normalization and Fusion of Multi-Objective Functions: To address the dimensional differences among the various sub-objectives, a linear normalization method is used to standardize each sub-objective, resulting in normalized sub-objectives. The overall optimization function is constructed using a linear weighted summation method. ; in, Let ω1 be the weight coefficient of each sub-objective, and let ω2 be the sum of 1, where ω1>ω3>ω4>ω2>ω5.

[0061] In this embodiment, the weight coefficients of each sub-target in step 3.4 can be dynamically adjusted according to the park inspection mode.

[0062] Step 3.5: Embed hard constraints of the park: Construct spatial constraints, permission constraints, and compliance constraints as mandatory constraints for model solving.

[0063] In this embodiment, step 3.5 constructs spatial constraints, permission constraints, and compliance constraints as mandatory constraints for model solving in step 4, and writes them into the feasible domain limit of the optimization model.

[0064] In this embodiment, in step 3.5, the spatial constraint is that the route must not enter the restricted area, as shown in the following formula: ; in, This is a set of coordinates for restricted areas within the park; routes that must pass through fire lanes must meet certain requirements. ∈ ,in This is a set of coordinates for fire escape routes.

[0065] In this embodiment, in step 3.5, the permission constraint is that the route of the inspection subject must match its access control permissions, that is, the set of access control points the route passes through. And it must belong to the set of permissions of the inspection entity. ,Right now .

[0066] In this embodiment, in step 3.5, the compliance constraint is that the key control areas traversed by the route must meet the requirements of the park's safety production regulations regarding the length of stay and the order of inspections.

[0067] Step 4: Based on the multi-objective optimization model and hard constraints, the genetic-ant colony hybrid algorithm, which is improved for the park inspection scenario, is used to solve the global optimal inspection route.

[0068] In this embodiment, step 4 is improved to address the shortcomings of conventional genetic algorithms, such as weak local search capabilities and premature convergence, and conventional ant colony algorithms, such as a lack of initial pheromones and slow convergence speed. This improvement is based on the hard constraints of the park inspection scenario and the core driving characteristics of the dynamic risk heat map.

[0069] In this embodiment, the genetic-ant colony hybrid algorithm in step 4 includes the following improvements: Improvement 1: The grid risk weights of the dynamic risk heatmap are used as the core factor of the heuristic function of the ant colony algorithm, replacing the fixed distance heuristic factor of the conventional algorithm, so that the algorithm naturally has the ability to avoid high-risk areas; Improvement 2: The hard constraints of the park are embedded in the tabu table of the algorithm, and solutions that do not meet the constraints are automatically eliminated during the iteration process, which greatly improves the solution efficiency and avoids invalid iterations; Improvement 3: The crossover and mutation operation of the genetic algorithm is used to optimize the initial pheromone distribution of the ant colony algorithm, solving the problem of insufficient initial pheromone in the conventional ant colony algorithm. At the same time, the positive feedback mechanism of the ant colony algorithm is used to optimize the local search capability of the genetic algorithm and avoid getting trapped in local optima.

[0070] In this embodiment, in the genetic-ant colony hybrid algorithm in step 4, the number of ants (m) is 1.5 times the number of guaranteed inspection points (N); the pheromone importance factor (γ=1.2) characterizes the influence of pheromones on path selection; the risk heuristic factor (δ=2.5) characterizes the influence of risk cost on path selection (a core improvement, replacing the conventional distance heuristic factor); the pheromone evaporation coefficient (ρ=0.1) ranges from [0,1]; the maximum number of iterations (MaxIter=200); and the genetic algorithm crossover probability (P) is... c =0.7, mutation probability: P m =0.05.

[0071] In this embodiment, step 4 involves the following steps to complete the iterative solution of the algorithm: Step 4.1: Based on the full permutation of the inspection points, generate an initial path population of size m. At the same time, embed the park's hard constraints into the population generation rules to ensure that all paths in the initial population meet the mandatory constraints.

[0072] Step 4.2: Using the overall optimization function F(R) from Step 3 as the fitness function, perform selection, crossover, and mutation operations on the initial population, iterating 50 times to obtain the optimized initial path population; based on the optimized initial path population, generate the initial pheromone distribution for the ant colony algorithm, as shown in the following formula: ; in, This is the initial pheromone base value. Let pheromone intensity be constant. The fitness value is the optimal path obtained by the genetic algorithm.

[0073] In this embodiment, steps 4.1 to 4.2 complete the generation of the initial population and the genetic algorithm preprocessing.

[0074] Step 4.3: Using the roulette wheel method and introducing a risk heuristic factor, calculate the state transition probability of ant k choosing the next visiting point j at point i. The formula is: ; in, The improved risk heuristic function, replacing the conventional inverse distance heuristic function, is defined as follows: =1 / ,in The total risk cost of the path from point i to point j is derived from the dynamic risk heatmap; Let be the set of points that ant k has not visited. Points that do not meet the hard constraints of the park are removed from this set, i.e., embedded in the tabu list.

[0075] In this embodiment, step 4.3 completes the path selection based on the risk heuristic function.

[0076] Step 4.4: After completing one iteration, an improved pheromone rule combining global and local methods is used for updating to avoid the algorithm getting trapped in local optima. Specifically, for local pheromone updates, a local update is performed immediately after each ant traverses a path (i,j), as shown in the following formula: ; The global pheromone update, after traversing all ants' paths, only performs a global update on the optimal path of the current iteration, as shown in the following formula: ; in, This is the fitness value of the optimal path in this iteration.

[0077] In this embodiment, step 4.4 completes the optimization of the pheromone update rule.

[0078] Step 4.5: Repeat steps 4.3 and 4.4 until the maximum number of iterations MaxIter is reached, or the fitness value of the optimal path remains unchanged for 30 consecutive iterations, at which point the iteration terminates; output the path with the smallest fitness value during the iteration process as the globally optimal inspection route, and simultaneously send it to the inspection execution end.

[0079] Step 5: Preset dynamic replanning trigger conditions to build a dynamic closed-loop control mechanism for the entire inspection process.

[0080] In this embodiment, step 5 includes the following steps: Step 5.1: During the inspection process, monitor the dynamic risk heat map of the park generated in Step 2 in real time, and track the inspection progress and location information simultaneously.

[0081] Step 5.2: Preset dynamic replanning trigger conditions. Replanning will be automatically triggered when any condition is met. Specifically, the dynamic replanning trigger conditions include: the risk value change in the core control area of ​​the park exceeds the preset threshold (20% recommended); the park adds a level 1 / level 2 high-risk alarm; the inspection execution progress deviates from the preset plan by more than the preset threshold; and the inspection execution terminal reports abnormal situations. Among them, the preset threshold for the risk value change in the core control area of ​​the park is 15%-30%, and the preset threshold for the inspection execution progress to deviate from the preset plan is 20-40 minutes.

[0082] Step 5.3: When replanning is triggered, steps 2 to 4 are executed again. Based on the latest dynamic risk heat map and inspection execution status, the updated global optimal inspection route is re-solved and generated.

[0083] Step 5.4: Send the updated inspection route to the inspection execution terminal in real time, and update the inspection task synchronously to complete the real-time closed loop of the entire process of "data collection → heat map update → route optimization → execution feedback".

[0084] This embodiment presents a patrol route planning method based on a dynamic risk heat map of a park. Steps 1-2 provide a dynamic risk heat map generation mechanism that does not rely on static heat maps, is driven by multi-source real-time data of the park, is spatially continuous across the entire area, and has the ability to predict risk diffusion, thus solving the problem of insufficient risk quantification capability in existing technologies. Steps 2-4 provide a globally optimal patrol route planning method that uses the risk weight of the entire heat map as the core cost function, integrates multi-dimensional control objectives, and embeds park-specific hard constraints, thus solving the problem of logical defects in route planning in existing technologies. Step 5 constructs a real-time closed loop for dynamic replanning of the entire route during the patrol execution process, enabling rapid response to sudden risks in the park and adaptive route adjustment, thus solving the problem of lack of dynamic adaptability in existing technologies. The method embeds the park's general and specific control rules into the entire process of the technical solution, adapting to various general smart park scenarios, greatly improving the feasibility and compatibility of the solution, and solving the problem of poor scenario adaptability in existing technologies.

[0085] This embodiment presents a method for planning inspection routes based on a dynamic risk heat map of a park. Step 3 involves a multi-objective fusion optimization model customized for the park inspection scenario. This model abandons the conventional approach of single-objective optimization and dual-route selection in existing technologies. It uses the real-time risk weight of the dynamic risk heat map as the core cost function and integrates five park-specific control objectives. The global optimal solution is achieved through normalized weighted fusion, taking into account both inspection safety and efficiency.

[0086] This embodiment presents an inspection route planning method based on a dynamic risk heat map of a park. Step 4 is an improved genetic-ant colony hybrid algorithm for park inspection scenarios. It uses dynamic risk weights as the core heuristic factor, embeds the park's hard constraints into the algorithm's taboo table, and optimizes the initial pheromone distribution and pheromone update rules. This solves the problems of premature convergence of conventional genetic algorithms and the lack of initial pheromone in conventional ant colony algorithms, and makes the algorithm naturally adaptable to the risk prevention and control requirements and management rules of park inspections.

[0087] To further illustrate the inspection route planning method based on a dynamic risk heat map of the park in this embodiment, the following scenario is used as an example: Application Scenario: A smart manufacturing industrial park covering 500,000 square meters includes 12 production plants, 2 comprehensive office buildings, 1 hazardous chemical storage area, 1 complete fire protection pipeline network, and 200 fixed inspection points. It is equipped with security systems, fire protection systems, equipment operation and maintenance systems, and video AI monitoring systems. The current inspection mode is fixed routes and manual inspections twice a day, which has problems such as lagging risk control, low inspection efficiency, and untimely response to sudden risks.

[0088] Using the method described in this invention, an inspection route planning system is constructed for the park. The specific implementation steps are as follows: Step A.1: Data acquisition and preprocessing.

[0089] A high-precision GIS map of the park was collected, and spatial grid division was completed. The grid size is 5m×5m, with a total of 200,000 effective grids. The spatial coordinates, building attributes, and control level of each grid were marked.

[0090] Enter the park's inspection and control rules: designate the hazardous chemical storage area, power distribution room, and fire control room as Level 1 control areas, the production plant as Level 2 control areas, and the office building as Level 3 control areas; clarify the mandatory requirements for 200 inspection points, the inspection time windows (day shift 8:00-18:00, night shift 20:00-6:00 the next day), and the permissions and load balancing rules for the 4 groups of inspection personnel.

[0091] It connects to the park's existing systems to collect real-time data on fire water pressure, smoke detection, perimeter alarms, power distribution cabinet operation, video AI recognition results, and meteorological data, with a collection frequency of 30 seconds per instance.

[0092] The collected multi-source data was cleaned, spatiotemporally aligned, and normalized. The unified spatial coordinate system was CGCS2000, and outliers and duplicate data were removed.

[0093] Step A.2: Construction of dynamic risk heat map of the park.

[0094] For each spatial grid, a risk weight calculation model is constructed, with the following four dimensions weighted as follows: real-time alarm level 40%, equipment anomaly probability 30%, area control level 20%, and historical risk frequency 10%. The real-time comprehensive risk weight of each grid is calculated, with the risk value ranging from 0 to 100. The higher the value, the higher the risk level.

[0095] Based on the topology of the park's fire protection pipeline network and power distribution pipeline network, a risk diffusion prediction model is constructed. For high-risk areas, the risk diffusion path is simulated for the next 15 minutes, and a prediction heat map is generated.

[0096] The heat map is set to update every minute, and the real-time heat map and the predicted heat map are updated simultaneously, and the visualization is completed on the park management platform.

[0097] Step A.3: Construction of multi-objective optimization model and solution of route.

[0098] Using real-time grid risk weights as the core cost function, a multi-objective optimization function is constructed. The priority of the core objectives is: minimum total risk exposure > 100% coverage of reachable points > satisfaction of time window constraints > shortest path length > load balancing.

[0099] Embedded hard constraints within the park: complete avoidance of restricted areas, priority access to fire lanes, mandatory access to level 1 controlled areas, and matching of access control permissions for inspection personnel.

[0100] An improved genetic-ant colony hybrid algorithm was adopted, with 200 iterations to solve the initial route problem. The globally optimal inspection routes were generated for the four groups of inspectors. Each route covered 50 must-reach inspection points. The total risk exposure of the route was reduced by 42% compared to the original fixed route, and the total inspection time was shortened by 28%, meeting all control constraints.

[0101] Step A.4: Dynamic replanning closed-loop execution.

[0102] During the inspection process, the dynamic risk heat map of the park is monitored in real time, and replanning trigger thresholds are set: the risk value of the first-level control area changes by more than 20%, a new first-level high-risk alarm is added, and the inspection progress deviates from the plan by more than 30 minutes.

[0103] During the inspection, the fire water pressure in a factory building in the park was abnormal, triggering a level one high-risk alarm. The risk value of the corresponding area rose from 15 to 85, meeting the conditions for replanning.

[0104] The system automatically triggers replanning, resolving and generating updated inspection routes based on the latest dynamic risk heat map, adjusting the routes of the corresponding inspection teams, prioritizing inspections of the high-risk areas, and ensuring that the time window constraints of other must-reach points are met. The entire replanning process takes 45 seconds, and the updated routes are sent to the handheld terminals of the inspection personnel in real time.

[0105] After the inspection personnel completed the handling of the anomaly, the system updated the heat map again, regenerated the optimal route, resumed the normal inspection task, and completed the closed-loop management of the entire process.

[0106] Example 2: An inspection route planning system based on a dynamic risk heat map of the park, such as Figure 2 As shown, the overall architecture adopts a five-layer linear structure driven by data flow, with the dynamic risk heat map of the park as the core driver, fully implementing the inspection route planning method described in this invention. The functions and data interaction logic of each level and module are as follows: Data Input Layer: Acquires raw data to provide basic data support for subsequent full-process calculations; specifically includes three types of data sources: basic static data of the park, real-time dynamic sensing data, and park inspection and control rules.

[0107] Basic static data of the park: covering high-precision GIS map of the park, building and pipeline topology, distribution of inspection points, spatial coordinates and attribute information of restricted areas / fire lanes / access control points.

[0108] Real-time dynamic perception data: covering real-time data of the park's fire protection system, alarm data of the perimeter security system, real-time data of equipment operation and maintenance, video AI recognition results, and park weather warning data.

[0109] Park inspection and control rules: covering key control area classification standards, mandatory inspection point requirements, inspection time window constraints, inspection personnel permissions, day / night / holiday inspection mode switching rules, and load balancing distribution rules.

[0110] Data Preprocessing Layer: The input end connects to the data input layer, and the output end connects to the dynamic risk heat map construction module of the core computing layer. This module receives all the raw data from the data input layer and performs data cleaning, outlier removal, spatiotemporal alignment, and normalization processing in sequence to unify the format and spatial coordinate system of all data and output standardized, highly consistent compliant data to ensure the timeliness and accuracy of subsequent calculations.

[0111] Core Computing Layer: The three core functional modules are sequentially connected according to the data flow to realize the entire process of calculation from risk quantification to optimal route solution; the three core functional modules include: dynamic risk heat map construction module, multi-objective optimization model construction module, and global optimal inspection route solution module.

[0112] Dynamic Risk Heatmap Construction Module: The input end connects to the data preprocessing layer, and the output end connects to the multi-objective optimization model construction module. After receiving standardized and compliant data, it sequentially completes the spatial grid division of the park, the calculation of multi-dimensional risk weights of the grids, the construction of the park risk diffusion prediction model, and the real-time update of the heatmap. Finally, it generates a continuous dynamic risk heatmap of the entire park space and outputs the real-time risk weight of each spatial grid as the core driving factor for subsequent route planning. Multi-objective optimization model construction module: This module corresponds to the core cost function construction module in the system. Its input end connects to the dynamic risk heat map construction module, and its output end connects to the global optimal inspection route solution module. It uses the real-time risk weight of the grid in the dynamic risk heat map as the core cost function for path planning, and constructs a multi-objective optimization function that integrates five core optimization objectives. At the same time, it embeds three hard constraints: park space constraints, permission constraints, and compliance constraints, forming a complete computational framework for route solution. The global optimal inspection route solution module: the input end connects to the multi-objective optimization model construction module, and the output end connects to the dynamic replanning closed-loop management module; based on the multi-objective optimization model and hard constraints, the improved genetic-ant colony hybrid algorithm for park inspection scenarios described in this invention is used to generate the global optimal inspection route through iterative solution.

[0113] Control and execution layer: realizes the issuance, execution, feedback and dynamic adjustment of inspection tasks; including: dynamic replanning closed-loop control module and inspection execution terminal.

[0114] The dynamic replanning closed-loop management module: its input end connects to the global optimal inspection route solving module and the inspection execution terminal, and its output end connects to both the inspection execution terminal and the dynamic risk heat map construction module; it receives the solved global optimal inspection route and sends it to the inspection execution terminal, while simultaneously monitoring the changes in the park's dynamic risk heat map and the inspection execution progress in real time; when the preset dynamic replanning trigger conditions are met, it immediately sends a replanning trigger signal to the dynamic risk heat map construction module, restarts the entire process calculation, generates an updated optimal inspection route and sends it out simultaneously, completing the dynamic closed-loop management of the entire inspection process; Inspection execution terminal: This is the execution carrier for inspection tasks. It receives inspection routes and tasks issued by the control module, performs on-site inspection operations, and synchronously transmits back inspection execution progress, location information, and on-site anomalies.

[0115] Visualized Control Layer: Corresponding to the park's intelligent control visualization platform, the input end synchronously connects to all modules of the core computing layer and the control execution layer, and synchronizes all data in real time, including dynamic risk heat maps, multi-objective optimization model calculation process, inspection route solution results, inspection task execution progress, replanning trigger records, etc., to realize the visualization display, task scheduling, anomaly alarm and data statistical analysis of the entire park inspection process.

[0116] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A method for planning inspection routes based on a dynamic risk heat map of a park, characterized in that, Includes the following steps: Step 1: Collect basic data and real-time sensing data of the park, and perform preprocessing; Step 2: Construct a continuous dynamic risk heat map of the entire park area; Step 3: Using the dynamic risk heat map of the park as the core driver, the real-time risk weight of the grid of the dynamic risk heat map is used as the core cost function for path planning, and a multi-objective optimization model and hard constraints are constructed for the park inspection scenario. Step 4: Based on the multi-objective optimization model and hard constraints, the genetic-ant colony hybrid algorithm improved for the park inspection scenario is used to solve the global optimal inspection route; Step 5: Preset dynamic replanning trigger conditions to build a dynamic closed-loop control mechanism for the entire inspection process.

2. The inspection route planning method based on a dynamic risk heat map of a park, as described in claim 1, is characterized in that... In step 1, the collected basic data and real-time sensing data of the park include: basic geographic information of the park, park inspection and control rules, and multi-source real-time sensing data; The basic geographic information of the park includes: the spatial coordinates and attribute information of the park's high-precision GIS map, building and pipeline topology, inspection point distribution, access control points, fire lanes, restricted areas, and restricted routes; The park's inspection and control rules include: key control area classification standards, mandatory inspection point requirements, inspection time window constraints, inspection personnel permissions, day / night / holiday inspection mode switching rules, and load balancing distribution rules. Multi-source real-time sensing data includes: real-time data from the park's fire protection system, alarm data from the perimeter security system, real-time data from equipment operation and maintenance, video AI recognition results, and park weather warning data; In step 1, data preprocessing includes: performing data cleaning, outlier removal, spatiotemporal alignment, and normalization on the collected multi-source data to unify the data format and spatial coordinate system, ensuring the timeliness and consistency of the data.

3. The inspection route planning method based on a dynamic risk heat map of a park, as described in claim 1, is characterized in that... Step 2 includes the following steps: Step 2.1: Divide the park into spatial grids: Based on the high-precision GIS map of the park, divide the entire park area into several continuous spatial grids of equal size, with each grid corresponding to unique spatial coordinates and attribute labels; Step 2.2 Calculate multi-dimensional risk weights: For each spatial grid, based on the preprocessed multi-source real-time sensing data and combined with the park's key area classification standards, calculate the grid's real-time comprehensive risk weight. Specifically, the risk weight is calculated by weighting and summing the real-time alarm level, the probability of equipment malfunction, the regional control level, and the frequency of historical risk occurrences. Step 2.3: Construct a risk diffusion prediction model: Based on the topology of buildings, pipelines, and circuits in the park, construct a risk diffusion prediction model for the park. Combine the current real-time risk weights to simulate the diffusion path and speed of high-risk areas and generate a park risk prediction heat map for the next 5-30 minutes. Step 2.4: Real-time update of heat map: According to the preset frequency, the real-time risk heat map and the predicted heat map of the park are updated synchronously, and the real-time risk value and the predicted risk value of each spatial grid are output. Through weighted fusion, a continuous dynamic risk heat map of the entire park space is formed, which provides basic data support for subsequent inspection route planning.

4. The inspection route planning method based on a dynamic risk heat map of a park, as described in claim 1, is characterized in that... Step 3 includes the following steps: Step 3.1: Construct decision variables: Divide the entire park area into M continuous spatial grids, each grid corresponding to a unique spatial coordinate (x, y). i ,y i ) and real-time risk weight r i ; Obtain the guaranteed inspection points for a single inspection task, with the corresponding coordinate set being P={p1,p2,...,p N }, where N is the number of mandatory inspection points; the inspection route is a closed or open path passing through all mandatory inspection points, represented by an ordered sequence of points as R={p s1 ,p s2 ,...,p sN }, where s1, s2, ..., s N It represents the permutation of 1 to N, corresponding to the access order of the inspection points; Step 3.2: Construct the core cost function: The core cost function for path planning is defined using the real-time risk weights of the grids in the dynamic risk heatmap. The unit path cost of the inspection route passing through a certain grid is defined as follows: ； in, Let be the unit path cost of the i-th grid; α is the risk weight coefficient, with a value range of [0.6, 0.9]. The real-time risk weight for the i-th grid is derived from the dynamic risk heatmap and has a value range of [0, 100]. The base distance cost per unit path is fixed at 1. Step 3.3: Constructing the multi-objective optimization function: Integrating five core inspection sub-objectives, including: minimizing the total risk exposure of the inspection route, minimizing the total length of the inspection route, achieving 100% coverage of the must-reach inspection points, maximizing the satisfaction of the inspection time window constraints, and maximizing the load balancing of multiple inspection entities; Constructing the overall optimization function through linear normalization and linear weighted summation. Step 3.4: Normalization and Fusion of Multi-Objective Functions: To address the dimensional differences among the various sub-objectives, a linear normalization method is used to standardize each sub-objective, resulting in normalized sub-objectives. The overall optimization function is constructed using a linear weighted summation method. ； in, Let ω1 be the weight coefficient of each sub-objective, and let ω2 be the sum of ω3 and ω4, where ω2 > ω5. Step 3.5: Embed hard constraints of the park: Construct spatial constraints, permission constraints, and compliance constraints as mandatory constraints for model solving.

5. The inspection route planning method based on a dynamic risk heat map of a park, as described in claim 4, is characterized in that... In step 3.3, the sub-objective function that minimizes the total risk exposure of the inspection route is as follows: ； Where L is the total number of grid cells traversed by the route. Let be the path length of the route within the i-th grid, and the optimization objective is . Minimum; The sub-objective function for minimizing the total length of the inspection route is as follows: ； in, The Euclidean distance between two adjacent inspection points is the optimization objective. Minimum; The sub-objective function for achieving 100% coverage of inspection points is as follows: ； in, The optimization objective is to determine the number of guaranteed reachable points covered by route R. =0; The sub-objective function that achieves the highest satisfaction of the inspection time window constraint is as follows: ； in,[ , For each guaranteed access point pi, there is a corresponding allowed inspection time window. The optimization objective is to... =0; The sub-objective function that yields the highest load balancing across multiple inspection entities is as follows: ； Among them, w k Given the workload of the k-th inspection entity among K inspection entities, the optimization objective is: Minimum, meaning the difference in workload among the various inspection entities is minimized; In step 3.5, the spatial constraint is that the route must not enter the restricted area, as shown in the following formula: ； in, This is a set of coordinates for restricted areas within the park; routes that must pass through fire lanes must meet certain requirements. ∈ ,in This is a set of coordinates for fire escape routes. Access control requires that the route taken by the patrol entity must match its access control permissions, i.e., the set of access control points the route passes through. And it must belong to the set of permissions of the inspection entity. ,Right now ; The compliance constraints are that the key controlled areas traversed by the route must meet the requirements of the park's safety production standards regarding the length of stay and the order of inspections.

6. The inspection route planning method based on a dynamic risk heat map of a park, as described in claim 1, is characterized in that... The genetic-ant colony hybrid algorithm in step 4 includes the following improvements: Improvement 1: The grid risk weights of the dynamic risk heatmap are used as the core factor of the heuristic function of the ant colony algorithm; Improvement 2: The hard constraints of the park are embedded in the tabu table of the algorithm, and solutions that do not meet the constraints are automatically eliminated during the iteration process; Improvement 3: The crossover and mutation operation of the genetic algorithm is used to optimize the initial pheromone distribution of the ant colony algorithm, and the local search capability of the genetic algorithm is optimized through the positive feedback mechanism of the ant colony algorithm.

7. The inspection route planning method based on a dynamic risk heat map of a park, as described in claim 6, is characterized in that... In the genetic-ant colony hybrid algorithm in step 4, the number of ants, m, is 1.5 ± 0.5 times the number of must-reach inspection points N; the pheromone importance factor, γ = 1.2 ± 0.2, characterizes the degree of influence of pheromones on path selection. Risk heuristic factor: δ = 2.5 ± 0.5, characterizing the degree of influence of risk cost on path selection; pheromone evaporation coefficient: ρ = 0.1 ± 0.02, ranging from [0,1]; maximum number of iterations: MaxIter = 200 ± 10; genetic algorithm crossover probability: P c =0.7±0.1, Probability of mutation: P m =0.05±0.

01.

8. The inspection route planning method based on a dynamic risk heat map of a park, as described in claim 7, is characterized in that... In step 4, the algorithm iteratively solves the problem through the following steps: Step 4.1: Based on the full permutation of the inspection points, generate an initial path population of size m. At the same time, embed the park's hard constraints into the population generation rules to ensure that all paths in the initial population meet the mandatory constraints. Step 4.2: Using the overall optimization function F(R) from Step 3 as the fitness function, perform selection, crossover, and mutation operations on the initial population, iterating 50 times to obtain the optimized initial path population; based on the optimized initial path population, generate the initial pheromone distribution for the ant colony algorithm, as shown in the following formula: ； in, This is the initial pheromone base value. Let pheromone intensity be constant. The fitness value is the result of optimization by the genetic algorithm for the optimal path. Step 4.3: Using the roulette wheel method and introducing a risk heuristic factor, calculate the state transition probability of ant k choosing the next visiting point j at point i. The formula is: ； in, The improved risk heuristic function is defined as follows: =1 / ,in The total risk cost of the path from point i to point j is derived from the dynamic risk heatmap; This is the set of points that ant k has not visited. Points that do not meet the hard constraints of the park are removed from this set, i.e., the tabu list is embedded. Step 4.4: After completing one iteration, an improved pheromone rule combining global and local methods is used for updating to avoid the algorithm getting trapped in local optima. Specifically, for local pheromone updates, a local update is performed immediately after each ant traverses a path (i,j), as shown in the following formula: ； The global pheromone update, after traversing all ants' paths, only performs a global update on the optimal path of the current iteration, as shown in the following formula: ； in, This is the fitness value of the optimal path in this iteration; Step 4.5: Repeat steps 4.3 and 4.4 until the maximum number of iterations MaxIter is reached, or the fitness value of the optimal path remains unchanged for 30 consecutive iterations, at which point the iteration terminates; output the path with the smallest fitness value during the iteration process as the globally optimal inspection route, and simultaneously send it to the inspection execution end.

9. The inspection route planning method based on a dynamic risk heat map of a park, as described in claim 1, is characterized in that... Step 5 includes the following steps: Step 5.1: During the inspection process, monitor the dynamic risk heat map of the park generated in Step 2 in real time, and track the inspection progress and location information simultaneously; Step 5.2: Preset dynamic replanning trigger conditions; replanning will be automatically triggered when any condition is met. Specifically, the triggering conditions for dynamic replanning include: the risk value change in the core control area of ​​the park exceeds the preset threshold; the park adds a level 1 / level 2 high-risk alarm; the inspection execution progress deviates from the preset plan by more than the preset threshold; and the inspection execution terminal reports abnormal situations. Among them, the preset threshold for the risk value change in the core control area of ​​the park is 15%-30%, and the preset threshold for the inspection execution progress deviating from the preset plan is 20-40 minutes. Step 5.3: When replanning is triggered, steps 2 to 4 are executed again. Based on the latest dynamic risk heat map and inspection execution status, the updated global optimal inspection route is re-solved and generated. Step 5.4: Send the updated inspection route to the inspection execution terminal in real time, update the inspection task synchronously, and complete the real-time closed loop of the entire process of "data collection → heat map update → route optimization → execution feedback".

10. A patrol route planning system based on a dynamic risk heat map of a park, characterized in that, include: Data input layer: Acquires raw data to provide basic data support for subsequent full-process calculations; Specifically, it includes three types of data sources: basic static data of the park, real-time dynamic sensing data, and park inspection and control rules; Data preprocessing layer: The input end connects to the data input layer, and the output end connects to the dynamic risk heat map construction module of the core computing layer; This module receives all the raw data from the data input layer and sequentially performs data cleaning, outlier removal, spatiotemporal alignment, and normalization processing to unify the format and spatial coordinate system of all data and output standardized, highly consistent, and compliant data to ensure the timeliness and accuracy of subsequent calculations. Core Computing Layer: The three core functional modules are sequentially connected according to the data flow to realize the entire process of calculation from risk quantification to optimal route solution; the three core functional modules include: dynamic risk heat map construction module, multi-objective optimization model construction module, and global optimal inspection route solution module; Control and execution layer: Enables the issuance, execution, feedback, and dynamic adjustment of inspection tasks; includes: dynamic replanning closed-loop control module and inspection execution terminal; Visualized Control Layer: Corresponding to the park's intelligent control visualization platform, the input end synchronously connects to all modules of the core computing layer and the control execution layer, and synchronizes all data in real time, including dynamic risk heat maps, multi-objective optimization model calculation process, inspection route solution results, inspection task execution progress, replanning trigger records, etc., to realize the visualization display, task scheduling, anomaly alarm and data statistical analysis of the entire park inspection process.

Images