An intelligent management system and method for refined oil transaction based on knowledge graph

Abstract

The application discloses a kind of intelligent management system and method of product oil transaction fusing knowledge graph, it is related to the technical field of intelligent management, including the following steps, step S100, based on nonlinear fork type bifurcation model, the stable boundary of entity association is obtained and entity is filtered, step S200, construct heterogeneous graph convolution network, based on graph query language and heterogeneous graph convolution network, obtain knowledge spectrum diagram, step S300, based on knowledge spectrum diagram identification current abnormal node and trace back first transaction path.The application is by introducing nonlinear dynamics method to process the stability screening of transaction entity, eliminate risk entity from the root, realize the abnormal tracing based on graph reasoning, provide the complete transaction path of explainable, upgrade knowledge graph from static knowledge base to dynamic analysis engine, promote product oil transaction management from experience-driven to data intelligent-driven, have significant technical advantages and application value in improving regulatory efficiency, preventing and controlling transaction risk, supporting tax compliance and the like.

Description

Technical Field

[0001] This invention relates to the technical field of intelligent management, and in particular to an intelligent management system and method for refined oil trading that integrates knowledge graphs. Background Technology

[0002] In recent years, refined oil transaction management technology has developed rapidly, covering the entire chain of wholesale, warehousing, transportation and retail, and integrating AI and quantum encryption technologies to create a smart supervision and transaction ecosystem. Through technologies such as privacy computing and blockchain, it has built a usable but invisible, controllable and measurable data circulation mechanism to support broader data sharing and value co-creation. Consumers can view information such as refueling volume, unit price and oil grade by scanning a code, solving the problem of having no evidence for complaints.

[0003] Currently, Chinese invention patent CN120729647A discloses a method, server, and system for processing oil and gas data assets based on a blockchain system. This method responds to transaction requests sent by business server in the blockchain system, verifies the business server initiating the transaction request, wherein the transaction request indicates the need to acquire oil and gas data assets. If the business server initiating the transaction request is verified, the encrypted oil and gas data assets indicated by the transaction request are obtained from the encrypted oil and gas data assets uploaded by different business servers, and the obtained encrypted oil and gas data assets are sent to the business server initiating the transaction request. However, the related technology manages and processes data based on transaction-related data, which can easily lead to redundant data classification and unclear semantics, which is not conducive to the rapid response and accuracy of large-scale data management and has certain limitations. Summary of the Invention

[0004] The technical problem solved by this invention is that in related technologies, data management based on transaction-related data can easily lead to redundant data classification and unclear semantics, which is not conducive to the rapid response and accuracy of large-scale data management and has certain limitations.

[0005] To solve the above technical problems, the present invention provides the following technical solution. In the first aspect, a method for intelligent management of refined oil transactions integrating knowledge graphs is provided, including the following steps: Step S100, based on historical transaction data, configuring entities and entity associations, configuring corresponding constraints for entity associations, obtaining stable boundaries of entity associations based on a nonlinear fork-type bifurcation model, and filtering entities according to the feature quantities of entity associations and corresponding constraints. Step S200: Set the filtered entities and entity associations as nodes, edges, node types and edge types. Based on nodes, edges, node types and edge types, construct a heterogeneous graph convolutional network. Based on graph query language and heterogeneous graph convolutional network, obtain a knowledge graph. Step S300: Identify the current abnormal node based on the knowledge graph, and trace the first transaction path based on the relationship of the current abnormal node.

[0006] As a preferred embodiment of the intelligent management method for refined oil transactions that integrates knowledge graphs as described in this invention, the entities include spatial entities and temporal entities; Entity associations include first entity associations, second entity associations, and third entity associations; The constraints include the first constraint, the second constraint, and the third constraint. Based on the transaction parties' IDs and GIS technology, the regions to which the transaction parties belong are obtained. Spatial entities are constructed according to the respective regions of the transaction parties. Time entities are constructed based on the transaction timestamp. First entity associations between time entities are constructed based on the transaction category and the average unit price of the transaction. Second entity associations between spatial entities are constructed based on the batch number of transported oil and the number of transported vehicles. Third entity associations between time entities and spatial entities are constructed based on the first entity association and the second entity association.

[0007] As a preferred embodiment of the intelligent management method for refined oil transactions that integrates knowledge graphs as described in this invention, the method for constructing spatial entities based on the transaction party IDs includes constructing time entities based on the transaction timestamps of the spatial entities.

[0008] As a preferred embodiment of the intelligent management method for refined oil transactions that integrates knowledge graphs as described in this invention, the method for constructing the first entity association includes determining whether the transaction category numbers of any two time entities are the same. When the transaction category numbers are the same, calculate the first difference in the average unit price of the transactions, select the absolute value of the first difference, represent the absolute value of the first difference as the third code, convert the third code into the first vector, and record the first vector as the first entity association; When the transaction category numbers are different, obtain the standard transaction volume corresponding to the transaction category, calculate the first ratio of the standard transaction volumes corresponding to the two time entities, calculate the first difference of the average transaction unit price, select the absolute value of the first difference, calculate the first quotient of the absolute value of the first difference and the first ratio, represent the first quotient as the third code, convert the third code into the first vector, and record the first vector as the first entity association; The construction method for the second entity association includes determining whether the batch numbers of oil transported by any two spatial entities are the same; When the batch numbers of transported oil products are the same, calculate the second difference in the transportation distances corresponding to these two spatial entities, select the absolute value of the second difference, represent the absolute value of the second difference as the fourth code, convert the fourth code into the second vector, and record the second vector as the second entity association; When the batch numbers of transported oil products are different, obtain the standard inspection pass rate corresponding to the refinery number, calculate the second ratio of the standard inspection pass rates corresponding to the two spatial entities, calculate the second difference of the transportation distance corresponding to the two spatial entities, select the absolute value of the second difference, calculate the second quotient of the absolute value of the second difference and the second ratio, represent the second quotient as the fourth code, convert the fourth code into the second vector, and record the second vector as the second entity association; The methods for constructing third entity associations include, Obtain the first vector and the second vector, calculate the vector sum based on the first vector and the second vector, and denote the vector sum as the third vector; The third vector is represented as a third entity association.

[0009] As a preferred embodiment of the intelligent management method for refined oil transactions that integrates knowledge graphs as described in this invention, the average value of the absolute value of the first difference is calculated, and the average value of the absolute value of the first difference is set as the standard value of the absolute value of the first difference. The average value of the first quotient is calculated and the average value of the first quotient is set as the standard value of the first quotient. Calculate the absolute value of any first difference and the third difference between it and the corresponding standard value, or calculate the fourth difference between any first quotient and the corresponding standard value; Set the first value as the third difference threshold and the second value as the fourth difference threshold; The first constraint is set as follows: when the third difference is greater than the first value, or when the fourth difference is greater than the second value, the original first entity association is retained; when the third difference is less than or equal to the first value, or when the fourth difference is less than or equal to the second value, the corresponding first entity association is set to a unit vector with the same direction as the original first vector. Calculate the average of the absolute values ​​of the second difference, set the average of the absolute values ​​of the second difference as the standard value of the absolute value of the second difference, calculate the average of the second quotient, and set the average of the second quotient as the standard value of the second quotient; Calculate the fifth difference between the absolute value of any second difference and its corresponding standard value, or calculate the sixth difference between any second quotient and its corresponding standard value; Set the third value as the third difference threshold and the fourth value as the fourth difference threshold; The second constraint is set as follows: when the fifth difference is greater than the third value, or the sixth difference is greater than the fourth value, the original second entity association is retained; when the fifth difference is less than or equal to the third value, or the sixth difference is less than or equal to the fourth value, the corresponding second entity association is set to a unit vector with the same direction as the original first vector.

[0010] As a preferred embodiment of the intelligent management method for refined oil transactions that integrates knowledge graphs as described in this invention, the stable boundary of entity association includes the stable boundary of the first entity association and the stable boundary of the second entity association. The method for constructing the stable boundary of entity association includes retrieving a new set of historical transaction data within the first historical time period and constraining the set of entity associations according to the constraints corresponding to the entity associations. Calculate new standard values ​​for entity associations after new historical transaction data is available; Calculate the balance difference based on the new standard value, set the fifth value as the balance threshold, calculate the difference between the balance difference and the fifth value, and record it as the seventh difference. Based on the nonlinear bifurcation model, the seventh difference, and the order of historical acquisition times within the first historical time period, the first expression is obtained; The imbalance points are obtained based on the value of the derivative of the first expression. The stability boundary of entity association is set based on the comparison between the number of imbalance points and 2.

[0011] As a preferred embodiment of the intelligent management method for refined oil transactions that integrates knowledge graphs as described in this invention, the filtered entities and entity associations are respectively set as nodes, edges, node types, and edge types; The method for constructing heterogeneous graph convolutional networks includes setting initial feature vectors for each node and setting initial weight matrices for each edge; For any given node, set other nodes connected to it via edges as its neighbors; The initial features of neighboring nodes are linearly transformed using an activation function. Based on the convolution operation, the initial feature values ​​of the neighboring nodes after linear transformation are convolved to obtain a new representation of the feature values ​​of the node. The iteration count of the heterogeneous graph convolutional network is set to the first number. When the first number is reached or the vector sum of the new feature quantities of the neighboring nodes and the node is the same as the corresponding entity, the final representation of the feature quantities of each node is output. At this time, the construction process of the knowledge graph is completed.

[0012] As a preferred embodiment of the intelligent management method for refined oil transactions that integrates knowledge graphs as described in this invention, the query target is obtained based on graph query language; The methods for obtaining the query target include constructing a meta-path based on the edges associated with the third entity. The meta-path is used to locate the abnormal transaction cluster under the current spatiotemporal conditions. Configure constraints for each metapath, retain metapaths that satisfy the constraints, and set index numbers for the retained metapaths.

[0013] As a preferred embodiment of the intelligent management method for refined oil transactions that integrates knowledge graphs as described in this invention, a mapping relationship between meta-paths and graph query language is established. This mapping is obtained through Cypher extended syntax; Based on the meta-path, trace the path corresponding to the abnormal node and set the traced path as the first transaction path.

[0014] Secondly, a refined oil trading intelligent management system integrating knowledge graphs includes a screening module, a construction module, and an identification module; The filtering module configures entities and entity associations based on historical transaction data, configures corresponding constraints for entity associations, obtains the stable boundary of entity associations based on a nonlinear fork-type bifurcation model, and filters entities according to the feature quantities of entity associations and the corresponding constraints. The construction module sets the filtered entities and entity associations as nodes, edges, node types, and edge types. Based on the nodes, edges, node types, and edge types, it constructs a heterogeneous graph convolutional network. Based on the graph query language and the heterogeneous graph convolutional network, it obtains a knowledge graph. The identification module identifies the current abnormal node based on the knowledge graph and traces the first transaction path based on the relationship between the current abnormal node.

[0015] The beneficial effects of this invention are as follows: It introduces a nonlinear dynamics method to process the stability screening of transaction entities, eliminating risky entities at the source; it uses heterogeneous graph learning technology to construct a highly expressive knowledge graph, preserving the complex structure and semantic information of the transaction network, realizing anomaly tracing based on graph reasoning, breaking through the limitations of traditional single-point detection, providing an interpretable complete transaction path, upgrading the knowledge graph from a static knowledge base to a dynamic analysis engine, and promoting the management of refined oil transactions from experience-driven to data intelligence-driven. It has significant technical advantages and application value in improving regulatory efficiency, preventing and controlling transaction risks, and supporting tax compliance. Attached Figure Description

[0016] Figure 1 This is a schematic diagram of the basic process of a refined oil trading intelligent management method that integrates knowledge graphs, provided as an embodiment of the present invention. Detailed Implementation

[0017] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.

[0018] It should be understood that the step numbers used herein are for ease of description only and are not intended to limit the order in which the steps are performed. It should also be understood that the terminology used in this specification is for the purpose of describing specific embodiments only and is not intended to limit the invention.

[0019] As used in this specification and the appended claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise.

[0020] The terms “comprising” and “including” indicate the presence of the described feature, whole, step, operation, element and / or component, but do not exclude the presence or addition of one or more other features, wholes, steps, operations, elements, components and / or collections thereof.

[0021] The term “and / or” refers to any combination of one or more of the associated listed items, as well as all possible combinations, and includes these combinations.

[0022] Example, refer to Figure 1 As an embodiment of the present invention, a method for intelligent management of refined oil transactions integrating knowledge graphs is provided, including the following steps: Step S100, based on historical transaction data, configure entities and entity associations, configure corresponding constraints for entity associations, obtain the stable boundary of entity associations based on a nonlinear fork-type bifurcation model, and filter entities according to the feature quantities of entity associations and corresponding constraints. Step S200: Set the filtered entities and entity associations as nodes, edges, node types and edge types. Based on nodes, edges, node types and edge types, construct a heterogeneous graph convolutional network. Based on graph query language and heterogeneous graph convolutional network, obtain a knowledge graph. Step S300: Identify the current abnormal node based on the knowledge graph, and trace the first transaction path based on the relationship of the current abnormal node.

[0023] More preferably, this invention introduces a nonlinear dynamics method to process the stability screening of transaction entities, eliminating risky entities at the source. It uses heterogeneous graph learning technology to construct a highly expressive knowledge graph, preserving the complex structure and semantic information of the transaction network, realizing anomaly tracing based on graph reasoning, breaking through the limitations of traditional single-point detection, providing an interpretable complete transaction path, and upgrading the knowledge graph from a static knowledge base to a dynamic analysis engine. This promotes the transformation of refined oil transaction management from experience-driven to data-intelligent driven, and has significant technical advantages and application value in improving regulatory efficiency, preventing transaction risks, and supporting tax compliance.

[0024] Entities include spatial entities and temporal entities; Entity associations include first entity associations, second entity associations, and third entity associations; The constraints include the first constraint, the second constraint, and the third constraint. Step S100 includes the following sub-steps: Step S101, extract historical transaction data from the refined oil transaction database; Historical transaction data includes the transaction parties' IDs, transport vehicle IDs, transaction category, transaction timestamp, transported oil batch number, and average transaction unit price.

[0025] More preferably, the transaction parties' numbers include the seller's number and the buyer's number; Transaction categories include refinery sales, wholesale distribution, gas station retail, inter-regional transfers, and emergency reserve allocation; The transaction timestamp includes the transaction year, transaction month, and transaction day; By identifying the transport vehicle's number, the license plate number, the transport company number, the transport train number, and the transport route can be obtained; The batch number of transported oil products represents the production batch number of the corresponding oil products at the production party when the transaction was conducted by the two parties at the corresponding transaction timestamp. Through the production batch number, the refinery number, production line number, feeding timestamp, and quality inspection report number can be identified. The average unit price of a transaction is expressed as the ratio of the total cost of oil and transportation paid by the buyer at the corresponding transaction timestamp to the weight of the oil. The unit of oil weight is kilograms. Step S102: Based on the transaction parties' ID numbers and GIS technology, obtain the regions to which the transaction parties belong, and construct spatial entities according to the respective regions to which the transaction parties belong; More preferably, the region is represented as a sequential combination of the provincial administrative division code, the municipal administrative division code, and the county administrative division code.

[0026] Step S103: Construct a time entity based on the transaction timestamp; Step S104: Based on the transaction category and the average transaction price, construct the first entity association between time entities and time entities; based on the transport oil batch number and the transport vehicle number, construct the second entity association between spatial entities and spatial entities; based on the first entity association and the second entity association, construct the third entity association between time entities and spatial entities. Step S105: Set the first constraint condition for the association of the first entity, the second constraint condition for the association of the second entity, and the third constraint condition for the association of the third entity.

[0027] The method for constructing a spatial entity includes obtaining any set of transaction party IDs, sequentially combining the region of the seller and the region of the buyer to obtain a first code, which is represented as a 12-character code. Iterate through the transaction parties' numbers in each group to obtain the first code corresponding to each group's transaction parties' numbers; Set each first code as a spatial entity; During the construction of spatial entities, they are automatically encoded according to the construction sequence. The encoding of a spatial entity is represented as follows: , where m is a natural number, representing the construction sequence number of the spatial entity.

[0028] More preferably, when the first code is repeated, the first code is set as an independent spatial entity; when the first code is not repeated, the first code is set as an independent spatial entity, that is, the spatial entities of different transactions are different.

[0029] The method for constructing time entities includes unifying the code lengths of the transaction year, transaction month, and transaction day in the transaction timestamps of each spatial transaction entity. The unification method is achieved by padding the transaction year, transaction month, and transaction day with zeros. Through the unification operation, the second code corresponding to each spatial transaction entity is obtained. Set the second code as a time entity; During the construction of time entities, they are automatically encoded according to the construction order. The encoding of a time entity is represented as follows: , where n is a natural number, representing the construction sequence number of the time entity.

[0030] More preferably, the expected value for the code length of the transaction year is 4, the expected value for the code length of the transaction month is 2, and the expected value for the code length of the transaction day is 2, until the code length of the transaction year, the code length of the transaction month, and the code length of the transaction day after the zero-padding operation reaches the corresponding expected value, then the zero-padding operation stops. Obtain the transaction timestamps of each spatial entity, unify the code length of the transaction year of each transaction timestamp to 4 characters, the code length of the transaction month of each transaction timestamp to 2 characters, and the code length of the transaction day of each transaction timestamp to 2 characters. After unifying the code lengths of the transaction year, transaction month, and transaction day, a second code corresponding to each spatial entity is obtained. The second code is represented as an 8-character code.

[0031] More preferably, when the second code is repeated, the second code is set as an independent spatial entity; when the second code is not repeated, the second code is set as an independent spatial entity, that is, the time entities of different transactions are different.

[0032] The method for constructing the first entity association includes numbering the transaction categories, where the transaction category number is represented as X. i where i is a natural number; Get the transaction category number and average transaction price for any two time entities; Determine if the transaction category numbers of these two time entities are the same; When the transaction category numbers are the same, calculate the first difference in the average unit price of the transactions, select the absolute value of the first difference, represent the absolute value of the first difference as the third code, convert the third code into the first vector, and record the first vector as the first entity association; When the transaction category numbers are different, obtain the standard transaction volume corresponding to the transaction category, calculate the first ratio of the standard transaction volumes corresponding to the two time entities, calculate the first difference of the average transaction price, select the absolute value of the first difference, calculate the first quotient of the absolute value of the first difference and the first ratio, represent the first quotient as the third code, convert the third code into the first vector, and record the first vector as the first entity association.

[0033] More preferably, the standard transaction volume is expressed as the average weight of oil products from all historical transactions corresponding to the transaction category; The first ratio is greater than 1, that is, the first ratio is the ratio between the standard trading volume with the larger value and the standard trading volume with the smaller value in the two time entities being compared. The rationale for calculating the first ratio is explained as follows: since the absolute value of the first difference is chosen, it indicates that the order of subtraction between the two time entities is not specified; it only indicates the difference in transaction volume between the two transactions. Therefore, by calculating the first ratio, it only indicates the ratio between the transaction volumes corresponding to the two time entities, that is, the multiple of the influence on the difference in transaction volume. Dividing the absolute value of the first difference with the first ratio effectively weakens the influence of the self-quantification of time entities of different transaction types on the first vector. That is, when the unit price of oil is high, the transaction volume of oil is limited due to the transaction type, making the quantitative value between unit price and transaction volume inaccurate. This makes the description of the relationship between heterogeneous time entities more unified and comparable.

[0034] More preferably, the third encoding is represented as binary encoding; The method of converting binary codes into vectors is an existing technology and will not be elaborated here.

[0035] The construction method for the second entity association includes obtaining the transport oil batch number corresponding to any two spatial entities and determining whether the transport oil batch number is the same. When the batch numbers of transported oil products are the same, extract the transport trajectory of the transport vehicle numbers of the two spatial entities, extract the transport distance corresponding to the transport trajectory, calculate the second difference between the transport distances corresponding to the two spatial entities, select the absolute value of the second difference, represent the absolute value of the second difference as the fourth code, convert the fourth code into the second vector, and record the second vector as the second entity association; When the batch numbers of transported oil products are different, obtain the refinery number corresponding to the batch numbers of transported oil products of the two spatial entities, obtain the standard inspection pass rate corresponding to the refinery number, calculate the second ratio of the standard inspection pass rates corresponding to the two spatial entities, calculate the second difference of the transport distances corresponding to the two spatial entities, select the absolute value of the second difference, calculate the second quotient of the absolute value of the second difference and the second ratio, represent the second quotient as the fourth code, convert the fourth code into the second vector, and record the second vector as the second entity association.

[0036] More preferably, the standard test pass rate is expressed as the average pass rate among the samples of oil products randomly tested by the testing party in each historical period corresponding to the refinery number; The second ratio is greater than 1, that is, the second ratio is the ratio between the standard test pass rate with the larger value and the standard test pass rate with the smaller value in the two spatial entities being compared. The rationale for calculating the second ratio is explained as follows: since the absolute value of the second difference is selected, it indicates that the order of subtraction between the two spatial entities is not specified. It only indicates the difference in transportation distance between the two transactions. Therefore, by calculating the second ratio, it only indicates the ratio between the refinery production quality corresponding to the two spatial entities, that is, the multiple of the influence on the difference in transportation distance. The division between the absolute value of the second difference and the second ratio effectively weakens the influence of the self-quantification of spatial entities of different refineries on the second vector. That is, when the refinery pass rate is low, it will lead to an increase in the amount of oil products that fail the test, thereby generating an increment in the transportation trajectory, including the inspection distance, making the description of the relationship between heterogeneous spatial entities more unified and comparable.

[0037] The fourth encoding is represented as binary encoding; The method of converting binary codes into vectors is an existing technology and will not be elaborated here.

[0038] The methods for constructing third entity associations include, Obtain the first vector and the second vector, calculate the vector sum based on the first vector and the second vector, and denote the vector sum as the third vector; The third vector is represented as a third entity association; The third entity association is represented as the entity association between any one of the two temporal entities corresponding to the first vector and any one of the two spatial entities corresponding to the second vector. For vector sums, a calculation method is configured, which includes setting the spatial entity corresponding to the starting point of the first vector as the first spatial entity and setting the spatial entity corresponding to the ending point of the first vector as the second spatial entity. Set the time entity corresponding to the starting point of the second vector as the first time entity, and set the time entity corresponding to the ending point of the first vector as the second time entity; When calculating the vector sum of the first spatial entity and the first temporal entity, the vector sum of the first vector and the second vector is calculated. When calculating the vector sum of the first spatial entity and the second temporal entity, the opposite vector of the second vector is obtained, and the vector sum of the opposite vectors of the first vector and the second vector is calculated. When calculating the vector sum of the second spatial entity and the first temporal entity, the opposite vector of the first vector is obtained, and the vector sum of the opposite vector of the first vector and the second vector is calculated. When calculating the vector sum of the second spatial entity and the second temporal entity, the opposite vector of the first vector and the opposite vector of the second vector are obtained, and the vector sum of the opposite vectors of the first vector and the opposite vectors of the second vector is calculated.

[0039] More preferably, by using the first vector, the second vector, the directionality of the first vector, and the directionality of the second vector, a third vector can be obtained between any temporal entity and any spatial vector. The combination of the three vectors not only achieves precise coupling in the spatiotemporal dimension, but also establishes a quantifiable, traceable, and verifiable dynamic association model between heterogeneous entities. The directional design enables bidirectional mapping between temporal evolution and spatial displacement, thereby supporting the collaborative analysis of quality fluctuations, transportation disturbances, and detection responses in the entire refining chain. Moreover, the entity types differ from previous knowledge graphs, which set nodes based on transaction IDs (i.e., a new node type is generated with each transaction). Instead, the spatiotemporal ontology is used as a unified anchor point to map multi-source heterogeneous entities such as equipment, processes, batches, and detection items to the same coordinate system, ensuring the semantic stability of nodes and the self-consistency of relational logic. This avoids graph redundancy and evolutionary distortion caused by transaction frequency, significantly improving the efficiency of anomaly tracing and inference accuracy.

[0040] The method for setting the first constraint includes obtaining the absolute value of the first difference corresponding to the first entity association or the first quotient corresponding to the first entity association; Calculate the average of the absolute values ​​of the first differences, set the average of the absolute values ​​of the first differences as the standard value of the absolute values ​​of the first differences, and calculate the average of the first quotients, set the average of the first quotients as the standard value of the first quotients. Calculate the absolute value of any first difference and the third difference between it and the corresponding standard value, or calculate the fourth difference between any first quotient and the corresponding standard value; Set the first value as the third difference threshold and the second value as the fourth difference threshold; The first constraint is set as follows: when the third difference is greater than the first value, or when the fourth difference is greater than the second value, the original first entity association is retained; when the third difference is less than or equal to the first value, or when the fourth difference is less than or equal to the second value, the corresponding first entity association is set to a unit vector with the same direction as the original first vector.

[0041] More preferably, by retaining or transforming the first vector into a unit vector in the same direction through the first constraint condition, noise interference can be effectively suppressed, the robustness of the vector representation can be enhanced, and the time-homogenized first vector can be avoided from having a side effect on the accuracy of the knowledge graph.

[0042] The method for setting the second constraint includes obtaining the absolute value of the second difference corresponding to the second entity association or the second quotient corresponding to the second entity association; Calculate the average of the absolute values ​​of the second difference, set the average of the absolute values ​​of the second difference as the standard value of the absolute value of the second difference, calculate the average of the second quotient, and set the average of the second quotient as the standard value of the second quotient; Calculate the fifth difference between the absolute value of any second difference and its corresponding standard value, or calculate the sixth difference between any second quotient and its corresponding standard value; Set the third value as the third difference threshold and the fourth value as the fourth difference threshold; The second constraint is set as follows: when the fifth difference is greater than the third value, or the sixth difference is greater than the fourth value, the original second entity association is retained; when the fifth difference is less than or equal to the third value, or the sixth difference is less than or equal to the fourth value, the corresponding second entity association is set to a unit vector with the same direction as the original first vector.

[0043] More preferably, by using the second constraint condition to retain or transform the second vector into a unit vector in the same direction, noise interference can be effectively suppressed, the robustness of the vector representation can be enhanced, and the spatially homogenized second vector can be avoided from having a side effect on the accuracy of the knowledge graph.

[0044] The method for setting the third constraint includes: when the first constraint sets any first vector to a unit vector with the same direction as the original first vector, or when the second constraint sets any second vector to a unit vector with the same direction as the original second vector, jumping to the corresponding third entity association, and reconstructing the third entity association according to the first vector constrained by the first constraint or the second vector constrained by the second constraint, and the vector sum calculation method, to obtain a new third entity association, and replacing the original third entity association with the new third entity association.

[0045] More preferably, the third constraint condition dynamically reconstructs the third entity association, calibrating the cross-modal alignment deviation while preserving the semantic topology. The third constraint condition is dynamically updated through the first and second constraints, forming a closed-loop constraint mechanism to ensure that the knowledge graph maintains semantic consistency in temporal evolution and spatial mapping.

[0046] The stable boundary of entity association includes the stable boundary of the first entity association and the stable boundary of the second entity association. More preferably, the method for constructing the stable boundary of the first entity association is the same as the method for constructing the stable boundary of the second entity association, and is collectively referred to as the method for constructing the stable boundary of the entity association.

[0047] The method for constructing the stable boundary of entity association includes retrieving a new set of historical transaction data within the first historical time period, and obtaining the entity association corresponding to the new set of historical transaction data based on the method for constructing entity association. And constrain the group of entity associations according to the constraints corresponding to the entity associations; Obtain the standard value corresponding to the original entity association, and calculate the new standard value of the entity association after adding constraints based on the calculation method of the standard value; The method for adding new standard values ​​is to continuously add new standard values ​​to the original standard values ​​according to the time sequence of the historical collection moments within the first historical time period, so as to continuously obtain each standard value value sequence until jumping to all the historical collection moments within the first historical time period. At this point, N standard value value sequences are constructed, and each standard value value sequence includes the original standard value and the corresponding newly added standard value. Calculate the difference between the new standard value and the original standard value, and record it as the balance difference. Set the fifth value as the balance threshold, and calculate the difference between the balance difference and the fifth value, and record it as the seventh difference. The nonlinear fork-shaped bifurcation model is retrieved, with the seventh difference as the vertical axis and the order of historical acquisition times within the first historical time period corresponding to the seventh difference as the horizontal axis. The seventh difference and the order of historical acquisition times within the first historical time period are substituted into the nonlinear fork-shaped bifurcation model to obtain the first expression. Differentiate the first expression to obtain its derivative function. Obtain the imbalance points based on the value of the derivative function. Set the stable boundary of entity association based on the comparison between the number of imbalance points and 2.

[0048] More preferably, the first expression is, ; in, It is random Gaussian white noise. It is a constant, usually The value is 0.05. This represents the order of historical data collection times within the first historical time period corresponding to the seventh difference. The seventh difference, is the differential symbol, and S is the value of the nonlinear bifurcation model; The derivative of the first expression is, ; in, This can be expressed as taking the partial derivative with respect to Z. is the value of the derivative of the first expression.

[0049] Setting the derivative of the first expression to 0, we obtain the equilibrium and unequilibrium points. The equilibrium point is denoted as... The imbalance point is denoted as ; Count the number of imbalance points; When the number of imbalance points is 2, the fifth value is continuously decreased until the number of imbalance points is less than 2. Then, the fifth value at this time is output and set as the standard balance value. When the number of imbalance points is greater than 2, the fifth value is continuously increased until the number of imbalance points is less than 2. Then, the fifth value at this time is output and set as the standard balance value. When the number of unbalanced points is less than 2, the fifth value is set as the standard balance value; Calculate the difference between the original standard value and the standard equilibrium value, set the difference between the original standard value and the standard equilibrium value as the first lower limit, calculate the sum of the first lower limit and the corresponding standard value, and set the sum of the first lower limit and the corresponding standard value as the lower limit of the stability boundary of entity association; Calculate the sum of the original standard value and the standard equilibrium value, set the sum of the original standard value and the standard equilibrium value as the first upper limit, calculate the sum of the first upper limit and the corresponding standard value, and set the sum of the first upper limit and the corresponding standard value as the upper limit of the stability boundary of entity association; Based on the lower bound and the upper bound of the stable boundary of entity association, construct a stable boundary corresponding to entity association. The stable boundary is a double-closed interval.

[0050] More preferably, the method for constructing the stable boundary of entity association is applicable to the first entity association and the second entity association, but not to the third entity association. The third entity association is constructed according to the construction methods of the filtered first entity association, the filtered second entity association, and the third entity association, and the third constraint condition is constrained.

[0051] The method for filtering entities includes obtaining any first entity association, comparing the absolute value or first quotient of the first difference corresponding to the first entity association with the stable boundary of the first entity association, and retaining the first entity association when the absolute value or first quotient of the first difference corresponding to the first entity association is distributed on the stable boundary of the first entity association. When the absolute value of the first difference or the first quotient corresponding to the first entity association is not distributed on the stable boundary of the first entity association, the first entity association is deleted. Get any second entity association, compare the absolute value or second quotient of the second difference corresponding to the second entity association with the stable boundary of the second entity association, and retain the second entity association when the absolute value or second quotient of the second difference corresponding to the second entity association is distributed on the stable boundary of the second entity association; When the absolute value of the second difference or the second quotient corresponding to the second entity association is not distributed on the stable boundary of the second entity association, the second entity association shall be deleted. When the first entity association is deleted or the second entity association is deleted, jump to the corresponding third entity association and delete the third entity association; At this point, the entity filtering step is complete.

[0052] More preferably, by filtering entities and deleting redundant third-party entity associations, the system disturbances caused by redundant associations can be effectively avoided, making the overall structure closer to the true state of dynamic equilibrium.

[0053] The filtered entities and their associations are set as nodes, edges, node types, and edge types, respectively. A node is represented as... ,and ; in, This is the value used to renumber the spatial entities in their original order after the filtering process is complete. For natural numbers, After filtering, the time entities are renumbered according to their original order. It is a natural number; Edges are represented as ,and ; and ; in, Let be the total number of edges. This represents the minimum value between the number of spatial entities selected after filtering and the number of temporal entities selected after filtering. Node type is represented as ,and ; in, For spatial node type, For time node type; Edge type is represented as ,and ; in, The edge type is associated with the first entity. The edge type associated with the second entity. The edge type associated with the third entity; The method for constructing heterogeneous graph convolutional networks includes setting an initial feature vector for each node, denoted as . For each edge, set an initial weight matrix, denoted as . ; For any given node, set other nodes connected to it via edges as its neighbors; The ReLU function is selected as the activation function, and the initial features of the neighboring nodes are linearly transformed through the activation function. According to the convolution operation, the initial feature values ​​of the neighboring nodes after linear transformation are convolved to obtain a new representation of the feature values ​​of the node. The new representation of the node is... ; in, This is represented as the feature quantity representation of a node in the l-th convolutional layer. For activation function, This is the initial weight matrix of the normalized neighbor nodes. Let be the weight matrix of the convolutional kernel in the l-th layer. This is a new representation of the feature values ​​of the node.

[0054] The iteration count of the heterogeneous graph convolutional network is set to the first number. When the first number is reached or the vector sum of the new feature quantities of the neighboring nodes and the node is the same as the corresponding entity, the final representation of the feature quantities of each node is output. At this time, the construction process of the knowledge graph is completed.

[0055] Based on graph query language, the query target is obtained.

[0056] More preferably, the expression for the query target is, ; in, The set of metapaths is obtained using a graph query language. For meta-path based The set of node pairs This represents a mapping relationship.

[0057] The method for obtaining the query target includes constructing sub-element paths and their sub-inverse paths based on the edges associated with the third entity. The sub-element paths and their sub-inverse paths are collectively referred to as meta-paths. Meta-paths are used to locate abnormal transaction clusters under the current spatiotemporal conditions. For each meta-path, configure constraints. The constraints of a meta-path are expressed as follows: for edges containing... For a meta-path of type 1, a distance constraint is set, which is expressed as follows: the sum of the transportation distances corresponding to the spatial nodes in the path is less than the first distance. For edge-containing types The metapath is set with time constraints, which are expressed as follows: the average time difference between time nodes in the path is less than a preset first duration. For edge-containing types The metapath is set with vector constraints, which are expressed as follows: the vector direction associated with the third entity in the path is consistent with the path traversal direction. Expanding the above metapaths using a Cartesian product generates sets of metapaths of different lengths. Set the maximum length of the metapath to L; Iterate through all possible combinations of length from 1 to L, and eliminate invalid paths that violate the constraints; Retain the meta-paths that satisfy the constraints, encode the retained meta-paths, and denote the code of the retained meta-path as P, where P is a natural number and P is the index number of the meta-path in the set.

[0058] Establish the mapping relationship between metapaths and graph query language; This mapping is obtained through Cypher extended syntax.

[0059] More preferably, the Cypher extended syntax is represented as follows: ; in, Represented as Nodes of type Represented as Nodes of type Represented as type of edge, Represented as type of edge, Represented as type of edge, , Each is bound to a node type. and Type of edge binding, and Edge binding of type.

[0060] Intermediate results are passed through the WITH clause, meta-paths are embedded through the WHERE clause, and finally, the graph query language corresponding to the meta-path is output through the RETURN clause. Construct a mapping relationship between metapaths and graph query language; By inputting graph query language into the mapping relationship, the corresponding meta-path is obtained. Based on the meta-path, the path corresponding to the abnormal node is traced, and the traced path is set as the first transaction path.

[0061] More preferably, by tracing abnormal data and abnormal paths, abnormal data can be quickly located, which is beneficial to the speed of refined oil management.

[0062] More preferably, this invention introduces a nonlinear dynamics method to process the stability screening of transaction entities, eliminating risky entities at the source. It uses heterogeneous graph learning technology to construct a highly expressive knowledge graph, preserving the complex structure and semantic information of the transaction network, realizing anomaly tracing based on graph reasoning, breaking through the limitations of traditional single-point detection, providing an interpretable complete transaction path, and upgrading the knowledge graph from a static knowledge base to a dynamic analysis engine. This promotes the transformation of refined oil transaction management from experience-driven to data-intelligent driven, and has significant technical advantages and application value in improving regulatory efficiency, preventing transaction risks, and supporting tax compliance.

[0063] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product implemented on one or more computer-usable storage media containing computer-usable program code. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Red-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk. These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0064] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the protection scope of the present invention.

Claims

1. A method for intelligent management of refined oil product transactions integrating knowledge graphs, characterized in that, The steps include: Step S100, configuring entities and entity associations based on historical transaction data, configuring corresponding constraints for entity associations, obtaining the stable boundary of entity associations based on a nonlinear fork-type bifurcation model, and filtering entities according to the characteristic quantities of entity associations and corresponding constraints. Step S200: Set the filtered entities and entity associations as nodes, edges, node types and edge types. Based on nodes, edges, node types and edge types, construct a heterogeneous graph convolutional network. Based on graph query language and heterogeneous graph convolutional network, obtain a knowledge graph. Step S300: Identify the current abnormal node based on the knowledge graph, and trace the first transaction path based on the relationship of the current abnormal node.

2. The intelligent management method for refined oil transactions integrating knowledge graphs as described in claim 1, characterized in that, Entities include spatial entities and temporal entities; Entity associations include first entity associations, second entity associations, and third entity associations; The constraints include the first constraint, the second constraint, and the third constraint. Based on the transaction parties' IDs and GIS technology, the regions to which the transaction parties belong are obtained. Spatial entities are constructed according to the respective regions of the transaction parties. Time entities are constructed based on the transaction timestamp. First entity associations between time entities are constructed based on the transaction category and the average unit price of the transaction. Second entity associations between spatial entities are constructed based on the batch number of transported oil and the number of transported vehicles. Third entity associations between time entities and spatial entities are constructed based on the first entity association and the second entity association.

3. The intelligent management method for refined oil product transactions integrating knowledge graphs as described in claim 2, characterized in that, The method for constructing a spatial entity based on the transaction parties' IDs includes constructing a time entity based on the transaction timestamp of the spatial entity.

4. The intelligent management method for refined oil product transactions integrating knowledge graphs as described in claim 1, characterized in that, The method for constructing the first entity association includes determining whether the transaction category numbers of any two time entities are the same; When the transaction category numbers are the same, calculate the first difference in the average unit price of the transactions, select the absolute value of the first difference, represent the absolute value of the first difference as the third code, convert the third code into the first vector, and record the first vector as the first entity association; When the transaction category numbers are different, obtain the standard transaction volume corresponding to the transaction category, calculate the first ratio of the standard transaction volumes corresponding to the two time entities, calculate the first difference of the average transaction unit price, select the absolute value of the first difference, calculate the first quotient of the absolute value of the first difference and the first ratio, represent the first quotient as the third code, convert the third code into the first vector, and record the first vector as the first entity association; The construction method for the second entity association includes determining whether the batch numbers of oil transported by any two spatial entities are the same; When the batch numbers of transported oil products are the same, calculate the second difference in the transportation distances corresponding to these two spatial entities, select the absolute value of the second difference, represent the absolute value of the second difference as the fourth code, convert the fourth code into the second vector, and record the second vector as the second entity association; When the batch numbers of transported oil products are different, obtain the standard inspection pass rate corresponding to the refinery number, calculate the second ratio of the standard inspection pass rates corresponding to the two spatial entities, calculate the second difference of the transportation distance corresponding to the two spatial entities, select the absolute value of the second difference, calculate the second quotient of the absolute value of the second difference and the second ratio, represent the second quotient as the fourth code, convert the fourth code into the second vector, and record the second vector as the second entity association; The methods for constructing third entity associations include, Obtain the first vector and the second vector, calculate the vector sum based on the first vector and the second vector, and denote the vector sum as the third vector; The third vector is represented as a third entity association.

5. The intelligent management method for refined oil product transactions integrating knowledge graphs as described in claim 1, characterized in that, Calculate the average of the absolute values ​​of the first differences, set the average of the absolute values ​​of the first differences as the standard value of the absolute values ​​of the first differences, and calculate the average of the first quotients, set the average of the first quotients as the standard value of the first quotients. Calculate the absolute value of any first difference and the third difference between it and the corresponding standard value, or calculate the fourth difference between any first quotient and the corresponding standard value; Set the first value as the third difference threshold and the second value as the fourth difference threshold; The first constraint is set as follows: when the third difference is greater than the first value, or when the fourth difference is greater than the second value, the original first entity association is retained; when the third difference is less than or equal to the first value, or when the fourth difference is less than or equal to the second value, the corresponding first entity association is set to a unit vector with the same direction as the original first vector. Calculate the average of the absolute values ​​of the second difference, set the average of the absolute values ​​of the second difference as the standard value of the absolute value of the second difference, calculate the average of the second quotient, and set the average of the second quotient as the standard value of the second quotient; Calculate the fifth difference between the absolute value of any second difference and its corresponding standard value, or calculate the sixth difference between any second quotient and its corresponding standard value; Set the third value as the third difference threshold and the fourth value as the fourth difference threshold; The second constraint is set as follows: when the fifth difference is greater than the third value, or the sixth difference is greater than the fourth value, the original second entity association is retained; when the fifth difference is less than or equal to the third value, or the sixth difference is less than or equal to the fourth value, the corresponding second entity association is set to a unit vector with the same direction as the original first vector.

6. The intelligent management method for refined oil product transactions integrating knowledge graphs as described in claim 1, characterized in that, The stable boundary of entity association includes the stable boundary of the first entity association and the stable boundary of the second entity association. The method for constructing the stable boundary of entity association includes retrieving a new set of historical transaction data within the first historical time period and constraining the set of entity associations according to the constraints corresponding to the entity associations. Calculate new standard values ​​for entity associations after new historical transaction data is available; Calculate the balance difference based on the new standard value, set the fifth value as the balance threshold, calculate the difference between the balance difference and the fifth value, and record it as the seventh difference. Based on the nonlinear bifurcation model, the seventh difference, and the order of historical acquisition times within the first historical time period, the first expression is obtained; The imbalance points are obtained based on the value of the derivative of the first expression. The stability boundary of entity association is set based on the comparison between the number of imbalance points and 2.

7. The intelligent management method for refined oil product transactions integrating knowledge graphs as described in claim 6, characterized in that, Set the filtered entities and entity associations to nodes, edges, node types, and edge types, respectively; The method for constructing heterogeneous graph convolutional networks includes setting initial feature vectors for each node and setting initial weight matrices for each edge; For any given node, set other nodes connected to it via edges as its neighbors; The initial features of neighboring nodes are linearly transformed using an activation function. Based on the convolution operation, the initial feature values ​​of the neighboring nodes after linear transformation are convolved to obtain a new representation of the feature values ​​of the node. The iteration count of the heterogeneous graph convolutional network is set to the first number. When the first number is reached or the vector sum of the new feature quantities of the neighboring nodes and the node is the same as the corresponding entity, the final representation of the feature quantities of each node is output. At this time, the construction process of the knowledge graph is completed.

8. The intelligent management method for refined oil product transactions integrating knowledge graphs as described in claim 7, characterized in that, Based on graph query language, the query target is obtained; The methods for obtaining the query target include constructing a meta-path based on the edges associated with the third entity. The meta-path is used to locate the abnormal transaction cluster under the current spatiotemporal conditions. Configure constraints for each metapath, retain metapaths that satisfy the constraints, and set index numbers for the retained metapaths.

9. The intelligent management method for refined oil product transactions integrating knowledge graphs as described in claim 8, characterized in that, Establish the mapping relationship between metapaths and graph query language; This mapping is obtained through Cypher extended syntax; Based on the meta-path, trace the path corresponding to the abnormal node and set the traced path as the first transaction path.

10. A refined oil trading intelligent management system integrating knowledge graphs, the system being used to execute the refined oil trading intelligent management method integrating knowledge graphs as described in claim 1, characterized in that, It includes a filtering module, a building module, and an identification module; The filtering module configures entities and entity associations based on historical transaction data, configures corresponding constraints for entity associations, obtains the stable boundary of entity associations based on a nonlinear fork-type bifurcation model, and filters entities according to the feature quantities of entity associations and the corresponding constraints. The construction module sets the filtered entities and entity associations as nodes, edges, node types, and edge types. Based on the nodes, edges, node types, and edge types, it constructs a heterogeneous graph convolutional network. Based on the graph query language and the heterogeneous graph convolutional network, it obtains a knowledge graph. The identification module identifies the current abnormal node based on the knowledge graph and traces the first transaction path based on the relationship between the current abnormal node.

Images