Horizontal well water control and plugging optimization method based on mapping knowledge domain

Abstract

The invention discloses a horizontal well water control and plugging optimization method based on a knowledge graph, which comprises the following steps: collecting multi-source data of oilfield development, and processing the data; s2, constructing a knowledge graph related to water control and water plugging of the horizontal well; designing an inference engine based on the water control and plugging knowledge graph of the horizontal well by utilizing the knowledge graph and the multi-source data; and according to an analysis result of the reasoning engine, a horizontal well list suitable for water control and plugging measures is recognized, a reasoning chain is displayed by using a visualization technology of a graph database, and a visualization report is generated. According to the method, the limitation of a traditional method is broken through by utilizing multi-source heterogeneous data and effectively integrating structured and unstructured data, efficient organization and use of the multi-source data are realized through the knowledge graph, and the comprehensiveness and accuracy of measure optimization are improved. Through automatic optimization analysis and scheme generation, scientific basis and decision support are provided for development of complex oil reservoirs, and economic benefits and social benefits of oil field development can be improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of horizontal well water control and plugging in oil field development, and in particular relates to an optimization method for horizontal well water control and plugging based on a knowledge graph. Background Art

[0002] Water control and shutoff technology is a crucial technical tool in oilfield development, aimed at maintaining efficient well production and reducing the risk of flooding. As oilfield development enters the middle and late stages, formation water intrusion into oil wells becomes increasingly serious, leading to increased water cut, decreased production efficiency, and even waterlogging and shutdown. Waterlogging is particularly problematic in horizontal well development due to the long wellbore trajectories and large reservoir contact areas. Through precise water control and shutoff, water control and shutoff technology can effectively suppress formation water intrusion, extend the production life of oil wells, increase reservoir recovery, and thus safeguard the economic benefits of the oilfield. Therefore, water control and shutoff technology is not only a core support for efficient oilfield development but also a key measure to address the challenge of waterlogging.

[0003] Existing water control and plugging solutions often rely on a single data source for analysis during the development process, such as structured data such as geological model parameters or production history records. While these methods can provide a certain degree of quantitative evidence, they lack the effective integration of multi-source heterogeneous data. Oilfield development involves a variety of data types, including geological exploration data, production dynamics data, technical reports, and expert review opinions, all of which come from diverse sources and formats. However, traditional solutions typically only utilize structured data, ignoring the potential value of other data sources, resulting in analysis results that fail to fully reflect the actual conditions of the oilfield. The limitations of this single data source significantly restrict the applicability and accuracy of water control and plugging solutions in complex reservoir conditions.

[0004] In actual oilfield development, different experts, based on their own experience and knowledge, may reach different conclusions on whether water control and shutoff measures are necessary for the same horizontal well. For example, some experts may focus more on unusual fluctuations in production data, while others may prioritize geological structural features. This divergence in judgment leads to the coexistence of multiple chains of reasoning, each potentially based on different data priorities and logical pathways. While this phenomenon enriches decision-making perspectives to a certain extent, it also complicates and uncertainties the development of water control and shutoff plans. How to extract consistent conclusions from these multiple chains of reasoning or select the optimal solution becomes a major challenge that affects the effectiveness of subsequent implementation.

[0005] Faced with the coexistence of multiple reasoning chains, traditional manual decision-making methods struggle to cope efficiently. Automating the reasoning of horizontal well water control and shutoff methods and generating corresponding reasoning reports has become a pressing issue. Existing water control and shutoff decision-making processes typically rely on manual analysis and empirical judgment by experts, which is inefficient and susceptible to subjective factors. Especially when the data volume is large and the reasoning paths are complex, manual methods struggle to fully integrate multi-source data and quickly generate scientific and reasonable solutions. Therefore, developing an intelligent system capable of automatic reasoning and generating visual reasoning reports requires not only the integration of multi-source heterogeneous data and the construction of a comprehensive knowledge system, but also intelligent analytical capabilities to support efficient decision-making in oilfield development. The implementation of these technologies will be a key breakthrough in improving the efficiency and accuracy of water control and shutoff solution formulation. Summary of the Invention

[0006] The present invention aims to provide a knowledge-graph-based optimization method for horizontal well water control and plugging. This method fully utilizes multi-source heterogeneous data, effectively integrates structured and unstructured data, and constructs a knowledge graph to achieve efficient organization and query of multidimensional data. Furthermore, an automated inference engine is used to optimize measures and generate solutions for horizontal wells, providing a scientific basis and decision-making support for the development of complex reservoirs.

[0007] In order to solve the above technical problems, the technical solution adopted by the present invention is: a horizontal well water control and plugging optimization method based on knowledge graph, comprising the following steps:

[0008] S1: Collect multi-source data of oilfield development, process the multi-source data, and store them in a database;

[0009] S2: Based on the multi-source data, a knowledge graph related to horizontal well water shutoff control is constructed; and a graph database is used to store the knowledge graph to achieve efficient query and analysis of the data;

[0010] S3: Using the knowledge graph constructed above and the multi-source data, design a reasoning engine based on the horizontal well water shutoff knowledge graph to achieve multi-dimensional analysis and automatic reasoning of the human-computer interaction knowledge graph;

[0011] S4: Based on the analysis results of the reasoning engine, a list of horizontal wells suitable for water control and plugging measures is identified, and the visualization technology of the graph database is used to display the reasoning chain and generate a visualization report to provide decision support for oilfield development.

[0012] Furthermore, the S1 includes the following steps:

[0013] S11: Collect structured and unstructured data from the oilfield development process;

[0014] S12: Use an outlier detection algorithm based on dynamic thresholds on structured data to identify and remove outlier data points;

[0015] S13: Regular expression matching technology is used on unstructured data to remove irrelevant symbols, redundant paragraphs, and non-text noise, retaining the core content related to water control and plugging;

[0016] S14: Unify data formats from different sources to conform to database storage standards.

[0017] Furthermore, the S2 includes the following steps:

[0018] S21: Constructing an entity state graph for storing oilfield dynamic data;

[0019] S22: Based on the historical experience and knowledge thinking chain, a logical diagram is constructed to diagnose whether horizontal well water control and plugging measures are necessary;

[0020] S23: Based on the entity state graph and the event knowledge graph, an event graph is generated to reflect the relationship between oilfield dynamic data and horizontal well water control and plugging knowledge, to assist the reasoning engine in reasoning.

[0021] Furthermore, the S3 includes the following steps:

[0022] S31: Using a stack structure to implement a depth-first search of the event graph rule reasoning;

[0023] S32: Integrate various microservice interfaces to realize dynamic condition status judgment;

[0024] S33: Use CF evidence theory to realize the calculation and transfer of confidence;

[0025] S34: Add an exception handling mechanism to ensure the integrity of reasoning, and implement a dynamic reasoning process that starts from the initial phenomenon, deduces through rule chains, and outputs the final conclusion and its confidence level.

[0026] Furthermore, the S31 includes the following steps:

[0027] S311: creating a rule stack for storing inference rules to be processed, wherein the rule includes a trigger condition set and a conclusion node;

[0028] S312: Loop and iterate until the top element of the stack is empty;

[0029] S313: Through the first-in, last-out feature of the stack, a depth-first traversal of the reasoning path is implemented.

[0030] Furthermore, the S312 includes the following steps:

[0031] S3121: Starting from the current phenomenon node, search for all triggerable rules, filter out the unstacked rules, and push them to the top of the stack;

[0032] S3122: Get the top rule in the stack and check its triggering condition status;

[0033] S3123: If all trigger conditions are met, rule reasoning is executed: the conclusion node state is set to true, the node attributes in the knowledge graph are updated, and the current rule is popped;

[0034] S3124: If the trigger condition is not met, find the rule associated with the unknown condition, push it to the top of the stack and update the current phenomenon node.

[0035] Furthermore, the S32 includes the following steps:

[0036] S321: Obtain real-time horizontal well data through the microservice interface and judge the phenomenon;

[0037] S322: Perform dynamic state evaluation on the conditional nodes of the rule;

[0038] S323: Through the reflection mechanism or direct calling of microservices, data interaction between the program and the external system is realized to ensure the real-time and accuracy of condition status judgment.

[0039] Furthermore, the S33 includes the following steps:

[0040] S331: Setting confidence calculation rules;

[0041] S332: Use Dempster-Shafer synthesis rule to fuse evidence of multiple rule conclusions;

[0042] S333: The confidence is transferred step by step in the reasoning chain through recursive calls.

[0043] Furthermore, the S34 includes the following steps:

[0044] S341: Anomaly detection and capture;

[0045] S342: Abnormal recovery strategy;

[0046] S343: Achieve robustness and traceability of the reasoning process through state writeback and log tracking.

[0047] Furthermore, the S4 includes the following steps:

[0048] S41: Using the inference engine to perform an in-depth analysis on the knowledge graph, and recommending horizontal wells suitable for water control and plugging based on the analysis results;

[0049] S42: Utilizing the visualization function of the graph database, the complete reasoning chain of the horizontal well is displayed, and the user can modify the parameters of the graph database nodes as needed, thereby enabling manual intervention in the reasoning chain;

[0050] S43: Generate a visualization report.

[0051] The advantages and positive effects of the present invention are:

[0052] This method leverages multi-source heterogeneous data, effectively integrating structured and unstructured data to overcome the limitations of traditional methods. It also enables efficient organization and utilization of multi-source data through knowledge graphs, improving the comprehensiveness and accuracy of optimal measures. Through automated optimal analysis and solution generation, it provides a scientific basis and decision-making support for the development of complex reservoirs, contributing to improved economic and social benefits of oilfield development. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] Figure 1 It is a schematic diagram of the overall process of an embodiment of the present invention.

[0054] Figure 2 It is a logical framework diagram for constructing a knowledge graph in the field of horizontal well water shutoff control according to an embodiment of the present invention.

[0055] Figure 3 It is a technical roadmap for the optimal method for horizontal well water shutoff control based on knowledge graph in an embodiment of the present invention.

[0056] Figure 4 4 is a structural diagram of the automated reasoning engine according to an embodiment of the present invention.

[0057] Figure 5 It is a class diagram of the core object class of the inference engine of an embodiment of the present invention.

[0058] Figure 6 This is a diagram of the microservice application architecture of an embodiment of the present invention. DETAILED DESCRIPTION

[0059] The technical solution of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the embodiments described are only some embodiments of the present invention, not all embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0060] The embodiments of the present invention are further described below with reference to the accompanying drawings:

[0061] like Figure 1 As shown, a method for optimizing water shutoff control in horizontal wells based on a knowledge graph includes the following steps.

[0062] S1: Collect multi-source data on oilfield development, including structured and unstructured data; clean and standardize the data and store it in a database to provide data support for the construction of the knowledge graph. Specifically, S1 includes the following steps.

[0063] S11: The system collects structured data from the oilfield development process, such as geological model parameters and production history, and unstructured data, such as technical reports, expert review opinions, construction logs, and other textual materials, through API interfaces, database connections, and file import. The data collection module supports multiple data formats to ensure comprehensiveness and diversity.

[0064] S12: Use the validity, completeness, and consistency of data quality rules to clean up the collected data, remove outliers, duplicates, and incomplete records, and ensure data quality. Specifically, it includes the following:

[0065] An outlier detection algorithm based on dynamic threshold is used for structured data to identify and remove outlier data points.

[0066] Regular expression matching technology is used for unstructured data to remove irrelevant symbols, redundant paragraphs and non-text noise, retaining the core content related to water control and plugging.

[0067] S13: Standardize the cleaned data and unify the data formats from different sources to make them conform to database storage standards. This embodiment designs and implements a database model for the intelligent analysis system for low-yield and low-efficiency wells based on the CNOOC data lake model.

[0068] S2: Based on historical experience knowledge and structured data, a knowledge graph related to horizontal well water control and shutoff is constructed; based on ontology and semantic network theory, OWL language is used for formal description, structured knowledge representation is constructed, and a graph database is used to store the knowledge graph to achieve efficient data query and analysis. Specifically, Figure 2 As shown, S2 includes the following steps.

[0069] S21: Constructing an entity state graph for storing dynamic data such as oilfield production data and test data. Specifically, S21 includes the following steps.

[0070] S211: Construct an entity state graph ontology, define entity types (such as well groups, layers, oil well production information, etc.) and their attributes, clarify the relationship types between entities (such as injection-production relationship, test relationship, etc.), and use OWL (Web Ontology Language) to formally describe the ontology.

[0071] S212: Automatically generate an entity state graph based on the ontology and the oilfield source database. Specifically, S212 includes the following steps:

[0072] S2121: Use ETL tools such as Kettle to monitor the structured data of the source database and regularly update the self-built Registry database.

[0073] S2122: Map the data of the self-built horizontal well database to the entity types and relationship types defined in the ontology in Neo4j to generate the initial horizontal well entity state map.

[0074] S22: Based on the historical experience knowledge thinking chain, a causal map is constructed for diagnosing whether horizontal well water control and plugging measures are necessary. Specifically, S22 includes the following steps.

[0075] S221: Construct a causal graph ontology, define four types of nodes: phenomena, rules, causes, measures, and the relationships between the four types of nodes, such as "inference", "trigger", "formulation", etc. Based on semantic network theory, ensure the logical consistency of graph nodes and relationships.

[0076] S222: Define the logical chain for horizontal well water control and shutoff decision-making. Build a knowledge base of historical data and, based on this knowledge base, construct a reasoning path to clarify the logical chain from phenomenon to cause to action. This provides a data foundation for constructing the logical chain for horizontal well water control and shutoff decision-making.

[0077] S223: Generate a cause-and-effect graph for horizontal well water shutoff control. Based on ontology definitions and decision logic chains, extract nodes of four categories: phenomena, rules, causes, and measures from the historical data knowledge base. These extracted nodes are connected by relationships and stored as a cause-and-effect graph, supporting efficient query and dynamic updates.

[0078] S23: Based on the entity state graph and the event knowledge graph, an event graph is generated to reflect the relationship between the oilfield dynamic data and the historical experience knowledge thinking chain, to assist the reasoning engine in reasoning. Specifically, S23 includes the following steps.

[0079] S231: Obtain all nodes of the event graph and entity nodes of the entity state graph.

[0080] S232: Node replication and association, connecting the nodes of the event graph with the corresponding entity nodes. Specifically including the following:

[0081] Phenomenon node replication, traverses each phenomenon node, creates a phenomenon copy based on its associated entity label and establishes associations.

[0082] The logic of rule node replication is similar to that of phenomenon nodes. It is replicated and associated based on the entity tags associated with the rule.

[0083] S233: Relationship migration, generating event graph. Specifically, S233 includes the following steps:

[0084] S2331: Obtain the relationship between the horizontal well water control and plugging event map.

[0085] S2332: Reconstruct the event relationship based on the above-mentioned relationship between things and generate an event map.

[0086] S3: Using the knowledge graph and structured data constructed above, we design a domain knowledge graph reasoning engine that combines forward reasoning (data-driven) with reverse verification (goal-driven), and encapsulate the phenomenon nodes of the event graph into microservices to support multi-dimensional analysis and automatic reasoning of the knowledge graph. Specifically, Figure 3-5 As shown, S3 includes the following steps:

[0087] S31: Design a reasoning engine based on domain knowledge graph, use stack structure to implement event graph rule reasoning, and perform reasoning path traversal through depth-first search (DFS). Specifically, S31 includes the following steps:

[0088] S311: Create a rule stack for storing inference rules to be processed, each rule including a trigger condition set and a conclusion node.

[0089] S312: Loop and iterate the top element of the stack until the stack is empty. Specifically, S312 includes the following steps:

[0090] S3121: Starting from the current phenomenon node, search for all triggerable rules, filter out the rules that have not been pushed into the stack, and push them to the top of the stack.

[0091] S3122: Get the top rule of the stack and check its trigger condition status.

[0092] S3123: If all trigger conditions are met, rule reasoning is executed: the conclusion node state is set to true, the node attributes in the knowledge graph are updated, and the current rule is popped out.

[0093] S3124: If the trigger condition is not met, find the rule associated with the unknown condition, push it to the top of the stack and update the current phenomenon node.

[0094] S313: Through the first-in, last-out feature of the stack, a depth-first traversal of the reasoning path is implemented to ensure the comprehensiveness and efficiency of reasoning.

[0095] S32: Integrate various microservice interfaces to realize dynamic condition status judgment. Specifically, it includes the following steps:

[0096] S321: Obtain real-time horizontal well data through the microservice interface and judge the phenomenon. Specifically, it includes the following:

[0097] For phenomenon nodes in the event graph that require human intervention, such as the static connectivity of well groups, we first call external APIs such as DeepSeek to conduct questions and answers, obtain judgment results, and return the results to ensure that the inference engine can provide inference results. The results of these phenomenon nodes that require human intervention can be modified by specialized technicians, thereby manually intervening in the inference results.

[0098] The system can automatically judge certain phenomenon nodes that are calculated only by formulas and return the results to the inference engine.

[0099] S322: Perform dynamic status evaluation on the conditional node of the rule, specifically including the following steps:

[0100] S3221: Call the threshold judgment service to update the status and confidence of the phenomenon node based on real-time data.

[0101] S3222: If the conditional node is associated with an external entity, such as a well group or layer, the associated node in the knowledge graph can be matched through the entity identifier (entity field) and its attributes can be dynamically loaded.

[0102] S323: Through the reflection mechanism or direct calling of microservices, data interaction between the program and the external system is realized to ensure the real-time and accuracy of condition status judgment.

[0103] S33: Use CF evidence theory to realize the calculation and transfer of confidence. Specifically, it includes the following steps:

[0104] S331: Define the confidence calculation rules; specifically include the following:

[0105] For each condition node of a rule, the minimum confidence is taken as the basic confidence for triggering the rule.

[0106] The temporary confidence of the conclusion node is calculated based on the confidence weight of the rule itself.

[0107] S332: Use Dempster-Shafer synthesis rule to fuse evidence of multiple rule conclusions, including the following:

[0108] If multiple rules support the same conclusion, the confidence is combined according to the following formula:

[0109] C combined =C1+C2-C1×C2

[0110] (When C1>0 and C2>0)

[0111]

[0112] (When C1 and C2 have opposite signs)

[0113] The final synthetic confidence is updated to the conclusion node to support the weight decision of subsequent reasoning.

[0114] S333: The confidence is transferred step by step in the reasoning chain through recursive calls.

[0115] S34: Add an exception handling mechanism to ensure the integrity of reasoning, and ultimately achieve a dynamic reasoning process that starts from the initial phenomenon, deduces through rule chains, and outputs the final conclusion and its confidence level. It is suitable for complex systems that require multi-level conditional judgment. Specifically, S34 includes the following steps:

[0116] S341: Anomaly detection and capture; specifically, including the following steps:

[0117] S3411: Include a try-catch block in the main inference loop to catch service call exceptions and logic errors.

[0118] S3412: When a conditional node of a rule cannot be inferred through existing rules, it is marked as an "anomaly node".

[0119] S342: Abnormal recovery strategy; specifically, including the following steps:

[0120] S3421: Force the status of the abnormal phenomenon node to be true and the confidence to be zero to avoid interruption of the reasoning chain.

[0121] S3422: Push the exception node back into the stack as a new trigger condition to ensure that subsequent rules can continue to be executed.

[0122] S343: Achieve robustness and traceability of the reasoning process through state writeback and log tracking.

[0123] Specifically, the reasoning engine of this embodiment performs multi-dimensional analysis and automatic reasoning on the knowledge graph, and uses the credibility theory to deal with the uncertainty in the horizontal well water control and shutoff analysis. Figure 6 As shown, specifically including:

[0124] Credibility theory is used to deal with the uncertainty in horizontal well water control and plugging analysis. The credibility of the rules and premise (evidence) is used as input to infer the credibility of the conclusion.

[0125] Credibility calculation of combined evidence:

[0126] When the evidence is the conjunction of multiple single pieces of evidence, that is,

[0127] E=E1∧E2∧…∧E n

[0128] The credibility is as follows,

[0129] CF(E)=min{CF(E1),CF(E2),…,CF(E n )}

[0130] When the evidence is a disjunction of multiple single pieces of evidence, i.e.

[0131] E=E1∨E2∨…∨E n

[0132] The credibility is as follows,

[0133] CF(E)=max{CF(E1),CF(E2),…,CF(E n )}

[0134] When there are multiple pieces of knowledge supporting the conclusion, and the credibility of the evidence and knowledge (rules) is known, the credibility of the conclusion is calculated as follows;

[0135] IF E1 THEN H(CF(H,E1))

[0136] CF1(H)=CF(H,E1)×max{0,CF(E1)}

[0137] IF E2 THEN H(CF(H,E2))

[0138] CF1(H)=CF(H,E2)×max{0,CF(E2)}

[0139]

[0140] S4: Based on the analysis results of the inference engine, a list of horizontal wells suitable for water control and plugging measures is identified, and the inference chain is displayed using graph database visualization technology. Finally, a visual report is generated to provide decision support for oilfield development. Specifically, S4 includes the following steps.

[0141] S41: Use the inference engine to conduct in-depth analysis of the knowledge graph and recommend horizontal wells suitable for water control and plugging based on the analysis results.

[0142] S42: Using the visualization function of the graph database, the complete reasoning chain of the horizontal well is displayed, and users can change the parameters of the graph database nodes according to their needs, so that manual intervention in the reasoning chain is possible. Specifically, it includes the following:

[0143] Provides a visual parameter adjustment interface, allowing users to modify node weights or add temporary constraints.

[0144] The inference engine is re-executed based on the modified parameters to generate updated recommendation results and provide real-time feedback.

[0145] Specifically, the built-in visualization tools of the Neo4j graph database are used to display the complete reasoning chain for each horizontal well. Users can intuitively understand the decision basis and modify node parameters through the interface (such as manually modifying the emulsification interval value) to manually intervene in the reasoning results.

[0146] S43: Generate a visual report. Specifically including the following:

[0147] The topological structure diagram of the reasoning path is exported through the graph database, and the key decision nodes and confidence distribution are marked.

[0148] Generate a detailed report on the optimal method for horizontal well water control and plugging, including horizontal well water control and plugging priority ranking, measure parameters, etc.

[0149] Specifically, the system of this embodiment automatically generates a visual report containing a list of preferred well locations, reasoning links, data analysis, and effect predictions, and supports export in multiple formats such as PDF and Excel, providing a scientific basis for oilfield development decisions.

[0150] The advantages and positive effects of the present invention are:

[0151] This method leverages multi-source heterogeneous data, effectively integrating structured and unstructured data to overcome the limitations of traditional methods. It also enables efficient organization and utilization of multi-source data through knowledge graphs, improving the comprehensiveness and accuracy of optimal measures. Through automated optimal analysis and solution generation, it provides a scientific basis and decision-making support for the development of complex reservoirs, contributing to improved economic and social benefits of oilfield development.

[0152] The above is a detailed description of an embodiment of the present invention. However, the content described is only a preferred embodiment of the present invention and should not be considered to limit the scope of the present invention. All equivalent changes and improvements made within the scope of the present invention should still fall within the scope of the patent coverage of the present invention.

Claims

1. A knowledge graph-based optimization method for horizontal well water shutoff control, characterized by: The following steps are included: S1: Collect multi-source data of oilfield development, process the multi-source data, and store them in a database; S2: Based on the multi-source data, a knowledge graph related to horizontal well water shutoff control is constructed; and a graph database is used to store the knowledge graph to achieve efficient query and analysis of the data; S3: Using the knowledge graph constructed above and the multi-source data, design a reasoning engine based on the horizontal well water shutoff knowledge graph to achieve multi-dimensional analysis and automatic reasoning of the human-computer interaction knowledge graph; S4: Based on the analysis results of the reasoning engine, a list of horizontal wells suitable for water control and plugging measures is identified, and the visualization technology of the graph database is used to display the reasoning chain and generate a visualization report to provide decision support for oilfield development.

2. The method for optimizing horizontal well water shutoff control based on knowledge graph according to claim 1, characterized in that: Said S1 comprises the following steps, S11: Collect structured and unstructured data from the oilfield development process; S12: Use an outlier detection algorithm based on dynamic thresholds on structured data to identify and remove outlier data points; S13: Regular expression matching technology is used on unstructured data to remove irrelevant symbols, redundant paragraphs, and non-text noise, retaining the core content related to water control and plugging; S14: Unify data formats from different sources to conform to database storage standards.

3. The method for optimizing horizontal well water shutoff control based on knowledge graph according to claim 1 or 2, characterized in that: Said S2 comprises the following steps, S21: Constructing an entity state graph for storing oilfield dynamic data; S22: Based on the historical experience and knowledge thinking chain, a logical diagram is constructed to diagnose whether horizontal well water control and plugging measures are necessary; S23: Based on the entity state graph and the event knowledge graph, an event graph is generated to reflect the relationship between oilfield dynamic data and horizontal well water control and plugging knowledge, to assist the reasoning engine in reasoning.

4. The method for optimizing horizontal well water shutoff control based on knowledge graph according to claim 3 is characterized by: Said S3 comprises the following steps, S31: Using a stack structure to implement a depth-first search of the event graph rule reasoning; S32: Integrate various microservice interfaces to realize dynamic condition status judgment; S33: Use CF evidence theory to realize the calculation and transfer of confidence; S34: Add an exception handling mechanism to ensure the integrity of reasoning, and implement a dynamic reasoning process that starts from the initial phenomenon, deduces through rule chains, and outputs the final conclusion and its confidence level.

5. The method for optimizing horizontal well water shutoff control based on knowledge graph according to claim 4 is characterized in that: S31 The following steps are included: S311: creating a rule stack for storing inference rules to be processed, wherein the rule includes a trigger condition set and a conclusion node; S312: Loop and iterate until the top element of the stack is empty; S313: Through the first-in, last-out feature of the stack, a depth-first traversal of the reasoning path is implemented.

6. The method for optimizing horizontal well water shutoff control based on knowledge graph according to claim 5, characterized in that: The S312 includes the following steps: S3121: Starting from the current phenomenon node, search for all triggerable rules, filter out the unstacked rules, and push them to the top of the stack; S3122: Get the top rule in the stack and check its triggering condition status; S3123: If all trigger conditions are met, rule reasoning is executed: the conclusion node state is set to true, the node attributes in the knowledge graph are updated, and the current rule is popped; S3124: If the trigger condition is not met, find the rule associated with the unknown condition, push it to the top of the stack and update the current phenomenon node.

7. A horizontal well water shutoff optimization method based on knowledge graph according to any one of claims 4 to 6, characterized in that: The S32 The following steps are included: S321: Obtain real-time horizontal well data through the microservice interface and judge the phenomenon; S322: Perform dynamic state evaluation on the conditional nodes of the rule; S323: Through the reflection mechanism or direct calling of microservices, data interaction between the program and the external system is realized to ensure the real-time and accuracy of condition status judgment.

8. A horizontal well water shutoff optimization method based on knowledge graph according to any one of claims 4 to 6, characterized in that: The S33 includes the following steps: S331: Setting confidence calculation rules; S332: Use Dempster-Shafer synthesis rule to fuse evidence of multiple rule conclusions; S333: The confidence is transferred step by step in the reasoning chain through recursive calls.

9. A horizontal well water shutoff optimization method based on knowledge graph according to any one of claims 4 to 6, characterized in that: The S34 includes the following steps: S341: Anomaly detection and capture; S342: Abnormal recovery strategy; S343: Achieve robustness and traceability of the reasoning process through state writeback and log tracking.

10. The method for optimizing horizontal well water shutoff control based on knowledge graph according to claim 1 or 2, characterized in that: Said S4 comprises the following steps, S41: Using the inference engine to perform an in-depth analysis on the knowledge graph, and recommending horizontal wells suitable for water control and plugging based on the analysis results; S42: Utilizing the visualization function of the graph database, the complete reasoning chain of the horizontal well is displayed, and the user can modify the parameters of the graph database nodes as needed, thereby enabling manual intervention in the reasoning chain; S43: Generate a visualization report.

Images