A liquid fuel multi-modal intermodal transport solution method based on dynamic divide-and-conquer strategy

Abstract

The present application relates to a kind of liquid fuel multimode intermodal transport solving method based on dynamic divide-and-conquer strategy, comprising: according to network parameter in cross-regional liquid fuel transport network and initial logistics plan, set the initial value of dynamic set;The time period of initial logistics plan is evenly split into multiple sub-time windows;According to the multiple sub-time windows after splitting, the parent model of mixed integer linear programming model is constructed, and the dynamic set of the transport batch of each node corresponding to the current iteration round is solved;Each sub-time window of previous round is further evenly split into several sub-time windows, and the sub-model of mixed integer linear programming model is further constructed, and the dynamic set of the transport batch of each node under the iteration round is solved;It is updated as parent model, and the division of sub-time window and the construction and solving of sub-model continue, until the dynamic set of transport batch solved by final sub-model meets the preset time accuracy requirement.

Description

Technical Field

[0001] This invention relates to the field of liquid fuel transportation technology, and in particular to a solution method for multimodal transport of liquid fuels based on a dynamic divide-and-conquer strategy. Background Technology

[0002] Cross-regional liquid fuel logistics, as a core link in the energy supply chain, heavily relies on the coordinated optimization of multimodal transport, including pipelines, railways, highways, and waterways. Among these, pipeline transport has become the main mode of transport due to its low cost and high efficiency. However, its unique multi-batch sequential transport process (such as the sequential transport of liquid fuels like gasoline, diesel, and aviation kerosene) leads to a strong coupling in the allocation of transport time and space resources. Meanwhile, point-to-point transport modes such as railways, highways, and waterways need to be closely integrated with dynamic constraints such as batch switching of pipeline transport, tank storage capacity, and loading and unloading time windows.

[0003] The existing multimodal transport problem for cross-regional fuel logistics requires handling mixed continuous-discrete decision variables, modeling it as a large-scale integrated linear programming (MILP) model to obtain the final point-to-point logistics plan and the corresponding transport modes and routes. Because liquid fuel transport involves the special phenomenon of "fixed vehicles and moving product batches" in pipeline transport, it is necessary to separate the pipeline transport scheme from the logistics plan obtained from the MILP model and gradually develop a pipeline batch transport plan based on manual experience. This process may involve multiple adjustments, and the plan development results are affected by the experience level of the personnel, resulting in low efficiency.

[0004] In existing technologies, to avoid the inefficiency caused by human experience, the experience of manually compiling pipeline batch transportation plans is transformed into a mathematical optimization model. Commercial optimization solvers (such as CPLEX, GUROBI, etc.) are then used to solve the logistics plan that includes the pipeline batch transportation plan in one go. However, since the pipeline batch transportation plan compilation model involves a large number of binary variables, once the number of nodes in the solution object is large and the time span is long, the rapid expansion of the model size when using commercial optimization solvers (such as CPLEX, GUROBI, etc.) to solve the logistics plan will cause the "curse of dimensionality" phenomenon, resulting in difficulty in algorithm convergence and failure to obtain the optimal solution or even a feasible solution for a long time.

[0005] Therefore, it is necessary to provide new solution methods for multimodal transport of liquid fuels across regions to address the problem of excessive time consumption. Summary of the Invention

[0006] To address the aforementioned problems, the purpose of this invention is to provide a solution method for multimodal transport of liquid fuels based on a dynamic divide-and-conquer strategy, solving the complexity of large-scale, long-cycle, cross-regional, and multi-variety transport planning in current logistics planning. This method effectively combines divide-and-conquer algorithms and dynamic adjustment mechanisms, decomposing the global optimization problem into multiple "parent-child" problems, significantly reducing the computational complexity of the model while ensuring optimization accuracy.

[0007] To achieve the above objectives, the present invention adopts the following technical solution:

[0008] In a first aspect, this application provides a solution method for multi-mode intermodal transport of liquid fuels based on a dynamic divide-and-conquer strategy, the method comprising:

[0009] (1) Based on the network parameters and initial logistics plan in the cross-regional liquid fuel transportation network, set the initial value of the dynamic set of transportation batches of each node in the transportation network; wherein, the network parameters include the type of transportation mode, the type of liquid fuel, the transportation mode, and several nodes in the transportation route of each transportation mode;

[0010] (2) First iteration: Divide the time period of the initial logistics plan into multiple sub-time windows evenly; construct the parent model of the mixed integer linear programming model based on the multiple sub-time windows after the division, and solve the dynamic set of transportation batches of each node corresponding to the current iteration based on the parent model;

[0011] (3) Second iteration: Each sub-time window of the previous round is further divided into several sub-time windows, and the sub-model of the mixed integer linear programming model is further constructed. Based on the sub-model, the dynamic set of transportation batches of each node under this iteration round is obtained.

[0012] (4) More rounds of iteration: update the sub-model to the parent model, further divide the sub-time window of the previous round, dynamically set the transportation batches of each node based on the parent model, and continue to divide the sub-time window and construct and solve the sub-model until the dynamic set of transportation batches solved by the final sub-model meets the preset time accuracy requirements.

[0013] In one implementation, in (1), the transportation mode includes at least pipeline transportation, and also includes one or more combinations of railway transportation, road transportation and waterway transportation; each node is distributed on each pipeline of pipeline transportation, or on the line of railway transportation, road transportation and waterway transportation, and the node types include supply points, distribution points and demand points.

[0014] The network parameters also include the position of the start and end of the batch of oil at each time point in the pipeline.

[0015] In one implementation, the parent or child model uses discrete-time representation, where t∈T represents the set of time points, and the number of time windows is one less than the number of time nodes; n∈N represents the set of nodes, where N... S Let N represent the set of supply points. D Let N = N S ∪N D ; o∈O represents the type of liquid fuel, O n The type of liquid fuel supplied or demanded by node n; z∈Z={train,truck,ship} represents the set of non-pipeline transportation modes;

[0016] For pipeline transportation patterns, p∈P represents the set of pipelines, and N p Let p represent the set of nodes on the pipe. Let p represent the set of supply points on pipeline p. Let p represent the set of demand points on the pipeline. I represents the batch set; I p I represents the dynamic set of batches that may pass through pipeline p throughout the entire cycle. p,n I represents the dynamic batch set that may pass through node n on pipeline p throughout the entire cycle. t,p,n This represents the set of batches that n nodes on pipeline p may inject / distribute within the t-th time window; dynamic set I p,n and I t,p,n It needs to be updated based on the batch position information of the parent model; using parameters. and This represents the position of the start and end points of batch i at each time point in the logistics plan of the parent model. If, in the logistics plan obtained from the parent model, the end point of batch i does not exceed node n on pipeline p at the start of the cycle, but the start point exceeds node n on pipeline p at the end of the cycle, then the dynamic batch set I of batch i at node n on pipeline p is... p,n In the middle, see equation (1); for each time window t<|T| of the sub-model, if the tail of batch i is in the corresponding time window td of the parent model t If a batch i starts before exceeding node n but ends after exceeding node n, then that batch i is in the dynamic set I. t,p,n See equation (2);

[0017]

[0018] For railway, highway, and waterway transportation modes, a dynamic node set N is also introduced. n,z,o Let n represent the set of demand points that can supply liquid fuel o via method z.

[0019] In one implementation, the objective function for solving the parent or child model is to minimize the total transportation cost under multimodal transport.

[0020] In one implementation, the objective function is expressed as:

[0021]

[0022] In the formula:

[0023] parameter The unit freight cost (CNY / m) from pipeline batch i to pipeline node n p. 3 );

[0024] Parameter c n,n′,z,o The unit freight cost (CNY / m²) for transporting liquid fuel p from supply point n to demand point n′ via method z. 3 );

[0025] variable Represents the volume (m³) of the injection / distribution batch i at node n on pipeline p within the t-th time window. 3 );

[0026] variable This represents the volume (m³) of liquid fuel o delivered by node n using method z within time window t. 3 ).

[0027] In one implementation, the constraints involved in the solution process of the parent model or the child model include: pipeline transportation constraints, and one or more of railway transportation constraints, highway transportation constraints, and waterway transportation constraints; it also includes node inventory constraints.

[0028] This invention achieves a breakthrough in efficiency through a multi-round dynamic divide-and-conquer strategy. The method divides the 30-day scheduling cycle into multiple progressive solutions. This innovative hierarchical "parent-child" problem-solving mechanism, combined with a dynamic divide-and-conquer strategy based on batch transport similarity, effectively controls the expansion of the variable space. The implementation of this technology has yielded significant practical benefits: in terms of computational efficiency, the total solution time is only 43.1 CPUs, reducing planning time by 95% compared to traditional methods; in terms of algorithm robustness, even with a 15-fold increase in the number of time windows, the variable increase can still be controlled within 38%. This method, through a dynamic divide-and-conquer strategy, maintains the solution accuracy of the original problem while significantly reducing computational complexity, providing a highly efficient solution for ultra-large-scale logistics scheduling problems that combines engineering practicality and theoretical innovation. Attached Figure Description

[0029] Figure 1 This is a schematic diagram illustrating the principle of a detailed embodiment of the method of this application;

[0030] Figure 2This is a schematic diagram of the transportation batch of pipeline P1 obtained from a calculation example in this application;

[0031] Figure 3 This is a schematic diagram of the transportation batch of pipeline P2 obtained from a calculation example in this application. Detailed Implementation

[0032] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the described embodiments of the present invention are within the scope of protection of the present invention.

[0033] To address the problems of existing technologies, embodiments of the present invention provide a solution method for multi-mode intermodal transport of liquid fuels based on a dynamic divide-and-conquer strategy, the method comprising:

[0034] (1) Based on the network parameters and initial logistics plan in the cross-regional liquid fuel transportation network, set the initial value of the dynamic set of transportation batches of each node in the transportation network; wherein, the network parameters include the type of transportation mode, the type of liquid fuel, the transportation mode, and several nodes in the transportation route of each transportation mode;

[0035] (2) First iteration: Divide the time period of the initial logistics plan into multiple sub-time windows evenly; construct the parent model of the mixed integer linear programming model based on the multiple sub-time windows after the division, and solve the dynamic set of transportation batches of each node corresponding to the current iteration based on the parent model;

[0036] (3) Second iteration: Each sub-time window of the previous round is further divided into several sub-time windows, and the sub-model of the mixed integer linear programming model is further constructed. Based on the sub-model, the dynamic set of transportation batches of each node under this iteration round is obtained.

[0037] (4) More rounds of iteration: update the sub-model to the parent model, further divide the sub-time window of the previous round, dynamically set the transportation batches of each node based on the parent model, and continue to divide the sub-time window and construct and solve the sub-model until the dynamic set of transportation batches solved by the final sub-model meets the preset time accuracy requirements.

[0038] The following is based on Figure 1 The principles of the method described above will be explained in one or more more detailed embodiments.

[0039] like Figure 1 The method includes the following steps:

[0040] Step (1), as follows Figure 1 As shown in the first line, for a logistics plan with a period of H, using... Break it down into time nodes. A time window, a set of time windows The length of each time window is Use dynamic collections This represents the set of batches that can be injected / distributed from station n on pipeline p within the t-th time window. Since the transport process of these batches is unknown in the first solution, therefore... This basically includes all batches injected into the station this month. Assume I... max,p Let p be the total number of batches. Based on this, a multimodal transport optimization model is constructed, considering constraints such as pipeline batch transport, road / rail / waterway transport channels, and inventory change constraints. This model is considered the "parent model." The optimal logistics plan is obtained using a solver, including pipeline batch transport plans, road transport plans, rail transport plans, and waterway transport plans. The pipeline batch transport plan includes the total volume of injection and distribution station operations along each pipeline within each time window, as well as the oil head position of each batch within the pipeline at each time point. and oil tail position Go to (2).

[0041] Step (2) involves splitting each time window of the "parent model" in step (1) into m (m ≥ 2, e.g., ...) Figure 1 (As shown in the second row), then the number of time windows becomes Time window set The length of each time window is 1 / m times that of the parent model, that is... The next step will be to build a "sub-model" based on the refined time window. The parameter td will be introduced. t The time window relationship representing the "parent-child model", td t This represents the time window number in the parent model corresponding to the t-th time window in the sub-model. For example, the first time window of the parent model contains the first to m time windows in the sub-model, DT1 = {1,..,m}. For each time window t∈T in the parent model... 1 oil nozzle position and oil tail position To determine the passing batches of each station within the corresponding time window of the sub-model, a dynamic set is used. Let represent the set of possible injection / distribution batches at nodes n in pipeline p within the t-th time window of the sub-model. Based on this, the sub-model for multimodal transport optimization is reconstructed. The optimal logistics plan is obtained using a solver, where the head and tail positions of the batches within the pipeline at each time point are denoted as . and oil tail position If the time window τ of the sub-model is greater than 24h, the accuracy of the logistics plan has not yet met the requirements of the daily plan, and proceed to (3); otherwise, end the calculation.

[0042] Step (3) converts the sub-model constructed in step (2) into a parent model, then proceeds to step (2) to further subdivide the time window of the new parent model, updates the time window length τ, and utilizes the time window set T of the sub-model. 2 Oil head and oil tail positions Update T to the parent model 1 , Based on this, new sub-models are constructed.

[0043] In a more detailed embodiment, the description is based on a mathematical model.

[0044] (1) Dynamic set

[0045] This model uses discrete-time representation, where t∈T represents the set of time points, and the number of time windows is one less than the number of time nodes. b∈N represents the set of nodes, where N... s Let N represent the set of supply points. D Let N = N S ∪N D o∈O represents the type of liquid fuel, O n This represents the type of liquid fuel supplied or demanded by node n. z∈Z={train,turck,ship} represents the set of non-pipeline transportation modes. Since railways, highways, and waterways use "container" and "point-to-point" transportation modes, which differ significantly from the continuous sequential transportation process of pipelines, they need to be modeled separately.

[0046] For pipelines, p∈P represents the set of pipelines, and N p Let p represent the set of nodes on the pipe. Let p represent the set of supply points on pipeline p. Let p represent the set of demand points on the pipeline. I represents the batch set. p I represents the dynamic set of batches that may pass through pipeline p throughout the entire cycle. p,n I represents the dynamic batch set that may pass through node n on pipeline p throughout the entire cycle. t,p,n This represents the set of batches that can be injected / distributed by node n on pipeline p within the t-th time window. Dynamic set I p,n and I t,p,n It needs to be updated based on the batch position information of the parent model. (Using parameters) and This represents the position of the start and end points of batch i at each time point in the logistics plan of the parent model. If, in the logistics plan obtained from the parent model, the end point of batch i does not exceed node n on pipeline p at the start of the cycle, but the start point exceeds node n on pipeline p at the end of the cycle, then the dynamic batch set I of batch i at node n on pipeline p is... p,n In the middle, see equation (1). For each time window t < |T| of the sub-model, if the batch i oil tail is in the corresponding time window td of the parent model t If a batch i starts before exceeding node n but ends after exceeding node n, then that batch i is in the dynamic set I. t,p,n See equation (2). Through batch dynamic set I t,p,n It can significantly reduce the number of binary variables related to the subscripts t, p, n, and i in the multimodal transport model, thereby reducing the complexity of model solving and improving the efficiency of solving.

[0047]

[0048] For transportation modes such as railways, highways, and waterways, a dynamic node set N is also introduced. n,z,o Let n represent the set of demand points that can supply liquid fuel o via method z.

[0049] (2) Objective function

[0050] The objective function of multimodal transport optimization is to minimize the transport cost C. The transport cost consists of pipeline, rail, road, and waterway transport costs, respectively; therefore, the total cost is the sum of the product of the unit cost and the corresponding transport volume. As shown in equation (3), the first term is the pipeline transport cost, and the second term is the sum of the non-pipeline transport (rail, water, road) costs. In sequential pipeline transport, oil products with similar properties are generally transported adjacently; therefore, the pipeline transport sequence is generally fixed. The type of liquid fuel transported in each batch is known, but the batch quantity will be adjusted according to supply and demand changes, requiring model-based decision-making. c n,n′,z,o This represents the unit freight cost for both pipeline and non-pipeline transport. For each mode of transport, the freight cost is calculated based on the actual volume received, therefore using the batch volume received at each demand point within each time window. as well as This indicates the transport volume by both pipeline and non-pipeline methods.

[0051]

[0052] In the formula:

[0053] parameter —Unit freight cost from pipeline batch i to pipeline node n p (CNY / m 3 );

[0054] Parameter c n,n′,z,o—Unit freight cost (CNY / m²) for supply point n to transport liquid fuel o to demand point n′ via method z. 3 );

[0055] variable —The volume (m³) of batch i injected / distributed at node n on pipeline p within the t-th time window. 3 );

[0056] variable —Node n′ receives the volume (m³) of liquid fuel o delivered by node n using method z within time window t. 3 ).

[0057] (3) Restrictions on rail, road and waterway transportation

[0058] Compared to the 1-2 m / s transport speed of pipelines, point-to-point transportation by rail, road, and waterway is punctual and efficient. Let the integer parameter Δt n,n′,z This represents the number of time windows required to transport liquid fuel from node n to node n′ via method z, i.e., the transit time. Specifically, node n transports liquid fuel via method z in time window t. If liquid fuel o reaches node n′, then node n′ will be at the t+Δt point. n,n′,z The same volume of liquid fuel is received within the time window, as shown in Equation (4). Equation (4) applies only to Δt within this scheduling cycle. n,n′,z The time window then takes effect, and for Δt n,n′,z Before the time window, the oil received by node n′ must have come from oil dispatched at a later time point in the previous scheduling cycle, and is a known product in transit. Therefore, equation (5) is used as a constraint. Equation (6) restricts the transport volume of each transport channel from exceeding its upper limit.

[0059] In the formula:

[0060] Parameter Δt n,n′,z —The number of time windows required to transport liquid fuel from node n to node n′ using method z;

[0061] variable —The volume (m³) of liquid fuel o sent by node n to node n′ in time window t using method z. 3 );

[0062] parameter —Node n′ receives the volume (m³) of liquid fuel o emitted by node n in mode z within time window t. 3 );

[0063] parameter —The maximum capacity (m) that node n sends to node n′ liquid fuel o using method z within a single time window. 3 ).

[0064] (4) Pipeline transportation constraints

[0065] RC t,p,i and LC t,p,i Let V represent the oil head (right coordinate) and oil tail (left coordinate) of batch i in pipe p at time point t. Equation (7) represents the batch volume V in the pipe. t,p,i This is equal to the difference in volume coordinates between the head and tail of the oil at the same time point. Since the compressibility of liquid fuel is negligible, liquid fuel always fills the entire pipeline. Equation (8) represents all batches i∈I of pipeline p. p The volume then equals the tube capacity at any given time point. Equation (9) indicates that the head coordinate of batch i in pipeline p is equal to the sum of the volumes of that batch and subsequent batches (i′≥i) in the same pipeline. Equation (10) is the mass conservation constraint for batch volume, where the volume of batch i in pipeline p at time t is equal to the volume at the previous time t-1 plus the volume of all supply points on the pipeline. The volume injected for this batch, minus all required points. The volume of this batch is distributed. This patent utilizes dynamic set I. t,p,n control variables The number of items. If batch i is not in the dynamic set I. t-1,p,n Inside, the corresponding variables If it does not exist, then the value is 0.

[0066]

[0067] In the formula:

[0068] Variable RC t,p,i —Coordinates of oil head volume (right coordinate) of batch i within pipeline p at time point t (m) 3 );

[0069] Variable LC t,p,i —Coordinates of the tail volume of batch i within pipeline p at time point t (left coordinate) (m) 3 );

[0070] variable V t,p,i —Volume of batch i within pipe p at time point t (m³) 3 );

[0071] parameter —pipe volume (m) 3 );

[0072] parameter —The pipe capacity (m) of pipeline batch i at the beginning of the scheduling cycle. 3 ).

[0073] Introducing binary variables This indicates whether node n in pipeline p is injected or transported in batches i within time window t. If Therefore, the oil tail coordinates of batch i at time node t do not exceed the station location σ. l,n (Equation (11)), and the oil head coordinate at time node t has exceeded σ. l,n (Equation (12)). Furthermore, Equation (13) constrains the operational flow rate to meet the upper limit requirement and not exceed it. Equation (14) limits the maximum volume of the distribution batch i of node n within the time window t. This volume cannot exceed the total volume of the pre-station batch plus the total injection volume of the upstream station of station n minus the total distribution volume.

[0074]

[0075] In the formula:

[0076] binary variable —Whether pipeline node n is injected or transported in batches i within time window t;

[0077] Parameter σ p,n —Volume coordinates of node n of pipe p (m) 3 );

[0078] Parameter τ—length of the time window (h);

[0079] parameter —Maximum injection / distribution flow rate at node n of pipeline p (m³) 3 / h).

[0080] Introducing variables This represents the transport volume of the downstream pipe segment of node n in pipeline p within time window t. For each time window, the transport volume of the downstream pipe segment of node n in pipeline p is equal to the volume of that node and the volume of the upstream node. The sum of the injection volumes minus the sum of the distribution volumes is shown in equation (15). Considering the pipeline's transportation capacity, equation (16) limits the pipeline's transport flow rate to no more than [a certain value].

[0081]

[0082] In the formula:

[0083] Variable FS t,p,n —Transport volume (m³) of the downstream pipe segment at node n of pipeline p within time window t. 3 );

[0084] parameter —The maximum transport flow rate (m³) of the downstream pipe segment of node n at node p within the time window t. 3 / h).

[0085] (5) Node inventory constraints

[0086] Equation (17) gives the inventory change trend of all nodes during the scheduling period. Parameters q represents the initial inventory of liquid fuel o at node n. t,n,o It is the production or consumption rate of liquid fuel o in time window t-1. Specifically, the inventory at time point t is equal to the inventory at the previous time point plus the total amount received by pipelines, railways, highways, and waterways within time window t-1, minus the amount shipped out within the same time window. Equation (18) constrains the reasonable range of inventory.

[0087]

[0088] In the formula:

[0089] variable —The inventory (m) of liquid fuel o at node n at time t 3 );

[0090] parameter —Initial inventory (m) of liquid fuel o at node n 3 );

[0091] Parameter q t-1,n,o — Rate of change of liquid fuel o at node n within the time window (m) 3 / h);

[0092] Binary parameter y p,i,o —Whether pipeline p batch i transports liquid fuel o;

[0093] parameter —Lower inventory limit of liquid fuel o at node n (m) 3 );

[0094] parameter —The upper limit of liquid fuel o at node n (m) 3 ).

[0095] The technical effects of the method in this application will be illustrated below using a more specific transportation network and actual numerical calculations.

[0096] This paper analyzes a logistics system in a certain northwest and southwest region of China. The region has 25 nodes, using two pipelines, a railway, and a highway to transport three types of liquid fuels. All nodes are on pipelines, but only liquid fuels #1 and #3 can be transported via pipeline; liquid fuel #2 can only be transported via railway and highway.

[0097] The basic information about the pipelines and nodes is shown in the table below.

[0098]

[0099]

[0100] Initially, pipeline P1 contains 5 old batches in the following order: 3#-1#-3#-1#-3#; pipeline P2 contains 3 old batches in the following order: 3#-1#-3#. During this cycle, both pipelines transport goods using an alternating injection method of 1# and 3#. The rate of change for each node is shown in the table below, where positive values ​​represent the generation rate of supply points and negative values ​​represent the consumption rate of demand points.

[0101] node <![CDATA[1# Fuel (m 3 / day)]]> <![CDATA[2# Fuel (m 3 / day)]]> <![CDATA[3# Fuel (m 3 / day)]]> N1 5629 0 10669 N2 -90 0 -1093 N3 0 0 -238 N4 448 0 795 N5 0 0 -159 N6 0 0 -557 N7 0 0 -199 N8 0 0 -1356 N9 -1903 0 -4271 N10 8098 1176 9347 N11 -298 0 -507 N12 -139 0 -675 N13 -188 0 -497 N14 -423 -134 -1193 N15 -497 0 -866 N16 -634 -134 -929 N17 -569 0 -1013 N18 1399 0 2166 N19 -669 -67 -625 N20 -2173 0 -3665 N21 -305 0 -364 N22 -199 0 -531 N23 -162 0 -587 N24 -162 0 -1101 N25 -2014 0 -3284

[0102] The planning cycle is approximately 30 days. The number of time nodes in the sub-model is set to 2, 5, and 3 respectively. Specifically, in the first round of calculation, there are 3 time nodes, 2 time windows, and a time window length of 360 hours; in the second round, there are 11 time nodes, 10 time windows, and a time window length of 72 hours; and in the third round, there are 31 time nodes, 30 time windows, and a time window length of 24 hours. The model size and calculation results are shown in Table 3. The total time consumed by the model solution is 43.1 CPUs, and the objective function for daily dispatch is 254,186 yuan. Compared with existing technologies, this invention reduces logistics planning time by 95%, significantly improving scheduling decision-making efficiency. This is because the number of binary variables in a single iteration of the model in this invention is far lower than that in the complete model. In each iteration, the proposed method uses a decomposition method to sequentially divide the time window of a large-scale complex logistics plan into several parent-child problems (the number of time windows in the parent problem is m times that in the child problem). Based on this, a dynamic divide-and-conquer strategy based on batch transport similarity is proposed. The batch transport information in the calculation results of the parent model is used to update the dynamic set of batches in the child model. When the time window increases, the increase of relevant binary variables and continuous variables is well controlled, which ultimately greatly reduces the variable search space of the original problem. Through the dynamic interaction of the results of multiple rounds of parent-child problems, efficient optimization of large-scale logistics plans is achieved.

[0103] The numerical results for comparison are shown in the table below:

[0104]

[0105] Furthermore, the more detailed results of this example are as follows:

[0106] The railway transportation plan is shown in the table below:

[0107] Shipping time Reception time supply points Demand points <![CDATA[Transport volume (m 3 )]]> Day 8 Day 10 N19 N20 1712 Day 9 Day 11 N19 N20 90 Day 10 Day 12 N19 N15 580 Day 10 Day 12 N19 N20 90 Day 11 Day 13 N19 N17 1982 Day 12 Day 14 N19 N17 180 Day 13 Day 15 N19 N17 180 Day 14 Day 16 N19 N17 180 Day 15 Day 17 N19 N17 180 Day 16 Day 18 N19 N17 99 Day 30 Day 32 N19 N17 180

[0108] In this example, the two pipelines P1 and P2 are respectively configured as follows: Figure 2 and Figure 3 .

[0109] In summary, the large-scale multimodal transport optimization method proposed in this invention achieves a breakthrough in efficiency through a multi-round dynamic divide-and-conquer strategy. This method divides the 30-day scheduling cycle into multi-stage progressive solutions. This innovative hierarchical "parent-child problem" solution mechanism, combined with a dynamic divide-and-conquer strategy based on batch transport similarity, effectively controls the expansion of the variable space. The implementation of this technology has yielded significant practical benefits: in terms of computational efficiency, the total solution time is only 43.1 CPUs, reducing the planning time by 95% compared to traditional methods; in terms of algorithm robustness, even with a 15-fold increase in the number of time windows, the variable increase can still be controlled within 38%. This method, through a dynamic divide-and-conquer strategy, maintains the solution accuracy of the original problem while significantly reducing computational complexity, providing a highly efficient solution for ultra-large-scale logistics scheduling problems that combines engineering practicality and theoretical innovation.

[0110] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working process of the above-mentioned system (device) and module unit can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0111] In the embodiments provided by this invention, it should be understood that the disclosed systems and methods can be implemented in other ways. For example, the system device embodiments described above are merely illustrative. For instance, the division of the above-described module units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be an indirect coupling or communication connection through some interfaces, devices, or units, and may be electrical, mechanical, or other forms.

[0112] The integrated units implemented as software functional units described above can be stored in a computer-readable storage medium. These software functional units, stored in a storage medium, include several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) or processor to execute some steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0113] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A liquid fuel multi-modal intermodal solution method based on dynamic divide-and-conquer strategy, characterized in that, The method comprises: (1) setting initial values of dynamic sets of transportation batches of each node in a transportation network according to network parameters and an initial logistics plan in the cross-regional liquid fuel transportation network; wherein the network parameters comprise types of transportation modes, types of liquid fuels, transportation modes, and a plurality of nodes in transportation lines of each transportation mode; (2) first iteration: uniformly splitting a time period of the initial logistics plan into a plurality of sub-time windows; constructing a parent model of a mixed integer linear programming model according to the plurality of sub-time windows after splitting, and solving the parent model to obtain dynamic sets of transportation batches of each node corresponding to the current iteration round; (3) second iteration: continuing to uniformly split each sub-time window of the previous round into a plurality of sub-time windows, continuing to construct a sub-model of the mixed integer linear programming model, and solving the sub-model to obtain dynamic sets of transportation batches of each node under the iteration round; (4) more rounds of iteration: updating the sub-model to the parent model, and further dividing the sub-time window of the previous round, obtaining the dynamic sets of transportation batches of each node based on the parent model, and continuing the division of the sub-time window and the construction and solving of the sub-model until the dynamic sets of transportation batches solved by the final sub-model meet the preset time accuracy requirement; In the (1), the transportation modes at least comprise pipeline transportation, and further comprise one or more combinations of railway transportation, highway transportation, and waterway transportation; each node is distributed on each line of pipeline transportation, or on lines of railway transportation, highway transportation, and waterway transportation, and the node types comprise supply points, distribution points, and demand points; The network parameters further comprise batch oil head and oil tail positions of each time node in the pipeline.

2. The dynamic divide-and-conquer strategy based multi-modal liquid fuel intermodal transport solution method according to claim 1, wherein, The parent model or the child model adopts a discrete time representation, and the time point set is represented by The number of time windows is less than the number of time nodes by one; The node set is represented by The supply point set is represented by The demand point set is represented by ; The kind of liquid fuel is represented by The node The kind of liquid fuel supplied or demanded by the node; The transport mode set of non-pipeline transportation is represented by For pipeline transportation mode Represents a set of pipes. Indicates pipeline The set of nodes on, Indicates pipeline The collection of supply points on the platform Indicates pipeline The set of demand points on the platform , Represents a batch set; This indicates that the pipeline may pass through during the entire cycle. Dynamic batch set, This indicates that the pipeline may pass through during the entire cycle. upper node Dynamic batch set, Indicates the first Within a time window On the pipe The set of batches that a node may inject / distribute; a dynamic set. and It needs to be updated based on the batch position information of the parent model; using parameters. and This represents the position of the first and last oil tankers in each time point of the batch in the logistics plan of the parent model; if in the logistics plan obtained from the parent model, the batch... The oil tail did not exceed the time at the start of the cycle. On the pipe At the node, its oil head exceeded [a certain value] at the end. On the pipe Node, then this batch exist On the pipe Dynamic batch set of nodes In the middle, see equation (1); for each time window of the sub-model In other words, if batch Oil tail in the corresponding time window of the parent model The start time has not exceeded the node. At the end, its oil head exceeded the node. Then this batch In dynamic collections See equation (2); (1) (2) For the railway, highway, waterway transport mode, the dynamic node set is also introduced representing supply points capable of passing by means delivering liquid fuel a set of demand points.

3. The dynamic divide-and-conquer strategy based multi-modal liquid fuel intermodal transport solution method according to claim 2, wherein, An objective function of the solving of the parent model or the sub-model is to minimize the total transportation cost under the combined transportation of multiple transportation modes.

4. The dynamic divide-and-conquer strategy based multi-modal liquid fuel intermodal transport solution method according to claim 3, wherein, The objective function is expressed as: In the formula: Parameter Representative pipe shipment batch To pipeline Node Unit freight, unit CNY / m 3 ; Parameter Representing supply point By way Delivering liquid fuel To demand point Unit freight, unit CNY / m 3 ; variable Representing the Within a time window On the pipe Node injection / batch distribution Volume, in m 3 ; variable Representative node In the time window Internal receiving node Method Transporting liquid fuel Volume, in m 3 .

5. The dynamic divide-and-conquer strategy based multi-modal liquid fuel intermodal transport solution method according to claim 4, wherein, The parent model or the sub-model involves constraint conditions in the solving process, which comprise one or more of pipeline transportation constraints, railway transportation constraints, highway transportation constraints, and waterway transportation constraints; and further comprise node inventory constraints.

Images