Chemical supply chain intelligent logistics scheduling optimization system based on big data analysis

Abstract

The application relates to the technical field of supply chain logistics scheduling, and particularly discloses a chemical supply chain intelligent logistics scheduling optimization system based on big data analysis, which belongs to the technical field of supply chain logistics scheduling. A data acquisition module is arranged in the system to obtain a multi-agent multi-scheduling cycle supply chain correlation data set, a supply and demand analysis module generates a supply and demand gap distribution map according to order demand and inventory state data, and a road condition evaluation module generates a road traffic cost matrix based on vehicle positioning and road traffic data. An improved Dijkstra algorithm with a node weight dynamic updating mechanism and an edge cost time-varying attenuation mechanism is adopted in a scheduling optimization module, joint path optimization is carried out on the two types of graph matrices, an optimal logistics scheduling scheme is generated, and a command issuing module generates and issues a vehicle and path planning instruction set according to the scheduling scheme, so that the dynamic intelligent logistics scheduling demand of the chemical supply chain is adapted.

Description

Technical Field

[0001] This invention relates to the field of supply chain logistics scheduling technology, specifically to a smart logistics scheduling and optimization system for the chemical supply chain based on big data analysis. Background Technology

[0002] Chemical supply chain logistics scheduling involves multiple stakeholders, including raw material suppliers, chemical plants, and product customers. The operation is divided into multiple scheduling cycles, and scheduling relies on various basic data such as order demand, inventory status, vehicle location, and real-time road conditions. Existing chemical supply chain logistics scheduling technologies mostly employ static data processing models, using only partial, single-dimensional data for scheduling analysis. This fails to comprehensively collect and integrate complete supply chain-related data across multiple stakeholders and scheduling cycles, and cannot simultaneously incorporate order, inventory, vehicle location, and road condition data for unified processing.

[0003] In the operation of existing technologies, supply and demand relationship analysis and road condition route assessment are carried out independently, and only simple data statistics are performed separately. It is impossible to complete a systematic supply and demand matching analysis based on order demand data and inventory status data, nor can it complete a quantitative route capacity assessment based on transport vehicle positioning data and real-time road traffic data. It is difficult to form a standardized distribution of supply and demand gaps and a quantitative representation of road traffic costs.

[0004] Conventional path optimization typically employs Dijkstra's algorithm with fixed parameters. This algorithm lacks the ability to dynamically update node weights and the time-varying decay of edge costs, making it unsuitable for adapting to the dynamic changes in the supply chain. Existing technologies do not perform joint optimization calculations based on the supply-demand gap distribution results and road traffic cost matrices, lacking a multi-dimensional data linkage optimization mechanism. Furthermore, they cannot automatically generate standardized scheduling instructions and complete targeted distribution, making it difficult to meet the application requirements of intelligent and interconnected logistics scheduling throughout the entire chemical supply chain. Summary of the Invention

[0005] The purpose of this invention is to provide an intelligent logistics scheduling and optimization system for the chemical supply chain based on big data analysis, so as to solve the problems mentioned above.

[0006] The objective of this invention can be achieved through the following technical solutions: A smart logistics scheduling and optimization system for the chemical supply chain based on big data analytics includes: The data acquisition module acquires a set of supply chain-related data from multiple chemical raw material suppliers, multiple chemical plants, and multiple chemical product customers in the chemical supply chain over multiple scheduling cycles. The set of supply chain-related data includes order demand data, inventory status data, transport vehicle location data, and real-time road traffic data for each scheduling cycle. The supply and demand analysis module performs supply and demand matching analysis on the order demand data and the inventory status data to generate a supply and demand gap distribution map for each scheduling cycle. The road condition assessment module performs path capacity assessment processing on the vehicle location data and the real-time road traffic data to generate a road traffic cost matrix for each scheduling cycle. The scheduling optimization module calls the improved Dijkstra algorithm to perform joint path optimization processing on the supply and demand gap distribution map and the road traffic cost matrix, and generates the optimal logistics scheduling scheme corresponding to the scheduling cycle. The improved Dijkstra algorithm is based on the dynamic update mechanism of node weights and the time-varying decay mechanism of edge costs. The instruction issuing module generates a set of vehicle dispatch instructions and a set of route planning instructions based on the optimal logistics scheduling scheme, and sends the set of vehicle dispatch instructions and the set of route planning instructions to the chemical supply chain scheduling center.

[0007] As a further aspect of the present invention, supply and demand matching analysis is performed on the order demand data and the inventory status data to generate a supply and demand gap distribution map for each scheduling cycle, including: Extract the product type demand and demand time window of each chemical product customer within the target scheduling cycle from the order demand data, and extract the available inventory of each chemical plant of various chemical products at the beginning of the target scheduling cycle from the inventory status data. For each chemical product, calculate the total demand for that chemical product from all customers, and the total available inventory of that chemical product from all chemical plants; The total demand is calculated by subtracting the total available inventory to obtain the global supply and demand gap for this chemical product. When the global supply-demand gap is positive, it is determined that replenishment needs to be made from the chemical raw material supplier to the chemical plant. When the global supply and demand imbalance is negative, it is determined that the chemical plant needs to ship goods to the chemical product customer. For each chemical plant, the available inventory of the plant is compared with the demand of chemical product customers within the plant's coverage area, and the types of products with insufficient inventory and the types of products with excess inventory are marked. The supply and demand gap distribution map is constructed by using the geographical coordinates of the chemical plant as nodes, the shortage quantity corresponding to the product type with insufficient inventory as the negative weight of the node, and the surplus quantity corresponding to the product type with excess inventory as the positive weight of the node.

[0008] As a further aspect of the present invention, path capacity assessment processing is performed on the vehicle positioning data and the real-time road traffic data to generate a road traffic cost matrix for each scheduling cycle, including: Extract the real-time latitude and longitude coordinates and driving speed values ​​of all transport vehicles in the current scheduling cycle from the transport vehicle positioning data, and extract the length value, number of lanes value, and current traffic flow value of each road segment in the road network from the real-time road traffic data. Based on the real-time latitude and longitude coordinates, each transport vehicle is matched to the nearest road segment to obtain the vehicle density value on each road segment; For each road segment, the vehicle density value, the traffic flow value, and the number of lanes are input into a preset road segment travel time estimation function to generate the estimated travel time value for that road segment. The estimated travel time is divided by the length of the road segment to obtain the unit distance travel cost of the road segment. The starting node identifiers, ending node identifiers, and corresponding unit distance travel costs of all road segments are organized in the form of node pairs to generate the road travel cost matrix.

[0009] As a further aspect of the present invention, the step of calling the improved Dijkstra algorithm to perform joint path optimization processing on the supply-demand gap distribution map and the road traffic cost matrix to generate the optimal logistics scheduling scheme corresponding to the scheduling cycle includes: Extract all positive weight nodes and all negative weight nodes from the supply and demand gap distribution map, and use each positive weight node as the source node for goods to be sent out and each negative weight node as the target node for goods to be received. For each source node, the surplus value corresponding to that source node is taken as the total amount of goods to be scheduled. Using the road traffic cost matrix as the initial edge cost between nodes, the improved Dijkstra algorithm is called to calculate the optimal transportation path from each source node to each target node and the total path cost corresponding to the optimal transportation path. Based on the total amount of goods to be scheduled at each source node and the total path cost from each source node to each target node, a transportation task allocation matrix is ​​constructed, wherein the rows of the transportation task allocation matrix correspond to the source nodes and the columns correspond to the target nodes. The minimum cost flow optimization process is performed on the transportation task allocation matrix to generate the number of goods scheduled from each source node to each target node. The number of goods scheduled from each source node to each target node and the corresponding optimal transportation path are taken as the optimal logistics scheduling scheme.

[0010] As a further aspect of the present invention, the improved Dijkstra algorithm, based on a dynamic node weight update mechanism and a time-varying edge cost decay mechanism, includes: The node weight dynamic update mechanism is as follows: During the path search process, whenever a node is moved from the open set to the closed set, the priority ranking value of the neighboring nodes is dynamically updated according to the weighted sum of the node's current accumulated path cost and the node weight value of the node in the supply and demand gap distribution map. When the node weight value is positive, the priority ranking value of the neighboring nodes is reduced, and when the node weight value is negative, the priority ranking value of the neighboring nodes is increased. The time-varying edge cost decay mechanism is as follows: during the path search process, whenever an edge is selected as part of the optimal path, the current edge cost of that edge is multiplied by a preset decay coefficient to generate the updated edge cost, and the updated edge cost is used for the re-evaluation of that edge in subsequent path searches.

[0011] As a further aspect of the present invention, the improved Dijkstra algorithm is invoked to calculate the optimal transportation path from each source node to each target node and the total path cost corresponding to the optimal transportation path, including: For each source node and target node pair, initialize the cumulative path cost of the current source node to zero, add the current source node as the current extended node to the open set, and initialize the cumulative path cost of all nodes except the current source node to infinity. Select the node with the smallest priority ranking value from the open set as the new current expansion node, and move the new current expansion node from the open set to the closed set; Obtain all neighbor nodes of the new current extended node. For each neighbor node, skip it if it is in the closed set; otherwise, calculate the candidate path cost from the current source node through the new current extended node to the neighbor node. The candidate path cost is equal to the cumulative path cost of the new current extended node plus the current edge cost of the edge from the new current extended node to the neighbor node. When the candidate path cost is less than the current cumulative path cost of the neighboring node, the cumulative path cost of the neighboring node is updated to the candidate path cost, the predecessor node of the neighboring node is pointed to the new current extension node, and then the priority ranking value of the neighboring node is recalculated based on the weighted sum of the updated cumulative path cost of the neighboring node and the node weight value of the neighboring node in the supply and demand gap distribution map. Repeat the operation of selecting the node with the smallest priority sorting value from the open set as the new current expansion node until the new current expansion node is equal to the current target node, then stop the search; Starting from the current target node, backtracking backward along the predecessor node to the current source node, the sequence of nodes traversed during the backtracking is taken as the optimal transportation path, and the cumulative path cost of the current target node is taken as the total path cost.

[0012] As a further aspect of the present invention, a vehicle dispatching instruction set and a route planning instruction set are generated based on the optimal logistics dispatching scheme, including: Iterate through the scheduling tasks between each source node and each target node in the optimal logistics scheduling scheme. Each scheduling task includes the quantity of goods to be scheduled, the type of goods, the origin of the goods, the destination of the goods, and the optimal transportation route. For each scheduling task, a set of candidate vehicles qualified to transport that type of cargo is selected from the transport vehicle database based on the cargo type. The required number of vehicles is determined from the candidate vehicle set based on the quantity of goods to be dispatched, and the currently idle vehicle that is closest to the departure point of the goods is selected as the dispatch vehicle. A vehicle dispatch instruction is generated for each dispatched vehicle. The vehicle dispatch instruction includes the vehicle's license plate number, cargo loading location, cargo loading time window, target unloading location, and target unloading time window. Based on the node sequence in the optimal transportation route, a route planning instruction is generated for each dispatched vehicle. The route planning instruction includes the sequence of intersection markers and road segment markers that the dispatched vehicle needs to pass through in sequence. All vehicle dispatch instructions are aggregated into the vehicle dispatch instruction set, and all route planning instructions for all dispatched vehicles are aggregated into the route planning instruction set.

[0013] As a further aspect of the present invention, after generating the optimal logistics scheduling scheme, the following steps are also included: Obtain historical logistics scheduling plans executed within the historical scheduling period and corresponding real-time road traffic data to build a historical scheduling experience database. The current optimal logistics scheduling scheme generated in the current scheduling cycle is compared with the historical logistics scheduling schemes in the historical scheduling experience base to filter out historical logistics scheduling schemes with a similarity exceeding a preset similarity threshold. Extract real-time traffic data of historical roads corresponding to the selected historical logistics scheduling schemes, and calculate the historical average congestion duration of each road segment under similar time windows. Based on the historical average congestion duration, the optimal transportation route in the current optimal logistics scheduling scheme is marked with congestion risk, and congestion-prone sections on the optimal transportation route whose congestion risk exceeds the risk threshold are marked. Remove the original transport route that passes through the congested section from the current optimal logistics scheduling scheme, re-invoke the improved Dijkstra algorithm to generate an alternative transport route that bypasses the congested section, and incorporate the alternative transport route into the current optimal logistics scheduling scheme.

[0014] As a further aspect of the present invention, after generating the optimal logistics scheduling scheme, the following steps are also included: Extract the actual order completion data and actual vehicle travel data after the end of each scheduling cycle from the aforementioned supply chain-related data set; The actual order completion data is compared with the expected order completion data in the optimal logistics scheduling scheme, and the order completion deviation rate of each scheduling task is calculated. The actual vehicle travel data is compared with the planned vehicle travel data in the optimal logistics scheduling scheme to calculate the path execution deviation rate for each scheduling task. When the order completion deviation rate exceeds the preset deviation threshold or the path execution deviation rate exceeds the preset deviation threshold, the corresponding scheduling task is marked as an abnormal scheduling task. Extract historical real-time road traffic data within the scheduling period corresponding to the abnormal scheduling task, identify the abnormal road segments and abnormal time periods that cause the deviation, increase the unit distance traffic cost of the abnormal road segments within the abnormal time periods by a penalty step, and generate a corrected road traffic cost matrix for path optimization processing in the next scheduling period.

[0015] As a further embodiment of the present invention, the system includes a processor and a memory, wherein the memory stores a computer program, and when the computer program is executed by the processor, it implements the above-described intelligent logistics scheduling and optimization system for the chemical supply chain based on big data analysis.

[0016] The beneficial effects of this invention are: The algorithm architecture incorporates a dynamic node weight update mechanism and a time-varying edge cost decay mechanism, adapting the Dijkstra algorithm for better compatibility. During algorithm operation, it can autonomously adjust weight parameters based on the real-time status of nodes in the supply chain network, while simultaneously performing adaptive decay iterations of road edge costs over time. The algorithm no longer maintains a fixed parameter operating mode, adapting to the continuously changing network state characteristics within the scheduling cycle. It relies on its own operational logic to autonomously iterate and update parameters, adapting to the path-solving logic requirements of complex scheduling scenarios.

[0017] Based on various types of supply chain-related data, supply and demand matching analysis and route capacity assessment are conducted to generate structured supply and demand gap distribution maps and road access cost matrices. These two types of quantitative results can serve as unified input conditions for computation, enabling hierarchical data processing and structured output. Each functional module is divided into processing logics according to data type. Raw data from different dimensions can enter the corresponding module to complete specialized analysis and computation, forming standardized intermediate representation data, providing regularized data support for subsequent joint optimization.

[0018] Based on an improved algorithm, joint path optimization is performed on the supply-demand gap distribution map and the road traffic cost matrix, breaking the single-dimensional independent planning operation mode and realizing collaborative computation and processing of supply-demand and road condition data. The integrated optimization mode can take into account the supply chain supply and demand arrangement and road traffic attributes, outputting a logistics scheduling plan adapted to the overall operating conditions. Based on the scheduling plan, corresponding vehicle scheduling instruction sets and route planning instruction sets are generated, completing the standardized collection and targeted transmission of instructions, and realizing modular and process-oriented operation of the entire logistics scheduling process. Attached Figure Description

[0019] The invention will now be further described with reference to the accompanying drawings.

[0020] Figure 1 This is a timeline diagram of the operation of a smart logistics scheduling and optimization system for the chemical supply chain based on big data analysis, as described in this invention. Figure 2 This is a flowchart of the route capacity assessment process and the generation of the road traffic cost matrix; Figure 3 It is a flowchart for generating joint path optimization and optimal logistics scheduling scheme. Detailed Implementation

[0021] 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 embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0022] See Figure 1 This invention is a smart logistics scheduling and optimization system for the chemical supply chain based on big data analysis, comprising: The system acquires a set of supply chain-related data from multiple chemical raw material suppliers, chemical plants, and chemical product customers across multiple scheduling cycles through a data acquisition module. This set includes order demand data, inventory status data, transport vehicle location data, and real-time road traffic data for each scheduling cycle. The supply and demand analysis module performs supply and demand matching analysis on the order demand data and inventory status data, generating a supply and demand gap distribution map with chemical plants as nodes and product shortages or surpluses as weights. Simultaneously, the road condition assessment module performs path capacity assessment on the transport vehicle location data and real-time road traffic data, generating a path cost matrix reflecting the real-time traffic costs of the road network. The scheduling optimization module, as the core, calls an improved Dijkstra algorithm. This algorithm integrates a dynamic node weight update mechanism and a time-varying edge cost decay mechanism to perform joint path optimization on the aforementioned supply and demand gap distribution map and road cost matrix, thereby generating the optimal logistics scheduling plan for the current scheduling cycle. This plan clarifies the quantity of goods allocated from surplus nodes (source nodes) to shortage nodes (target nodes) and the specific transportation routes. The instruction issuance module parses and generates a set of vehicle dispatch instructions and a set of route planning instructions based on the optimal logistics scheduling scheme, and issues them to the chemical supply chain scheduling center to drive physical logistics operations.

[0023] In one embodiment of the present invention, in the supply and demand matching analysis process, the product type demand and demand time window of each chemical product customer within the target scheduling cycle are extracted from the order demand data. The available inventory of each chemical product of each chemical plant at the beginning of the target scheduling cycle is extracted from the inventory status data. For each chemical product, the total demand of all chemical product customers for that chemical product and the total available inventory of all chemical plants for that chemical product are calculated. The difference between the total demand and the total available inventory is calculated to obtain the global supply and demand difference of that chemical product. When the global supply and demand difference is positive, it is determined that replenishment needs to be made from the chemical raw material supplier to the chemical plant. When the global supply and demand difference is negative, it is determined that the chemical plant needs to ship goods to the chemical product customer. For each chemical plant, the available inventory of the plant is compared with the demand of chemical product customers within the plant's coverage area. Product types with insufficient inventory and product types with excess inventory are marked. Using the geographical coordinates of the chemical plant as nodes, the shortage of product types with insufficient inventory is used as the negative weight of the node, and the surplus of product types with excess inventory is used as the positive weight of the node, thus constructing a supply and demand gap distribution map that visually represents the supply and demand contradiction.

[0024] In the process of route capacity assessment, refer to Figure 2The system extracts the real-time latitude and longitude coordinates and driving speed of all transport vehicles within the current scheduling cycle from the vehicle location data. It also extracts the length, number of lanes, and current traffic flow of each road segment from the real-time road traffic data. Based on the real-time latitude and longitude coordinates, each transport vehicle is matched to the nearest road segment to obtain the vehicle density value for each road segment. For each road segment, the vehicle density value, traffic flow value, and number of lanes are input into a preset road segment travel time estimation function to generate the estimated travel time value for that road segment. The estimated travel time value is then divided by the length value of the road segment to obtain the unit distance travel cost for that road segment. Finally, the start node identifiers, end node identifiers, and corresponding unit distance travel costs of all road segments are organized into node pairs to generate a road traffic cost matrix.

[0025] In practical implementation, consider an example scenario involving chemical plant A, chemical plant B, chemical product customer C, and chemical product customer D. Order demand data records that chemical product customer C's demand for chemical product P1 within the target scheduling period is 100 tons, with a demand window from 8:00 AM to 12:00 PM; chemical product customer D's demand for chemical product P1 is 80 tons, with a demand window from 2:00 PM to 6:00 PM. Inventory status data records that at the start of the target scheduling period, chemical plant A's available inventory of chemical product P1 is 150 tons, and chemical plant B's available inventory of chemical product P1 is 50 tons. For chemical product P1, the total demand from all chemical product customers is calculated to be 180 tons, and the total available inventory from all chemical plants for chemical product P1 is 200 tons. The total demand is then compared with the total available inventory. The difference between the supply and demand of chemical product P1 is calculated to be -20 tons. When the global supply and demand difference is negative, it is determined that the chemical plant needs to ship the chemical product to the customer. For chemical plant A, the available inventory of 150 tons of chemical plant A is compared with the demand of 100 tons of chemical product customer C within the coverage area of ​​chemical plant A. The surplus of product type P1 with excess inventory is marked as 50 tons. For chemical plant B, the available inventory of 50 tons of chemical plant B is compared with the demand of 80 tons of chemical product customer D within the coverage area of ​​chemical plant B. The shortage of product type P1 with insufficient inventory is marked as 30 tons. The supply and demand gap distribution map is constructed with the geographical coordinates of chemical plant A as the node and the surplus of 50 tons as the positive weight of the node, and the geographical coordinates of chemical plant B as the node and the shortage of 30 tons as the negative weight of the node.

[0026] In practical implementation, the route capacity assessment process involves road segments L1 and L2 in the road network. The vehicle positioning data provides the real-time latitude and longitude coordinates (116.5, 39.9) and driving speed of vehicle V1 (60 km / h). The real-time road traffic data provides the length of road segment L1 (5 km), the number of lanes (3), and the current traffic flow (200 vehicles / hour); the length of road segment L2 (8 km), the number of lanes (2), and the current traffic flow (150 vehicles / hour). Based on the real-time latitude and longitude coordinates, vehicle V1 is matched to the nearest road segment L1, resulting in a vehicle density of 40 vehicles / km on road segment L1 and 20 vehicles / km on road segment L2. For road segment L1, the vehicle density of 40 vehicles / km, the traffic flow of 200 vehicles / hour, and the number of lanes of 3 are input into a preset road segment travel time estimation function to generate the estimated travel time for road segment L1.

[0027] In some embodiments, the preset road segment travel time estimation function is expressed as: in: Indicates the estimated travel time. Indicates the length of the road segment. Indicates the free flow velocity. and It is a correction factor. Indicates vehicle density. Indicates the maximum vehicle density. Indicates traffic flow. Indicates maximum traffic volume. Free-flow speed. Set to 60 km / h, maximum vehicle density Set to 100 vehicles / km, maximum traffic flow Set to 500 vehicles / hour, correction factor and The values ​​are 0.5 and 0.3 respectively. For road segment L1, the road segment length L = 5 km, vehicle density ρ = 40 vehicles / km, and traffic flow q = 200 vehicles / hour are substituted into the formula to calculate the estimated travel time T as 0.12 hours. The estimated travel time value of 0.12 hours is divided by the length value of road segment L1 of 5 km to obtain the unit distance travel cost of road segment L1 as 0.024 hours / km.

[0028] In some embodiments, for road segment L2, the vehicle density value of 20 vehicles / km, the traffic flow value of 150 vehicles / hour, and the number of lanes of 2 are input into a preset road segment travel time estimation function to calculate the estimated travel time value. The estimated travel time value is divided by the length value of road segment L2 of 8 km to obtain the unit distance travel cost of road segment L2. The start node identifiers, end node identifiers, and corresponding unit distance travel costs of all road segments are organized in the form of node pairs to generate a road travel cost matrix.

[0029] Optionally, in the supply and demand matching analysis, a global supply and demand gap calculation is performed separately for various chemical products. For example, if the total demand for chemical product P2 is 120 tons and the total available inventory is 100 tons, the global supply and demand gap is positive 20 tons. When the global supply and demand gap is positive, it is determined that replenishment needs to be made from chemical raw material suppliers to chemical plants. The geographical coordinates of the chemical raw material suppliers are used as nodes, and the replenishment quantity is used as the positive weight of the node, which is then integrated into the supply and demand gap distribution map. Optionally, the vehicle matching process in the route capacity assessment is based on a geographic information system, mapping the latitude and longitude coordinates of transport vehicle positioning data to the nearest road segment in the digital road network. The vehicle density value is obtained by statistically analyzing the number of vehicles matched to the same road segment and dividing by the road segment length. It can be understood that the preset road segment travel time estimation function can be adjusted according to the actual traffic model, for example, by introducing the number of lanes as a denominator to reflect road capacity. It can be understood that the supply and demand gap distribution map is stored in a graph data structure, and the node attributes include geographical coordinates and weight values. A positive weight value indicates surplus, and a negative weight value indicates shortage.

[0030] In one embodiment of the present invention, in the joint path optimization processing of the scheduling optimization module, see [reference needed]. Figure 3 From the supply-demand gap distribution map, all positive-weight nodes and all negative-weight nodes are extracted. Each positive-weight node is designated as the source node for goods dispatch, and each negative-weight node as the target node for goods reception. For each source node, the corresponding surplus value is taken as the total amount of goods to be scheduled. Using the road traffic cost matrix as the initial edge cost between nodes, an improved Dijkstra's algorithm is applied to calculate the optimal transportation path from each source node to each target node and the corresponding total path cost. A transportation task allocation matrix is ​​constructed based on the total amount of goods to be scheduled from each source node and the total path cost from each source node to each target node. Rows in this matrix correspond to source nodes, and columns correspond to target nodes. Minimum-cost flow optimization is performed on the transportation task allocation matrix to generate the quantity of goods to be scheduled from each source node to each target node. The quantity of goods to be scheduled from each source node to each target node and the corresponding optimal transportation path are taken as the optimal logistics scheduling scheme.

[0031] In calculating the optimal transportation path from a single source node to a target node, for each source-target node pair, the cumulative path cost of the current source node is initialized to zero. The current source node is then placed into the open set as the current expansion node, and the cumulative path costs of all nodes except the current source node are initialized to infinity. The node with the smallest priority ranking value from the open set is selected as the new current expansion node. This new current expansion node is then moved from the open set to the closed set. All neighbor nodes of the new current expansion node are obtained. For each neighbor node, if it is in the closed set, it is skipped; otherwise, the candidate path cost from the current source node through the new current expansion node to that neighbor node is calculated. This candidate path cost equals the cumulative path cost of the new current expansion node plus the current edge cost of the edge from the new current expansion node to that neighbor node. When the candidate path cost is less than the current cumulative path cost of the neighbor node, the cumulative path cost of the neighbor node is updated to the candidate path cost. The predecessor node of the neighbor node is then pointed to the new current expansion node. Finally, the priority ranking value of the neighbor node is recalculated based on the weighted sum of its updated cumulative path cost and its node weight in the supply-demand gap distribution graph. Repeatedly execute the operation of selecting the node with the smallest priority ranking value from the open set as the new current expansion node until the new current expansion node is equal to the current target node. Then stop the search and backtrack from the current target node back to the current source node. The sequence of nodes backtracked is taken as the optimal transportation path, and the cumulative path cost of the current target node is taken as the total path cost.

[0032] In practical implementation, the joint path optimization process is based on an example network containing source node factory A, source node factory B, target node warehouse C, and target node warehouse D. The supply and demand gap distribution map records the geographical coordinates of factory A as node N1 with a positive weight of +50, factory B as node N2 with a positive weight of +30, warehouse C as node N3 with a negative weight of -40, and warehouse D as node N4 with a negative weight of -40. All positive weight nodes N1 and N2 and all negative weight nodes N3 and N4 are extracted from the supply and demand gap distribution map. Positive weight nodes N1 and N2 are used as source nodes for goods to be dispatched, and negative weight nodes N3 and N4 are used as target nodes for goods to be received. For source node N1, the surplus value of 50 corresponding to source node N1 is used as the total amount of goods to be dispatched. For source node N2, the surplus value of 30 corresponding to source node N2 is used as the total amount of goods to be dispatched. The road cost matrix serves as the initial edge cost between nodes, including an edge cost of 8 from node N1 to node N3, 5 from node N1 to node N4, 4 from node N2 to node N3, and 6 from node N2 to node N4. The improved Dijkstra algorithm is then used to calculate the optimal transportation path and total path cost for each source node to each target node. Taking the calculation of the path from source node N1 to target node N3 as an example, the cumulative path cost of source node N1 is initialized to zero, and source node N1 is added to the open set as the current expanding node. The cumulative path costs of nodes N2, N3, and N4 are initialized to infinity.

[0033] In some embodiments, the search process for calculating the optimal transportation path from source node N1 to target node N3 is as follows: Select node N1 with the smallest priority ranking value from the open set as the new current expansion node, move node N1 from the open set to the closed set, and obtain the neighbor nodes N3 and N4 of node N1. For neighbor node N3, calculate the candidate path cost from source node N1 through node N1 to neighbor node N3. The candidate path cost is equal to the cumulative path cost 0 of node N1 plus the current edge cost 8 of the edge from node N1 to neighbor node N3, thus obtaining the candidate path cost 8. Since the current cumulative path cost of neighbor node N3 is infinite, the candidate path cost 8 is less than infinite, so update the cumulative path cost of neighbor node N3 to 8, point the predecessor node of neighbor node N3 to node N1, and then recalculate the priority ranking value of neighbor node N3 based on the weighted sum of the updated cumulative path cost 8 of neighbor node N3 and the node weight value of neighbor node N3 in the supply and demand gap distribution diagram -40. A similar operation is performed on neighbor node N4: its cumulative path cost is updated to 5, its predecessor node points to node N1, and its priority ranking value is recalculated. Then, node N4 with the smallest priority ranking value is selected from the open set as the new current expanding node. Node N4 is moved to the closed set, and its neighbor nodes are evaluated. The operation of selecting the node with the smallest priority ranking value from the open set as the new current expanding node is repeated until the new current expanding node equals the target node N3. The search then stops when the target node N3 is found to be backtracking from its predecessor node back to the source node N1. The backtracked node sequence N1->N3 is taken as the optimal transportation path, and the cumulative path cost of the target node N3, 8, is taken as the total path cost.

[0034] In some embodiments, the priority ranking value in the node weight dynamic update mechanism The calculation formula is: in: Represents a node priority sorting value, Represents a node Current cumulative path cost, Represents a node The node weight values ​​in the supply and demand gap distribution map (surplus is positive, shortage is negative). It is a preset weight baseline value used to assign weight values ​​to nodes. Normalization is performed. It is an adjustment factor, accumulating path cost. It is the core metric in the path optimization process, the node weight value. This indicates the quantity of goods. (In the formula) When the sign is negative, it means that a node with a negative weight (out of stock) can have its priority ranking value reduced, making it rank higher in the open set and thus be explored first.

[0035] Optionally, after calculating the optimal path between all node pairs, the total path cost from source node N1 to target node N3 is 8, and the path is N1->N3; the total path cost from source node N1 to target node N4 is 5, and the path is N1->N4; the total path cost from source node N2 to target node N3 is 4, and the path is N2->N3; and the total path cost from source node N2 to target node N4 is 6, and the path is N2->N4. A transportation task allocation matrix is ​​constructed based on the total amount of goods to be scheduled at each source node and the total path cost from each source node to each target node. The rows of the transportation task allocation matrix correspond to source nodes N1 and N2, and the columns correspond to target nodes N3 and N4. The matrix elements represent the cost of transporting a unit of goods from a source node to a target node. The transportation task allocation matrix is ​​optimized using minimum-cost flow processing. Under the constraints of the total supply from source nodes and the total demand from target nodes, the scheduling scheme with the minimum total transportation cost is sought. This generates the quantity of goods to be scheduled from each source node to each target node. For example, source node N1 schedules 0 tons of goods to target node N3 and 40 tons to target node N4; source node N2 schedules 30 tons of goods to target node N3 and 0 tons to target node N4. The above-mentioned quantity of goods and the corresponding optimal transportation path are taken as the optimal logistics scheduling scheme. Optional, the weight baseline value... It can be set to the sum of the absolute values ​​of the weights of all nodes, or it can be set to a fixed constant value, such as 100 tons, as an adjustment coefficient. The value can be adjusted based on the algorithm's performance on the test set. It's understandable that minimum cost flow optimization can be solved using the network simplex method or the Hungarian algorithm, with the goal of minimizing the total transportation cost while satisfying supply and demand constraints. It's also understandable that recording the node predecessor pointers allows backtracking from the target node to the source node at the end of the search, thus fully reconstructing the node sequence of the optimal transportation path.

[0036] In one embodiment of the present invention, the node weight dynamic update mechanism on which the improved Dijkstra algorithm is based specifically involves the following: during path search, whenever a node is moved from the open set to the closed set, the priority ranking value of the node's neighboring nodes is dynamically updated based on the weighted sum of the node's current accumulated path cost and its node weight value in the supply-demand gap distribution map. Specifically, when a node's weight value is positive, the priority ranking value of neighboring nodes is decreased; when a node's weight value is negative, the priority ranking value of neighboring nodes is increased. This mechanism allows the algorithm to consider not only path distance or travel cost during optimization but also to proactively guide paths towards target nodes with negative weights (i.e., shortages) or to expand outwards more quickly from source nodes with positive weights (i.e., surpluses).

[0037] The improved Dijkstra algorithm is based on a time-varying edge cost decay mechanism. Specifically, during path search, whenever an edge is selected as part of the optimal path, its current edge cost is multiplied by a preset decay coefficient to generate an updated edge cost. This updated cost is then used to re-evaluate the edge in subsequent path searches. This mechanism simulates real-world transportation, where frequent use or planning of a route can reduce its implicit "familiarity" or "pre-planning" costs. This encourages the formation of relatively concentrated transportation corridors in subsequent global optimization, helping to reduce the dispersion and management complexity of vehicle scheduling.

[0038] In practical implementation, consider a simplified road network example containing four nodes (node ​​A, node B, node C, and node D). Node A is the source node, and node D is the target node. The supply and demand gap distribution map assigns node A a node weight of +10 (positive weight, indicating a surplus of goods), node C a node weight of -5 (negative weight, indicating a shortage of goods), and nodes B and D a node weight of zero. The initial edge costs between nodes are as follows: edge (AB) has an edge cost of 4, edge (AC) has an edge cost of 3, edge (BC) has an edge cost of 2, edge (BD) has an edge cost of 5, and edge (CD) has an edge cost of 1. The improved Dijkstra algorithm searches for the optimal path to the target node D starting from the source node A. The search process maintains an open set and a closed set, and uses a dynamic node weight update mechanism and a time-varying edge cost decay mechanism.

[0039] In practice, the dynamic node weight update mechanism operates as follows: During the initialization phase of path search, the cumulative path cost of source node A is set to 0 and placed in the open set. The cumulative path costs of nodes B, C, and D are initialized to infinity. When node A moves from the open set to the closed set as the current expanding node, the priority ranking values ​​of node A's neighboring nodes (nodes B and C) are dynamically updated based on the weighted sum of node A's current cumulative path cost of 0 and node weight value + 10 in the supply-demand gap distribution map. When a node weight value is positive, the priority ranking value of neighboring nodes is decreased; when a node weight value is negative, the priority ranking value of neighboring nodes is increased. The update formula for the node priority ranking value is: in: Represents a node priority sorting value, Represents a node Current cumulative path cost, Represents a node The node weight values ​​in the supply-demand gap distribution map It is a preset weighted normalization benchmark. It is a positive adjustment parameter. Cumulative path cost. It is a core metric in the optimization process, and has a time or common cost unit. Node weight value. Indicates the surplus or shortage of goods, weighted normalized benchmark. Typically, the absolute or maximum value of the node weights in the system is taken. For node C, its initial cumulative path cost becomes 3 after being updated through node A, assuming a weight normalization baseline. Set the value to 20 and adjust the parameter. The node weight of node C is 1 (with the same dimensions as the cumulative path cost). If it is -5, then the priority ranking value is... The dynamic node weight update mechanism uses this formula to ensure that the calculated priority ranking value of node C with negative weight (out of stock) is higher than its cumulative path cost, thus ranking it higher in the open set and allowing it to be explored first.

[0040] In some embodiments, the edge cost time-varying decay mechanism operates as follows: During the path search process, whenever an edge is selected as part of the current optimal path, the current edge cost of that edge is multiplied by a preset decay coefficient to generate the updated edge cost. The formula for the updated edge cost is: in: This represents the cost of the current edge before it was selected. This indicates the preset attenuation coefficient (0 < 0). <1), This represents the cost of updating the next edge. Assume a preset decay coefficient. The cost is 0.9. During a search, if edge (AC) is selected first, its cost is updated from 3 to 3 * 0.9 = 2.7. This updated cost of 2.7 will be used to re-evaluate the edge in subsequent path searches. If edge (AC) is traversed again in subsequent searches, its cost will be calculated starting from 2.7. Refer to Table 1, which shows the process of each edge being selected and its cost decaying during a complete path search.

[0041] Table 1: The process of each edge being selected and its cost decay. Optional, adjust parameters The value can be set to 2 to enhance the influence of node weight on priority ranking. In the calculation, the priority ranking value of node C is... Optional, preset attenuation coefficient. It can be set to 0.8 to reflect the more significant decrease in toll costs after the route is planned and used. This is understandable, as it represents the weighted normalization baseline. The choice of [value] affects the relative influence of node weight values ​​in priority calculation, and its value can be set according to the scale of cargo volume in the specific scenario. It can be understood that the time-varying edge cost decay mechanism simulates the path dependency effect; once an edge is included in the planned path, its subsequent evaluation cost will decrease, which helps to form a relatively stable transportation corridor in multiple scheduling processes.

[0042] In one embodiment of the present invention, when the instruction issuing module generates instructions based on the optimal logistics scheduling scheme, it iterates through the scheduling tasks between each source node and each target node in the optimal logistics scheduling scheme. Each scheduling task includes the quantity of goods to be scheduled, the type of goods, the originating location of the goods, the receiving location of the goods, and the optimal transportation route. For each scheduling task, a set of candidate vehicles with the qualifications to transport that type of goods is selected from the transportation vehicle database based on the type of goods. The required number of vehicles is determined from the candidate vehicle set based on the quantity of goods to be scheduled, and the currently idle vehicle closest to the originating location of the goods is selected as the scheduling vehicle. A vehicle scheduling instruction is generated for each scheduling vehicle, which includes the license plate identifier of the scheduling vehicle, the loading location of the goods, the loading time window of the goods, the target unloading location, and the target unloading time window. Simultaneously, based on the node sequence in the optimal transportation route, a route planning instruction is generated for each scheduling vehicle, which includes the sequence of intersection identifiers and road segment identifiers that the scheduling vehicle needs to pass through in sequence. All vehicle scheduling instructions of all scheduling vehicles are summarized into a vehicle scheduling instruction set, and all route planning instructions of all scheduling vehicles are summarized into a route planning instruction set, thus completing the instruction encapsulation.

[0043] In a specific implementation, consider a specific scheduling task derived from an optimal logistics scheduling plan. The optimal logistics scheduling plan includes a scheduling task from the source node factory X to the target node warehouse Y. This scheduling task includes the quantity of goods to be scheduled, which is 20 tons, the type of goods is chemical product P3, the goods departure location is the loading area of factory X, the goods receiving location is the unloading area of warehouse Y, and the optimal transportation route is factory X -> intersection M1 -> section R1 -> intersection M2 -> warehouse Y. The instruction issuing module traverses the scheduling tasks between each source node and each target node in the optimal logistics scheduling plan and processes the above scheduling tasks. For this scheduling task, a set of candidate vehicles with the qualification to transport chemical product P3 is selected from the transportation vehicle database according to the type of goods, chemical product P3. The transportation vehicle database records vehicle license plates, vehicle locations, vehicle statuses, and the categories of chemical products that can be carried. The selected set of candidate vehicles includes three transportation vehicles with license plates "Beijing A12345", "Beijing B67890", and "Beijing C24680". All three transportation vehicles are marked as capable of safely transporting chemical product P3. The required number of vehicles is determined from the set of candidate vehicles according to the quantity of goods to be scheduled, which is 20 tons. The maximum load capacity of each transportation vehicle is 10 tons. Therefore, the number of vehicles required to schedule 20 tons of goods is 2. The vehicles that are currently idle and closest to the loading area of the goods departure location, factory X, are selected as the scheduling vehicles. The real-time locations and statuses of the three vehicles are queried. Vehicle "Beijing A12345" is 5 kilometers away from factory X and is in an idle state. Vehicle "Beijing B67890" is 8 kilometers away from factory X and is in an idle state. Vehicle "Beijing C24680" is 12 kilometers away from factory X and is in a driving state. Therefore, vehicle "Beijing A12345" and vehicle "Beijing B67890" are selected as the scheduling vehicles.

[0044] In some embodiments, a vehicle scheduling instruction is generated for each scheduling vehicle. The vehicle scheduling instruction generated for vehicle "Beijing A12345" includes the license plate identification "Beijing A12345", the goods loading location "Factory X - Loading Area 3rd Position", the goods loading time window "08:00 - 09:00", the target unloading location "Warehouse Y - Unloading Platform A", and the target unloading time window "10:30 - 11:30". The vehicle scheduling instruction generated for vehicle "Beijing B67890" includes the license plate identification "Beijing B67890", the goods loading location "Factory X - Loading Area 5th Position", the goods loading time window "08:15 - 09:15", the target unloading location "Warehouse Y - Unloading Platform B", and the target unloading time window "10:45 - 11:45". The determination of the vehicle loading time window is based on a calculation model that is used to estimate the time required for the vehicle to travel from the current location to the loading location. The start time of the time window The calculation formula is , where Indicates the start time of the time window, Indicates the current time when the system generates an instruction, Indicates the distance from the current location of the dispatched vehicle to the goods loading location, Indicates the estimated average driving speed of the dispatched vehicle. For vehicle "Beijing A12345", the current system time is 07:30, and the distance is 5 kilometers, and the estimated average speed is 30 kilometers per hour. The start time of the calculated time window is obtained, so the loading time window is set to 08:00 - 09:00.

[0045] Optionally, according to the node sequence in the optimal transportation path, a path planning instruction is generated for each dispatched vehicle. The optimal transportation path factory X -> Intersection M1 -> Road Section R1 -> Intersection M2 -> Warehouse Y contains a series of nodes. The path planning instruction generated for vehicle "Beijing A12345" contains the sequence of intersection identifiers "Intersection M1, Intersection M2" that the dispatched vehicle needs to pass through in turn and the sequence of road section identifiers "Road Section R1". The path planning instruction generated for vehicle "Beijing B67890" contains the same sequence of intersection identifiers and road section identifiers. Refer to Table 2 for the details of the instructions generated for the two dispatched vehicles.

[0046] Table 2: Instruction Table for Two Dispatched Vehicles In some embodiments, the setting of the target unloading time window needs to consider the transportation time from the loading location to the receiving location, and the transportation time is calculated based on the length of the optimal transportation path and the estimated average road section passing speed. Optionally, the vehicle dispatching instructions of all dispatched vehicles are aggregated into a vehicle dispatching instruction set, which is a list containing multiple vehicle dispatching instruction data structures. The path planning instructions of all dispatched vehicles are aggregated into a path planning instruction set, which is a list containing multiple path planning instruction data structures. It can be understood that vehicle dispatching instructions and path planning instructions are two types of instructions that are logically related but independent in content. Vehicle dispatching instructions focus on task and resource allocation, while path planning instructions focus on specific driving guidance. It can be understood that when screening vehicles from the transportation vehicle database, in addition to considering qualifications and distances, the remaining load capacity of the vehicle and the estimated time to complete the current task of the vehicle can also be combined for comprehensive sorting and selection.

[0047] In one embodiment of the present invention, after generating the optimal logistics scheduling scheme, the system acquires historical logistics scheduling schemes executed within the historical scheduling period and corresponding historical road real-time traffic data to construct a historical scheduling experience base. The system compares the current optimal logistics scheduling scheme generated in the current scheduling period with the historical logistics scheduling schemes in the historical scheduling experience base, filtering out historical logistics scheduling schemes with similarity exceeding a preset similarity threshold. The system extracts the historical road real-time traffic data corresponding to the filtered historical logistics scheduling schemes, and calculates the historical average congestion duration for each road segment within a similar time window. Based on the historical average congestion duration, the optimal transportation route in the current optimal logistics scheduling scheme is marked with congestion risk, identifying easily congested road segments on the optimal transportation route whose congestion risk exceeds a risk threshold. The original transportation routes passing through easily congested road segments are deleted from the current optimal logistics scheduling scheme, and an improved Dijkstra algorithm is invoked to generate alternative transportation routes bypassing the easily congested road segments. These alternative transportation routes are then incorporated into the current optimal logistics scheduling scheme, thereby enhancing the robustness of the scheduling scheme to historical congestion patterns.

[0048] After generating the optimal logistics scheduling plan, the system also extracts the actual order completion data and actual vehicle travel data at the end of each scheduling cycle from the supply chain related data set. It compares the actual order completion data with the expected order completion data in the optimal logistics scheduling plan to calculate the order completion deviation rate for each scheduling task, and compares the actual vehicle travel data with the planned vehicle travel data in the optimal logistics scheduling plan to calculate the path execution deviation rate for each scheduling task. When the order completion deviation rate or the path execution deviation rate exceeds a preset deviation threshold, the corresponding scheduling task is marked as an abnormal scheduling task. The system extracts the historical real-time road traffic data within the scheduling cycle corresponding to the abnormal scheduling task, identifies the abnormal road segments and abnormal time periods that caused the deviation, and increases the unit distance travel cost of the abnormal road segments within the abnormal time periods by a penalty step, generating a corrected road traffic cost matrix for path optimization processing in the next scheduling cycle, achieving adaptive correction of model parameters based on feedback data.

[0049] In practice, the system acquires historical logistics scheduling schemes executed within historical scheduling cycles and corresponding real-time road traffic data to construct a historical scheduling experience database. This database records schemes and data for scheduling cycles T1, T2, and T3. The optimal logistics scheduling scheme generated in the current scheduling cycle T4 includes a transportation path from factory M to warehouse N, passing through nodes M->P->Q->N in sequence. The system compares the similarity of the optimal logistics scheduling scheme generated in the current scheduling cycle T4 with the historical logistics scheduling schemes in the database, filtering out those with a similarity exceeding a preset similarity threshold. The similarity comparison is achieved by calculating the path overlap ratio, defined as the percentage of the length of the same node sequence in two paths relative to the total path length. The preset similarity threshold is set to 0.7. After comparison, the optimal logistics scheduling scheme in historical scheduling cycle T2, containing the path M->P->Q->N, is completely identical to the current scheme, with a similarity of 1.0, exceeding the preset similarity threshold of 0.7. Therefore, the scheme from historical scheduling cycle T2 is selected. Historical real-time traffic data corresponding to the selected historical logistics scheduling schemes were extracted. The real-time traffic data for historical scheduling period T2 recorded the traffic status of each road segment on the path M->P->Q->N within a specific time window. The historical average congestion duration for each road segment within similar time windows was calculated. For road segment PQ, the historical average congestion duration was 20 minutes between 9:00 AM and 10:00 AM. Based on the historical average congestion duration, congestion risk was labeled for the optimal transportation route in the current optimal logistics scheduling scheme. A congestion risk threshold of 15 minutes was set, and easily congested road segments exceeding the risk threshold were marked on the optimal transportation route. Since the historical average congestion duration of road segment PQ was 20 minutes, exceeding the 15-minute risk threshold, road segment PQ was marked as an easily congested road segment. Remove the original transport route M->P->Q->N that passes through the congested road segment PQ from the current optimal logistics scheduling scheme. Re-invoke the improved Dijkstra algorithm to generate an alternative transport route that bypasses the congested road segment PQ. The improved Dijkstra algorithm is recalculated based on the updated road network to generate the alternative transport route M->P->R->S->N. The alternative transport route M->P->R->S->N is then incorporated into the current optimal logistics scheduling scheme.

[0050] In some embodiments, the formula for calculating the similarity of optimal logistics scheduling schemes is: in: This represents the overall similarity between two scheduling schemes. This indicates the number of task pairs scheduled in the plan. It is the first The weighting coefficients of each task pair Indicates the first in the current scheme The transportation route for each task. Indicating the first in the historical scheme The transportation path for each task, function Calculate the node sequence matching degree between the two paths. Weight coefficients. It can be set according to the proportion of the task transportation volume.

[0051] In practice, actual order completion data and actual vehicle travel data are extracted from the supply chain-related data set after each scheduling cycle. For scheduling cycle T3, the actual order completion data shows that the expected quantity of goods to be completed for the scheduling task from factory M to warehouse N is 100 tons, and the actual quantity is 80 tons. The actual vehicle travel data shows that the planned path is M->X->Y->N, and the actual travel path is M->X->Z->N. The actual order completion data is compared with the expected order completion data in the optimal logistics scheduling plan to calculate the order completion deviation rate for each scheduling task. The order completion deviation rate is calculated as (expected completion amount - actual completion amount) / expected completion amount. For the scheduling task from factory M to warehouse N, the order completion deviation rate is (100-80) / 100 = 0.2. The actual vehicle travel data is compared with the planned vehicle travel data in the optimal logistics scheduling scheme to calculate the path execution deviation rate for each scheduling task. The path execution deviation rate is obtained by calculating the node difference between the actual path and the planned path. For this task, the actual path M->X->Z->N deviates from the planned path M->X->Y->N at the third node, with a node difference of 1 / 3. When the order completion deviation rate or the path execution deviation rate exceeds the preset deviation threshold, the corresponding scheduling task is marked as an abnormal scheduling task. The preset deviation threshold is set to 0.15. The order completion deviation rate of 0.2 exceeds 0.15, therefore the task is marked as an abnormal scheduling task.

[0052] Optionally, historical real-time road traffic data within the scheduling period corresponding to the abnormal scheduling task can be extracted to identify the abnormal road segments and abnormal time periods that caused the deviation. Analyzing the historical real-time road traffic data of scheduling period T3, it was found that a traffic accident occurred on road segment Y between 2 pm and 3 pm, resulting in complete closure. This time period and road segment were identified as abnormal time periods and abnormal road segments.

[0053] Optionally, a penalty step can be added to the unit distance travel cost of the abnormal road segment during the abnormal time period to generate a corrected road travel cost matrix for path optimization processing in the next scheduling cycle. The original unit distance travel cost of road segment Y from 2 pm to 3 pm is 0.5 hours / km, the penalty step is set to 0.5, and the corrected unit distance travel cost is updated to 0.5 + 0.5 = 1.0 hours / km. The corrected cost will be updated in the road travel cost matrix.

[0054] It is understandable that the historical average congestion duration statistics can be calculated smoothly based on data from multiple similar historical dates to reduce the impact of random fluctuations.

[0055] It is understandable that the identification and cost correction of deviations is a continuous learning process. Through iterative correction, the road traffic cost matrix can better reflect the real and dynamic traffic conditions.

[0056] The foregoing has provided a detailed description of one embodiment of the present invention, but this description is merely a preferred embodiment and should not be construed as limiting the scope of the invention. All equivalent variations and modifications made within the scope of the claims of this invention should still fall within the patent coverage of this invention.

Claims

1. A smart logistics scheduling and optimization system for the chemical supply chain based on big data analysis, characterized in that, The system includes: The data acquisition module acquires a set of supply chain-related data from multiple chemical raw material suppliers, multiple chemical plants, and multiple chemical product customers in the chemical supply chain over multiple scheduling cycles. The set of supply chain-related data includes order demand data, inventory status data, transport vehicle location data, and real-time road traffic data for each scheduling cycle. The supply and demand analysis module performs supply and demand matching analysis on the order demand data and the inventory status data to generate a supply and demand gap distribution map for each scheduling cycle. The road condition assessment module performs path capacity assessment processing on the vehicle location data and the real-time road traffic data to generate a road traffic cost matrix for each scheduling cycle. The scheduling optimization module calls the improved Dijkstra algorithm to perform joint path optimization processing on the supply and demand gap distribution map and the road traffic cost matrix, and generates the optimal logistics scheduling scheme corresponding to the scheduling cycle. The improved Dijkstra algorithm is based on the dynamic update mechanism of node weights and the time-varying decay mechanism of edge costs. The instruction issuing module generates a set of vehicle dispatch instructions and a set of route planning instructions based on the optimal logistics scheduling scheme, and sends the set of vehicle dispatch instructions and the set of route planning instructions to the chemical supply chain scheduling center.

2. The intelligent logistics scheduling and optimization system for the chemical supply chain based on big data analysis according to claim 1, characterized in that, Perform supply and demand matching analysis on the order demand data and the inventory status data to generate a supply and demand gap distribution map for each scheduling cycle, including: Extract the product type demand and demand time window of each chemical product customer within the target scheduling cycle from the order demand data, and extract the available inventory of each chemical plant of various chemical products at the beginning of the target scheduling cycle from the inventory status data. For each chemical product, calculate the total demand for that chemical product from all customers, and the total available inventory of that chemical product from all chemical plants; The total demand is calculated by subtracting the total available inventory to obtain the global supply and demand gap for this chemical product. When the global supply-demand gap is positive, it is determined that replenishment needs to be made from the chemical raw material supplier to the chemical plant. When the global supply and demand imbalance is negative, it is determined that the chemical plant needs to ship goods to the chemical product customer. For each chemical plant, the available inventory of the plant is compared with the demand of chemical product customers within the plant's coverage area, and the types of products with insufficient inventory and the types of products with excess inventory are marked. The supply and demand gap distribution map is constructed by using the geographical coordinates of the chemical plant as nodes, the shortage quantity corresponding to the product type with insufficient inventory as the negative weight of the node, and the surplus quantity corresponding to the product type with excess inventory as the positive weight of the node.

3. The intelligent logistics scheduling and optimization system for the chemical supply chain based on big data analysis according to claim 1, characterized in that, The route capacity assessment process is performed on the vehicle location data and the real-time road traffic data to generate a road traffic cost matrix for each scheduling cycle, including: Extract the real-time latitude and longitude coordinates and driving speed values ​​of all transport vehicles in the current scheduling cycle from the transport vehicle positioning data, and extract the length value, number of lanes value, and current traffic flow value of each road segment in the road network from the real-time road traffic data. Based on the real-time latitude and longitude coordinates, each transport vehicle is matched to the nearest road segment to obtain the vehicle density value on each road segment; For each road segment, the vehicle density value, the traffic flow value, and the number of lanes are input into a preset road segment travel time estimation function to generate the estimated travel time value for that road segment. The estimated travel time is divided by the length of the road segment to obtain the unit distance travel cost of the road segment. The starting node identifiers, ending node identifiers, and corresponding unit distance travel costs of all road segments are organized in the form of node pairs to generate the road travel cost matrix.

4. The intelligent logistics scheduling and optimization system for the chemical supply chain based on big data analysis according to claim 1, characterized in that, The improved Dijkstra algorithm is invoked to perform joint path optimization on the supply-demand gap distribution map and the road traffic cost matrix to generate the optimal logistics scheduling scheme corresponding to the scheduling cycle, including: Extract all positive weight nodes and all negative weight nodes from the supply and demand gap distribution map, and use each positive weight node as the source node for goods to be sent out and each negative weight node as the target node for goods to be received. For each source node, the surplus value corresponding to that source node is taken as the total amount of goods to be scheduled. Using the road traffic cost matrix as the initial edge cost between nodes, the improved Dijkstra algorithm is called to calculate the optimal transportation path from each source node to each target node and the total path cost corresponding to the optimal transportation path. Based on the total amount of goods to be scheduled at each source node and the total path cost from each source node to each target node, a transportation task allocation matrix is ​​constructed, wherein the rows of the transportation task allocation matrix correspond to the source nodes and the columns correspond to the target nodes. The minimum cost flow optimization process is performed on the transportation task allocation matrix to generate the number of goods scheduled from each source node to each target node. The number of goods scheduled from each source node to each target node and the corresponding optimal transportation path are taken as the optimal logistics scheduling scheme.

5. The intelligent logistics scheduling and optimization system for the chemical supply chain based on big data analysis according to claim 4, characterized in that, The improved Dijkstra algorithm is based on a dynamic node weight update mechanism and a time-varying edge cost decay mechanism, including: The node weight dynamic update mechanism is as follows: During the path search process, whenever a node is moved from the open set to the closed set, the priority ranking value of the neighboring nodes is dynamically updated according to the weighted sum of the node's current accumulated path cost and the node weight value of the node in the supply and demand gap distribution map. When the node weight value is positive, the priority ranking value of the neighboring nodes is reduced, and when the node weight value is negative, the priority ranking value of the neighboring nodes is increased. The time-varying edge cost decay mechanism is as follows: during the path search process, whenever an edge is selected as part of the optimal path, the current edge cost of that edge is multiplied by a preset decay coefficient to generate the updated edge cost, and the updated edge cost is used for the re-evaluation of that edge in subsequent path searches.

6. The intelligent logistics scheduling and optimization system for the chemical supply chain based on big data analysis according to claim 4, characterized in that, The improved Dijkstra algorithm is invoked to calculate the optimal transportation path from each source node to each target node and the total path cost corresponding to that optimal transportation path, including: For each source node and target node pair, initialize the cumulative path cost of the current source node to zero, add the current source node as the current extended node to the open set, and initialize the cumulative path cost of all nodes except the current source node to infinity. Select the node with the smallest priority ranking value from the open set as the new current expansion node, and move the new current expansion node from the open set to the closed set; Obtain all neighbor nodes of the new current extended node. For each neighbor node, skip it if it is in the closed set; otherwise, calculate the candidate path cost from the current source node through the new current extended node to the neighbor node. The candidate path cost is equal to the cumulative path cost of the new current extended node plus the current edge cost of the edge from the new current extended node to the neighbor node. When the candidate path cost is less than the current cumulative path cost of the neighboring node, the cumulative path cost of the neighboring node is updated to the candidate path cost, the predecessor node of the neighboring node is pointed to the new current extension node, and then the priority ranking value of the neighboring node is recalculated based on the weighted sum of the updated cumulative path cost of the neighboring node and the node weight value of the neighboring node in the supply and demand gap distribution map. Repeat the operation of selecting the node with the smallest priority sorting value from the open set as the new current expansion node until the new current expansion node is equal to the current target node, then stop the search; Starting from the current target node, backtracking backward along the predecessor node to the current source node, the sequence of nodes traversed during the backtracking is taken as the optimal transportation path, and the cumulative path cost of the current target node is taken as the total path cost.

7. The intelligent logistics scheduling and optimization system for the chemical supply chain based on big data analysis according to claim 1, characterized in that, Based on the optimal logistics scheduling scheme, a set of vehicle scheduling instructions and a set of route planning instructions are generated, including: Iterate through the scheduling tasks between each source node and each target node in the optimal logistics scheduling scheme. Each scheduling task includes the quantity of goods to be scheduled, the type of goods, the origin of the goods, the destination of the goods, and the optimal transportation route. For each scheduling task, a set of candidate vehicles qualified to transport that type of cargo is selected from the transport vehicle database based on the cargo type. The required number of vehicles is determined from the candidate vehicle set based on the quantity of goods to be dispatched, and the currently idle vehicle that is closest to the departure point of the goods is selected as the dispatch vehicle. A vehicle dispatch instruction is generated for each dispatched vehicle. The vehicle dispatch instruction includes the vehicle's license plate number, cargo loading location, cargo loading time window, target unloading location, and target unloading time window. Based on the node sequence in the optimal transportation route, a route planning instruction is generated for each dispatched vehicle. The route planning instruction includes the sequence of intersection markers and road segment markers that the dispatched vehicle needs to pass through in sequence. All vehicle dispatch instructions are aggregated into the vehicle dispatch instruction set, and all route planning instructions for all dispatched vehicles are aggregated into the route planning instruction set.

8. The intelligent logistics scheduling and optimization system for the chemical supply chain based on big data analysis according to claim 1, characterized in that, After generating the optimal logistics scheduling plan, the process also includes executing: Obtain historical logistics scheduling plans executed within the historical scheduling period and corresponding real-time road traffic data to build a historical scheduling experience database. The current optimal logistics scheduling scheme generated in the current scheduling cycle is compared with the historical logistics scheduling schemes in the historical scheduling experience base to filter out historical logistics scheduling schemes with a similarity exceeding a preset similarity threshold. Extract real-time traffic data of historical roads corresponding to the selected historical logistics scheduling schemes, and calculate the historical average congestion duration of each road segment under similar time windows. Based on the historical average congestion duration, the optimal transportation route in the current optimal logistics scheduling scheme is marked with congestion risk, and congestion-prone sections on the optimal transportation route whose congestion risk exceeds the risk threshold are marked. Remove the original transport route that passes through the congested section from the current optimal logistics scheduling scheme, re-invoke the improved Dijkstra algorithm to generate an alternative transport route that bypasses the congested section, and incorporate the alternative transport route into the current optimal logistics scheduling scheme.

9. The intelligent logistics scheduling and optimization system for the chemical supply chain based on big data analysis according to claim 1, characterized in that, After generating the optimal logistics scheduling plan, the process also includes executing: Extract the actual order completion data and actual vehicle travel data after the end of each scheduling cycle from the aforementioned supply chain-related data set; The actual order completion data is compared with the expected order completion data in the optimal logistics scheduling scheme, and the order completion deviation rate of each scheduling task is calculated. The actual vehicle travel data is compared with the planned vehicle travel data in the optimal logistics scheduling scheme to calculate the path execution deviation rate for each scheduling task. When the order completion deviation rate exceeds the preset deviation threshold or the path execution deviation rate exceeds the preset deviation threshold, the corresponding scheduling task is marked as an abnormal scheduling task. Extract historical real-time road traffic data within the scheduling period corresponding to the abnormal scheduling task, identify the abnormal road segments and abnormal time periods that cause the deviation, increase the unit distance traffic cost of the abnormal road segments within the abnormal time periods by a penalty step, and generate a corrected road traffic cost matrix for path optimization processing in the next scheduling period.

10. A device, characterized in that, It includes a processor and a memory, wherein the memory stores a computer program, and when the computer program is executed by the processor, it implements a chemical supply chain intelligent logistics scheduling and optimization system based on big data analysis as described in any one of claims 1 to 9.

Images