Transport vehicle in-transit queuing optimization method based on big data analysis

Abstract

The invention discloses a transport vehicle in-transit queuing optimization method based on big data analysis. The method comprises the following steps: S1, obtaining and preprocessing transport vehicle interaction data; s2, the average loading period of each crane position is calculated, a vehicle category-crane position efficiency matrix is constructed, and a beat estimation model is established; s3, estimating arrival time by adopting a weighted moving average method, and calculating an optimal entrance time window; s4, an emission grading model is constructed through big data hierarchical clustering, and basic priority scoring is completed; s5, calculating a priority scheduling factor, and generating a dynamic scheduling sequence; s6, the admission time is pushed, and the scheduling system dynamically allocates crane positions according to the real-time state and updates the queue; and S7, when the vehicle approaches the electronic fence area, the scheduling system automatically checks and triggers an entrance instruction, and a scheduling execution closed loop is ensured. According to the invention, through data acquisition, beat modeling, emission grading, priority scheduling and crane position distribution, intelligent optimization from order receiving to loading of the transport vehicle is realized.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent scheduling and transportation management, and in particular to a method for optimizing in-transit queuing of transport vehicles based on big data analysis. Background Art

[0002] In existing logistics and port dispatch management systems, in-transit vehicle scheduling, queuing for entry, and on-site operations often rely on traditional manual control, simple timed scheduling mechanisms, or static rule configurations. This is particularly true in concentrated transportation scenarios for materials like petrochemicals, bulk cargo, coal, and crude oil. Port loading bays are limited, and the sheer number of vehicles, each with varying emission levels, loading capacities, and driving behaviors, leads to uneven vehicle entry and significant fluctuations in loading efficiency, severely impacting overall transportation efficiency and port turnover.

[0003] Although some ports or large storage yards have initially introduced information systems for queue management, such as setting up simple vehicle registration systems, electronic fences, and loading space management interfaces, these systems are generally based on static time windows or manual queuing rules, and lack the ability to dynamically model the actual efficiency of loading. At the same time, most systems ignore the impact of real-time traffic conditions and vehicle emission characteristics on scheduling priorities, and still use the "first come, first served" or "appointment order" queuing logic. This method can easily cause a large number of vehicles to gather outside the station area and traffic congestion during peak hours, and even lead to idle crane positions or waste of resources due to inconsistent loading rhythms, further reducing overall operational efficiency. In addition, existing technologies often use a fixed rhythm to arrange vehicle entry, lack the ability to dynamically adjust according to real-time loading data, and the loading rhythm is seriously disconnected from the actual on-site operating status.

[0004] In terms of intelligent dispatching, although some systems have attempted to introduce big data or dispatching algorithms for queue optimization, the algorithm design still tends to simplify the model and fails to fully integrate multi-source heterogeneous data, such as historical vehicle loading records, emission levels, escort behavior characteristics and other in-depth information. Existing systems are mostly used only to screen for compliance in emission level management, and do not use emission factors as the basis for calculating dispatch priority, making it difficult to meet the comprehensive requirements of the port area's green environmental protection policies and classified dispatch management. In addition, traditional systems lack effective time prediction capabilities, and most of the arrival time estimates are based on a rough estimate of linear distance divided by the static average speed. They cannot adapt to real-time road traffic changes and differences in vehicle driving habits, further affecting the accuracy and rationality of the queue time window.

[0005] In terms of takt scheduling, current research has focused on takt-based scheduling models. However, most models rely on fixed takt cycles set empirically and lack the ability to respond sensitively to fluctuations in loading efficiency. During the actual loading process, the takt cycle is affected by a variety of factors, such as the busyness of the loading crane, the frequency of material type switching, and the rhythm of personnel handovers, all of which can cause deviations between the actual takt and the preset takt. Existing technologies often lack the ability to dynamically identify takt deviations and reconstruct the takt model in real time, resulting in a serious disconnect between the recommended scheduling time window and actual operating capacity, further exacerbating the uneven utilization of vehicle waiting and loading resources.

[0006] At the same time, existing queuing and scheduling systems generally ignore the ability to collaboratively optimize multi-dimensional factors. For example, vehicle arrival time prediction and loading cycle modeling are two independent modules, without data sharing and model fusion. Vehicle prioritization also mostly relies on a single dimension (such as arrival time), lacking the modeling and integration of comprehensive indicators such as emission levels, resource availability, and loading demand type. Furthermore, most existing systems lack the ability to self-learn and make real-time corrections to queue order, and lack feedback mechanisms, making it difficult to adapt to intelligent evolution under complex and changing scenarios.

[0007] Furthermore, the vehicle entry scheduling process often faces issues such as delayed entry signals and drivers not receiving timely information. This results in scheduling plans being generated but not being executed efficiently, impacting the stability of the scheduling closed loop. Although some systems have introduced push notifications or outbound call mechanisms for notifications, these still suffer from issues such as a mismatch between notification cadence and real-time status, simplistic notification policy rules, and a lack of deep integration with the beat scheduling model, which creates the risk of failure at the execution level.

[0008] Therefore, how to provide a method for optimizing in-transit queuing of transport vehicles based on big data analysis is an urgent problem that those skilled in the art need to solve. Summary of the Invention

[0009] One purpose of the present invention is to propose a method for optimizing in-transit queuing of transport vehicles based on big data analysis. The present invention makes full use of technologies such as big data modeling, dynamic beat prediction, emission level hierarchical clustering, and multi-factor scheduling and sorting, and describes in detail how to achieve intelligent optimization of the transport vehicle entry queuing process through real-time loading efficiency modeling, accurate prediction of vehicle arrival times, and a green priority scheduling mechanism. This method has the advantages of high scheduling efficiency, high queuing accuracy, high resource utilization, and strong environmental compliance. It can effectively solve the problems existing in the existing system, such as unstable loading beats, waiting vehicles in clusters, lack of emission control, and disconnected scheduling execution. It is suitable for a variety of logistics and transportation scenarios such as ports, oil stations, and bulk loading and unloading areas.

[0010] A method for optimizing in-transit queuing of transport vehicles based on big data analysis according to an embodiment of the present invention includes the following steps:

[0011] S1. Obtain and preprocess the interaction data of transport vehicles to generate a standardized data set;

[0012] S2. Calculate the average loading cycle for each crane position per unit time based on the standardized data set, construct a vehicle category-crane loading efficiency matrix and a takt estimation model for each category of vehicle, and dynamically adjust it based on the latest interaction data within the sliding time window to generate a loading takt prediction value for each category of vehicle;

[0013] S3. Using big data to drive all vehicles to be dispatched, the weighted moving average method is used to estimate the time required for each vehicle to arrive at the station based on the real-time location and road speed of each vehicle. Combined with the loading cycle prediction value, the optimal entry time window for the vehicle is calculated;

[0014] S4. Use a big data hierarchical clustering algorithm to build a vehicle emission classification model, score the emission levels of all vehicles in the queue, divide the emission levels into different level intervals according to environmental protection requirements, and assign a basic priority value to each vehicle. The basic priority value is inversely proportional to the vehicle's emission level;

[0015] S5. Combine the optimal entry time window and the basic priority value to jointly calculate the priority dispatch factor. Based on the priority dispatch factor, all vehicles are sorted to generate the optimal dispatch sequence, and all parameters of the beat estimation model and vehicle emission classification model are updated in real time.

[0016] S6. Push the corresponding recommended entry time to each vehicle. The station dispatching system dynamically allocates loading slots and updates vehicle queue information based on the optimal dispatching sequence and the real-time loading slot status of the station area.

[0017] S7. When a vehicle approaches the electronic fence, the real-time location of the vehicle, the recommended entry time and the station area scheduling status are compared. When the entry conditions are met, the corresponding entry scheduling instruction is triggered, and the scheduling system guides the vehicle to enter the station for loading.

[0018] Optionally, the interactive data includes the real-time location of the vehicle, basic information of the vehicle, emission level of the vehicle, loading start and end time, historical loading records, usage status of the loading crane positions in the station area and the number of vehicles that can be accommodated, and the preprocessing includes time window division and abnormal data elimination.

[0019] Optionally, the S2 specifically includes:

[0020] S21. Extract the historical loading records of each loading crane from the standardized data set, and calculate the average loading cycle of each crane within a unit time based on the loading start and end times:

[0021]

[0022] Among them, Ti represents the average loading cycle of the i-th crane position, n i Indicates the number of valid loading records of the i-th crane position within the statistical time window, and They represent the loading start and end time of the i-th crane position in the j-th record respectively;

[0023] S22. Based on the vehicle category information in the historical loading records, a vehicle category-crane position loading efficiency matrix is ​​constructed:

[0024]

[0025] Among them, E kl represents the average loading efficiency of vehicle type k at crane position l, M kl is the total number of vehicle loading records of category k at crane position l, is the number of vehicles loaded in the mth record, and Respectively represent the start and end time of the mth record;

[0026] S23. Set the sliding time window length to W, perform real-time update operations on the latest interactive data in the window, and reconstruct the loading efficiency matrix And the data of different periods are fused using the weighted average method:

[0027]

[0028] Among them, ω t represents the time weight at time t, λ is the time attenuation coefficient, T now is the current station area dispatching system time, t∈W is each time point in the sliding window;

[0029] S24, based on the updated Construct a loading cycle estimation model for each type of vehicle:

[0030]

[0031] Among them, B k It represents the loading cycle prediction value of vehicle category k, which represents the loading time required for each vehicle. is the updated loading efficiency, φ kl is the current available state factor of crane position l, φ kl ∈{0,1}, the value is 1 means the crane position is available, L k represents the number of all available crane slots for vehicle category k.

[0032] Optionally, the S3 specifically includes:

[0033] S31. Calculate the Euclidean distance between the vehicle v and the station area based on the real-time positioning coordinates of each transport vehicle and the fixed coordinates of the station area:

[0034]

[0035] Among them, D v Indicates the current straight-line distance between vehicle v and the station area, (x v ,y v ) represents the current position of the vehicle, (x s ,y s ) represents the coordinates of the center of the station area;

[0036] S32. Obtain the average road speed S in the vehicle's driving path. v , combined with the straight-line distance D v , use the weighted moving average method to estimate the time required for the vehicle to arrive at the station:

[0037]

[0038] in, is the estimated arrival time of vehicle v, is the length of the i-th segment in the path, is the current speed of the road section, α i is the time period weight of the road section, and n represents the number of path segments;

[0039] S33, combined with the loading cycle prediction value B k , according to the category k to which the vehicle belongs, With B k Jointly calculate the optimal entry time window for vehicle v

[0040]

[0041] in, is the earliest recommended entry time for vehicle v, is the latest recommended entry time for vehicle v, δ is the time window adjustment coefficient, and are the earliest and latest recommended entry times for vehicle v, respectively.

[0042] Optionally, the S4 specifically includes:

[0043] S41. Collect emission-related data of all vehicles waiting in line and establish a vehicle emission vector set:

[0044]

[0045] in, is the vehicle emission vector set, Rv represents the emission vector of vehicle v, are the specific emissions of nitrogen oxides, particulate matter, carbon monoxide and hydrocarbons of the vehicle, is the collection of all vehicles currently waiting in line;

[0046] S42. Vehicle emission vector set Use a big data hierarchical clustering algorithm based on a Gaussian mixture model to build a vehicle emission level model and output a cluster center vector set:

[0047]

[0048] in, is the cluster center vector set, K is the total number of clusters of emission levels, are the mean values ​​of cluster center j on vehicle nitrogen oxides, particulate matter, carbon monoxide and hydrocarbon indicators, C j represents the cluster center of the jth emission level;

[0049] S43. For any vehicle v, calculate the emission vector R of vehicle v. v With each cluster center C j The Euclidean distance of:

[0050]

[0051] Among them, d vj is the Euclidean distance between vehicle v and cluster center j;

[0052] Assign vehicle class labels based on the minimum distance principle:

[0053]

[0054] Among them, L v ∈{1,2,…,K} is the emission level label corresponding to vehicle v. The smaller the value, the better the level. is the category of independent variable j when the emission level label value is the smallest;

[0055] S44, according to vehicle grade label L v Assign a basic priority value, which is inversely proportional to the vehicle level:

[0056]

[0057] Among them, P v is the basic priority value of vehicle v. The larger the value, the higher the scheduling priority and the higher the emission level, that is, the heavier the pollution. v The larger the P v The lower it is, the more environmentally friendly scheduling priority strategy can be implemented.

[0058] Optionally, the S5 specifically includes:

[0059] S51. Based on the optimal entry time window Calculate the desired entry center moment and estimate the entry time offset:

[0060]

[0061] in, is the earliest recommended entry time for vehicle v, is the latest recommended entry time for vehicle v, is the recommended entry center time point of vehicle v, that is, the midpoint of the entry time window, T now is the current station area dispatching system time, ΔT v represents the degree of entry delay of vehicle v;

[0062] S52. Combine the loading cycle prediction value, basic priority value, and entry time offset to jointly calculate the priority scheduling factor:

[0063]

[0064] Among them, v represents the comprehensive scheduling priority factor of vehicle v, B k is the loading cycle prediction value of vehicle category k, P v is the basic priority value of vehicle v, γ and β are scheduling weight parameters, and ∈ is a small constant to avoid the denominator being zero;

[0065] S53. Sort all vehicles according to their scheduling priority factors to obtain the optimal scheduling sequence:

[0066]

[0067] in, Indicates the priority factor Ψ v The vehicle queue is sorted from high to low, is the set of all vehicles currently waiting in line, ↓ indicates descending order, and Sort is the sorting operation;

[0068] S54, station area dispatching system according to the queue The cluster centers in the loading cycle estimation model and the vehicle emission classification model are updated in the order of , and the update rule uses the exponentially weighted moving average for dynamic correction:

[0069]

[0070] in, is the beat value of vehicle category k observed in the current period, is the beat value of vehicle category k predicted in the previous period, is the beat prediction value of vehicle category k after weighted update, is the emission mean vector of cluster center j observed in the current period, is the emission mean vector of cluster center j predicted in the previous period, is the updated emission mean vector of cluster center j, and α is the updated weight factor.

[0071] The beneficial effects of the present invention are:

[0072] First, the present invention introduces a big data dynamic modeling method based on a sliding time window to construct a real-time updated loading rhythm estimation model and a vehicle category-crane position loading efficiency matrix, overcoming the defects of the existing technology that the rhythm model is static and the loading efficiency fluctuations cannot be dynamically reflected. It realizes the adaptive adjustment of the rhythm prediction value as the operation status changes, effectively improves the matching degree between the scheduling rhythm and the on-site loading capacity, and avoids the occurrence of idle resources or waiting congestion.

[0073] Secondly, the present invention uses a weighted moving average algorithm and a multi-segment path traffic speed fusion strategy to accurately estimate the time it takes for vehicles to arrive at the station. At the same time, it introduces emission levels as key indicators into the scheduling priority system, and constructs an emission classification model and a green priority scheduling mechanism based on hierarchical clustering. While ensuring fairness and efficiency in queuing, it actively guides low-emission vehicles to enter first, effectively responding to the policy requirements of green ports and environmental protection management, and improving the intelligence and sustainability of system scheduling.

[0074] Finally, the present invention establishes a unified scheduling priority factor calculation model based on comprehensive consideration of beat prediction, arrival time offset and emission priority, realizes multi-dimensional factor collaborative sorting and dynamic scheduling, and dynamically drives the scheduling queue and updates the beat model and emission model parameters in combination with the scheduling results, forming a scheduling optimization closed loop with self-learning and self-adjusting capabilities. Through this mechanism, the queuing process of transport vehicles is more accurate and efficient, the scheduling response speed is significantly improved, the crane utilization rate and port traffic efficiency are improved simultaneously, which is significantly better than traditional scheduling. BRIEF DESCRIPTION OF THE DRAWINGS

[0075] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0076] Figure 1 This is a flow chart of a method for optimizing in-transit queuing of transport vehicles based on big data analysis proposed by the present invention. DETAILED DESCRIPTION

[0077] The present invention will now be described in further detail with reference to the accompanying drawings, which are simplified schematic diagrams that illustrate the basic structure of the present invention in a schematic manner.

[0078] refer to Figure 1 A method for optimizing in-transit queuing of transport vehicles based on big data analysis includes the following steps:

[0079] S1. Obtain and preprocess the interaction data of transport vehicles to generate a standardized data set;

[0080] S2. Calculate the average loading cycle for each crane position per unit time based on the standardized data set, construct a vehicle category-crane loading efficiency matrix and a takt estimation model for each category of vehicle, and dynamically adjust it based on the latest interaction data within the sliding time window to generate a loading takt prediction value for each category of vehicle;

[0081] S3. Using big data to drive all vehicles to be dispatched, the weighted moving average method is used to estimate the time required for each vehicle to arrive at the station based on the real-time location and road speed of each vehicle. Combined with the loading cycle prediction value, the optimal entry time window for the vehicle is calculated;

[0082] S4. Use a big data hierarchical clustering algorithm to build a vehicle emission classification model, score the emission levels of all vehicles in the queue, divide the emission levels into different level intervals according to environmental protection requirements, and assign a basic priority value to each vehicle. The basic priority value is inversely proportional to the vehicle's emission level;

[0083] S5. Combine the optimal entry time window and the basic priority value to jointly calculate the priority dispatch factor. Based on the priority dispatch factor, all vehicles are sorted to generate the optimal dispatch sequence, and all parameters of the beat estimation model and vehicle emission classification model are updated in real time.

[0084] S6. Push the corresponding recommended entry time to each vehicle. The station dispatching system dynamically allocates loading slots and updates vehicle queue information based on the optimal dispatching sequence and the real-time loading slot status of the station area.

[0085] S7. When a vehicle approaches the electronic fence, the real-time location of the vehicle, the recommended entry time and the station area scheduling status are compared. When the entry conditions are met, the corresponding entry scheduling instruction is triggered, and the scheduling system guides the vehicle to enter the station for loading.

[0086] The present invention optimizes the entire chain of the in-transit queuing process of transport vehicles, constructs a complete technical system from interactive data collection, dynamic beat modeling, accurate arrival time prediction, emission level grading assessment to multi-factor priority scheduling, and realizes the real-time, adaptive and environmentally friendly nature of the scheduling strategy. Through the above method, not only the accuracy of vehicle queuing and the utilization efficiency of crane resources are significantly improved, but also the congestion and long waiting time problems around the station area are effectively alleviated. At the same time, low-emission vehicles are guided to enter first, the overall green operation level and scheduling intelligence capabilities are improved, and the comprehensive beneficial effects of high efficiency, safety and environmental protection are achieved.

[0087] In this embodiment, the interactive data includes the real-time location of the vehicle, basic information of the vehicle, emission level of the vehicle, loading start and end time, historical loading records, the usage status of the loading crane positions in the station area and the number of vehicles that can be accommodated. The preprocessing includes time window division and abnormal data elimination.

[0088] The present invention significantly improves the accuracy and timeliness of data by performing pre-processing operations such as time window division and abnormal data elimination on multi-dimensional interactive data including the real-time location of vehicles, basic information of vehicles, vehicle emission levels, loading start and end times, historical loading records, and the usage status of loading crane positions in the station area and the number of vehicles that can be accommodated. It provides a reliable foundation for subsequent beat modeling and scheduling optimization, effectively ensures the scheduling model's sensitive response capability to actual on-site operating conditions, and improves the scheduling accuracy and robustness of the overall system.

[0089] In this embodiment, S2 specifically includes:

[0090] S21. Extract the historical loading records of each loading crane from the standardized data set, and calculate the average loading cycle of each crane within a unit time based on the loading start and end times:

[0091]

[0092] Among them, T i represents the average loading cycle of the i-th crane position, n i Indicates the number of valid loading records of the i-th crane position within the statistical time window, and They represent the loading start and end time of the i-th crane position in the j-th record respectively;

[0093] S22. Based on the vehicle category information in the historical loading records, a vehicle category-crane position loading efficiency matrix is ​​constructed:

[0094]

[0095] Among them, E kl represents the average loading efficiency of vehicle type k at crane position l, Mkl is the total number of vehicle loading records of category k at crane position l, is the number of vehicles loaded in the mth record, and Respectively represent the start and end time of the mth record;

[0096] S23. Set the sliding time window length to W, perform real-time update operations on the latest interactive data in the window, and reconstruct the loading efficiency matrix And the data of different periods are fused using the weighted average method:

[0097]

[0098] Among them, ω t represents the time weight at time t, λ is the time attenuation coefficient, T now is the current station area dispatching system time, t∈W is each time point in the sliding window;

[0099] S24, based on the updated Construct a loading cycle estimation model for each type of vehicle:

[0100]

[0101] Among them, B k It represents the loading cycle prediction value of vehicle category k, which represents the loading time required for each vehicle. is the updated loading efficiency, φ kl is the current available state factor of crane position l, φ kl ∈{0,1}, the value is 1 means the crane position is available, L k represents the number of all available crane slots for vehicle category k.

[0102] This method dynamically mines and models multidimensional data from a standardized dataset, including loading start and end times, vehicle types, and crane slot resources, to establish a mapping relationship between vehicle type and crane slot efficiency. This method then updates the model in real time based on a sliding time window, enabling the loading cycle estimation model to adapt to dynamic fluctuations in on-site operational efficiency. This method avoids the disconnect between scheduling and actual operations caused by the use of static averages, effectively improving the accuracy and timeliness of cycle predictions, providing a reliable data foundation for subsequent scheduling, and significantly enhancing the scientific nature of vehicle scheduling and the efficiency of crane slot matching.

[0103] In this embodiment, S3 specifically includes:

[0104] S31. Calculate the Euclidean distance between the vehicle v and the station area based on the real-time positioning coordinates of each transport vehicle and the fixed coordinates of the station area:

[0105]

[0106] Among them, D v Indicates the current straight-line distance between vehicle v and the station area, (x v ,y v ) represents the current position of the vehicle, (x s ,y s ) represents the coordinates of the center of the station area;

[0107] S32. Obtain the average road speed S in the vehicle's driving path. v , combined with the straight-line distance D v , use the weighted moving average method to estimate the time required for the vehicle to arrive at the station:

[0108]

[0109] in, is the estimated arrival time of vehicle v, is the length of the i-th segment in the path, is the current speed of the road section, α i is the time period weight of the road section, and n represents the number of path segments;

[0110] S33, combined with the loading cycle prediction value B k , according to the category k to which the vehicle belongs, With B k Jointly calculate the optimal entry time window for vehicle v

[0111]

[0112] in, is the earliest recommended entry time for vehicle v, is the latest recommended entry time for vehicle v, δ is the time window adjustment coefficient, and are the earliest and latest recommended entry times for vehicle v, respectively.

[0113] This invention integrates real-time positioning data, road speeds, and historical route characteristics, and uses a weighted moving average algorithm to estimate the arrival time of each vehicle. This allows the system to accurately determine the time distribution of vehicle arrivals at the port. Furthermore, combined with the beat prediction values ​​for each type of vehicle, the optimal entry time window for each vehicle is scientifically set, effectively balancing the pace of vehicle entry and loading capacity. This effectively avoids congestion and waiting caused by concentrated entry, reduces ineffective queuing time, improves traffic efficiency within and outside the station area, and enhances the execution stability of the overall scheduling plan and the rationality of resource allocation.

[0114] In this embodiment, the S4 specifically includes:

[0115] S41. Collect emission-related data of all vehicles waiting in line and establish a vehicle emission vector set:

[0116]

[0117] in, is the vehicle emission vector set, R v represents the emission vector of vehicle v, are the specific emissions of nitrogen oxides, particulate matter, carbon monoxide and hydrocarbons of the vehicle, is the collection of all vehicles currently waiting in line;

[0118] S42. Vehicle emission vector set Use a big data hierarchical clustering algorithm based on a Gaussian mixture model to build a vehicle emission level model and output a cluster center vector set:

[0119]

[0120] in, is the cluster center vector set, K is the total number of clusters of emission levels, are the mean values ​​of cluster center j on vehicle nitrogen oxides, particulate matter, carbon monoxide and hydrocarbon indicators, C j represents the cluster center of the jth emission level;

[0121] S43. For any vehicle v, calculate the emission vector R of vehicle v. v With each cluster center C j The Euclidean distance of:

[0122]

[0123] Among them, d vj is the Euclidean distance between vehicle v and cluster center j;

[0124] Assign vehicle class labels based on the minimum distance principle:

[0125]

[0126] Among them, L v ∈{1,2,…,K} is the emission level label corresponding to vehicle v. The smaller the value, the better the level. is the category of independent variable j when the emission level label value is the smallest;

[0127] S44, according to vehicle grade label L v Assign a basic priority value, which is inversely proportional to the vehicle level:

[0128]

[0129] Among them, P v is the basic priority value of vehicle v. The larger the value, the higher the scheduling priority and the higher the emission level, that is, the heavier the pollution. v The larger the P v The lower it is, the more environmentally friendly scheduling priority strategy can be implemented.

[0130] This invention establishes a vehicle emission level model through a big data-driven hierarchical clustering algorithm, enabling quantitative scoring and classification of the emission characteristics of queued vehicles. It also converts emission levels into basic priority values, establishing a green scheduling mechanism centered on environmental compliance. This step not only meets the requirements for responding to environmental policies but also enables dynamic control of vehicle emission levels, enabling the system to proactively prioritize low-emission vehicles, effectively promoting the green transformation of the station's vehicle structure and enhancing the system's environmental friendliness, intelligence, and policy adaptability.

[0131] In this embodiment, the S5 specifically includes:

[0132] S51. Based on the optimal entry time window Calculate the desired entry center moment and estimate the entry time offset:

[0133]

[0134] in, is the earliest recommended entry time for vehicle v, is the latest recommended entry time for vehicle v, is the recommended entry center time point of vehicle v, that is, the midpoint of the entry time window, T now is the current station area dispatching system time, ΔT v represents the degree of entry delay of vehicle v;

[0135] S52. Combine the loading cycle prediction value, basic priority value, and entry time offset to jointly calculate the priority scheduling factor:

[0136]

[0137] Among them, v represents the comprehensive scheduling priority factor of vehicle v, B k is the loading cycle prediction value of vehicle category k, P v is the basic priority value of vehicle v, γ and β are scheduling weight parameters, and ∈ is a small constant to avoid the denominator being zero;

[0138] S53. Sort all vehicles according to their scheduling priority factors to obtain the optimal scheduling sequence:

[0139]

[0140] in, Indicates the priority factor Ψ v The vehicle queue is sorted from high to low, is the set of all vehicles currently waiting in line, ↓ indicates descending order, and Sort is the sorting operation;

[0141] S54, station area dispatching system according to the queue The cluster centers in the loading cycle estimation model and the vehicle emission classification model are updated in the order of , and the update rule uses the exponentially weighted moving average for dynamic correction:

[0142]

[0143] in, is the beat value of vehicle category k observed in the current period, is the beat value of vehicle category k predicted in the previous period, is the beat prediction value of vehicle category k after weighted update, is the emission mean vector of cluster center j observed in the current period, is the emission mean vector of cluster center j predicted in the previous period, is the updated emission mean vector of cluster center j, and α is the updated weight factor.

[0144] This invention comprehensively quantifies the optimal vehicle entry time window and basic priority value, constructs a unified priority scheduling factor, completes the scheduling sorting of all waiting vehicles, and generates a dynamically executable optimal scheduling sequence, further improving the intelligence and precision of scheduling execution. Simultaneously, the scheduling system implements real-time parameter updates for the beat estimation model and emission classification model on this basis, forming a closed-loop self-regulating mechanism for scheduling prediction and execution. This method can continuously optimize model accuracy and scheduling strategy adaptability during actual operation, effectively improving the scheduling system's response speed and robustness to operational dynamics, and ensuring the continuous optimization and long-term stable operation of the scheduling strategy.

[0145] Example 1:

[0146] In order to verify the feasibility of the present invention in implementation, the present invention is applied to the liquid bulk cargo shipping scenario of a large coastal comprehensive port. The port undertakes the transshipment task of raw materials and refined oil products for multiple petrochemical enterprises in the region, with an average annual material throughput of more than 6 million tons, an average daily loading vehicle of more than 300 vehicles, and up to 500 vehicles during peak periods. The loading operation in the port area relies on 12 fixed crane positions. For a long time, due to the use of traditional timed queuing scheduling methods, there have been problems such as concentrated arrival of vehicles, disorderly queuing, large fluctuations in loading efficiency, and uneven resource utilization. Especially during the peak transportation period in summer, the periphery of the station area is seriously congested, some drivers have to wait for more than 3 hours, the actual utilization rate of the loading operation crane position is less than 80%, the system scheduling efficiency is out of line with the actual operating capacity, and it seriously restricts the overall operational efficiency of the port area.

[0147] In the above scenario, the system deploys the in-transit queuing optimization method for transport vehicles based on big data analysis proposed in the present invention. All vehicles entering the port are connected to the scheduling system after the pick-up plan is generated. The system automatically collects data such as the real-time location, emission level, vehicle type, historical loading behavior, etc. of the vehicle, and links the port loading operation platform to obtain the historical loading efficiency, current available status and operation queue of each crane position. The system first processes the historical loading cycle through a sliding window mechanism, filters out abnormal data caused by faults or abnormal operations, and constructs a loading efficiency matrix between vehicle categories and crane positions. Taking a certain dispatch as an example, the system constructed independent loading beat models for three typical vehicles, among which the average beat of Class A vehicles in the past 7 days was 21.3 minutes, that of Class B vehicles was 25.7 minutes, and that of Class C vehicles was 18.9 minutes. Combined with the latest 3-hour window data, the system automatically adjusts the beat estimate and controls the beat prediction error within ±1.2 minutes.

[0148] The system simultaneously introduces an emission graded scheduling mechanism, which constructs a 5-level emission grade by performing data standardization and cluster analysis on the emission parameters of all vehicles waiting in line, and assigns a corresponding basic priority value to each vehicle. During the execution of the scheduling strategy, the system combines the predicted arrival time of the vehicle, the beat estimation model and the emission priority, and comprehensively calculates the scheduling priority factor to form a scheduling sorting sequence for all vehicles, and pushes personalized recommended entry time to each vehicle. In the case that the crane position is about to be idle, the system will trigger the pre-scheduling mechanism to notify the next vehicle to be loaded in advance to ensure seamless resource connection. During the entire scheduling process, the scheduling sequence will be dynamically updated in real time based on the latest interactive data. At the same time, the system will perform parameter self-learning on the beat model and the emission model to improve the model accuracy and adaptability. The following table shows the changes in key indicators of the port area before and after the application of the method of the present invention:

[0149] Table 1 Comparison data before and after vehicle intelligent scheduling optimization

[0150]

[0151] After 30 consecutive days of actual testing, the system handled a total of 9,843 transport vehicles in the port area, and the average waiting time in queues was shortened from 92 minutes before deployment to 36 minutes, a reduction of 60.87%. At the same time, the average utilization rate of loading crane positions increased from the original 78.2% to 91.5%, and the idle rate dropped significantly. Within the 3-kilometer electronic fence area outside the port area, the instantaneous number of vehicles dropped from 148 during the peak period to 64, and road traffic efficiency was significantly improved. The response rate of drivers receiving dispatch notifications increased from 68% of the original system to 94%, and the deviation time of dispatch command execution was controlled within ±3 minutes. In terms of environmental scheduling, during the execution of the system, the average waiting time of low-emission vehicles (emission levels 1 and 2) was shortened by about 19.4 minutes compared with high-emission vehicles (levels 4 and 5), which significantly demonstrated the effect of the green priority strategy.

[0152] It can be seen from the above implementation results that the application of the present invention has significantly improved the intelligence level of in-transit queue management of transport vehicles, achieved a comprehensive improvement in dispatching efficiency, precise matching of station resources and effective implementation of environmental protection policies, and has good promotion value and application prospects.

[0153] The above description is only a preferred specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any technician familiar with the technical field, within the technical scope disclosed by the present invention, who makes equivalent replacements or changes based on the technical solution and inventive concept of the present invention, should be covered by the scope of protection of the present invention.

Claims

1. A method for optimizing in-transit queuing of transport vehicles based on big data analysis, characterized in that: The steps include: S1. Obtain and preprocess the interaction data of transport vehicles to generate a standardized data set; S2. Calculate the average loading cycle per unit time for each crane position based on the standardized data set, construct a vehicle category-crane position loading efficiency matrix and a cycle time estimation model, and generate a predicted loading cycle time for each category of vehicle; S3. Based on the real-time location of vehicles and road speeds, the weighted moving average method is used to estimate the time required for each vehicle to arrive at the station. Combined with the loading cycle prediction value, the optimal entry time window is determined; S4. Build an emission classification model using a big data hierarchical clustering algorithm to score vehicles and classify them into different levels, generating a basic priority value that is inversely proportional to the emission level. S5. Combine the optimal entry time window and the basic priority value to jointly calculate the priority scheduling factor, sort and generate the optimal scheduling sequence, and update all parameters of the beat estimation model and vehicle emission classification model in real time; S6. Push the recommended entry time to each vehicle. The station dispatching system dynamically allocates loading slots and simultaneously updates the vehicle queue information based on the optimal dispatching sequence and real-time loading slot status. S7. When a vehicle approaches the electronic fence, the station dispatching system verifies whether the vehicle's real-time position matches the recommended entry time. If the conditions are met, an entry instruction is issued to guide the vehicle into the loading operation process.

2. The method for optimizing in-transit queuing of transport vehicles based on big data analysis according to claim 1, characterized in that: The interactive data includes the real-time location of the vehicle, basic information of the vehicle, emission level of the vehicle, loading start and end time, historical loading records, the usage status of the loading crane positions in the station area and the number of vehicles that can be accommodated. The preprocessing includes time window division and abnormal data elimination.

3. The method for optimizing in-transit queuing of transport vehicles based on big data analysis according to claim 1, characterized in that: The S2 specifically includes: S21. Extract the historical loading records of each loading crane from the standardized data set, and calculate the average loading cycle of each crane within a unit time based on the loading start and end times: Among them, T i represents the average loading cycle of the i-th crane position, n i Indicates the number of valid loading records of the i-th crane position within the statistical time window, and They represent the loading start and end time of the i-th crane position in the j-th record respectively; S22. Based on the vehicle category information in the historical loading records, a vehicle category-crane position loading efficiency matrix is ​​constructed: Among them, E kl represents the average loading efficiency of vehicle type k at crane position l, M kl is the total number of vehicle loading records of category k at crane position l, is the number of vehicles loaded in the mth record, and Respectively represent the start and end time of the mth record; S23. Set the sliding time window length to W, perform real-time update operations on the latest interactive data in the window, and reconstruct the loading efficiency matrix And the data of different periods are fused using the weighted average method: Among them, ω t represents the time weight at time t, λ is the time attenuation coefficient, T now is the current station area dispatching system time, t∈W is each time point in the sliding window; S24, based on the updated Construct a loading cycle estimation model for each type of vehicle: Among them, B k It represents the loading cycle prediction value of vehicle category k, which represents the loading time required for each vehicle. is the updated loading efficiency, φ kl is the current available state factor of crane position l, φ kl ∈{0,1}, the value is 1 means the crane position is available, L k represents the number of all available crane slots for vehicle category k.

4. The method for optimizing in-transit queuing of transport vehicles based on big data analysis according to claim 1, characterized in that: The S3 specifically includes: S31. Calculate the Euclidean distance between the vehicle v and the station area based on the real-time positioning coordinates of each transport vehicle and the fixed coordinates of the station area: Among them, D v Indicates the current straight-line distance between vehicle v and the station area, (x v ,y v ) represents the current position of the vehicle, (x s ,y s ) represents the coordinates of the center of the station area; S32. Obtain the average road speed S in the vehicle's driving path. v , combined with the straight-line distance D v , use the weighted moving average method to estimate the time required for the vehicle to arrive at the station: in, is the estimated arrival time of vehicle v, is the length of the i-th segment in the path, is the current speed of the road section, α i is the time period weight of the road section, and n represents the number of path segments; S33, combined with the loading cycle prediction value B k , according to the category k to which the vehicle belongs, With B k Jointly calculate the optimal entry time window for vehicle v in, is the earliest recommended entry time for vehicle v, is the latest recommended entry time for vehicle v, δ is the time window adjustment coefficient, and are the earliest and latest recommended entry times for vehicle v, respectively.

5. The method for optimizing in-transit queuing of transport vehicles based on big data analysis according to claim 1 is characterized in that: The S4 specifically includes: S41. Collect emission-related data of all vehicles waiting in line and establish a vehicle emission vector set: in, is the vehicle emission vector set, R v represents the emission vector of vehicle v, are the specific emissions of nitrogen oxides, particulate matter, carbon monoxide and hydrocarbons of the vehicle, is the collection of all vehicles currently waiting in line; S42. Vehicle emission vector set Use a big data hierarchical clustering algorithm based on a Gaussian mixture model to build a vehicle emission level model and output a cluster center vector set: in, is the cluster center vector set, K is the total number of clusters of emission levels, are the mean values ​​of cluster center j on vehicle nitrogen oxides, particulate matter, carbon monoxide and hydrocarbon indicators, C j represents the cluster center of the jth emission level; S43. For any vehicle v, calculate the emission vector R of vehicle v. v With each cluster center C j The Euclidean distance of: Among them, d vj is the Euclidean distance between vehicle v and cluster center j; Assign vehicle class labels based on the minimum distance principle: Among them, L v ∈{1,2,…,K} is the emission level label corresponding to vehicle v. The smaller the value, the better the level. is the category of independent variable j when the emission level label value is the smallest; S44, according to vehicle grade label L v Assign a basic priority value, which is inversely proportional to the vehicle level: Among them, P v is the basic priority value of vehicle v. The larger the value, the higher the scheduling priority and the higher the emission level, that is, the heavier the pollution. v The larger the P v The lower it is, the more environmentally friendly scheduling priority strategy can be implemented.

6. The method for optimizing in-transit queuing of transport vehicles based on big data analysis according to claim 1, characterized in that: The S5 specifically includes: S51. Based on the optimal entry time window Calculate the desired entry center moment and estimate the entry time offset: in, is the earliest recommended entry time for vehicle v, is the latest recommended entry time for vehicle v, is the recommended entry center time point of vehicle v, that is, the midpoint of the entry time window, T now is the current station area dispatching system time, ΔT v represents the degree of entry delay of vehicle v; S52. Combine the loading cycle prediction value, basic priority value, and entry time offset to jointly calculate the priority scheduling factor: Among them, v represents the comprehensive scheduling priority factor of vehicle v, B k is the loading cycle prediction value of vehicle category k, P v is the basic priority value of vehicle v, γ and β are scheduling weight parameters, and ∈ is a small constant to avoid the denominator being zero; S53. Sort all vehicles according to their scheduling priority factors to obtain the optimal scheduling sequence: in, Indicates the priority factor Ψ v The vehicle queue is sorted from high to low, is the set of all vehicles currently waiting in line, ↓ indicates descending order, and Sort is the sorting operation; S54, station area dispatching system according to the queue The cluster centers in the loading cycle estimation model and the vehicle emission classification model are updated in the order of , and the update rule uses the exponentially weighted moving average for dynamic correction: in, is the beat value of vehicle category k observed in the current period, is the beat value of vehicle category k predicted in the previous period, is the beat prediction value of vehicle category k after weighted update, is the emission mean vector of cluster center j observed in the current period, is the emission mean vector of cluster center j predicted in the previous period, is the updated emission mean vector of cluster center j, and α is the updated weight factor.

Images