Intelligent scheduling and operation analysis system and method for bulk grain terminal based on knowledge graph

Abstract

The present application relates to the technical field of port intelligent scheduling and operation management, in particular to a bulk grain terminal intelligent scheduling and operation analysis system and method based on a knowledge graph. The method comprises the following steps: bulk grain terminal field knowledge graph construction oriented by job cost; graph correlation mapping of cost drivers and scheduling resources; multi-dimensional intelligent scheduling scheme generation driven by job cost drivers; multi-dimensional operation cost analysis based on graph tracing. The present application realizes the deep integration of scheduling and cost control, reduces the complexity of path reasoning and shortens the response time through bulk grain terminal field knowledge graph construction oriented by job cost, graph correlation mapping of cost drivers and scheduling resources, local path incremental reasoning driven by job cost drivers and multi-dimensional operation cost analysis based on graph tracing, achieves the collaborative optimization of efficiency and cost, and at the same time realizes the full-link cost tracing, improves the collection accuracy and analysis dimension.

Description

Technical Field

[0001] This invention relates to the field of port intelligent scheduling and operation management technology, and more specifically, to a knowledge graph-based intelligent scheduling and operation analysis system and method for bulk grain terminals. Background Technology

[0002] As a core hub in the grain supply chain, bulk grain terminals are undergoing intelligent and digital transformation, which has become an inevitable trend in the industry as the scale of global grain circulation continues to expand. Intelligent scheduling and operational analysis technology for bulk grain terminals is a core technological means to improve terminal operation efficiency, optimize resource allocation, and reduce operating costs. This technology typically integrates terminal operation data, resource data, and cost data, using algorithmic models to achieve automated scheduling of operational processes and quantitative analysis of operational status. Its core key lies in the coordinated management and control of operational processes, resource elements, and cost drivers. The technical challenge lies in multi-objective optimization under dynamic operating conditions and end-to-end cost traceability, which has significant industry value in supporting the refined and efficient operation of bulk grain terminals.

[0003] The industry generally strives to achieve multi-dimensional intelligent scheduling and full-chain operational cost analysis for bulk grain terminals driven by operational costs. Ideally, this technology should be able to dynamically adjust scheduling schemes based on real-time operational cost drivers, while simultaneously achieving accurate collection and multi-dimensional traceability analysis of operational costs. Currently, the industry has not yet formed a technical system that deeply integrates operational scheduling and cost control, making it difficult to simultaneously meet multi-dimensional optimization goals such as efficiency and cost under dynamic operating conditions. This is a common core contradiction restricting the further improvement of the intelligence level of bulk grain terminals.

[0004] Currently, there are two main technical problems in the field of bulk grain terminal scheduling and operation analysis: First, scheduling decision-making and cost control are disconnected. Existing intelligent scheduling technologies mostly focus on operational efficiency and resource utilization as the sole optimization objectives, failing to establish a quantitative mapping relationship between operational cost drivers and scheduling resources. This makes it impossible to dynamically adjust operational paths based on real-time cost driver changes, resulting in high computational complexity for global path reasoning and hindering the synergistic optimization of efficiency and cost under dynamic operating conditions. Second, operational cost analysis lacks end-to-end data support. Existing cost analysis technologies are mostly based on traditional relational databases for data storage and processing. Data relationships are fixed and lack cross-stage traceability, making it impossible to accurately collect operational costs and perform multi-dimensional correlation analysis. Operational analysis results lag behind the operational process, making it difficult to support refined operational decisions. These technical problems lead to insufficient cost controllability in bulk grain terminal scheduling schemes and limited depth of operational analysis, restricting the improvement of refined operation levels at bulk grain terminals. Therefore, we propose a knowledge graph-based intelligent scheduling and operation analysis system and method for bulk grain terminals. Summary of the Invention

[0005] The purpose of this invention is to provide a knowledge graph-based intelligent scheduling and operation analysis system and method for bulk grain terminals, in order to solve the problems mentioned in the background art, such as the disconnect between scheduling decision-making and cost control, the difficulty in achieving synergistic optimization of efficiency and cost under dynamic operating conditions, and the lack of full-link data support for operation cost analysis, which makes it impossible to achieve accurate collection of operating costs and multi-dimensional correlation analysis.

[0006] To address the aforementioned technical problems, one objective of this invention is to provide a knowledge graph-based intelligent scheduling and operation analysis method for bulk grain terminals, comprising the following steps: S1. Construction of a knowledge graph for bulk grain terminals based on activity-based costing: Combining the cost aggregation logic of activity-based costing, a knowledge graph for bulk grain terminals is constructed. The knowledge graph for bulk grain terminals includes the operational links, resource elements, cost drivers, and the relationships between these elements. S2. Graph Relationship Mapping between Cost Drivers and Scheduling Resources: Based on the entity association capability of the knowledge graph in the bulk grain terminal field, and combined with the operational cost driver analysis method, establish the mapping relationship between the scheduling resources of bulk grain terminals and operational cost elements. S3. Generation of multi-dimensional intelligent scheduling scheme driven by operation cost drivers: Based on the mapping relationship built in S2 and real-time operation cost driver data, through the dynamic path reasoning method of knowledge graph driven by operation cost drivers, only the local operation paths affected by cost driver changes are updated to generate an intelligent scheduling scheme for bulk grain terminals that meets multi-dimensional optimization objectives. S4. Multidimensional analysis of operating costs based on knowledge graph traceability: Relying on the full-link data traceability capability of knowledge graph in the bulk grain terminal field, combined with the results of operation cost collection, the operation status of bulk grain terminals is analyzed in multiple dimensions and the corresponding operation analysis results are output.

[0007] As a further improvement to this technical solution, in step S1, the process of constructing a knowledge graph for the bulk grain terminal sector includes the following steps: S1.1 Based on the cost aggregation logic of activity-based costing, define a dedicated ontology architecture for the knowledge graph of bulk grain terminals. The ontology architecture is adapted to the business attributes and association rules of bulk grain terminal operation links, resource elements, and cost drivers. S1.2 Based on the defined ontology architecture, accurately extract the entities of bulk grain terminal operation links, resource elements, and cost drivers from the bulk grain terminal business data source. S1.3 Based on the cost aggregation logic of activity-based costing, explore the business relationships and cost coupling relationships between entities, and construct the relationship mapping links between entities; S1.4. The extracted entities and the constructed relationships are integrated, standardized, and stored in a solidified manner to generate a complete knowledge graph of the bulk grain terminal domain.

[0008] As a further improvement to this technical solution, in step S2, the process of establishing the mapping relationship between bulk grain terminal scheduling resources and operating cost elements includes the following steps: S2.1 Based on the operational cost driver analysis method, sort out the attribute characteristics of the scheduling resources and operational cost elements of bulk grain terminals, as well as the corresponding rules between the two types of elements; S2.2 Utilize the entity association capability of the bulk grain terminal domain knowledge graph to determine the corresponding node positions of the bulk grain terminal scheduling resource entity and the operation cost element entity within the graph; S2.3. According to the corresponding rules obtained from the analysis, establish the connection links between the corresponding nodes in the map to complete the initial mapping and binding of bulk grain terminal scheduling resources and operating cost elements; S2.4. Perform logical consistency verification and correction on the established association links to solidify a stable mapping relationship.

[0009] As a further improvement to this technical solution, in step S3, the process of generating an intelligent scheduling scheme for bulk grain terminals includes the following steps: S3.1. Based on historical operational data of bulk grain terminals, iteratively update the benchmark parameters of operational cost drivers and the benchmark cost of standard operating routes; S3.2. Combine real-time operation cost driver data to quantitatively calculate the impact coefficient of each operation cost driver on the total operation cost. S3.3 Based on the deviation between real-time operating conditions and baseline operating conditions, adaptively adjust the weight ratio of each operation cost driver in the dynamic path reasoning of the knowledge graph. S3.4. Based on the preset change threshold, determine the degree of fluctuation of the operating cost driver, and only perform incremental updates on the affected local operating paths to generate a smart scheduling scheme for bulk grain terminals that meets multi-dimensional optimization objectives.

[0010] As a further improvement to this technical solution, the process of iteratively updating the baseline parameters of the activity cost driver and the baseline cost of the standard activity path in S3.1 includes the following steps: S3.11. Extract complete operational and cost data from the previous day's data warehouse at fixed time intervals each day; S3.12. Based on the extracted data, statistically calculate the baseline parameters of each activity cost driver and the baseline cost of each standard activity path; S3.13. Based on the historical relationship between the total cost of the entire operation and the cost value of individual operation cost drivers, calculate the benchmark sensitivity coefficient of each operation cost driver; S3.14. Update the calculated benchmark parameters, benchmark costs, and benchmark sensitivity coefficients to the corresponding node attributes in the bulk grain terminal domain knowledge graph; S3.15. Pre-calculate and store the baseline total cost of all standard operating procedures as the basis for online inference.

[0011] As a further improvement to this technical solution, the process of quantitatively calculating the impact coefficients of each activity cost driver on the total activity cost in S3.2 includes the following steps: S3.21. Obtain real-time data on the cost drivers of each operational stage at the bulk grain terminal; S3.22 Determine the sensitivity coefficient of activity-based cost drivers The activity cost driver sensitivity coefficient Indicates the first In the first stage of bulk grain terminal operations The extent to which each activity cost driver affects the total cost of the entire activity chain; S3.23. Based on the relationship between the total cost of the entire operation and the cost value of individual operation cost drivers, calculate the sensitivity coefficient corresponding to each operation cost driver. ; S3.24, Calculate the sensitivity coefficient As the impact coefficient of this operation cost driver on the total operation cost; S3.25. Update the calculated influence coefficients to the corresponding node attributes in the knowledge graph of the bulk grain terminal domain.

[0012] As a further improvement to this technical solution, the process of adaptively adjusting the weight ratio of each activity cost driver in the dynamic path reasoning of the knowledge graph in S3.3 includes the following steps: S3.31, The activity cost driver sensitivity coefficient calculated in real time. The sensitivity coefficient is compared with the benchmark sensitivity coefficient stored in the knowledge graph of the bulk grain terminal field, and the sensitivity coefficient deviation value is calculated. S3.32 Determine the dynamic weighting factor The dynamic weighting factor express Time of the first In the first stage of bulk grain terminal operations The weighting of each activity cost driver in path reasoning; S3.33. Based on the deviation between the baseline weight and the sensitivity coefficient, calculate the dynamic weight factor at the current moment according to the preset linear adjustment rule. ; S3.34, Based on the calculated dynamic weighting factor Update the weight ratio of each activity cost driver in the dynamic path reasoning of the knowledge graph; S3.35. Store the updated weight percentages in the node attributes corresponding to the bulk grain terminal domain knowledge graph.

[0013] As a further improvement to this technical solution, the process of generating an intelligent scheduling scheme for bulk grain terminals that satisfies multi-dimensional optimization objectives in step S3.4 includes the following steps: S3.41. Compare the real-time operational cost driver data with the benchmark parameters stored in the bulk grain terminal domain knowledge graph, and calculate the rate of change of each operational cost driver. S3.42 Determine whether the rate of change of the activity cost driver exceeds the preset change threshold; S3.43 If the rate of change exceeds the threshold, locate all operation paths within the preset association range centered on the operation cost driver node in the knowledge graph of the bulk grain terminal domain. S3.44, Based on the adjusted dynamic weighting factor Recalculate the total cost of the work paths within the preset associated range, and retain the original calculation results of the unaffected work paths; S3.45. Sort all operation paths according to multi-dimensional optimization objectives to generate a smart scheduling scheme for bulk grain terminals that meets the multi-dimensional optimization objectives.

[0014] As a further improvement to this technical solution, in step S4, the process of performing multidimensional analysis of operating costs based on map tracing includes the following steps: S4.1. Call the knowledge graph of the bulk grain terminal domain, enable the full-link data traceability capability, and extract the relevant data of the entire business link of the bulk grain terminal; S4.2 Obtain the activity cost aggregation results; S4.3. Combine the data related to the entire business chain with the results of the collection of operating costs to determine the analytical dimensions of the bulk grain terminal's operational status; S4.4. According to the established analysis dimensions, conduct an analysis of the operational status of the bulk grain terminal item by item to obtain the analysis content corresponding to each dimension; S4.5 Integrate the analysis content from various dimensions to generate and output the operational analysis results of the bulk grain terminal.

[0015] The second objective of this invention is to provide a knowledge graph-based intelligent scheduling and operation analysis system for bulk grain terminals. The computer program of this system, when running, executes the aforementioned knowledge graph-based intelligent scheduling and operation analysis method for bulk grain terminals, including: A knowledge graph construction unit is used to construct a knowledge graph for the bulk grain terminal field by combining the cost aggregation logic of activity-based costing. The association mapping unit relies on the entity association capability of the knowledge graph in the bulk grain terminal field and combines the operation cost driver analysis method to establish the mapping relationship between the scheduling resources and operation cost elements of the bulk grain terminal. A multi-dimensional intelligent scheduling scheme generation unit, based on mapping relationships and real-time operational cost driver data, uses a knowledge graph-driven dynamic path reasoning method driven by operational cost drivers to update only the local operational paths affected by changes in cost drivers, generating an intelligent scheduling scheme for bulk grain terminals that meets multi-dimensional optimization objectives. This unit includes an offline baseline parameter pre-configuration module, a real-time weight quantization calculation module, a dynamic weight adaptive correction module, and a local path incremental update module, wherein: The offline benchmark parameter pre-configuration module iteratively updates the benchmark parameters of the operating cost drivers and the benchmark cost of the standard operating path based on historical operating data of the bulk grain terminal. The real-time weight quantification calculation module is used to combine real-time operation cost driver data to quantify and calculate the impact coefficient of each operation cost driver on the total operation cost. The dynamic weight adaptive correction module is used to adaptively adjust the weight ratio of each operation cost driver in the dynamic path reasoning of the knowledge graph based on the deviation between the real-time operating conditions and the baseline operating conditions. The local path incremental update module judges the degree of fluctuation of the operating cost driver based on a preset change threshold, and only performs incremental updates on the affected local operating paths to generate a smart scheduling scheme for bulk grain terminals that meets multi-dimensional optimization objectives. The multi-dimensional analysis unit for operating costs is used to conduct multi-dimensional analysis of the operating status of bulk grain terminals by relying on the full-link data traceability capability of the knowledge graph in the bulk grain terminal field and combining the results of operation cost collection, and output the corresponding operation analysis results.

[0016] Compared with the prior art, the beneficial effects of the present invention are as follows: 1. This invention constructs a knowledge graph for bulk grain terminals based on operational cost, integrating operational cost drivers as core entities into the knowledge graph system, and establishing a quantitative mapping relationship between scheduling resources and operational cost elements. Through a dynamic path reasoning method driven by operational cost drivers in the knowledge graph, only local operational paths affected by changes in cost drivers are updated, reducing the computational complexity of global path re-reasoning, shortening the response time of scheduling schemes, and realizing the synergistic optimization of operational efficiency and operational cost under dynamic operating conditions. 2. This invention relies on the full-link entity association capability of the knowledge graph in the bulk grain terminal field, breaking through the limitations of the fixed table structure of traditional relational databases, and realizing full-link traceability of cost data across operational links; combined with the operational cost collection results, it conducts multi-dimensional operational status analysis, improves the accuracy of operational cost collection, enriches the dimensions of operational analysis, and keeps the operational analysis results synchronized with the operational process, providing data support for refined operational decision-making of bulk grain terminals. Attached Figure Description

[0017] Figure 1 This is a schematic diagram of the overall method steps of the present invention; Figure 2 This is a schematic diagram of the overall process of generating the multi-dimensional intelligent scheduling scheme in this invention; Figure 3 This is a schematic diagram of the overall system framework of the present invention; The meanings of the labels in the diagram are as follows: 1. Knowledge graph construction unit; 2. Association mapping unit; 3. Multi-dimensional intelligent scheduling scheme generation unit; 4. Multidimensional analysis unit for operating costs. Detailed Implementation

[0018] 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 of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0019] like Figures 1-2 As shown, this embodiment provides a knowledge graph-based intelligent scheduling and operation analysis method for bulk grain terminals, including: S1. Construction of a Knowledge Graph for Bulk Grain Terminals Based on Activity-Based Costing: Combining the cost aggregation logic of activity-based costing, a knowledge graph for bulk grain terminals is constructed. This knowledge graph includes operational processes, resource elements, cost drivers, and the relationships between these elements. Specifically, it includes the following steps: S1.1 Based on the cost aggregation logic of activity-based costing, define a dedicated ontology architecture for the knowledge graph of bulk grain terminals. The ontology architecture is adapted to the business attributes and association rules of bulk grain terminal operation links, resource elements, and cost drivers. Specifically, the first step is to clarify which types of entities will be represented in the graph and the semantic relationships between them. This step defines a dedicated ontology architecture for the bulk grain terminal domain based on the aggregation logic of activity-based costing: "resource consumption is aggregated to activities, and activity costs are allocated to cost objects."

[0020] Furthermore, the entity types include four core ontology categories: The first category is operational processes, corresponding to various operational activities in the bulk grain terminal's technological flow. Based on business characteristics, these are divided into four primary types: unloading operations, horizontal transport operations, warehousing operations, and distribution operations. Unloading operations are further subdivided into sub-nodes such as unloading machine grabbing, hopper receiving, and belt conveyor operation. Warehousing operations are subdivided into sub-nodes such as warehousing, warehousing, and transfer. Each operational node is configured with attributes such as standard time consumption, rated energy consumption, and designed capacity.

[0021] Second, there are resource elements, which correspond to the resources consumed or occupied during the operation. These are categorized by resource form into equipment resources (ship unloaders, belt conveyors, bucket elevators, truck loaders, etc.), human resources (ship unloader drivers, inspectors, central control operators, etc.), energy resources (electricity, fuel oil), and material resources (dust removal consumables, lubricating grease). Equipment resource nodes carry attributes such as rated power, design life, and cumulative operating time; human resource nodes carry attributes such as skill level and hourly rate.

[0022] Thirdly, there are cost drivers, which correspond to various driving factors that cause changes in costs. Based on the nature of the drivers, they are divided into workload drivers (tonnage unloaded, tonnage transported), efficiency drivers (equipment operating rate, failure rate, waiting time), price drivers (peak and off-peak electricity prices, fuel price per unit), and condition drivers (equipment age, weather conditions, tidal periods).

[0023] Fourthly, there are cost object classes, which serve as the target carriers for the final cost collection, such as single ship operations, single batch of goods, and single customer orders. Attribute fields such as total cost and unit cost are set on the nodes.

[0024] The relationships between the above ontology classes are defined synchronously: the relationship between operational stages and resource elements is "consumption," indicating which resources are used in a particular operation; the relationship between resource elements and cost drivers is "driving," indicating which factors drive changes in the cost of a certain type of resource; the relationship between adjacent operational stages is "connection," forming a sequential link in the process flow; the relationship between cost drivers and operational stages is "influence," indicating that fluctuations in drivers are transmitted to operational costs through this relationship; and the relationship between cost objects and operational stages is "aggregation," indicating the ultimate goal of cost aggregation. These five types of relationships constitute the semantic skeleton of cost aggregation at bulk grain terminals. Once the ontology architecture is defined, the schema of the graph is determined.

[0025] S1.2 Based on the defined ontology architecture, accurately extract the entities of bulk grain terminal operation links, resource elements, and cost drivers from the bulk grain terminal business data source. Specifically, based on the ontology architecture defined in S1.1, entity instances are extracted from various business data sources at the bulk grain terminal. These data sources cover the terminal's production management system, equipment management system, energy management system, human resources system, and financial system. Details are as follows: For structured data such as equipment ledgers, work order records, and energy meters, a mapping relationship is directly established between data table fields and ontology class attributes, transforming each record into a corresponding entity instance. For example, a ship unloader record in the equipment ledger is directly mapped to an equipment resource entity, with its rated capacity, installed power, and other fields written into the corresponding attributes.

[0026] For semi-structured data such as job logs and operation records, key fields are extracted using rule templates, such as job type, start and end time, and job quantity, to generate job process entity instances.

[0027] For unstructured texts such as shift handover records and fault reports, a named entity recognition model is used to automatically extract information such as equipment names and abnormal statuses. After manual review and confirmation, entities are added.

[0028] During the extraction process, each entity node records the source system, extraction time, and original record identifier to support subsequent data traceability.

[0029] S1.3 Based on the cost aggregation logic of activity-based costing, explore the business relationships and cost coupling relationships between entities, and construct the relationship mapping links between entities; Specifically, once entity instances are in place, the association links between entities are constructed according to the semantic relationships defined in the ontology architecture. The details are as follows: The first type of relationship is "consumption". By linking work orders with equipment operation records, we can determine which equipment, energy and manpower were consumed in each work process instance. In the graph, the corresponding work process nodes and resource element nodes are connected by the "consumption" relationship, and the consumption amount or consumption duration is recorded on the edge.

[0030] The second type of relationship is "driving force". By analyzing the correlation between historical operational data and cost data, we can identify which drivers influence the cost of each resource element. Equipment costs are mainly driven by operational volume and efficiency drivers, while energy costs are mainly driven by operational volume and price drivers. Following this rule, resource element nodes are connected to their corresponding cost driver nodes using "driving force" relationship edges.

[0031] The third type of relationship is "connection". Based on the fixed process flow sequence of the bulk grain terminal, namely unloading → horizontal transportation → storage or distribution, the upstream and downstream relationships of adjacent operation links are determined by analyzing the operation sequence and material flow data, and "connection" relationship edges are established to form an operation link that runs through the entire process.

[0032] The fourth type of relationship is "influence". An "influence" relationship edge is established between the cost driver node and the operational node affected by it. The edge records the degree of influence of the cost driver on the operational cost, providing a basis for cost transmission in subsequent path reasoning.

[0033] After the four types of relationships are constructed, the graph forms a three-dimensional network with the operational links as the vertical backbone and resource consumption and cost drivers as the horizontal links. The aggregation logic of activity-based costing is solidified into the topology of the graph.

[0034] S1.4. The extracted entities and the constructed relationships are integrated, standardized, and stored in a solidified manner to generate a complete knowledge graph of the bulk grain terminal domain.

[0035] Specifically, after entities and relationships are constructed, the graph is fused, regularized, and persistently stored. Among these: The fusion and normalization process primarily addresses entity conflicts and relation redundancy issues arising from different data sources. First, entity alignment is performed, merging entities pointing to the same real-world object from different systems (e.g., the same ship unloader recorded in both the equipment management system and the energy management system) by matching key attributes such as equipment number, thus merging the attribute values ​​from both sides. Second, relation deduplication is performed, merging duplicate relation edges between the same pair of entities caused by multiple extractions. Finally, conflict resolution is implemented; when different data sources provide inconsistent values ​​for the same attribute, the data source reliability priority is used to determine the order of priority, with original equipment technical parameters taking precedence over operational monitoring data, and production records taking precedence over financial allocation records.

[0036] Graph databases are chosen as the storage medium. Graph databases use nodes and edges as their native storage units, and their storage structure is consistent with the graph model of knowledge graphs. This makes them far more efficient than relational databases in handling multi-hop path traversal and subgraph queries, providing fundamental performance assurance for the subsequent dynamic path reasoning in S3 and full-link tracing in S4. All merged and normalized nodes and relation edges are imported into the graph database. Node labels are created according to ontology classes, indexes are created according to relation types, and attribute indexes are established for frequently queried attributes.

[0037] After storage is complete, perform an integrity check: check whether all job link nodes are connected to the process link, check whether all resource element nodes are connected to at least one job link node and one cost driver node, randomly select several complete paths for end-to-end traversal, and confirm that the association relationship is correct and the path is reachable.

[0038] With this, the knowledge graph for the bulk grain terminal sector has been completed. This graph, with its internal structure based on the aggregation logic of activity-based costing and its physical carrier being a graph database, provides a unified semantic data foundation for the mapping and association between scheduling resources and cost drivers in S2, the dynamic path reasoning based on changes in cost drivers in S3, and the cross-link, end-to-end cost tracing in S4.

[0039] S2. Graph Relationship Mapping between Cost Drivers and Scheduling Resources: Leveraging the entity association capabilities of the knowledge graph in the bulk grain terminal domain, and combining it with operational cost driver analysis methods, a mapping relationship is established between the scheduling resources and operational cost elements of bulk grain terminals. Specifically, this includes the following steps: S2.1 Based on the operational cost driver analysis method, sort out the attribute characteristics of the scheduling resources and operational cost elements of bulk grain terminals, as well as the corresponding rules between the two types of elements; Specifically, based on the operational cost driver analysis method, the scheduling resources and operational cost elements of bulk grain terminals are systematically sorted out to clarify their respective attributes and the corresponding rules between them.

[0040] Among them, dispatchable resources refer to production factors that can be allocated and managed by the dispatch system during bulk grain terminal operations. These mainly include equipment resources (ship unloaders, belt conveyors, bucket elevators, car loaders, etc.) and human resources (ship unloader drivers, inspection personnel, central control operators, etc.). Each type of dispatchable resource has characteristics such as status attributes (in use / idle / faulty), capacity attributes (rated efficiency, applicable cargo types), and cost attributes (unit labor hour rate, unit energy consumption). Operating cost elements refer to cost accounting items in the operating cost accounting system. In the knowledge graph constructed by S1, these cost attributes correspond to cost driver classes and related operating process classes, mainly including unit operating costs, energy costs, labor costs, and equipment depreciation costs for each operating process.

[0041] Building upon this foundation, the correspondence between scheduled resources and operational cost elements is further clarified. When a scheduled resource is used, it enters an operational state, and resource consumption is simultaneously converted into operational costs. This correspondence follows two basic rules: First, the direct consumption rule: scheduled resources directly consume energy and materials during operation, with consumption directly proportional to the workload; the corresponding cost elements are energy cost and material cost. Second, the occupancy and depreciation rule: scheduled resources incur time-related costs during occupancy, including equipment depreciation and labor hour costs; the corresponding cost elements are depreciation cost and labor cost. Through this clarification, a clear correspondence logic is established between scheduled resources and operational cost elements.

[0042] S2.2 Utilize the entity association capability of the bulk grain terminal domain knowledge graph to determine the corresponding node positions of the bulk grain terminal scheduling resource entity and the operation cost element entity within the graph; Specifically, after sorting out the corresponding rules, the entity association capability of the bulk grain terminal domain knowledge graph constructed by S1 is invoked to accurately locate the node positions of scheduling resource entities and operation cost element entities in the graph.

[0043] Specifically, in the knowledge graph constructed by S1, scheduling resource entities are located under the equipment resource and human resource subclasses of the resource element class. Each entity is assigned a unique node identifier upon entry into the database. Taking the ship unloader as an example, its nodes in the graph, in addition to carrying basic attributes such as rated power, design life, and cumulative operating time, are also connected to energy resource nodes through "consumption" relationship edges and to cost driver nodes through "drive" relationship edges, forming a preliminary association network in the graph. Operational cost element entities are distributed in the cost attributes of the cost driver class and the operational link class, such as the energy consumption cost attribute of the ship unloading operation and the equipment depreciation attribute captured by the ship unloader.

[0044] During the localization process, the node label index and attribute index of the graph database are used to quickly retrieve and confirm the node identifiers and storage locations of each scheduled resource entity and operation cost element entity by using scheduling resource type and cost element type as query conditions. The search results are summarized to form a node correspondence table, providing accurate start and end node information for the next step of building the association link.

[0045] S2.3. According to the corresponding rules obtained from the analysis, establish the connection links between the corresponding nodes in the map to complete the initial mapping and binding of bulk grain terminal scheduling resources and operating cost elements; Specifically, after determining the node location, the corresponding rules outlined in S2.1 are used to establish the associated links between resource entity nodes and activity cost element entity nodes in the graph one by one.

[0046] For direct consumption rules, a "cost mapping" relationship is established between the scheduling resource node and the corresponding energy consumption cost or material cost node. For example, the unloader node is connected to the electricity consumption cost node of the unloading operation, and the edge records the consumption coefficient, which represents the baseline value of electricity consumption generated by the equipment per unit of operation. This coefficient is derived from the statistical analysis of historical operation data and can be differentiated according to equipment model and operating conditions. For occupancy loss rules, a "cost mapping" relationship is established between the scheduling resource node and the corresponding depreciation cost or labor cost node. For example, the unloader node is connected to the equipment depreciation cost node, and the edge records the depreciation rate per unit time; the unloader driver node is connected to the labor cost node, and the edge records the labor hour rate.

[0047] After the above operations, the initial mapping and binding between scheduling resources and operational cost elements is completed. At this point, when the system queries a scheduling resource node, it can directly obtain all associated cost element information through multi-hop traversal of the graph, thereby enabling real-time consideration of the cost dimension in the scheduling decision-making process.

[0048] S2.4. Perform logical consistency verification and correction on the established association links to solidify a stable mapping relationship.

[0049] Specifically, after the initial mapping and binding are completed, in order to ensure the accuracy and integrity of the mapping relationship, logical consistency verification and correction processing are carried out on the established association links.

[0050] The verification process is being carried out systematically in three aspects, as follows: First, integrity verification involves checking each type of scheduled resource to ensure that it is associated with at least one direct consumption cost mapping and one occupancy loss cost mapping, guaranteeing comprehensive coverage of cost elements. Any uncovered items discovered during the check are promptly supplemented with the corresponding associated links. Secondly, consistency verification checks whether the associated cost element types and mapping coefficients of different instances of the same type of scheduled resources remain consistent within a reasonable range. Instances with deviations exceeding the preset range are manually checked to determine whether the deviation is due to differences in business scenarios or data entry errors. If it is indeed a data error, it is corrected. Third, path reachability verification: Starting from the scheduling resource node, traverse the "cost mapping" relationship edge hop by hop to verify whether the corresponding operation cost element node can be reached within a preset number of steps. If a path interruption is found during the traversal, investigate the cause and supplement the missing relationship edge to ensure that each mapping link is reachable.

[0051] After verification, the corrected and confirmed association links are solidified. The type, attributes and mapping coefficients of the relationship edges are stored in the graph database with immutable version tags to prevent accidental modification in subsequent operations.

[0052] At this point, the mapping relationship between the scheduling resources and operational cost elements of the bulk grain terminal is stably established. This mapping relationship provides a quantitative cost transmission channel for dynamic path reasoning based on real-time cost driver data in S3, enabling the scheduling scheme generation process to simultaneously consider both resource allocation and cost impact. It also provides a correlation basis for tracing back the cost composition from the perspective of scheduling decisions in S4.

[0053] S3. Generation of a multi-dimensional intelligent scheduling scheme driven by operational cost drivers: Based on the mapping relationship constructed in S2 and real-time operational cost driver data, a knowledge graph dynamic path reasoning method driven by operational cost drivers is used to update only the local operational paths affected by changes in cost drivers, generating an intelligent scheduling scheme for bulk grain terminals that meets multi-dimensional optimization objectives; specifically, it includes the following steps: S3.1. Based on historical operational data of the bulk grain terminal, iteratively update the baseline parameters of operational cost drivers and the baseline cost of standard operating routes. The offline phase is executed daily from 02:00 to 04:00, utilizing the complete operational and cost data accumulated the previous day to iteratively update the baseline parameters of operational cost drivers, the baseline cost of standard operating routes, and the driver sensitivity coefficients. It also pre-calculates the baseline total cost of all standard routes, preparing a static reference basis for online inference. Specifically, this includes the following steps: S3.11. Extract complete operational and cost data from the previous day's data warehouse at fixed time intervals each day; In this step, from 02:00 to 04:00 each day, the system retrieves the previous day's (00:00:00 to 23:59:59) operational and cost data from the bulk grain terminal data warehouse. The data covers the start and end times, workload, equipment and personnel call records from the production management system; equipment runtime, fault alarms, and energy consumption monitoring values ​​from the equipment management system; time-of-use electricity consumption and peak / valley electricity prices from the energy management system; and the already calculated operational cost records from the financial system.

[0054] After extraction, the number of records and key fields of each data source are checked for completeness. If data is missing or the number of records deviates from the historical average for the same period by more than a preset threshold, the supplementary data collection process is automatically triggered to ensure the quality of input data.

[0055] S3.12. Based on the extracted data, statistically calculate the baseline parameters of each activity cost driver and the baseline cost of each standard activity path; In this step, based on the extracted data, the baseline parameters for each activity cost driver are updated. For activity volume drivers (such as tonnage unloaded), the mean and standard deviation are calculated for different time periods and work lines, and the baseline values ​​and fluctuation ranges are updated. For efficiency drivers (such as equipment operating rate and failure rate), the average operating rate and downtime percentage of various types of equipment are calculated, and the baseline efficiency parameters are updated. For price drivers (such as electricity price), the weighted average price actually executed the previous day is taken as the new baseline price parameter.

[0056] Simultaneously, the baseline cost of each standard operating procedure (SOP) is updated. A SOP refers to a complete sequence of operational steps connected by "connection" relationships in the knowledge graph, from the starting point of the self-unloading vessel to the destination of the evacuation. Let a certain path consist of... It consists of several links, of which the first link is... The benchmark cost for each stage is Then the baseline total cost of this path is: ; In the formula: Indicates the first The baseline cost for each operational step is the sum of the energy cost, labor cost, and equipment depreciation cost of that step on the previous day, derived from actual consumption data. To mitigate daily fluctuations, the baseline cost is updated using an index-weighted moving average method, which proportionally integrates the previous day's actual statistical value with historical baseline values. Recent weights can be adjusted based on operational needs. This indicates the total number of operation steps included in the path; Indicates the first [item] on this path The baseline cost of each work process is the sum of the energy consumption cost, labor cost, and equipment depreciation cost actually incurred in that process on the previous day, and is derived from the consumption data of the corresponding process.

[0057] S3.13. Based on the historical relationship between the total cost of the entire operation and the cost value of individual operation cost drivers, calculate the benchmark sensitivity coefficient of each operation cost driver; In this step, after updating the baseline parameters and baseline costs, the baseline sensitivity coefficient for each activity cost driver is calculated. This coefficient quantifies the relative impact of a single cost driver on the total cost of the entire activity chain. Based on historical changes in the total cost of the entire chain and the costs of each cost driver over the past 30 days, the baseline sensitivity coefficient is calculated. Defined as: ; In the formula: Indicates the first Within the sample period, the first In the first step of the operation The cost value corresponding to each cost driver; Indicates the first The total cost of the entire operation chain for each sample time period; express In all The arithmetic mean of the samples; express In all The arithmetic mean of the samples; This represents the total number of sample time periods.

[0058] In the formula, the numerator of the first fraction is the covariance (dimension [unit]) of the driver cost deviation and the total cost deviation. ² The denominator is the variance of the cost deviation (dimension [yuan]). ² The first fraction, after dividing the second, is a dimensionless regression coefficient; the second fraction is the ratio of the mean driver cost to the mean total cost, also dimensionless. Multiplying the two yields the elastic form. This allows for direct comparison of the sensitivity of cost drivers with different dimensions and properties. The larger the absolute value, the more significant the impact of the driver on the total cost.

[0059] S3.14. Update the calculated benchmark parameters, benchmark costs, and benchmark sensitivity coefficients to the corresponding node attributes in the bulk grain terminal domain knowledge graph; In this step, the baseline parameters and baseline cost obtained in S3.12, and the baseline sensitivity coefficient obtained in S3.13 are used. The corresponding node attributes in the bulk grain terminal domain knowledge graph are updated through the transaction write mechanism of the graph database. Operational node updates include their baseline energy consumption cost, baseline labor cost, and baseline equipment depreciation cost; cost driver nodes update their baseline values, fluctuation ranges, and... Standard operation path node update Each update synchronously records the timestamp and version number, retaining a number of recent historical versions to support backtracking.

[0060] Before writing, the change range is checked: if the daily update range of a certain parameter exceeds the preset safety threshold (such as the daily fluctuation of the benchmark cost exceeding 30%), the system will suspend the update and trigger an alarm. The write will be performed after manual verification to prevent abnormal data from interfering with the benchmark system.

[0061] S3.15. Pre-calculate and store the baseline total cost of all standard operating procedures as the basis for online inference.

[0062] In this step, after the update is complete, traverse all complete job paths in the knowledge graph connected by "connection" edges, and for each path, accumulate the latest values ​​of each stage according to the formula in S3.12. To obtain the updated path And simultaneously extract all cost driver nodes on the path. This forms the baseline sensitivity vector for the path. The pre-calculated results are written to the graph database cache with the path identifier as the key. They can be directly read during subsequent dynamic path reasoning without repeatedly aggregating multi-node attributes, thus effectively shortening the time required to generate the scheduling scheme.

[0063] S3.2. Combining real-time activity cost driver data, quantitatively calculate the impact coefficient of each activity cost driver on the overall activity cost; specifically, this includes the following steps: S3.21. Obtain real-time data on the cost drivers of each operational stage at the bulk grain terminal; In this step, the system continuously collects real-time cost driver data for each operational stage from the SCADA, MES, and ERP systems of the bulk grain terminal. The collected data covers: instantaneous workload of each operational stage (e.g., the current tonnage grabbed by the unloader, the current tonnage conveyed by the belt conveyor), real-time equipment operating speed and fault status signals, current electricity price and real-time electricity consumption data pushed by the energy management system, and on-duty personnel and work hour records fed back by the manual check-in system.

[0064] Meanwhile, the data collection cycle is set according to the type of driving factor. The data collection cycle for equipment status is at the second level, the data collection cycle for workload is at the minute level, and the price data is triggered to be updated when the electricity price time period changes or the unit price changes.

[0065] Finally, the collected raw data undergoes outlier detection and cleaning to remove outliers caused by sensor jumps and communication interruptions, forming an effective sample stream that can be used for subsequent calculations.

[0066] S3.22 Determine the sensitivity coefficient of activity-based cost drivers Activity-based cost driver sensitivity coefficient Indicates the first In the first stage of bulk grain terminal operations The extent to which each activity cost driver affects the total cost of operations across the entire value chain; specifically... Compared with the baseline sensitivity coefficient calculated based on long-term historical data in S3.13 Different, here Based on real-time data within a recent short-term window, it reflects the cost impact intensity of each driver under the current operating conditions. Its physical meaning is still the elasticity coefficient, that is, the relative change in the total cost of the entire link caused by a unit relative change in the cost value of the driver.

[0067] S3.23. Based on the relationship between the total cost of the entire operation and the cost value of individual operation cost drivers, calculate the sensitivity coefficient corresponding to each operation cost driver. ; In this step, based on the relationship between the total cost of the entire operation and the cost value of individual operation cost drivers within a short time window, the real-time sensitivity coefficient corresponding to each operation cost driver is calculated. .

[0068] Set the duration of the sliding time window to (For example, take 15 minutes), the current time is The sample window is Assume a total of [number] samples are collected within the window. The first valid sample, the... In the nth sample The first step of the operation The cost value of each cost driver is The total cost of the end-to-end operation at the corresponding time is First, calculate the mean of each variable within the window: ; ; In the formula: Indicates the first in the window In the first step of the operation The arithmetic mean of the cost values ​​of each cost driver is expressed in yuan. This represents the arithmetic mean of the total cost of all operations within the window, expressed in yuan. This represents the number of valid samples within the sliding window; it is dimensionless.

[0069] Based on the window mean, the real-time sensitivity coefficient The calculation formula is: ; In the formula: Indicates the first [number]th ... In the first step of the operation The arithmetic mean of the cost values ​​of each cost driver is expressed in yuan. This represents the arithmetic mean of the total cost of all operations within the sliding window, expressed in yuan. Formula structure and baseline sensitivity coefficient in S3.13 Consistency: The numerator of the first fraction is the covariance of the driver cost and the total cost within the window, and the denominator is the variance of the driver cost; dividing them yields the regression coefficient. The second fraction is the ratio of the mean driver cost to the mean total cost, converting the regression coefficient into an elastic form. Both the numerator and denominator are dimensional (e.g., units of yuan). ² The result after division is dimensionless. Ultimately, it is a dimensionless elastic coefficient, which can be compared with... Direct comparison.

[0070] Additionally, window length The selection of the window balances sensitivity and stability: if the window is too short, there will be insufficient samples. It is susceptible to interference from single-point anomalies; an excessively long window will result in a delayed response to changes in operating conditions. Based on the operational rhythm of the bulk grain terminal, It can be set within a range of 15 to 30 minutes and dynamically adjusted according to the peak and off-peak periods of terminal operations.

[0071] S3.24, Calculate the sensitivity coefficient This serves as the impact coefficient of the cost driver of this operation on the overall operation cost; specifically, it represents the real-time sensitivity coefficient calculated in S3.23. The impact coefficient of this activity as a direct cost driver on the overall cost of the entire operation chain. This impact coefficient reflects the impact of the first activity as a cost driver under the current operating conditions. The first step The marginal elasticity of the impact of each driver on total cost, with the magnitude and direction of the impact representing the strength and direction of increase or decrease, respectively.

[0072] S3.25. Update the calculated influence coefficients to the corresponding node attributes in the knowledge graph of the bulk grain terminal domain.

[0073] In this step, the influence coefficient determined in S3.24 is... The current sensitivity attribute field of the corresponding cost driver node in the bulk grain terminal domain knowledge graph is updated in real time. The update operation uses single-node attribute writing in the graph database, recording a timestamp for each update and retaining a historical value sequence of the most recent updates for analyzing sensitivity change trends during subsequent weight adjustments. After the update is completed, the driver node can be directly read by the S3.3 dynamic weight adaptive correction step and participate in the weight factor calculation.

[0074] S3.3. Based on the deviation between real-time operating conditions and baseline operating conditions, adaptively adjust the weight ratio of each operation cost driver in the dynamic path reasoning of the knowledge graph; specifically, this includes the following steps: S3.31, The activity cost driver sensitivity coefficient calculated in real time. The sensitivity coefficient is compared with the benchmark sensitivity coefficient stored in the knowledge graph of the bulk grain terminal field, and the sensitivity coefficient deviation value is calculated. In this step, the sensitivity coefficients calculated in real time by S3.2 are... Compared with the baseline sensitivity coefficient stored in the knowledge graph of S3.1 Compare the two and calculate the deviation: ; Indicates the first In the first step of the operation The sensitivity coefficient deviation value of each cost driver is dimensionless.

[0075] when When this occurs, it indicates that the current impact of the driving force on the total cost of the entire link is higher than the historical baseline level, and more attention should be paid to it in path reasoning; when When the influence of a certain factor decreases, its weight can be reduced accordingly.

[0076] S3.32 Determine the dynamic weighting factor Dynamic weighting factor express Time of the first In the first stage of bulk grain terminal operations The weighting of each activity cost driver in path reasoning; specifically... This is a dimensionless proportional value, and the sum of the weights of each driver within the same operational stage satisfies the normalization constraint. This factor will directly participate in the weighted calculation of the total path cost, and the larger the weight of the driver, the stronger its influence on path optimization.

[0077] S3.33. Based on the deviation between the baseline weight and the sensitivity coefficient, calculate the dynamic weight factor at the current moment according to the preset linear adjustment rule. ; In this step, based on the deviation between the baseline weight and the sensitivity coefficient, the dynamic weight factor at the current moment is calculated according to a preset linear adjustment rule. Let the first... In the first step of the operation The baseline weight of each cost driver is Adjust the step size coefficient to The calculation formula is: ; The baseline weight representing the cost driver is determined by the average proportion of the cost driver in the historical cost structure and is dimensionless. This represents the preset weight adjustment step size coefficient, used to control the response magnitude of the weights to sensitivity deviations; it is dimensionless. The deviation value of the sensitivity coefficient calculated by S3.31 is dimensionless.

[0078] The value of needs to balance response speed and weight stability: if the value is too small, the weight will be insensitive to changes in operating conditions, and the adjustment effect will be insignificant; if the value is too large, the weight may fluctuate violently due to short-term volatility. Based on experience in bulk grain terminal operations, The recommended value is between 0.1 and 0.5. In actual deployment, the value can be determined after simulation and optimization based on historical data.

[0079] To prevent weight drift from exceeding a reasonable range, the calculated... Apply constraints: if If the value is below the preset lower bound, the lower bound value is used; if the value is above the preset upper bound, the upper bound value is used. After constraint, the weights of each driver within the same work process are normalized to ensure that the sum of the weights is 1, maintaining the consistency of the contribution ratio of each process in the path cost calculation.

[0080] S3.34, Based on the calculated dynamic weighting factor Update the weight ratio of each activity cost driver in the dynamic path reasoning of the knowledge graph; In this step, the dynamic weighting factors are calculated and constrained based on S3.33. This update operation resizes the weight of each cost driver in the dynamic path reasoning of the knowledge graph. The update operation replaces the current reasoning weight of each driver node with... In subsequent calculations of total path cost, the cost item corresponding to this driver will be weighted using the new weights.

[0081] S3.35. Store the updated weight percentages in the node attributes corresponding to the bulk grain terminal domain knowledge graph.

[0082] In this step, the dynamic weighting factor obtained in S3.33 is... Write the current weight attribute field of the corresponding cost driver node in the bulk grain terminal domain knowledge graph. Write the operation record timestamp to retain the historical weight sequence for subsequent analysis of weight drift trends and optimization adjustment step size coefficients.

[0083] Once the storage is complete, each driving node will have weight information reflecting the current operating condition, which can be directly called by S3.4 during local path incremental updates.

[0084] S3.4. Based on a preset change threshold, determine the degree of fluctuation of operating cost drivers, and only perform incremental updates on the affected local operating paths to generate an intelligent scheduling scheme for bulk grain terminals that meets multi-dimensional optimization objectives. Specifically, this includes the following steps: S3.41. Compare the real-time operational cost driver data with the benchmark parameters stored in the bulk grain terminal domain knowledge graph, and calculate the rate of change of each operational cost driver. In this step, the real-time collected operational cost driver data is compared with the baseline parameters of that driver stored in the knowledge graph of the bulk grain terminal sector, and the rate of change of each cost driver is calculated one by one. For the first... In the first step of the operation One cost driver, let its real-time value be... The baseline value is Then the rate of change The calculation formula is: ; express Time of the first In the first step of the operation The real-time value of each cost driver is given, with the unit depending on the type of driver (e.g., the unit for volume drivers is tons, and the unit for price drivers is yuan / kilowatt-hour). The baseline value representing this cost driver is calculated by S3.12 and stored in the knowledge graph, with units equal to... Consistent; This represents the rate of change of the cost driver, expressed as a percentage; a positive value indicates an increase, and a negative value indicates a decrease.

[0085] For different types of cost drivers The methods for determining the values ​​differ: the workload driver is the average workload within the current sliding window; the efficiency driver is the instantaneous or short-term average of the current equipment operating rate; the price driver is the current electricity price or fuel price per unit; and the status driver is the current status indicator value (e.g., 1 for equipment fault status and 0 for normal status). After being uniformly converted into dimensionless rates of change using the above formulas, the fluctuation levels of different types of drivers can be compared on the same scale.

[0086] S3.42 Determine whether the rate of change of the activity cost driver exceeds the preset change threshold; In this step, the rate of change of each cost driver calculated in S3.41 is used. The value is compared with a preset change threshold. The preset change threshold is set based on experience in bulk grain terminal operation management, with a typical value of ±5%. That is, when the real-time value of the driving factor deviates from the benchmark value by less than ±5%, it is considered a normal fluctuation and does not trigger a path update; when the deviation exceeds ±5%, it is considered a significant fluctuation and triggers an incremental path update.

[0087] The logic for determining the threshold condition is as follows: like (in If a preset change threshold (e.g., 5%) is set, then the fluctuation of the driver is within an acceptable range, and the original cost calculation results of the driver's associated path are retained without adjustment. like If the fluctuation of the driving force exceeds the normal range, the cost of the local path of its influence needs to be recalculated.

[0088] By using threshold filtering, the system only responds to factors that have indeed undergone significant changes, avoiding frequent full recalculations due to minor fluctuations. This effectively controls the amount of inference computation while ensuring the cost sensitivity of the scheduling scheme.

[0089] S3.43 If the rate of change exceeds the threshold, locate all operation paths within the preset association range centered on the operation cost driver node in the knowledge graph of the bulk grain terminal domain. In this step, for cost driver nodes whose rate of change exceeds a threshold, all operational paths within a preset association range centered on that node are located in the bulk grain terminal domain knowledge graph.

[0090] The localization method employs multi-hop neighborhood expansion of a graph. Starting with a cost driver node exceeding a threshold, it finds operational nodes directly affected by that driver along the "influence" relationship edges, and then expands upstream and downstream along the "connection" relationship edges. The preset association range is based on the number of hops. Measurement, based on the transmission characteristics of the bulk grain terminal operation chain, A typical value is 3. The expansion proceeds simultaneously along both the positive and negative directions of the "connection" relationship, ensuring coverage of the transmission effects of this driving force change in both upstream and downstream directions throughout the process.

[0091] After the expansion is complete, all traversed job paths within the range are collected. These paths constitute a subset of local paths affected by the fluctuation of this driver. The remaining paths not within this range are considered unaffected by this fluctuation, and their cost calculation results remain unchanged. If multiple drivers exceed the threshold simultaneously, expansion is performed with each driver node as the center, and the union of the path subsets covered by each is taken to obtain the set of all affected paths.

[0092] S3.44, Based on the adjusted dynamic weighting factor Recalculate the total cost of the work paths within the preset associated range, and retain the original calculation results of the unaffected work paths; In this step, the affected path subset located in S3.43 is adjusted based on the dynamic weight factor updated in S3.3. Recalculate the total cost for each path. The formula for calculating the total cost for the affected paths is: ; express The weighted total cost of this path at any given time, in yuan; This indicates the total number of operation steps included in the path; Indicates the first The number of cost drivers involved in cost calculation in each operational stage; express Time of the first In the first step of the operation The dynamic weighting factors of each cost driver are calculated by S3.33 and are dimensionless. express The real-time cost value corresponding to this driver is calculated by multiplying the real-time workload by the resource unit price, and the unit is yuan.

[0093] For unaffected paths, the baseline total cost pre-calculated in S3.15 or the calculation results retained from the previous update are directly used, without the need for recalculation. This "local incremental update" strategy compresses the scope of path cost recalculation from all candidate paths to a subset of affected paths. The computational load is only related to the size of the neighborhood graph of the node with the changed driving force, decoupling it from the total number of paths in the entire terminal process, effectively ensuring the response speed in real-time scheduling scenarios.

[0094] S3.45. Sort all operation paths according to multi-dimensional optimization objectives to generate a smart scheduling scheme for bulk grain terminals that meets the multi-dimensional optimization objectives.

[0095] In this step, after updating the local path cost, all candidate operation paths are comprehensively sorted according to multi-dimensional optimization objectives to generate a smart scheduling scheme for bulk grain terminals that meets the requirements.

[0096] The multi-dimensional optimization objectives include at least: lowest total operating cost, shortest total operating time, and most balanced equipment utilization. The quantification method for each objective is as follows: total operating cost is directly taken as... The total operation time is derived by combining the standard time of each operation step along the path with the real-time efficiency correction coefficient; the equipment utilization balance is measured by the load variance of the equipment involved in the path. These multi-objectives are then converted into a comprehensive path score using a weighted scalarization method. ; This represents the overall score of the path, which is dimensionless; the lower the score, the better the overall performance. This represents the normalized total path cost index, which is dimensionless. This represents the normalized total path duration, and is dimensionless. This represents the normalized equipment load balancing index, which is dimensionless. These represent the weighting coefficients for cost, time, and equilibrium, respectively, satisfying... The specific value is determined by the scheduling strategy configuration.

[0097] Each indicator is normalized using the min-max standardization method, mapping the original indicator values ​​of different dimensions to the [0,1] interval. Sorting is done by... The paths are sorted from lowest to highest score, and the top few paths (e.g., the top 3) with the best scores are output as recommended solutions. After the scheduler selects an execution plan from the recommended plans, the system issues the selected scheduling instruction to the dock production management system. At the same time, the path information, weights of each driving factor, and cost estimation data corresponding to the plan are written back to the data warehouse, forming a data loop for the next S3.1 iteration.

[0098] S4. Multidimensional Analysis of Operating Costs Based on Knowledge Graph Traceability: Leveraging the end-to-end data traceability capabilities of the knowledge graph in the bulk grain terminal sector, and combining it with operational cost aggregation results, this analysis performs multidimensional analysis of the operational status of bulk grain terminals and outputs corresponding operational analysis results. Specifically, it includes the following steps: S4.1. Call the knowledge graph of the bulk grain terminal domain, enable the full-link data traceability capability, and extract the relevant data of the entire business link of the bulk grain terminal; Specifically, when the analysis starts, it first calls the knowledge graph of the bulk grain terminal domain built and stored by S1, enables its full-link data traceability capability, and extracts the related data of the entire business link of the bulk grain terminal.

[0099] The end-to-end traceability capability of knowledge graphs stems from their graph storage structure. Unlike traditional relational databases that require multiple table joins, graph databases use nodes and edges as native storage units, and data tracing across links can be completed through multi-hop traversal along relational edges.

[0100] In this step, the system starts the query from a cost object node (such as a single ship operation or a batch of goods), traces backward along the "aggregation" relationship edge to each operational link node, then extends along the "consumption" relationship edge to the resource element node, further extends along the "driving" relationship edge to the cost driver node, and simultaneously reconstructs the complete operational process sequence along the "connection" relationship edge. Through this traversal process, all operational links, resource consumption, and cost driver data associated with the cost object are extracted, forming a complete business link associated dataset. The extracted data includes attribute snapshots of each node and the type and attributes of the relationship edges, fully reproducing the traceability path from the cost object to the underlying drivers.

[0101] S4.2 Obtain the activity cost aggregation results; Specifically, while extracting the link-related data, the activity cost aggregation results for the cost object are obtained. The aggregation results originate from the actual cost data calculated by the financial system according to activity-based costing after the activity is completed. This includes the actual total cost of each activity stage, the unit activity cost, and detailed breakdowns of energy consumption costs, labor costs, equipment depreciation costs, and material costs by cost driver. The aggregation results are provided in structured record format, with each record corresponding to an activity stage node in the knowledge graph through a unique identifier, ensuring consistency between financial accounting data and graph business data at the entity level.

[0102] S4.3. Combine the data related to the entire business chain with the results of the collection of operating costs to determine the analytical dimensions of the bulk grain terminal's operational status; Specifically, by combining the full business chain correlation data extracted in S4.1 with the operational cost aggregation results obtained in S4.2, the specific dimensions of this operational status analysis are determined. The determination of the analysis dimensions is based on the terminal management needs and historical operational characteristics. In this embodiment, the following four dimensions are analyzed by default.

[0103] First, the cost dimension of the operational process. Taking the operational process as the unit of analysis, we compare the cost proportion of each operational process within the same cost object to identify the bottleneck process with the greatest cost consumption; Second, the cost driver dimension. Using cost drivers as the analysis unit, the actual contribution rate of each driver to the total cost is statistically analyzed to identify the core driving factors affecting costs under the current operating conditions; Third, the efficiency dimension of resource allocation. Taking resource allocation as the unit of analysis, the actual utilization rate of each equipment resource and human resource, the unit operating cost, and the deviation from the benchmark efficiency are calculated to evaluate the rationality of resource allocation; Fourth, the end-to-end trend dimension. Using time as the horizontal axis, track the total cost and unit cost changes of the same type of cost object over multiple consecutive operating cycles to determine the direction of operational efficiency evolution.

[0104] It is understandable that the analysis dimensions are not limited to the four items mentioned above, and can be flexibly added, reduced or adjusted according to the actual needs of the port.

[0105] S4.4. According to the established analysis dimensions, conduct an analysis of the operational status of the bulk grain terminal item by item to obtain the analysis content corresponding to each dimension; Specifically, according to the dimensions defined in S4.3, the graph query and path traversal capabilities of the graph are invoked one by one, and the actual cost values ​​in the aggregated results are combined to perform quantitative analysis on each dimension. The details are as follows: In terms of operational cost, the system traverses all operational nodes associated with the cost object along the "aggregation" relationship, reads the actual cost aggregation data of each stage, calculates the proportion of each stage's cost in the total cost, and sorts them in descending order of proportion. For stages with a cost proportion significantly higher than the benchmark level, the system drills down along the "consumption" relationship to the specific details of equipment, manpower, and energy consumption to pinpoint the specific reasons for the high cost. For example, if the cost proportion of a certain unloading operation is abnormally high, further drilling can be performed to find that it is due to increased electricity consumption costs, and then the system can trace the fluctuations in electricity price drivers or equipment efficiency drivers along the "driving" relationship.

[0106] At the cost driver level, the system aggregates the actual cost contribution values ​​of the same type of cost driver across various operational stages, calculating the contribution rate of each driver to the total cost across the entire chain. The actual contribution rate is compared with the baseline sensitivity coefficient stored in the knowledge graph to identify drivers whose influence in the current operation significantly deviates from the historical baseline. Drivers with large deviations are marked as critical drivers and given early warnings, suggesting focused attention or targeted control measures in subsequent scheduling.

[0107] In terms of resource scheduling efficiency, the system uses equipment resources and human resources nodes as the starting point for queries, reads the cumulative operating time, idle time, and failure time of each resource within the statistical period, and calculates the actual utilization rate. Simultaneously, it extracts the unit operating cost of each resource from the collected results and performs deviation analysis compared to the benchmark cost. For resources with low utilization rates or high unit costs, it analyzes the associated operational processes and cost drivers to determine whether the issue stems from an unreasonable scheduling scheme or performance degradation of the resource itself.

[0108] Under the end-to-end trend dimension, the system extracts the total cost and unit cost series of the same type of cost object over several recent periods and plots cost trend curves. Combining the workload, equipment status, and price driver data for each period, the system analyzes the driving factors of cost changes and determines whether the cost trend is seasonal fluctuation, a trend of increase, or an occasional anomaly, providing a basis for adjusting operational strategies.

[0109] S4.5 Integrate the analysis content from various dimensions to generate and output the operational analysis results of the bulk grain terminal.

[0110] Specifically, during the integration process, anomalies, key drivers, and efficiency bottlenecks identified across various dimensions are cross-referenced to form comprehensive operational diagnostic conclusions. For example, if the operational dimension reveals high costs in the unloading process, the cost driver dimension reveals an increased contribution rate of electricity price drivers, and the scheduling resource efficiency dimension reveals high utilization rates of unloading machines during peak electricity price periods, the cross-reference reveals that the concentrated scheduling of unloading operations during peak electricity price periods is the reason for the high costs. From this, optimization suggestions can be given to shift some adjustable operations to off-peak electricity periods.

[0111] In terms of output format, operational analysis results are provided simultaneously in two forms: visual reports and structured data. Visual reports, intended for terminal management and dispatchers, present analytical conclusions and key indicators across various dimensions in chart format, facilitating rapid understanding of operational status. Structured data is output in JSON or CSV format for further processing by downstream operational analysis systems or data warehouses. The output results are synchronously written back to the corresponding cost object node's analysis result attribute fields in the knowledge graph, forming a historical analysis archive. This provides data accumulation for subsequent operational comparative analysis and S3.1 benchmark parameter iteration.

[0112] like Figure 3As shown, this embodiment also provides a knowledge graph-based intelligent scheduling and operation analysis system for bulk grain terminals. The computer program of this system, when running, executes the steps of the aforementioned knowledge graph-based intelligent scheduling and operation analysis method for bulk grain terminals, including: Knowledge Graph Construction Unit 1 is used to construct a knowledge graph for the bulk grain terminal field by combining the cost aggregation logic of activity-based costing. Association Mapping Unit 2: Based on the entity association capability of the knowledge graph in the bulk grain terminal field, and combined with the operation cost driver analysis method, Association Mapping Unit 2 establishes the mapping relationship between the scheduling resources and operation cost elements of the bulk grain terminal. The multi-dimensional intelligent scheduling scheme generation unit 3, based on mapping relationships and real-time operational cost driver data, uses a knowledge graph dynamic path reasoning method driven by operational cost drivers to update only the local operational paths affected by changes in cost drivers, generating an intelligent scheduling scheme for bulk grain terminals that meets multi-dimensional optimization objectives. This unit includes an offline baseline parameter pre-configuration module, a real-time weight quantization calculation module, a dynamic weight adaptive correction module, and a local path incremental update module. The offline baseline parameter pre-configuration module iteratively updates the baseline parameters of operating cost drivers and the baseline cost of standard operating routes based on historical operating data of bulk grain terminals. The real-time weight quantification calculation module is used to combine real-time operation cost driver data to quantify the impact coefficient of each operation cost driver on the total operation cost. The dynamic weight adaptive correction module is used to adaptively adjust the weight ratio of each operation cost driver in the dynamic path reasoning of the knowledge graph based on the deviation between the real-time operating conditions and the baseline operating conditions. The local path incremental update module judges the degree of fluctuation of the operating cost driver based on the preset change threshold, and only performs incremental updates on the affected local operating paths to generate a smart scheduling scheme for bulk grain terminals that meets multi-dimensional optimization objectives. The multi-dimensional analysis unit 4 for operating costs is used to conduct multi-dimensional analysis of the operating status of bulk grain terminals by relying on the full-link data traceability capability of the knowledge graph in the bulk grain terminal field and combining the results of operation cost collection, and output the corresponding operation analysis results.

[0113] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely preferred examples and are not intended to limit the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the claimed invention.

Claims

1. A knowledge graph-based intelligent scheduling and operation analysis method for bulk grain terminals, characterized in that, include: S1. Construction of a knowledge graph for bulk grain terminals based on activity-based costing: Combining the cost aggregation logic of activity-based costing, a knowledge graph for bulk grain terminals is constructed. The knowledge graph for bulk grain terminals includes the operational links, resource elements, cost drivers, and the relationships between these elements. S2. Graph Relationship Mapping between Cost Drivers and Scheduling Resources: Based on the entity association capability of the knowledge graph in the bulk grain terminal field, and combined with the operational cost driver analysis method, establish the mapping relationship between the scheduling resources of bulk grain terminals and operational cost elements. S3. Generation of multi-dimensional intelligent scheduling scheme driven by operation cost drivers: Based on the mapping relationship built in S2 and real-time operation cost driver data, through the dynamic path reasoning method of knowledge graph driven by operation cost drivers, only the local operation paths affected by cost driver changes are updated to generate an intelligent scheduling scheme for bulk grain terminals that meets multi-dimensional optimization objectives. S4. Multidimensional analysis of operating costs based on knowledge graph traceability: Relying on the full-link data traceability capability of knowledge graph in the bulk grain terminal field, combined with the results of operation cost collection, the operation status of bulk grain terminals is analyzed in multiple dimensions and the corresponding operation analysis results are output.

2. The method for intelligent scheduling and operation analysis of bulk grain terminals based on knowledge graphs according to claim 1, characterized in that, In S1, the process of constructing a knowledge graph for the bulk grain terminal domain includes the following steps: S1.1 Based on the cost aggregation logic of activity-based costing, define a dedicated ontology architecture for the knowledge graph of bulk grain terminals. The ontology architecture is adapted to the business attributes and association rules of bulk grain terminal operation links, resource elements, and cost drivers. S1.2 Based on the defined ontology architecture, accurately extract the entities of bulk grain terminal operation links, resource elements, and cost drivers from the bulk grain terminal business data source. S1.3 Based on the cost aggregation logic of activity-based costing, explore the business relationships and cost coupling relationships between entities, and construct the relationship mapping links between entities; S1.

4. The extracted entities and the constructed relationships are integrated, standardized, and stored in a solidified manner to generate a complete knowledge graph of the bulk grain terminal domain.

3. The method for intelligent scheduling and operation analysis of bulk grain terminals based on knowledge graphs according to claim 1, characterized in that, In S2, the process of establishing the mapping relationship between bulk grain terminal scheduling resources and operating cost elements includes the following steps: S2.1 Based on the operational cost driver analysis method, sort out the attribute characteristics of the scheduling resources and operational cost elements of bulk grain terminals, as well as the corresponding rules between the two types of elements; S2.2 Utilize the entity association capability of the bulk grain terminal domain knowledge graph to determine the corresponding node positions of the bulk grain terminal scheduling resource entity and the operation cost element entity within the graph; S2.

3. According to the corresponding rules obtained from the analysis, establish the connection links between the corresponding nodes in the map to complete the initial mapping and binding of bulk grain terminal scheduling resources and operating cost elements; S2.

4. Perform logical consistency verification and correction on the established association links to solidify a stable mapping relationship.

4. The method for intelligent scheduling and operation analysis of bulk grain terminals based on knowledge graphs according to claim 1, characterized in that, In step S3, the process of generating an intelligent scheduling scheme for bulk grain terminals includes the following steps: S3.

1. Based on historical operational data of bulk grain terminals, iteratively update the benchmark parameters of operational cost drivers and the benchmark cost of standard operating routes; S3.

2. Combine real-time operation cost driver data to quantitatively calculate the impact coefficient of each operation cost driver on the total operation cost. S3.3 Based on the deviation between real-time operating conditions and baseline operating conditions, adaptively adjust the weight ratio of each operation cost driver in the dynamic path reasoning of the knowledge graph. S3.

4. Based on the preset change threshold, determine the degree of fluctuation of the operating cost driver, and only perform incremental updates on the affected local operating paths to generate a smart scheduling scheme for bulk grain terminals that meets multi-dimensional optimization objectives.

5. The method for intelligent scheduling and operation analysis of bulk grain terminals based on knowledge graphs according to claim 4, characterized in that, The process of iteratively updating the baseline parameters of the activity cost driver and the baseline cost of the standard activity path in S3.1 includes the following steps: S3.

11. Extract complete operational and cost data from the previous day's data warehouse at fixed time intervals each day; S3.

12. Based on the extracted data, statistically calculate the baseline parameters of each activity cost driver and the baseline cost of each standard activity path; S3.

13. Based on the historical relationship between the total cost of the entire operation and the cost value of individual operation cost drivers, calculate the benchmark sensitivity coefficient of each operation cost driver; S3.

14. Update the calculated benchmark parameters, benchmark costs, and benchmark sensitivity coefficients to the corresponding node attributes in the bulk grain terminal domain knowledge graph; S3.

15. Pre-calculate and store the baseline total cost of all standard operating procedures as the basis for online inference.

6. The intelligent scheduling and operation analysis method for bulk grain terminals based on knowledge graphs according to claim 4, characterized in that, The process of quantitatively calculating the impact coefficient of each activity cost driver on the total activity cost in S3.2 includes the following steps: S3.

21. Obtain real-time data on the cost drivers of each operational stage at the bulk grain terminal; S3.22 Determine the sensitivity coefficient of activity-based cost drivers The activity cost driver sensitivity coefficient Indicates the first In the first stage of bulk grain terminal operations The extent to which each activity cost driver affects the total cost of the entire activity chain; S3.

23. Based on the relationship between the total cost of the entire operation and the cost value of individual operation cost drivers, calculate the sensitivity coefficient corresponding to each operation cost driver. ; S3.24, Calculate the sensitivity coefficient As the impact coefficient of this operation cost driver on the total operation cost; S3.

25. Update the calculated influence coefficients to the corresponding node attributes in the knowledge graph of the bulk grain terminal domain.

7. The method for intelligent scheduling and operation analysis of bulk grain terminals based on knowledge graphs according to claim 4, characterized in that, The adaptive adjustment of the weight ratio of each activity cost driver in the dynamic path reasoning of the knowledge graph in S3.3 includes the following steps: S3.31, The activity cost driver sensitivity coefficient calculated in real time. The sensitivity coefficient is compared with the benchmark sensitivity coefficient stored in the knowledge graph of the bulk grain terminal field, and the sensitivity coefficient deviation value is calculated. S3.32 Determine the dynamic weighting factor The dynamic weighting factor express Time of the first In the first stage of bulk grain terminal operations The weighting of each activity cost driver in path reasoning; S3.

33. Based on the deviation between the baseline weight and the sensitivity coefficient, calculate the dynamic weight factor at the current moment according to the preset linear adjustment rule. ; S3.34, Based on the calculated dynamic weighting factor Update the weight ratio of each activity cost driver in the dynamic path reasoning of the knowledge graph; S3.

35. Store the updated weight percentages in the node attributes corresponding to the bulk grain terminal domain knowledge graph.

8. The intelligent scheduling and operation analysis method for bulk grain terminals based on knowledge graphs according to claim 4, characterized in that, The process of generating an intelligent scheduling scheme for bulk grain terminals that satisfies multi-dimensional optimization objectives in S3.4 includes the following steps: S3.

41. Compare the real-time operational cost driver data with the benchmark parameters stored in the bulk grain terminal domain knowledge graph, and calculate the rate of change of each operational cost driver. S3.42 Determine whether the rate of change of the activity cost driver exceeds the preset change threshold; S3.43 If the rate of change exceeds the threshold, locate all operation paths within the preset association range centered on the operation cost driver node in the knowledge graph of the bulk grain terminal domain. S3.44, Based on the adjusted dynamic weighting factor Recalculate the total cost of the work paths within the preset associated range, and retain the original calculation results of the unaffected work paths; S3.

45. Sort all operation paths according to multi-dimensional optimization objectives to generate a smart scheduling scheme for bulk grain terminals that meets the multi-dimensional optimization objectives.

9. The method for intelligent scheduling and operation analysis of bulk grain terminals based on knowledge graphs according to claim 1, characterized in that, In S4, the process of performing a multidimensional analysis of operational costs based on map tracing includes the following steps: S4.

1. Call the knowledge graph of the bulk grain terminal domain, enable the full-link data traceability capability, and extract the relevant data of the entire business link of the bulk grain terminal; S4.2 Obtain the activity cost aggregation results; S4.

3. Combine the data related to the entire business chain with the results of the collection of operating costs to determine the analytical dimensions of the bulk grain terminal's operational status; S4.

4. According to the established analysis dimensions, conduct an analysis of the operational status of the bulk grain terminal item by item to obtain the analysis content corresponding to each dimension; S4.5 Integrate the analysis content from various dimensions to generate and output the operational analysis results of the bulk grain terminal.

10. A knowledge graph-based intelligent scheduling and operation analysis system for bulk grain terminals, characterized in that: When the computer program of the knowledge graph-based intelligent scheduling and operation analysis system for bulk grain terminals is executed, it performs the steps of the knowledge graph-based intelligent scheduling and operation analysis method for bulk grain terminals as described in any one of claims 1-9, including: Knowledge graph construction unit (1), which is used to construct a knowledge graph in the bulk grain terminal field by combining the cost collection logic of activity costing method; The association mapping unit (2) relies on the entity association capability of the knowledge graph in the bulk grain terminal field and combines the operation cost driver analysis method to establish the mapping relationship between the scheduling resources and operation cost elements of the bulk grain terminal; The multi-dimensional intelligent scheduling scheme generation unit (3) generates a multi-dimensional intelligent scheduling scheme for bulk grain terminals based on mapping relationships and real-time operation cost driver data. It uses a knowledge graph dynamic path reasoning method driven by operation cost drivers to update only the local operation paths affected by changes in cost drivers, thus generating a multi-dimensional optimization objective. The unit includes an offline baseline parameter pre-configuration module, a real-time weight quantization calculation module, a dynamic weight adaptive correction module, and a local path incremental update module. The offline benchmark parameter pre-configuration module iteratively updates the benchmark parameters of the operating cost drivers and the benchmark cost of the standard operating path based on historical operating data of the bulk grain terminal. The real-time weight quantification calculation module is used to combine real-time operation cost driver data to quantify and calculate the impact coefficient of each operation cost driver on the total operation cost. The dynamic weight adaptive correction module is used to adaptively adjust the weight ratio of each operation cost driver in the dynamic path reasoning of the knowledge graph based on the deviation between the real-time operating conditions and the baseline operating conditions. The local path incremental update module judges the degree of fluctuation of the operating cost driver based on a preset change threshold, and only performs incremental updates on the affected local operating paths to generate a smart scheduling scheme for bulk grain terminals that meets multi-dimensional optimization objectives. The multidimensional analysis unit (4) for operating costs is used to conduct multidimensional analysis of the operating status of bulk grain terminals based on the full-link data traceability capability of the knowledge graph in the bulk grain terminal field and in combination with the results of operation cost collection, and output the corresponding operation analysis results.

Images