Tunnel traffic volume and air demand prediction method based on BP neural network

Abstract

The present application relates to the technical field of tunnel ventilation control, and particularly relates to a tunnel traffic volume and air requirement volume prediction method based on BP neural network. The method comprises the following steps: according to tunnel structure parameters and traffic flow data, a multilayer sensor array of a tunnel vertical section is laid out; environment parameter collection along the vertical direction of the tunnel is carried out by using the multilayer sensor array, and traffic flow data is integrated to obtain environment-traffic coupling data; environment-traffic working condition feature extraction and correlation are carried out according to the environment-traffic coupling data to obtain a vertical feature layer division graph and a working condition-pollution response correlation matrix; pollution migration dynamics calculation under the condition of heat and humidity is carried out according to the vertical feature layer division graph and the working condition-pollution response correlation matrix to obtain a pollution migration dynamics parameter set. By constructing a tunnel vertical layered pollution load index, the present application predicts and optimizes air requirement volume distribution, and significantly improves the intelligent level and operation efficiency of the tunnel ventilation system.

Description

Technical Field

[0001] The present invention relates to the technical field of tunnel ventilation control, and in particular to a method for predicting tunnel traffic volume and required air volume based on a BP neural network. Background Art

[0002] Existing tunnel ventilation methods typically employ single-point or limited-point monitoring, ignoring the vertical distribution differences of pollutants caused by vertical temperature and humidity stratification within the tunnel. In particular, pollutant accumulation at the driver's breathing height (approximately 1.5 meters) cannot be accurately captured, resulting in a blind spot in air quality assessment and impacting driving safety. Traditional air volume calculation methods, often based on empirical formulas or simple threshold controls, fail to effectively account for the complex migration characteristics and accumulation patterns of pollutants under varying temperature and humidity conditions, and fail to fully utilize historical data for intelligent forecasting. This leads to irrational allocation of ventilation resources, a failure to guarantee air quality in critical areas, and energy waste. Existing tunnel management systems typically treat ventilation control and traffic management as independent systems, lacking effective coordination mechanisms. When tunnel air quality deteriorates, relying solely on increasing ventilation volume without considering traffic control measures increases energy consumption while failing to control pollution emissions at the source, resulting in overall system inefficiency.

[0003] In summary, existing technologies have problems such as insufficient understanding of the characteristics of the vertical stratified environment in tunnels, inaccurate prediction of required air volume, and lack of ventilation-traffic coordinated control mechanism, which need to be urgently addressed. Summary of the Invention

[0004] Based on this, it is necessary to provide a tunnel traffic volume and required air volume prediction method based on BP neural network to solve at least one of the above technical problems.

[0005] To achieve the above purpose, a tunnel traffic volume and required air volume prediction method based on BP neural network includes the following steps:

[0006] Step S1: Obtain and deploy a multi-layer sensor array in the vertical section of the tunnel based on tunnel structural parameters and traffic flow data; use the multi-layer sensor array to collect environmental parameters along the vertical direction of the tunnel from the road surface to the top, and integrate traffic flow data to obtain environmental-traffic coupled data;

[0007] Step S2: Extracting and correlating environment-traffic operating condition characteristics based on the environment-traffic coupling data to obtain a vertical feature layer partition map and an operating condition-pollution response correlation matrix; calculating pollutant migration dynamics under thermal and humid conditions based on the vertical feature layer partition map and the operating condition-pollution response correlation matrix to obtain a pollutant migration dynamics parameter set; extracting respiratory layer pollutant accumulation characteristics from the pollutant migration dynamics parameter set to obtain a vertically layered pollutant load index;

[0008] Step S3: Obtain historical ventilation operation records and simulate the stratified airflow distribution characteristics based on the vertical stratified pollutant load index to obtain the stratified ventilation efficiency coefficient; predict the required air volume based on the BP neural network and the vertical stratified pollutant load index to obtain the overall required air volume prediction value; optimize the stratified air volume distribution based on the overall required air volume prediction value, the stratified ventilation efficiency coefficient, and the vertical stratified pollutant load index to obtain the stratified air volume matrix;

[0009] Step S4: construct a three-dimensional evaluation system based on the layered air volume demand matrix; perform ventilation-traffic coordinated control based on the three-dimensional evaluation system to obtain a two-way control execution plan.

[0010] By densely deploying multiple layers of sensors (including DFOS and point sensors) vertically in key tunnel areas, particularly increasing the number of monitoring points near the driver's breathing height, this method enables refined and continuous acquisition of the vertical distribution of environmental parameters such as temperature, humidity, and pollutant concentrations within the tunnel. This overcomes the limitations of traditional single-point or limited-point monitoring, which cannot capture vertical stratification. Furthermore, by integrating traffic flow data and particulate matter characteristic data, a "multidimensional environment-traffic coupling dataset" containing multi-source heterogeneous information is constructed. This provides a comprehensive, high-quality data foundation for subsequent in-depth analysis of the coupling relationship between environment and traffic, the vertical migration patterns of pollutants, and precise stratified control, significantly improving the comprehensiveness and targeted nature of data collection. Through in-depth analysis of the environment-traffic coupling data, characteristic layers of the tunnel's vertical space are identified and quantified, as well as the response patterns of pollutant concentrations in each layer under different traffic conditions, revealing the correlation between traffic emissions and vertically stratified pollution distribution. More importantly, based on the calculation of pollutant migration dynamics under hot and humid conditions and the extraction of pollutant accumulation characteristics in the breathing layer, a "vertical stratification pollutant load index" is constructed that can quantitatively describe the concentration and changing trends of pollutants at different altitudes. This index comprehensively considers pollutant concentrations, health risks, migration characteristics, and environmental stability, providing scientific and refined input for subsequent air volume demand prediction and allocation, particularly highlighting air quality conditions at the driver's breathing level. By analyzing historical ventilation performance and simulating airflow distribution under different ventilation schemes, a "stratified ventilation efficiency coefficient" was calculated and derived, quantifying the differences in the ability of a unit of air volume to remove pollutants at different altitudes (particularly at the breathing level). This provides an important basis for optimizing air volume allocation. Furthermore, using a BP neural network model, combined with the vertically stratified pollutant load index and real-time traffic environment data, an intelligent and dynamic prediction of the tunnel's overall air volume demand was achieved, improving both accuracy and foresight. Furthermore, based on the differences in pollutant load and ventilation efficiency at each level, the overall air volume demand was optimally allocated within technical constraints. A "highly sensitive air volume demand matrix" was generated, containing differentiated allocation recommendations for different altitudes. This overcomes the limitation of traditional methods that only predict overall air volume demand and provides a concrete and feasible solution for achieving precise stratified ventilation. By constructing a "three-dimensional evaluation system" encompassing energy consumption, air quality (emphasizing the breathing zone), and traffic efficiency, a quantitative standard is provided for evaluating the comprehensive performance of different control strategies, enabling a shift from single-objective optimization (such as solely satisfying air quality) to multi-objective collaborative optimization. A more crucial benefit lies in the generation and screening of "control strategy candidates" that balance various objectives based on this evaluation system and the collaborative control feasible domain. Furthermore, through a dynamic optimization algorithm, a "two-way control execution plan" is generated, encompassing both precise ventilation system control instructions and traffic management control recommendations.This achieves the linkage and coordination between the ventilation system and traffic management. It not only improves the efficiency of air quality improvement and energy utilization by optimizing ventilation strategies, but also reduces pollutant emissions from the source or optimizes tunnel traffic conditions through traffic control. This maximizes the overall benefits of tunnel operations while ensuring air quality at all height levels (especially the breathing layer) and improving driving safety. This is the key to achieving intelligent and refined control in this method.

[0011] Therefore, the present invention provides a method for predicting tunnel traffic volume and air volume demand based on a BP neural network. By constructing a multi-layer sensor array and a distributed fiber optic sensing system, accurate collection of tunnel vertical profile environmental data is achieved. A vertically layered pollutant load index is constructed based on the analysis of pollutant migration dynamics under thermal and humid conditions. BP neural network technology is used to accurately predict air volume demand and optimize its distribution. At the same time, a three-dimensional evaluation system of energy consumption, air quality, and traffic efficiency is established, and a two-way control execution plan for the ventilation system and traffic management is generated, thereby achieving accurate assessment and efficient control of air quality at different heights in the tunnel (especially the driver's breathing height), significantly improving the intelligence level and operating efficiency of the tunnel ventilation system. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] Figure 1 The figure is a flow chart of the steps of a method for predicting tunnel traffic volume and required air volume based on BP neural network;

[0013] Figure 2 Schematic diagram of the detailed implementation steps of step S4 in the present invention.

[0014] The purpose, features and advantages of the present invention will be further described with reference to the accompanying drawings and in conjunction with the embodiments. DETAILED DESCRIPTION

[0015] The following is a clear and complete description of the technical method of the present invention in conjunction with the accompanying drawings. It is obvious that the embodiments described are part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without making any creative efforts are within the scope of protection of the present invention.

[0016] In addition, the accompanying drawings are merely schematic illustrations of the present invention and are not necessarily drawn to scale. Identical reference numerals in the figures denote identical or similar parts, and thus repetitive descriptions thereof will be omitted. Some of the block diagrams shown in the accompanying drawings are functional entities that do not necessarily correspond to physically or logically separate entities. These functional entities may be implemented in software, in one or more hardware modules or integrated circuits, or in different network and / or processor and / or microcontroller approaches.

[0017] It should be understood that although the terms "first," "second," and the like may be used herein to describe various elements, these elements should not be limited by these terms. These terms are used solely to distinguish one element from another. For example, a first element may be referred to as a second element, and similarly, a second element may be referred to as a first element, without departing from the scope of the exemplary embodiments. The term "and / or" as used herein includes any and all combinations of one or more of the listed associated items.

[0018] In the embodiment of the present invention, reference Figure 1 FIG. 1 is a flow chart of the steps of a method for predicting tunnel traffic volume and required air volume based on a BP neural network according to the present invention. In this example, the method for predicting tunnel traffic volume and required air volume based on a BP neural network includes the following steps:

[0019] Step S1: Obtain and deploy a multi-layer sensor array in the vertical section of the tunnel based on tunnel structural parameters and traffic flow data; use the multi-layer sensor array to collect environmental parameters along the vertical direction of the tunnel from the road surface to the top, and integrate traffic flow data to obtain environmental-traffic coupled data;

[0020] In this embodiment of the present invention, key monitoring areas are identified based on tunnel structure and historical traffic data. Distributed fiber optic sensors (for continuous temperature and humidity profiling) and multilayer point sensors (for gas and particulate matter concentration and size distribution) are installed vertically within these areas (for example, at 0.5-meter intervals, particularly near a breathing height of 1.5 meters). This forms a multilayer sensor array for real-time vertical environmental data collection. Simultaneously, an integrated traffic monitoring system acquires traffic flow data such as vehicle volume, vehicle type, and speed, and estimates vehicle emission characteristics. The collected environmental data is then analyzed for particulate matter characteristics (particle size distribution and chemical composition). Finally, environmental and traffic data from different sources and at varying temporal and spatial resolutions are temporally aligned, spatially interpolated, and fused to form a "multidimensional environment-traffic coupled dataset" containing environmental parameters at each altitude, particulate matter characteristics, and corresponding traffic parameters.

[0021] Step S2: Extracting and correlating environment-traffic operating condition characteristics based on the environment-traffic coupling data to obtain a vertical feature layer partition map and an operating condition-pollution response correlation matrix; calculating pollutant migration dynamics under thermal and humid conditions based on the vertical feature layer partition map and the operating condition-pollution response correlation matrix to obtain a pollutant migration dynamics parameter set; extracting respiratory layer pollutant accumulation characteristics from the pollutant migration dynamics parameter set to obtain a vertically layered pollutant load index;

[0022] In this embodiment of the present invention, a multidimensional environment-traffic coupled dataset is used to cluster analyze vertical temperature and humidity profiles to divide the tunnel vertical space into vertical spatial layers, including ground, lower, breathing, upper, and top. The environmental statistics and temporal variation characteristics of each layer are extracted to form a "vertical feature layer division map." Traffic data is divided into typical traffic conditions based on flow rate, vehicle type, speed, etc., and the pollutant concentrations (CO, , , The response characteristics (peak value, arrival time, decay rate, etc.) of traffic parameters are analyzed, and the correlation and causal relationship between traffic parameters and pollutant concentrations are calculated to generate a "working condition-pollution response correlation matrix". Based on the vertical temperature and humidity gradient field and the basic characteristics of pollutants, the migration kinetic parameters of pollutants such as diffusion coefficient, sedimentation rate, and phase change rate under different temperatures and humidities are calculated. The pollutant data of the breathing layer (1.2 meters-1.8 meters) are analyzed in detail, its relative enrichment coefficient is calculated, and the accumulation characteristics (accumulation rate, duration) under different working conditions and environments are analyzed. Typical accumulation patterns are extracted and time series predictions are made. A multi-factor weighted model is constructed based on the pollutant concentration, health risks, migration characteristics (retention or removal difficulty) and environmental stability of each layer. The "vertical stratified pollutant load index" of each vertical characteristic layer is calculated to quantify the concentration degree and change trend of pollutants in each layer.

[0023] Step S3: Obtain historical ventilation operation records and simulate the stratified airflow distribution characteristics based on the vertical stratified pollutant load index to obtain the stratified ventilation efficiency coefficient; predict the required air volume based on the BP neural network and the vertical stratified pollutant load index to obtain the overall required air volume prediction value; optimize the stratified air volume distribution based on the overall required air volume prediction value, the stratified ventilation efficiency coefficient, and the vertical stratified pollutant load index to obtain the stratified air volume matrix;

[0024] In this embodiment of the present invention, historical ventilation operation records (fan status, air volume) and changes in the vertically layered pollutant load index over corresponding periods are analyzed to evaluate the effectiveness of different ventilation schemes in removing pollutants from each layer, generating "stratified ventilation effectiveness evaluation data." Based on tunnel structural parameters and ventilation equipment characteristics, a simplified tunnel geometry model is constructed. Different ventilation scheme scenarios are set up, and computational fluid dynamics (CFD) methods are used to simulate the airflow distribution within the tunnel. Vertical velocity profiles, turbulence characteristics, and dead / high-speed zone distributions are obtained, creating a "vertically layered airflow characteristic map." Combining airflow characteristics with historical ventilation performance, the dilution capacity of unit air volume for pollutants in each layer is calculated, generating a "stratified ventilation efficiency coefficient table." Using the vertically layered pollutant load index (particularly the breathing zone), real-time traffic flow, and environmental parameters as input, a three-layer BP neural network model is constructed and trained to predict the "overall air volume forecast" required to maintain air quality standards and perform short-term trend forecasts. Based on the predicted overall air volume demand, the current load index for each floor, the target load index, and the stratified ventilation efficiency coefficient, an optimization model is established to determine the optimal air volume distribution plan, taking into account the technical constraints of the ventilation equipment. This plan is then converted into specific ventilation equipment control parameter settings (fan speed, angle, etc.). The optimized stratified air volume demand, equipment parameter plan, and estimated air quality improvement are integrated to construct a "highly sensitive air volume demand matrix."

[0025] Step S4: constructing a three-dimensional evaluation system based on the layered air volume demand matrix; performing ventilation-traffic coordinated control based on the three-dimensional evaluation system to obtain a two-way control execution plan;

[0026] In this embodiment of the present invention, tunnel ventilation equipment parameters and traffic management system parameters are acquired to evaluate the ventilation system's maximum and minimum capacity, fan adjustment range, and control measures (speed limits, lane separation, vehicle type restrictions) that can be implemented by the traffic management system, along with their impact on traffic flow. This determines a "coordinated control feasible domain diagram." A "three-dimensional evaluation system" is constructed for energy consumption, air quality (focusing on the respiratory load index), and traffic efficiency (average speed and delay), with weights and scoring criteria assigned to each indicator. Within the coordinated control feasible domain, a series of combinations of ventilation control parameters and traffic control parameters are generated as candidate strategies. An estimation model is used to calculate the impact of each candidate strategy on energy consumption, air quality at each level, and traffic efficiency after implementation. The comprehensive benefit index of each strategy is then calculated based on the three-dimensional evaluation system. Several strategies with the highest comprehensive benefit index are selected as "control strategy candidate solutions." Based on real-time environmental and traffic data, a dynamic optimization algorithm is used to select or fine-tune the candidate solutions, generating specific ventilation system control instructions (fan setpoints) and traffic management control recommendations (speed limits, lane adjustments, etc.). This creates a "two-way control execution plan" that is sent to the corresponding control system for execution, achieving closed-loop optimization through continuous monitoring and feedback.

[0027] Preferably, step S1 is specifically:

[0028] Determine the key monitoring area based on tunnel structural parameters and traffic flow data, and arrange vertically layered monitoring points in the key monitoring area to obtain a multi-layer sensor array;

[0029] Use distributed fiber optic sensing technology and multi-layer sensor arrays to collect tunnel environmental data and obtain vertical profile monitoring data;

[0030] Extracting traffic flow characteristic parameters from traffic flow data;

[0031] Analyze the particle characteristics of different altitude layers on the vertical profile monitoring data to obtain particle characteristic data;

[0032] The particle characteristic data and traffic flow characteristic parameters are analyzed to correlate particle and traffic, and a layered particle characteristic spectrum is obtained;

[0033] The vertical profile monitoring data, traffic flow characteristic parameters and layered particulate matter characteristic spectra are temporally and spatially unified and fused to obtain environment-traffic coupling data.

[0034] In this embodiment of the present invention, key monitoring areas within the tunnel are identified based on analysis of known tunnel structural parameters (e.g., a 5-kilometer tunnel length, a single-hole, two-lane structure, a 0.5% longitudinal slope at the entrance, a flat slope in the middle, and a -1% longitudinal slope at the exit) and historical traffic flow data (e.g., peak hour traffic volume, a high proportion of heavy vehicles, and low nighttime traffic). Key monitoring areas are selected: the slope (e.g., the middle of the 0.5% uphill section at the entrance), the vicinity of the ventilation shaft (e.g., within 50 meters of the ventilation shaft), and the vicinity of the service area entrance and exit, for a total of three key monitoring areas. Within each key monitoring area, metal brackets are used to vertically install multiple layers of sensor mounting points extending upward from the road surface. Sensor fixtures are installed at heights of 0 meters (road surface), 0.5 meters, 1.0 meters, 1.4 meters, 1.5 meters, 1.6 meters, 2.0 meters, 2.5 meters, 3.0 meters, and 4.0 meters, for a total of 10 vertical heights, forming a vertically layered array of monitoring points, forming a multi-layer sensor array.

[0035] Distributed fiber optic sensing (DFOS) technology is used to continuously collect temperature and humidity. On the sensor bracket of each key monitoring area, a temperature-sensing optical fiber and a humidity-sensing optical fiber are wound or fixed in the vertical direction. Connect both ends of the optical fiber to the distributed fiber optic sensor demodulator. The demodulator sends laser pulses to the optical fiber, and calculates the temperature distribution along the optical fiber by analyzing the Raman scattering or Brillouin scattering signal intensity and frequency shift along the optical fiber; and calculates the humidity distribution along the optical fiber by analyzing the attenuation or phase shift change of the special coated optical fiber. The demodulator outputs the temperature and humidity profile data along the length of the optical fiber (i.e., the vertical direction) once a minute. At the same time, point sensors are installed at each installation point (10 height points) of the multi-layer sensor array: electrochemical gas sensors are installed at heights of 0.5 meters, 1.5 meters, and 2.5 meters to collect carbon monoxide (CO) and nitrogen oxides ( ) concentration; optical particulate matter sensors are installed at heights of 1.5 meters and 3.0 meters to collect and Mass concentration; an aerosol particle size spectrometer is installed at a height of 1.5 meters to collect the particle size distribution. All point sensors are connected to the data acquisition server via industrial Ethernet, and monitoring data with timestamps are collected and uploaded every minute to obtain data including temperature, humidity (DFOS), CO, 、 、 , original monitoring data of vertical profile of particulate matter particle size distribution.

[0036] Traffic flow data is collected from the tunnel traffic monitoring system, which consists of loop detectors at key sections of the tunnel (e.g., near each key monitoring area) and a video analysis system. The loop detectors output the total number of vehicles passing through each section and their average speed every minute. The video analysis system processes the video stream, using image recognition algorithms to identify the types of vehicles passing through, classifying them into four categories: small passenger cars, large passenger cars, light trucks, and heavy trucks. The system also counts the number of different vehicle types per minute. Simultaneously, the video analysis system estimates the exhaust height of each type of vehicle based on the vehicle identification results and a database of vehicle types (e.g., approximately 0.3 meters for small passenger cars and 0.8 meters for heavy trucks). This data is aggregated to generate a set of traffic flow characteristic parameters, including vehicle volume (vehicles / min), vehicle type composition (percentage of each vehicle type), average speed (km / h), and estimated exhaust heights for each type of vehicle.

[0037] The particle data in the original vertical profile monitoring data were analyzed. The particle size distribution data collected by the aerosol particle size spectrometer at a height of 1.5 meters were used to calculate the number concentration, mass concentration and key parameters of the particle size distribution at that height, such as the median diameter ( ) and the geometric standard deviation ( Assuming that the simple online chemical component analyzer can measure the black carbon (BC) and organic carbon (OC) content in particulate matter in real time, calculate the proportion of BC and OC. This analysis process is repeated, and combined with data from optical particle sensors at different heights, the particle size distribution parameters and the proportions of the main chemical components at different heights are estimated to obtain particle characteristic data.

[0038] Correlate and analyze the particle characteristic data with the traffic flow characteristic parameter set collected simultaneously. For example, calculate the Pearson correlation coefficient between the black carbon concentration at a height of 1.5 meters and the heavy truck traffic volume to evaluate the impact of truck traffic on black carbon. The relationship between mass concentration and average speed of different vehicle types was studied to study the effect of vehicle speed on road dust. The characteristic parameters of particulate matter (such as The quantified relationship model between the median diameter (PMD) and traffic parameters (such as the average speed of all vehicles) is constructed. Based on these correlation analysis results, a layered particle characteristic spectrum is generated, which describes the typical characteristics of the particle size distribution and chemical composition of particles at different altitudes under different traffic conditions.

[0039] In view of the differences in sampling frequency and spatial distribution of different data sources, spatiotemporal unification and data fusion are performed. DFOS data provides continuous vertical temperature / humidity distribution (spatial resolution of about 0.1 meters), point sensor data provides data at 10 fixed height points (spatially discontinuous), and the time resolution of traffic data and point sensor data is 1 minute. First, the DFOS data is interpolated to the same 10 height points as the point sensors in the vertical direction and aligned with the point sensor data for time stamps. Then, all data are unified to a time resolution of 1 minute. For intermediate heights (such as 0.2 meters, 0.7 meters, etc.) where no point sensors are set in space, the CO, 、 、 Concentration and particle characteristic parameters are used to construct a more continuous environmental parameter distribution curve in the vertical direction. Finally, the environmental parameters (temperature, humidity, CO, 、 、 , particulate matter particle size parameters, chemical composition ratio) and the corresponding traffic flow characteristic parameters (traffic volume, vehicle type composition, average speed) are integrated into a database table. Each row of records in the database table contains a timestamp, a vertical height point and all environmental and traffic-related parameter values ​​of the point at this timestamp, forming a multidimensional environment-traffic coupling data set. For example, the database table contains columns: timestamp, height (m), temperature (℃), humidity (%), CO (ppm), (ppm), (µg / m ), (µg / m ), _ (µm), _BC (%), total traffic volume (vehicles / min), proportion of heavy trucks (%), average speed (km / h), etc.

[0040] Preferably, the extraction and association of environment-traffic condition features in step S2 are specifically as follows:

[0041] The vertical space of the tunnel is divided into layers according to the environment-traffic coupling data to obtain vertical space layers, which include the ground layer, lower layer, breathing layer, upper layer and top layer.

[0042] Extract environmental features from the vertical spatial layer to obtain a vertical feature layer division map;

[0043] Traffic data is classified into different traffic condition types. According to the vertical characteristic layer division map, the response characteristics of pollutant concentrations in each vertical characteristic layer are analyzed for different traffic condition types to obtain the condition-pollution response correlation matrix.

[0044] In this embodiment of the present invention, long-term vertical temperature and humidity profile data are extracted from a multidimensional environment-traffic coupled dataset (containing temperature, humidity, pollutant concentrations, and corresponding traffic parameters at different times and altitudes). These profile data are analyzed using the K-means clustering algorithm. For example, assuming the number of clusters K = 5, the algorithm aims to group data points with similar vertical temperature and humidity profile characteristics into a single cluster. The algorithm iteratively calculates the centroid of each cluster (i.e., the average temperature and humidity distribution of all profiles within that cluster) and assigns the vertical profile at each time point to the cluster with the closest centroid. By analyzing the temperature and humidity distribution characteristics of each cluster in the clustering results, several typical vertical temperature and humidity stratification patterns can be identified. For example, one cluster represents a stratification pattern characterized by "warm and humid lower part and cool and dry upper part," while another cluster represents a pattern characterized by "uniform temperature and humidity along the vertical direction." Based on these typical stratification patterns, the vertical space of the tunnel is divided into logical vertical spatial layers. For example, the area between 0 and 0.5 meters is divided into the ground layer, the area between 0.5 and 1.2 meters is divided into the lower layer, the area between 1.2 and 1.8 meters is divided into the breathing layer, the area between 1.8 and 3 meters is divided into the upper layer, and the area above 3 meters is divided into the top layer. These division boundaries can be fine-tuned based on the heights where the temperature and humidity gradients change significantly in the clustering results. This results in a vertical spatial layer division scheme that includes these predefined vertical spatial layers.

[0045] Environmental features are extracted for each vertical spatial layer to generate a vertical feature layer partitioning map. For each vertical spatial layer (e.g., ground layer, lower layer, etc.), statistical features such as average temperature, average humidity, temperature coefficient of variation, and humidity coefficient of variation are calculated for all monitoring points within that layer during the time period. Temporal variations in environmental parameters within that layer are also analyzed, such as the daily average, daily maximum, nighttime minimum, and weekly variation. These statistical and temporal variation features are used as the environmental feature vector for that vertical spatial layer. The vertical feature layer partitioning map records the spatial range (e.g., altitude range) of each vertical spatial layer and the environmental feature vector associated with that layer. For example, for the breathing layer (1.2-1.8 meters), the map is structured as follows: [Average temperature: 25°C, Average humidity: 60%, Temperature coefficient of variation: 0.1, Humidity coefficient of variation: 0.15, Diurnal variation (temperature): 5°C, ...].

[0046] The traffic data in the multidimensional environment-traffic coupling dataset are classified according to different traffic conditions. Based on the three main indicators of total vehicle flow, proportion of heavy trucks and average speed, typical conditions are defined. For example: Condition A: low traffic (<300 vehicles / h), high speed (>80km / h), low proportion of heavy trucks (<10%); Condition B: peak traffic (>1500 vehicles / h), low speed (<40km / h), high proportion of heavy trucks (>30%); Condition C: flat peak traffic (600-1000 vehicles / h), medium speed (50-70km / h), medium proportion of heavy trucks (10%-30%). The map is divided according to the vertical feature layer, and for each traffic condition type, the response characteristics of the pollutant concentration in each vertical feature layer are analyzed. For example, for Condition B (peak, low speed, high trucks), the CO and Concentration changes. Calculate the average value, maximum value, time required to reach the peak value, and time required for the pollutant concentration to decay to a certain level after the operation is completed during the duration of the working condition. Use the Pearson correlation coefficient to calculate the relationship between different traffic parameters (such as the number of heavy trucks) and the pollutant concentration in each vertical characteristic layer (such as the breathing layer). The Granger causality test is used to analyze whether traffic parameters are the Granger cause of changes in pollutant concentrations. These analysis results are integrated into a working condition-pollution response correlation matrix. The rows of this matrix represent different traffic conditions, and the columns represent different vertical characteristic layers and pollutant types. The elements in the matrix quantify the response characteristics of specific pollutants in specific vertical layers under specific traffic conditions (such as average concentration, peak value, correlation coefficient, causal relationship strength, etc.). For example, one item in the correlation matrix represents: Working condition B-Respiratory layer -[Average concentration: 100µg / m , Peak value: 150µg / m , time to reach peak: 15min, correlation coefficient with the number of heavy trucks: 0.85].

[0047] Preferably, the pollutant migration kinetics calculation under hot and humid conditions in step S2 is specifically as follows:

[0048] Construct vertical temperature and humidity gradient field based on vertical characteristic layer division atlas;

[0049] Determine pollutant-based characteristics of the operating condition-pollution response correlation matrix;

[0050] The temperature and humidity correction coefficients are calculated based on the vertical temperature and humidity gradient field and the basic characteristics of pollutants to obtain the dynamic correction coefficient matrix;

[0051] The vertical diffusion coefficient matrix is ​​calculated based on the kinetic correction coefficient matrix;

[0052] The sedimentation rate and phase change rate of the particles are calculated based on the vertical diffusion coefficient matrix to obtain the phase change sedimentation rate table;

[0053] The pollutant vertical flux equations are constructed based on the phase change sedimentation rate table and the vertical diffusion coefficient matrix;

[0054] The pollutant vertical flux equations are numerically solved and the migration characteristics are extracted to obtain the pollutant migration kinetic parameter set.

[0055] In this embodiment of the present invention, a vertical temperature and humidity gradient field is constructed based on a vertical feature layer partitioning map (which records the height range and environmental feature vectors of each vertical spatial layer) and real-time temperature and humidity data from a multidimensional environment-traffic coupled dataset. For example, at a given moment, the dataset provides temperature and humidity values ​​at altitudes of 0, 0.5, 1.0, ..., and 4.0 meters. By vertically interpolating these discrete temperature and humidity values ​​(e.g., using cubic spline interpolation), a continuous vertical temperature profile curve T(z) and vertical humidity profile curve H(z) are obtained, where z represents the vertical height. The temperature gradient dT / dz and humidity gradient dH / dz are calculated at different altitudes. For example, in the breathing layer (1.2-1.8 meters), the average temperature gradient and average humidity gradient within this layer are calculated. The vertical temperature and humidity gradient field describes the rate and direction of temperature and humidity change with altitude at different altitudes within the tunnel.

[0056] Determine the basic characteristics of each pollutant involved in the operation-pollution response matrix. For example, consider CO, 、 (Main components are black carbon and organic carbon), (including road dust, etc.) and other pollutants. Determine the molecular weight, gas diffusion coefficient (at standard temperature and pressure), particle density, particle size distribution characteristics (e.g. The median diameter of ), and temperature and humidity related properties, such as Henry's law constant of water-soluble gases, hygroscopicity of particles (hygroscopic growth factor), and volatility. These basic properties can be obtained by consulting chemical handbooks, physical property databases, or through laboratory measurements. For example, the molecular weight of CO is 28 g / mol. The density of medium black carbon is about 1.8g / cm .

[0057] Based on the constructed vertical temperature and humidity gradient field and the determined basic characteristics of the pollutants, the temperature and humidity correction coefficients are calculated to obtain the dynamic correction coefficient matrix. The diffusion coefficient of the gas changes with temperature and pressure. The diffusion, sedimentation and phase change rates of the particles are affected by temperature, humidity, air pressure and the characteristics of the particles themselves. For example, the molecular diffusion coefficient D0 of the gas can be corrected by the Stokes-Einstein equation, considering the influence of temperature T and air pressure P: D = ×(T / T0) ×(P0 / P), where T0 and P0 are standard temperature and pressure, and D0 is the diffusion coefficient under standard conditions. For particles, calculate their Brownian diffusion coefficient DB (d p ,T) with particle size d p and temperature T. Calculate the hygroscopic growth factor GF(RH) of the particles, which describes the increase in particle size at relative humidity RH. For example, using the κ-Köhler theoretical model, calculate GF=(1-κ×ln(RH))^(-1 / 3)×( / ), where κ is the hygroscopicity parameter, is the dry particle size, and ^(-1 / 3) represents taking the cube root of the base (1-κ×ln(RH)) and then taking the reciprocal. The kinetic correction coefficient matrix contains correction factors or directly corrected values ​​for the diffusion coefficients, sedimentation rates, and phase change rates of various pollutants at different vertical heights and different temperature and humidity conditions. For example, the matrix contains the following: At height z = 1.5m, temperature = 26℃, and humidity = 70%, the CO diffusion coefficient correction factor is 1.1. The hygroscopic growth factor is 1.3.

[0058] The vertical diffusion coefficient matrix is ​​calculated based on the dynamic correction coefficient matrix. The vertical diffusion coefficient includes not only molecular diffusion and Brownian diffusion, but also turbulent diffusion in the tunnel. Turbulent diffusion coefficient K t (z) varies with the height z and the turbulence intensity of the airflow in the tunnel. The turbulence intensity can be estimated by simulating the airflow distribution characteristics (the content of step 3, here as input) or by empirical formulas. The total vertical diffusion coefficient D v (z) is the molecular diffusion coefficient D(z) (for gases) or the Brownian diffusion coefficient DB(z) (for particles) and the turbulent diffusion coefficient K t Superposition of (z): D v (z)=D(z)+K t (z) or D v (z)=DB(z)+K t (z). The vertical diffusion coefficient matrix records the vertical total diffusion coefficient of different pollutant types (taking into account their particle size and phase) at different vertical heights. For example, the matrix contains: height z = 1.5m, pollutant = CO, Dv =0.1m / s; height z = 1.5m, pollutant = (d p =0.5µm),D v =0.05m / s.

[0059] According to the vertical diffusion coefficient matrix and the determined basic characteristics of pollutants, the sedimentation rate and phase change rate of the particles are calculated to obtain the phase change sedimentation rate table. p ) mainly depends on its particle size d p and density ρ p :V_g=(ρ p ×g×d p 2 ) / (18×μ), where g is the acceleration due to gravity and μ is the air dynamic viscosity. The phase change rate mainly refers to the hygroscopic growth / evaporation rate of particulate matter or the condensation / evaporation rate of volatile organic compounds, which depends on humidity, temperature and the chemical composition of the particulate matter. For example, for hygroscopic particulate matter, when the relative humidity is higher than its critical relative humidity, hygroscopic growth will occur, and its particle size will increase, resulting in an increase in sedimentation rate; when the relative humidity decreases, the particles will evaporate water and the particle size will decrease. The phase change sedimentation rate table records the effective sedimentation rate (particle size after taking into account hygroscopic growth) and phase change rate of particles of different sizes at different vertical heights and different temperature and humidity conditions. For example, the table contains: height z=1.5m, temperature=26℃, humidity=70%, particle size= (Drying p =0.5µm), after moisture absorption growth d p =0.65µm, V_g=0.0001m / s, phase change rate=0 (assuming constant humidity).

[0060] Based on the constructed phase change sedimentation rate table and vertical diffusion coefficient matrix, the pollutant vertical flux equations are constructed. The vertical migration of pollutants is mainly affected by diffusion (including turbulence and molecular / Brownian diffusion) and sedimentation (for particulate matter). For gaseous pollutants (such as CO, ), whose vertical flux J_gas(z) is mainly driven by diffusion: J_gas(z)=-D v _gas(z)×dC_gas / dz, where C_gas is the gas concentration. For particulate matter, its vertical flux J_part(z) includes diffusion flux and deposition flux: J_part(z)=-D v_part(z)×dC_part / dz-V_eff_part(z)×C_part, where C_part is the particle concentration and V_eff_part is the effective deposition velocity taking into account phase change. The pollutant vertical flux equations describe the relationship between the vertical flux of different pollutants at different vertical heights and the pollutant concentration gradient at that height, the vertical diffusion coefficient, and the effective deposition velocity.

[0061] The pollutant vertical flux equations are numerically solved and the migration characteristics are extracted to obtain the pollutant migration dynamics parameter set. The pollutant vertical flux equations are combined with the mass conservation equation to form a partial differential equation that describes the vertical distribution of pollutants over time. , where S(z,t) is the source term at that height (such as vehicle emissions). The partial differential equation is usually discretized using numerical methods such as the finite difference method or the finite volume method, and numerically solved under given boundary conditions (such as source terms at the road surface, top boundary conditions) and initial conditions to obtain the distribution of pollutant concentration C(z,t) over time and height. The migration characteristics of pollutants at different vertical heights are extracted from the solution results, such as: the average residence time of pollutants in the breathing layer, the time required for pollutants to diffuse upward from the ground layer to the breathing layer, the effective deposition rate of particulate matter in the breathing layer, the degree of influence of humidity changes on the vertical distribution profile of specific pollutants, etc. These extracted parameters constitute the pollutant migration kinetics parameter set, which quantifies the vertical migration behavior of different pollutants under specific environmental conditions. For example, the parameter set includes: the average diffusion rate of CO in the breathing layer is Xm / s, The effective deposition rate in the respiratory layer is Ym / s, and a 10% increase in humidity leads to The concentration in the breathing layer increases by Z%, etc.

[0062] Preferably, the extraction of respiratory layer pollutant accumulation characteristics in step S2 is specifically as follows:

[0063] Based on the pollutant migration dynamics parameter set, the breathing layer boundary is located and data is extracted to obtain the breathing layer data subset;

[0064] Calculate the relative enrichment coefficient of the respiratory layer data subset to obtain the pollutant enrichment coefficient table;

[0065] Perform working condition correlation cumulative characteristic analysis on the pollutant enrichment coefficient table and the working condition-pollution response correlation matrix to obtain the working condition cumulative characteristic table;

[0066] Quantify the impact of environmental factors based on the cumulative characteristic table of working conditions and obtain the environmental impact coefficient matrix;

[0067] Based on the environmental impact coefficient matrix and the working condition cumulative feature table, the typical cumulative pattern of the breathing layer data subset is extracted to obtain the cumulative pattern characteristics;

[0068] Perform time series analysis on the cumulative pattern characteristics to obtain time series characteristic parameters;

[0069] According to the time series characteristic parameters, a short-term cumulative trend forecast is conducted to obtain a short-term cumulative forecast result;

[0070] Integrate the short-term cumulative prediction results and cumulative pattern characteristics to obtain the respiratory layer pollution accumulation feature set;

[0071] The vertically stratified pollutant load index is constructed based on the respiratory layer pollution accumulation characteristic set, the pollutant migration dynamics parameter set and the vertical characteristic layer division map to obtain the vertically stratified pollutant load index.

[0072] In this embodiment of the present invention, based on the breathing layer height range (e.g., 1.2 to 1.8 meters) determined in the vertical characteristic layer partitioning map, all environmental, traffic, and migration dynamics parameters within that height range (e.g., data at heights of 1.2, 1.4, 1.5, 1.6, and 1.8 meters) are extracted from a multidimensional environment-traffic coupled dataset and a pollutant migration dynamics parameter set (including pollutant concentrations, vertical diffusion coefficients, and effective deposition velocities at different times and heights). This creates a breathing layer data subset. This subset specifically contains detailed environmental and migration information for the breathing layer at different time points.

[0073] The relative enrichment coefficients of the respiratory layer data subset are calculated to obtain the pollutant enrichment coefficient table. ), calculate the ratio of its concentration in the breathing layer (for example, 1.5 meters height) to the average concentration of its upper and lower adjacent layers (for example, 1.0 meters and 2.0 meters height). For example, If the pollutant concentration in the breathing layer is significantly higher than that in adjacent layers, the coefficient is greater than 1, indicating enrichment. The ratio of the pollutant concentration in the breathing layer to the average pollutant concentration in the tunnel (averaged across all heights) is also calculated. The Pollutant Enrichment Coefficient Table records the relative enrichment coefficients of different pollutants in the breathing layer at different points in time, quantifying the relative concentration of pollutants in the breathing layer.

[0074] Based on the pollutant enrichment coefficient table and the operating condition-pollution response correlation matrix, the operating condition correlation cumulative characteristic analysis is performed to obtain the operating condition cumulative characteristic table. The data in the pollutant enrichment coefficient table is correlated with the corresponding traffic conditions (such as the peak low speed high truck operating condition). The characteristics of the enrichment coefficient of the respiratory layer pollutants under different traffic conditions are analyzed. For example, under the peak low speed high truck operating condition, Whether the CO enrichment factor remains consistently above 1.2, and whether the CO enrichment factor increases rapidly within a specific time period. Calculate indicators such as the average enrichment factor, maximum enrichment factor, and duration of high enrichment (e.g., enrichment factor > 1.3) of major pollutants in the breathing layer during each typical traffic condition. The Condition Accumulation Characteristics table summarizes typical characteristics of pollutant accumulation in the breathing layer (manifested by high enrichment factors) under different traffic conditions.

[0075] Based on the operating condition accumulation characteristic table and the environmental parameters (temperature, humidity) in the breathing layer data subset, the environmental factors are quantified to obtain the environmental impact coefficient matrix. The impact of different temperature and humidity conditions on the accumulation characteristics of breathing layer pollutants (such as enrichment coefficient and accumulation speed) under the same traffic operating condition is analyzed. For example, under high temperature and high humidity conditions, The hygroscopic growth of pollutants in the respiratory layer leads to a decrease in their effective deposition rate, thereby intensifying their accumulation, which is manifested as a higher enrichment coefficient and a faster accumulation rate. Through regression analysis or variance analysis, the degree of influence of temperature and humidity on the enrichment coefficient and accumulation rate of pollutants in the respiratory layer is quantified, and the corresponding environmental impact coefficient is calculated. The environmental impact coefficient matrix records the quantitative impact of environmental factors such as temperature and humidity on the accumulation characteristics of different pollutants in the respiratory layer. For example, the matrix contains: for every 1°C increase in temperature, The enrichment coefficient increases by 0.02; for every 1% increase in humidity, the CO accumulation rate decreases by 0.5%.

[0076] Based on the environmental impact coefficient matrix and the operating condition accumulation feature table, typical accumulation patterns are extracted from the breathing layer data subset to obtain accumulation pattern characteristics. Combined with historical data, typical accumulation patterns of breathing layer pollutant concentrations changing over time are identified. For example, the "rapid rise-platform-slow decline" pattern (corresponding to relief after peak congestion), the "continuous slow rise" pattern (corresponding to long-term medium traffic), etc. These patterns are jointly affected by traffic conditions and environmental factors. Clustering algorithms (such as spectral clustering) are used to analyze the time series data of breathing layer pollutant concentrations, and time periods with similar change trends are grouped together to extract several representative accumulation patterns. The accumulation pattern characteristics describe the key parameters of each typical accumulation pattern, such as shape, duration, peak intensity, rise / fall rate, etc., and are associated with the typical traffic conditions and environmental conditions that produce the pattern.

[0077] Time series analysis is performed on the accumulation pattern characteristics to obtain time series characteristic parameters. For each typical accumulation pattern extracted, time series analysis is performed. For example, time series characteristic parameters such as the pattern's periodicity (e.g., daily or weekly), trend, and autocorrelation are calculated. Using time series analysis methods such as ARIMA or state-space models, historical accumulation pattern data is modeled to extract parameters describing the pattern's dynamics. These time series characteristic parameters capture the temporal evolution of the accumulation pattern of pollutants in the respiratory layer.

[0078] According to the time series characteristic parameters, a short-term cumulative trend forecast is made to obtain the short-term cumulative forecast results. The time series model (established based on the time series characteristic parameters) is used to extrapolate the current respiratory layer pollutant concentration data. For example, based on the current concentration value and the time series characteristics of the identified typical accumulation pattern, the main respiratory layer pollutants (CO, The short-term cumulative forecast result is a quantitative estimate of the change in the concentration of pollutants in the respiratory layer in the very short term in the future.

[0079] The short-term accumulation prediction results and accumulation pattern characteristics are integrated to obtain the respiratory layer pollution accumulation feature set. The short-term accumulation prediction results (future concentration prediction values) are integrated with the identified typical accumulation pattern characteristics (such as pattern shape, rate) and the accumulation trend information corrected by environmental factors. The respiratory layer pollution accumulation feature set contains a comprehensive description of the current and future short-term accumulation status of respiratory layer pollutants, such as: The concentration is 80µg / m , is in a "rapid rise" mode and is expected to rise to 120µg / m in the next 10 minutes , reaching 150µg / m within the next 30 minutes .

[0080] Based on the pollution accumulation characteristic set of the breathing layer, the pollutant migration dynamics parameter set (including the vertical diffusion coefficient, sedimentation rate, phase change rate, etc. of each layer), and the vertical characteristic layer division map (including the height range and environmental characteristics of each layer), a vertical layered pollutant load index is constructed. For each vertical characteristic layer (ground layer, lower layer, breathing layer, upper layer, top layer), the pollutant load index is calculated by comprehensively considering the following factors: the concentration of the main pollutants in the layer (especially the current and short-term predicted concentrations of the breathing layer from the accumulation characteristic set), the ratio of the pollutant concentration in the layer to the health standard threshold (health risk), the vertical migration rate of pollutants under the temperature and humidity conditions of the layer (diffusion / sedimentation / phase change, from the migration dynamics parameter set, reflecting the retention or removal difficulty of pollutants in the layer), and the environmental characteristics of the layer (such as temperature and humidity variability, reflecting the stability of the environment in the layer). A multi-factor weighted summation model is constructed to calculate the load index. For example, for the breathing layer, the load index can be calculated using the formula: Load index_breathing layer = w1×(C_breathing layer / standard value_breathing layer) + w2×(1 / V v _eff_ breathing layer)+w3×(1 / D v _breathing layer) + w4 × V_temp_breathing layer + w5 × V_humidity_breathing layer. i is the weight coefficient, V v _eff is the effective vertical velocity (considering vertical motion caused by sedimentation and phase change), D v is the vertical diffusion coefficient, while V_temp and V_humidity are the coefficients of variation of temperature and humidity. Weighting coefficients can be determined through historical data regression analysis or expert experience, for example, assigning higher weights to breathing layer concentrations and effective vertical velocity (negatively correlated, meaning slower velocity indicates higher load). Ensure that the index value is within the range of 0-100, with higher values ​​indicating heavier loads. Calculate the corresponding pollutant load index for all vertical characteristic layers to form a complete vertically stratified pollutant load index distribution.

[0081] Preferably, the simulation of the stratified airflow distribution characteristics in step S3 is specifically as follows:

[0082] Analyze ventilation effect data based on historical ventilation operation records and vertical stratified pollutant load index to obtain stratified ventilation effect evaluation data;

[0083] A simplified tunnel geometry model is constructed based on tunnel structural parameters to obtain a tunnel grid model;

[0084] Arrange ventilation equipment parameters based on the stratified ventilation effect evaluation data to obtain a ventilation equipment characteristic table; set ventilation plan scenarios based on the ventilation equipment characteristic table;

[0085] Determine the calculation boundary condition table for the ventilation scenario; solve the simplified airflow equation for the tunnel grid model based on the boundary condition table to obtain the airflow field calculation results;

[0086] Extract vertical velocity profile sets from airflow field calculation results;

[0087] Based on the calculation results of the airflow field, the turbulence characteristics at different locations and heights in the tunnel are analyzed to obtain a turbulence characteristics distribution table;

[0088] Based on the airflow field calculation results and the vertical velocity profile set, the airflow dead zone and high-speed zone are identified, and the distribution map of the dead zone and high-speed zone is obtained;

[0089] Construct a vertical stratified airflow characteristic map based on the vertical velocity profile set, turbulence characteristic distribution table and dead zone high-speed area distribution map;

[0090] The stratified ventilation efficiency coefficient is calculated based on the vertical stratified airflow characteristic diagram and stratified ventilation effect evaluation data.

[0091] In an embodiment of the present invention, ventilation effect data analysis is performed based on historical ventilation operation records (including the number, model, speed, jet angle, etc. of fans turned on at different time points) and corresponding vertical stratified pollutant load index data to obtain stratified ventilation effect evaluation data. The changes in the pollutant load index of each vertical characteristic layer (ground layer, lower layer, breathing layer, upper layer, top layer) before and after the implementation of a specific ventilation scheme (for example, turning on the jet fan of a certain ventilation shaft with a speed of 70%) are analyzed. The rate of decrease of the pollutant load index of each layer or the time required to maintain stability under different ventilation schemes is calculated. For example, when the traffic flow is medium, turn on fan A, the breathing layer It takes 10 minutes for the load index to drop from 60 to 40; turning on both fan A and fan B takes 5 minutes to drop to 40. This historical data is accumulated to form a stratified ventilation effectiveness evaluation database, which records the relationship between different ventilation inputs (fan status) and the pollutant load index response of each layer.

[0092] A simplified tunnel geometry model is constructed based on tunnel structural parameters (such as tunnel length, cross-sectional shape, number of lanes, ventilation shaft location, and jet fan installation position and angle) to produce a tunnel mesh model. Computer-aided design (CAD) tools are used to create a three-dimensional tunnel geometry model, including the tunnel body, pavement, walls, roof, as well as structures such as ventilation shafts, jet fans, and fresh air inlets. This geometry is then imported into a meshing tool. Based on computational accuracy requirements and computational resource constraints, the three-dimensional space is discretized into a large number of mesh elements (e.g., hexahedral or tetrahedral meshes). The mesh is refined in areas requiring detailed analysis (e.g., near the jet fan outlet and the breathing zone), while the mesh can be sparse in other areas. This results in a tunnel mesh model for fluid dynamics calculations.

[0093] The ventilation equipment parameters are organized based on the stratified ventilation effect evaluation data to generate a ventilation equipment characteristic table. The performance parameters of each fan model are extracted from historical ventilation records, including air volume, air pressure, jet velocity, and range at different speeds. If the jet fan supports angle adjustment, the jet characteristics at different angles are recorded. The cross-sectional area and location information of the fresh air inlet / exhaust outlet are organized. The ventilation equipment characteristic table summarizes the key operating parameters and performance data of all ventilation equipment in the tunnel. Based on the ventilation equipment characteristic table, ventilation plan scenarios are set. Each scenario represents a specific ventilation system operating state, for example: Scenario 1: All fans are turned off; Scenario 2: The jet fan unit at the tunnel entrance is turned on at a speed of 50% and an angle of 0°; Scenario 3: The central ventilation shaft is turned on for exhaust, and the jet fan at the entrance has a speed of 80% and an angle of +10° (upward tilt). Multiple ventilation plan scenarios are set to cover common tunnel operating modes.

[0094] Determine the calculation boundary condition table for the ventilation scheme scenario. For each set ventilation scheme scenario, convert it into the boundary conditions required for fluid dynamics calculation. For example, for the scenario where the jet fan is turned on, set the fan outlet as the velocity inlet boundary condition, and the velocity magnitude and direction are determined according to the fan characteristic table; for the scenario where the ventilation shaft is used as the exhaust outlet, set it as the pressure outlet boundary condition; the tunnel entrance and exit can be set as pressure boundary conditions (such as atmospheric pressure) or flow boundary conditions (taking into account the piston wind effect, but here we focus on the role of the ventilation system and can simplify it). The road surface and wall are set as no-slip wall boundary conditions. The calculation boundary condition table describes in detail the fluid dynamics parameter settings on each boundary of the tunnel geometric model under each ventilation scheme scenario.

[0095] Based on the boundary condition table, simplified airflow equations are solved for the tunnel mesh model to obtain the airflow field calculation results. A computational fluid dynamics (CFD) solver is used to solve the tunnel mesh model using simplified airflow equations based on the Navier-Stokes equations. Given the large scale of the tunnel, simplified calculations using a two-dimensional or quasi-three-dimensional model can be used to improve efficiency, or a more efficient turbulence model (such as the standard k-ε model) can be used for a three-dimensional solution. Based on the specified boundary conditions, the solver calculates the air velocity vector (including vertical velocity) and pressure within each mesh cell in the tunnel under steady-state or unsteady-state conditions. This results in a three-dimensional airflow field calculation describing the airflow distribution within the tunnel.

[0096] A set of vertical velocity profiles is extracted from the airflow field calculation results. Within the airflow field calculation results, multiple representative sections are selected along the tunnel axis (e.g., near key monitoring areas, downstream of the fan, etc.). Within each section, vertical velocity components are extracted at different heights along the vertical direction (from the road surface to the roof). These vertical velocities are plotted as a function of height to form vertical velocity profiles. The vertical velocity profile set contains the distribution characteristics of vertical air movement at different tunnel locations under different ventilation schemes. For example, a significant upward or downward vertical velocity is observed downstream of the jet fan.

[0097] Based on the calculation results of the airflow field, the turbulence characteristics at different locations and heights in the tunnel are analyzed to obtain a turbulence characteristic distribution table. Parameters such as turbulent kinetic energy k and turbulent dissipation rate ε are extracted from the results of the CFD solver, or the turbulence intensity I=u is directly calculated. / U, where u is the root mean square velocity fluctuation, and U is the local average velocity. The spatial distribution of these turbulence parameters within the tunnel is analyzed, with particular attention paid to turbulence intensity at different vertical heights. Areas of high turbulence intensity facilitate mixing and dilution of pollutants. The turbulence characteristic distribution table records turbulence intensity or related parameters at different locations and heights within the tunnel under different ventilation schemes.

[0098] Based on the airflow field calculation results and the vertical velocity profile set, dead zones and high-speed zones are identified, generating a dead zone and high-speed zone distribution map. Dead zones are areas with extremely low air velocity (close to zero), where pollutants easily accumulate. High-speed zones are areas where air velocity is much higher than the average velocity, facilitating the rapid transport of pollutants. Based on the magnitude of the velocity vector in the airflow field calculation results, speed thresholds are set to identify dead zones and high-speed zones. For example, areas with velocities less than 0.1 m / s are marked as dead zones, and areas with velocities greater than 5 m / s are marked as high-speed zones. The identified dead zones and high-speed zones are visualized on the tunnel geometry model to generate a dead zone and high-speed zone distribution map. This map shows the areas where pollutants stagnate or are rapidly transported under a specific ventilation scheme, as well as their vertical distribution.

[0099] Based on the vertical velocity profile set, turbulence characteristic distribution table, and dead zone high-speed area distribution map, a vertical stratified airflow characteristic map is constructed. Combining the previously extracted airflow characteristics, the airflow characteristics of each vertical characteristic layer (ground layer, lower layer, breathing layer, upper layer, and top layer) are described under different ventilation schemes. For example, under one ventilation scheme, the breathing layer exhibits an average axial velocity of 1 m / s, an average vertical velocity of 0.05 m / s (upward), an average turbulence intensity of 0.12, and the presence of small localized dead zones. The vertical stratified airflow characteristic map quantifies the impact of different ventilation schemes on the airflow state of each vertical layer.

[0100] The stratified ventilation efficiency coefficient is calculated based on the vertical stratified airflow characteristic diagram and the stratified ventilation effect evaluation data. The stratified ventilation efficiency coefficient quantifies the removal capacity of a specific pollutant (or pollutant load index) in a specific vertical layer per unit air volume or unit ventilation energy consumption. For example, for the breathing layer, its The total ventilation volume (or energy consumption) required to reduce the load index by 1 unit. The reciprocal of this value is the ventilation efficiency coefficient for that layer. Analyze the differences in ventilation efficiency coefficients for each layer under different ventilation schemes. For example, an upward-angled jet improves the efficiency of the upper layers but reduces the efficiency of the ground layer; increasing the fan speed improves the efficiency of all layers but sharply increases energy consumption. Consider the differences in the removal efficiency of different pollutants (gases and particulates) under different airflow conditions. The layered ventilation efficiency coefficient table records the ventilation efficiency coefficients for each vertical characteristic layer and each major pollutant under different ventilation schemes, reflecting the current ventilation scheme's effectiveness in removing pollutants at different height levels.

[0101] Preferably, the required air volume prediction in step S3 is specifically as follows:

[0102] The vertically layered pollutant load index, real-time traffic flow data, and environmental condition parameters were screened and preprocessed to obtain a normalized feature data set.

[0103] Perform historical data association analysis on the normalized feature data set to obtain a feature importance ranking table;

[0104] Design a three-layer BP neural network structure based on the feature importance ranking table and normalized feature data set;

[0105] Construct training and validation datasets based on historical ventilation operation records and normalized feature datasets;

[0106] The training and validation data sets are used to train and optimize the three-layer BP neural network structure to obtain a training complete model;

[0107] The trained model is used to perform real-time prediction and short-term trend prediction to obtain the overall air demand forecast value.

[0108] In this embodiment of the present invention, characteristic variables are screened and preprocessed based on vertically layered pollutant load indices (including load indices for the ground layer, lower layer, breathing layer, upper layer, and top layer), real-time traffic flow data (including total vehicle volume, proportion of each vehicle type, and average speed), and environmental condition parameters (including tunnel entrance / exit temperature, humidity, air pressure, and average temperature and humidity of each layer within the tunnel) to produce a normalized characteristic dataset. Characteristic variables highly correlated with air volume demand prediction are selected, such as the breathing layer pollutant load index, total vehicle volume, proportion of heavy trucks, average tunnel temperature, and average tunnel humidity. Data cleaning is performed on the selected characteristic variables to remove outliers and missing values. Normalization is then performed, for example, using the minimum-maximum scaling method to scale all characteristic variable values ​​to the range [0, 1]. The normalization formula is: X_normalized = (X - X_min) / (X_max - X_min), where X is the original value and X_min and X_max are the historical minimum and maximum values ​​of the characteristic variable. The normalized feature variables constitute a normalized feature dataset, which is used as the input of the neural network.

[0109] Perform historical data association analysis on the normalized feature data set to obtain a feature importance ranking table. Use the historically collected normalized feature data set and the corresponding historical total ventilation data to analyze the correlation strength between each feature variable and the total ventilation volume. Use the Pearson correlation coefficient, mutual information, or tree-based feature importance evaluation method (such as random forest feature importance) to quantify the impact of each feature variable on the total ventilation volume. Sort the feature variables according to the calculated correlation strength or importance score to generate a feature importance ranking table. For example, the table shows: breathing layer The load index is important at 0.9, the total traffic volume is important at 0.8, the proportion of heavy trucks is important at 0.75, the average tunnel temperature is important at 0.6, and the average tunnel humidity is important at 0.55. This table guides the selection of input features and the design of the structure of the subsequent neural network model.

[0110] Based on the feature importance ranking table and the normalized feature data set, a three-layer BP neural network structure is designed. Feature variables with higher importance are selected as the input of the neural network. For example, according to the feature importance ranking table, the CO load index of the breathing layer, the respiratory layer Load index, total traffic volume, proportion of heavy trucks, average tunnel temperature, and average tunnel humidity are used as input features. The number of input layer neurons is 6. A three-layer BP neural network with one hidden layer is designed. The number of hidden layer neurons is determined based on an empirical formula (e.g., the square root of the sum of the number of input and output layer neurons or a constant) or through cross-validation experiments. For example, the number of hidden layer neurons is set to 10. The number of output layer neurons is 1, indicating the total air demand to be predicted (e.g., in m2). The neural network structure is: input layer (6 neurons) → hidden layer (10 neurons) → output layer (1 neuron). The activation function uses the ReLU function in the hidden layer and the linear function in the output layer.

[0111] A training and validation dataset is constructed based on historical ventilation operation records and a normalized feature dataset. A long-term normalized feature dataset (input) and the corresponding historical total ventilation volume data (output) are extracted from the historical database. Total ventilation volume data can be calculated based on historical fan operating states and fan characteristic tables. This historical data is divided into a training set and a validation set in chronological order, for example, the first 80% of the data is used for training and the last 20% for validation. The training set is used to adjust the weights and biases of the neural network, while the validation set is used to evaluate the model's generalization ability and select optimal model parameters to avoid overfitting.

[0112] The three-layer BP neural network structure is trained and optimized using the training and validation datasets to obtain a trained model. The backpropagation algorithm is used as the training algorithm. The weights and biases of the neural network are initialized. The training set data is input into the neural network, and the error between the network's output value and the actual historical total ventilation volume is calculated (for example, using the mean squared error (MSE) as the loss function). By backpropagating the error, the connection weights and biases between neurons in each layer of the network are adjusted to minimize the loss function. The Adam optimizer is used for weight updates. The number of training iterations (epochs) and batch size (batchsize) are set. During training, the performance of the model is regularly evaluated using the validation set. The loss changes on the validation set are monitored, and training is stopped when the validation set loss no longer decreases significantly to avoid overfitting. After the training process, a set of optimal weight and bias parameters are obtained, forming a trained BP neural network model.

[0113] The trained model is used for real-time prediction and short-term trend prediction to obtain the predicted value of the total air volume demand. The real-time normalized feature data set (including the currently collected load index of each layer, traffic flow, and environmental parameters) is input into the trained BP neural network model. The model outputs a predicted value through forward propagation calculation, which is the total ventilation volume required at the current moment. In order to make short-term trend predictions, the cumulative pattern features and time series feature parameters identified in the historical data can be used to predict the changing trends of the vertical stratified pollutant load index, traffic flow and environmental condition parameters in the next few minutes (for example, the next 10 minutes, 20 minutes, and 30 minutes). These predicted feature data of the future moments are used as input and input into the trained model to obtain the predicted value of the total air volume demand at the corresponding moment in the future. For example, it is predicted that the air volume demand at the current moment is 150m / s, and it is predicted that the required air volume will rise to 180m3 in 10 minutes. / s. These forecast values ​​together constitute the overall air demand forecast value.

[0114] Preferably, the optimized distribution of required air volume by layer in step S3 is as follows:

[0115] Set up a stratified quality target table based on the vertical stratified pollutant load index;

[0116] According to the layered ventilation efficiency coefficient, the difference in ventilation efficiency between layers is analyzed to obtain the efficiency difference table between layers;

[0117] The preliminary allocation coefficient table is calculated based on the inter-layer efficiency difference table and the layer quality target table;

[0118] Determine the technical constraint condition set of the tunnel ventilation system based on the predicted value of the total air demand;

[0119] The hierarchical optimized air volume table is calculated based on the preliminary allocation coefficient table, the technical constraint condition set and the overall air volume forecast value;

[0120] Perform ventilation mode parameter conversion on the stratified optimized air volume table to obtain the equipment parameter solution table;

[0121] Estimate the stratified air volume effect based on the equipment parameter plan table and stratified ventilation efficiency coefficient table to obtain the expected effect evaluation table;

[0122] Integrate the stratified optimized air volume table, equipment parameter plan table and expected effect evaluation table to construct a stratified air volume requirement matrix.

[0123] In the embodiment of the present invention, a stratified quality target table is set based on the real-time vertical stratified pollutant load index. For example, the health standard stipulates that the CO concentration shall not exceed 90ppm. The concentration shall not exceed 75µg / m . These standards are converted into load index target values, for example, when CO and When concentrations reach the upper limit of the standard, the load index is set to 80. Considering the higher air quality requirements for the driver's breathing layer, a more stringent quality target can be set for it. For example, when the CO concentration in the breathing layer reaches 80% of the upper limit of the standard, the load index target value is set to 70. For other layers, the target values ​​can be appropriately relaxed or aligned with the general standard. The layered quality target table records the pollutant load index target values ​​that each vertical characteristic layer (ground layer, lower layer, breathing layer, upper layer, and top layer) should strive to achieve at different times and operating conditions. For example, a target load index of 60 for the breathing layer and 75 for the ground layer.

[0124] According to the stratified ventilation efficiency coefficient table (which records the ventilation efficiency coefficients of each vertical layer and each pollutant under different ventilation schemes), an analysis of the differences in ventilation efficiency between layers is performed to obtain an inter-layer efficiency difference table. For example, under the longitudinal ventilation mode, the efficiency of the upper layer is much higher than that of the ground layer and the breathing layer; and under certain special ventilation modes (such as the use of ventilation shafts for local supply and exhaust), the efficiency of the breathing layer is significantly improved. Calculate the ratio of each vertical characteristic layer to the overall average ventilation efficiency under different ventilation schemes. For example, if the overall average ventilation efficiency is E_avg and the efficiency of the breathing layer is E_respir, then the efficiency difference coefficient of the breathing layer is E_respir / E_avg. The inter-layer efficiency difference table quantifies the difference in the difficulty of removing pollutants in each vertical layer relative to the overall situation under different ventilation schemes.

[0125] Based on the inter-layer efficiency difference table and the stratified quality target table, a preliminary allocation coefficient table is calculated. The goal is to preliminarily determine the proportion of ventilation resources allocated to each layer based on the current pollutant load of each layer and the set quality target, combined with the difference in removal efficiency of each layer. For example, if the breathing layer load index is high and the target value is strict, and the ventilation efficiency of this layer is relatively low, more air volume will need to be allocated. The preliminary allocation coefficient can be calculated by a simple ratio: preliminary allocation coefficient ᵢ = (current load index i -Target load index i ) / efficiency difference coefficient i The preliminary distribution coefficients for all layers are then normalized so that their sum is 1. The preliminary distribution coefficient table gives the ideal air volume distribution ratio based on demand and efficiency differences without considering equipment constraints. For example: the preliminary distribution coefficient for the breathing layer is 0.4, the ground layer is 0.15, the lower layer is 0.2, the upper layer is 0.15, and the top layer is 0.1.

[0126] According to the total air volume forecast value (for example, the total air volume forecasted by the BP neural network at the current moment is 150m / s), determine the technical constraints of the tunnel ventilation system. These constraints include: the total ventilation volume cannot exceed the maximum total air supply capacity of all fans in the tunnel (for example, the maximum total air volume of all fans fully opened is 200m / s); minimum / maximum speed limits for individual fans (e.g., speed range of 30%-100%); jet fan jet angle adjustment range (e.g., -15° to +15°); opening limits for fresh air inlets and outlets; switching times and applicable conditions between different ventilation modes (longitudinal, semi-transverse, and full-transverse); and minimum air speed requirements to ensure safe and stable airflow organization. The set of technical constraints details the feasible operating parameter ranges and mode restrictions for the ventilation system under the current forecast total air demand.

[0127] Based on the preliminary distribution coefficient table, technical constraint condition set and the total air volume forecast value, the hierarchical optimized air volume table is calculated. This is an optimization problem. The goal is to reduce the total air volume (150m / s) are optimally distributed to each ventilation device (fan, damper, etc.), so that the air volume combination of each device can most effectively reduce the pollutant load index of each vertical characteristic layer, especially give priority to meeting the quality target of the breathing layer. Solve using optimization algorithms (such as linear programming, nonlinear programming or genetic algorithm). The objective function can be to minimize the sum of the gaps between the load index of each layer and the target index (weighted, the breathing layer has the highest weight), or to maximize the sum of the pollutant removal rates of each layer. The constraints are that the total air volume is equal to the predicted value and the operating parameters of each device are within the technical constraints. The solution result is the optimal air volume that each ventilation device (such as a jet fan or the exhaust volume of a ventilation shaft) should provide. The air volume of these devices is converted into the effective contribution air volume or removal capacity to each vertical characteristic layer according to their position and function to form a layered optimized air volume table. For example, the layered optimized air volume table shows that in order to reduce the load index of the breathing layer to the target value, a specific fan combination is required to provide the air volume Xm equivalent to the removal capacity of this layer. / s, providing Ym for the ground floor / s, etc.

[0128] The ventilation mode parameters of the layered optimized air volume table are converted to obtain the equipment parameter scheme table. The optimal air volume or state of each ventilation device calculated by the optimization algorithm is converted into a device control instruction that can be directly sent to the ventilation control system. For example, if the optimization result requires a jet fan to provide 80m / s of air volume, which corresponds to an 85% speed according to the fan's characteristic table. If the airflow organization needs to be adjusted to improve breathing layer ventilation, the optimization results indicate that a certain jet fan needs to adjust the jet angle to +10°. The equipment parameter solution table lists the specific control parameter setting values ​​for each ventilation device required to achieve stratified optimized air volume, for example: Jet fan No. 1 speed 85%, angle +10°; Jet fan No. 2 speed 60%, angle 0°; ventilation shaft exhaust valve No. 3 opening degree 70%, etc.

[0129] According to the equipment parameter scheme table and the stratified ventilation efficiency coefficient table, the stratified air volume effect is estimated to obtain the expected effect evaluation table. Using the ventilation scheme set by the equipment parameter scheme table, refer to the stratified ventilation efficiency coefficient table (which quantifies the removal capacity of each layer under different ventilation schemes) to estimate how the pollutant load index of each vertical characteristic layer will change after the implementation of the scheme, and how long it will take to reach the target load index. For example, according to the fan status set in the scheme, look up or interpolate the stratified ventilation efficiency coefficient table to obtain the ventilation load index of the breathing layer under the scheme. The removal efficiency coefficient is E_ _respir. If the current breathing layer The load index is 70 and the target is 60. It is estimated that the target can be achieved within a certain period of time through this plan, and the approximate time required is calculated. The expected effect evaluation table records the expected effect of improving the air quality of each vertical characteristic layer after implementing the current equipment parameter plan, such as: breathing layer The load index is expected to drop to 60 within 15 minutes, and the CO load index is expected to drop to 50 within 10 minutes; ground floor The load index is expected to drop to 70 within 20 minutes.

[0130] The stratified air demand matrix is ​​constructed by integrating the stratified optimized air volume table, the equipment parameter plan table, and the expected effect evaluation table. The stratified air demand matrix is ​​a comprehensive output that contains all key decision-making information for the current air demand forecast and allocation. The main contents of the matrix include: the predicted overall air demand value; the "equivalent" air demand or removal capacity requirements allocated to each vertical characteristic layer to achieve air quality goals; the specific ventilation equipment control parameter settings required to achieve these requirements (fan speed, angle, air valve opening, etc.); and the expected effect of implementing the plan on improving air quality in each vertical layer (such as the load index change trend and the time required to achieve the target). This matrix not only provides the overall air demand value, but more importantly, it provides a detailed implementation plan for achieving stratified precise ventilation control, namely, recommendations for differentiated allocation of ventilation resources for different height layers and corresponding equipment operating instructions.

[0131] As an example of the present invention, refer to Figure 2 As shown, in this example, step S4 includes:

[0132] Step S41: Obtain and evaluate equipment capacity and traffic control range based on tunnel ventilation equipment parameters and traffic management system parameters to obtain a collaborative control feasible domain diagram;

[0133] Step S42: constructing a three-dimensional evaluation system of energy consumption, air quality, and traffic efficiency based on the collaborative control feasible region diagram and the layered air demand matrix;

[0134] Step S43: Generate and evaluate collaborative control strategies based on the three-dimensional evaluation system and the collaborative control feasible region diagram to obtain candidate control strategy solutions;

[0135] Step S44: Dynamically optimize and execute the control strategy candidate solutions to obtain a bidirectional control execution solution.

[0136] In an embodiment of the present invention, the equipment capacity and traffic control range are evaluated based on the tunnel ventilation equipment parameters and traffic management system parameters to obtain a collaborative control feasible domain diagram. The tunnel ventilation equipment parameters include the rated power of each fan, the maximum air supply volume, the minimum / maximum rotation speed, the jet angle adjustable range, the ventilation shaft air valve opening range, etc. The traffic management system parameters include the display capability of the variable message sign (VMS), the setting range of the speed limit sign (for example, the minimum speed limit is 60km / h, the maximum speed limit is 100km / h), the feasibility of variable lane control, the travel time / lane restriction strategy capability of different vehicle types (such as heavy trucks), the information release coverage of traffic broadcasts and mobile phone APPs, and the driver response time estimation. Evaluate the total maximum air supply capacity of the ventilation system (for example, the total air supply volume of 200m when all fans are fully opened) / s) and minimum air supply capacity (for example, the minimum fan combination air supply required to meet basic ventilation needs). Evaluate the control potential of the traffic management system under different traffic flows. For example, when the flow is close to saturation, speed limits or lane separation have a greater impact on traffic efficiency; at low flows, the impact is smaller. The feasible domain diagram of coordinated control is a multidimensional space, whose dimensions represent ventilation control parameters (such as total air volume, fan combination, angle) and traffic control parameters (such as speed limit, heavy truck traffic ratio). The feasible domain contains all technically and operationally feasible ventilation-traffic control parameter combinations. For example, the feasible domain limits the total air volume to 50m / s to 200m / s, the speed limit is between 60km / h and 100km / h, and the proportion of heavy trucks is limited between 0% and 100%.

[0137] A three-dimensional evaluation system of energy consumption, air quality and traffic efficiency is constructed for the collaborative control feasible domain diagram and the layered air demand matrix (including the predicted overall air demand, target demand of each layer, equipment parameter scheme and expected effect). Define energy consumption indicators: for example, the total input power (kW) of all running fans under the current ventilation scheme is used as the main energy consumption indicator, and the number of equipment starts and stops can also be considered. Define air quality indicators: for example, the average difference or maximum difference between the current and predicted pollutant load index and the target load index of each vertical characteristic layer (especially the breathing layer) is used as the air quality indicator. It can be further refined, such as calculating the CO and The weighted average of the ratios of concentrations to health-related thresholds. Define traffic efficiency indicators: For example, average tunnel speed, number of vehicles passing per unit time, and vehicle delay time (relative to free flow) can be used as traffic efficiency indicators. Construct a comprehensive benefit index, which is the weighted sum of energy consumption, air quality, and traffic efficiency. For example, the comprehensive benefit index = w_AQ × AQ_score - w_E × E_score - w_T × T_score, where AQ_score is the air quality score (higher is better), E_score is the energy consumption score (higher means lower energy consumption), and T_score is the traffic efficiency score (higher is better). w_AQ, w_E, and w_T are weighting coefficients, and w_AQ + w_E + w_T = 1. The weighting reflects the priority of the control objectives; for example, breathing zone air quality has the highest weight. The three original indicators are converted to scores on a 0-100 scale using standardization or a utility function, and then the weighted sum is used to obtain the comprehensive benefit index. The three-dimensional evaluation system provides a standard for quantitatively evaluating the effectiveness of different control strategies.

[0138] Based on the three-dimensional evaluation system and the collaborative control feasible domain diagram, the collaborative control strategy is generated and evaluated to obtain the control strategy candidate scheme. In the collaborative control feasible domain, a series of discrete ventilation-traffic control parameter combinations are generated as candidate strategies. For example, strategy 1: total air volume 120m / s, fan combination A+B, angle 0°, speed limit 80km / h, no restrictions on heavy trucks. Strategy 2: total air volume 100m / s, fan combination A+C, angle +15°, speed limit 70 km / h, and heavy trucks restricted to specific lanes. For each generated candidate strategy, a prediction model (built based on historical data and a simplified model) is used to calculate indicators such as tunnel energy consumption, air quality in each vertical layer (especially changes in the pollutant load index in the breathing layer), and traffic efficiency (average speed, delays) over a period of time under that strategy. The comprehensive benefit index of each strategy is then calculated based on a three-dimensional evaluation system. For example, Strategy 1 results in lower energy consumption and higher traffic efficiency, but with slower improvement in breathing layer air quality; Strategy 2 effectively improves breathing layer air quality, but with higher energy consumption and reduced traffic efficiency. All candidate strategies are ranked according to the comprehensive benefit index, and the strategies with the highest comprehensive benefits are selected as candidate control strategies. These strategies represent the optimal control options under different trade-offs.

[0139] Dynamically optimize and execute candidate control strategies to obtain a two-way control execution plan. Based on real-time environmental monitoring data (pollutant load index at each layer) and traffic monitoring data (vehicle volume, speed, vehicle type), dynamically evaluate the solutions in the control strategy candidate solution library. For example, if the real-time monitoring shows that the pollutant load index of the respiratory layer is rising rapidly, the system will give priority to those strategies with high scores in air quality indicators. According to the actual traffic conditions (such as whether congestion has occurred), the feasibility and impact of traffic control measures are judged. Use dynamic optimization algorithms (such as the framework of model predictive control MPC) to make real-time selections or fine-tuning among candidate solutions. In each control cycle (for example, 1 minute), the algorithm searches for the optimal ventilation-traffic parameter combination in the feasible domain based on the current state and short-term forecast (from the demand air volume forecast) to maximize the comprehensive benefit index in the future. The optimization results generate a specific two-way control execution plan, which consists of two parts: one is precise control instructions for the ventilation system, such as instructions sent to the ventilation control system: set the speed of fan No. 1 to 88%, the jet angle to +12°; set the speed of fan No. 2 to 65%, the jet angle to 0°; and set the opening of the exhaust valve of ventilation shaft No. 3 to 75%. The other is control suggestions for traffic management, such as instructions sent to the traffic guidance system: the tunnel entrance VMS displays "congestion ahead, speed limit 70km / h"; heavy trucks are advised to drive in the right lane. The system sends these instructions and suggestions to the corresponding control execution units to achieve coordinated control of ventilation and traffic. At the same time, it continuously receives real-time monitoring feedback data after execution for dynamic optimization of the next control cycle, forming a closed-loop control.

[0140] Preferably, the collaborative control strategy generation and evaluation in step S4 is specifically as follows:

[0141] Generate control parameter combinations based on the collaborative control feasible domain diagram;

[0142] Calculate the air quality impact of the control parameter combination to obtain an air quality impact table;

[0143] Conduct energy consumption benefit analysis on the control parameter combination to obtain an energy consumption analysis table;

[0144] Carry out traffic efficiency evaluation on the control parameter combination to obtain the traffic impact evaluation table;

[0145] Calculate the comprehensive benefit index table for the air quality impact table, energy consumption analysis table, and traffic impact assessment table based on the three-dimensional evaluation system;

[0146] The comprehensive benefit index table is used to evaluate and screen the control parameter combinations to obtain candidate control strategy solutions.

[0147] In this embodiment of the present invention, control parameter combinations are generated based on a coordinated control feasible domain diagram (which defines the permissible ranges for ventilation and traffic parameters). Within this feasible domain, a series of parameter combinations representing different ventilation and traffic control states are generated through discretization or sampling. For example, representative values ​​for control dimensions such as total air volume, fan combination, jet angle, speed limit, and heavy truck ratio limit can be selected and then combined.

[0148] For example, the following control parameter combinations are generated:

[0149] Combination 1: Total air volume 120 m³ / s (achieved by fan combination A+B), jet angle 0°, speed limit 80 km / h, no restrictions for heavy trucks.

[0150] Combination 2: Total air volume 150 m³ / s (achieved by fan combination A+B+C), jet angle +10°, speed limit 80 km / h, no restrictions for heavy trucks.

[0151] Combination 3: Total air volume 100 m³ / s (achieved by fan combination A+C), jet angle +15°, speed limit 70 km / h, heavy trucks restricted to the right lane.

[0152] Combination 4: Total air volume 80 m³ / s (achieved by fan combination A), jet angle 0°, speed limit 60 km / h, heavy trucks are restricted to the right lane and are advised to use it in different time periods.

[0153] Generate a sufficient number of control parameter combinations covering typical control states in the feasible domain to form a set of strategies to be evaluated.

[0154] For each control parameter combination, the air quality impact is calculated in combination with the current real-time environmental status (pollutant load index of each layer) and traffic status (vehicle volume, vehicle type composition), and an air quality impact table is obtained. Using the layered ventilation efficiency coefficient table and pollutant migration dynamics model constructed in step 3, it is estimated that under the current environmental and traffic conditions, after the implementation of the control parameter combination, the pollutant concentration and load index of each vertical characteristic layer (especially the breathing layer) will change in the future (for example, the next 30 minutes) and the predicted value. For example, for combination 3 (total air volume 100m³ / s, angle +15°, speed limit 70, trucks restricted to the right), its impact on the breathing layer is estimated. The removal rate of the load index is X1, and the removal rate of the CO load index is , for the ground floor The clearance rate of the loading index is Calculate whether the strategy can reduce the breathing layer pollutant load index below the target value within the predicted timeframe, and the time required to achieve the target. Quantify these estimated air quality improvements, such as the predicted reduction in the load index at each layer, the time required to reach the target load index, and the predicted peak load index. The air quality impact table records the estimated quantitative impact of each control parameter combination on the air quality of each vertical characteristic layer.

[0155] Perform an energy efficiency analysis for each control parameter combination to generate an energy consumption analysis table. Based on the fan combination and operating parameters (such as speed) set in the control parameter combination, as well as the power characteristic curve of each fan (a function of power changing with speed), calculate the total input power of all running fans under this combination. For example, the total input power corresponding to combination 2 (total air volume 150m³ / s, fan combination A+B+C) is At the same time, the additional energy consumption caused by switching ventilation modes or starting and stopping the fan can be considered. The energy consumption analysis table records the total energy consumption or power consumption corresponding to each control parameter combination.

[0156] Traffic efficiency is evaluated for each control parameter combination, resulting in a traffic impact assessment table. Based on the traffic control measures (such as speed limits and heavy truck restrictions) specified in the control parameter combination, as well as the current traffic volume, vehicle type composition, and average speed, a traffic flow model (such as a macroscopic flow model or a microscopic simulation model) is used to estimate the impact of these measures on tunnel traffic efficiency. For example, a 70 km / h speed limit (combination 3) reduces the average speed from 85 km / h to 75 km / h while also reducing traffic fluctuations. Restricting heavy trucks to the right lane improves traffic efficiency in the left lane but causes localized congestion in the right lane or a slight decrease in overall capacity. Evaluation metrics include predicted average tunnel speed, travel time, number of vehicles passing through per unit time, and vehicle delay rate. The traffic impact assessment table records the quantitative impact of each control parameter combination on tunnel traffic efficiency.

[0157] Based on the three-dimensional evaluation system (which includes scoring criteria and weightings for energy consumption, air quality, and traffic efficiency), a comprehensive benefit index table is calculated for the air quality impact table, energy consumption analysis table, and traffic impact assessment table. The estimated effects (e.g., load index reduction and time to standard) in the air quality impact table are converted into an air quality score (AQ_score). For example, a greater reduction and a shorter time to standard are associated with a higher score, and the breathing layer score is weighted more heavily. Power consumption in the energy consumption analysis table is converted into an energy consumption score (E_score). For example, lower power consumption is associated with a higher score. Efficiency indicators (e.g., average speed and delay) in the traffic impact assessment table are converted into a traffic efficiency score (T_score). For example, higher average speed and fewer delays are associated with a higher score. The comprehensive benefit index formula defined in the three-dimensional evaluation system (e.g., weighted summation formula: Comprehensive Benefit Index = w_AQ × AQ_score - w_E × E_score - w_T × T_score) is then used to calculate the comprehensive benefit index for each control parameter combination. The comprehensive benefit index table lists each control parameter combination and its corresponding quantitative comprehensive benefit score.

[0158] The comprehensive benefit index table is used to evaluate and screen the control parameter combinations to obtain candidate control strategy solutions. According to the comprehensive benefit index table, all control parameter combinations are sorted from high to low according to the comprehensive benefit index. Several control parameter combinations with the highest comprehensive benefit index (for example, the top 5) are selected as the current control strategy candidate solutions. These solutions represent the control options that can relatively best balance energy consumption, air quality (especially the breathing layer) and traffic efficiency under the current environment. For example, the selected candidate solutions include: strategies that can quickly improve the air quality of the breathing layer while controlling energy consumption and the impact on traffic within an acceptable range. These candidate solutions will be used for subsequent dynamic optimization and execution.

[0159] The present invention is therefore intended to be illustrative and non-restrictive in all respects, with the scope of the invention being defined by the appended claims rather than the foregoing description, and all changes that come within the meaning and range of equivalents of the application documents are intended to be embraced therein.

[0160] The foregoing description is intended only to provide specific embodiments of the present invention, which will enable those skilled in the art to understand and implement the present invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention is not intended to be limited to the embodiments shown herein, but is to be construed in the widest possible manner consistent with the principles and novel features disclosed herein.

Claims

1. A tunnel traffic volume and air volume prediction method based on BP neural network, characterized in that: The following steps are involved: Step S1: Obtain and deploy a multi-layer sensor array in the vertical section of the tunnel based on tunnel structural parameters and traffic flow data; use the multi-layer sensor array to collect environmental parameters along the vertical direction of the tunnel from the road surface to the top, and integrate traffic flow data to obtain environmental-traffic coupled data; Step S2: Extracting and correlating environment-traffic operating condition characteristics based on the environment-traffic coupling data to obtain a vertical feature layer partition map and an operating condition-pollution response correlation matrix; calculating pollutant migration dynamics under thermal and humid conditions based on the vertical feature layer partition map and the operating condition-pollution response correlation matrix to obtain a pollutant migration dynamics parameter set; extracting respiratory layer pollutant accumulation characteristics from the pollutant migration dynamics parameter set to obtain a vertically layered pollutant load index; Step S3: Obtain historical ventilation operation records and simulate the stratified airflow distribution characteristics based on the vertical stratified pollutant load index to obtain the stratified ventilation efficiency coefficient; predict the required air volume based on the BP neural network and the vertical stratified pollutant load index to obtain the overall required air volume prediction value; optimize the stratified air volume distribution based on the overall required air volume prediction value, the stratified ventilation efficiency coefficient, and the vertical stratified pollutant load index to obtain the stratified air volume matrix; Step S4: Constructing a three-dimensional evaluation system based on the layered air volume demand matrix; performing ventilation-traffic coordinated control based on the three-dimensional evaluation system to obtain a two-way control execution plan; The specific extraction and association of environment-traffic condition features in step S2 is as follows: The vertical space of the tunnel is divided into layers according to the environment-traffic coupling data to obtain vertical space layers, which include the ground layer, lower layer, breathing layer, upper layer and top layer. Extract environmental features from the vertical spatial layer to obtain a vertical feature layer division map; Traffic data is classified into different traffic condition types. Based on the vertical characteristic layer division map, the response characteristics of pollutant concentrations in each vertical characteristic layer are analyzed for different traffic condition types to obtain the condition-pollution response correlation matrix. The extraction of respiratory layer pollutant accumulation features in step S2 is specifically as follows: Based on the pollutant migration dynamics parameter set, the breathing layer boundary is located and data is extracted to obtain the breathing layer data subset; Calculate the relative enrichment coefficient of the respiratory layer data subset to obtain the pollutant enrichment coefficient table; Perform working condition correlation cumulative characteristic analysis on the pollutant enrichment coefficient table and the working condition-pollution response correlation matrix to obtain the working condition cumulative characteristic table; Quantify the impact of environmental factors based on the cumulative characteristic table of working conditions and obtain the environmental impact coefficient matrix; Based on the environmental impact coefficient matrix and the working condition cumulative feature table, the typical cumulative pattern of the breathing layer data subset is extracted to obtain the cumulative pattern characteristics; Perform time series analysis on the cumulative pattern characteristics to obtain time series characteristic parameters; According to the time series characteristic parameters, a short-term cumulative trend forecast is conducted to obtain a short-term cumulative forecast result; Integrate the short-term cumulative prediction results and cumulative pattern characteristics to obtain the respiratory layer pollution accumulation feature set; The vertically stratified pollutant load index is constructed based on the respiratory layer pollution accumulation characteristic set, the pollutant migration dynamics parameter set and the vertical characteristic layer division map to obtain the vertically stratified pollutant load index.

2. The tunnel traffic volume and required air volume prediction method based on BP neural network according to claim 1 is characterized in that: Step S1 is specifically as follows: Determine the key monitoring area based on tunnel structural parameters and traffic flow data, and arrange vertically layered monitoring points in the key monitoring area to obtain a multi-layer sensor array; Use distributed fiber optic sensing technology and multi-layer sensor arrays to collect tunnel environmental data and obtain vertical profile monitoring data; Extracting traffic flow characteristic parameters from traffic flow data; Analyze the particle characteristics of different altitude layers on the vertical profile monitoring data to obtain particle characteristic data; The particle characteristic data and traffic flow characteristic parameters are analyzed to correlate particle and traffic, and a layered particle characteristic spectrum is obtained; The vertical profile monitoring data, traffic flow characteristic parameters and layered particulate matter characteristic spectra are temporally and spatially unified and fused to obtain environment-traffic coupling data.

3. The tunnel traffic volume and required air volume prediction method based on BP neural network according to claim 1 is characterized in that: The specific calculation of pollutant migration dynamics under hot and humid conditions in step S2 is: Construct vertical temperature and humidity gradient field based on vertical characteristic layer division atlas; Determine pollutant-based characteristics of the operating condition-pollution response correlation matrix; The temperature and humidity correction coefficients are calculated based on the vertical temperature and humidity gradient field and the basic characteristics of pollutants to obtain the dynamic correction coefficient matrix; The vertical diffusion coefficient matrix is ​​calculated based on the kinetic correction coefficient matrix; The sedimentation rate and phase change rate of the particles are calculated based on the vertical diffusion coefficient matrix to obtain the phase change sedimentation rate table; The pollutant vertical flux equations are constructed based on the phase change sedimentation rate table and the vertical diffusion coefficient matrix; The pollutant vertical flux equations are numerically solved and the migration characteristics are extracted to obtain the pollutant migration kinetic parameter set.

4. The tunnel traffic volume and required air volume prediction method based on BP neural network according to claim 1 is characterized in that: The simulation of the stratified airflow distribution characteristics in step S3 is specifically as follows: Analyze ventilation effect data based on historical ventilation operation records and vertical stratified pollutant load index to obtain stratified ventilation effect evaluation data; A simplified tunnel geometry model is constructed based on tunnel structural parameters to obtain a tunnel grid model; Arrange ventilation equipment parameters based on the stratified ventilation effect evaluation data to obtain a ventilation equipment characteristic table; set ventilation plan scenarios based on the ventilation equipment characteristic table; Determine the calculation boundary condition table for the ventilation scenario; solve the simplified airflow equation for the tunnel grid model based on the boundary condition table to obtain the airflow field calculation results; Extract vertical velocity profile sets from airflow field calculation results; Based on the calculation results of the airflow field, the turbulence characteristics at different locations and heights in the tunnel are analyzed to obtain a turbulence characteristics distribution table; Based on the airflow field calculation results and the vertical velocity profile set, the airflow dead zone and high-speed zone are identified, and the distribution map of the dead zone and high-speed zone is obtained; Construct a vertical stratified airflow characteristic map based on the vertical velocity profile set, turbulence characteristic distribution table and dead zone high-speed area distribution map; The stratified ventilation efficiency coefficient is calculated based on the vertical stratified airflow characteristic diagram and stratified ventilation effect evaluation data.

5. The tunnel traffic volume and required air volume prediction method based on BP neural network according to claim 1, characterized in that: The required air volume prediction in step S3 is specifically as follows: The vertically layered pollutant load index, real-time traffic flow data, and environmental condition parameters were screened and preprocessed to obtain a normalized feature data set. Perform historical data association analysis on the normalized feature data set to obtain a feature importance ranking table; Design a three-layer BP neural network structure based on the feature importance ranking table and normalized feature data set; Construct training and validation datasets based on historical ventilation operation records and normalized feature datasets; The training and validation data sets are used to train and optimize the three-layer BP neural network structure to obtain a training complete model; The trained model is used to perform real-time prediction and short-term trend prediction to obtain the overall air demand forecast value.

6. The method for predicting tunnel traffic volume and required air volume based on BP neural network according to claim 1, characterized in that: The optimized distribution of required air volume by layer in step S3 is specifically as follows: Set up a stratified quality target table based on the vertical stratified pollutant load index; According to the layered ventilation efficiency coefficient, the difference in ventilation efficiency between layers is analyzed to obtain the efficiency difference table between layers; The preliminary allocation coefficient table is calculated based on the inter-layer efficiency difference table and the layer quality target table; Determine the technical constraint condition set of the tunnel ventilation system based on the predicted value of the total air demand; The hierarchical optimized air volume table is calculated based on the preliminary allocation coefficient table, the technical constraint condition set and the overall air volume forecast value; Perform ventilation mode parameter conversion on the stratified optimized air volume table to obtain the equipment parameter solution table; Estimate the stratified air volume effect based on the equipment parameter plan table and stratified ventilation efficiency coefficient table to obtain the expected effect evaluation table; Integrate the stratified optimized air volume table, equipment parameter plan table and expected effect evaluation table to construct a stratified air volume requirement matrix.

7. The method for predicting tunnel traffic volume and required air volume based on BP neural network according to claim 1, characterized in that: Step S4 is specifically as follows: Step S41: Obtain and evaluate equipment capacity and traffic control range based on tunnel ventilation equipment parameters and traffic management system parameters to obtain a collaborative control feasible domain diagram; Step S42: constructing a three-dimensional evaluation system of energy consumption, air quality, and traffic efficiency based on the collaborative control feasible region diagram and the layered air demand matrix; Step S43: Generate and evaluate collaborative control strategies based on the three-dimensional evaluation system and the collaborative control feasible region diagram to obtain candidate control strategy solutions; Step S44: Dynamically optimize and execute the control strategy candidate solutions to obtain a bidirectional control execution solution.

8. The method for predicting tunnel traffic volume and required air volume based on BP neural network according to claim 7, characterized in that: The generation and evaluation of the collaborative control strategy in step S4 is specifically as follows: Generate control parameter combinations based on the collaborative control feasible domain diagram; Calculate the air quality impact of the control parameter combination to obtain an air quality impact table; Conduct energy consumption benefit analysis on the control parameter combination to obtain an energy consumption analysis table; Carry out traffic efficiency evaluation on the control parameter combination to obtain the traffic impact evaluation table; Calculate the comprehensive benefit index table for the air quality impact table, energy consumption analysis table, and traffic impact assessment table based on the three-dimensional evaluation system; The comprehensive benefit index table is used to evaluate and screen the control parameter combinations to obtain candidate control strategy solutions.

Images