A service area real-time congestion research method, device, medium and program product

Abstract

The application discloses a service area real-time congestion research and judgment method and device, medium and program product, the method comprises the following steps: acquiring the three core dimensions and auxiliary dimension indexes and associated reference data of the highway service area parking space, toilet position and charging service in real time; after each index is preprocessed by the edge, the associated rule algorithm and the space-time matching algorithm are fused on the fusion analysis platform, and then the congestion research and judgment model optimized by the random forest algorithm is input, the basic index of the prediction is optimized by using the correction coefficient K, and the dynamic congestion index Z and the grade are obtained; when Z exceeds the congestion threshold, the associated reference data is called to locate the congestion cause, and differentiated services are provided to the management end and the driver and passenger end distinguished by vehicle type and power type. The application can realize the fusion analysis of the multi-scene data of the parking space, the toilet position and the charging service, the specific adaptation to the new energy vehicle, the complete data time processing process, the accurate positioning of the congestion cause and the early prediction of the congestion clue.

Description

Technical Field

[0001] This invention relates to the field of intelligent transportation technology, and in particular to a method, equipment, medium, and program product for real-time congestion assessment in service areas. Background Technology

[0002] Service areas, as core hubs connecting transportation networks, serve as mobile rest stops for millions of drivers and passengers, providing them with supplies and rest during their journeys. Their operational efficiency directly impacts the public's travel experience and road network safety. In recent years, with the continuous increase in motor vehicle ownership and the surge in demand for cross-regional tourism and logistics transportation, the oversaturation of service areas has become increasingly prominent. During peak holiday periods, it is common to see vehicles queuing up at the entrances of some service areas and spilling onto the main road. Scenes such as intertwined traffic in parking lots, long queues for restrooms, and a shortage of charging stations have become the norm.

[0003] Such congestion not only significantly reduces service quality but also easily leads to minor accidents, exacerbates road traffic pressure, and even causes public concerns such as energy supply disruptions. Against this backdrop, accurate assessment of service area congestion is crucial to addressing these issues, as its core value is always rooted in the travel experience of drivers and passengers, directly impacting public satisfaction with transportation.

[0004] Currently, service area congestion assessment has gradually moved beyond traditional manual inspections and entered a digital monitoring phase supported by IoT technology. By deploying cameras at service area entrances, the number of vehicles, vehicle types, and entry times are identified in real time, allowing for precise control of vehicle inflow patterns. Sensors such as human body sensors and toilet occupancy sensors are deployed in key service areas like restrooms to dynamically collect data on the number of people using the facilities and their usage status, providing fundamental data support for congestion assessment. Based on this, three typical congestion assessment models have emerged: Parking Space Mode: This mode defines congestion solely by comparing the number of currently parked vehicles with the total number of parking spaces in the service area—when the number of parked vehicles approaches or exceeds the total number of parking spaces, it is considered congested. This method ignores key variables: the parking space requirements of different vehicle types (passenger cars, trucks, RVs) vary significantly, and the intensity of space occupation by vehicles stopping briefly to pick up or drop off passengers differs greatly from that of vehicles resting for extended periods. Furthermore, it fails to differentiate between power types, making it unsuitable for the charging needs of new energy vehicles. This results in the inability to identify resource conflicts between new energy vehicles occupying charging spaces and conventional parking of gasoline vehicles. For example, even if a service area's total number of parking spaces is not saturated, concentrated parking of trucks may lead to a shortage of passenger car spaces, or new energy vehicles may occupy conventional parking spaces while waiting to charge, causing resource misallocation. This mode cannot identify such structural congestion, easily leading to distorted judgments.

[0005] The toilet stall mode is based on the relationship between the current number of users and the total number of toilet stalls. A congestion warning is triggered when the number of users approaches or exceeds the total number of stalls. However, this mode fails to consider dynamic spatiotemporal factors and does not establish a mechanism linking toilet stalls with ancillary facilities such as washbasins and mother-and-baby rooms. This can easily lead to misjudgments of hidden congestion—for example, toilet stalls may not be saturated, but ancillary facilities may be congested, or there may be a momentary surge in people but efficient turnover may prevent actual congestion. This mode's static comparison logic, and its failure to consider the impact of ancillary facilities on the judgment results, makes it difficult to accurately reflect the actual congestion situation in the service area.

[0006] Charging service mode: The core judgment criterion is the difference between the number of occupied charging piles and the total number of charging piles. When the number of occupied charging piles approaches or exceeds the total number, a charging congestion warning is triggered. This mode has two limitations: First, it is a static judgment, focusing only on the current occupancy status and failing to combine charging time and power to predict the changing trend of the charging waiting queue, nor does it consider the differences in charging interface charging, which can easily lead to problems such as improper charging pile interface adaptation and resource mismatch; Second, the data is isolated, not linked with parking space and entry power type data, making it impossible to identify the derivative congestion and distinguish the causes of congestion, resulting in low accuracy of warnings and a lack of targeted diversion measures.

[0007] Based on the practical scenarios and current technological applications of service area congestion monitoring, existing technologies have significant limitations in terms of accuracy, foresight, and comprehensiveness in congestion assessment, specifically in the following four aspects: 1. Limited judgment dimensions, prone to misjudgment by substituting local data for overall congestion: Existing technologies mostly rely on single-scenario data for congestion judgment, failing to cover the coordinated operation characteristics of core service scenarios such as parking, restroom use, and energy replenishment in service areas. Furthermore, the judgment indicators are limited and fail to differentiate between power types, charging parameter differences, and the usage status of auxiliary facilities, which can easily lead to structural congestion misjudgments, resulting in a disconnect between early warnings and actual needs, and failing to support a complete data processing flow.

[0008] 2. Static data processing, lacking dynamic predictive capabilities: Existing technologies use threshold-triggered static judgment logic, which is a post-event discovery mode. It does not integrate historical and real-time data, does not build a dynamic congestion index generation model, and does not form a complete data processing flow. It is unable to predict congestion trends and misses the best intervention opportunity.

[0009] 3. Significant data silos and lack of collaborative analysis capabilities: Data from various dimensions are scattered across different hardware and management modules, lacking a unified fusion analysis platform. The integration of correlation rules and spatiotemporal matching has not been achieved, making it impossible to determine the congestion transmission path, locate the causes of complex congestion, and ensure that the mitigation measures are not targeted enough.

[0010] 4. Incomplete vehicle model identification system and lack of technical implementation details: The system simply distinguishes vehicle models without subdividing the energy replenishment needs of new energy vehicles and does not specify the details of data processing procedures. This results in an inaccurate match between vehicle parking and energy replenishment needs, and the technical solutions lack feasibility.

[0011] In essence, all three types of models are threshold-triggered judgments, which fail to achieve integrated analysis of data from multiple scenarios such as parking spaces, toilets, and charging services. They also lack targeted adaptation for new energy vehicles, have not formed a standardized dynamic congestion index generation mechanism, and do not follow a complete data time processing flow. Therefore, they cannot accurately locate the causes of congestion, let alone predict the signs of congestion in advance. Summary of the Invention

[0012] This invention provides a method, device, medium, and program product for real-time congestion assessment in service areas, which solves the significant limitations of existing technologies in terms of accuracy, foresight, and comprehensiveness of congestion assessment, as well as the core problems of non-standard data processing procedures and lack of power type adaptation.

[0013] In a first aspect, the present invention provides a method for real-time congestion assessment of service areas, comprising: Real-time acquisition of indicators and related reference data for three core dimensions—parking spaces, toilets, and charging services—of highway service areas, as well as auxiliary dimensions. The parking space dimension includes inbound traffic flow rate, outbound traffic flow rate, vehicle type ratio, power type ratio, parking space turnover rate, and parking space occupancy time. The toilet dimension includes toilet usage rate, toilet turnover rate, and toilet passenger density. The charging service dimension includes charging pile usage rate, charging wait time, charging pile occupancy time, and charging pile compatibility rate. The auxiliary dimension includes a comprehensive index K2 integrating weather and holiday indices. The parking space and charging service dimensions are linked by the power type ratio, with related reference data including the site load-bearing pressure index. Utilization rate of ancillary facilities; After each indicator is preprocessed at the edge, it is fused in the fusion analysis platform through association rule algorithm and spatiotemporal matching algorithm, and then input into the congestion judgment model optimized by random forest algorithm. The basic index is optimized by correction coefficient K to obtain dynamic congestion index Z and level; where K=K1×K2, K1 is the ratio of real-time and historical scenario baseline; When Z exceeds the congestion threshold, the relevant reference data is retrieved to locate the cause of congestion, and then differentiated services are provided to the management end and the driver and passenger ends distinguished by vehicle type and power type.

[0014] Optionally, the fusion analysis platform can be used to fuse data through association rule algorithms and spatiotemporal matching algorithms. Specific implementation steps include: Spatial alignment: Establish a 10m×10m unified spatial grid model, associate each monitoring device of parking space, toilet stall, and charging service dimension with a unique grid number, and classify the index data of the same grid into the same spatial dataset. Associate the monitoring devices of toilet stalls with the sinks and mother and baby room ancillary facilities with the same grid to achieve collaborative analysis of spatial dimensions. Time synchronization: Based on the NTP protocol, the timestamp error of each monitoring device is ensured to be ≤1 second. Data with different sampling frequencies is interpolated and aligned, and the sampling frequency is unified to ensure the synchronization of indicator data. Feature fusion: Extract the temporal and spatial features of each indicator data to form a fused feature vector, and complete the data fusion of toilet space dimension with parking space and charging service dimension, as well as the data association between toilet space dimension and ancillary facilities.

[0015] Optionally, input the congestion assessment model optimized by the random forest algorithm, use the correction coefficient K to optimize the predicted base index, and obtain the dynamic congestion index Z and level. The specific implementation steps are as follows: Baseline Construction: Based on at least one year of historical monitoring data, six types of historical scenario baselines were constructed, including daily commuting, peak hours, concentrated entry of new energy vehicles into the area, extreme weather emergencies, large-scale event gatherings and dispersals, and nighttime rest and support. Model prediction: Using six historical scenario baselines as training samples, the congestion assessment model driven by the random forest algorithm learns the weight distribution and data correlation patterns of 14 indicators under different scenarios, forming a scenario-based assessment capability; in the real-time assessment stage, the 14 pre-processed and fused indicators are input, and the congestion assessment model generates the basic index for the corresponding scenario based on the learned baseline features. Output results: After optimizing the predicted base index using the correction coefficient K=K1×K2, the final output is the dynamic congestion index Z in the range of 0-100. The levels are divided as follows: Z<30 is smooth traffic, 30≤Z<60 is slow traffic, 60≤Z<80 is congested traffic, and Z≥80 is severe congestion.

[0016] Optionally, the calculation process for K1 specifically includes: Scene matching: Based on real-time acquired indicator data, automatically match the corresponding historical scene baseline; Baseline extraction: Extract the baseline value sequence of 14 indicators from the matched historical scene baseline, denoted as B1, B2, ..., B 14 Simultaneously, the real-time value sequences of 14 preprocessed and fused indicators are extracted and denoted as R1, R2, ..., R 14 ; Weighted calculation: Based on the indicator weights, the real-time values ​​and baseline values ​​of each indicator are weighted and summed, using the formula K1=Σ(R i / B i ×W i ), i=1 to 14, Wi Let ΣW be the weight of the i-th indicator. i =1; Threshold constraint: The calculated K1 is constrained within the range of [0.5, 2.0]. Values ​​below 0.5 are counted as 0.5, and values ​​above 2.0 are counted as 2.0.

[0017] Optionally, when Z exceeds the congestion threshold, relevant reference data is retrieved to pinpoint the cause of congestion. The specific implementation steps are as follows: Congestion area identification: When Z≥60, it is determined to be congested; by tracing the contribution of each regional indicator back through the random forest algorithm, the core congested area is identified, with a positioning accuracy error ≤5 meters; Anomaly analysis of related indicators: 14 indicators and related reference data of the core congestion area are extracted; then they are compared with the corresponding historical scenario baseline. If the deviation value exceeds the threshold, it is judged as abnormal data, including abnormal indicators and abnormal related reference data; then, based on the correlation model, strong correlation indicators related to the abnormal data are found. Cause type matching and verification: Pre-set single congestion cause and compound congestion cause type libraries, integrate abnormal data and strongly correlated indicators together and match them with the type libraries, ensuring a matching degree of ≥85%, output the matched congestion cause, and then combine real-time video monitoring to verify the accuracy of the congestion cause.

[0018] Optionally, differentiated services can be provided to the management team and to drivers and passengers differentiated by vehicle type and powertrain, specifically: Provides management with functions such as visual display of congestion status, tracing the causes of congestion, recommending diversion strategies, and sending tiered early warnings. General information and information specific to new energy vehicles are released based on the vehicle type and powertrain type for both drivers and passengers.

[0019] Optional, site bearing pressure index The calculation formula is as follows:

[0020] In the above formula, Indicates the number of vehicles entering the area. Indicates the number of vehicles leaving the area. This indicates the average parking space occupancy time. This indicates the total number of parking spaces. This indicates the vehicle model correction factor.

[0021] Secondly, the present invention also provides a computer device, including a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of the aforementioned service area real-time congestion assessment method.

[0022] Thirdly, the present invention also provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of the aforementioned service area real-time congestion assessment method.

[0023] Fourthly, the present invention also provides a computer program product, including a computer program, which, when executed by a processor, implements the steps of the aforementioned service area real-time congestion assessment method.

[0024] One or more technical solutions provided by this invention have at least the following technical effects or advantages: This invention provides a multi-dimensional, multi-indicator, dynamic, collaborative, fully adaptable, and implementable real-time congestion assessment system for service areas. Its core is to standardize data processing procedures and achieve closed-loop management of the entire data chain. Specifically, it includes: 1. Overcoming the limitations of single-dimensional data and compatibility with new energy vehicles: Integrating data from three core dimensions—parking spaces, toilet spaces, and charging services—and supplementing this with auxiliary dimensions such as weather and holidays, the system clarifies the entire data processing flow through the deep integration of IoT devices and intelligent algorithms. It refines 14 indicators, including parking space turnover rate, power type ratio, and charging pile compatibility rate. By using the power type ratio to link the two dimensions of parking spaces and charging services, it accurately adapts to the energy replenishment and parking needs of new energy vehicles, avoids misjudgments of structural congestion, provides standardized support for the data collection process, and completely solves the problems of misjudgment from single-dimensional data and insufficient compatibility with new energy vehicles.

[0025] 2. Break down data collaboration barriers and improve algorithm fusion mechanisms: Build a unified data fusion and analysis platform to break down data silos across multiple scenarios. Through association rule algorithms and spatiotemporal matching algorithms, achieve deep fusion of multi-source data. It can associate data from multiple scenarios, link toilet and parking spaces, charging service dimensions and related reference data, establish multi-dimensional data relationships, accurately locate congestion transmission paths and causes, improve data fusion links, provide data support for differentiated services for management and drivers, and completely solve the problems of data dispersion and inability to conduct collaborative analysis.

[0026] 3. Real-time congestion assessment and pre-warning mechanism: The 14 pre-processed and integrated indicators are input into the congestion assessment model optimized by the random forest algorithm. The basic index is optimized using the correction coefficient K, and a dynamic congestion index is automatically generated and classified. It can predict the development trend of congestion and trigger an early warning before the demand for vehicles, pedestrians, and energy replenishment approaches the congestion threshold. This realizes the transformation from post-event handling to pre-event warning and completely solves the problem of being unable to predict congestion trends and missing the opportunity for intervention.

[0027] 4. Improve data application processes and form a closed-loop process: Based on accurate early warning and congestion cause location results, further improve data application processes and provide differentiated services to the management end and drivers and passengers differentiated by vehicle type and power type, forming a closed-loop process. Attached Figure Description

[0028] Figure 1 A schematic diagram of the entire data processing flow for a real-time congestion assessment method for service areas; Figure 2 A schematic diagram illustrating the construction logic of a multi-dimensional indicator system; Figure 3 This is a processing logic diagram of each monitoring device in the multi-source sensing process. Figure 4 A detailed processing logic diagram for the hierarchical transmission and edge preprocessing stages; Figure 5 This is a diagram illustrating the specific processing logic of the data fusion process. Figure 6 A diagram illustrating the specific processing logic of the intelligent modeling and analysis process; Figure 7 A diagram illustrating the specific processing logic for differentiated service steps; Figure 8 A system interface diagram showing the trend of the number of service areas where toilets are at saturation, in a specific example; Figure 9 The system interface diagram of the top ten service areas with the longest duration of toilet saturation, in a specific example; Figure 10 The system interface diagram shows the number of service areas with a daily charging pile utilization rate exceeding 40% in a specific example. Figure 11 The system interface diagram is shown in a specific example for the top ten service areas with a daily charging pile utilization rate of over 40%. Detailed Implementation

[0029] This invention provides a method, device, medium, and program product for real-time congestion assessment in service areas, which solves the significant limitations of existing technologies in terms of the accuracy, foresight, and comprehensiveness of congestion assessment, as well as the core problems of non-standard data time processing procedures and lack of power type adaptation.

[0030] To better understand, a detailed description will be provided below with reference to the accompanying drawings and specific embodiments. Obviously, the embodiments described in this invention are only a part of the embodiments of this invention, not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without inventive effort are within the scope of protection of this invention.

[0031] like Figure 1As shown, the present invention provides a real-time congestion assessment method for service areas, which is applied to a real-time congestion assessment system for service areas. The system follows a full-process data processing flow of "multi-source perception - hierarchical transmission and edge preprocessing - data fusion - intelligent modeling and assessment - differentiated services" to achieve real-time perception, accurate assessment, early warning, and congestion cause localization for congestion in multiple scenarios. The core content is as follows: I. System Architecture Design As shown in Table 1, the system adopts a four-layer architecture of "perception layer - transmission layer - data layer - application layer". The perception layer realizes multi-source perception, the transmission layer realizes data transmission and edge preprocessing, the data layer realizes data fusion, and the application layer realizes intelligent modeling, analysis and differentiated services, ensuring low latency and high reliability operation of the entire link and accurately matching each link in the data processing process.

[0032] Table 1 System Architecture Design Table

[0033] II. Operation of each stage of the entire data processing workflow 1. Construction of a multi-dimensional indicator system: Please refer to... Figure 2 The document clarifies the real-time data collection standards and the collection scope of 14 indicators to support multi-source sensing and ensure the targeted and standardized nature of data collection. Parking spaces, toilet spaces, and charging services are the three core dimensions, with auxiliary dimensions including weather and holiday indices. The power type percentage among the 14 indicators is used to link parking space occupancy and charging demand. In other words, the power type percentage acts as a bridge indicator connecting the parking space and charging service dimensions, enabling data linkage between the two core dimensions. Specific indicators are as follows: Parking space dimensions (6 items, core): inbound traffic flow rate, outbound traffic flow rate, vehicle type ratio, power type ratio, parking space turnover rate, and parking space occupancy time. See Table 2 for details.

[0034] Table 2. Explanation of 6 Indicators for Parking Spaces

[0035] In Table 2, the vehicle type ratio and power type ratio complement each other to support congestion assessment. The core differences between the two are as follows: In terms of statistical attributes, the vehicle type ratio focuses on vehicle categories (classified by size and purpose), while the power type ratio focuses on vehicle power attributes (statistical proportion of new energy vehicles); In terms of functional positioning, the vehicle type ratio is used to determine the compatibility between parking spaces and vehicle structure, while the power type ratio is used to assess the parking space occupancy and charging demand of new energy vehicles, predict charging pressure, and identify complex congestion problems.

[0036] Toilet stall dimension (3 items, core): toilet stall utilization rate, toilet stall turnover rate, and toilet stall passenger density, see Table 3 for details. The three indicators under the toilet stall dimension are also linked to the handwashing stations and mother and baby room facilities. The utilization rate of the facilities is used as the reference data for the linkage, but is not included in the 14 indicators. This is used to help verify the accuracy of the toilet stall dimension congestion assessment, locate the causes of congestion, and provide a reference for coordinated traffic management.

[0037] Table 3. Explanation of the three indicators for toilet stalls

[0038] Charging service dimensions (4 items, core): charging pile utilization rate, charging waiting time, charging pile occupancy time, and charging pile compatibility rate, see Table 4 for details.

[0039] Table 4. Explanation of the four indicators in the charging service dimension

[0040] Auxiliary dimension (1 item, auxiliary): Comprehensive index K2, which integrates weather index and holiday index, quantifies the interference of weather and time period on the traffic efficiency of service area, and is only used for the basic index optimization of dynamic congestion index, and is not included in the core weight allocation.

[0041] 2. Vehicle Feature Recognition: A dual recognition model consisting of a license plate recognition module and a camera is adopted. Through initial license plate judgment and secondary verification of vehicle body features, accurate recognition of power type and vehicle type is achieved. The core steps are license plate information extraction, initial judgment of power type, verification of vehicle body features, and result output. The recognition accuracy rate is ≥96%, ensuring the accuracy of collected data and supporting subsequent data processing.

[0042] 3. Algorithm Fusion Mechanism: In the data fusion stage, association rule algorithms and spatiotemporal matching algorithms compatible with the congestion assessment model are adopted to associate data from multiple scenarios, with a focus on linking toilet and parking spaces, charging services, and related reference data of ancillary facilities. This breaks down data silos, accurately locates the congestion transmission path and causes, and improves the data fusion process.

[0043] 4. Congestion Assessment Model: The constructed congestion assessment model, optimized using the random forest algorithm, is compatible with the association rule algorithm in the data fusion stage, balancing adaptability and flexibility. Based on nearly one year of historical monitoring data, six historical scenario baselines are constructed, including daily commuting, peak hours, concentrated entry of new energy vehicles into the area, extreme weather emergencies, large-scale event gatherings and dispersals, and nighttime rest and support. Using 14 pre-processed and fused indicators as input, a dynamic congestion index ranging from 0 to 100 is generated through feature extraction, dynamic weight allocation (the weights of the three core dimensions are fine-tuned ±5% according to the scenario, while the weight of the toilet stall dimension remains stable at around 30%), and basic index optimization. This corresponds to four warning levels (smooth traffic <30, slow traffic 30-60, congested 60-80, severe congestion ≥80), with an output delay of ≤2 seconds. Simultaneously, related reference data (site carrying capacity index) is retrieved. (Utilization rate of ancillary facilities) to help locate the causes of congestion, ensure that the analysis results are consistent with reality, and complete the core operations of data modeling and analysis.

[0044] Among them, the site bearing pressure index The calculation formula is as follows:

[0045] In the above formula, This indicates the number of vehicles entering the area. The number of vehicles entering the area = the rate of traffic flow entering the area × the duration of the statistical period. This indicates the number of vehicles leaving the area. The number of vehicles leaving the area = the outbound traffic flow rate × the statistical period duration. The average parking space occupancy time is calculated by weighting the parking space occupancy time of all vehicles by type within the statistical period. This indicates the total number of parking spaces, specifically the total number of legally designated parking spaces in the service area (including dedicated parking spaces for small passenger cars, trucks, and RVs), which is a basic parameter in the parking space dimension. This represents the vehicle type correction factor, which is dynamically adjusted according to the proportion of each vehicle type. A factor of 1.0 is used for passenger cars, 1.8 for trucks, and 2.5 for motorhomes. The weighted calculation method is as follows: .

[0046] It should be noted that the 14 indicators after preprocessing and fusion were used as input data for the model, including the site bearing pressure index. The auxiliary indicators derived from the parking space dimension have the following specific logic: Data source: Site bearing capacity index Based on the 14 indicators, including the inbound traffic flow rate, outbound traffic flow rate, and parking space occupancy time, the model was trained using these 14 indicators, allowing the model to learn the complex relationship between multiple indicators and the final congestion. Core application: Site bearing capacity index It is not used as input for congestion assessment models, but is used to visually display the real-time parking space capacity pressure to administrators, such as the site capacity pressure index. The system indicates that parking space pressure is too high and is also used to help pinpoint the causes of congestion, such as linking indicators like charging pile utilization rate and power type ratio, to verify whether parking space congestion is caused by overload. It does not participate in the result verification of the congestion assessment model. Link Closed Loop: 14 indicators support the generation of basic indices for congestion assessment models, including site carrying capacity pressure index. In addition, a deeper assessment of parking space dimensions is provided. The combination of these two approaches achieves the dual goals of macro-level congestion assessment and micro-level bottleneck identification, spanning the entire data processing process.

[0047] Please continue to refer to the following. Figure 1 This paper will provide a detailed introduction to the specific implementation steps of a service area real-time congestion assessment method of the present invention, following the full-process data processing flow of "multi-source perception - hierarchical transmission and edge preprocessing - data fusion - intelligent modeling and judgment - differentiated services".

[0048] 1. Multi-source sensing The perception layer is the data source of the entire system. Its core objective is to achieve real-time data collection with full coverage, high precision, and low latency in the service area. It is necessary to strictly follow the collection standards of 14 indicators. In addition to collecting 3 indicators for the toilet stall dimension, it is also necessary to collect the usage rate of the auxiliary facilities of the handwashing station and mother and baby room as related reference data. The specific equipment deployment and functions are shown in Table 5.

[0049] Table 5 Equipment Deployment and Function Table

[0050] Please refer to the following. Figure 3 The monitoring equipment for each dimension is described as follows: 1.1 Parking Space Dimensional Monitoring Equipment 1.1.1 Parking space status monitoring Cameras and parking space analysis modules are deployed in the parking lot to capture the parking space occupancy status in real time, providing basic data support for parking space-related indicators such as parking space turnover rate and parking space occupancy time. The sampling frequency is 1 time / 3 seconds to ensure data time sequence consistency and provide standardized data for subsequent dynamic processes.

[0051] 1.1.2 Dual Recognition Model A dual recognition model consisting of a license plate recognition module and cameras is deployed at checkpoints. The model's recognition accuracy is trained using 80,000 samples (30% of which are new energy vehicles), achieving a final overall recognition accuracy of ≥96%. This meets the core needs of service area congestion assessment, simplifies the vehicle and power type recognition process, and reduces deployment complexity while ensuring core recognition accuracy. The dual recognition model collects data on the number of vehicles entering and leaving the service area, the proportion of different vehicle types, the proportion of different power types, entry time, and exit time. This data is integrated in real-time, avoiding the bias of single-device data collection and overcoming the limitations of traditional power type compatibility. The specific implementation steps are as follows: (1) License plate information extraction and initial determination of power type: The license plate recognition module extracts the license plate information first, and quickly determines the power type, such as fuel or electric, by the color and number segment of the new energy special license plate, such as green plate or exclusive number segment. At the same time, it extracts the vehicle registration information associated with the license plate to help confirm the vehicle type, such as small passenger car, truck, or RV.

[0052] (2) Vehicle body feature verification and result output: The camera simultaneously collects the vehicle body appearance features, such as the location of the charging interface and the new energy vehicle sticker, and performs secondary verification on the license plate recognition results to exclude misjudgment scenarios such as fuel vehicles using new energy vehicle license plates or modified vehicles, and outputs the verification results to enhance the reliability of recognition.

[0053] 1.2 Monitoring equipment for toilet stall dimensions Cameras are installed at the entrances and exits of restrooms to count the density of people entering and exiting the toilets. Figure 8 The system interface diagram shows the trend of the number of service areas where toilets reach saturation, using a specific example. Figure 8 It can be seen that the number of service areas in saturation reached its peak on January 1, 2026, and then declined. On January 8, 2026, the number of service areas in saturation rebounded slightly. Figure 9 Provide a system interface diagram of the top ten service areas with the longest duration of toilet saturation, using a specific example. Figure 9 It can be seen that Yongxiu ranks first in total duration and uphill duration in terms of saturation time, but lowest in downhill duration, indicating that toilet resources need to be optimized in its uphill service area.

[0054] Infrared toilet occupancy sensors are installed in each toilet stall to collect real-time data on the usage status of the stalls and calculate the toilet turnover rate based on the time people spend in the toilet.

[0055] The usage rates of ancillary facilities such as mother and baby rooms and washbasins are collected synchronously by infrared sensors, door magnetic sensors, and video counters. These data are used as reference data and stored in a tagged manner according to "facility type-area number-equipment ID", laying the foundation for the correlation analysis between toilet stalls and ancillary facilities.

[0056] 1.3 Monitoring Equipment for Charging Service Power type compatibility detection sensors and charging usage monitoring equipment work together to achieve multi-indicator data collection across charging services: accurately identifying vehicle charging interface standards (such as national standard GB / T, Tesla Supercharger, CCS, etc.) and comparing them with charging pile interface types, outputting a "compatible / incompatible" judgment signal; reading the rated charging power of electric vehicles (such as 30kW slow charging, 120kW fast charging), matching the output power range of charging piles, and determining whether it meets the vehicle's rapid energy replenishment needs; and simultaneously collecting charging pile utilization rate, real-time charging power, charging time, and charging waiting time.

[0057] Figure 10 The system interface diagram shows the number of service areas with a daily charging pile utilization rate exceeding 40%, as illustrated in a specific example. Figure 10 It can be seen that from January 1, 2026 to January 3, 2026, the number of new energy vehicles entering the area on a single day remained at a high level, and the number of service areas with a daily charging pile utilization rate of over 40% also remained at a high level during this period. However, after January 4, 2026, the number dropped sharply and remained at a low level. Figure 11 The system interface diagram is shown in a specific example for the top ten service areas with the highest daily charging pile utilization rate exceeding 40%. Figure 11 It is evident that Lushan ranks first in both the total number of days with a charging pile utilization rate exceeding 40% and the number of days with a charging pile utilization rate exceeding 40% for both uplink and downlink charging piles, indicating an urgent need to optimize charging service resources.

[0058] 1.4 Auxiliary Dimension Monitoring Equipment 1.4.1 Weather Index Collection: By connecting to the public API interface of the National Meteorological Administration, such as the China Weather Network API, the system obtains real-time weather type (sunny, cloudy, rainy, snowy, etc.), precipitation intensity, and wind force data of the service area. The system automatically maps the weather type to the corresponding weather index (sunny=1.0, cloudy=1.1, rain / snow=1.3). When the precipitation intensity is greater than or equal to moderate rain and the wind force is greater than or equal to level 6, the weather index is increased by 0.1. The system is updated every 10 minutes every day to ensure the timeliness of the data.

[0059] 1.4.2 Holiday Index Collection: The system has a built-in database of national statutory holidays (including adjusted workdays) and weekend calendars. It pre-enters the holiday periods throughout the year (accurate to the start and end dates and specific time periods, such as statutory holidays and weekends, which are 00:00-24:00 every day). It automatically identifies the current time period type and maps it to the corresponding holiday index (weekdays = 1.0, weekends = 1.2, statutory holidays = 1.4). It also supports manual addition of temporary holidays and regional major event periods (such as local festivals) to flexibly adapt to special scenarios.

[0060] 1.4.3 Comprehensive Index Calculation: K2 = Weather Index × Holiday Index.

[0061] 2. Hierarchical transmission and edge preprocessing The transport layer performs dual functions of hierarchical transmission and edge preprocessing. It employs a hybrid transmission architecture of "edge gateway + 5G / NB-IoT / fiber optic," connecting the multi-source sensing layer and the cloud data layer to achieve low-latency, highly reliable synchronous data transmission. This addresses the issues of high latency in large-volume data transmission and heavy cloud data processing pressure, ensuring data timeliness and processing efficiency. Please refer to the following: Figure 4 The specific steps are as follows: 2.1 Edge Gateway Deployment Deploy an edge computing gateway locally in the service area as a core hardware node of the transport layer, with the following responsibilities: 2.1.1 Multi-source data aggregation: Access data collected by all monitoring devices in the perception layer, focusing on aggregating toilet stalls and related reference data for handwashing stations and mother-and-baby rooms, and simultaneously integrating data from parking spaces and charging services.

[0062] 2.1.2 Protocol Standardization Conversion: The private protocols of heterogeneous devices (such as sensor serial port protocols, camera SDK protocols, and charging pile communication protocols) are uniformly converted into the MQTT / HTTP standardized protocol, and core fields such as device ID, timestamp, and spatial tag are added to lay the foundation for subsequent data transmission and processing.

[0063] 2.1.3 Pre-defined transmission strategies: Based on data type and real-time requirements, a hierarchical transmission strategy is pre-defined to provide rule support for subsequent data transmission.

[0064] 2.2 Hierarchical Transmission Strategy For data with high real-time requirements, such as inbound and outbound traffic flow rates, toilet turnover rates, and charging pile compatibility rates, data is transmitted to the cloud in real time via 5G / NB-IoT. For non-real-time data, such as historical traffic trends, equipment operation logs, and historical usage rates of ancillary facilities, data is transmitted in batches via fiber optics to reduce network bandwidth consumption and ensure that different types of data transmission meet their timeliness requirements. All transmitted data is accompanied by a CRC checksum. The cloud verifies the integrity of the data upon receipt, and if the verification fails, a breakpoint resume mechanism is triggered to prevent data loss.

[0065] 2.3 Edge Preprocessing After completing data aggregation and protocol conversion, the edge gateway performs edge-specific preprocessing operations, handling only tasks with high real-time requirements and low computational load to avoid duplication with cloud functions. The specific steps are as follows: 2.3.1 Uniform equipment sampling frequency Preset the sampling frequency according to the type of monitoring equipment in the sensing layer to ensure a consistent data collection rhythm, such as toilet passenger flow density once every 5 seconds, charging pile usage rate once every 1 second, vehicle flow rate entering the area once every 2 seconds, and weather index once every 10 minutes.

[0066] 2.3.2 Dynamic denoising and missing value completion (executed only at the edge, not repeated in the cloud) (1) Data windowing: Real-time data is divided into 1 minute / window, and each window contains continuous sampling data of each indicator.

[0067] (2) Outlier removal: The sliding window midpoint method is used to remove data that exceed the threshold range [M-2σ,M+2σ]; Gradient detection method is used simultaneously to determine that when the data change rate is >50% / second and there is no traffic flow data to support it, it is an anomaly and is directly removed.

[0068] (3) Missing value completion: For short-term missing data (e.g., missing data ≤ 3 data points), linear interpolation is used; for long-term missing data (e.g., missing data > 3 data points), historical similar scenario completion is used to complete the data, ensuring the integrity of the edge basic dataset.

[0069] 3. Data Fusion The data layer is the core processing unit of the system. By building a unified fusion analysis platform, it enables correlation analysis of data from various dimensions, connecting edge preprocessing with intelligent modeling and judgment. Please refer to the following. Figure 5 The specific steps are as follows: 3.1 Database Construction The platform integrates real-time and historical databases to enable the categorized storage of collected data.

[0070] 3.1.1 Real-time Database It stores high-frequency collected data within 5 minutes, including real-time data collected from various dimensions, to support real-time analysis.

[0071] 3.1.2 Historical Database It stores nearly a year's worth of monitoring data, including data on the correlation between indicators of various dimensions and the congestion index, to build baselines for six types of historical scenarios, including daily commuting, peak hours, concentrated entry of new energy vehicles into the area, extreme weather emergencies, large-scale event gatherings and dispersals, and nighttime rest and support.

[0072] 3.1.3 Adding new data fields Power type coding (0=gasoline vehicle, 1=new energy vehicle, 2=hybrid vehicle), charging interface type (0=DC fast charging, 1=AC slow charging, 2=dual interface), charging pile interface specifications, charging time by vehicle type, congestion contribution of electric vehicle, toilet area coding, handwashing station and mother and baby room facility coding, etc., to ensure the storage and retrieval of indicator data and related reference data in various dimensions.

[0073] 3.2 Indicator Normalization Process Min-Max normalization is used to map the 14 indicators uploaded from the edge to the [0,1] interval to adapt to the input requirements of the random forest model and ensure the fairness of the weight calculation of each indicator.

[0074] 3.3 Spatiotemporal matching and multi-dimensional fusion Deep data fusion is achieved by combining association rule algorithms compatible with congestion assessment models with spatiotemporal matching algorithms. Specifically: 3.3.1 Spatial Alignment: Establish a 10m×10m unified spatial grid model, associate all monitoring devices with a unique grid number, and group data from the same grid into the same spatial dataset. Focus on associating monitoring devices for toilet stalls, handwashing stations, and mother-and-baby room facilities with the same grid to achieve collaborative analysis of spatial dimensions.

[0075] 3.3.2 Time Synchronization: Based on the NTP protocol, ensure that the timestamp error of each monitoring device is ≤1 second, interpolate and align data with different sampling frequencies, and unify the sampling frequency to 1 time / second to ensure data synchronization across all dimensions.

[0076] 3.3.3 Feature Fusion: Extract the temporal features (mean, variance, peak value) and spatial features (spatial distribution density, correlation between adjacent grids) of each indicator to form a fused feature vector. The focus is on completing the data fusion of parking space dimension and charging service dimension, as well as the data correlation of toilet space dimension and ancillary facilities.

[0077] 4. Intelligent Modeling and Analysis The algorithm model library is the core processing unit of the data layer. The application layer calls the algorithm model library, using 14 pre-processed and fused indicators as the core input, and links them to the site bearing pressure index. To assist in locating the causes, the emphasis is on strengthening the weighting and analytical role of indicators across various dimensions, constructing a complete chain of "input adaptation - model calculation - result output - cause localization." Please refer to the following. Figure 6 The specific steps are as follows: 4.1 Model Input Adaptation 4.1.1 Baseline Adaptation: Based on real-time monitoring data, the historical scenario baseline is automatically matched. For example, the historical scenario baseline during the National Day holiday is called, and the historical scenario baseline for concentrated entry of new energy vehicles into the area is called when the power type accounts for more than 40%.

[0078] 4.1.2 Data Linkage: The 14 pre-processed and integrated indicators are input into the congestion assessment model, and the site carrying capacity index is retrieved simultaneously. The utilization rate of ancillary facilities is used as a supplementary reference for causal identification.

[0079] 4.1.3 Optimization of correction coefficient: The basic index for prediction of the congestion assessment model is optimized using the correction coefficient K=K1×K2.

[0080] Wherein, K1 is the ratio of real-time to historical scenario baselines, calculated using a weighted average ratio. The specific implementation steps are as follows: ① Scenario matching: Based on real-time acquired indicator data, such as the proportion of power type, time period, and weather index, automatically match the corresponding historical scenario baseline in the historical database. For example, if the proportion of power type is 45%+ and it is weekday evening peak+ and it is sunny, match the historical scenario baseline of concentrated entry of new energy vehicles into the area; ② Baseline extraction: Extract the baseline value sequence of 14 indicators from the matched historical scenario baseline, denoted as B1, B2, ..., B 14 The baseline value sequence corresponds to the historical average level of 14 indicators. Simultaneously, the real-time value sequences of the 14 indicators after preprocessing and fusion are extracted and denoted as R1, R2, ..., R... 14 ③ Weighted Calculation: Based on the indicator weights, such as 40% for parking spaces, 30% for toilet stalls, 25% for charging services, and 5% for auxiliary services, the real-time and baseline values ​​of each indicator are weighted and summed. The formula is K1=Σ(R i / B i ×W i ), i=1 to 14, W i Let ΣW be the weight of the i-th indicator. i =1, this calculation method does not require additional normalization and the scale is more stable; ④ Threshold constraint: constrain the calculated K1 in the range of [0.5, 2.0], below 0.5 is counted as 0.5, above 2.0 is counted as 2.0, to avoid excessive correction due to extreme data.

[0081] Among them, the comprehensive index K2 = weather index × holiday index.

[0082] 4.2 Core Logic of Congestion Assessment Model The system incorporates a congestion assessment model optimized using a random forest algorithm. The core logic is as follows: Based on nearly one year of historical monitoring data, six historical scenario baselines are constructed. These baselines serve as training samples, driving the congestion assessment model optimized by the random forest algorithm to learn the weight distribution and data correlation patterns of 14 indicators under different scenarios, thus forming scenario-based assessment capabilities. In the real-time assessment phase, the 14 pre-processed and fused indicators are input. Based on the learned baseline features, the congestion assessment model generates a basic index for the corresponding scenario. The weight of the toilet stall dimension is consistently around 30%, balanced with the weights of parking spaces and charging services. After optimizing the predicted basic index using a correction coefficient K=K1×K2, a dynamic congestion index Z ranging from 0 to 100 is finally output. The index levels are divided as follows: Z<30 for smooth traffic, 30≤Z<60 for slow traffic, 60≤Z<80 for congestion, and Z≥80 for severe congestion, completely solving the problems of lag and insufficient scenario adaptability in traditional static threshold judgments. (Site carrying capacity pressure index) As an auxiliary indicator, it is used to help managers understand the parking space occupancy status and is not involved in the model result verification and correction.

[0083] 4.3 Identifying the Causes of Congestion 4.3.1 Congestion Area Locking: Based on the output dynamic congestion index Z, congestion is determined when Z≥60; the contribution of each regional indicator is traced back through the random forest algorithm to extract the core congestion area, such as the toilet indicator contribution of 35% in area A and the parking space indicator contribution of 40% in area B. Combined with 10m×10m spatial grid data, the specific range is locked, such as "toilet area A" and "parking area B + adjacent charging pile group", with a positioning accuracy error ≤5 meters.

[0084] 4.3.2 Anomaly Analysis of Correlation Indicators: After locating the core congestion area identified in the previous step, two types of key data for that area are extracted—14 indicators and related reference data. Then, these are compared with the corresponding historical baseline. If the deviation between the two types of key data and the historical average exceeds the threshold, they are identified as abnormal data ("abnormal indicators + abnormal related reference data"). The system will use the correlation model to find strongly correlated indicators related to the abnormal data, clarify the inherent correlation between each anomaly, and provide core data support for subsequent cause type matching.

[0085] For example, the parking space turnover rate of a certain parking area is 0.3 times / minute, which is lower than the historical average of 0.6 times / minute for the corresponding scenario, and the site load pressure index is also high. =0.78 > threshold 0.7. Based on the correlation model, the system searched for strong correlation indicators related to abnormal data and finally found that: the charging pile utilization rate was as high as 90% (almost full load, vehicles did not leave in time after charging) and the proportion of electric vehicles reached 45% (electric vehicles take a long time to charge, occupying parking spaces) – these two indicators are the cause of low parking space turnover. Strongly correlated indicators whose values ​​exceed the standard.

[0086] For example, the toilet stall turnover rate in a certain restroom area was only 0.2 times / minute, far below the historical average of 0.7 times / minute, which is an abnormal indicator. Upon checking the related reference data of the surrounding facilities, it was found that the washbasin usage rate was 92% and the mother and baby room usage rate was 88%, which are also abnormal reference data. Both are close to full capacity. This confirms that the washbasins and mother and baby room are too crowded, causing people to be unable to leave the toilet stall in time after use, thus resulting in a low toilet stall turnover rate.

[0087] 4.3.3 Cause Type Matching and Verification: A pre-defined type library is prepared for single congestion causes (such as "insufficient dedicated electric vehicle parking spaces", "parking space overload", "toilet shortage", "insufficient handwashing facilities leading to pedestrian congestion") and compound congestion causes (such as "charging vehicles occupying spaces + parking space overload", "inefficient toilet turnover + handwashing station congestion"). The combination of "abnormal indicators + abnormal correlation reference data" (i.e., abnormal data) formed in step 4.3.3 and strongly correlated indicators are integrated together and matched with the type library. The matching degree must be ≥85%. The matched congestion causes are output, and the accuracy of the congestion causes is verified by combining real-time video monitoring.

[0088] 4.3.4 Output of Causes and Generation of Relief Suggestions: Outputs a report containing the causes of congestion and strongly correlated indicators. Based on the one-to-one correspondence between congestion causes and relief strategies, it automatically matches relief strategies. For example, for toilet congestion, it dispatches cleaning staff to speed up toilet turnover and guides non-urgent users to use the toilets at off-peak times; for handwashing station congestion, it increases cleaning staff and opens alternative handwashing areas; site carrying capacity index... If the value is abnormal, temporary parking spaces will be opened and non-charging vehicles will be diverted; if the proportion of power type is high, portable charging piles will be allocated and cross-regional charging will be guided.

[0089] 5. Differentiated services At the application level, differentiated services are provided to two user groups: managers and drivers (categorized by vehicle type and powertrain). The focus is on showcasing the application value of congestion analysis across various dimensions of the service area, connecting intelligent modeling and analysis processes to realize the value of the solution and complete the closed-loop data processing workflow. On one hand, it provides managers with targeted traffic management suggestions to reduce congestion duration and secondary problems, ultimately enabling the service area to transform from passive to proactive service. On the other hand, it pushes real-time congestion status, parking space and charging pile compatibility information, and alternative solutions to drivers and passengers of different vehicle types and powertrains, assisting them in making reasonable parking decisions. Please refer to the following: Figure 7 The specific steps are as follows: 5.1 Management Terminal Functions Develop a service area congestion monitoring and management platform for managers to support precise control. 5.1.1 Visual display of congestion status: Real-time display of the dynamic congestion index levels in each area, with a focus on differentiating the congestion status in three core dimensions of parking spaces, toilet cubicles, and charging services. Simultaneously display the distribution of electric vehicle models, the usage rate and waiting time trend charts of chargers by vehicle type, the usage rate of toilet cubicles, the real-time data of the usage rates of washbasins and mother and baby rooms, and the site bearing pressure index 。

[0090] 5.1.2 Tracing the causes of congestion: Automatically locate the strongly associated indicators of congestion through steps 4.3.1 - 4.3.3 above, and identify the causes of congestion in each dimension, such as inefficient toilet cubicle turnover, insufficient washbasin facilities, etc.

[0091] 5.1.3 Recommended congestion mitigation strategies: Combine the identified causes of congestion and targetedly push differentiated congestion mitigation strategies to achieve precise matching of "cause - strategy". For example: For congestion in the toilet cubicle dimension (including associated congestion in washbasins and mother and baby rooms), push "Dispatch cleaning staff to accelerate toilet cubicle turnover, guide non-urgent personnel to use at staggered times, and simultaneously open alternative wash areas"; for congestion in the charging dimension, push "Deploy portable chargers, guide vehicles to refuel across regions, and optimize the priority of charger usage"; for congestion in the parking space dimension (including a relatively high value of the site bearing pressure index and congestion caused by the concentrated entry of new energy vehicles into the area), push "Open temporary parking spaces, guide non-charging vehicles to divert, and standardize the parking order of electric vehicles."

[0092] 5.1.4 Hierarchical early warning push: Relying on the hierarchical standards of the dynamic congestion index and combining the visual monitoring results on the management side, for the three core dimensions of toilet cubicles, parking spaces, and charging services, achieve the synchronous implementation of hierarchical early warning and precise push. The early warning information will be accompanied by the causes of congestion and preliminary mitigation suggestions in the corresponding area, and will be pushed to the management personnel via text messages, platform pop-ups, etc., to ensure that the management personnel can respond quickly and handle precisely.

[0093] 5.2 Functions for drivers and passengers Based on the precise handling on the management side, synchronously connect to the navigation APP and the LED information board at the service area entrance, and real-time publish differentiated service information and alternative suggestions to drivers and passengers, realizing two-way linkage between the management side's handling and the drivers and passengers' guidance. Specifically including: 5.2.1 General information: The overall congestion status of the service area, the remaining number of parking spaces, the availability of toilet cubicles, the usage status of washbasins and mother and baby rooms.

[0094] 5.2.2 Information specific to new energy vehicles: charging pile availability status, interface type compatibility prompts, charging waiting time, and guidance for dedicated electric vehicle lanes, such as "Current service area electric passenger vehicle charging pile waiting time is 15 minutes, recommended next stop XX service area (charging pile available)", "Your electric truck requires a charging pile of 120kW or higher, currently there is only 1 available in the service area, recommended YY service area 30 kilometers away", "Electric passenger vehicles please use lane 2 to directly reach the fast charging area".

[0095] In a second aspect, the present invention provides a computer device, including a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of a service area real-time congestion assessment method.

[0096] The memory can be volatile or non-volatile, or a combination of both. Non-volatile memory can be read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), or flash memory. Volatile memory can be random access memory (RAM), used as an external cache. By way of example, but not limitation, many forms of RAM are available, such as Static Random Access Memory (SRAM), Dynamic Random Access Memory (DRAM), Synchronous DRAM (SDRAM), Double Data Rate SDRAM (DDRSDRAM), Enhanced SDRAM (ESDRAM), Synchlink DRAM (SLDRAM), and Direct Rambus RAM (DRRAM). The memory of this invention is intended to include, but is not limited to, these and any other suitable types of memory.

[0097] The processor can be an integrated circuit chip with signal processing capabilities. In implementation, each step of the above method can be completed by integrated logic circuits in the processor's hardware or by instructions in software form. The processor can be a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. It can implement or execute the methods, steps, and logic block diagrams disclosed in this invention. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the methods disclosed in this invention can be directly embodied in the execution of a hardware decoding processor, or executed by a combination of hardware and software modules in the decoding processor. The software modules can reside in random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, registers, or other mature storage media in the art. This storage medium is located in memory; the processor reads information from the memory and, in conjunction with its hardware, completes the steps of the above method.

[0098] The method steps of this invention can be implemented in hardware, software, firmware, middleware, microcode, or a combination thereof. For hardware implementation, the processing unit can be implemented in one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field-programmable gate arrays (FPGAs), general-purpose processors, controllers, microcontrollers, microprocessors, other electronic units for performing the functions described in this application, or combinations thereof.

[0099] Software implementation can be achieved by executing functional modules (such as procedures, functions, etc.). Software code can be stored in memory and executed by the processor. Memory can be implemented in the processor or outside the processor.

[0100] Thirdly, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of the aforementioned service area real-time congestion assessment method.

[0101] Computer storage media can include various media that can store program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0102] Fourthly, the present invention provides a computer program product, including a computer program, wherein when the computer program is executed by a processor, it implements the steps of a service area real-time congestion assessment method.

[0103] Specifically, computer program products include: data signals, data signals embodied in a carrier wave, or computer-readable storage media.

[0104] It should be noted that the technical solutions described in this invention can be combined arbitrarily without conflict.

[0105] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if such modifications and variations fall within the scope of the claims of this invention and their equivalents, then this invention also includes such modifications and variations.

Claims

1. A method for real-time congestion assessment in service areas, characterized in that, include: Real-time acquisition of indicators and related reference data for three core dimensions—parking spaces, toilets, and charging services—of highway service areas, as well as auxiliary dimensions. The parking space dimension includes inbound traffic flow rate, outbound traffic flow rate, vehicle type ratio, power type ratio, parking space turnover rate, and parking space occupancy time. The toilet dimension includes toilet usage rate, toilet turnover rate, and toilet passenger density. The charging service dimension includes charging pile usage rate, charging wait time, charging pile occupancy time, and charging pile compatibility rate. The auxiliary dimension includes a comprehensive index K2 integrating weather and holiday indices. The parking space and charging service dimensions are linked by the power type ratio, with related reference data including the site load-bearing pressure index. Utilization rate of ancillary facilities; After each indicator is preprocessed at the edge, it is fused in the fusion analysis platform through association rule algorithm and spatiotemporal matching algorithm, and then input into the congestion judgment model optimized by random forest algorithm. The basic index is optimized by correction coefficient K to obtain dynamic congestion index Z and level; where K=K1×K2, K1 is the ratio of real-time and historical scenario baseline; When Z exceeds the congestion threshold, the relevant reference data is retrieved to locate the cause of congestion, and then differentiated services are provided to the management end and the driver and passenger ends distinguished by vehicle type and power type.

2. The method as described in claim 1, characterized in that, The fusion analysis platform integrates association rule algorithms and spatiotemporal matching algorithms. Specific implementation steps include: Spatial alignment: Establish a 10m×10m unified spatial grid model, associate each monitoring device of parking space, toilet stall, and charging service dimension with a unique grid number, and classify the index data of the same grid into the same spatial dataset. Associate the monitoring devices of toilet stalls with the sinks and mother and baby room ancillary facilities with the same grid to achieve collaborative analysis of spatial dimensions. Time synchronization: Based on the NTP protocol, the timestamp error of each monitoring device is ensured to be ≤1 second. Data with different sampling frequencies is interpolated and aligned, and the sampling frequency is unified to ensure the synchronization of indicator data. Feature fusion: Extract the temporal and spatial features of each indicator data to form a fused feature vector, and complete the data fusion of toilet space dimension with parking space and charging service dimension, as well as the data association between toilet space dimension and ancillary facilities.

3. The method as described in claim 1, characterized in that, Input the congestion assessment model optimized by the random forest algorithm, use the correction coefficient K to optimize the predicted base index, and obtain the dynamic congestion index Z and its level. The specific implementation steps are as follows: Baseline Construction: Based on at least one year of historical monitoring data, six types of historical scenario baselines were constructed, including daily commuting, peak hours, concentrated entry of new energy vehicles into the area, extreme weather emergencies, large-scale event gatherings and dispersals, and nighttime rest and support. Model prediction: Using six historical scenario baselines as training samples, the congestion assessment model driven by the random forest algorithm learns the weight distribution and data correlation patterns of 14 indicators under different scenarios, forming a scenario-based assessment capability; in the real-time assessment stage, the 14 pre-processed and fused indicators are input, and the congestion assessment model generates the basic index for the corresponding scenario based on the learned baseline features. Output results: After optimizing the predicted base index using the correction coefficient K=K1×K2, the final output is the dynamic congestion index Z in the range of 0-100. The levels are divided as follows: Z<30 is smooth traffic, 30≤Z<60 is slow traffic, 60≤Z<80 is congested traffic, and Z≥80 is severe congestion.

4. The method as described in claim 3, characterized in that, The calculation process for K1 specifically includes: Scene matching: Based on real-time acquired indicator data, automatically match the corresponding historical scene baseline; Baseline extraction: Extract the baseline value sequence of 14 indicators from the matched historical scene baseline, denoted as B1, B2, ..., B 14 Simultaneously, the real-time value sequences of 14 preprocessed and fused indicators are extracted and denoted as R1, R2, ..., R 14 ; Weighted calculation: Based on the indicator weights, the real-time values ​​and baseline values ​​of each indicator are weighted and summed, using the formula K1=Σ(R i / B i ×W i ), i=1 to 14, W i Let ΣW be the weight of the i-th indicator. i =1; Threshold constraint: The calculated K1 is constrained within the range of [0.5, 2.0]. Values ​​below 0.5 are counted as 0.5, and values ​​above 2.0 are counted as 2.

0.

5. The method as described in claim 1, characterized in that, When Z exceeds the congestion threshold, relevant reference data is retrieved to pinpoint the cause of congestion. The specific implementation steps are as follows: Congestion area identification: When Z≥60, it is determined to be congested; by tracing the contribution of each regional indicator back through the random forest algorithm, the core congested area is identified, with a positioning accuracy error ≤5 meters; Anomaly analysis of related indicators: 14 indicators and related reference data of the core congestion area are extracted; then they are compared with the corresponding historical scenario baseline. If the deviation value exceeds the threshold, it is judged as abnormal data, including abnormal indicators and abnormal related reference data; then, based on the correlation model, strong correlation indicators related to the abnormal data are found. Cause type matching and verification: Pre-set single congestion cause and compound congestion cause type libraries, integrate abnormal data and strongly correlated indicators together and match them with the type libraries, ensuring a matching degree of ≥85%, output the matched congestion cause, and then combine real-time video monitoring to verify the accuracy of the congestion cause.

6. The method as described in claim 1, characterized in that, Provide differentiated services to management and drivers / passengers based on vehicle model and powertrain type, specifically: Provides management with functions such as visual display of congestion status, tracing the causes of congestion, recommending diversion strategies, and sending tiered early warnings. General information and information specific to new energy vehicles are released based on the vehicle type and powertrain type for both drivers and passengers.

7. The method as described in claim 1, characterized in that, Site bearing pressure index The calculation formula is as follows: In the above formula, Indicates the number of vehicles entering the area. Indicates the number of vehicles leaving the area. This indicates the average parking space occupancy time. This indicates the total number of parking spaces. This indicates the vehicle model correction factor.

8. A computer device, comprising a memory, a processor, and a computer program stored in the memory, characterized in that, The processor executes the computer program to implement the steps of the method according to any one of claims 1-7.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1-7.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1-7.

Images