An agent construction method of a pipeline cathodic protection operation and maintenance decision type knowledge base

Abstract

The application discloses an agent construction method of a pipeline cathodic protection operation and maintenance decision-making knowledge base, relates to the technical field of decision-making knowledge base ontology modeling, and the method is characterized in that, in the agent construction process of the pipeline cathodic protection operation and maintenance, the pipeline cathodic protection operation and maintenance decision-making knowledge base is subjected to ontology model modeling, the constructed ontology model is subjected to qualification verification and correction, after the ontology model modeling is qualified, the fusion and correlation of cathodic protection data are carried out, the fusion quality of the cathodic protection data is evaluated and calibrated to establish effective correlation of cross-modal data, after the fusion and correlation of the cathodic protection data are qualified, the simulation result of the pipeline cathodic protection operation and maintenance decision-making knowledge base is evaluated, if the simulation result is qualified, closed-loop intelligent decision-making is carried out and a decision-making result is obtained, the decision-making effect is tested and iterated, if the simulation result is unqualified, an agent needs to be artificially reconstructed, and a prompt is sent, thereby solving the problem that the safety of the pipeline cathodic protection operation and maintenance decision-making cannot be guaranteed in the prior art.

Description

Technical Field

[0001] This invention relates to the field of decision-making knowledge base ontology modeling technology, and in particular to a method for constructing an intelligent agent for a pipeline cathodic protection operation and maintenance decision-making knowledge base. Background Technology

[0002] As a core infrastructure for transporting strategic materials such as energy and chemicals, the safe and stable operation of pipelines is directly related to the continuity of industrial production and public safety. Pipeline cathodic protection technology, as a key means to inhibit the corrosion of steel pipelines, needs to achieve the anti-corrosion goal by precisely controlling the pipeline potential.

[0003] With the continuous increase in pipeline mileage and service life, as well as the increasing complexity of working conditions such as changes in soil resistivity, stray current interference, and damage from third-party construction, the traditional operation and maintenance model that relies on manual inspection and experience-based judgment can no longer meet the intelligent transformation requirements of "real-time response, accurate decision-making, and compliant operation." The industry's demand for intelligent upgrades in cathodic protection operation and maintenance is extremely urgent.

[0004] Current explorations in intelligent operation and maintenance of pipeline cathodic protection begin with ontology modeling, such as OWL (Web Ontology Language) and RDF (Resource Description Framework), to formally define equipment types, monitoring parameters, fault modes, protection principles, operation and maintenance specifications, and judgment criteria within the cathodic protection field. This constructs a machine-understandable semantic framework. Based on this, information from industry standards, historical operation and maintenance cases, fault databases, equipment ledgers, and expert experience is integrated to build a professional knowledge graph covering the relationships between pipelines, equipment, parameters, faults, handling measures, and operational effects. Subsequently, multimodal data fusion technology is used to uniformly access and process structured monitoring data, semi-structured operation and maintenance records, and unstructured on-site images, videos, documents, drawings, and other information in cathodic protection operation and maintenance scenarios, completing data cleaning, spatiotemporal alignment, and anomaly removal. Beyond missing information completion, the rule engine solidifies industry standards, threshold criteria, and fault judgment conditions into executable logic. This enables automatic identification and tiered early warning of typical issues such as abnormal potential, insulation failure, and grounding faults. It also integrates with a knowledge graph to quickly match fault causes, historical cases, and recommended handling measures. Furthermore, it incorporates a large language model to achieve deep understanding, information extraction, and knowledge transformation capabilities for unstructured text. This allows for automatic parsing of inspection reports, standard clauses, and technical documents. Simultaneously, it supports maintenance personnel in using natural language to conduct status queries, fault consultations, and solution generation. The agent-centric reasoning mechanism enables continuous perception of multi-source fusion data and can work in conjunction with SCADA (Supervisory Control And Data Acquisition) systems, inspection systems, and management platforms to execute early warning commands and maintenance scheduling.

[0005] However, existing technologies, due to their simple integration of mature technologies and lack of customized development and detailed design between technical modules in the field of cathodic protection, suffer from several drawbacks. First, general AI (Artificial Intelligence) lacks professional understanding of cathodic protection mechanisms and lacks domain-specific ontology knowledge support. It cannot establish equipment topology relationships, distinguish the meaning of professional parameters, or understand domain causal logic. As a result, maintenance suggestions are divorced from professional mechanisms, leading to low accuracy in fault location. Second, due to the lack of professional mechanism support, general AI cannot build cross-modal data association mechanisms adapted to the field of cathodic protection. This prevents the effective fusion and association of multi-source heterogeneous data. It also lacks cross-modal data association mechanisms, cannot automatically extract the spatiotemporal event characteristics of unstructured data, and cannot establish causal mappings between unstructured data and structured data or geospatial data. Fault diagnosis lacks complete context and is prone to misjudgment and omission. Third, decision outputs lack constraints and physical feasibility verification. Purely data-driven AI models do not incorporate the rigid constraints of industry standards, posing a "model illusion" risk. They are prone to generating maintenance suggestions that violate regulations, potentially leading to safety hazards such as overprotection, and cannot guarantee the compliance and safety of pipeline cathodic protection maintenance decisions. Summary of the Invention

[0006] This invention provides a method for constructing an intelligent agent for a pipeline cathodic protection operation and maintenance decision-making knowledge base. The technical solution provided by this application is as follows: A method for constructing an intelligent agent for a pipeline cathodic protection operation and maintenance decision-making knowledge base is provided. The specific implementation of this method is as follows: An ontology model is constructed for the pipeline cathodic protection operation and maintenance decision-making knowledge base, and the constructed ontology model is verified and corrected for its conformity. After the ontology model is qualified, cathodic protection data is fused and correlated, and the quality of cathodic protection data fusion is evaluated and calibrated to establish effective correlation of cross-modal data. After the cathodic protection data fusion and correlation are qualified, the simulation results of the pipeline cathodic protection operation and maintenance decision-making knowledge base are evaluated. If the simulation results are qualified, closed-loop intelligent decision-making is carried out and the decision results are obtained. The decision effect is verified and iterated. If the simulation results are unqualified, a prompt that the intelligent agent needs to be manually reconstructed is sent to the preset personnel.

[0007] The beneficial effects of the technical solutions provided in the embodiments of the present invention include at least the following: 1. Ontology modeling of the pipeline cathodic protection operation and maintenance decision-making knowledge base, followed by qualification verification and correction of the constructed ontology model, helps to build a complete, semantically consistent, and logically rigorous professional knowledge framework for the intelligent agent in the field of cathodic protection. This enables the intelligent agent to accurately understand the intrinsic relationship between physical entities, environmental factors, and fault events, laying a solid cognitive foundation for subsequent intelligent reasoning. The fusion and correlation of cathodic protection data, along with the evaluation and calibration of the fusion quality to establish effective cross-modal data correlation, helps to break down information silos between structured time-series data, unstructured operation and maintenance logs, and geospatial data, providing the intelligent agent with accurate, comprehensive, and traceable decision-making basis. The simulation results of the pipeline cathodic protection operation and maintenance decision-making knowledge base are evaluated to verify the feasibility and safety of the intelligent agent's output scheme in a real physical environment. This ensures that decision recommendations not only meet regulatory and standard requirements but also conform to electrochemical mechanisms and field operating condition constraints, fundamentally eliminating secondary risks caused by simulation deviations and guaranteeing the long-term stable operation of the pipeline cathodic protection management system.

[0008] 2. This solution conducts qualification verification and correction on the constructed ontology model, building a professional knowledge framework for the intelligent agent that accurately maps physical entities, environmental entities, and event entities in the cathodic protection management system and their complex internal relationships. This ensures from the root that the intelligent agent has the core domain cognitive ability to deeply understand the equipment topology, parameter semantic connotations, and electrochemical causal logic, effectively establishing complete and reliable equipment topology relationships, accurately distinguishing the actual meaning of various professional parameters, truly understanding and applying the causal logic in the domain, ensuring that the generated operation and maintenance suggestions always conform to professional mechanisms, significantly improving the accuracy and reliability of fault location, and ultimately achieving a comprehensive improvement in the depth of cognition, the accuracy of reasoning, and the professionalism of decision-making in the intelligent operation and maintenance of the cathodic protection system.

[0009] 3. This solution assesses and calibrates the quality of cathodic protection data fusion. Compared to existing general AI technologies, which lack professional mechanism support and cannot build a cross-modal data association mechanism adapted to the field of cathodic protection, resulting in the ineffective fusion and association of multi-source heterogeneous data, the lack of cross-modal data association mechanism, the inability to automatically extract the spatiotemporal event features of unstructured data, and the inability to establish a causal mapping between unstructured data and structured data and geospatial data, fault diagnosis lacks complete context and is prone to misjudgment and omission. This solution can achieve precise alignment and semantic association in the spatiotemporal dimension, break down data silos, and provide intelligent agents with comprehensive, consistent, and traceable decision-making basis, fundamentally improving the accuracy and completeness of fault diagnosis and effectively reducing the risk of misjudgment and omission.

[0010] 4. Optimize the reasoning logic or feedback mechanism. Compared with existing technologies, which rely on purely data-driven AI models that do not incorporate rigid industry standard constraints, leading to the risk of "model illusion" and the generation of maintenance suggestions that violate regulations, potentially causing overprotection and other safety hazards, and failing to guarantee the compliance and safety of pipeline cathodic protection maintenance decisions, this solution ensures that every decision in the pipeline cathodic protection maintenance process strictly follows professional standards and safety requirements. While improving the scientific and rational nature of maintenance decisions, it effectively guarantees the compliance of maintenance operations and the safety of system operation, achieving an organic unity between data-driven approaches and industry mechanisms and standard constraints. This provides reliable, controllable, and traceable intelligent decision-making support for the safe and stable operation of pipeline cathodic protection throughout its entire lifecycle. Attached Figure Description

[0011] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0012] Figure 1 This is a flowchart of an intelligent agent construction method for a pipeline cathodic protection operation and maintenance decision-making knowledge base provided in an embodiment of the present invention; Figure 2 This is a schematic diagram of the structure of an intelligent agent construction method for a pipeline cathodic protection operation and maintenance decision-making knowledge base provided in an embodiment of the present invention; Figure 3 This is an ontology modeling architecture diagram of an intelligent agent construction method for a pipeline cathodic protection operation and maintenance decision-making knowledge base provided in an embodiment of the present invention; Figure 4 This is a diagram illustrating the fusion and association architecture of cathodic protection data in an intelligent agent construction method for a pipeline cathodic protection operation and maintenance decision-making knowledge base provided in this embodiment of the invention. Figure 5 This is a schematic diagram of the workflow of a pipeline cathodic protection operation and maintenance decision-making knowledge base intelligent agent provided in an embodiment of the present invention. Detailed Implementation

[0013] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the implementation methods of this application will be further described in detail below with reference to the accompanying drawings.

[0014] Before providing a detailed explanation of the embodiments of this application, the application scenarios of these embodiments will be described first.

[0015] Example 1: This embodiment of the invention provides a method for constructing an intelligent agent for a pipeline cathodic protection operation and maintenance decision-making knowledge base. For example... Figure 1The flowchart shown is a method for constructing an intelligent agent for a pipeline cathodic protection operation and maintenance decision-making knowledge base. The processing flow of this method may include the following steps: The system acquires the device topology matching degree and determines whether it is greater than or equal to the preset device topology matching degree. If so, it sends a device topology matching qualified prompt to the preset personnel. Otherwise, it acquires the parameter semantic recognition accuracy and causal logic correctness, and performs cathodic protection data fusion and association. When the cathodic protection data alignment accuracy is greater than the preset cathodic protection data alignment accuracy, the cathodic protection feature extraction accuracy is greater than the preset cathodic protection feature extraction accuracy, and the cathodic protection data association success rate is greater than the preset cathodic protection data association success rate, it constructs compliance constraints and determines whether the potential simulation deviation is less than or equal to the preset potential simulation deviation. If so, it performs closed-loop intelligent decision-making and feedback updates based on the thought chain. Otherwise, it implements physical feasibility optimization. When any one of the cathodic protection data alignment accuracy, cathodic protection feature extraction accuracy, and cathodic protection data association success rate is less than or equal to the corresponding cathodic protection data standard parameter, it calibrates the cathodic protection data fusion quality.

[0016] like Figure 2 The diagram shown is a structural schematic of an intelligent agent construction method for a pipeline cathodic protection operation and maintenance decision-making knowledge base provided in an embodiment of the present invention. In the process of constructing the intelligent agent for pipeline cathodic protection operation and maintenance, firstly, an ontology model is built for the pipeline cathodic protection operation and maintenance decision-making knowledge base. The constructed ontology model is then validated and corrected to ensure that the intelligent agent possesses professional domain cognitive capabilities. The pipeline cathodic protection operation and maintenance decision-making knowledge base refers to a systematic collection of knowledge used to store and manage professional knowledge in the field of cathodic protection, including a conceptual system, entity attributes, semantic relationships, and historical cases built based on the ontology model. This process can build a professional knowledge skeleton for the intelligent agent that accurately maps the physical entities, environmental entities, and event entities of the cathodic protection system and their inherent relationships, enabling the intelligent agent to have deep domain cognitive capabilities to understand equipment topology connections, distinguish the semantic connotations of parameters, and grasp electrochemical causal logic. For example, the ontology model refers to a structured knowledge representation model built based on professional knowledge in the field of pipeline cathodic protection, used to uniformly represent physical entities, environmental entities, event entities, and their physical connection relationships, causal influence relationships, and spatial mapping relationships in the cathodic protection management system.

[0017] After the ontology model is successfully modeled, the cathodic protection data is fused and correlated, and the quality of the cathodic protection data fusion is calibrated to establish effective correlation of cross-modal data. Cathodic protection data refers to the full amount of operation and maintenance data collected during the operation and maintenance of pipeline cathodic protection, which comes from different sources and has significant differences in data structure and presentation. This includes structured time-series data, unstructured logs, and GIS (Geographic Information System) data. This process can provide the agent with comprehensive, consistent, and traceable cross-modal decision-making basis, fundamentally improving the accuracy and completeness of fault diagnosis.

[0018] After the cathodic protection data is successfully integrated and correlated, the simulation results of the pipeline cathodic protection operation and maintenance decision-making knowledge base are evaluated. If the simulation results are satisfactory, closed-loop intelligent decision-making is carried out and the decision results are obtained. The decision effect is tested and iterated to evaluate the accuracy, effectiveness and adaptive optimization capability of the intelligent agent. If the simulation results are unsatisfactory, a prompt that the intelligent agent needs to be manually reconstructed is sent to the preset personnel. This process can provide accurate, reliable and adaptive intelligent decision-making support for the long-term safe and stable operation of pipeline cathodic protection.

[0019] In this application, a database storing various types of preset data is established before designing an intelligent agent construction method for a pipeline cathodic protection operation and maintenance decision-making knowledge base. The database includes, but is not limited to, preset equipment topology matching degree, preset parameter semantic recognition accuracy, preset causal logic correctness, preset cathodic protection data alignment accuracy, preset cathodic protection feature extraction accuracy, and preset cathodic protection data association success rate.

[0020] These data are mainly based on current national and industry standards and specifications, such as "GB / T21448 Technical Specification for Cathodic Protection of Buried Steel Pipelines" and "SY / T0086 Measurement Method of Cathodic Protection Parameters for Buried Steel Pipelines". They are also combined with long-term engineering experience and electrochemical mechanism literature in the field of cathodic protection. After statistical analysis of historical operation and maintenance data through an expert system, benchmark thresholds are extracted and dynamically corrected by simulation and laboratory test results.

[0021] In terms of data structure, the database adopts a storage model that combines key-value pairs and relational tables. Each type of preset indicator is stored as an independent data entry, including the indicator's unique identifier, indicator name, qualified threshold range, applicable scenario labels such as "underprotection judgment", "data alignment", "feature extraction", data source basis such as referenced standard clause number or expert experience description, and metadata attributes such as version number and effective timestamp, so that the agent can quickly retrieve and match during the reasoning process.

[0022] In terms of storage, the database adopts a hybrid architecture that combines distributed relational databases with memory caching. Frequently accessed preset indicators are kept in memory to improve query response speed. At the same time, persistent storage and transactional management of data are achieved through relational databases, and a B+ tree-based index structure is established to support multi-dimensional fast retrieval by scenario tags and indicator names, ensuring that the intelligent agent can obtain the required qualification criteria in real time and accurately during the decision-making process.

[0023] Furthermore, the specific process of modeling the ontology model of the pipeline cathodic protection operation and maintenance decision-making knowledge base is as follows: The device topology matching degree, parameter semantic recognition accuracy, and causal logic correctness are obtained and compared with the corresponding qualified standard values ​​to determine whether the ontology model construction is qualified. The total number of physical entity connections defined in the ontology model and the total number of physical entity connections in the cathodic protection management system are obtained. At the same time, the number of entity connections in the ontology model that are completely consistent with the actual connection relationship, connection method, and hierarchical association is counted. The ratio of the total number of physical entity connections defined in the ontology model to the number of physical entity connections in the cathodic protection management system to the total number of actual entity connections is expressed as the device topology matching degree used to quantify the accuracy of the ontology model in restoring the physical association of field equipment.

[0024] Specifically, the equipment layout diagram and connection diagram of the pipeline cathodic protection management system are retrieved through the GIS geographic information system to extract the connection relationships, connection methods and hierarchical associations of the actual physical entities, and to count the total number of actual entity connections. At the same time, the connection information of all physical entities defined in the model is exported through the ontology model editing terminal, and the total number of entity connections in the model and the number of connections that are completely consistent with the actual situation are counted. If the connection type (such as series or parallel), connection node and hierarchical level of the physical entities in the ontology model are completely consistent with the actual situation on site, they are judged as consistent connections; otherwise, they are judged as inconsistent connections.

[0025] Determine whether the equipment topology matching degree is greater than or equal to the preset equipment topology matching degree set by the preset personnel. If so, send a qualified equipment topology matching prompt to the preset personnel. Otherwise, obtain the parameter semantic recognition accuracy used to quantify the consistency of the core electrochemical parameters of cathodic protection in the ontology model and the causal logic correctness used to characterize the consistency between the causal logic relationship in the ontology model and the actual electrochemical mechanism on site.

[0026] The specific process for obtaining the accuracy of parameter semantic recognition and the correctness of causal logic is as follows: Obtain the total number of core electrochemical parameters for cathodic protection defined in the ontology model; obtain the number of parameters whose semantic interpretation is completely consistent with the preset judgment criteria definition obtained from the database, and express the ratio of the corresponding number to the total number of core electrochemical parameters for cathodic protection as the parameter semantic recognition accuracy.

[0027] The process involves retrieving core electrochemical parameters for cathodic protection through data acquisition terminals, such as SCADA systems, and determining the total number of core parameters. Semantic explanations of each parameter in the ontology model are obtained based on SPARQL, compared with the parameter definitions in the preset judgment criteria, and the number of parameters with completely consistent semantic explanations is counted. If the name, unit, and definition range of a parameter in the ontology model are completely consistent with the preset judgment criteria, it is determined that the semantic explanation is consistent; otherwise, it is determined that it is inconsistent.

[0028] Obtain the total number of causal relationships defined in the cathodic protection domain in the ontology model; obtain the number of causal relationships in the ontology model that are completely consistent with the actual cathodic protection electrochemical mechanism, and express the ratio of the corresponding number to the total number of causal relationships in the cathodic protection domain as the causal logic correctness rate.

[0029] The study involved collecting data on the influence relationships of various parameters in the cathodic protection management system using an electrochemical tester and a soil resistivity tester. Combined with industry mechanistic literature, the total number of causal relationships in the cathodic protection field was statistically analyzed. Furthermore, each set of causal relationships defined in the ontology model was verified using a fault diagnosis instrument, and the number of matching causal relationships was statistically analyzed. A causal relationship refers to the logical connection in the cathodic protection management system where a change in one entity, such as an interference source, environmental factors, or equipment status, or the occurrence of an event, directly leads to corresponding changes in another entity, such as pipeline potential, anti-corrosion layer integrity, or equipment operating status. This includes causal chains based on electrochemical mechanisms, such as interference sources causing potential fluctuations, soil resistivity affecting IR drop, anti-corrosion layer damage causing underprotection, and aging of the anode bed leading to a decrease in output current.

[0030] The accuracy rate of parameter semantic recognition and the correctness rate of causal logic are compared with the corresponding qualified standard values. If the accuracy rate of parameter semantic recognition is greater than or equal to the preset accuracy rate of parameter semantic recognition, and the correctness rate of causal logic is greater than or equal to the preset correctness rate of causal logic, it indicates that the ontology model for cathodic protection is qualified and cathodic protection data fusion and correlation can be carried out. The preset accuracy rate of parameter semantic recognition is represented by the average of the accuracy rates of parameter semantic recognition over a historical period, and the preset correctness rate of causal logic is represented by the average of the correctness rates of causal logic over a historical period. Conversely, if they are not, it indicates that the ontology model for cathodic protection is unqualified and needs to be corrected.

[0031] In this embodiment, the process quantitatively evaluates the accuracy of the ontology model in three dimensions: physical structure mapping, parameter semantic uniformity, and causal logic fit. This ensures that the constructed ontology model for the field of cathodic protection has the ability to reproduce the operating rules of the real system with high fidelity, providing a solid and reliable professional knowledge framework for the agent's subsequent reasoning and decision-making.

[0032] Furthermore, the specific process for revising the constructed ontology model is as follows: Based on the spatial positioning function of the GIS system and combined with the cathodic protection data collected by the inspection instrument, the physical connection edges in the ontology model are automatically corrected, missing equipment relationships are supplemented, and the topology matching degree is improved.

[0033] Specifically, by comparing the spatial coordinates of equipment in the GIS geographic information system with the field equipment connection photos and connection node data collected by the pipeline cathodic protection management system's inspection instrument, a three-dimensional comparison is performed on the equipment spatial coordinates (X / Y / Z) in the GIS system and the relative positional relationships of the equipment in three-dimensional space. The cathodic protection management system automatically identifies connection relationships in the model that do not match reality, generates correction instructions, and synchronously updates the topology structure of the ontology model; based on NLP (Natural Language Processing). The Natural Language Processing (NLP) model automatically calibrates the semantics of parameters in the ontology model, eliminating erroneous interpretations and ensuring consistency with preset judgment criteria. The parameter definitions in the preset judgment criteria are compiled into a dedicated corpus, which is then input into the NLP model for fine-tuning and training, resulting in a dedicated semantic recognition model for cathodic protection. This model is used to calibrate the semantic interpretations of all parameters in the ontology model one by one, automatically correcting interpretations that do not conform to the standards. Simultaneously, a manual verification step is introduced to sample and check the calibrated parameter semantics. Based on electrochemical mechanism literature and field measurement data, the causal relationships in the ontology model are verified and corrected, eliminating logical relationships that do not conform to the mechanism and supplementing missing causal associations. The connection node data refers to the structured data of the actual physical connection relationships between field equipment, such as test piles, anodes, and reference electrodes, including equipment number, connection type, connection location, interface status, and docking relationship.

[0034] For example, entity category definition: Create three major categories of core node tags to cover all elements of the cathodic protection management system: Physical entities: pipelines, potentiostats, anode ground beds, reference electrodes, test piles, insulating joints, solid decouplers, etc.; Environmental entities: soil resistivity, interference sources (attributes: DC / AC, frequency), meteorological conditions (temperature, humidity), etc.; Event entities: underprotection, overprotection, corrosion protection layer damage, equipment offline, AC interference, DC interference, etc.

[0035] Specific semantic relationship definition: Construct three types of core edges to restore the physical mechanism and logical association of the cathodic protection management system: Physical connection edge: (potential constant) - [power supply] -> (anode ground bed), (reference electrode) - [monitoring] -> (pipeline); Causal influence edge: (interference source) - [cause] -> (potential fluctuation), (soil resistivity) - [influence] -> (IR drop), (corrosion protection layer damage) - [cause] -> (underprotection); Spatial mapping edge: (test pile) - [located in] -> (GIS coordinates), (pipeline segment) - [contains] -> (test pile).

[0036] Core attribute feature assignment: The key electrochemical parameters of cathodic protection are defined as nodal numerical attributes, which makes it easy for subsequent intelligent agents to directly call the values ​​for calculation.

[0037] It should be added that, such as Figure 3 The diagram shown is an ontology modeling architecture diagram of an intelligent agent construction method for a pipeline cathodic protection operation and maintenance decision-making knowledge base provided by an embodiment of the present invention. The method involves obtaining the equipment topology matching degree, determining whether the equipment topology matching degree is greater than or equal to a preset equipment topology matching degree, and if so, sending a qualified equipment topology matching prompt to a preset personnel. If not, it involves obtaining the parameter semantic recognition accuracy rate and causal logic correctness rate, and determining whether both the parameter semantic recognition accuracy rate and causal logic correctness rate are greater than or equal to the corresponding qualified standard value. If not, the constructed ontology model is corrected; if so, the cathodic protection data is fused and associated.

[0038] In this embodiment, this step is used to ensure that the ontology model can accurately reproduce the real structure and operating mechanism of the cathodic protection system, thereby fundamentally ensuring that the subsequent reasoning and decision-making based on the model has a high degree of credibility and accuracy.

[0039] Furthermore, the specific process for integrating and associating the negative data is as follows: Based on the alignment accuracy, feature extraction accuracy, and association success rate of the CAMP data obtained after preprocessing, it is determined whether to implement fusion strategy optimization to improve the fusion effect of CAMP data.

[0040] The specific process of preprocessing negative pressure protection data is as follows: The potential and current data of the cathodic protection management system are timestamped; the inspection logs, such as PDFs and handwritten record images, are subjected to OCR text recognition, and NLP models are used for entity recognition and word segmentation to extract time features, spatial features and event features.

[0041] Time features refer to time description information related to the time of occurrence of operation and maintenance events, which is identified and extracted from unstructured data (such as inspection logs, maintenance records, and handwritten reports) through NLP models. Specifically, it includes absolute time points, such as "March 10, 2025, 10:30", relative time expressions, such as "30 minutes after the inspection was completed", or time periods, such as "during night duty". After standardization, it is converted into a unified timestamp format.

[0042] Spatial features refer to spatial descriptive information related to the location of operation and maintenance events that is identified and extracted from unstructured data through NLP models. Specifically, these include explicit equipment numbers, such as "TP-108 test pile", mileage markers, such as "K102+500", geographical area descriptions, such as "crossing river section" or "third work area", or relative location descriptions, such as "500 meters downstream of the potentiostat". After coordinate transformation by the GIS system, these features are mapped to specific spatial coordinate points or spatial ranges and used as spatial dimension identifiers in the "spatiotemporal anchor point" mechanism.

[0043] Event features refer to the core semantic information identified and extracted from unstructured data through NLP models, describing abnormal states, fault phenomena, or operation and maintenance activities that occur during the operation of a cathodic protection system. Specifically, they include event types, such as "underprotection," "corrosion protection layer damage," and "equipment offline," event severity, such as "emergency alarm" and "general defect," event descriptive elements, such as "abnormal potential fluctuations" and "reference electrode failure," and associated equipment or environmental entities, such as "affected by AC interference" and "sudden increase in soil resistivity." After standardization, these features are transformed into structured event entity nodes.

[0044] Geographic data association establishes a mapping relationship between GIS coordinates and test pile numbers and pipeline section mileage to support spatial range retrieval.

[0045] The specific process of establishing the mapping relationship is as follows: Extract the design coordinates of the test piles and the design trajectory of the pipeline centerline from the design drawings or as-built data of the cathodic protection management system. Simultaneously, use a handheld GPS (Global Positioning System) terminal or Beidou positioning device to conduct on-site measurements of the test pile installation locations and pipeline direction, obtaining real-time coordinate data with timestamps. For buried pipeline sections, a pipeline detector can be used to track the pipeline direction on the ground, collecting pipeline centerline coordinate points at regular intervals to form a continuous pipeline trajectory line. Convert the WGS84 latitude and longitude coordinates into plane projection coordinates suitable for length calculation to spatially register with the design coordinates. Correct system deviations using algorithms such as the least squares method. After coordinate correction, define the pipeline centerline as a linear reference line, with the starting mileage set as the design starting point, following the pipeline direction. The mileage is incremented by one meter, and the collected pipeline centerline coordinate points are connected sequentially to form a broken line. Each broken line segment is assigned a mileage attribute. For the measured coordinate point of each test pile, its perpendicular foot point on the pipeline centerline broken line is calculated. Then, the cumulative length along the pipeline from the starting point to the perpendicular foot point is obtained to get the pipeline mileage value of the test pile. The calculated mileage value is compared and calibrated with the design mileage or the on-site marked mileage to ensure that the deviation is within the allowable range. Finally, the test pile number, GIS coordinates, pipeline segment number, pipeline mileage, coordinate source, collection time and other information are stored in a database table or knowledge graph to form a structured mapping relationship.

[0046] Spatiotemporal anchor point establishment: Anchor point definition: The combination of "time T-space L" is used as the core anchor point; Threshold configuration: Set the time window Δt and the spatial radius r; Cross-modal association mapping: Data retrieval: The cathodic protection management system automatically retrieves potential / current time series data fragments with time in the range of [T-Δt, T+Δt] and location in the L radius r based on the anchor point (T,L); Knowledge graph association: Generate "interpretive association" edges in the ontology model to establish association between "event nodes" in unstructured logs, such as "corrosion layer damage", and structured "cathodic protection data waveform nodes" and "spatial nodes"; Causal index construction: Automatically record the association relationship of "event-data-location" to form a fault causal knowledge base, supporting subsequent reverse retrieval, such as quickly matching historical events through similar potential waveforms. Among them, the cathodic protection data waveform node refers to a sampling point in the potential or current waveform data arranged in time series collected in real time or at regular intervals from cathodic protection monitoring equipment, such as intelligent test piles, which is the smallest data unit that constitutes the entire waveform.

[0047] The cathodic protection data alignment accuracy is defined as the reciprocal of the time deviation and the reciprocal of the spatial deviation after the cathodic protection data is aligned based on spatiotemporal anchor points. This is used to quantify the degree of matching of cathodic protection data in the spatiotemporal dimensions. Structured time-series data includes continuous monitoring data with timestamps, such as potential, current, and potentiostat output parameters collected by the SCADA system. Unstructured logs include text data such as inspection reports in PDF / Word format, handwritten maintenance records, and equipment maintenance logs. GIS geospatial data includes geographic data with spatial attributes, such as pipeline segment coordinates, test pile locations, anode bed distribution, and soil environmental zoning. The reciprocal of the time deviation is obtained by calibrating the timestamps of the cathodic protection management system through the SCADA system time synchronizer, selecting the timestamps of the cathodic protection data corresponding to the same maintenance event, calculating the absolute value of the time difference between any two types of data, and taking the reciprocal of the maximum value as the reciprocal of the time deviation.

[0048] The spatial coordinates of physical entities or record points corresponding to each data source related to the same maintenance event are collected by a GIS locator. The straight-line distance between the record locations of any two types of data sources is calculated, and the reciprocal of the maximum value is taken as the reciprocal of the spatial deviation.

[0049] For example, suppose there is an abnormal pipeline potential. SCADA data shows: 10:00:00, equipment K102 (coordinate A) has an abnormal potential. The inspection log records: 10:02:00, inspector Xiao Li recorded "smelled an odor, suspected leak" at coordinate B. GIS data shows: there is a third-party construction marker at coordinate C. The accuracy of cathodic protection feature extraction is defined as the percentage of key features (including time, space, event type, etc.) correctly extracted from unstructured data out of the total number of features actually extracted. The success rate of cathodic protection data association is defined as the percentage of cathodic protection data groups that successfully establish associations based on spatiotemporal anchor points out of the total number of cathodic protection data groups that attempt association.

[0050] The number of data sets for cloud protection refers to the number of multi-source heterogeneous cloud protection data combinations participating in cross-modal association matching, using the same spatiotemporal anchor point as the retrieval center. Each set contains at least two types of data from structured time-series data, unstructured log data, and GIS geospatial data. When the cloud protection data alignment accuracy, cloud protection feature extraction accuracy, and cloud protection data association success rate are all greater than preset values, the fusion and association quality of the multi-source heterogeneous data is considered acceptable, and the process can proceed to the next step of compliance constraint construction. Conversely, if any one of the cloud protection data alignment accuracy, cloud protection feature extraction accuracy, or cloud protection data association success rate is less than or equal to the corresponding value... If the data fusion and association quality does not meet the required standards, it indicates that the data fusion and association quality has not met the requirements and calibration of the data fusion quality is necessary. The data fusion standard parameters refer to the preset data alignment accuracy, preset data feature extraction accuracy, and preset data association success rate. These are all reference values ​​obtained by the preset personnel based on historical experience and through repeated experiments. The preset data alignment accuracy includes the preset reciprocal value of time deviation and the preset reciprocal value of spatial deviation. If the data fusion alignment accuracy is greater than the preset data alignment accuracy, that is, the preset reciprocal value of time deviation is greater than or equal to the preset reciprocal value of spatial deviation, the preset reciprocal value of spatial deviation is greater than or equal to the preset reciprocal value of spatial deviation.

[0051] It should be added that, such as Figure 4 The diagram shown is an architecture diagram of cathodic protection data fusion and association in an intelligent agent construction method for a pipeline cathodic protection operation and maintenance decision-making knowledge base provided by an embodiment of the present invention. The method involves fusing and associating cathodic protection data, obtaining the alignment accuracy, feature extraction accuracy, and association success rate of the cathodic protection data, and determining whether all three parameters are greater than the corresponding standard parameters for the cathodic protection data. If not, the quality of the cathodic protection data fusion is evaluated and calibrated; if so, compliance constraints are constructed.

[0052] In this embodiment, the process precisely aligns and merges fragmented structured time-series data, unstructured logs, and geospatial data into a cross-modal associated entity with a unified spatiotemporal reference. This provides the agent with a complete, traceable, and highly reliable holographic fault context for subsequent anomaly perception and reasoning decisions, fundamentally improving data utilization efficiency and fault diagnosis accuracy.

[0053] Furthermore, the quality of the Yinbao data fusion was evaluated and calibrated. The specific process is as follows: the Yinbao data was double-corrected based on BeiDou time synchronization and GIS spatial calibration.

[0054] Specifically, this refers to real-time calibration of the timestamps of the SCADA system, inspection equipment, and GIS system based on the time synchronization signal of the BeiDou positioning system to correct the inverse value of time deviation. Real-time coordinate data collected by the GIS positioning instrument is combined with the equipment layout diagram of the pipeline cathodic protection management system, and a coordinate interpolation algorithm is used to correct the spatial coordinates of the cathodic protection data to compensate for the inverse value of spatial deviation. Based on a preset calibration cycle, generally every 60 seconds, the alignment accuracy of the cathodic protection data is reacquired.

[0055] It should be explained that real-time coordinate data refers to dynamic location data with timestamps and spatial latitude and longitude information, which is obtained on-site at the moment of data collection through GIS positioning instruments, Beidou positioning terminals or mobile inspection equipment. It is used to reflect the precise spatial location of physical entities or operation and maintenance events in the cathodic protection system, such as inspection records and fault points at the current moment.

[0056] Semantic correction is performed on the identified text based on a specialized vocabulary database for cathodic protection, and erroneous features are corrected. Erroneous features refer to issues such as terminological errors, expression deviations, and semantic ambiguities in the identified text, including but not limited to time, space, and event type. Based on a graph database-based association indexing method, the retrieval strategy is optimized according to time windows and spatial radii, automatically matching different types of cathodic protection data within the same spatiotemporal range, and introducing a historical association case library to supplement missing association relationships by referring to data association patterns in similar scenarios. The historical association case library refers to a collection of instances that record successful association patterns of multi-source heterogeneous data, formed through manual confirmation or intelligent agent verification during the historical operation of the cathodic protection management system.

[0057] Specifically, optimizing the retrieval strategy means setting dynamic time windows based on cathodic protection data, such as pipeline inspection cycles, potential monitoring frequency, and fault handling timeliness. For example, for potential anomaly alarm data, the time window is narrowed to 1 hour before and after the alarm occurs, while for daily inspection data, it is widened to the same day. At the same time, the window boundaries can be dynamically adjusted according to the time correlation patterns of similar scenarios in the historical case library, such as pipeline corrosion problems in a certain area often occurring within 3 days after rain, to filter data from irrelevant time periods.

[0058] Centered on the target retrieval object, such as the station number of a certain pipeline or the location of the anode bed, a hierarchical spatial radius is set in combination with the pipeline route, soil environment, and cathodic protection system topology. The core retrieval range is 50 meters around the target location, and the extended range is 200 meters. At the same time, discrete data with no topological relationship in space, such as irrelevant anode data across regions, are removed.

[0059] By introducing matching rules from a historical case database, verified association patterns from the cases are embedded into the retrieval strategy. Priority is given to matching data types with high frequency associations in the cases, and missing association dimensions are supplemented. For example, when only potential data is retrieved, the retrieval of soil resistivity and current output data within the same spatial range is automatically triggered, and retrieval paths marked as invalid associations in the cases are removed, thereby improving retrieval accuracy and efficiency.

[0060] Determine whether the alignment accuracy, feature extraction accuracy, and association success rate of the newly acquired cybersecurity data after evaluation and calibration of the cybersecurity data fusion quality are all greater than the corresponding cybersecurity data standard parameters. If so, construct compliance constraints; otherwise, send a data fusion anomaly alert to the designated personnel.

[0061] In this embodiment, this step provides a complete, traceable, and highly reliable holographic fault context for the agent's subsequent anomaly perception and reasoning decisions, fundamentally improving data utilization efficiency and fault diagnosis accuracy.

[0062] Furthermore, the specific process for constructing compliance constraints is as follows: The potential simulation deviation is obtained and compared with the corresponding preset potential simulation deviation to obtain a consistency index for quantifying the physical feasibility of decision recommendations. The potential simulation deviation refers to the absolute value of the difference between the pipeline potential value calculated by the physical mechanism simulation model and the actual monitored pipeline potential value, which is used to quantify the physical feasibility of decision recommendations. It is then determined whether the potential simulation deviation is less than or equal to the preset potential simulation deviation, which is represented by the average value of potential simulation deviations over historical time periods. If so, a closed-loop intelligent decision-making and feedback update based on the thought chain is carried out; otherwise, physical feasibility optimization is implemented to ensure the feasibility of decision recommendations.

[0063] In this embodiment, the process enables decision recommendations to accurately match the actual on-site environment, thereby further strengthening the engineering safety bottom line of decision-making on the basis of compliance constraints, and ensuring that every instruction output by the intelligent agent has both theoretical correctness and on-site executability.

[0064] Furthermore, the specific process for optimizing physical feasibility is as follows: Based on the physical mechanism simulation model running on the simulation server, the initial decision suggestions generated by the agent are iteratively verified, the physical boundaries of the decision parameters are standardized, the decision outputs that violate the electrochemical principles are avoided, and the model input parameters are optimized to adapt to changes in actual working conditions.

[0065] Specifically, the initial decision suggestions generated by the intelligent agent, such as "adjust the energizing potential to -1.30V", are first used as input, along with cathodic protection data obtained from the SCADA system, such as real-time soil resistivity and pipeline coating resistivity, and then input into the physical mechanism simulation model.

[0066] It should be added that the construction process of the physical mechanism simulation model is as follows: First, basic engineering data for the target pipeline is collected, including pipeline material, diameter, wall thickness, type and performance parameters of anti-corrosion coating, anode type and location, and soil or water resistivity. This data forms the foundational data source for model construction. Based on this, finite element or finite difference numerical methods are used to construct a geometric and meshed model consistent with the actual pipeline structure, reconstructing the spatial topological relationships of physical entities such as pipelines, anodes, media, and structures. Then, the electrochemical polarization equations, namely the Butler–Volmer equations, are applied. Physical mechanisms such as equations, dielectric conductivity equations, and Ohm's law conduction laws are embedded in the model to form corresponding physical field control equations. Pre-set acceptable potential ranges, hydrogen evolution critical potentials, and protection potential thresholds are written into the model's decision logic as rigid constraints. Then, the simulation model is calibrated and corrected using on-site measured potential, current, and dielectric parameters. By adjusting key parameters such as polarization curves, dielectric resistivity, and corrosion layer leakage resistance, the simulation results are matched with on-site measured data. Finally, the simulation results under multiple different operating conditions are compared with the measured results to verify the calculation accuracy, convergence, and stability of the physical mechanism simulation model. This ensures that the physical mechanism simulation model can accurately reflect the potential distribution law and physical change process of the pipeline cathodic protection system, ultimately forming a physical mechanism simulation model that can be directly used for decision simulation iteration and has clear inputs, outputs, and calculation logic.

[0067] The specific process of running the physical mechanism simulation model is as follows: Using real-time pipeline operating parameters, cathodic protection system operating parameters, environmental medium parameters, and structural geometric parameters as unified inputs, the physical mechanism simulation model quantitatively solves the potential distribution and values ​​at each monitoring point and key node along the pipeline under the influence of current decision parameters through its built-in potential field solution equations and numerical calculation methods. Then, the simulation results are automatically compared with the standard acceptable range. If the simulation results show that the potential at a key location exceeds the acceptable range, such as being below the hydrogen evolution critical point of -1.20V, the current decision recommendation is deemed to have an overprotection risk and its physical feasibility is deemed unacceptable. At this point, the physical mechanism simulation model will automatically iteratively correct the control parameters in the decision recommendation according to constraint rules, such as gradually adjusting the output current and optimizing the output voltage. The corrected parameters are then substituted back into the mechanism model to conduct a new round of simulation calculations, repeating the closed-loop iterative process of potential solution, compliance judgment, and parameter fine-tuning until the simulated potentials at all key locations along the entire pipeline stably fall within the standard acceptable range. Finally, an optimized decision recommendation that meets both the rigid constraints of industry standards and has physical feasibility is output. Real-time parameters of pipeline operation include: type of conveyed medium, temperature, pressure, flow rate, flow velocity, moisture content, sand content, pipe wall temperature, etc. Cathodic protection system operating parameters include: pipeline natural potential, energized potential, de-energized potential, protection current, current density, constant potential meter output voltage, anode grounding resistance, insulation resistance, etc. Environmental media parameters include: soil resistivity, pH value, oxidation-reduction potential, chloride ion content, groundwater level, ambient temperature, moisture content, microbial content, etc. Structural geometric parameters include: pipe diameter, wall thickness, length, burial depth, pipe material, anti-corrosion layer thickness and type, elbows, tees, valve structural dimensions, etc.

[0068] For example, in an area with high soil resistivity, the agent initially suggests adjusting the output voltage of the potentiostat to 50V. However, after calculation by the physical mechanism simulation model, it is found that this will cause the potential of the remote pipeline to drop below -1.20V. The model then iteratively optimizes the process, gradually reducing the output voltage to 42V. After resimulation, the remote potential recovers to a qualified state of -0.95V, thus outputting a recommendation. Based on the optimized decision recommendation, a closed-loop intelligent decision-making and feedback update based on the thought chain is carried out.

[0069] In this embodiment, this process is used to ensure that the intelligent agent can continuously evolve and reliably support the intelligent decision-making needs of the cathodic protection system in the actual operation and maintenance environment.

[0070] Furthermore, the specific process of closed-loop intelligent decision-making and feedback updates based on the thought chain is as follows: The quality of the agent construction for the pipeline cathodic protection operation and maintenance decision-making knowledge base is judged by acquiring the fault location accuracy, decision execution success rate, and knowledge base update efficiency. The fault location accuracy is quantified by the percentage of cases that perfectly match the actual fault description determined by the agent based on closed-loop reasoning, out of the total number of inferred fault cases. The decision execution success rate is quantified by the percentage of operation and maintenance decision instructions that successfully resolve pipeline cathodic protection faults, out of the total number of output instructions. The speed at which successful cases and fault causes are updated to the knowledge base after decision execution is also measured. The knowledge base update efficiency is used to quantify the adaptive optimization capability of the intelligent agent. The fault location accuracy is compared with the corresponding qualified standard value, the decision execution success rate is compared with the corresponding qualified standard value, and the knowledge base update efficiency is compared with the corresponding qualified standard value. The qualified standard value for fault location accuracy is greater than or equal to 90%, the qualified standard value for decision execution success rate is greater than or equal to 85%, and the knowledge base update efficiency is required to be greater than or equal to 10 items / hour. Among them, the thought chain is the structured, traceable, multi-step progressive logical reasoning sequence followed by the intelligent agent when making pipeline cathodic protection operation and maintenance decisions, from initial observation information, problem identification, rule or knowledge matching, reasoning and deduction, scheme generation to decision output.

[0071] Within a preset monitoring period, typically one month, if the fault location accuracy, decision execution success rate, and knowledge base update efficiency all meet the corresponding qualification standards, it indicates that the constructed intelligent agent possesses accurate, effective, and continuously optimized decision-making capabilities and can be deployed in actual operation and maintenance applications. A qualified construction notification for the pipeline cathodic protection intelligent agent will be sent to the preset personnel. Conversely, if the fault location accuracy, decision execution success rate, and knowledge base update efficiency all meet the corresponding qualification standards, it indicates that the intelligent agent's closed-loop decision-making and learning capabilities are deficient and the reasoning logic or feedback mechanism needs to be optimized.

[0072] In this embodiment, the process enhances the accuracy, feasibility, and adaptive optimization capabilities of the intelligent agent's decision-making, realizes the standardization and efficiency of the intelligent agent's operation and maintenance decision-making, promotes the intelligent upgrade of pipeline cathodic protection operation and maintenance, and allows substandard intelligent agents to improve their performance through optimization, thus ensuring the reliability of operation and maintenance decisions.

[0073] Furthermore, the specific process for optimizing the reasoning logic or feedback mechanism is as follows: Based on the iterative verification of the physical mechanism simulation model, simulation boundary condition constraints are added, such as dynamic changes in soil resistivity and aging factors of pipeline coatings, to generate preliminary solutions. The preliminary solutions are then subjected to a preset number of simulation rounds, generally 3-5 rounds, until the simulated potential distribution falls completely within the acceptable range. The preliminary solution refers to the initial decision-making suggestion that has not been verified by physical mechanism simulation and compliance with regulations and rules. It includes specific operational instructions for the current fault, such as adjusting the output current and voltage values ​​of the potentiostat, the expected operating parameter targets, such as the target energized potential, and the recommended maintenance actions, such as replacing the reference electrode and excavating to inspect the anti-corrosion layer.

[0074] It should be added that the specific implementation method for adding simulation boundary condition constraints is as follows: Dynamic boundary parameters under current operating conditions are obtained from the real-time database or historical trend database of the cathodic protection management system. These parameters include soil resistivity change curves transmitted in real time by a soil resistivity monitor, coating aging degradation factors calculated based on pipeline service life and coating type, and the intensity and frequency of interference sources collected by a stray current monitoring device. These dynamic boundary parameters are then loaded as additional input conditions into the physical mechanism simulation model constructed based on the boundary element method, replacing the original static default values ​​and making the simulation environment closer to the actual field conditions. Subsequently, preliminary decision suggestions generated by the intelligent agent, such as adjusting the output current or voltage of the potentiostat, are input into the simulation model along with the aforementioned dynamic boundary parameters. The model then begins to calculate the potential distribution along the pipeline under the current constraints. If the initial simulation results show that the potential at critical locations, such as the far end of the pipeline, the damaged point of the anti-corrosion layer, or the insulating joint, exceeds the acceptable range, such as being lower than the hydrogen evolution critical point of -1.20V or higher than the underprotection critical point of -0.85V, the system will automatically trigger an iterative optimization process: according to the preset step size rules, such as adjusting the key parameters in the decision suggestion by 0.5A of output current or 1V of output voltage each time, and re-substitute them into the simulation model for calculation. After each simulation, the results will be compared with the acceptable range.

[0075] After completing the above steps, the closed-loop intelligent decision-making and feedback update based on the thought chain is re-entered, and the fault location accuracy, decision execution success rate and knowledge base update efficiency are re-evaluated in the next monitoring cycle until all three indicators meet the corresponding qualification standards or reach the preset maximum number of optimizations, which is generally 3 times. If the standards are still not met after the maximum number of optimizations, a prompt for manual reconstruction of the intelligent agent is sent to the preset personnel for in-depth intervention and reconstruction of the cathodic protection management system.

[0076] In this embodiment, the process can ensure that the agent can converge to the qualified standard within a preset optimization period, thereby providing a stable and reliable decision output for the cathodic protection management system.

[0077] Example 2: Building upon Example 1, in the actual operation and maintenance of pipeline cathodic protection, statistical AI models are easily interfered with, lacking the "counterfactual inference" capabilities of experts, and unable to perform quantitative deductions such as "how much the potential will recover if a certain interference source is forcibly isolated." Secondly, SCADA data from older pipelines often involves low-frequency sampling and high packet loss rates. Purely data-driven models may be subject to interference under poor-quality data and struggle to bridge the semantic gap between numerical waveforms and text logs. Finally, large single-unit models are prone to outputting control commands that violate physical safety regulations under complex operating conditions. Industrial field edge control cabinets do not allow the operation of complex Python code environments. Without a functional safety interlocking mechanism based on physical hardware, illegal pressure regulation commands generated by cloud-based AI, such as those causing extreme voltages that lead to hydrogen stripping of corrosion-resistant coatings, could trigger accidents. The entire process requires the intelligent agent to repeatedly log in, input, and query between different systems. Any omission in information transmission at any stage can lead to process interruption or decision failure. The specific process of closed-loop intelligent decision-making and feedback updates based on the thought chain is as follows: The average response time of the cathodic protection management system within a preset coordination period is obtained by the system operation monitoring device and expressed as the response delay of each agent to reflect the real-time performance of cross-system collaborative invocation. It is determined whether the response delay of each agent is less than or equal to the preset response delay threshold. If the response delay of all systems is less than or equal to the preset response delay threshold, it indicates that the cross-system collaboration is in good condition and the agents can distribute tasks and synchronize their status according to the established process. Otherwise, it indicates that there is an anomaly in the construction of the agents, and optimization of the inference logic or feedback mechanism needs to be triggered.

[0078] The specific process for optimizing the reasoning logic or feedback mechanism is as follows: First, a directed acyclic graph (DAG) is constructed in the cloud-based graph database, containing causal variables X (potential constant output), outcome variables Y (pipe-to-soil potential), and confounding factors U (such as dynamic soil resistivity). To reduce subjective biases that may arise during the predefined process, the system introduces a dynamic causal discovery algorithm based on continuous optimization. This algorithm transforms the traditional discrete graph search, which focuses on finding connections between nodes, into a continuous algebraic matrix constraint optimization problem. By minimizing the reconstruction loss of historical SCADA observation data, it automatically solves for the causal adjacency matrix between nodes. During this process, the system uses the Pipeline Standard & Instrumentation Diagram (P&ID) as a hard prior constraint input, strictly suppressing the generation of any pseudo-causal edges that violate physical topology.

[0079] Secondly, in the fusion of waveform logic propositions and log propositions extracted by NLP, the system uses the shortest physical connectivity path based on geographic information systems to complete spatial distance matching, and introduces a dynamic time warping algorithm in the time dimension. This algorithm minimizes the warping cost to offset the polarization delay and waveform distortion caused by stray currents during pipeline propagation, thereby generating a high-fidelity joint causal clue pool, providing reliable data support for subsequent inference.

[0080] Subsequently, the compliance rules verified by SMT in the cloud are automatically generated into structured text underlying code conforming to the IEC61131-3 standard through a model-driven industrial compiler, and then sent to the field programmable logic controller for execution, realizing the practical application of the compliance rules.

[0081] Finally, the system utilizes a causal retrieval enhancement mechanism to achieve knowledge self-evolution: the verified fault symptoms, identified true confounding factors, and successful voltage regulation commands are transformed into text sequences, mapped into high-dimensional semantic feature vectors through a text embedding model, and used as self-evolving knowledge units of the verified causal chain, which are then mounted onto the corresponding physical entity nodes in a graph database, such as Neo4j. When similar operating conditions occur subsequently, the system directly calls the vector node through cosine similarity retrieval to update the prior probability distribution of the knowledge graph, achieving experience reuse and model iterative optimization.

[0082] like Figure 5The diagram illustrates the workflow of a pipeline cathodic protection operation and maintenance decision-making knowledge base intelligent agent provided in this embodiment of the invention. S100: Construct a multimodal hybrid knowledge base, completing ontology modeling, data alignment, and regulatory constraint construction to provide a domain knowledge foundation for subsequent decision-making. S200: Real-time data perception and anomaly monitoring, accessing multi-source data such as SCADA systems, environmental data, and operation and maintenance logs to conduct real-time perception and anomaly identification of the pipeline cathodic protection status. S300: Determine if intelligent decision-making is triggered. If an anomaly is detected or the decision triggering conditions are met, proceed to the intelligent decision-making stage; otherwise, return to S200 to continue monitoring. If triggered, proceed to S400: Retrieval and reasoning based on the Chain of Thought (CoT), relying on the knowledge base and using the CoT reasoning mechanism to perform in-depth analysis and knowledge retrieval of abnormal scenarios. S500: Generate preliminary operation and maintenance suggestions. Based on the reasoning results, generate preliminary operation and maintenance decision suggestions including parameter adjustments and operation and maintenance dispatch. S600: Dual verification of physical mechanism and regulations. The preliminary solution undergoes physical mechanism simulation (verifying electrochemical feasibility) and compliance verification. If verification fails (e.g., due to violations or overprotection risks), it returns to S400 for re-inference and optimization; if verification passes, it proceeds to the next step. S700: Outputs final decision instructions, including executable final operation and maintenance instructions such as potentiostat parameter adjustment and fault operation and maintenance dispatch. S800: Adaptive feedback update based on decision execution effect, updating knowledge base weights and optimizing model parameters to form a closed-loop iteration, ultimately feeding back to S100 to complete the knowledge base update.

[0083] In this embodiment, an ontology model is constructed for the pipeline cathodic protection operation and maintenance decision-making knowledge base. The constructed ontology model undergoes qualification verification and correction. This helps to build a structurally complete, semantically consistent, and logically rigorous professional knowledge framework for the intelligent agent in the field of cathodic protection. This enables the intelligent agent to accurately understand the intrinsic relationships between physical entities, environmental factors, and fault events, laying a solid cognitive foundation for subsequent intelligent reasoning. The fusion and correlation of cathodic protection data are carried out, and the quality of the fused data is evaluated and calibrated to establish effective cross-modal data correlation. This helps to break down information silos between structured time-series data, unstructured operation and maintenance logs, and geospatial data, providing the intelligent agent with accurate, comprehensive, and traceable decision-making basis. The simulation results of the pipeline cathodic protection operation and maintenance decision-making knowledge base are evaluated, which helps to verify the feasibility and safety of the intelligent agent's output scheme in a real physical environment. This ensures that the decision recommendations not only meet regulatory and standard requirements but also conform to electrochemical mechanisms and field operating condition constraints, fundamentally eliminating secondary risks caused by simulation deviations and guaranteeing the long-term stable operation of the pipeline cathodic protection management system.

[0084] The above-disclosed embodiments are merely some examples of the present invention and should not be construed as limiting the scope of the present invention. Therefore, any equivalent variations made in accordance with the claims of the present invention are still within the scope of the present invention.

Claims

1. A method for constructing an intelligent agent for a pipeline cathodic protection operation and maintenance decision-making knowledge base, characterized in that, The method includes: An ontology model was created for the pipeline cathodic protection operation and maintenance decision-making knowledge base, and the constructed ontology model was validated and corrected for compliance. After the ontology model is successfully modeled, the fusion and correlation of cathodic protection data are carried out, and the quality of cathodic protection data fusion is evaluated and calibrated in order to establish effective correlation of cross-modal data. After the cathodic protection data is successfully integrated and correlated, the simulation results of the pipeline cathodic protection operation and maintenance decision-making knowledge base are evaluated. If the simulation results are satisfactory, closed-loop intelligent decision-making is carried out and the decision results are obtained. The decision effect is then tested and iterated. If the simulation results are unsatisfactory, a prompt requiring manual reconstruction of the intelligent agent is sent to the preset personnel.

2. The method for constructing an intelligent agent for a pipeline cathodic protection operation and maintenance decision-making knowledge base as described in claim 1, characterized in that, The specific process of modeling the ontology model of the pipeline cathodic protection operation and maintenance decision-making knowledge base is as follows: The device topology matching degree, parameter semantic recognition accuracy, and causal logic correctness are obtained respectively, and compared with the corresponding qualified standard values ​​to determine whether the ontology model construction is qualified. Obtain the total number of physical entity connections defined in the ontology model and the total number of physical entity connections in the cathodic protection management system. At the same time, count the number of entity connections that are completely consistent in the ontology model. The ratio of the number of consistent connections to the total number of actual entity connections represents the device topology matching degree. Determine whether the device topology matching degree is greater than or equal to the preset device topology matching degree; If yes, a device topology matching success notification is sent to the designated personnel; otherwise, the accuracy of parameter semantic recognition and the correctness of causal logic are obtained. The specific process for obtaining the semantic recognition accuracy and causal logic correctness of the parameters is as follows: Obtain the total number of core electrochemical parameters for cathodic protection defined in the ontological model; The number of parameters whose semantic interpretation of the core electrochemical parameters of cathodic protection is completely consistent with the definition of the preset judgment criteria is obtained, and the ratio of the corresponding number to the total number of core electrochemical parameters of cathodic protection is expressed as the parameter semantic recognition accuracy. Obtain the total number of causal relationships in the cathodic protection domain defined in the ontology model; The number of causal relationships in the ontology model that are completely consistent with the actual electrochemical mechanism of cathodic protection is obtained, and the ratio of the corresponding number to the total number of causal relationships in the field of cathodic protection is expressed as the causal logic accuracy rate. The accuracy of parameter semantic recognition and the correctness of causal logic are compared with the corresponding qualified standard values. If the accuracy rate of parameter semantic recognition is greater than or equal to the preset accuracy rate of parameter semantic recognition, and the correctness rate of causal logic is greater than or equal to the preset correctness rate of causal logic, then the fusion and association of the protection data will be carried out. Conversely, the constructed ontology model is revised.

3. The method for constructing an intelligent agent for a pipeline cathodic protection operation and maintenance decision-making knowledge base as described in claim 2, characterized in that, The specific process for modifying the constructed ontology model is as follows: By comparing the spatial coordinates of the equipment in the GIS system with the field equipment connection information in the cathode protection data in three dimensions, the system automatically identifies connection relationships in the ontology model that do not match the actual physical connections, generates correction instructions, synchronously updates the topology of the ontology model, and supplements the missing equipment association relationships. Based on the NLP model, the semantics of parameters in the ontology model are automatically calibrated to eliminate erroneous interpretations and ensure consistency with the preset judgment criteria. The parameter definitions in the preset judgment criteria are compiled into a dedicated corpus, which is then input into the NLP model for fine-tuning and training to obtain a dedicated semantic recognition model for the field of cathodic protection. Based on this dedicated semantic recognition model, the semantic interpretation of all parameters in the ontology model is calibrated one by one, and interpretations that do not conform to the standard are automatically corrected.

4. The method for constructing an intelligent agent for a pipeline cathodic protection operation and maintenance decision-making knowledge base as described in claim 2, characterized in that, The specific process for integrating and associating the negative energy protection data is as follows: Based on the alignment accuracy, feature extraction accuracy, and association success rate of the data obtained after data preprocessing, it is determined whether to implement fusion strategy optimization to improve the fusion effect of data. The alignment accuracy of the data is defined as the reciprocal of the time deviation and the reciprocal of the spatial deviation after the data is aligned based on the spatiotemporal anchor points. It is used to quantify the degree of matching of the data in the spatiotemporal dimension. The method for obtaining the reciprocal value of the time deviation is as follows: calibrate the timestamp of the cathodic protection management system, select the timestamp of the cathodic protection data corresponding to the same operation and maintenance event, calculate the absolute value of the time difference between any two types of data, and take the reciprocal value of the maximum value as the reciprocal value of the time deviation. Collect the spatial coordinates of physical entities or record points corresponding to each data source related to the same operation and maintenance event, calculate the straight-line distance between the record locations of any two types of data sources, and take the reciprocal of the maximum value as the reciprocal of the spatial deviation. The accuracy of feature extraction is defined as the percentage of key features correctly extracted from unstructured data out of the total number of features actually extracted. The success rate of data association for cloud protection is defined as the percentage of cloud protection data groups that successfully establish associations based on spatiotemporal anchors out of the total number of cloud protection data groups that attempt to be associated. When the alignment accuracy of the cybersecurity protection data is greater than the preset alignment accuracy of the cybersecurity protection data, the accuracy rate of the cybersecurity protection feature extraction is greater than the preset accuracy rate of the cybersecurity protection feature extraction, and the success rate of the cybersecurity protection data association is greater than the preset success rate of the cybersecurity protection data association, the next step of compliance constraint construction can be carried out. When any one of the following parameters—account alignment accuracy, account feature extraction accuracy, and account association success rate—is less than or equal to the corresponding account data standard parameter, the account data fusion quality is calibrated.

5. The method for constructing an intelligent agent for a pipeline cathodic protection operation and maintenance decision-making knowledge base as described in claim 4, characterized in that, The specific process for calibrating the quality of cathodic protection data fusion is as follows: The data is double-corrected based on BeiDou time synchronization and GIS spatial calibration. Based on the time synchronization signal of the Beidou positioning system, the timestamps of the SCADA system, inspection equipment, and GIS system are calibrated in real time to correct the reciprocal value of the time deviation. The real-time coordinate data collected by the GIS positioning instrument is combined with the equipment layout diagram of the pipeline cathodic protection management system, and the spatial coordinates of the cathodic protection data are corrected by the coordinate interpolation algorithm. The alignment accuracy of the cathodic protection data is re-acquired based on the preset calibration cycle. Perform semantic correction on the recognized text and correct any erroneous features extracted; Based on the graph database-based association indexing method, the retrieval strategy is optimized according to the time window and spatial radius, and the different types of data in the same spatiotemporal range are automatically matched. Determine whether the alignment accuracy, feature extraction accuracy, and association success rate of the newly acquired CAMP data after calibration of the CAMP data fusion quality are all greater than the corresponding CAMP data standard parameters. If so, construct compliance constraints; otherwise, send a data fusion anomaly alert to the designated personnel.

6. The method for constructing an intelligent agent for a pipeline cathodic protection operation and maintenance decision-making knowledge base as described in claim 5, characterized in that, The specific process for constructing the compliance constraints is as follows: The potential simulation deviation is obtained and compared with the corresponding preset potential simulation deviation to obtain a consistency index. Determine whether the potential simulation deviation is less than or equal to the preset potential simulation deviation; If so, then conduct closed-loop intelligent decision-making and feedback updates based on the thought chain; otherwise, implement physical feasibility optimization to ensure the feasibility of the decision recommendations.

7. The method for constructing an intelligent agent for a pipeline cathodic protection operation and maintenance decision-making knowledge base as described in claim 6, characterized in that, The specific process of optimizing physical feasibility is as follows: Based on the physical mechanism simulation model running on the simulation server, the preliminary decision suggestions generated by the agent are iteratively verified. The initial decision suggestions generated by the intelligent agent are used as input, along with the cathodic protection data obtained from the SCADA system, and then input into the physical mechanism simulation model. Based on the optimized decision recommendations, we conduct closed-loop intelligent decision-making and feedback updates based on the thought chain.

8. The method for constructing an intelligent agent for a pipeline cathodic protection operation and maintenance decision-making knowledge base as described in claim 7, characterized in that, The specific process of closed-loop intelligent decision-making and feedback update based on the thought chain is as follows: The accuracy of fault location, the success rate of decision execution, and the efficiency of knowledge base update are used to determine whether the construction of the intelligent agent of the pipeline cathodic protection operation and maintenance decision-making knowledge base is qualified. The fault location accuracy is expressed as the percentage of cases that perfectly match the actual fault situation, determined by the agent based on closed-loop reasoning, relative to the total number of fault cases it infers. The percentage of the number of operation and maintenance decision instructions output by the intelligent agent that successfully resolve pipeline cathodic protection faults after execution, out of the total number of output instructions, is expressed as the decision execution success rate. The speed at which information is updated to the knowledge base after a decision is implemented is called the knowledge base update efficiency. Compare the fault location accuracy with the corresponding pass / fail standard value, compare the decision execution success rate with the corresponding pass / fail standard value, and compare the knowledge base update efficiency with the corresponding pass / fail standard value. If the fault location accuracy, decision execution success rate, and knowledge base update efficiency all meet the corresponding qualification standards during the preset monitoring period, a qualified prompt for the construction of the pipeline cathodic protection intelligent agent will be sent to the preset personnel. Conversely, the reasoning logic or feedback mechanism should be optimized.

9. The method for constructing an intelligent agent for a pipeline cathodic protection operation and maintenance decision-making knowledge base as described in claim 8, characterized in that, The specific process for optimizing the reasoning logic or feedback mechanism is as follows: Based on the iterative verification of the physical mechanism simulation model, simulation boundary condition constraints are added to generate a preliminary scheme. The preliminary scheme is then subjected to a preset number of simulation verifications until the simulation potential distribution falls completely within the qualified range. After completing the above steps, the closed-loop intelligent decision-making and feedback update based on the thinking chain is re-entered, and the fault location accuracy, decision execution success rate and knowledge base update efficiency are evaluated again in the next monitoring cycle until all three indicators meet the corresponding qualification standards or reach the preset maximum number of optimizations. If the target is still not met after the maximum number of optimization attempts, a prompt indicating that the agent needs to be manually reconstructed will be sent to the designated personnel.

10. The method for constructing an intelligent agent for a pipeline cathodic protection operation and maintenance decision-making knowledge base as described in claim 7, characterized in that, The specific process of closed-loop intelligent decision-making and feedback update based on the thought chain is as follows: The average response time of the cathodic protection management system within a preset coordination period is obtained and expressed as the response delay of each agent. Determine whether the response delay of each agent is less than or equal to a preset response delay threshold; If all system response delays are less than or equal to the preset response delay threshold, the agent can distribute tasks and synchronize states according to the established process. Conversely, optimization of the reasoning logic or feedback mechanism needs to be triggered.

Images