Method for generating beforehand prevention and control strategy for traffic safety risk of highway network in mountainous area

Abstract

The invention relates to the technical field of traffic safety, in particular to a beforehand prevention and control strategy generation method for traffic safety risks of a highway network in a mountainous area. Comprising the following steps: risk diagnosis and data preparation: collecting road basic information, historical traffic flow data, meteorological data and accident-prone point data of a mountainous area expressway network; constructing a risk assessment model: selecting road alignment, traffic flow, weather and environmental toughness indexes, determining index weights by adopting an improved analytic hierarchy process, constructing the risk assessment model through a fuzzy comprehensive evaluation method, and calculating the traffic safety risk level of each road section; a prevention and control strategy is generated; and strategy verification and iteration. According to the method, the total factor evaluation model is constructed by integrating the road network topological structure, the traffic flow characteristics and the meteorological sensitive section data, so that the problem of one-sided risk identification caused by single factor analysis in the prior art is solved, and systematic description of the mountainous area highway composite risk scene is realized.

Description

Technical Field

[0001] The present invention relates to the field of traffic safety technology, and in particular to a method for generating an advance prevention and control strategy for traffic safety risks in a mountainous expressway network. Background Art

[0002] Mountainous expressway networks, characterized by dense bridges and tunnels and steep longitudinal slopes, coupled with the dynamic nature of sudden increases in traffic volume during holidays, create a complex multi-dimensional risk scenario characterized by terrain constraints, sudden changes in traffic volume, and meteorological sensitivity. Existing prevention and control measures often rely on single road alignment indicators or historical accident data, lacking a coordinated analysis of network topology (such as node betweenness centrality), meteorologically sensitive areas (such as those prone to frequent fog), and holiday traffic flow characteristics. This results in one-sided risk assessments and static strategy generation, making it difficult to adapt to the ever-changing weather conditions and the short-term concentrated traffic demands of mountainous areas.

[0003] For example, Chinese patent CN202310130333.5 discloses a method for proactively preventing and controlling traffic safety on national and provincial trunk roads. This method generates strategies through accident data collection and risk stratification. However, this method does not address the unique topological structure of mountain highways, which often includes long downhill slopes and tunnel clusters, nor does it incorporate the spatiotemporal characteristics of traffic flows during holiday traffic surges. Consequently, it is unable to quantify the threat level of complex risk scenarios. Another example is Chinese patent CN201810783019.6, which discloses a method for predicting accident risks for traffic participants based on ensemble learning. While this method can identify high-risk individuals, it lacks correlation analysis between road network topological resilience (such as network connectivity indicators) and meteorologically sensitive periods. Furthermore, it lacks a closed-loop iterative mechanism for "risk assessment-strategy implementation-real-time feedback," making it difficult to dynamically adjust prevention and control strategies in the event of extreme weather events on mountain highways.

[0004] While the aforementioned technical solutions possess their own design advantages, they suffer from the following technical drawbacks: First, they lack in-depth multi-source data fusion: they fail to couple the unique mountainous road network topology parameters (such as tunnel spacing and longitudinal slope) and meteorological sensitivity thresholds with holiday traffic characteristics (such as the spatiotemporal distribution of traffic density), making it impossible to construct a multi-dimensional risk assessment system encompassing terrain, weather, and traffic. Second, they lack a dynamic response mechanism: strategy generation relies on pre-set rules and lacks the closed-loop optimization capability based on real-time data, making it difficult to cope with the complex holiday scenarios of sudden increases in traffic volume and sudden changes in weather on mountainous highways. In light of these factors, we propose a method for generating pre-emptive prevention and control strategies for traffic safety risks on mountainous highway networks. Summary of the Invention

[0005] The purpose of the present invention is to provide a method for generating an advance prevention and control strategy for traffic safety risks in mountainous expressway networks, so as to solve the problems of insufficient multi-source data fusion depth and lack of dynamic response mechanism proposed in the above background technology.

[0006] To solve the above technical problems, the present invention provides a method for generating a pre-emptive prevention and control strategy for traffic safety risks in a mountainous expressway network, comprising the following steps: S100, Risk Diagnosis and Data Preparation: Collect basic road information, historical traffic flow data, meteorological data, and accident-prone location data for mountain expressway networks. Combined with on-site surveys, clarify the road network topology and identify road alignment defects, holiday traffic characteristics, and core traffic safety risk factors in meteorologically sensitive sections. S200, Risk Assessment Model Construction: Select road alignment, traffic flow, meteorological and environmental resilience indicators, use the improved analytic hierarchy process to determine the indicator weights, construct a risk assessment model using the fuzzy comprehensive evaluation method, and calculate the traffic safety risk level of each road section; S300, Prevention and Control Strategy Generation: Based on the differences in traffic safety risk levels across road sections, formulate pre-emptive prevention and control strategies covering speed limit management, facility optimization, and emergency response layout, and clarify the applicable scenarios and implementation constraints of each pre-emptive prevention and control strategy; S400, Strategy Verification and Iteration: Verify the effectiveness of pre-emptive prevention and control strategies through traffic flow simulation, collect real-time traffic and meteorological data after the implementation of pre-emptive prevention and control strategies, dynamically adjust strategy parameters based on the collected real-time traffic and meteorological data, and build a closed-loop strategy iteration mechanism driven by risk status feedback.

[0007] As a further improvement of the present technical solution, in S100, basic road information, historical traffic flow data, meteorological data, and accident-prone point data of the mountain expressway network are collected, and the road network topology is determined in combination with on-site investigation, including the following steps: S110.1. Use a GIS system to collect road parameters, which shall include at least horizontal curve radius, longitudinal slope gradient, bridge span, tunnel clearance dimensions, and roadbed width; S110.2. Use a combination of loop detectors, microwave detectors, and video analysis equipment to collect traffic flow data, recording holiday OD traffic volume, vehicle type composition (large vehicles / small vehicles / hazardous goods vehicles), road section saturation, and vehicle speed percentiles; S110.3. Obtain precipitation intensity, visibility, road surface temperature, and icing warning index through weather stations; S110.4. Construct a road network topology model based on complex network theory. Nodes include toll stations, service areas, bridges and tunnels. Lines are associated with attribute parameters such as horizontal curve radius and longitudinal slope. Calculate node betweenness centrality and line connectivity efficiency.

[0008] As a further improvement of the present technical solution, in S100, identifying the core coercive factors of traffic safety risks of holiday traffic characteristics includes the following steps: S120.1. Set the holiday range to statutory holidays and peak tourist seasons (e.g., Spring Festival, National Day, and Spring Festival travel period), and extract the sub-data for the corresponding period of the historical traffic flow data collected in S100; S120.2. Extract traffic flow fluctuation characteristics, vehicle type mix, and continuous following distance distribution from traffic flow data collected by S100; S120.3. Construct a generalized ordered logit model, using accident severity (minor / moderate / serious) as the dependent variable and vehicle type mix and the probability of exceeding the continuous following distance limit as independent variables, to analyze the factors contributing to traffic safety risks unique to holidays and their weight rankings.

[0009] As a further improvement of the present technical solution, in S100, identifying the core threat factors of traffic safety risks in the meteorologically sensitive section includes the following steps: S130.1. Based on the meteorological data and accident-prone location data collected by S100, screen meteorologically sensitive periods when precipitation intensity reaches a preset sensitivity threshold, visibility is below a preset critical value, and road surface temperature is below a preset icing risk temperature; S130.2. Construct a spatiotemporal correlation analysis model, overlay the meteorologically sensitive periods with the road network topology (e.g., bridges, long downhill sections), and identify weather-road coupling risk areas; S130.3. Use the Bayesian network model, with meteorological parameters (precipitation intensity, road surface temperature) and road parameters (longitudinal slope, flat curve radius) as input and accident probability as output, to quantify the risk level of meteorologically sensitive sections.

[0010] Furthermore, the preset sensitive threshold can be set according to historical data. For example, in the high-speed scene in the mountainous area of ​​Yunnan, the sensitive threshold of precipitation intensity is set to ≥5mm / h, the critical visibility value is set to <500m, and the road surface temperature icing risk temperature is set to ≤0℃; for example, in the northwestern mountainous scene, it can be adjusted to precipitation intensity ≥3mm / h and visibility <800m to adapt to the meteorological characteristics of different regions.

[0011] As a further improvement of the present technical solution, in S200, determining the indicator weights using the improved analytic hierarchy process includes the following steps: S210.1. Constructing index order relationship based on G1 method: Building a set of indicators ,in, The total number of indicators (such as weather, road alignment, etc.); the importance order of indicators is determined through expert consultation ,in Indicates the important indicators; Adjacent sorting index and , define the importance proportional coefficient ,in, Indicator The weight of , which is used to balance the importance differences of adjacent indicators, avoid the deviation caused by imbalance in the subjective weight calculation, and provide reasonable constraints for accurately constructing the indicator weight system for risk assessment of mountain highway networks; By recursive formula , calculate the indicator weight vector ; The formula is: ; in, Indicates that from to Multiply the proportional coefficients of adjacent indicators; Indicates the sum of the consecutive multiplication results; S210.2. Calculate objective weights based on the entropy weight method: right evaluation object Indicator data matrix ;in, Indicates the The evaluation object is in The original data of each indicator; Calculate the Normalized value of an indicator (Eliminate the dimension effect and make different indicators comparable): ; Calculate the The entropy value of the indicator (Reflects the degree of data dispersion: the smaller the entropy value, the stronger the indicator's ability to distinguish): ; in, Normalize the entropy value to [0 1] interval (otherwise the entropy value range varies changes, making horizontal comparison impossible); Defining Metrics The coefficient of variation , normalize the difference coefficient and get the objective weight vector : ; S210.3. Fusion of subjective and objective weights: Define weight fusion coefficient , calculate the comprehensive weight by linear weighting formula : ; in, ; It is used to reflect the relative importance of subjective judgment in the comprehensive weight, and to determine the optimal value through the goodness of fit between historical accident data and model prediction results.

[0012] As a further improvement of this technical solution, the environmental resilience indicators include network topology resilience, transportation service resilience, and environmental adaptability resilience, where: The network topology resilience is based on the road network topology model constructed by S110.4 to calculate the node betweenness centrality , line connectivity efficiency ; The traffic service resilience is based on the historical traffic flow data collected in step S100 to calculate the peak period traffic capacity redundancy. , the time it takes to restore the road network after the accident; The environmental adaptability resilience is based on the meteorologically sensitive sections identified in S130.1, and the proportion of meteorologically sensitive sections is calculated. , slope stability coefficient .

[0013] As a further improvement of the present technical solution, in S200, constructing a risk assessment model by using a fuzzy comprehensive evaluation method includes the following steps: S220.1. Construct an intuitionistic fuzzy evaluation matrix: In view of the uncertainty of mountainous meteorology and the mixed characteristics of traffic flow, the risk level is divided into four levels: low, medium, high and very high. The relationship between the characterization indicator and the risk level, where: Indicates the degree of membership of the indicator to a certain risk level; Indicates the non-membership degree of the indicator to a certain risk level; Indicates hesitation, and , used to quantify the uncertainty in mountainous scenarios; S220.2. Calculate the comprehensive evaluation vector: Based on the comprehensive weight determined in S210.3 , the intuitionistic fuzzy numbers of each indicator are fused through the weighted average operator, and the formula is: ; in, Indicates the The degree of membership of an indicator to a certain risk level; Indicates the The non-membership degree of an indicator to a certain risk level; Indicates the total number of indicators; represents the intuitionistic fuzzy comprehensive evaluation vector after weighted fusion; S220.3. Determine the risk level: Using dual function sorting rules to analyze comprehensive evaluation vectors : Score function : Defined as , the larger the value, the higher the risk level (e.g. >S , the risk level is higher); Exact function : Defined as , used to solve the sorting when the scores are equal (if ,but The smaller (higher the hesitation), the higher the risk level, e.g. Compare higher risk).

[0014] As a further improvement of this technical solution, in S300, formulating a pre-emptive prevention and control strategy includes the following steps: S310.1. Adopt a speed limit adjustment and facility reinforcement strategy on low-risk road sections; this applies to non-holiday and normal weather conditions. S310.2: On medium-risk roads, adopt a strategy of speed limit adjustment and facility reinforcement; this strategy is applicable during peak holiday periods or mildly weather-sensitive scenarios. S310.3: Implement mandatory speed limits, facility renovation, and emergency stationing strategies on high-risk roads; this strategy is applicable to scenarios with high passenger flow during extreme weather or holidays. S310.4: Temporary closures and long-term rerouting should be implemented in extremely high-risk sections. This strategy is applicable to areas with geological hazards or major linear defects.

[0015] As a further improvement of this technical solution, in S400, the effectiveness of the pre-emptive prevention and control strategy is verified through traffic flow simulation, specifically including: Based on the road network topology data (including linear parameters) from S200, the traffic fluctuation characteristics (vehicle type ratio, peak flow) from S120.2, and the meteorological sensitive period data (fog / icing parameters) from S130.1, a digital twin model including lane-level geometric models was constructed using professional simulation software. Based on the S300's pre-emptive prevention and control strategy, extreme weather and holiday peak scenarios were simulated to verify the traffic stability, accident risk and traffic efficiency after the strategy was implemented.

[0016] As a further improvement of the present technical solution, in S400, dynamically adjusting the policy parameters and building a policy closed-loop iteration mechanism specifically include: Real-time data collection: Utilizing loop detectors, microwave detectors, and video analysis equipment, we collect real-time traffic flow data (volume, vehicle speed, and lane occupancy). We also obtain visibility, precipitation intensity, and road surface temperature from weather stations to create a minute-by-minute dynamic data set. Dynamic adaptation of strategy parameters: When real-time traffic exceeds the preset peak threshold, lane occupancy exceeds the congestion threshold, visibility falls below the meteorological sensitivity threshold, or road surface temperature falls below the freezing risk temperature, the corresponding level of prevention and control strategy is automatically triggered; Risk-strategy closed-loop iteration: Real-time data is input into the risk assessment model, the risk level of the road section is recalculated, and the triggering conditions and implementation parameters of the prevention and control strategy are adjusted according to the changes in the risk level to form a dynamic optimization mechanism.

[0017] Compared with the prior art, the present invention has the following beneficial effects: 1. This invention integrates road network topology, traffic flow characteristics, and meteorologically sensitive section data to construct a comprehensive assessment model covering "roads, traffic, and the environment." This model addresses the one-sided risk identification problem caused by single-factor analysis in existing technologies and systematically characterizes complex risk scenarios on mountainous highways. 2. This invention establishes a closed-loop process of "risk assessment-strategy generation-real-time verification." It dynamically adjusts prevention and control parameters (such as speed limit thresholds and emergency response layouts) based on real-time traffic flow and meteorological data. This addresses the problem that traditional static strategies are insufficiently adaptable to sudden weather changes and traffic surges in mountainous areas, thereby improving the timeliness and accuracy of policy responses. 3. This invention incorporates indicators such as network topology resilience and transportation service resilience. Based on the terrain characteristics of mountain highways, which include dense bridges and tunnels and steep longitudinal slopes, it develops differentiated strategies for "low-to-very-high" risk levels (e.g., mandatory speed limits and emergency stations on high-risk sections). This addresses the disconnect between existing prevention and control measures and terrain characteristics, resulting in a precise prevention and control solution tailored to mountainous scenarios. 4. This invention constructs a digital twin system that includes a lane-level geometric model and combines it with simulations of extreme weather and holiday peak scenarios to verify traffic flow stability and accident risks after the strategy is implemented. This avoids the lack of empirical evidence for the effectiveness of strategies in existing technologies and provides quantifiable technical support for practical applications. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 Schematic diagram of the method steps of the present invention. DETAILED DESCRIPTION

[0019] The following will provide a clear and complete description of the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0020] Example 1 like Figure 1 As shown, this embodiment provides a method for generating a pre-emptive prevention and control strategy for traffic safety risks in a mountainous expressway network, including the following steps: S100, Risk Diagnosis and Data Preparation: Collect basic road information, historical traffic flow data, meteorological data, and accident-prone location data for mountain expressway networks. Combined with on-site surveys, clarify the road network topology and identify road alignment defects, holiday traffic characteristics, and core traffic safety risk factors in meteorologically sensitive sections. In this step, in S100, basic road information, historical traffic flow data, meteorological data, and accident-prone point data of the mountain highway network are collected, and the road network topology is determined by combining on-site investigation, including the following steps: S110.1. Use a GIS system to collect road parameters, which shall include at least horizontal curve radius, longitudinal slope gradient, bridge span, tunnel clearance dimensions, and roadbed width; S110.2. Use a combination of loop detectors, microwave detectors, and video analysis equipment to collect traffic flow data, recording holiday OD traffic volume, vehicle type composition (large vehicles / small vehicles / hazardous goods vehicles), road section saturation, and vehicle speed percentiles; S110.3. Obtain precipitation intensity, visibility, road surface temperature, and icing warning index through weather stations; S110.4. Construct a road network topology model based on complex network theory. Nodes include toll stations, service areas, bridges and tunnels. Lines are associated with attribute parameters such as horizontal curve radius and longitudinal slope. Calculate node betweenness centrality and line connectivity efficiency.

[0021] As a further illustration of this step, this embodiment can fuse GIS road parameters, traffic flow data, and meteorological data using a spatiotemporal alignment algorithm. For example, using 15-minute time slices, segment-level traffic flow data (e.g., saturation, speed percentiles) is matched with meteorological parameters for the corresponding time period (e.g., precipitation intensity, visibility), forming a "road-traffic-weather" triplet data unit. Furthermore, this embodiment can employ a data cleaning strategy: interpolation (e.g., using linear interpolation or a KNN algorithm) is performed for time periods with a missing traffic flow data rate exceeding 20%, and outliers (e.g., speed percentiles exceeding ±3 standard deviations from the mean) are flagged and manually reviewed and corrected.

[0022] Furthermore, in this embodiment, the fusion of GIS road parameters, traffic flow data, and meteorological data through the spatiotemporal alignment algorithm includes the following steps: First, the timestamps of traffic flow detection equipment and weather stations are uniformly converted to Unix timestamps to eliminate format differences and ensure consistent time bases. For single-second sampling missing in traffic flow data (overall missing rate ≤ 5%), the short-term continuity characteristics of traffic flow are utilized, and a forward filling method is used to replace the missing value with valid sampling data 1 second before the missing moment. For 5-minute slice missing in meteorological data (overall missing rate ≤ 3%), the neighbor interpolation method is used based on the spatiotemporal continuity of mountainous meteorological changes to fill the missing value with meteorological data of the previous valid time slice. This preliminary regularization of homologous data lays a quality foundation for the subsequent spatiotemporal alignment of cross-source data. Then, after completing data preprocessing, we proceeded to align the multi-source data in the time dimension, using dynamic data aggregation to unify the time bases of each data. We first analyzed the 300-second (or 5-minute) update cycle of the weather station and the 1-second sampling cycle of the traffic flow detection equipment, and selected the lowest common multiple of these two cycles, 300 seconds, as the length of the time slice. The advantage of this is that it ensures that each time slice contains both a complete piece of meteorological data (covering a 5-minute monitoring period) and the aggregation of 300 continuous traffic flow sampling data (this data can be used to calculate the statistical laws of dynamic characteristics such as average vehicle speed, flow rate, and occupancy rate within the interval). Next, based on the completion of the time dimension alignment work, we will promote the alignment of multi-source data in the spatial dimension to achieve spatial matching of static road parameters and dynamic monitoring data. Based on the 100-meter pile identification of GIS roads (such as the road pile number K12+300, which is a common road identification), the mountain highway is divided into 100-meter granularity spatial units (for example, the section K12+300-K12+400 is the 123rd unit), and each spatial unit is associated with the horizontal curve radius. (used to reflect the curvature of the curve, the unit is meter), longitudinal slope (Reflects the absolute value of the slope, the unit is %), roadbed width (Reflects the number and width of lanes, in meters) and other static road parameters; Traffic flow detectors can be directly matched to corresponding spatial units using their own latitude and longitude coordinates. For meteorological stations in mountainous areas, which are usually spaced 5 km or more apart, inverse distance weighted interpolation is used to assign meteorological parameters (such as precipitation intensity, wind speed, temperature, etc.) to each spatial unit. The interpolation formula is as follows: ;in, Meteorological parameter values ​​representing the target spatial unit; It is Observations from weather stations; For the target unit to The straight-line distance between weather stations, in meters, raised to a power of 2 (a well-known value that strikes a balance between distance decay and computational complexity); Represents the number of weather stations involved in the spatial interpolation calculation. Through spatial discretization and interpolation calculation, a multi-source data unit with consistent time and space is constructed; Finally, we conduct fusion consistency verification to ensure that the data after time and space alignment is logically self-consistent. We select at least 100 time slices (covering different weather conditions such as sunny, rainy, and snowy days, as well as peak and off-peak traffic conditions) and calculate the Pearson correlation coefficient between indirect meteorological indicators derived from traffic flow (such as the standard deviation of vehicle speed on wet roads, which can be used to reflect the impact of precipitation on driving) and direct observations from weather stations. , the calculation formula is: ; among them, among them, It is an indirect indicator of traffic flow; is the direct observation value of the weather station; 、 are the means of the indirect indicators of traffic flow and the direct observations of weather stations, respectively; To verify the sample size; is the sample mean of the indirect indicator of traffic flow; is the sample mean of the direct observations from the weather station; Is the sample number, used to traverse each set of data involved in the verification. , the spatiotemporal alignment is deemed valid; if it fails to meet this standard, the system retrospectively checks parameters such as the time slice length and spatial interpolation power, makes corrections, and then revalidates the data. Consistency-verified fused data provides subsequent risk diagnosis models with input datasets that strictly match the spatiotemporal dimensions, ensuring controllable accuracy during model training and prediction.

[0023] In this step, in S100, identifying the core coercive factors of traffic safety risks of holiday traffic characteristics includes the following steps: S120.1. Set the holiday range to statutory holidays and peak tourist seasons (e.g., Spring Festival, National Day, and Spring Festival travel period), and extract the sub-data for the corresponding period of the historical traffic flow data collected in S100; S120.2. Extract traffic flow fluctuation characteristics, vehicle type mix, and continuous following distance distribution from traffic flow data collected by S100; S120.3. Construct a generalized ordered logit model, using accident severity (minor / moderate / serious) as the dependent variable and vehicle type mix and the probability of exceeding the continuous following distance limit as independent variables, to analyze the factors contributing to traffic safety risks unique to holidays and their weight rankings.

[0024] As a further illustration of this step, statutory holidays and peak tourist seasons are divided into analysis periods. The quantification method of traffic fluctuation characteristics is as follows: First, calculate the percentage difference between holiday traffic and weekday traffic , extract continuously for 30 minutes The period exceeding ±30% is considered as a high-risk period; among them, Traffic flow of the road section during holidays; The traffic volume of the same road section during the same period on weekdays; Then, the modified Herfindahl-Hirschman index was used to calculate the vehicle type complexity. : ,in, For the The ratio of vehicle models of the same type. The closer the value is to 1, the more uniform the vehicle model is and the lower the risk is. is the total number of vehicle types classified; Finally, a generalized ordered logit model was constructed, with the severity of the accident as the dependent variable, the vehicle type mix and the probability of exceeding the following distance as the independent variables, and significant variables were screened by stepwise regression method ( ), and the Hosmer-Lemeshow test was used to assess the goodness of fit of the model.

[0025] In this step, in S100, identifying the core threat factors of traffic safety risks in weather-sensitive sections includes the following steps: S130.1. Based on the meteorological data and accident-prone location data collected by S100, screen meteorologically sensitive periods when precipitation intensity reaches a preset sensitivity threshold, visibility is below a preset critical value, and road surface temperature is below a preset icing risk temperature; Furthermore, the preset sensitive threshold can be set according to historical data. For example, in the high-speed scene in the mountainous area of ​​Yunnan, the sensitive threshold of precipitation intensity is set to ≥5mm / h, the critical visibility value is set to <500m, and the road surface temperature icing risk temperature is set to ≤0℃; for example, in the northwestern mountainous scene, it can be adjusted to precipitation intensity ≥3mm / h and visibility <800m to adapt to the meteorological characteristics of different regions.

[0026] As a further illustration of this step, this embodiment uses the quantile method to set the threshold based on 5-10 years of historical data: Precipitation intensity threshold: The 90th percentile of the precipitation intensity at the time of the statistical accident (i.e., the precipitation intensity at 90% of the accidents is ≤ this value); Visibility threshold: Take the 10th percentile of visibility during the period of high accident incidence (i.e., visibility is ≤ this value when 10% of accidents occur); Pavement temperature threshold: determined by combining the lower temperature limit corresponding to the pavement icing accident with the effective temperature of the snow-melting agent.

[0027] Furthermore, the missing data of the weather stations are repaired by spatiotemporal interpolation method, and the spatial dimension is repaired by inverse distance weighting method, the formula is: ; in, Indicates interpolation points Estimated values ​​of meteorological parameters at the location (e.g. precipitation intensity, visibility); Indicates a known weather station The measured value at ) indicates a weather station With interpolation points spatial distance; represents the distance weight coefficient; Indicates the number of neighboring weather stations involved in the interpolation (usually 3 to 5, adjusted according to the station density).

[0028] S130.2. Construct a spatiotemporal correlation analysis model, overlay the meteorologically sensitive periods with the road network topology (e.g., bridges, long downhill sections), and identify weather-road coupling risk areas; As a further explanation of this step, this embodiment implements the coupling risk identification between meteorological sensitive periods and road network topology through the following steps: First, the GIS system is used to transform the key structures of mountain highways into spatial elements: Bridge: Defined as a surface feature, with attributes including span, elevation, and terrain (e.g., valley / ridge) Long downhill road section: defined as a line feature, with the longitudinal slope gradient, slope length and the coordinates of the safe lane position marked; Then, the filtered meteorological sensitive period data (such as precipitation intensity ≥ 5 mm / h) are converted into 100m×100m grids, and each grid value corresponds to the meteorological parameters within the period, forming a continuous meteorological field; Finally, the ArcGIS Spatial Join tool was used to perform the following operations: ; Among them, "∩" represents the spatial intersection operation; for example, the grid area with "precipitation intensity ≥ 5mm / h and located within 200 meters of the bridge body and both sides" is extracted as the "precipitation-bridge" coupling risk section.

[0029] S130.3. Use the Bayesian network model, with meteorological parameters (precipitation intensity, road surface temperature) and road parameters (longitudinal slope, flat curve radius) as input and accident probability as output, to quantify the risk level of meteorologically sensitive sections.

[0030] As a further explanation of this step, in this embodiment, a Bayesian network model is used to implement a quantitative assessment of weather-road coupling risk, which is specifically implemented through the following whole-stage process: First, construct a Bayesian network topology containing input nodes and output nodes: The input nodes include meteorological parameters (precipitation intensity is divided into low / medium / high, corresponding to <3 mm / h, 3-5 mm / h, and ≥5 mm / h; road surface temperature is divided into >0°C / ≤0°C) and road parameters (longitudinal slope is divided into <3% / ≥3% with a critical value of 3%; flat curve radius is divided into >500m / ≤500m with a critical value of 500m). The output node is the accident probability (divided into three levels: low / medium / high, corresponding to the probability intervals [0, 0.3), [0.3, 0.7), and [0.7, 1]). The K2 algorithm is used for structural learning, calculating conditional probabilistic dependencies between nodes based on 5-10 years of historical meteorological data, road parameters, and accident records. For example, data training determines the direct dependency chain between the "accident probability" node and the "precipitation intensity" and "longitudinal slope" nodes, forming a directed acyclic graph (DAG) structure to ensure that the network topology conforms to the causal logic of "weather-road-accident"; Then, a conditional probability table between nodes is constructed using the maximum likelihood estimation method, and the joint probability distribution of each node state is calculated using historical accident data as a sample. For example, regarding the impact of "precipitation intensity" and "longitudinal slope gradient" on "accident probability", the conditional probability value of "high" accident probability when precipitation intensity is "high" (≥5mm / h) and longitudinal slope gradient is ≥3% is calculated based on historical accident data statistics of typical sections of mountain highways. For the joint conditional probability of multiple nodes, the chain rule is used to decompose it into the product of conditional probabilities, such as: ,pass and Calculate the correlation between the two conditions (assuming independence, actual verification needs to be combined with statistical laws); ensure that each conditional probability value is based on the statistical laws of the measured data; Next, we perform posterior probability inference based on Bayes’ theorem, and the formula is: ; in, is the posterior probability of an accident under known weather and road conditions, is the likelihood of the corresponding condition when the accident occurs, is the prior probability of historical accidents (e.g., the annual accident rate of a certain road section is 0.05), is the joint probability of weather and road conditions (evidence factor). During the inference process, the probability distribution of each node state is calculated through forward propagation. For example, when the input precipitation intensity is 6mm / h (high) and the longitudinal slope is 4% (≥3%), the accident probability calculated based on the conditional probability table is 0.62, corresponding to the "medium risk" level (0.3≤0.62<0.7); Finally, 10-fold cross-validation was used to evaluate model performance. Historical data was randomly divided into 10 subsets, with nine selected for model training and one for testing. After iterations, the average AUC (area under the curve) was calculated, requiring a value of ≥0.75 to ensure the accuracy of risk level predictions. To address the regional characteristics of different mountainous areas (e.g., rainy southwest, foggy northwest, and snowy northeast), the model was adapted to regional variations by adjusting meteorological parameter thresholds (e.g., critical values ​​for precipitation intensity, visibility, and road surface temperature) and retraining the Bayesian network's conditional probability table based on local historical data. For example, in the foggy northwest mountainous areas, when the visibility threshold was adjusted to <800m, the conditional probability relationship between the "low visibility" node and the "accident probability" node was simultaneously updated to ensure that the risk assessment results were consistent with the local weather-road coupling characteristics.

[0031] S200, Risk Assessment Model Construction: Select road alignment, traffic flow, meteorological and environmental resilience indicators, use the improved analytic hierarchy process to determine the indicator weights, construct a risk assessment model using the fuzzy comprehensive evaluation method, and calculate the traffic safety risk level of each road section; In this step, in S200, determining the indicator weights using the improved analytic hierarchy process includes the following steps: S210.1. Constructing index order relationship based on G1 method: Building a set of indicators ,in, The total number of indicators (such as weather, road alignment, etc.); the importance order of indicators is determined through expert consultation ,in Indicates the important indicators; It is worth noting that when conducting risk assessments for complex risk scenarios on mountainous expressways, we invited five or more experts with the title of Senior Road Engineering Engineer and experience in at least two mountainous expressway safety assessment projects (covering diverse mountainous environments, such as those in the challenging mountainous areas of Northwest China and the foggy mountainous areas of Yunnan) to independently score each risk indicator using a 1–9 scale: The scores of those rated as "high risk" are averaged and converted into "high risk membership"; The scores of those rated “not high risk” are averaged and converted into “non-high risk membership”; The sum of the two scores is required to be ≤ 9 points, and the remaining score corresponds to the "hesitation" (the uncertainty interval of the expert's risk judgment); Finally, the average of all experts’ scores is taken to construct an intuitive fuzzy judgment matrix as the input basis for risk assessment.

[0032] Adjacent sorting index and , define the importance proportional coefficient ,in, Indicator The weight of , which is used to balance the importance differences of adjacent indicators, avoid deviations caused by imbalanced proportions in subjective weight calculations, and provide reasonable constraints for accurately constructing an indicator weight system for risk assessment of mountain highway networks; although there are differences in the importance of risk assessment indicators for mountain highways (such as road alignment, weather, traffic flow, environmental resilience, etc.), the degree of difference is usually "gradual" rather than "extreme". For example, although the "proportion of meteorologically sensitive sections" and the "longitudinal slope" are of different importance, the former will not be "extremely important" than the latter (such as the importance ratio exceeds 3 times); conversely, they will not be completely equally important (the ratio is less than 1). Setting It can fit the actual correlation of such indicators - reflecting the order relationship of "the previous indicator is more important than the next indicator", while avoiding exaggerating the differences and causing weight imbalance, which meets the assessment needs of multi-factor coupling risks in mountainous scenarios.

[0033] By recursive formula , calculate the indicator weight vector ; The formula is: ; in, Indicates that from to Multiply the proportional coefficients of adjacent indicators; Indicates the sum of the consecutive multiplication results; In this step, for example, three indicators For example, if ,but: ; ; .

[0034] S210.2. Calculate objective weights based on the entropy weight method: right evaluation object Indicator data matrix ;in, Indicates the The evaluation object is in The original data of each indicator; Calculate the Normalized value of an indicator (Eliminate the dimension effect and make different indicators comparable): ; Calculate the The entropy value of the indicator (Reflects the degree of data dispersion: the smaller the entropy value, the stronger the indicator's ability to distinguish): ; in, Normalize the entropy value to [0 1] interval (otherwise the entropy value range varies changes, making horizontal comparison impossible); Defining Metrics The coefficient of variation , normalize the difference coefficient and get the objective weight vector : ; As a further explanation of this step, in this embodiment, first evaluation object Indicator data matrix , first use the interquartile range (IQR) method to deal with outliers: calculate the Quartiles of indicators , the outlier threshold is and , the data exceeding the threshold are repaired by interpolation of adjacent means; then the Normalized value of an indicator , calculate the The entropy value of the indicator , and define the coefficient of variation , and finally get the objective weight vector .

[0035] S210.3. Fusion of subjective and objective weights: Define weight fusion coefficient , calculate the comprehensive weight by linear weighting formula : ; in, ; It is used to reflect the relative importance of subjective judgment in the comprehensive weight, and to determine the optimal value through the goodness of fit between historical accident data and model prediction results.

[0036] As a further explanation of this step, in this embodiment, the weight fusion coefficient is defined as , determine the optimal value through leave-one-out cross-validation: divide the historical data into copy, traverse Calculate the Kappa coefficient between the predicted risk level and the actual accident level, and select the Kappa coefficient that is suitable for reflecting the subjective judgment. The comprehensive weight formula is: ; Need to meet and passed the consistency test ( , where CI is the consistency index and RI is the random consistency index).

[0037] In this step, the environmental resilience indicators include network topology resilience, transportation service resilience, and environmental adaptability resilience, where: The network topology resilience is based on the road network topology model constructed by S110.4 to calculate the node betweenness centrality , line connectivity efficiency ; The traffic service resilience is based on the historical traffic flow data collected in step S100 to calculate the peak period traffic capacity redundancy. , the time it takes to restore the road network after the accident; The environmental adaptability resilience is based on the meteorologically sensitive sections identified in S130.1, and the proportion of meteorologically sensitive sections is calculated. , slope stability coefficient .

[0038] As a further explanation of this step, the quantification method of the environmental resilience index in this embodiment is as follows: First, from the perspective of network topology resilience, the node betweenness centrality and line connectivity efficiency are calculated based on the road network topology model, where the node betweenness centrality The calculation formula is: ,in, Representation node arrive The total number of shortest paths, Indicates passing through node The number of shortest paths is used to reflect the criticality of nodes and the efficiency of line connectivity. The calculation formula is: , For road sections arrive The shortest distance of the road network is taken as the indicator; Then, for traffic service resilience, the capacity redundancy during peak hours and the time it takes for the road network to recover after an accident are calculated based on historical traffic flow data. The calculation formula is: ,in, For peak hour traffic during holidays, The traffic capacity of the road section is designed, and the recovery time after the accident is based on the average handling time of similar accidents in history; Finally, in terms of environmental resilience, the proportion of meteorologically sensitive sections and slope stability coefficient are calculated by combining meteorological and geological data (such as through on-site drilling sampling). The calculation formula is: ;in, is the total length of the meteorological sensitive section, is the total length of the road network; slope stability coefficient The limit equilibrium method is used for calculation, and the formula is: ;in, For the The cohesion of the soil strips, is the weight of the soil strip, is the slope angle, is the pore water pressure, is the internal friction angle, For the The length of a soil strip; the parameter value is taken from the geological survey report.

[0039] In this embodiment, in S200, constructing a risk assessment model using the fuzzy comprehensive evaluation method includes the following steps: S220.1. Construct an intuitionistic fuzzy evaluation matrix: In view of the uncertainty of mountainous meteorology and the mixed characteristics of traffic flow, the risk level is divided into four levels: low, medium, high and very high. The relationship between the characterization indicator and the risk level, where: Indicates the degree of membership of the indicator to a certain risk level; Indicates the non-membership degree of the indicator to a certain risk level; Indicates hesitation, and , used to quantify the uncertainty in mountainous scenarios; S220.2. Calculate the comprehensive evaluation vector: Based on the comprehensive weight determined in S210.3 , the intuitionistic fuzzy numbers of each indicator are fused through the weighted average operator, and the formula is: ; in, Indicates the The degree of membership of an indicator to a certain risk level; Indicates the The non-membership degree of an indicator to a certain risk level; Indicates the total number of indicators; represents the intuitionistic fuzzy comprehensive evaluation vector after weighted fusion; S220.3. Determine the risk level: Using dual function sorting rules to analyze comprehensive evaluation vectors : Score function : Defined as , the larger the value, the higher the risk level (e.g. , the risk level is higher); Exact function : Defined as , used to solve the sorting when the scores are equal (if ,but The smaller (higher the hesitation), the higher the risk level, e.g. Compare higher risk).

[0040] As a further explanation of this step, first, this embodiment uses the frequency statistics method to determine the membership of quantitative indicators (such as precipitation intensity and longitudinal slope). : ; Among them, the sensitivity threshold is determined based on S130.1 (e.g. 5mm / h in mountainous areas of Yunnan); Secondly, this embodiment adopts an expert scoring method for qualitative indicators (e.g. “slope stability”): 3-5 geotechnical engineering experts are organized to score the degree to which the indicator belongs to a certain risk level on a 10-point scale, and the normalized score is used as the membership degree (e.g. 8 points corresponds to =0.8).

[0041] As a further explanation of this step, the non-membership degree in this embodiment Experts assign values ​​based on the uncertainty of mountain environments, which must meet the following requirements: For example: regular road section: (Hesitation ); sections with frequent fog: (Enhanced uncertainty representation, ).

[0042] As a further explanation of this step, in the calculation of the comprehensive evaluation vector, three indicators are used Take the evaluation of the "high risk" level as an example: Indicator weight: ; Intuitionistic fuzzy numbers: ; Comprehensive evaluation vector calculation: ; Score function: , corresponding to the "high risk" level .

[0043] As a further explanation of this step, the risk level classification in this embodiment is based on industry standards and historical accident statistics, as follows: First, this embodiment refers to the "Highway Project Safety Assessment Specification" (JTGB05-2015) and the "Mountain Highway Traffic Engineering and Along-line Facilities Design Specification" (JTG / TD81-2017), and divides the risk level into four levels: low, medium, high, and very high. , as the basis for judgment: Low risk: , corresponding to the daily inspection needs of regular road sections; Medium risk: , dynamic speed limit and other control measures need to be activated; High risk: , lane control or emergency station deployment is required; Very high risk: , triggering emergency responses such as main line closure.

[0044] Secondly, the accident rate association logic in this embodiment is as follows: the risk level is positively correlated with the historical accident rate. For example, the accident rate range corresponding to the high-risk level refers to the typical value of the five-year statistics of a certain mountain highway (such as 0.3-0.8 times / year / km). In specific applications, the level threshold needs to be calibrated according to the historical data of the project location to ensure that the classification standard matches the local safety level.

[0045] Furthermore, for emergency stations on “extremely high-risk sections” of mountain highways (single-side spacing ≤ 5km), the following equipment can be deployed: Anti-skid sand reserve: 1m3 per kilometer for each road section with an average annual snowy weather of ≥10 days 3 , single station reserve 5m 3 (Adopt waterproof modular sand storage box for quick access); Trailer specifications: Choose a four-wheel drive heavy-duty trailer (e.g., rated traction capacity of 30 tons, maximum climbing ability of 25% slope), suitable for towing on steep slopes in mountainous areas; Warning and detection equipment: Solar strobe light (life ≥ 72 hours, automatic light-sensing start and stop); Foldable temporary speed limit signs (including 20km / h and 40km / h signs, with supporting magnetic fixing devices); Portable road friction coefficient tester (response time ≤ 1 minute, real-time feedback on anti-skid performance); Communication guarantee: Equipped with emergency satellite communication terminals (supporting short message transmission in no-signal areas) to ensure information exchange under extreme conditions.

[0046] Furthermore, in order to achieve accuracy in risk assessment, this embodiment makes the following settings: First, a 5-fold cross-validation was used to ensure the reliability of the risk assessment model: the historical accident data was evenly divided into 5 subsets according to the time series. 4 subsets were selected as training sets each time to construct the intuitive fuzzy evaluation matrix and calibrate the indicator weights; the remaining 1 subset was used as a test set to verify the consistency between the risk level prediction results and actual accidents. The prediction accuracy was quantified by calculating the Kappa coefficient, which is calculated as follows: ; among them, among them is the consistency rate between the predicted and actual grades, is the expected consistency rate of random predictions; and , ensuring the stability and universality of the model on different data subsets; Then, based on the geographical and climatic characteristics of mountainous areas, this embodiment dynamically adjusts the model parameters through the parameter calibration-effect verification-iterative optimization process: In rainy mountainous areas in the southwest: Because cluster fog and heavy rainfall have a significant impact on driving safety, the weight of the "precipitation intensity" indicator will be increased by 10% (for example, from 0.2 to 0.22), and the hesitation index will be adjusted for cluster fog scenes. Set to 0.3 (i.e. ), strengthen the representation of meteorological uncertainty; In the snowy mountainous areas of Northeast China, the icing risk threshold is set at a road surface temperature of -5°C or less. The membership of the "road surface temperature" indicator is reconstructed based on the local historical frequency of icing accidents. , and adjust the scoring function threshold for extremely high risk levels to , adapted to the rapid evolution of accident causes in ice and snow environments; Foggy mountainous areas in the northwest: Based on the local fog distribution pattern, the visibility sensitivity threshold is set to less than 800m, and the conditional probability table of the "low visibility" node in the Bayesian network is simultaneously updated to improve the model's ability to identify foggy accidents.

[0047] Furthermore, this embodiment achieves regional adaptation through a three-step approach: data collection, expert consultation, and parameter calibration. First, historical meteorological data (at least five years), road network design parameters, and accident records for the target mountainous area are collected to form a standardized dataset. Next, a team of at least three experts in transportation engineering and geology is assembled to make targeted adjustments to indicator weights, meteorological thresholds, and intuitive fuzzy numbers. For example, in the foggy mountainous areas of the northwest, the visibility sensitivity threshold is set to <800 meters. Finally, the adapted model's prediction accuracy is verified through backtesting of historical data. When the Kappa coefficient improves by 15% or more, the adapted parameters are considered effective, ensuring that the model output aligns with the evolution of local traffic safety risks.

[0048] S300, Prevention and Control Strategy Generation: Based on the differences in traffic safety risk levels across road sections, formulate pre-emptive prevention and control strategies covering speed limit management, facility optimization, and emergency response layout, and clarify the applicable scenarios and implementation constraints of each pre-emptive prevention and control strategy; In this step, in S300, formulating a pre-emptive prevention and control strategy includes the following steps: S310.1. Adopt a speed limit adjustment and facility reinforcement strategy on low-risk road sections; this applies to non-holiday and normal weather conditions. As a further explanation of this step, low-risk road sections are applicable to non-holidays (e.g., average daily traffic volume < 80% of the designed capacity) and normal weather conditions (e.g., precipitation < 3 mm / h, visibility > 800 m, road surface temperature > 5°C). Specifically: the 85% operating speed of the road section is calculated through traffic flow detection data. (That is, 85% of the vehicles on the road do not exceed this speed, representing the stable driving state of most vehicles). A safety correction factor is introduced to formulate the speed limit: ; Among them, the coefficient This will reduce vehicle speeds by 10%, creating a safety margin to handle unexpected situations in mountainous areas (such as falling rocks and sharp curves). Simultaneously, sections with insufficient roadside clear zones will be upgraded to SB-class corrugated beam guardrails in accordance with the "Highway Guardrail Safety Performance Evaluation Standard" (JTGB05-01). Warning sign spacing will be shortened to ≤300m on long straight sections (compared to the standard 500m spacing), reinforcing the "Drive with Caution in Mountainous Areas" message. During implementation, speed limit adjustments will be verified for their impact on traffic efficiency using traffic simulation tools such as VISSIM, and facility reinforcements will be prioritized during maintenance windows.

[0049] S310.2: On medium-risk roads, adopt a strategy of speed limit adjustment and facility reinforcement; this strategy is applicable during peak holiday periods or mildly weather-sensitive scenarios. As a further explanation of this step, medium-risk sections are suitable for holiday peaks (e.g., average daily traffic volume ≥ 80% of the designed capacity) or mildly weather-sensitive scenarios (e.g., precipitation 3-5 mm / h, visibility 500-800 m, road surface temperature 0-5°C). Specifically: speed limit on low-risk sections Based on this, the meteorological correction factor is introduced Optimize speed limit: ;in, To optimize speed limits, LED delineators will be installed at bridge and tunnel entrances and exits (e.g., spacing ≤50m to enhance nighttime visibility and assist drivers in identifying changes in roadway shape). Intelligent "curve warning + speed limit reminder" signs will be deployed 200m before curves (e.g., automatically switching warning content based on weather monitoring data to dynamically adapt to current weather risks). During implementation, speed limit adjustments will require variable information boards to disseminate traffic guidance information. Facility renovations will prioritize construction during low-traffic hours between midnight and 6:00 AM on holidays.

[0050] S310.3: Implement mandatory speed limits, facility renovation, and emergency stationing strategies on high-risk roads; this strategy is applicable to scenarios with high passenger flow during extreme weather or holidays. As a further explanation of this step, high-risk road sections are suitable for extreme weather conditions (such as precipitation ≥ 5mm / h, visibility < 500m, road surface temperature ≤ 0℃) or high passenger flow scenarios during holidays (average daily traffic volume ≥ 120% of the designed capacity). Specifically: Based on the emergency braking safety distance, the mandatory speed limit is inferred: ;in, is the acceleration due to gravity (standard value of the gravitational field strength on the earth's surface), is the road friction coefficient (0.4 for wet roads and 0.2 for icy roads, reflecting the braking resistance under different road surface conditions), m represents the emergency braking distance specified in the "Highway Project Safety Assessment Specifications" (to ensure vehicles brake before an emergency occurs). Simultaneously, on long downhill sections, additional escape lanes (≤5 km apart) will be added within 1 km downstream of the crest in accordance with the "Highway Interchange Design Regulations" (JTG / TD21) to provide emergency shelter for out-of-control vehicles. On bridge sections, anti-throw nets with a mesh size of ≤50mm×50mm will be installed outside the guardrails to prevent falling objects from threatening traffic below or oncoming traffic. Furthermore, emergency stations will be deployed every ≤10 km within a "5-minute response circle" and equipped with simple rescue equipment such as anti-skid sand, emergency power supplies, tow trucks, and warning signs to ensure rapid response to emergencies. When implemented, mandatory speed limits must be registered with the traffic police (to ensure a basis for enforcement). Facility renovations must be included in the annual maintenance plan (to ensure funding and construction time). Emergency stations will be located within existing service areas or toll booths (to reduce construction costs).

[0051] S310.4: Temporary closures and long-term rerouting should be implemented in extremely high-risk sections. This strategy is applicable to areas with geological hazards or major linear defects.

[0052] As a further explanation of this step, extremely high-risk sections are suitable for geological disaster hazards or major linear defects (horizontal curve radius <250m and longitudinal slope ≥5%, exceeding the limit of the Highway Route Design Specifications, which may easily induce vehicle loss of control). Specifically: when the daily change in slope displacement is ≥5mm (which can be obtained through geological monitoring equipment to reflect the dynamic stability of the slope) or When the road is closed, a temporary closure is initiated, and traffic is directed through traffic control signs such as "Closed ahead, please detour" and intelligent guidance systems; if the annual accident rate of the road section is ≥ 2 times / year / km (based on historical accident statistics, quantifying the severity of the risk consequences), a temporary closure is immediately implemented. When long-term route changes are being discussed, a cost-effective approach is established. : ;in, is the average annual accident loss before the route change (including compensation for casualties, road facility repairs, traffic delays, etc., estimated through historical data statistics and industry quotas), The investment in the line change project (based on the estimated value of the feasibility study phase, including the construction costs of roadbed, bridges, tunnels, etc.); compare the full life cycle costs of "line change", "tunnel / bridge reconstruction" and "addition of risk avoidance facilities" (taking into account the expenditures of the construction, maintenance, and operation phases), and give priority to A plan should be developed to ensure the rerouting is economically feasible. During implementation, temporary closures must be coordinated by a joint command center established with traffic police and road administration (to ensure coordinated traffic control). The rerouting plan must be reviewed by provincial transportation authorities (to ensure it complies with regional road network planning and technical standards).

[0053] S400, Strategy Verification and Iteration: Verify the effectiveness of pre-emptive prevention and control strategies through traffic flow simulation, collect real-time traffic and meteorological data after the implementation of pre-emptive prevention and control strategies, dynamically adjust strategy parameters based on the collected real-time traffic and meteorological data, and build a closed-loop strategy iteration mechanism driven by risk status feedback.

[0054] In this step, in S400, the effectiveness of the pre-emptive prevention and control strategy is verified through traffic flow simulation, specifically including: Based on the road network topology data (including linear parameters) from S200, the traffic fluctuation characteristics (vehicle type ratio, peak flow) from S120.2, and the meteorological sensitive period data (fog / icing parameters) from S130.1, a digital twin model including lane-level geometric models was constructed using professional simulation software. Based on the S300's pre-emptive prevention and control strategy, extreme weather and holiday peak scenarios were simulated to verify the traffic stability, accident risk and traffic efficiency after the strategy was implemented.

[0055] To further illustrate this step, this embodiment uses professional simulation software such as VISSIM to construct a lane-level digital twin model based on S200 road network topology data (including linear parameters such as horizontal curve radius and longitudinal slope), S120.2 traffic fluctuation characteristics (vehicle type ratio and peak hour traffic), and S130.1 meteorological sensitive period data (fog occurrence conditions and freezing temperature thresholds). The model includes the following process: First, the roadbed width, lane divisions, and sign and marking positions are restored on a 1:1 basis; Then, based on historical data, behavioral parameters such as the truck ratio (e.g., 20% for mountain highways) and the vehicle-following reaction time (e.g., 1.5s) are set. Then, parameters such as the road section affected by fog (such as visibility <500m) and the friction coefficient of icy road surface (such as 0.2) are imported through the API.

[0056] Finally, the strategy was validated during peak holiday periods (e.g., traffic volume = 120% of the designed capacity) and extreme weather conditions (e.g., precipitation ≥ 5 mm / h and road surface temperature ≤ 0°C) using the following indicators: traffic stability (average speed standard deviation); accident risk (number of critical conflicts, i.e., the number of events where the following distance is less than the safe braking distance); and traffic efficiency (average delay time D on the road section).

[0057] It is worth noting that when constructing the lane-level digital twin model for mountainous highways, the following parameters were configured based on the Appendix of the Highway Project Safety Evaluation Specifications and the characteristics of mountainous scenarios: Truck following distance: Refer to the recommended values ​​in the regulations and dynamically adjust the longitudinal slope in mountainous areas. When going uphill (longitudinal slope > 5%), the distance is extended to ensure safety, and when going downhill (longitudinal slope < -3%), the distance is shortened to adapt braking efficiency. Road friction coefficient: Based on real-time weather dynamic matching, the normal value is used for dry roads, the coefficient is reduced for wet roads (precipitation intensity > 0.5mm / h), and it is further lowered for ice-covered roads (temperature < 0℃ and snowfall), accurately simulating driving resistance in different weather conditions.

[0058] By configuring the above parameters, we can reproduce real traffic conditions on mountain highways (such as following vehicles in foggy sections and braking on icy slopes), and thus verify the effectiveness of risk prevention and control strategies.

[0059] In this step, in S400, dynamically adjusting policy parameters and building a policy closed-loop iteration mechanism specifically include: Real-time data collection: Utilizing loop detectors, microwave detectors, and video analysis equipment, we collect real-time traffic flow data (volume, vehicle speed, and lane occupancy). We also obtain visibility, precipitation intensity, and road surface temperature from weather stations to create a minute-by-minute dynamic data set. Dynamic adaptation of strategy parameters: When real-time traffic exceeds the preset peak threshold, lane occupancy exceeds the congestion threshold, visibility falls below the meteorological sensitivity threshold, or road surface temperature falls below the freezing risk temperature, the corresponding level of prevention and control strategy is automatically triggered; Risk-strategy closed-loop iteration: Real-time data is input into the risk assessment model, the risk level of the road section is recalculated, and the triggering conditions and implementation parameters of the prevention and control strategy are adjusted according to the changes in the risk level to form a dynamic optimization mechanism.

[0060] As a further explanation of this step, in this embodiment, the flow threshold is set and meteorological thresholds (using the standards of S130.1), where To design the traffic capacity, calculate it according to the Highway Engineering Technical Standards; when the real-time data meets , or when the weather exceeds the limit, the following strategies are triggered according to the following logic: Strategy Level = ; in, is the real-time flow rate (collected by a loop coil detector), and These are the low-medium and medium-high risk traffic flow dividing points, which are set based on the correlation between traffic flow and accident rate on mountain highways.

[0061] The dynamic correction formula of speed limit is: ;in, is the adjusted speed limit value; As the basic speed limit; The adjustment factor is set based on engineering practice and controls the speed limit reduction range; Real-time lane occupancy can be collected by detectors to reflect the degree of road congestion; is the free flow occupancy rate, which is a typical value when the road is clear; It is the critical occupancy rate for congestion. Traffic congestion is likely to occur if this value is exceeded.

[0062] It should be added that, in this embodiment, the risk-strategy closed-loop iteration realizes the dynamic upgrade of the prevention and control strategy through the cycle of "real-time data feedback - risk level recalculation - strategy parameter optimization". The specific process is as follows: First, start the iteration trigger mechanism: under normal circumstances, it is automatically triggered every hour, and the iteration is started based on the latest 12 hours of real-time data (traffic flow, weather, geology); in an emergency, when the real-time accident rate ,in is the actual accident rate in the past hour, The accident rate is immediately triggered when the risk is higher than the historical average accident rate for the same period, ensuring a rapid response when the risk increases abnormally; Subsequently, the risk level is recalculated: real-time traffic flow data (volume, speed, lane occupancy, etc.), meteorological data (visibility, precipitation intensity, road surface temperature), and geological monitoring data (slope displacement, stability coefficient) are input into the S200 risk assessment model to recalculate the road section risk level (low / medium / high / very high), and the dominant factors leading to the level change (such as excessive precipitation or sudden increase in traffic) are identified; Then, implement strategic parameter adjustments: Targeted adjustments are made based on the recalculated risk level. If the risk increases (e.g., from medium to high), the speed limit is lowered (e.g., from 80 km / h to 60 km / h) and the emergency station response time is shortened (e.g., from 5 minutes to 3 minutes). If the risk decreases (e.g., from high to medium), the reverse adjustments are made. Safety is also verified using the braking distance formula: ,in For safe braking distance, is the adjusted speed limit value, 254 is the unit conversion coefficient, is the road friction coefficient), Actual safety distance of the road section (e.g. 200m for a long downhill section); Next, simulation verification was carried out: the digital twin model (including road network topology, traffic flow, and meteorological parameters) was called to verify the adjusted indicators, including: traffic flow stability (such as the average speed standard deviation decreased by ≥20% compared with the pre-adjustment level), accident risk (such as the number of critical conflicts decreased by ≥30%), and traffic efficiency (such as the average delay). minutes) to ensure the strategy is effective and feasible; Finally, the strategy library is updated and the loop is closed: If verification passes, the adjusted parameters are written to the strategy library, replacing the old ones. If not, the algorithm returns to the parameter adjustment step and optimizes again until the requirements are met. The iteration data (parameters before and after adjustment, verification results) is simultaneously stored in the cloud (retained for ≥5 years) to provide a basis for the next iteration, forming a complete closed loop.

[0063] Those skilled in the art will appreciate that the process of implementing all or part of the steps of the above embodiments may be accomplished by hardware, or by instructing related hardware through a program.

[0064] The above shows and describes the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The above embodiments and descriptions are merely preferred examples of the present invention and are not intended to limit the present invention. Various changes and improvements may be made to the present invention without departing from the spirit and scope of the present invention. Such changes and improvements fall within the scope of the present invention. The scope of protection claimed in the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for generating pre-emptive prevention and control strategies for traffic safety risks in mountainous expressway networks, characterized by: The steps include: S100, Risk Diagnosis and Data Preparation: Collect basic road information, historical traffic flow data, meteorological data, and accident-prone location data for mountain expressway networks. Combined with on-site surveys, clarify the road network topology and identify road alignment defects, holiday traffic characteristics, and core traffic safety risk factors in meteorologically sensitive sections. S200, Risk Assessment Model Construction: Select road alignment, traffic flow, meteorological and environmental resilience indicators, use the improved analytic hierarchy process to determine the indicator weights, construct a risk assessment model using the fuzzy comprehensive evaluation method, and calculate the traffic safety risk level of each road section; S300, Prevention and Control Strategy Generation: Based on the differences in traffic safety risk levels across road sections, formulate pre-emptive prevention and control strategies covering speed limit management, facility optimization, and emergency response layout, and clarify the applicable scenarios and implementation constraints of each pre-emptive prevention and control strategy; S400, Strategy Verification and Iteration: Verify the effectiveness of pre-emptive prevention and control strategies through traffic flow simulation, collect real-time traffic and meteorological data after the implementation of pre-emptive prevention and control strategies, dynamically adjust strategy parameters based on the collected real-time traffic and meteorological data, and build a closed-loop strategy iteration mechanism driven by risk status feedback.

2. The method for generating a pre-emptive prevention and control strategy for traffic safety risks in a mountainous expressway network according to claim 1 is characterized in that: In S100, basic road information, historical traffic flow data, meteorological data, and accident-prone point data of the mountain expressway network are collected, and the road network topology is determined by combining on-site investigation, including the following steps: S110.

1. Use a GIS system to collect road parameters, which shall include at least horizontal curve radius, longitudinal slope gradient, bridge span, tunnel clearance dimensions, and roadbed width; S110.

2. Use a combination of loop detectors, microwave detectors, and video analysis equipment to collect traffic flow data, recording holiday OD traffic volume, vehicle type composition, road section saturation, and vehicle speed percentiles; S110.

3. Obtain precipitation intensity, visibility, road surface temperature, and icing warning index through weather stations; S110.

4. Construct a road network topology model based on complex network theory. Nodes include toll stations, service areas, bridges and tunnels. Lines are associated with attribute parameters such as horizontal curve radius and longitudinal slope. Calculate node betweenness centrality and line connectivity efficiency.

3. The method for generating a pre-emptive prevention and control strategy for traffic safety risks in a mountainous expressway network according to claim 2, characterized in that: In S100, identifying the core threat factors of traffic safety risks in holiday traffic characteristics includes the following steps: S120.

1. Set the holiday range to statutory holidays and peak tourist seasons, and extract the sub-data for the corresponding period of the historical traffic flow data collected in S100; S120.

2. Extract traffic flow fluctuation characteristics, vehicle type mix, and continuous following distance distribution from traffic flow data collected by S100; S120.

3. Construct a generalized ordered logit model with accident severity as the dependent variable and vehicle type mix and the probability of exceeding the continuous following distance limit as independent variables to analyze the factors contributing to traffic safety risks unique to holidays and their weight ranking.

4. The method for generating a pre-emptive prevention and control strategy for traffic safety risks in a mountainous expressway network according to claim 3 is characterized in that: In S100, identifying the core threat factors of traffic safety risks in weather-sensitive areas includes the following steps: S130.

1. Based on the meteorological data and accident-prone location data collected by S100, screen meteorologically sensitive periods when precipitation intensity reaches a preset sensitivity threshold, visibility is below a preset critical value, and road surface temperature is below a preset icing risk temperature; S130.

2. Construct a spatiotemporal correlation analysis model, overlay the meteorologically sensitive periods with the road network topology, and identify weather-road coupling risk areas; S130.

3. Use the Bayesian network model, with meteorological parameters and road parameters as input and accident probability as output, to quantify the risk level of meteorologically sensitive sections.

5. The method for generating a pre-emptive prevention and control strategy for traffic safety risks in a mountainous expressway network according to claim 4, characterized in that: In S200, determining the indicator weights using the improved analytic hierarchy process includes the following steps: S210.

1. Constructing index order relationship based on G1 method: Building a set of indicators ,in, is the total number of indicators; the order of importance of indicators is determined through expert consultation ,in Indicates the important indicators; Adjacent sorting index and , define the importance proportional coefficient ,in, Indicator The weight of , used to balance the importance differences of adjacent indicators; By recursive formula , calculate the indicator weight vector ; S210.

2. Calculate objective weights based on the entropy weight method: right evaluation object Indicator data matrix ;in, Indicates the The evaluation object is in The original data of each indicator; Calculate the Normalized value of an indicator ; Calculate the The entropy value of the indicator ; Defining Metrics The coefficient of variation , normalize the difference coefficient and get the objective weight vector ; S210.

3. Fusion of subjective and objective weights: Define weight fusion coefficient , calculate the comprehensive weight by linear weighting formula : ； in, ; It is used to reflect the relative importance of subjective judgment in the comprehensive weight, and to determine the optimal value through the goodness of fit between historical accident data and model prediction results.

6. The method for generating a pre-emptive prevention and control strategy for traffic safety risks in a mountainous expressway network according to claim 5, characterized in that: The environmental resilience indicators include network topology resilience, transportation service resilience, and environmental adaptability resilience, among which: The network topology resilience is based on the road network topology model constructed by S110.4 to calculate the node betweenness centrality , line connectivity efficiency ; The traffic service resilience is based on the historical traffic flow data collected in step S100 to calculate the peak period traffic capacity redundancy. , the time it takes to restore the road network after the accident; The environmental adaptability resilience is based on the meteorologically sensitive sections identified in S130.1, and the proportion of meteorologically sensitive sections is calculated. , slope stability coefficient .

7. The method for generating a pre-emptive prevention and control strategy for traffic safety risks in a mountainous expressway network according to claim 6, characterized in that: In S200, constructing a risk assessment model using the fuzzy comprehensive evaluation method includes the following steps: S220.

1. Construct an intuitionistic fuzzy evaluation matrix: In view of the uncertainty of mountainous meteorology and the mixed characteristics of traffic flow, the risk level is divided into four levels: low, medium, high and very high. The relationship between the characterization indicator and the risk level, where: Indicates the degree of membership of the indicator to a certain risk level; Indicates the non-membership degree of the indicator to a certain risk level; Indicates hesitation, and , used to quantify the uncertainty in mountainous scenarios; S220.

2. Calculate the comprehensive evaluation vector: Based on the comprehensive weight determined in S210.3 , the intuitionistic fuzzy numbers of each indicator are fused through the weighted average operator, and the formula is: ； in, Indicates the The degree of membership of an indicator to a certain risk level; Indicates the The non-membership degree of an indicator to a certain risk level; Indicates the total number of indicators; represents the intuitionistic fuzzy comprehensive evaluation vector after weighted fusion; S220.

3. Determine the risk level: Using dual function sorting rules to analyze comprehensive evaluation vectors : Score function : Defined as , the larger the value, the higher the risk level; Exact function : Defined as , used to solve the sorting when the scores are equal.

8. The method for generating a pre-emptive prevention and control strategy for traffic safety risks in a mountainous expressway network according to claim 7, characterized in that: In S300, formulating a pre-emptive prevention and control strategy includes the following steps: S310.

1. Adopt a speed limit adjustment + facility reinforcement strategy on low-risk road sections; S310.

2. Adopt a speed limit adjustment + facility reinforcement strategy on medium-risk sections; S310.

3. Adopt mandatory speed limits, facility renovation, and emergency stationing strategies on high-risk sections of roads; S310.

4. A temporary closure plus long-term rerouting strategy will be adopted for extremely high-risk sections.

9. The method for generating a pre-emptive prevention and control strategy for traffic safety risks in a mountainous expressway network according to claim 8, characterized in that: In S400, the effectiveness of the pre-emptive prevention and control strategy is verified through traffic flow simulation, specifically including: Based on the road network topology data from S200, the traffic fluctuation characteristics from S120.2, and the meteorological sensitive period data from S130.1, a digital twin model including lane-level geometric models was constructed using professional simulation software. Based on the S300's pre-emptive prevention and control strategy, extreme weather and holiday peak scenarios were simulated to verify the traffic stability, accident risk and traffic efficiency after the strategy was implemented.

10. The method for generating a pre-emptive prevention and control strategy for traffic safety risks in a mountainous expressway network according to claim 9, characterized in that: In S400, dynamically adjusting policy parameters and building a closed-loop policy iteration mechanism specifically include: Real-time data collection: Using loop detectors, microwave detectors, and video analysis equipment to collect real-time traffic flow data, and using weather stations to obtain visibility, precipitation intensity, and road surface temperature, to form a minute-by-minute dynamic data set; Dynamic adaptation of strategy parameters: When real-time traffic exceeds the preset peak threshold, lane occupancy exceeds the congestion threshold, visibility falls below the meteorological sensitivity threshold, or road surface temperature falls below the freezing risk temperature, the corresponding level of prevention and control strategy is automatically triggered; Risk-strategy closed-loop iteration: Real-time data is input into the risk assessment model, the risk level of the road section is recalculated, and the triggering conditions and implementation parameters of the prevention and control strategy are adjusted according to the changes in the risk level to form a dynamic optimization mechanism.

Images