Dynamic Control Method and System for Energy Conservation Optimization in Tunnels

Abstract

This invention relates to the field of tunnel energy consumption optimization technology, and particularly to a dynamic control method and system for tunnel energy conservation optimization. The dynamic control method includes: dividing the tunnel into m segments based on tunnel geometric parameters and traffic flow characteristics; calculating the pollutant concentration of the next segment at the next time step; inputting the characteristic data and pollutant concentration of the target segment into a pre-constructed airflow prediction model; outputting the predicted total airflow; and generating fan control commands based on the predicted total airflow. The dynamic control method for tunnel energy conservation optimization provided by this invention intelligently generates control strategies based on predicted airflow and fan capacity. Compared with the traditional "timed and quantitative, full-area air supply" mode, this method achieves "on-demand air supply and precise control" through data-driven approaches, avoiding ineffective ventilation and energy waste in low-density pollutant areas, greatly improving ventilation efficiency, and significantly reducing energy consumption and fan hardware redundancy configuration costs.

Description

Technical Field

[0001] This invention relates to the field of tunnel energy consumption optimization technology, specifically to a dynamic control method and system for tunnel energy conservation optimization. Background Technology

[0002] In existing tunnel air purification systems, ventilation volume control mainly relies on sensor-based air quality monitoring systems. Specifically, a multi-point distributed monitoring network is deployed along the tunnel's longitudinal direction to collect real-time concentration data of key pollutants such as CO, NOx, and PM2.5. Theoretical dilution airflow is calculated using a pollutant diffusion model, which in turn controls axial flow fans for forced ventilation.

[0003] Chinese Patent CN118070569B discloses a comprehensive optimization system and method for ventilation calculation in high-altitude tunnels, aiming to improve the efficiency and performance of ventilation systems. This method integrates empirical formulas and numerical simulations to evaluate and optimize the design of the ventilation system by calculating parameters such as ventilation volume, wind speed, and wind pressure. First, based on the tunnel's characteristics and actual needs, a suitable calculation method is selected, and the calculation results from empirical formulas and numerical simulations are integrated. Second, ventilation parameters, including ventilation volume and wind speed, are calculated to assess whether the ventilation effect meets requirements. Finally, based on the evaluation results, the calculation parameters are adjusted to optimize the ventilation scheme. This invention's method can more accurately calculate ventilation parameters, improve the efficiency of the ventilation system, reduce energy consumption, and achieve the goal of energy conservation and emission reduction.

[0004] The existing technologies described above have the following problems when in use: First, the sensor units that collect data are limited by their inherent accuracy and are subject to interference from complex environments such as high humidity and turbulent airflow in high-altitude tunnels, resulting in insufficient reliability of the monitoring data; Second, in order to achieve effective coverage, a set of monitoring terminals needs to be deployed every 200-300 meters, which will result in high costs for hardware procurement, wiring installation, and subsequent maintenance and use in long tunnel scenarios; Furthermore, the excessive number of sensor units will generate excessive energy consumption, thus causing the existing systems to generally suffer from excessive energy consumption and energy waste. Summary of the Invention

[0005] To address the aforementioned problems, this invention provides a dynamic control method for optimizing tunnel energy conservation.

[0006] The present invention adopts the following technical solution:

[0007] Dynamic control methods for tunnel energy conservation optimization include:

[0008] Step 1: Based on the tunnel's geometric parameters and traffic flow characteristics, an improved K-means clustering algorithm is used to dynamically divide the tunnel into m segments based on the segment length.

[0009] Step 2: Obtain the migration time of pollutants from one segment to the next;

[0010] Step 3: Based on the migration time and mass conservation of pollutants, calculate the pollutant concentration at the next moment in the next segment;

[0011] Step 4: Based on the obtained pollutant concentration of the next segment at the next time moment, mark the next segment as the target segment, obtain the feature data of the target segment, input the feature data and pollutant concentration of the target segment into the pre-built air volume prediction model, and output the predicted total air volume used to dilute to a safe concentration.

[0012] Step 5: Generate fan control commands based on the predicted pollutant concentration in the target section and the predicted total air volume.

[0013] As a further description of the above technical solution: the method for dividing the tunnel into m segments includes:

[0014] Collect tunnel geometric parameters, including the total length and slope of the tunnel; collect tunnel traffic flow characteristics, including average vehicle speed variation, vehicle density, and vehicle type ratio within the tunnel.

[0015] Preset vehicle speed gradient thresholds, classify traffic conditions into congested, smooth, and transitional based on vehicle speed, and initially set the segment length based on the traffic conditions;

[0016] The initial section length is corrected based on the tunnel slope, a slope correction factor is defined, and the corrected section length is obtained based on the slope correction factor.

[0017] Based on the corrected segment length, the dividing points between segments are obtained, and the improved K-means clustering algorithm is used to dynamically divide the segment length.

[0018] As a further description of the above technical solution: the vehicle type ratio information includes the proportion of gasoline vehicles, electric vehicles, and diesel vehicles. The traffic flow density information and vehicle type ratio information are obtained by collecting the passing frequency, speed, and vehicle type information of each vehicle in real time through the ETC system in the tunnel and calculating them.

[0019] As a further description of the above technical solution: the method of obtaining the division points between segments based on the corrected segment length and dynamically dividing the segment length using the improved K-means clustering algorithm includes:

[0020] Based on the corrected segment length, feature vectors are extracted for each division point. The feature vectors include: average vehicle speed, traffic density, proportion of gasoline vehicles, proportion of electric vehicles, and proportion of diesel vehicles. According to the importance of their impact on pollutant diffusion, the weights of each feature vector are set, and the weighted Euclidean distance between two division points is calculated.

[0021] Divide the total length of the tunnel by the corrected segment length and round to the nearest integer to obtain the initial number of clusters K. Randomly select K from the partition points as initial cluster centers. For each partition point, calculate its weighted Euclidean distance to each initial cluster center and assign each partition point to the nearest cluster center. For each cluster, calculate the mean of the features of all its member points as the new cluster center position. If the change of all cluster centers is less than a set threshold, or the maximum number of iterations is reached, the iteration is terminated.

[0022] Based on the obtained cluster centers as the dividing points, the tunnel is divided into m segments. The minimum length of each segment is preset. When the length of a segment after division is less than the minimum length, a segment merging operation is performed to merge it into an adjacent segment.

[0023] As a further description of the above technical solution: the characteristic data of the target section includes the length and slope of the target section, which are obtained through tunnel BIM model or CAD design drawings.

[0024] As a further description of the above technical solution: the training method of the air volume prediction model includes:

[0025] Acquire historical purification data of the tunnel, which includes tunnel characteristic data and pollutant concentration data, as well as the actual total air volume diluted to a safe concentration.

[0026] Perform data cleaning and standardization on historical purification data;

[0027] The collected historical purification data is converted into a corresponding set of feature vectors;

[0028] Each set of feature vectors is used as input to the air volume prediction model. The air volume prediction model outputs the predicted total air volume corresponding to the feature data and pollutant concentration data of each set of tunnels, and uses the actual total air volume of the feature data and pollutant concentration data of each set of tunnels as the prediction target. The training objective is to minimize the loss function value of the air volume prediction model. Training stops when the loss function value of the air volume prediction model is less than or equal to the preset target loss value.

[0029] As a further description of the above technical solution: the method for generating the wind turbine control command includes:

[0030] Obtain the rated air volume of each fan in the target section, divide the predicted total air volume by the rated air volume and round up to obtain the number of fans that are turned on, and mark it as the target number of fans.

[0031] Obtain road segment feature data of the target section, divide the target section into pollutant hot spot areas and regular areas based on the road segment feature data, and sort the pollutant hot spot areas according to the degree of pollutant aggregation.

[0032] According to the pollutant aggregation degree from large to small, the smallest closed fan cluster surrounding the pollutant hot spot area is activated in sequence to form a local high-speed airflow to scour the tunnel section and concentrate the pollutants to be discharged.

[0033] Once the number of activated fans reaches the target number, fan activation stops. If the number of activated fans does not reach the target number, a preset spacing length threshold is used to obtain the length range of unactivated fans in the normal area. If the preset spacing length threshold is exceeded, a group of fans is activated in the middle of the target section until the number of activated fans reaches the target number.

[0034] As a further description of the above technical solution: the method for obtaining road segment feature data of the target segment and dividing the target segment into pollutant hotspot areas and conventional areas based on the road segment feature data includes:

[0035] Based on the changes in slope and curvature of the target section, the target section is divided into multiple sub-regions;

[0036] Areas with increased slope, areas with changes in curvature, or a combination of both are designated as pollutant hotspot areas.

[0037] Areas with reduced slope and level road sections are marked as regular areas.

[0038] As a further description of the above technical solution: the method for sorting pollutant hotspot regions according to the degree of pollutant aggregation includes:

[0039] The characteristic data of each pollutant hot spot region are obtained sequentially. The characteristic data of the pollutant hot spot region includes the length, slope and curvature radius of the pollutant hot spot region.

[0040] The characteristic data of the hot spot region of pollutants are calculated using formulas to obtain the pollutant aggregation coefficient;

[0041] Based on the magnitude of the pollutant convergence coefficient, each pollutant hotspot region is sorted from largest to smallest.

[0042] A tunnel energy-saving optimization dynamic control system, used to implement the aforementioned tunnel energy-saving optimization dynamic control method, includes:

[0043] The segment division module, based on tunnel geometric parameters and traffic flow characteristics, uses an improved K-means clustering algorithm to dynamically divide the segment length, dividing the tunnel into m segments;

[0044] The time acquisition module acquires the migration time taken for pollutants to migrate from one segment to the next.

[0045] The concentration prediction module calculates the pollutant concentration in the next segment at the next moment based on the pollutant's migration time and mass conservation.

[0046] The air volume prediction module, based on the pollutant concentration of the next segment at the next moment, marks the next segment as the target segment, obtains the feature data of the target segment, inputs the feature data of the target segment and the pollutant concentration into the pre-built air volume prediction model, and outputs the predicted total air volume used to dilute to a safe concentration.

[0047] The instruction generation module generates fan control instructions based on the predicted pollutant concentration in the target section and the predicted total air volume.

[0048] Beneficial effects:

[0049] The tunnel energy-saving optimization dynamic control method provided by this invention introduces traffic flow characteristics and tunnel geometric parameters, and uses an improved K-means clustering algorithm to realize dynamic segmentation of the tunnel, making the regional division more in line with the characteristics of pollutant accumulation; by calculating the migration time of pollutants between different segments, and on this basis, combined with mass conservation, it realizes accurate prediction of the pollutant concentration of the next segment at the next time.

[0050] Furthermore, by inputting pollutant concentration and road segment characteristic data into the airflow prediction model, the model dynamically outputs the predicted total airflow required to reach a safe concentration. Finally, a control strategy is intelligently generated based on the predicted total airflow and fan capacity. Compared to the traditional "timed and quantitative, full-area air supply" mode, this method achieves "on-demand air supply and precise control" through data-driven approaches. This avoids ineffective ventilation and energy waste in low-density pollutant areas, significantly improving ventilation efficiency, significantly reducing energy consumption and fan hardware redundancy configuration costs, and eliminating the need to deploy numerous sensors in the tunnel for pollutant concentration collection, thus reducing equipment setup and energy costs. Optionally, during use, several segments can be randomly selected to set up sensor groups for pollutant concentration collection and verification.

[0051] Finally, by combining the characteristics of road sections such as slope, curvature, and length, a minimum closed fan cluster is formed to carry out directional flushing of areas with high pollutant concentration. While achieving the air purification goal, the efficiency of fan use is significantly improved. Compared with the linear opening mode of equally spaced fans commonly used in existing technologies, traditional solutions lack the ability to perceive the distribution of pollutants. They often over-supply air in non-critical areas, while under-supply air in key areas such as uphill sections and curves where pollutants are prone to accumulate. This leads to the accumulation and slow diffusion of pollutants, resulting in problems such as purification blind spots and energy waste. Attached Figure Description

[0052] The present invention will be further explained below with reference to the accompanying drawings and embodiments:

[0053] Figure 1 The flowchart is a dynamic control method for tunnel energy saving optimization provided in Embodiment 1 of the present invention;

[0054] Figure 2 A flowchart illustrating the method for generating wind turbine control commands provided in Embodiment 2 of the present invention;

[0055] Figure 3 This is a module connection diagram of the tunnel energy-saving optimization dynamic control system provided in Embodiment 3 of the present invention;

[0056] Figure 4 The flowchart is shown for the tunnel energy-saving optimization dynamic control method provided in Embodiment 1 of the present invention. Detailed Implementation

[0057] To make the technical means, creative features, objectives, and effects of this invention readily understandable, the invention is further described below with reference to specific illustrations. It should be noted that, unless otherwise specified, the embodiments and features described in these embodiments can be combined with each other.

[0058] Example 1:

[0059] Please see Figure 1 and Figure 4 This invention provides a technical solution: a dynamic control method for optimizing tunnel energy conservation, specifically including:

[0060] Step 1: Based on the tunnel's geometric parameters (length, gradient) and traffic flow characteristics (average vehicle speed, vehicle type ratio, traffic density), an improved K-means clustering algorithm is used to dynamically divide the tunnel into m segments.

[0061] Specifically, methods for dividing a tunnel into m segments include:

[0062] Collect tunnel geometric parameters, including the total length and slope of the tunnel; collect tunnel traffic flow characteristics, including average vehicle speed variation, traffic density, and vehicle type ratio within the tunnel; the vehicle type ratio includes the proportion of gasoline vehicles, electric vehicles, and diesel vehicles.

[0063] It should be noted that the traffic density information and vehicle type ratio information can be collected in real time through the ETC system (electronic toll collection system) inside the tunnel. This system can collect information on the frequency of passage, speed, and vehicle type (such as the proportion of passenger cars, trucks, and diesel vehicles) of each vehicle, and further accurately estimate the amount of traffic density information and the proportion of diesel vehicles.

[0064] Preset vehicle speed gradient thresholds, classify traffic conditions into congested, smooth, and transitional based on vehicle speed, and initially set the segment length based on the traffic conditions;

[0065] The initial section length is adjusted based on the tunnel slope, and a slope correction factor is defined. The corrected segment length is obtained based on the slope correction factor;

[0066] The slope correction factor The calculation method is as follows: In the formula, For weighting coefficients, the preferred ones are... Take 0.2; It is the tangent of the tunnel slope angle θ (dimensionless). It directly reflects the slope of the tunnel;

[0067] The formula for calculating the corrected segment length based on the slope correction factor is as follows: In the formula, For the corrected segment length, The initial segment length, This is the slope correction factor.

[0068] It should be noted that tunnel gradient affects pollutant retention and vehicle speed. Gradient correction is used to dynamically adjust the section length based on the impact of tunnel gradient on vehicle driving conditions and pollutant retention characteristics. For example, when going uphill, vehicles decelerate and emissions increase, making it easier for pollutants to accumulate. Therefore, it is necessary to shorten the section to improve monitoring and prediction accuracy. When going downhill, vehicles accelerate and pollutants diffuse more quickly. The section length can be appropriately extended, and the section length can be linearly adjusted with the gradient by introducing a gradient correction factor.

[0069] Based on the corrected segment length, the division points between segments are obtained, and the improved K-means clustering algorithm is used to dynamically divide the segment length;

[0070] Based on the corrected segment length, the dividing points between segments are obtained. An improved K-means clustering algorithm is used. Methods for dynamically dividing segment lengths include:

[0071] Based on the corrected segment length, feature vectors are extracted for each division point. The feature vectors include: average vehicle speed, traffic density, proportion of gasoline vehicles, proportion of electric vehicles, and proportion of diesel vehicles. According to the importance of their impact on pollutant diffusion, the weights of each feature vector are set, and the weighted Euclidean distance between two division points is calculated.

[0072] The formula for calculating weighted Euclidean distance is: In the formula, To represent the dividing point and The weighted Euclidean distance between the sampling points represents the distance between the sampling points. and In the The values ​​taken on each feature dimension Indicates the first The weights of each feature;

[0073] Divide the total length of the tunnel by the corrected segment length and round to the nearest integer to obtain the initial number of clusters K. Randomly select K from the partition points as initial cluster centers. For each partition point, calculate its weighted Euclidean distance to each initial cluster center and assign each partition point to the nearest cluster center. For each cluster, calculate the mean of the features of all its member points as the new cluster center position. If the change of all cluster centers is less than a set threshold, or the maximum number of iterations is reached, the iteration is terminated.

[0074] Based on the obtained cluster centers as the dividing points, the tunnel is divided into m segments. The minimum length of each segment is preset. When the length of a segment after division is less than the minimum length, a segment merging operation is performed to merge it into an adjacent segment.

[0075] In this embodiment, the segmentation method is based on weighted Euclidean distance and dynamic clustering, which can more accurately reflect the differences in traffic conditions inside the tunnel and avoid the coarseness of traditional fixed-length segmentation. By assigning different weights to different indicators (such as vehicle speed, density, and vehicle type ratio), the key factors affecting pollutant diffusion are meticulously captured, improving the accuracy of pollutant concentration prediction. At the same time, a slope correction factor is introduced to consider the impact of terrain changes on pollutant retention and vehicle speed, making the segmentation more consistent with the actual operating characteristics of the tunnel. The number of clusters is dynamically set, reducing subjective human intervention and achieving a more intelligent segmentation strategy.

[0076] Step 2: Obtain the migration time of pollutants from one segment to the next;

[0077] Step 3: Based on the migration time and mass conservation of pollutants, calculate the pollutant concentration at the next moment in the next segment;

[0078] The formula for calculating the migration time is as follows: In the formula, The migration time (in seconds) for a pollutant to travel from one section to the next. The distance between the two sections (unit: meters). The airflow velocity between the two sections includes natural wind and forced ventilation (unit: meters per second). It is 10% of the section length (unit: meters). The diffusion coefficient of pollutants is used to represent the ability of pollutants to diffuse naturally in the air (unit: m / s), and can be dynamically estimated by a turbulence model (k-ε model);

[0079] in, Represents the time that has passed with the wind. This represents the time it takes for pollutants to "slowly drift away" due to diffusion. Combined, these two values ​​represent the total time required for the pollutants to migrate.

[0080] Methods for calculating the pollutant concentration at the next time step in the next segment include:

[0081] ;

[0082] Current time Next section Pollutant concentration (unit: ppm) The dilution attenuation coefficient (dimensionless, value between 0 and 1) represents the proportion of pollutants that are diluted and dispersed during movement. It is set by human experience, with a preferred value of 0.3. For the previous segment p in time Delayed pollutant concentration (unit: ppm) Emission factor, describing the amount of pollutants produced per vehicle per second (unit: ppm per vehicle). -1 ·Second -1 ), For time Traffic flow (unit: vehicles / second). This represents the current percentage of diesel vehicles. For diesel vehicle pollution amplification function, optionally, This indicates that diesel vehicles pollute more than regular vehicles; the higher the proportion of diesel vehicles, the heavier the pollution.

[0083] In the formula, It reflects pollution transmitted from the previous section, but takes into account intermediate losses. This represents the pollutants emitted by newly added vehicles in this section.

[0084] In this embodiment, the concentration of pollutants in the next section is predicted based on the concentration of pollutants in the previous section. This reduces the need for setting up pollution sensors in the tunnel. Only a few sections need to be set up for verification to collect and verify the pollutant concentration. Furthermore, this predictive method can identify high-risk sections in time before pollutants exceed the standard, and take measures such as ventilation and jet intervention in advance to avoid passive response and improve safety. According to the predicted pollution diffusion trend, the start and stop of fans, air volume, and jet direction can be dynamically optimized to ensure air quality, reduce the number of sensors to collect data, and reduce energy consumption and operating costs.

[0085] Step 4: Based on the obtained pollutant concentration of the next segment at the next time moment, mark the next segment as the target segment, obtain the feature data of the target segment, input the feature data and pollutant concentration of the target segment into the pre-built air volume prediction model, and output the predicted total air volume used to dilute to a safe concentration.

[0086] The characteristic data of the target section include the length and slope of the target section;

[0087] The training method for the air volume prediction model includes:

[0088] Acquire historical purification data of the tunnel, which includes tunnel characteristic data and pollutant concentration data, as well as the actual total air volume diluted to a safe concentration.

[0089] It should be noted that the pollutant concentration collection method involves rationally arranging multiple pollutant concentration sensors within the tunnel, eliminating outliers, and obtaining the average value as the pollutant concentration.

[0090] Perform data cleaning and standardization on historical purification data;

[0091] The data cleaning process includes checking the collected data for missing values, outliers, etc. For missing values, interpolation methods (such as linear interpolation and spline interpolation) can be used to fill them in; for outliers, statistical methods (such as the 3σ principle) can be used to identify and process them.

[0092] The standardization process includes: standardizing the input features of tunnel length, slope, and pollutant concentration data, converting them into a distribution with a mean of 0 and a standard deviation of 1, using the Z-score standardization method.

[0093] The collected historical purification data is converted into a corresponding set of feature vectors;

[0094] Each set of feature vectors is used as the input to the air volume prediction model. The air volume prediction model outputs the predicted total air volume corresponding to the feature data and pollutant concentration data of each set of tunnels, and uses the actual total air volume of the feature data and pollutant concentration data of each set of tunnels as the prediction target. The training objective is to minimize the loss function value of the air volume prediction model. Training stops when the loss function value of the air volume prediction model is less than or equal to the preset target loss value.

[0095] The loss function value of the air volume prediction model is the mean square error. Preferably, the air volume prediction model is a neural network model, such as a multilayer perceptron (MLP). MLP has a strong nonlinear fitting capability and can handle the complex nonlinear relationship between target section feature data, pollutant concentration and required air volume.

[0096] Other model parameters of the wind volume prediction model, including the target loss value, optimization algorithm, ratio of training set, test set and validation set, and optimization of loss function, are all obtained through actual engineering implementation and continuous experimental tuning.

[0097] Step 5: Generate fan control commands based on the predicted pollutant concentration in the target section and the predicted total air volume.

[0098] The method for generating fan control commands includes: obtaining the rated air volume of each fan in the target section, dividing the predicted total air volume by the rated air volume and rounding up to obtain the number of fans to be turned on, marking it as the target number of fans, and turning on the fans at equal intervals in the target section so that the number of fans reaches the target number.

[0099] In this embodiment, by introducing traffic flow characteristics and tunnel geometric parameters, an improved K-means clustering algorithm is used to dynamically segment the tunnel, making the regional division more closely match the characteristics of pollutant accumulation. Furthermore, by calculating the migration time of pollutants between different segments, and by inputting the pollutant concentration of the target segment and the segment's characteristic data into the airflow prediction model, the predicted total airflow required to reach a safe concentration is dynamically output. Finally, a control strategy is intelligently generated based on the predicted total airflow and fan capacity. Compared with the traditional "timed and quantitative, full-area air supply" mode, this method achieves "on-demand air supply and precise control" through data-driven approaches, avoiding ineffective ventilation and energy waste in low-density pollutant areas, greatly improving ventilation efficiency, significantly reducing energy consumption and fan hardware redundancy configuration costs, and eliminating the need to deploy a large number of sensors in the tunnel for pollutant concentration collection, thus reducing equipment setup and energy costs. Optionally, during use, several segments can be randomly selected to set up sensor groups for pollutant concentration collection and verification.

[0100] Example 2:

[0101] Please see Figure 2 This invention provides another method for generating fan control commands;

[0102] The method for generating the wind turbine control commands can also be:

[0103] Obtain the rated air volume of each fan in the target section, divide the predicted total air volume by the rated air volume and round up to obtain the number of fans that need to be turned on, and mark it as the target number of fans.

[0104] Obtain road segment feature data of the target section, divide the target section into pollutant hot spot areas and regular areas based on the road segment feature data, and sort the pollutant hot spot areas according to the degree of pollutant aggregation.

[0105] According to the pollutant aggregation degree from large to small, the smallest closed fan cluster surrounding the pollutant hot spot area is activated in sequence to form a local high-speed airflow to scour the tunnel section and concentrate the pollutants to be discharged.

[0106] For example, if a pollutant hotspot covers a 20-meter area, only three nearby fans need to be activated, rather than activating the entire fan linearly.

[0107] Once the number of activated fans reaches the target number, fan activation stops. Conversely, if the number of activated fans does not reach the target number, a preset spacing length threshold is used to obtain the length range of unactivated fans in the normal area. When the preset spacing length threshold is exceeded, a group of fans is activated in the middle of the target segment until the number of activated fans reaches the target number.

[0108] Methods for obtaining road segment feature data of the target section and dividing the target section into pollutant hotspot areas and regular areas based on the road segment feature data include:

[0109] Based on the changes in slope and curvature of the target section, the target section is divided into multiple sub-regions;

[0110] Areas with increased slope, areas with changes in curvature, or a combination of both are designated as pollutant hotspot areas.

[0111] Areas with reduced slope and level road sections are marked as regular areas.

[0112] Inside the tunnel, the distribution and accumulation of pollutants are influenced by a variety of factors, and they are usually concentrated in specific areas, including uphill areas and curved areas.

[0113] In uphill areas, the engine load of vehicles (especially heavy-duty diesel vehicles) increases significantly, resulting in a substantial increase in exhaust emissions (CO emissions can be 2-3 times higher than on flat roads). Hot flue gas flows upwards due to buoyancy, overlapping with natural ventilation, and wind speeds at slope changes easily create eddies, hindering pollutant diffusion.

[0114] In the bend and curve area, the airflow forms a secondary circulation due to centrifugal force at the bend, and a low-speed zone appears near the outer wall (the wind speed can be reduced by 40%), which causes pollutants to be trapped. The smaller the bend radius, the more obvious the accumulation of pollutants.

[0115] Methods for sorting pollutant hotspot regions according to the degree of pollutant accumulation include:

[0116] The characteristic data of each pollutant hot spot region are obtained sequentially. The characteristic data of the pollutant hot spot region includes the length, slope and curvature radius of the pollutant hot spot region.

[0117] The characteristic data of the hot spot region of pollutants are calculated using formulas to obtain the pollutant aggregation coefficient;

[0118] Based on the magnitude of the pollutant convergence coefficient, each pollutant hotspot region is sorted from largest to smallest.

[0119] The formula for calculating the pollutant aggregation coefficient is as follows: In the formula, Here, the pollutant convergence coefficient is... The larger the size, the more pollutants accumulate. This represents the length of the pollutant hotspot region. The longer the pollutant hotspot region, the more pollutants accumulate. The slope is considered the gradient; the steeper the slope, the more pollutants accumulate. The radius of curvature is the bending radius; the smaller the bending radius, the more pollutants accumulate. , , These are the weighting coefficients, and , and All are greater than 0.

[0120] It should be noted that the formulas mentioned above are all dimensionless calculations, derived from software simulation using a large amount of data to approximate the real situation. The weighting coefficients and preset thresholds in the formulas are set by those skilled in the art based on actual conditions or obtained through large-scale data simulation. The magnitude of the weighting coefficient is a specific value obtained by quantifying each parameter to facilitate subsequent comparison. The magnitude of the weighting coefficient depends on the amount of sample data and the processing coefficients initially set by those skilled in the art for each set of sample data; as long as it does not affect the proportional relationship between the parameter and the quantified value, it is acceptable.

[0121] In this embodiment, by combining the characteristics of road sections such as slope, curvature, and length, a minimal closed-loop fan cluster is formed to directionally flush areas with high pollutant concentration. This significantly improves fan efficiency while achieving air purification goals. Compared to the linear operation mode of equally spaced fans commonly used in existing technologies, traditional solutions lack the ability to perceive pollutant distribution, often over-ventilating non-critical areas while under-ventilating key areas such as uphill sections and curves where pollutants tend to accumulate. This leads to pollutant accumulation, slow diffusion, purification blind spots, and energy waste. This technology can concentrate small-scale, high-intensity airflow in areas with high pollutant density. Compared to traditional large-scale, low-intensity dilution methods, the purification effect is more significant, the response is more timely, and the overall system energy consumption is lower. By dynamically optimizing the fan operation position and number, "precise airflow and targeted purification" are achieved, effectively improving the intelligence and precision of tunnel air quality management.

[0122] Example 3:

[0123] Please see Figure 3 This invention provides a tunnel energy-saving optimization dynamic control system for implementing the aforementioned tunnel energy-saving optimization dynamic control method, comprising:

[0124] The segment division module, based on tunnel geometric parameters and traffic flow characteristics, uses an improved K-means clustering algorithm to dynamically divide the segment length, dividing the tunnel into m segments;

[0125] The time acquisition module acquires the migration time taken for pollutants to migrate from one segment to the next.

[0126] The concentration prediction module calculates the pollutant concentration in the next segment at the next moment based on the pollutant's migration time and mass conservation.

[0127] The air volume prediction module, based on the pollutant concentration of the next segment at the next moment, marks the next segment as the target segment, obtains the feature data of the target segment, inputs the feature data of the target segment and the pollutant concentration into the pre-built air volume prediction model, and outputs the predicted total air volume used to dilute to a safe concentration.

[0128] The instruction generation module generates fan control instructions based on the predicted pollutant concentration in the target section and the predicted total air volume.

[0129] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of this invention is defined by the appended claims and their equivalents.

Claims

1. A dynamic control method for optimizing tunnel energy conservation, characterized in that, include: Step 1: Based on the tunnel's geometric parameters and traffic flow characteristics, an improved K-means clustering algorithm is used to dynamically divide the tunnel into m segments based on the segment length. Step 2: Obtain the migration time of pollutants from one segment to the next; Step 3: Based on the migration time and mass conservation of pollutants, calculate the pollutant concentration at the next moment in the next segment; Step 4: Based on the obtained pollutant concentration of the next segment at the next time moment, mark the next segment as the target segment, obtain the feature data of the target segment, input the feature data and pollutant concentration of the target segment into the pre-built air volume prediction model, and output the predicted total air volume used to dilute to a safe concentration. Step 5: Generate fan control commands based on the predicted pollutant concentration in the target section and the predicted total air volume; The method for dividing the tunnel into m segments includes: Collect tunnel geometric parameters, including the total length and slope of the tunnel; collect tunnel traffic flow characteristics, including average vehicle speed variation, vehicle density, and vehicle type ratio within the tunnel. Preset vehicle speed gradient thresholds, classify traffic conditions into congested, smooth, and transitional based on vehicle speed, and initially set the segment length based on the traffic conditions; The initial section length is corrected based on the tunnel slope, a slope correction factor is defined, and the corrected section length is obtained based on the slope correction factor. Based on the corrected segment length, the division points between segments are obtained, and the improved K-means clustering algorithm is used to dynamically divide the segment length; The method for dynamically dividing segment lengths based on the corrected segment lengths, obtaining the dividing points between segments, and using an improved K-means clustering algorithm includes: Based on the corrected segment length, feature vectors are extracted for each division point. The feature vectors include: average vehicle speed, traffic density, proportion of gasoline vehicles, proportion of electric vehicles, and proportion of diesel vehicles. According to the importance of their impact on pollutant diffusion, the weights of each feature vector are set, and the weighted Euclidean distance between two division points is calculated. Divide the total length of the tunnel by the corrected segment length and round to the nearest integer to obtain the initial number of clusters K. Randomly select K from the partition points as initial cluster centers. For each partition point, calculate its weighted Euclidean distance to each initial cluster center and assign each partition point to the nearest cluster center. For each cluster, calculate the mean of the features of all its member points and use it as the new cluster center position. If the change of all cluster centers is less than a set threshold, or the maximum number of iterations is reached, the iteration is terminated. Based on the obtained cluster centers as the dividing points, the tunnel is divided into m segments. The minimum length of each segment is preset. When the length of a segment after division is less than the minimum length, a segment merging operation is performed to merge it into an adjacent segment. The method for generating the wind turbine control commands includes: Obtain the rated air volume of each fan in the target section, divide the predicted total air volume by the rated air volume and round up to obtain the number of fans that are turned on, and mark it as the target number of fans. Obtain road segment feature data of the target section, divide the target section into pollutant hot spot areas and regular areas based on the road segment feature data, and sort the pollutant hot spot areas according to the degree of pollutant aggregation. According to the pollutant aggregation degree from large to small, the smallest closed fan cluster surrounding the pollutant hot spot area is activated in sequence to form a local high-speed airflow to scour the tunnel section and concentrate the pollutants to be discharged. Once the number of activated fans reaches the target number, fan activation stops. If the number of activated fans does not reach the target number, a preset spacing length threshold is used to obtain the length range of unactivated fans in the normal area. If the preset spacing length threshold is exceeded, a group of fans is activated in the middle of the target section until the number of activated fans reaches the target number.

2. The tunnel energy-saving optimization dynamic control method according to claim 1, characterized in that, The vehicle type ratio information includes the proportion of gasoline vehicles, electric vehicles, and diesel vehicles. The traffic flow density information and vehicle type ratio information are obtained by collecting the passing frequency, speed, and vehicle type information of each vehicle in real time through the ETC system in the tunnel and calculating them.

3. The tunnel energy-saving optimization dynamic control method according to claim 1, characterized in that, The characteristic data of the target section includes the length and slope of the target section, which are obtained through tunnel BIM model or CAD design drawings.

4. The tunnel energy-saving optimization dynamic control method according to claim 3, characterized in that, The training method for the air volume prediction model includes: Acquire historical purification data of the tunnel, which includes tunnel characteristic data and pollutant concentration data, as well as the actual total air volume diluted to a safe concentration. Perform data cleaning and standardization on historical purification data; The collected historical purification data is converted into a corresponding set of feature vectors; Each set of feature vectors is used as input to the air volume prediction model. The air volume prediction model outputs the predicted total air volume corresponding to the feature data and pollutant concentration data of each set of tunnels, and uses the actual total air volume of the feature data and pollutant concentration data of each set of tunnels as the prediction target. The training objective is to minimize the loss function value of the air volume prediction model. Training stops when the loss function value of the air volume prediction model is less than or equal to the preset target loss value.

5. The tunnel energy-saving optimization dynamic control method according to claim 1, characterized in that, The method for obtaining road segment feature data of the target segment and dividing the target segment into pollutant hotspot areas and conventional areas based on the road segment feature data includes: Based on the changes in slope and curvature of the target section, the target section is divided into multiple sub-regions; Areas with increased slope, areas with changes in curvature, or a combination of both are designated as pollutant hotspot areas. Areas with reduced slope and level road sections are marked as regular areas.

6. The tunnel energy-saving optimization dynamic control method according to claim 5, characterized in that, The method for sorting pollutant hot spot regions according to the degree of pollutant aggregation includes: The characteristic data of each pollutant hot spot region are obtained sequentially. The characteristic data of the pollutant hot spot region includes the length, slope and curvature radius of the pollutant hot spot region. The characteristic data of the hot spot region of pollutants are calculated using formulas to obtain the pollutant aggregation coefficient; Based on the magnitude of the pollutant convergence coefficient, each pollutant hotspot region is sorted from largest to smallest.

7. A tunnel energy-saving optimization dynamic control system, used to implement the tunnel energy-saving optimization dynamic control method according to any one of claims 1-6, characterized in that, include: The segment division module, based on tunnel geometric parameters and traffic flow characteristics, uses an improved K-means clustering algorithm to dynamically divide the segment length, dividing the tunnel into m segments; The time acquisition module acquires the migration time taken for pollutants to migrate from one segment to the next. The concentration prediction module calculates the pollutant concentration in the next segment at the next moment based on the pollutant's migration time and mass conservation. The air volume prediction module, based on the pollutant concentration of the next segment at the next moment, marks the next segment as the target segment, obtains the feature data of the target segment, inputs the feature data of the target segment and the pollutant concentration into the pre-built air volume prediction model, and outputs the predicted total air volume used to dilute to a safe concentration. The instruction generation module generates fan control instructions based on the predicted pollutant concentration in the target section and the predicted total air volume.

Images