A multi-modal bulk order transportation shipping method with optimal price matching

Abstract

The present application belongs to the technical field of logistics and machine learning, and discloses a multimodal transport batch order transportation and shipping method for optimal price matching, comprising: S1, multimodal transport price data acquisition and standardized processing; S2, comprehensive optimal price matching and candidate carrier screening; S3, intelligent batch order splitting of orders; S4, integrated scheduling of multimodal transport resources; S5, global cost accounting optimization and multimodal transport scheme generation; S6, dynamic adaptive adjustment of the shipping scheme. The present application overcomes the technical problems that the regulation and control mechanism is difficult to adapt to the complex demands of multimodal transport, and when the transport capacity is tight or the node transport capacity is insufficient, it is difficult to timely regulate and control, which easily causes supply and demand imbalance and node congestion.

Description

Technical Field

[0001] This invention belongs to the field of logistics and machine learning technology, specifically relating to a multimodal transport bulk order shipping method with optimal price matching. Background Technology

[0002] Currently, for optimizing the transportation and dispatch of bulk orders in multimodal transport, the industry has developed a technical solution that combines dynamic pricing and capacity scheduling. For example, patent document CN117217480A discloses a method for intelligent scheduling and pricing of multimodal transport, which realizes technical functions such as multi-source freight rate collection, basic dynamic pricing, unified capacity scheduling and transportation route optimization. It is the mainstream technical approach in the logistics field to solve the problem of multimodal transport scheduling and matching.

[0003] However, current technologies rely solely on freight rates as the basis for carrier selection and ranking, failing to comprehensively consider factors such as batch size, timeliness requirements, and carrier fulfillment capabilities. The established control mechanisms are ill-suited to the complex demands of multimodal transport. When capacity is tight or node capacity is insufficient, timely adjustments are difficult. Furthermore, multimodal transport must consider potential congestion at transit points, easily leading to supply-demand imbalances and node congestion. Addressing these issues involves fragmented multimodal transport resource scheduling, lacking integrated coordination; adjusting capacity and transit nodes, but unreasonable order splitting hinders node adaptation; the complex multimodal transport process currently lacks global cost optimization, resulting in excessively high costs; and the involvement of multiple modes of transport weakens overall dynamic adjustment capabilities, hindering real-time monitoring and rapid response to logistics conditions. Summary of the Invention

[0004] The purpose of this invention is to provide a multimodal transport bulk order shipping method with optimal price matching, in order to solve the technical problems of existing technologies that cannot comprehensively consider the influencing factors involved in logistics regulation, such as batch size, time requirements, and carrier fulfillment capabilities. The established regulation mechanism is difficult to adapt to the complex needs of multimodal transport. When there is a shortage of transport capacity or insufficient capacity at nodes, it is difficult to make timely adjustments. For multimodal transport, the possibility of congestion at node transfers also needs to be considered, which easily leads to supply and demand imbalance and node congestion.

[0005] The optimal price matching multimodal transport bulk order shipping method includes the following steps: S1. Multimodal transport freight rate data collection and standardization processing; S2. Perform comprehensive optimal price matching and candidate carrier screening; This step includes the following sub-steps: S2.1 Establish a formula for the comprehensive optimal matching freight rate, which should at least incorporate factors such as batch size, mode of transport, time requirement, and number of transfers. S2.2 Establish an adaptive price matching threshold system to initially screen carrier quotations based on allowable deviation thresholds; S2.3 Calculate a comprehensive score for each candidate quote based on the optimal matching freight rate combined with the multi-factor intermodal transport supply and demand tension index; S2.4 Sort the preliminary candidate set from low to high according to comprehensive score, and select the top N candidates as the final candidate set; S3. Perform intelligent batch order splitting on orders; S4, integrated scheduling of multimodal transport resources; S5. Generation of shipping schemes based on global cost accounting optimization and multi-objective optimization; S6. The shipping plan is dynamically and adaptively adjusted.

[0006] Preferably, in step S2.1, the basic comprehensive optimal matching freight rate formula is used, as follows:

[0007] in, This represents the overall optimal matching freight rate for the basic type. This represents the standardized base freight rate. Indicates the batch size factor. Indicates the mode of transport factor. Indicates the timeliness requirement factor. Indicates the number of transfers factor; Batch size factor, range of values Based on historical bulk freight rate data, a combination of piecewise functions and nonlinear regression is used to fit the bulk size curve, thereby determining the bulk size factor. The mode of transport factor characterizes the differences between different modes of transport. Timeliness requirement factor, range of values By analyzing the relationship between the timeliness level and premium level of historical orders, a nonlinear function that monotonically increases with the timeliness level is constructed for calculation to obtain the timeliness requirement factor; Transfer frequency factor, range of values It also characterizes transit service fees and transit risk costs, and applies a penalty coefficient to schemes that exceed the experience threshold for the number of transits.

[0008] Preferably, in step S2.1, a performance factor and a carbon emission factor are introduced into the basic comprehensive optimal matching freight rate formula to form an extended comprehensive optimal matching freight rate formula, as follows:

[0009] in, This represents the extended, comprehensive, optimal matching freight rate. Indicates the performance factor. Indicates carbon emission factor; Performance factor, range of values Scoring is based on the carrier's historical on-time performance, damage rate, and claims history; Carbon emission factor, range of values The carbon emissions per unit of transport capacity for each mode of transportation are converted into carbon costs.

[0010] Preferably, in step S2.2, the allowable deviation threshold between the candidate freight rate and the optimal matching freight rate is automatically adjusted based on market volatility, carrier concentration, and order type. The permissible deviation range of market volatility intensity is calculated based on the volatility estimation model, and the daily volatility of freight rates is calculated using the moving average volatility algorithm. The formula is as follows:

[0011] in, P represents volatility. i and P i-1 These represent the historical freight rates for day i and day i-1, respectively. Based on the adaptively adjusted allowable deviation threshold, all carrier quotes with deviations from the optimal matching freight rate within the threshold range are selected to form a preliminary candidate set.

[0012] Preferably, in step S2.3, the comprehensive scoring function for the carrier's quotation is defined as follows:

[0013] Here, "Score" represents the overall score. Provide a quote for the carrier. This represents the extended optimal matching rate; the basic optimal matching rate can also be used. , The Multi-Factor Intermodal Transport Supply and Demand Tension Index reflects the supply and demand status of transportation lines and the carrying capacity of key nodes. , Indicates corresponding to respectively The weighting coefficients of the multi-factor intermodal transport supply and demand tension index, This indicates the degree of difference between the quoted price and the best matching freight rate; The formula for calculating the multi-factor intermodal transport supply and demand tension index is:

[0014] Among them, freight rate deviation characteristics , The current freight rate, and These are the minimum and maximum freight rates for this route over the past 30 days. Characteristics of supply sufficiency , Indicates remaining transport capacity. Indicates total transport capacity; Node load characteristics , For node utilization, To adjust the parameters; Congestion risk characteristics , ; The corresponding freight rate deviation characteristics are as follows: Characteristics of supply sufficiency Node load characteristics Congestion risk characteristics The weighting coefficients.

[0015] Preferably, step S3 includes: S3.1 Set splitting constraints, including constraints on cargo dimension, transportation capacity dimension, timeliness dimension, cost dimension, and node capacity. The node capacity constraint is: the loading and unloading capacity, storage capacity, and parking space / berth capacity of each transfer node within a given time window shall not be exceeded by the superposition of splitting schemes. For nodes that are close to saturation, further diversion shall be automatically suppressed by the penalty coefficient. S3.2 Order splitting based on clustering and heuristic correction techniques; S3.3 Sub-order merging and stability optimization: After obtaining a feasible splitting scheme, a second merging optimization is performed on sub-orders with similar destinations, consistent transportation time, and matching freight rates after splitting. Step S3.2 includes the following steps: S3.2.1. Using K-means or density clustering algorithms, with cargo destination, transportation timeliness, freight rate gradient, and capacity adaptability as clustering features, batch orders are pre-divided into several initial sub-order clusters; S3.2.2 For each sub-order cluster, based on the integer programming model and metaheuristic algorithm, calculate the optimal loading combination under the current transportation capacity resource conditions to ensure that the loading rate and transportation capacity constraints are not exceeded. S3.2.3. Perform cost and timeliness calculations on the split sub-orders. If the total transportation cost exceeds the constraint threshold or the node capacity is limited, iterative corrections are made by adjusting the cluster center and merging / splitting local clusters to form a convergent split solution.

[0016] Preferably, step S4 includes: S4.1 Abstract the transportation capacity resources of each mode of transportation into a three-layer structure of capacity node - transportation link - spatiotemporal state, and construct a spatiotemporal extended graph model; S4.2 Optimization of node connection time; S4.3 Batch capacity allocation: Based on the optimization results of the capacity resource map and node connection, a combination strategy of greedy algorithm, integer programming and rolling correction is adopted to allocate the split sub-orders to the optimal intermodal capacity resources.

[0017] Preferably, in step S4.1, the node attributes include loading and unloading capacity, storage capacity, working hours, and historical congestion index; the transportation link represents the feasible transportation path between nodes, and the link attributes include transportation mode, mileage, standardized freight rate, average transit time, transit time fluctuation range, and carbon emission level; a visualized multimodal transport capacity resource map is formed based on the spatiotemporal extended graph model, and dynamic information such as real-time freight rate, remaining capacity, and node queuing time is continuously superimposed; a time series prediction module is introduced on the multimodal transport capacity resource map, and the corresponding prediction method includes: performing sliding window modeling on the time series including freight rate, remaining capacity, and queuing time of key trunk lines and hub nodes to obtain short-term predicted values ​​and confidence intervals; the prediction results and the current observation values ​​are used as inputs for scheduling optimization.

[0018] Preferably, step S4.2 addresses the pain points of long waiting times and poor connections in multimodal transport cross-node connections by designing a node connection time optimization algorithm; this step specifically includes: S4.2.1. With the goals of minimizing the total transportation time of sub-orders, minimizing transit waiting time, and balancing node load, establish a time window constraint model based on a spatiotemporal expansion graph. S4.2.2. Based on the transportation route of the sub-order, accurately match the departure time of short-distance road transfer, railway freight train, waterway shipping schedule, and air cargo space, and control the deviation between the arrival time of short-distance road transfer trucks at railway freight stations and the departure time of railway freight trains within a set time window, and control the deviation between the arrival time of waterway shipping cargo and the arrival time of road transshipment capacity within another set time window. S4.2.3 For time periods when nodes are about to become congested, the algorithm automatically balances the load through time shifting, path detours, and early / late shipments to avoid peak accumulation at transit nodes.

[0019] Preferably, step S5 includes: S5.1 Optimize global cost accounting, establish a global cost accounting system for multimodal transport batch orders, comprehensively consider cost factors in all stages, and introduce multidimensional cost vectors; S5.2 Multi-objective optimization of shipping schemes: The multi-objective optimization functions are global total transportation cost optimization, transportation time efficiency optimization, capacity resource utilization improvement and carbon emission control. The non-dominated sorting genetic algorithm is used to optimize and solve the candidate capacity combination and transportation route. S5.3 Shipping plan decision generation: Based on the user's core needs, weights are set for each objective. The shipping plan is comprehensively scored using a weighted scoring method or analytic hierarchy process. A scenario-adaptive weight adjustment mechanism is introduced to finally generate the shipping plan.

[0020] The technical advantages of this invention are as follows: This invention constructs a comprehensive optimal price matching algorithm that integrates batch size factors, transportation mode factors, timeliness requirement factors, and transshipment frequency factors, forming an optimal price matching model, thereby comprehensively considering these influencing factors involved in logistics regulation. A multi-factor intermodal transport supply and demand tension index is introduced in the price matching stage to comprehensively reflect the supply and demand status of routes and node carrying capacity, achieving a quantitative assessment of the degree of supply and demand tension of routes, node load, and congestion risk. It can automatically identify and eliminate low-price, high-risk carriers and highly congested routes; thus solving the technical problem in existing technologies where the regulation mechanism is difficult to adapt to the complex needs of multimodal transport, and when capacity is tight or node carrying capacity is insufficient, it is difficult to regulate in a timely manner, easily leading to supply and demand imbalance and node congestion. The model also includes an adaptive price matching threshold system, which can effectively improve the accuracy of price matching.

[0021] This invention achieves multi-objective optimization of multimodal transport bulk order shipments based on an optimal price matching model, thereby realizing integrated scheduling of multimodal transport resources. This significantly improves the utilization rate of intermodal resources and reduces total transportation costs. After improving the order splitting method, this invention significantly increases the effective loading rate of logistics and reduces congestion rates at transit nodes. This invention can also effectively perform global cost accounting for multimodal transport bulk order shipments while ensuring accounting accuracy, which also helps reduce total costs. Combined with the dynamic re-matching and re-scheduling steps of this invention, it can also respond promptly, reduce losses caused by abnormal events, and effectively quantify the adjustment effects. Attached Figure Description

[0022] Figure 1 This is a flowchart illustrating a multimodal bulk order transportation and shipment method with optimal price matching according to the present invention. Detailed Implementation

[0023] The following detailed description of the embodiments, with reference to the accompanying drawings, will further illustrate the specific implementation of the present invention, in order to help those skilled in the art to have a more complete, accurate, and in-depth understanding of the inventive concept and technical solution of the present invention.

[0024] Existing technologies mostly employ fixed freight rates, failing to consider dynamic factors such as batch size, timeliness requirements, and carrier fulfillment capabilities. The lack of a suitable supply and demand control mechanism easily leads to supply-demand imbalances during periods of capacity shortage. Multimodal transport involves multiple modes of transportation, including road, rail, waterway, and air, but current related scheduling systems are often independent, resulting in severe information silos. Current logistics control relies solely on cargo category or weight, neglecting freight rate tiers, capacity capacity, and node capacity constraints, leading to congestion at transit points. Furthermore, complex logistics control results in a lack of overall cost optimization and significantly insufficient accuracy in overall cost accounting. Finally, because some modes of transport are prone to freight rate fluctuations and capacity changes, complex multimodal transport systems currently lack real-time monitoring mechanisms, leading to delays in anomaly detection and difficulty in ensuring the effectiveness of adjustments.

[0025] This invention provides a method for shipping multimodal bulk orders with optimal price matching, aiming to solve the problems of inefficient price matching, imbalanced allocation of transportation resources, and high cost of shipping schemes in multimodal transport scenarios. It achieves optimal price matching between orders and intermodal resources and global optimization of bulk shipments, thereby improving the cost-effectiveness and transportation efficiency of multimodal transport.

[0026] like Figure 1 As shown, the optimal price matching multimodal transport bulk order shipping method includes the following steps: S1. Multimodal transport freight rate data collection and standardization processing.

[0027] This step includes the following sub-steps: S1.1 Multi-source freight rate data collection.

[0028] This step involves building a freight rate data collection terminal for multimodal transport scenarios. It adopts a multi-channel access combined with streaming data collection and edge preprocessing to achieve unified access to the official freight rate interfaces of highway carriers, railway freight platforms, waterway and shipping platforms, and air freight service providers. At the same time, it introduces third-party freight matching platforms, spot freight rate index platforms, and carrier self-quoting terminals to collect data such as real-time floating freight rates, bulk order freight rate discounts, transshipment service fees, and surcharges (such as hazardous materials surcharges and remote area surcharges) from the freight market.

[0029] In the time dimension, an incremental data collection mechanism with a frequency of minutes or higher is adopted to record the timestamps, validity periods, and expiration conditions of freight rates. In the spatial dimension, the origin-destination OD pairs, transit nodes, and transportation mileage are finely labeled. A historical freight rate database is built simultaneously to record the freight rate fluctuation patterns during different seasons / time periods, holidays, peak and off-peak seasons, and major events. External influencing factors such as weather, port congestion index, and highway congestion index are also linked to form a spatiotemporally integrated, multi-source, heterogeneous, multi-dimensional freight rate database.

[0030] S1.2 Standardization of freight rate data.

[0031] This step unifies multi-source heterogeneous freight rate data into a standardized freight rate feature set with high comparability and reliability, providing a data foundation for subsequent price matching and cost optimization.

[0032] Freight rate data standardization algorithms include: Unified pricing benchmark conversion: The freight rates of different modes of transportation (by ticket, by vehicle, by ton, by container, etc.), different carriers, and different transportation segments are uniformly converted into a standardized freight rate index per unit weight / volume per unit distance, while retaining the original pricing rules as metadata for subsequent reverse restoration.

[0033] Discount and surcharge normalization: Bulk discounts, long-term agreement discounts, temporary promotional discounts, fuel surcharges, peak season surcharges, etc. are decomposed and modeled independently, and converted into stackable discount coefficients and surcharge coefficients, forming structured features that can be directly used in model calculations.

[0034] Intelligent identification of abnormal fares: Combining box plot statistical rules with Bayesian anomaly detection methods based on historical sequences, fares with extreme high prices, extreme low prices, and those suspected of being entered incorrectly are marked and corrected. Fares with obvious jumps but supported by reasonable external events (such as sudden flight restrictions or road closures) are marked as "abnormal but valid" for use by the dynamic adjustment module.

[0035] Quality assessment and confidence weighting: Data quality scores are calculated for different data sources based on data integrity, timeliness, and historical stability, and confidence weights are assigned to each fare record to provide a weighted calculation basis for the base fare in the subsequent optimal price matching model.

[0036] S1.3, Enhanced freight rate features and scenario labeling.

[0037] After completing the basic standardization, in order to improve the ability of the subsequent matching and optimization modules to characterize the differences in scenarios, the fare data is further enhanced and labeled.

[0038] After feature enhancement and tagging, the freight rate data yields the following features: Time period / rhythm pattern characteristics: The fare time series of the same route within the day, week, and month are clustered, and patterns such as "morning and evening peak, weekend, holiday, peak season" are encoded as discrete time period labels, and corresponding fluctuation intensity characteristics are introduced.

[0039] Event-sensitive features: Using an abnormal event knowledge base, add event tags to fare records affected by special events such as "traffic restrictions, flight closures, and large-scale events", and distinguish them from ordinary fluctuations in subsequent modeling.

[0040] Regional and route classification characteristics: Based on the freight volume, carrier concentration, and node level, the routes are divided into "trunk lines, branch lines, terminal lines" and "first-tier hub cities / second-tier nodes", forming a structured label that can be used to assess supply and demand tension.

[0041] Historical stability characteristics: The variance and extreme value ratio of freight rates for each route within a certain window are statistically analyzed to form three levels of stability labels: "stable / medium / high volatility", which provide input for subsequent adaptive threshold and robust matching strategies.

[0042] The aforementioned enhanced features, together with the basic standardized features, constitute the input feature space for subsequent model training and online matching, enabling price matching and scheduling decisions to more finely distinguish the behavioral patterns of different routes and time periods.

[0043] S2. Perform comprehensive optimal price matching and candidate carrier screening.

[0044] This step includes the following sub-steps: S2.1 Establish the formula for the comprehensive optimal matching freight rate. This step can set two formulas for the comprehensive optimal matching freight rate: a basic type that integrates factors such as batch size, mode of transport, timeliness requirements, and number of transshipments, as shown in the following formula:

[0045] in, This indicates the comprehensive best-matched freight rate for the basic type (unit: yuan / ton·km or yuan / cubic meter·km). This represents the standardized base freight rate. Indicates the batch size factor. Indicates the mode of transport factor. Indicates the timeliness requirement factor. This represents the number of transfers factor.

[0046] The standardized base freight rate is a real-time average freight rate weighted by confidence level for the corresponding transportation route and mode of transportation. It is obtained by summing the freight rates from each data source according to their quality weights.

[0047] Batch size factor, range of values Based on historical bulk freight rate data, a bulk size curve is fitted using a combination of piecewise functions and nonlinear regression to determine the bulk size factor. Specifically, this includes: 1) Segment Node Division: The historical order batch size is divided into several (5) intervals using the percentile method. These intervals are then combined with the pre-defined business boundary values, sorted, and deduplicated to obtain monotonically increasing segment nodes, forming several (5) intervals (left-open, right-closed intervals or closed intervals), thus determining the segment boundaries. For example: [0, 50 tons], (50, 200 tons], (200, 500 tons], (500, 1000 tons], (1000 tons, +∞), corresponding to discount coefficient baseline values ​​of 1.0, 0.9, 0.75, 0.65, and 0.6, respectively. The batch size factor reflects the price discount.

[0048] 2) Nonlinear regression: using an exponential function within each segmented interval. Perform fitting, where For batch size, Here, a, b, and c represent the discount factors, and a, b, and c represent the corresponding coefficients (including constants). The fitting method employs constrained nonlinear least squares minimization, with the following constraints: , , The fitted curve for that segment is obtained.

[0049] 3) Inter-segment connection: For any batch size, if the difference between adjacent segments at the common boundary exceeds the preset tolerance (e.g., 0.02), then linear interpolation is performed on the results of the two adjacent segments in the boundary neighborhood to obtain the final fitting result.

[0050] 4) Tiered batch rewards: For orders whose batch size exceeds the reward threshold, an additional discount reward will be given, with the maximum reward not exceeding 0.1.

[0051] The transportation mode factor characterizes the differences in unit transport cost, timeliness stability, and accessibility among different transportation modes; carbon emission-related impacts are represented by an independent carbon emission factor. Use standardized methods to avoid redundant measurements.

[0052] Timeliness requirement factor, range of values By analyzing the relationship between the timeliness level and premium level of historical orders, a nonlinear function that monotonically increases with the timeliness level is constructed for calculation to obtain the timeliness requirement factor. The methods for calculating the aforementioned nonlinear functions include: 1) Timeliness Level Classification: Order timeliness is divided into several (5) levels, specifying the total timeliness required for the entire order process. (Days) are mapped to levels. For example: Normal (7-15 days), Express (3-7 days), Super Express (1-3 days), Super Express (within 24 hours), Time Limit (within 12 hours).

[0053] 2) Constructing a nonlinear function: using a logarithmic function Calculation, where This is the standard timeframe (10 days). For practical timeliness requirements, .

[0054] 3) Adjust peak suppression factor: Add suppression factor for urgent orders (within 24 hours). For time-limited items (within 12 hours), an inhibitory factor is superimposed. To prevent excessively high timeliness premiums.

[0055] Transfer frequency factor, range of values It also characterizes transit service fees and transit risk costs (probability of damage and delay), and applies a penalty coefficient to schemes that exceed the experience threshold in terms of the number of transits.

[0056] Another type is the extended type, which introduces performance factors and carbon emission factors on the basis of the basic type's comprehensive optimal matching freight rate formula.

[0057] Performance Factor Based on the carrier's historical on-time performance rate, damage rate, and claims record, the score is converted into a performance factor. This allows for a reasonable cost incentive for highly reliable carriers. ), and impose cost penalties on carriers with poor reliability ( ).

[0058] carbon emission factors The carbon emissions per unit of transport capacity for each mode of transportation are converted into carbon costs and then transformed into carbon emission factors. By introducing carbon emission factors to control green preference, a synergistic optimization of cost and low carbon emissions can be achieved.

[0059] Thus, the extended comprehensive optimal matching freight rate formula is:

[0060] in, This represents the extended overall optimal matching freight rate (unit: yuan / ton·km or yuan / cubic meter·km).

[0061] S2.2 Establish an adaptive price matching threshold system to initially screen carrier quotations based on allowable deviation thresholds.

[0062] This step automatically adjusts the candidate freight rates and the best-matched freight rate (e.g., extended best-matched freight rate) based on market volatility, carrier concentration, and order type. The permissible deviation threshold between ( ). Specifically, it includes: The permissible deviation range for market volatility intensity is calculated based on a volatility estimation model. The daily volatility of historical 30-day freight rates is calculated using the Moving Average Volatility (MAV) algorithm, with the following formula:

[0063] in, P represents volatility. i and P i-1 Let these represent the historical freight rates for day i and day i-1, respectively. For example: when At that time, it was determined to be for market stabilization, and adopted... Threshold; when At that time, it was determined to be for market stabilization, and adopted... Threshold; when At that time, it was determined to be a market with severe fluctuations, and adopted... Threshold.

[0064] The allowable deviation threshold is adjusted based on carrier concentration. For example, when the CR4 (market share of the top 4 carriers) is... At that time, the threshold is automatically relaxed. .

[0065] Adjust the allowable deviation threshold based on order type. For example, for orders requiring special control or orders under major customer agreements, the threshold is tightened. .

[0066] Based on the adaptively adjusted allowable deviation threshold, all carrier quotes with deviations from the optimal matching freight rate within the threshold range are selected to form a preliminary candidate set.

[0067] S2.3 Calculate a comprehensive score for each candidate quote based on the optimal matching freight rate combined with the multi-factor intermodal transport supply and demand tension index.

[0068] This step comprehensively reflects the supply and demand status of the transportation lines and the carrying capacity of nodes during the price matching stage through a multi-factor intermodal transport supply and demand tension index. The factor characteristics involved are selected from the characteristics of the corresponding lines and their associated nodes within a given time window, and are first normalized to ensure consistency in the dimensions and clear range of values ​​for each characteristic. Specifically: Freight rate deviation characteristics : Use min-max normalization, ,in The current freight rate, and These are the minimum and maximum freight rates for this route over the past 30 days.

[0069] Characteristics of supply sufficiency : Directly use the remaining capacity percentage .

[0070] Node load characteristics : Use sigmoid normalization, ,in For node utilization, To adjust the parameter (usually set to 10).

[0071] Congestion risk characteristics A probabilistic model based on historical congestion records. .

[0072] Based on the above normalized factor characteristics, a tension index is constructed, ensuring that its value range and direction are consistent (i.e., a larger value indicates tighter supply and demand and higher connection risk). The calculation formula is as follows:

[0073] in, This represents the multi-factor intermodal transport supply and demand tension index. The corresponding freight rate deviation characteristics are as follows: Characteristics of supply sufficiency Node load characteristics Congestion risk characteristics The weighting coefficients. ,i=1,2,3,4 and .

[0074] The weighting method can employ historical data playback fitting: the gradient descent method is used to minimize the prediction error of historical matching results to obtain the optimal weight combination; for each historical waybill, a four-dimensional vector is extracted at the time of shipment. ,Label A value of 1 indicates that the order experienced a significant delay or additional charge within the subsequent observation window (a threshold can be defined according to company rules); otherwise, it is 0. (Under constraints) , Minimize logical loss ,in It is sigmoid. ; Projected gradient descent is used (after each step, the gradient is adjusted). (Perform non-negative normalized projection) Iterate until convergence to obtain data-driven weights.

[0075] The weight determination method can also adopt the enterprise strategy configuration: the default weight is... , , , And adjust dynamically according to the company's risk appetite.

[0076] based on Establish robust matching rules: when When the value is too high, it indicates that the line supply is tight or the node is under high load, and it should be demoted or removed from the candidate ranking; when If the price is too low and there is both an abnormally low price and a low performance score, it will be judged as "low price, high risk" and will be downgraded accordingly; other routes within a reasonable range will be given priority.

[0077] As a specific implementation of robust matching rules, this step constructs a comprehensive scoring function based on the optimal matching freight rate, and... A comprehensive scoring function is incorporated for candidate screening and ranking. The comprehensive scoring function for carrier quotes is defined as follows:

[0078] Here, "Score" represents the overall score. Provide a quote for the carrier. This represents the extended optimal matching rate; the basic optimal matching rate can also be used. , The Multi-Factor Intermodal Transport Supply and Demand Tension Index reflects the supply and demand status of transportation lines and the carrying capacity of key nodes. , Indicates corresponding to respectively The weighting coefficients of the multi-factor intermodal transport supply and demand tension index, This indicates the degree of difference between the quoted price and the best matching freight rate. and A lower score indicates "smaller price deviation and more robust supply and demand," and is used to rank candidates for the final set. For example: , .

[0079] S2.4 Sort the preliminary candidate set from low to high according to comprehensive score, and select the top N candidates (usually N=5) as the final candidate set.

[0080] Ultimately, the above process forms a multi-dimensional weighted price matching candidate set, ensuring that the candidate set has both price advantages and supply and demand robustness.

[0081] The parameter calibration method for the comprehensive optimal price matching model formed in step S2 is as follows: Calibration data: Retain the "quote-transaction-performance result" triplet from the past few months and define the business loss function (such as weighted average price difference loss + delay penalty).

[0082] Adjustment rules: Fine-tune using grid search or Bayesian optimization on the validation set. (Keep the sum to 1) to minimize losses; record the version number and effect for each adjustment, and use it for online A / B testing or canary release.

[0083] Cold start: When a new line has no history, the group's default parameters are used until the sample size exceeds the threshold. (e.g., after 100 orders) a weight reassessment will be automatically triggered.

[0084] The training, validation, and deployment update process for the comprehensive optimal price matching model is as follows: Dataset Construction: Stratified sampling is performed by route (OD), mode of transport, and time period to construct a training / validation / test set (e.g., 7:2:1); the smallest sample granularity is the "order-candidate quote-result" triple, containing... And the final transaction and performance results.

[0085] Feature preprocessing: Calculate each factor according to the explicit procedure in 4.2.1; fill missing fields with the median of "same line, same week, same mode"; truncate outliers by 1% / 99% quantile and retain outlier markers.

[0086] Parameter initialization: Initialize to 0.7 / 0.3; Initialize based on historical playback fitting results; if there are insufficient samples, use default weights.

[0087] Objective function definition: Construct a joint loss using "price deviation loss + default penalty + risk penalty":

[0088] in It can also be configured by corporate strategy.

[0089] Offline calibration: Perform grid search or Bayesian optimization on the validation set, joint search. (Threshold levels), and when necessary , choose to The minimum set of parameters that meets business constraints (such as the lower limit of the timeliness target achievement rate).

[0090] Access control system launch: New parameters will only be released in a phased manner after meeting the thresholds of "price matching accuracy does not decrease, supply and demand imbalance rate does not increase, and cost indicators improve"; during the phased release, the system will be grouped by line or customer, and the old version rollback switch will be retained.

[0091] Online update cycle: executed according to "weekly minor updates, monthly major updates"; weekly updates only update thresholds and weight fine-tuning, monthly updates allow for retraining factor regression curves (e.g., (The piecewise fitting parameters); each update records the version number, sample window, and index changes, and these can be traced.

[0092] S3. Perform intelligent batch order splitting on orders.

[0093] This step targets batch orders with different specifications, timeliness, and destinations. It combines the transportation capacity, freight rate gradient, timeliness level, and node capacity constraints of the price matching candidate set to design an intelligent splitting algorithm under multiple constraints. Under the premise of ensuring service constraints, it achieves the optimal overall cost after order splitting.

[0094] This step includes the following sub-steps: S3.1 Set splitting constraints to further refine the constraint set based on traditional constraints.

[0095] The specific constraints are as follows: Goods dimension constraints: Sub-order goods categories cannot be mixed (except for special categories), and can be further subdivided into combinable sets under the same category based on hazard level, temperature control requirements, whether reinforcement is required, etc.

[0096] Capacity constraints: The weight / volume of sub-orders shall not exceed the maximum carrying capacity of the corresponding intermodal transport capacity. At the same time, the minimum economic loading rate of vehicles / carriages / cabins shall be considered, and penalties shall be imposed on splitting schemes with excessively low loading rates of sub-orders. Timeliness constraints: The delivery time of sub-orders must be the same as that of the original order or within an acceptable time window (e.g., ±0.5 days). For splitting and combining orders with a high risk of exceeding the time limit, a prohibition constraint shall be set. Cost constraints: The total transportation cost of the split sub-orders shall be less than or equal to the upper limit of the estimated transportation cost of the original bulk order, and under the same cost conditions, the combination with fewer transit times and higher reliability shall be given priority. Node capacity constraints: The loading and unloading capacity, storage capacity, and parking / berth capacity of each transfer node within a given time window must not be exceeded by the superposition of split schemes. For nodes that are close to saturation, further diversion is automatically suppressed through a penalty coefficient.

[0097] S3.2 Order splitting based on clustering and heuristic correction techniques. This includes the following steps: S3.2.1. Using K-means or density clustering algorithms, with cargo destination, transportation timeliness, freight rate gradient, and capacity adaptability as clustering features, batch orders are pre-divided into several initial sub-order clusters.

[0098] S3.2.2 For each sub-order cluster, based on the integer programming model and metaheuristic algorithm, calculate the optimal loading combination under the current capacity resource conditions to ensure that the loading rate and capacity constraints are not exceeded.

[0099] To improve the efficiency of the solution, this step can introduce metaheuristic algorithms such as simulated annealing or tabu search to perform global search and local fine-tuning on the initial clustering results, thus avoiding getting trapped in local optima.

[0100] At the level of integer programming modeling, this invention formalizes the order splitting problem as follows:

[0101]

[0102]

[0103] in, This refers to the total quantity of the original batch order. Indicates allocation to the first The quantity of goods in a single order / capacity unit; This represents the corresponding cost function; For load-bearing capacity; Load factor indicators (such as average load factor or minimum load factor). The minimum economic load factor threshold; For risk or node load indicators; This serves as a timeliness indicator. By using the clustering results as the initial solution to constrain the feasible region, and combining this with heuristic fine-tuning, a splitting process of coarse division followed by fine-tuning is achieved.

[0104] S3.2.3. Perform cost and timeliness calculations on the split sub-orders. If the total transportation cost exceeds the constraint threshold or the node capacity is limited, iterative corrections are made by adjusting the cluster center, merging / splitting local clusters, etc., to form a convergent split solution.

[0105] S3.3 Sub-order merging and stability optimization.

[0106] After obtaining feasible splitting solutions, a second round of merging optimization is performed on sub-orders with similar destinations, consistent transit times, and matching freight rates. This specifically includes the following: Prioritize merging sub-orders with the same destination, the same timeliness level, and low load factor into combinations with higher load factors to reduce capacity input and transshipment links.

[0107] For scattered orders along multiple similar routes, a "merging-diversion" design is adopted, which means merging them into a unified transportation capacity at intermediate nodes and then diverting them to the final destination at terminal regional nodes.

[0108] By comparing the evaluation indicators of cost, timeliness, and node load before and after the merger, only the merger results where Pareto is superior to or no worse than the original solution are retained, ensuring that the overall solution is improved in both stability and economy.

[0109] S4, integrated scheduling of multimodal transport resources.

[0110] This step, based on the optimal price matching candidate set and the intelligently split sub-orders, constructs an integrated scheduling platform for multimodal transport resources, realizing global scheduling and node-based collaborative connection of road, rail, waterway, and air transport capacity.

[0111] This step includes the following sub-steps: S4.1 Abstract the transportation capacity resources of each mode of transportation into a three-layer structure of capacity node, transportation link, and spatiotemporal state, and construct a spatiotemporal extended graph model.

[0112] The three-layer structure of capacity node—transportation link—spatiotemporal state includes: Transportation capacity nodes include freight stations, ports, airports, logistics parks, distribution centers, etc. Node attributes include loading and unloading capacity, warehousing capacity, working hours, historical congestion index, etc.

[0113] A transport link represents a feasible transport route between nodes. Link attributes include transport mode, mileage, standardized freight rate, average transit time, transit time fluctuation range, carbon emission level, etc.

[0114] Based on this, a spatiotemporal extended graph model is constructed, which discretizes the transportation link into multiple selectable dispatch time slots on the time axis, providing fine spatiotemporal constraints for subsequent scheduling algorithms.

[0115] Based on the above model, a visualized multimodal transport capacity resource map is formed, and dynamic information such as real-time freight rates, remaining capacity, and node queuing time are continuously superimposed to achieve online characterization of the capacity resource status.

[0116] To enhance forecasting capabilities, a simple and interpretable time-series forecasting module is introduced on top of the multimodal transport capacity resource map. The corresponding forecasting methods include: performing sliding window modeling on the time series of freight rates, remaining capacity, and queuing times of key trunk lines and hub nodes to obtain short-term forecast values ​​and confidence intervals; and using the forecast results and current observations as inputs for scheduling optimization, so that scheduling is not limited to the "current perspective" but has a certain degree of foresight, and can perform route detours or advance dispatches in advance during periods when congestion or capacity shortages are expected.

[0117] S4.2 Optimization of node connection time.

[0118] This step addresses the pain points of long waiting times and poor connectivity in multimodal transport by designing a node connectivity time optimization algorithm. The node connectivity time optimization algorithm includes the following steps: S4.2.1. With the goals of minimizing the total transportation time of sub-orders, minimizing transit waiting time, and balancing node load, establish a time window constraint model based on a spatiotemporal extension graph.

[0119] S4.2.2. Based on the transportation route of the sub-order, accurately match the departure time of short-distance road transport, railway freight train, waterway shipping schedule, and air cargo space. Control the deviation between the arrival time of short-distance road transport trucks at the railway freight station and the departure time of railway freight trains within a set time window (e.g., 1 hour), and control the deviation between the arrival time of waterway cargo at the port and the arrival time of road transshipment capacity within another set time window (e.g., 2 hours).

[0120] S4.2.3 For time periods when nodes are about to become congested, the algorithm automatically balances the load through time shifting, path detours, and early / late shipments to avoid peak accumulation at transit nodes.

[0121] S4.3, Batch capacity allocation.

[0122] This step, based on the optimized results of the capacity resource map and node connectivity, employs a combined strategy of greedy algorithms, integer programming, and rolling correction to allocate the split sub-orders to the optimal intermodal capacity resources. Specifically, it includes the following steps: S4.3.1 First, perform heuristic sorting based on indicators such as unit cost, timeliness, and node load level, and then use a greedy algorithm to generate an initial allocation scheme.

[0123] S4.3.2. Then, construct an integer programming model constrained by cost, timeliness, and resource utilization rate, and perform global optimization and feasibility verification on the initial scheme.

[0124] S4.3.3 Finally, in the actual execution process, a rolling optimization mechanism is adopted in combination with the real-time capacity status to make local adjustments to the capacity allocation of some sub-orders, so as to ensure that the plan maintains a comprehensive optimization status during the execution period.

[0125] Through the above methods, the integrated and optimized allocation of multimodal transport resources, from static planning to dynamic execution, can be achieved.

[0126] S5. Generation of shipping schemes based on global cost accounting optimization and multi-objective optimization.

[0127] This step includes the following sub-steps: S5.1 Perform global cost accounting optimization.

[0128] Establish a global cost accounting system for multimodal transport bulk orders, comprehensively considering cost factors across all stages, including transportation costs, transshipment service fees, loading and unloading fees, insurance premiums, idle capacity costs, timeliness breach of contract costs, carbon emission costs, and management and coordination costs. A multi-dimensional cost vector is introduced, expressed as follows:

[0129] in, Represents a multidimensional cost vector; : No. The transportation cost of a shipment can be further broken down into a basic freight rate and additional costs for oversized / special services. : No. The transit service fee for each transit node can be set with differentiated rates based on the node level and the operating period; The total cost of loading and unloading cargo is related to the number of loading and unloading operations, the intensity of loading and unloading, and the duration of the operation. Cargo transportation insurance premiums can be calculated in segments based on cargo value, route risk level, and historical accident rate; Costs incurred due to idle capacity, including empty vehicle / cargo runs and empty ship cabins, can be calculated based on empty mileage and empty cabin duration. The time-lapse cost of failing to complete transportation on time can be converted into a default price through the Service Level Agreement (SLA) agreed with the customer. The environmental cost calculated based on carbon emissions can be determined by the unit carbon emission coefficient and carbon price for different modes of transportation. The management and tool usage costs associated with multi-carrier and multi-node collaboration can be estimated based on the number of collaborative participants, the number of interfaces, and the frequency of collaboration.

[0130] Based on this, a scalarized global cost is defined for cost optimization, and the calculation formula is as follows:

[0131] in, Total global shipping cost for bulk orders.

[0132] To improve the accuracy and adaptability of the accounting, a cost estimation model driven by historical execution data is introduced. The construction method of the cost estimation model is as follows: 1) Construct a regression or segmented statistical model for each cost item, and use "transportation mode, node level, time period type, service level" as independent variables to fit the corresponding cost level.

[0133] 2) For cost items with large fluctuations (such as carbon emission costs and management coordination costs), additional confidence intervals or risk premium coefficients are provided to provide input for risk constraints in subsequent multi-objective optimization.

[0134] 3) Set up a cost weight vector based on the company's business strategy. After normalizing each cost item, a weighted summation is performed to obtain a "comprehensive cost score" that facilitates comparison of cost items with different dimensions:

[0135] in, For cost items The normalized value, and When enterprises need to consider both cost structure and risk (e.g., prioritize default risk or carbon costs), they should prioritize adopting [the appropriate approach]. As an optimization objective or constraint input, it makes the expression clearer and the units more consistent.

[0136] S5.2 Multi-objective optimization of shipping plan.

[0137] This step uses multi-objective optimization functions, namely optimizing total global transportation cost, improving transportation timeliness, increasing capacity utilization, and controlling carbon emission levels, and employs a non-dominated sorting genetic algorithm (NSGA-Ⅲ) to optimize candidate capacity combinations and transportation routes. The optimization solution method includes the following steps: 1) Encode each feasible "sub-order-route-carrier-shipment time" combination as a chromosome.

[0138] 2) Using indicators such as cost, timeliness, capacity utilization, and carbon emissions as multi-objective fitness functions, the population is continuously evolved through genetic operations such as crossover and mutation.

[0139] 3) By using non-dominated sorting and crowding distance to maintain population diversity, and eliminating suboptimal solutions that are significantly inferior to other schemes, we finally obtain a set of Pareto optimal shipping schemes with different trade-offs between different objectives.

[0140] S5.3, Shipping plan decision generation.

[0141] After obtaining the Pareto optimal solution set, this step assigns weights to each objective based on the user's core needs (such as cost priority, timeliness priority, comprehensive benefit priority, and low carbon priority), uses a weighted scoring method or analytic hierarchy process (AHP) to comprehensively score the shipping plan, and introduces a scenario-adaptive weight adjustment mechanism.

[0142] When a user selects "cost priority", the initial weight of the cost target is increased, and reasonable lower limits are set for timeliness and carbon emissions. In scenarios with drastic fluctuations in freight rates or tight cash flow, the cost weight can be further dynamically increased.

[0143] When a user selects "Timeliness First", secondary optimizations are performed on cost and resource utilization while meeting timeliness constraints. In emergency scenarios such as pre-holiday promotions and disaster relief supplies, a strategy of "increasing timeliness weight and decreasing cost weight" can be automatically triggered.

[0144] When users select "overall benefits first", a multi-objective balanced weighting is adopted to select the scheme with the highest overall score; in green and low-carbon demonstration projects or policy-oriented scenarios, the carbon emission target weight can be increased to a level comparable to cost and timeliness.

[0145] Finally, the multimodal transport bulk order transportation and dispatch scheme that meets the constraints and has the best comprehensive score is selected, and a structured dispatch instruction set is generated, which clarifies the key information of each sub-order, such as the mode of transport, carrier, transport route, dispatch time, transit node, and freight rate standard, so as to achieve seamless integration with the execution platform.

[0146] S6. The shipping plan is dynamically and adaptively adjusted.

[0147] Build a dynamic adaptive adjustment module for shipping plans to monitor in real time market freight rate fluctuations, changes in transportation capacity resources, adjustments in order demand, and abnormal transportation nodes (such as congestion, shutdowns, and severe weather), enabling rapid adaptive adjustments to shipping plans.

[0148] This step includes the following sub-steps: S6.1 Dynamic Information Perception.

[0149] This step collects real-time dynamic information on freight rates, capacity, orders, and nodes through multiple channels, including IoT devices, freight platform interfaces, video and sensor data collection, and manual feedback. An anomaly event knowledge base is then built to categorize and label the information. When the magnitude of information changes exceeds preset thresholds (e.g., freight rate fluctuations ≥10%, temporary capacity reductions ≥20%, node congestion index exceeding a set level) or matches a major anomaly event rule, an automatic adjustment mechanism is triggered.

[0150] S6.2, Fast Rematching and Rescheduling.

[0151] Once the adjustment mechanism is triggered, this step performs a fast recalculation in a tiered manner, as follows: Price layer: Invoke the optimal price matching model to recalculate the comprehensive optimal matching freight rate for the affected routes or carriers, and automatically remove or downweight carriers that no longer meet the price threshold.

[0152] Capacity layer: Update the remaining capacity and available time slots of affected capacity nodes and transportation links, and prioritize the reallocation of sub-orders that have been locked but not yet actually shipped.

[0153] Path layer: Search for feasible alternative paths for the affected sub-orders and evaluate the changes in cost, timeliness, and risk of the alternative paths.

[0154] Dispatch layer: For sub-orders already in transit, try to keep the front-end transportation unchanged, and only optimize and adjust the subsequent transit and last-mile delivery links to reduce the disturbance to the execution site.

[0155] Through the aforementioned hierarchical rapid recalculation mechanism, a new executable adjustment plan can be generated within minutes, and the changes in key indicators before and after the adjustment can be compared to ensure that the adjustment has positive benefits or is within an acceptable range.

[0156] S6.3, Scheme Updates and Synchronization The adjusted shipping plan is synchronized in real time to carriers, logistics parks, cargo owners and other relevant parties in the form of structured messages. Notifications are sent through multiple channels, including operating terminals, API interfaces and SMS / email. The confirmation status and feedback information of each party are recorded, and adjustment logs and tracking numbers are automatically generated to support subsequent auditing and effect evaluation. This ensures the coordinated execution of each link and guarantees the stability and traceability of bulk order transportation and shipping.

[0157] Based on the above steps, the present invention provides the following specific embodiments: A logistics company has a bulk order for 500 tons of industrial raw materials destined for Guangzhou, with a transit time of 7 days. The planned transportation method is a multimodal transport system combining road, rail, and road, originating in Jinan. Based on the optimal price matching model of this invention, the following steps are performed for this order: S1. Multimodal transport freight rate data collection and standardization processing: By using a multi-source freight rate collection terminal, real-time freight rates for highways and railways between Jinan and Guangzhou are obtained. The collected raw data includes various formats such as "charge per vehicle", "charge per ton", and "charge per container".

[0158] The system automatically invokes the freight standardization module to convert different billing methods into a standardized freight rate in yuan / ton-kilometer, resulting in: Basic Standardized Freight Rate for Highways. Yuan / ton·km, basic standardized railway freight rate Yuan / ton·km.

[0159] Simultaneously, discount curves for 500-ton batch orders are extracted from the historical order database and fitted with current market discount strategies to obtain the batch size factor. Based on the established plan of using two transit points for the current order with a standard delivery time, the delivery time factor is obtained. transit frequency factor The transportation mode factor is given by a preset parameter table: Highway ,railway .

[0160] S2. Perform comprehensive optimal price matching and candidate carrier screening: Historical on-time performance and damage rates of candidate road and rail carriers were extracted from the carrier performance database to calculate the performance factor: This study selected a group of road carriers... Railway carrier group .

[0161] Based on the company's "balancing cost and low carbon" strategy, the carbon emission factor calculation module was invoked: carbon emission factor for road transportation. Carbon emission factors of railway transportation .

[0162] Substituting into the extended price formula, we obtain the example calculation result: Optimal matching extended freight rates for highways: Yuan / ton·km.

[0163] Optimal matching extended freight rates for railways: Yuan / ton·km.

[0164] Effective matching filtering under adaptive thresholds: Based on the historical freight rate volatility (MAV=1.8%) of the current route, Jinan-Guangzhou is judged to be a relatively stable trunk line in terms of freight rates. Therefore, the price matching threshold is set to... .

[0165] by Centered on yuan / ton·km, The railway carriers' quotations were screened within the range, resulting in 3 railway carriers being included in the candidate set.

[0166] by Centered on yuan / ton·km, Within the specified range, quotes from highway shuttle carriers were screened, resulting in two highway carriers being included in the candidate set.

[0167] A comprehensive score was calculated for the above candidate carriers: Sort by rating from lowest to highest.

[0168] Technical Results: Compared to the control scheme (which uses a fixed freight rate standard for the same route and does not introduce a rule engine with tension index and adaptive threshold), this model improves price matching accuracy by 32%, reduces average transportation costs by 18%, and reduces the supply-demand mismatch rate from 35% to 8%.

[0169] S3. Perform intelligent batch order splitting: Based on the price matching candidate set obtained in step S1, the specific steps for intelligent splitting of the aforementioned 500-ton batch order are as follows: Constraints and parameter settings: Capacity constraints: Railway freight trains have a single carriage load capacity of 60 tons, and highway connecting freight trains have a load capacity of 30 tons; at the same time, a minimum economic loading rate threshold of 80% is set to avoid a large number of sub-orders with low loading rates.

[0170] Cost constraint: The total transportation cost after splitting the sub-orders shall not exceed the original estimated cost of 800,000 yuan.

[0171] Time constraints: The overall delivery time shall not exceed 7 days, and the time deviation of a single sub-order shall not exceed ±0.5 days.

[0172] Node capacity constraints: The loading and unloading capacity and warehousing capacity of Jinan Railway Freight Station and Guangzhou Railway Freight Station within the planned shipping window shall not be exceeded by the superposition of split plans.

[0173] Splitting based on clustering pregrouping: Treating 500-ton orders as batch orders of the same type of goods, destination, and timeliness, and using "capacity adaptability, loading rate target, and node load level" as feature inputs, K-means is used to pre-group "loading combinations of single-carriage / single-car truck".

[0174] The initial clustering results show 8 cluster centers on the railway side, corresponding to the load combination of "60 tons × 7 + 40 tons × 1"; and 14 cluster centers on the highway side, corresponding to the combination of "30 tons × 13 + 10 tons × 1".

[0175] Sub-orders with low loading rates of 40 tons and 10 tons are automatically marked as "pending merging" for optimization in the subsequent heuristic correction phase.

[0176] Heuristic correction and merging optimization: An integer programming model is constructed with the goal of "maximizing loading rate, avoiding node overload, and meeting timeliness and cost constraints", and the K-means result is introduced as the initial solution.

[0177] Through heuristic search and local adjustments, the remaining capacity of the 40-ton sub-order on the railway side was recombined with that of other sub-orders, and the 10-ton sub-order on the highway side was merged with other sub-orders. Finally, a splitting scheme of 8 railway sub-orders (all with a loading rate of no less than 90%) and 12 highway sub-orders (all with a loading rate of no less than 85%) was obtained.

[0178] Technical Results: Compared to the control scheme (which only splits by category / weight, without combined freight rate tiers and node capacity constraints, with a historical load factor of approximately 65%), this scheme increases the load factor to 88%, reduces unit transportation costs by 12%, and lowers the congestion rate at transit nodes from 20% to 5%.

[0179] S4. Integrated scheduling of multimodal transport resources: Next, for the 500-ton batch orders under the above split plan, multimodal transport resources will be integrated and scheduled. The specific steps are as follows: Integrated capacity scheduling and node connection optimization: On the multimodal transport capacity resource map, specific carriers and departure schedules are matched for 8 railway sub-orders and 12 road sub-orders: Upstream highway connection: Match local 30-ton heavy-duty trucks in Jinan and allocate each highway sub-order to different truck trips according to time windows.

[0180] Main railway line: Match 7 60-ton and 1 40-ton railway freight train carriages (Jinan-Guangzhou) for 8 railway sub-orders.

[0181] Downstream road connection: Matching local Guangzhou road freight trucks to transport goods from Guangzhou Railway Freight Station to different factory addresses.

[0182] The node connection algorithm automatically calculates the latest departure time of the upstream highway connection based on the fixed departure timetable of the railway freight train, so that the deviation between the arrival time of the highway freight train at Jinan Railway Freight Station and the departure time of the railway freight train is controlled within 40 minutes.

[0183] For the Guangzhou transit node, by locking in the warehouse space and loading and unloading time in advance, the road connecting transport capacity is in place 30 minutes before the train's expected arrival time, achieving a seamless connection of "loading and unloading upon arrival and dispatching immediately after loading and unloading".

[0184] S5. Generation of shipping plans based on global cost accounting optimization and multi-objective optimization: After completing order splitting and initial capacity allocation, the above solution undergoes global cost accounting and multi-objective optimization, as detailed below: Cost accounting for all stages: The costs of rail and road transportation are calculated separately: rail costs are approximately 420,000 yuan, and road costs are approximately 180,000 yuan.

[0185] The transit service fee is calculated at 15,000 yuan per transit point, for each transit point in Jinan and Guangzhou, totaling 30,000 yuan.

[0186] The loading and unloading fee is calculated at 40,000 yuan based on the frequency and tonnage of loading and unloading, and the insurance premium is calculated at 20,000 yuan based on the value of the goods and the risk level of the route.

[0187] The cost of idle transport capacity is estimated at 0.5 million yuan based on the empty running and empty load time of railway carriages and highway freight cars under the current scheme.

[0188] The cost of breaching the time limit is 0 under the current plan (all sub-orders meet the time limit requirements).

[0189] Without explicitly factoring in carbon emission costs and collaboration costs, the initial total global transportation cost is: Ten thousand yuan.

[0190] NSGA-Ⅲ Multi-Objective Optimization Solution: The decision variables such as "which carrier to choose, which train to select, whether to allow the merging or splitting of some sub-orders, and whether to adjust some shipping time windows" are encoded as chromosomes to construct a population of candidate shipping schemes.

[0191] Using "minimum total cost, shortest total transportation time, highest capacity utilization, and controllable carbon emissions" as the multi-objective fitness, the corresponding indicators are calculated for each candidate scheme.

[0192] After several generations of iterative evolution, the NSGA-Ⅲ algorithm eliminates suboptimal solutions that are significantly inferior to other solutions: for example, it eliminates a railway carrier with significantly high freight rates and mediocre performance, and replaces it with a railway carrier that offers better prices and comparable delivery times; at the same time, it reduces the empty running of one idle freight car and improves the overall loading rate by recombining the components on the highway connection side.

[0193] Solution decision and result output: For multiple options at the Pareto frontier, based on the company's preference for "cost priority with consideration of timeliness", the cost target is set as the highest weight, while timeliness and utilization rate are secondary targets.

[0194] After comprehensive evaluation, one of the proposed solutions was selected as the final implementation plan. Under this plan, the total transportation cost was reduced to approximately RMB 658,000, the overall transportation time was shortened to 6.5 days (meeting the 7-day requirement), and the utilization rate of intermodal resources was increased to approximately 96%.

[0195] Technical results: Compared with the control scheme (which only calculates trunk line freight and does not include all aspects such as transshipment / loading / unloading / idleness, with a historical calculation deviation rate of about 15% to 20%), the cost accounting accuracy of this scheme (the consistency between the estimated value and the actual settlement value of all aspects) is improved to about 95%, the overall total transportation cost is reduced by 5.3%, and the utilization rate of intermodal resources is improved to about 96% (about 21% higher than the control group, the specific figure is based on the enterprise's baseline).

[0196] The solution is broken down into a set of executable shipping instructions, which automatically generates information such as carrier selection, train / carriage number, shipping time, arrival time, and transit node operation arrangements for each sub-order, and pushes it to the relevant execution device or platform.

[0197] S6. Shipping plan dynamically and adaptively adjusted: During the execution of the above-mentioned shipping plan, the railway freight rate for the Jinan-Guangzhou section increased by 12% due to market fluctuations, triggering the dynamic plan adjustment mechanism. The specific process is as follows: Dynamic information perception and trigger determination: The freight rate monitoring module continuously obtains trunk railway freight rates from the railway freight rate interface and third-party index platforms. When it detects that the freight rate of the line has increased by 12% cumulatively within 1 hour, it matches the event with the preset rule of "freight rate fluctuation ≥ 10%" and judges it as a major cost risk event.

[0198] At the same time, the list of affected orders was checked and it was found that some sub-orders of the current 500-ton batch order had not yet been actually shipped (they were in the state of waiting to be loaded), which met the "adjustable" condition and automatically triggered the recalculation process for this batch of orders.

[0199] Fast rematching and rescheduling computation: Price layer: Re-invoke the extended price matching model to adjust the base railway freight rate after the price increase. Standardize the data and calculate the new optimal matching extended freight rate for railways. Two railway carriers with relatively small freight rate fluctuations and good performance were selected to enter the new candidate set under the new price threshold.

[0200] Batch size factor adjustment: Considering that enterprises hope to hedge against the cost increase, based on the negotiation strategy with carriers, the single-vehicle loading rate is appropriately increased, and the batch discount curve is refitted to obtain a new batch size factor. .

[0201] Capacity and Route Layer: For sub-orders that have not yet been shipped, perform a local NSGA-Ⅲ optimization again on the new candidate railway carriers and road connection resources. Without changing the overall time constraints, adjust the shipping time windows and carrier combinations of some sub-orders to find more cost-effective alternative routes or schedules.

[0202] For orders already en route, the existing train schedules will remain unchanged. Only at the terminal highway connection point, advance reservations and vehicle reallocation will be used to reduce the risk of potential congestion.

[0203] Plan adjustments, implementation, and coordination are carried out simultaneously: A comparative analysis of the revised shipping plan and the original plan revealed that the total transportation cost under the new plan was controlled at approximately 672,000 yuan, which is about 28,000 yuan lower than the expected cost of directly accepting the price increase under the original plan. The overall delivery time still meets the 7-day requirement.

[0204] Technical benefits: Compared to the control process (which relies on manual modification of the plan after freight rate anomalies, with a historical response time of about 4 to 8 hours), this plan automatically recalculates and reduces the response time to less than 5 minutes, reduces cost losses (the incremental cost relative to "not adjusting but executing at the original price after the price increase") by about 65%, and achieves a 98% success rate in plan adjustment.

[0205] Subsequently, adjustment instructions are automatically generated, and a railway carrier change notice and a new loading plan are issued through the carrier interface. At the same time, a road connection time adjustment plan is issued to the relevant fleets and logistics parks.

[0206] The adjustment results are synchronized to the cargo owner via terminal and SMS / email, and the complete adjustment process and effect data are recorded for subsequent dynamic strategy optimization and performance evaluation, ensuring the smooth implementation of the plan and its traceability.

[0207] The following are the definitions of technical effect indicators in the embodiments and their correspondence with the specific steps of the invention.

[0208] To meet the requirement of verifiable technical effects in the inventive step demonstration, this section provides a unified statistical definition for the quantitative expressions appearing in the embodiments and their correspondence with the specific features of the invention (as shown in Table 1). During implementation, historical data can be replayed or online A / B verification can be performed under the same definition. 1. Explanation of standardized experimental procedures and control baselines To avoid disputes caused by "incomparable effect data" or "inconsistent standards," the percentages in the examples are calculated using the following unified standards: 1) Sample scope: Select batch order samples within the same business domain, the same corridor type (trunk line / branch line), and the same type of goods; it is recommended that the sample size be no less than 300 orders and the statistical period be no less than 3 consecutive months.

[0209] 2) Grouping method: Control group: Using the company's original rules (fixed freight rate or simple minimum price, with manual adjustment of the process); Experimental group: The scheme of this invention (including 4.2 price matching model and its threshold, scoring, and tension index mechanism) is adopted.

[0210] 3) Consistency constraints: The two groups are comparable in terms of order size range, timeliness level, and distribution of major routes; orders under extreme force majeure (such as temporary policy-related lockdowns) are excluded according to unified rules.

[0211] 4) Statistical caliber: All indicators are calculated according to the definition in Table 5.5, and the "group mean / group percentage" are compared within the same time window; when "increase / decrease" is involved, the calculation is uniformly based on the relative control group.

[0212] 5) Significance Recommendation: When conditions permit, conduct significance tests (such as t-tests or non-parametric tests) on key indicators (such as price matching accuracy and average transportation cost), and retain the test results for future reference.

[0213] 6) Traceability requirements: Retain the sample list (de-identified), parameter version number, operation log, and indicator calculation script version to ensure that the technical effect can be verified.

[0214] Table 1. Correspondence between statistical scope and specific features of the invention.

[0215] To strengthen the causal relationship between innovation points and technical effects, a one-to-one correspondence is established between the specific features of the invention and the quantitative effects shown in the implementation results. The correspondence between the specific features of the invention and the quantitative effects is shown in Table 2.

[0216] Table 2. Correspondence between specific features of the invention and its quantifiable effects

[0217] The present invention has been described above by way of example with reference to the accompanying drawings. Obviously, the specific implementation of the present invention is not limited to the above-described manner. Any non-substantial improvements made using the inventive concept and technical solution of the present invention, or the direct application of the inventive concept and technical solution of the present invention to other occasions without modification, are all within the protection scope of the present invention.

Claims

1. A multimodal transport bulk order shipping method with optimal price matching, characterized in that, Includes the following steps: S1. Multimodal transport freight rate data collection and standardization processing; S2. Perform comprehensive optimal price matching and candidate carrier screening; This step includes the following sub-steps: S2.1 Establish a formula for the comprehensive optimal matching freight rate, which should at least incorporate factors such as batch size, mode of transport, time requirement, and number of transfers. S2.2 Establish an adaptive price matching threshold system to initially screen carrier quotations based on allowable deviation thresholds; S2.3 Calculate a comprehensive score for each candidate quote based on the optimal matching freight rate combined with the multi-factor intermodal transport supply and demand tension index; S2.4 Sort the preliminary candidate set from low to high according to comprehensive score, and select the top N candidates as the final candidate set; S3. Perform intelligent batch order splitting on orders; S4, integrated scheduling of multimodal transport resources; S5. Generation of shipping schemes based on global cost accounting optimization and multi-objective optimization; S6. The shipping plan is dynamically and adaptively adjusted.

2. The multimodal transport bulk order shipping method with optimal price matching according to claim 1, characterized in that, In step S2.1, the basic comprehensive optimal matching freight rate formula is used, as follows: in, This represents the overall optimal matching freight rate for the basic type. This represents the standardized base freight rate. Indicates the batch size factor. Indicates the mode of transport factor. Indicates the timeliness requirement factor. Indicates the number of transfers factor; Batch size factor, range of values Based on historical bulk freight rate data, a combination of piecewise functions and nonlinear regression is used to fit the bulk size curve, thereby determining the bulk size factor. The mode of transport factor characterizes the differences between different modes of transport. Timeliness requirement factor, range of values By analyzing the relationship between the timeliness level and premium level of historical orders, a nonlinear function that monotonically increases with the timeliness level is constructed for calculation to obtain the timeliness requirement factor; Transfer frequency factor, range of values It also characterizes transit service fees and transit risk costs, and applies a penalty coefficient to schemes that exceed the experience threshold for the number of transits.

3. The multimodal transport bulk order shipping method with optimal price matching according to claim 2, characterized in that, In step S2.1, a performance factor and a carbon emission factor are introduced into the basic comprehensive optimal matching freight rate formula to form an extended comprehensive optimal matching freight rate formula, as follows: in, This represents the extended, comprehensive, optimal matching freight rate. Indicates the performance factor. Indicates carbon emission factor; Performance factor, range of values Scoring is based on the carrier's historical on-time performance, damage rate, and claims history; Carbon emission factor, range of values The carbon emissions per unit of transport capacity for each mode of transportation are converted into carbon costs.

4. The multimodal transport bulk order shipping method with optimal price matching according to claim 3, characterized in that, In step S2.2, the allowable deviation threshold between the candidate freight rate and the optimal matching freight rate is automatically adjusted based on market volatility, carrier concentration, and order type. The permissible deviation range of market volatility intensity is calculated based on the volatility estimation model, and the daily volatility of freight rates is calculated using the moving average volatility algorithm. The formula is as follows: in, P represents volatility. i and P i-1 These represent the historical freight rates for day i and day i-1, respectively. Based on the adaptively adjusted allowable deviation threshold, all carrier quotes with deviations from the optimal matching freight rate within the threshold range are selected to form a preliminary candidate set.

5. The multimodal transport bulk order shipping method with optimal price matching according to claim 4, characterized in that, In step S2.3, the comprehensive scoring function for the carrier's quotation is defined as follows: Here, "Score" represents the overall score. Provide a quote for the carrier. This represents the extended optimal matching rate; the basic optimal matching rate can also be used. , The Multi-Factor Intermodal Transport Supply and Demand Tension Index reflects the supply and demand status of transportation lines and the carrying capacity of key nodes. , Indicates corresponding to respectively The weighting coefficients of the multi-factor intermodal transport supply and demand tension index, This indicates the degree of difference between the quoted price and the best matching freight rate; The formula for calculating the multi-factor intermodal transport supply and demand tension index is: Among them, freight rate deviation characteristics , The current freight rate, and These are the minimum and maximum freight rates for this route over the past 30 days. Characteristics of supply sufficiency , Indicates remaining transport capacity. Indicates total transport capacity; Node load characteristics , For node utilization, To adjust the parameters; Congestion risk characteristics , ; The corresponding freight rate deviation characteristics are as follows: Characteristics of supply sufficiency Node load characteristics Congestion risk characteristics The weighting coefficients.

6. The multimodal transport bulk order shipping method with optimal price matching according to claim 1, characterized in that, Step S3 includes: S3.1 Set splitting constraints, including constraints on cargo dimension, transportation capacity dimension, timeliness dimension, cost dimension, and node capacity. The node capacity constraint is: the loading and unloading capacity, storage capacity, and parking space / berth capacity of each transfer node within a given time window shall not be exceeded by the superposition of splitting schemes. For nodes that are close to saturation, further diversion shall be automatically suppressed by the penalty coefficient. S3.2 Order splitting based on clustering and heuristic correction techniques; S3.3 Sub-order merging and stability optimization: After obtaining a feasible splitting scheme, a second merging optimization is performed on sub-orders with similar destinations, consistent transportation time, and matching freight rates after splitting. Step S3.2 includes the following steps: S3.2.

1. Using K-means or density clustering algorithms, with cargo destination, transportation timeliness, freight rate gradient, and capacity adaptability as clustering features, batch orders are pre-divided into several initial sub-order clusters; S3.2.2 For each sub-order cluster, based on the integer programming model and metaheuristic algorithm, calculate the optimal loading combination under the current transportation capacity resource conditions to ensure that the loading rate and transportation capacity constraints are not exceeded. S3.2.

3. Perform cost and timeliness calculations on the split sub-orders. If the total transportation cost exceeds the constraint threshold or the node capacity is limited, iterative corrections are made by adjusting the cluster center and merging / splitting local clusters to form a convergent split solution.

7. The multimodal transport bulk order shipping method with optimal price matching according to claim 1, characterized in that, Step S4 includes: S4.1 Abstract the transportation capacity resources of each mode of transportation into a three-layer structure of capacity node - transportation link - spatiotemporal state, and construct a spatiotemporal extended graph model; S4.2 Optimization of node connection time; S4.3 Batch capacity allocation: Based on the optimization results of the capacity resource map and node connection, a combination strategy of greedy algorithm, integer programming and rolling correction is adopted to allocate the split sub-orders to the optimal intermodal capacity resources.

8. The multimodal transport bulk order shipping method with optimal price matching according to claim 7, characterized in that, In step S4.1, node attributes include loading and unloading capacity, storage capacity, working hours, and historical congestion index; the transportation link represents the feasible transportation path between nodes, and the link attributes include transportation mode, mileage, standardized freight rate, average transit time, transit time fluctuation range, and carbon emission level; a visualized multimodal transport capacity resource map is formed based on the spatiotemporal extended graph model, and dynamic information such as real-time freight rate, remaining capacity, and node queuing time is continuously superimposed; A time series forecasting module is introduced on the multimodal transport capacity resource map. The corresponding forecasting methods include: sliding window modeling of time series including freight rates, remaining capacity, and queuing time of key trunk lines and hub nodes to obtain short-term forecast values ​​and confidence intervals. The prediction results and the current observations are used together as inputs for scheduling optimization.

9. The multimodal transport bulk order shipping method with optimal price matching according to claim 7, characterized in that, Step S4.2 addresses the pain points of long waiting times and poor connections in multimodal transport cross-node connections by designing a node connection time optimization algorithm; this step specifically includes: S4.2.

1. With the goals of minimizing the total transportation time of sub-orders, minimizing transit waiting time, and balancing node load, establish a time window constraint model based on a spatiotemporal expansion graph. S4.2.

2. Based on the transportation route of the sub-order, accurately match the departure time of short-distance road transfer, railway freight train, waterway shipping schedule, and air cargo space, and control the deviation between the arrival time of short-distance road transfer trucks at railway freight stations and the departure time of railway freight trains within a set time window, and control the deviation between the arrival time of waterway shipping cargo and the arrival time of road transshipment capacity within another set time window. S4.2.3 For time periods when nodes are about to become congested, the algorithm automatically balances the load through time shifting, path detours, and early / late shipments to avoid peak accumulation at transit nodes.

10. The multimodal transport bulk order shipping method with optimal price matching according to claim 1, characterized in that, Step S5 includes: S5.1 Optimize global cost accounting, establish a global cost accounting system for multimodal transport batch orders, comprehensively consider cost factors in all stages, and introduce multidimensional cost vectors; S5.2 Multi-objective optimization of shipping schemes: The multi-objective optimization functions are global total transportation cost optimization, transportation time efficiency optimization, capacity resource utilization improvement and carbon emission control. The non-dominated sorting genetic algorithm is used to optimize and solve the candidate capacity combination and transportation route. S5.3 Shipping plan decision generation: Based on the user's core needs, weights are set for each objective. The shipping plan is comprehensively scored using a weighted scoring method or analytic hierarchy process. A scenario-adaptive weight adjustment mechanism is introduced to finally generate the shipping plan.

Images