Offshore wind power supply chain anomaly detection method, device and equipment and storage medium

Abstract

The invention discloses an offshore wind power supply chain anomaly detection method and device, equipment and a storage medium, and the method comprises the steps: collecting wind power supply chain data and multi-source verification data in response to a detection instruction for an offshore wind power supply chain; performing structural decomposition on the wind power supply chain data to generate structural data fragments; generating a content data fragment in combination with the multi-source verification data and the structure data fragment; identifying the content data fragment based on a preset rule index to generate initial detection data; performing risk classification based on the initial detection data, and determining a risk category; determining an early warning strategy corresponding to the risk category; and executing the early warning strategy. According to the embodiment of the invention, the analysis granularity of the offshore wind power supply chain can be refined, dynamic monitoring and intelligent early warning of abnormal risks are realized by fusing multi-source data, and the emergency response efficiency is improved.

Description

Technical Field

[0001] This application relates to the field of offshore wind power supply chain risk management technology, and in particular to an offshore wind power supply chain anomaly detection method, an offshore wind power supply chain anomaly detection device, an electronic device, and a computer-readable storage medium. Background Technology

[0002] As the global energy structure shifts towards cleaner energy, offshore wind power, as an important form of renewable energy, is developing rapidly. The offshore wind power supply chain encompasses multiple professional areas, including wind farm planning, turbine development, grid connection and transmission, and operation and maintenance services, involving various risk factors such as raw materials, components, ship equipment, and policy and market conditions. Currently, supply chain risk management methods are mostly concentrated on general models used in traditional manufacturing, lacking a systematic consideration of the unique characteristics of offshore wind power. Most methods remain at the equipment or system level, failing to delve into the micro-level of components, processes, and production capacity, making it difficult to pinpoint specific bottlenecks and leading to delayed early warnings or false alarms. Summary of the Invention

[0003] In view of the above problems, embodiments of this application are proposed to provide a method for detecting anomalies in the offshore wind power supply chain, a device for detecting anomalies in the offshore wind power supply chain, an electronic device, and a computer-readable storage medium to overcome or at least partially solve the above problems.

[0004] To address the aforementioned problems, in the first aspect of this application, an embodiment discloses a method for detecting anomalies in the offshore wind power supply chain, comprising: In response to inspection directives targeting the offshore wind power supply chain, collect wind power supply chain data and multi-source verification data; The wind power supply chain data is structurally decomposed to generate structural data fragments; By combining the multi-source verification data and the structured data fragment, a content data fragment is generated; The content data fragments are identified based on preset rule indicators to generate initial detection data; Risk classification is performed based on the initial detection data to determine the risk category; Determine the early warning strategy corresponding to the risk category; Implement the aforementioned early warning strategy.

[0005] Optionally, the step of structurally decomposing the wind power supply chain data to generate structural data fragments includes: Based on a pre-defined supply chain decomposition model, the wind power supply chain data is structurally decomposed to determine the business links, structural levels, and data attributes corresponding to the wind power supply chain data. The business process, the structural hierarchy, and the data attributes are combined to generate a structured data fragment.

[0006] Optionally, the step of generating a content data fragment by combining the multi-source verification data and the structured data fragment includes: Determine the confidence labels of the multi-source verification data; The confidence labels are associated with the structured data fragments to generate content data fragments.

[0007] Optionally, the preset rule indicators include capacity matching degree, yield indicator, and logistics indicator. The step of identifying the content data fragments based on the preset rule indicators and generating initial detection data includes: The content data fragments are identified based on preset rule indicators, and the target rule indicators corresponding to the content data fragments are determined. The target rule index is invoked to perform matching calculations on the content data fragments to generate initial detection data.

[0008] Optionally, the step of identifying the content data segment based on preset rule indicators and determining the target rule indicator corresponding to the content data segment includes: Identify the business dimensions of the content data fragments; Based on the business dimension, one or more target rule indicators are matched from the preset rule indicators.

[0009] Optionally, the step of classifying risks based on the initial detection data and determining risk categories includes: Based on the initial detection data, identify the risk identification code corresponding to the initial detection data; Based on the risk identification code, the association between the initial detection data and the preset risk decomposition structure is constructed; Based on the aforementioned relationships, risk categories are matched.

[0010] Optionally, the step of determining the early warning strategy corresponding to the risk category includes: The initial detection data is classified according to the risk category to determine the warning level; A corresponding early warning strategy is matched based on the risk category and the early warning level.

[0011] In a second aspect, this application discloses an anomaly detection device for the offshore wind power supply chain, comprising: The data acquisition module is used to collect wind power supply chain data and multi-source verification data in response to testing commands for the offshore wind power supply chain. The generation module is used to perform structural decomposition on the wind power supply chain data and generate structural data fragments. The module is used to combine the multi-source verification data and the structured data fragments to generate content data fragments; The identification module is used to identify the content data fragments based on preset rule indicators and generate initial detection data; The classification module is used to classify risks based on the initial detection data and determine the risk category; The determination module is used to determine the early warning strategy corresponding to the risk category; The execution module is used to execute the warning strategy.

[0012] In a third aspect of this application, embodiments of this application disclose an electronic device, including a processor, a memory, and a computer program stored in the memory and capable of running on the processor. When the computer program is executed by the processor, it implements the offshore wind power supply chain risk anomaly detection method as described above.

[0013] In a fourth aspect of this application, embodiments of this application disclose a computer-readable storage medium storing a computer program that, when executed by a processor, implements the offshore wind power supply chain risk anomaly detection method as described above.

[0014] The embodiments of this application have the following advantages: This application embodiment, in response to a detection command targeting the offshore wind power supply chain, collects wind power supply chain data and multi-source verification data; performs structural decomposition on the wind power supply chain data to generate structural data fragments; combines the multi-source verification data and the structural data fragments to generate content data fragments; identifies the content data fragments based on preset rule indicators to generate initial detection data; classifies risks based on the initial detection data to determine risk categories; determines the early warning strategy corresponding to the risk category; and executes the early warning strategy. This application embodiment can refine the granularity of offshore wind power supply chain analysis based on structural decomposition, integrate multi-source data to achieve dynamic monitoring and intelligent early warning of abnormal risks, and improve emergency response efficiency. Attached Figure Description

[0015] Figure 1 This is a flowchart illustrating the steps of an embodiment of an anomaly detection method for offshore wind power supply chain according to this application; Figure 2 This is a flowchart illustrating the steps of another embodiment of the offshore wind power supply chain anomaly detection method of this application; Figure 3 This is a schematic diagram illustrating an example of confidence level verification in this application; Figure 4 This is a structural block diagram of an embodiment of an anomaly detection device for offshore wind power supply chain according to this application; Figure 5 This is a structural block diagram of an electronic device provided in an embodiment of this application; Figure 6 This is a structural block diagram of a storage medium provided in an embodiment of this application. Detailed Implementation

[0016] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, the application will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0017] Reference Figure 1 The diagram illustrates a flowchart of an embodiment of an anomaly detection method for offshore wind power supply chains according to this application. The anomaly detection for offshore wind power supply chains may specifically include the following steps: Step 101: In response to the inspection command for the offshore wind power supply chain, collect wind power supply chain data and multi-source verification data; When the system receives a detection command for the offshore wind power supply chain, it initiates concurrent data acquisition requests to all data sources related to the target unit based on a pre-configured communication protocol and data source list. This involves collecting wind power supply chain data and multi-source verification data. Specifically, high-concurrency data transfer can be achieved through the Kafka message bus, supporting various industrial and internet communication protocols such as Modbus, OPC-UA (Open Platform Communications – Unified Architecture), HTTP (Hypertext Transfer Protocol), and SFTP (Secure File Transfer Protocol). This includes collecting business data (supply chain data) related to orders, capacity, inventory, production plans, and delivery cycles from internal enterprise and supplier ERP (Enterprise Resource Planning), MES (Manufacturing Execution System), and WMS (Warehouse Management System), as well as from SCADA (Supervisory Control and Data Acquisition) systems related to the target components / equipment via industrial protocols such as Modbus and OPC-UA. The Acquisition (Data Acquisition and Monitoring Control) system collects real-time status and operational data from sensors, construction vessel and machinery operation records, and submarine cable monitoring equipment. It also collects environmental data related to the target unit, such as industry policy texts, raw material market prices, port logistics status, and weather and sea condition forecasts, from external databases through API interfaces or web crawlers. This is known as multi-source verification data.

[0018] Step 102: Decompose the wind power supply chain data into structural data fragments; The system incorporates a decomposition model defined by three-dimensional coordinates. This model is used to perform structural analysis on each piece of raw data collected, generating structured data fragments containing technical directions, supply chain levels, and information elements. The specific decomposition process can employ a 5×N×M three-dimensional decomposition method. The first dimension represents five major technical directions, i.e., P={P1…P5}, such as P1 wind farm planning, design, and construction; P2 wind turbine design and R&D; P3 offshore grid connection and transmission systems; P4 offshore wind power operation and maintenance; and P5 integrated development of offshore wind power. The second dimension represents N=3 levels of supply. The chain layer is L={L1,L2,L3}, such as L1 major category (e.g., main unit of the wind turbine), L2 medium category (e.g., wind turbine yaw system), L3 minor category (e.g., yaw bearing). The third dimension: M=9 types of information elements, namely E={E1…E9}, such as design theory, design standards, R&D tools, production process, production tooling, quality control, capacity, test standards, and test tools. Through the three-dimensional Cartesian product, 5×3×9=135 subdivided units are generated, and each unit is assigned a unique ID (e.g., P2-L3-E4 represents wind turbine - main shaft forging - production process).

[0019] Step 103: Combine the multi-source verification data and the structure data fragment to generate a content data fragment; The system first assigns a basic confidence level to the multi-source verification data based on its collection source. Data directly from official documents or national standards is marked with the highest level (3); data from authoritative third-party research reports or certification bodies is marked with 2; data obtained through expert interviews or enterprise-designated channels is marked with 1; and data automatically scraped from unverified channels such as publicly available online information is marked with the lowest level (0). The system then identifies the multi-source verification data with confidence level labels and adds them to the corresponding structured data segments, forming enhanced content data segments with quantifiable credibility attributes. The confidence level label is used when multiple sources of data describe the same fact (such as the production capacity of a component). The system compares the consistency of the values: if the data are consistent, the data with the highest confidence level label is adopted as the valid value; if there are significant conflicts, the data point is marked as pending review. Furthermore, regardless of whether there are conflicts, if the confidence level label of the main data is <1, the system will automatically trigger a manual review process to ensure that low-confidence data does not directly enter the subsequent analysis stage without review.

[0020] Step 104: Identify the content data fragments based on preset rule indicators to generate initial detection data; The system loads predefined business rule indicators such as capacity matching rules, localization rate rules, and logistics bottleneck rules, and inputs these rules and corresponding content data fragments into the bottleneck diagnosis engine for matching and calculation. This automatically generates a diagnostic card, i.e., initial detection data. For example, for a content data fragment containing the demand forecast for the next 12 months and the current capacity value, the system executes the capacity matching rule, calculating the ratio of demand forecast to current capacity. If this ratio is greater than a preset threshold (e.g., >1.2), the rule is triggered, and the system immediately generates an initial detection data point. This data is a structured diagnostic card that can contain a unique issue ID (issue_id), a description, a list of affected projects (impact_projects), and associated predefined risk codes (risk_ids). The entire process is driven by a rule engine (such as Drools) without manual intervention, achieving automatic transformation from data to preliminary risk clues and providing clear input for subsequent comprehensive risk assessment and early warning.

[0021] Step 105: Based on the initial detection data, perform risk classification to determine the risk category; The system pre-defines an RBS knowledge base consisting of three levels of codes: risk category, risk subcategory, and risk event. For example, code R31 represents the capacity shortage subcategory under the capacity risk category. When initial detection data is generated, the system automatically parses the pre-associated risk codes. Based on these codes, the system precisely maps them to the corresponding categories and subcategories within the six major risk systems (policy, technology, capacity, quality, standards, and environment). For instance, when the initial detection data is {"risk_ids":["R31 capacity shortage", "R11 import restrictions"]}, code R31 is mapped to capacity risk, and R11 is mapped to the trade barrier subcategory under policy risk. This mapping process is achieved by querying the risk event-category relationships stored in the built-in risk classification dictionary or knowledge graph. Specifically, for the spindle forging shortage problem, the output will clearly indicate that the problem involves both capacity risk and policy risk.

[0022] Step 106: Determine the early warning strategy corresponding to the risk category; First, based on the risk classification results, the system can also determine the quantitative indicator value and risk level corresponding to the risk category, and input the above data into the preset Drools rule engine. The rule engine matches the real-time data with the dynamically updated business rules. For example, for a problem that has been classified as a capacity risk, the system will call the capacity matching index indicator and its current dynamic threshold. This threshold can be adaptively adjusted according to technology iteration and market conditions. If the indicator value exceeds the threshold, the corresponding level of early warning strategy will be triggered immediately. Specifically, based on a preset early warning classification mechanism, the system automatically classifies early warnings into four levels, from Level I (red) to Level IV (blue), by combining the early warning probability predicted by the LSTM-CNN model and the expected impact level, such as the number of times the project's core schedule is affected. The engine uses the current risk status (including risk category and level) as the "state space" and calculates the maximum reward function (i.e., the weighted difference between risk mitigation rate and disposal cost) from the preset "action space" (such as joint reserve, technology substitution, and capacity transfer). It iteratively optimizes and recommends the current optimal disposal strategy combination, i.e., the early warning strategy. The early warning strategy generated by the system is a complete instruction package containing a clear risk level, triggering rules, recommended disposal measures, and push targets, realizing an intelligent closed loop from risk identification to disposal suggestions.

[0023] Step 107: Execute the aforementioned early warning strategy.

[0024] Based on the warning strategy's defined levels and preset role-permission matrix, the system sends warning information to relevant responsible persons in parallel through an integrated communication interface. Simultaneously, it searches the emergency response knowledge base, automatically matching historical successful cases or standard operating procedures based on risk category, impact scope, and response strategy. For example, for an orange warning triggered by capacity risk, the system might simultaneously issue instructions to send a warning to the procurement manager and recommend initiating a joint reserve agreement and evaluating a second supplier as the preferred response. It can also create a response tracking task for each triggered warning, continuously monitoring changes in relevant risk indicators and the execution status based on manual feedback. Level I (red) warnings will simultaneously send strong alerts via SMS, platform pop-ups, and API integration to the command center's large screen; while Level IV (blue) warnings may only send a trend report to the planning department via email. The push notifications not only include a risk description but also directly attach a list of recommended response strategies generated by a reinforcement learning engine (e.g., Measure A: Initiate joint reserve agreement; Measure B: Negotiate capacity transfer). The system fully records the key nodes, execution results, and actual effectiveness of the entire handling process. This data is fed back to the data lake to update the risk assessment model. Each early warning not only solves the immediate problem but also continuously enhances the system's early warning and decision-making capabilities through iteration, achieving an evolution from static response to dynamic self-optimization.

[0025] Reference Figure 2 The diagram illustrates a flowchart of another embodiment of the offshore wind power supply chain anomaly detection method of this application. The offshore wind power supply chain anomaly detection may specifically include the following steps: Step 201: In response to the inspection command for the offshore wind power supply chain, collect wind power supply chain data and multi-source verification data; The system receives supply chain testing instructions initiated by users or scheduled tasks. It collects information from multiple dimensions through market research, data collection, and expert consultation. It establishes information collection templates for the entire process of equipment design, R&D, manufacturing, and testing. It collects relevant information from dimensions such as design theory, design standards, R&D tools or software (database), production processes, production tooling, quality control, production capacity, testing standards, and testing tools or testing platforms. This includes collecting business data (supply chain data) on orders, capacity, inventory, production plans, and delivery cycles, as well as real-time status, operational data, and environmental data (multi-source verification data).

[0026] Step 202: Decompose the wind power supply chain data into structural data fragments; The system has a built-in three-dimensional structural decomposition model that uses a 5×3×9 Cartesian product to break down the offshore wind power supply chain into 135 uniquely identifiable sub-units (ID format such as P2-L3-E4). This is used to perform structured parsing of the collected raw data and generate standardized data fragments that include business links, structural levels, and data attributes.

[0027] In an optional embodiment of this application, the step of structurally decomposing the wind power supply chain data to generate structural data fragments includes: Sub-step S2021: Based on the preset supply chain decomposition model, the wind power supply chain data is structurally decomposed to determine the business links, structural levels and data attributes corresponding to the wind power supply chain data. Sub-step S2022: Combine the business process, the structural hierarchy, and the data attributes to generate a structural data fragment.

[0028] Data from multiple wind power supply chains is acquired, and specific components are identified from 135 sub-units. This data is then matched against a pre-defined decomposition model to clarify its technical direction (business process) and supply chain level (structural level). Core objects are defined for each data segment. Based on the matching results, a corresponding three-dimensional unit ID is generated, such as P2-L3-E4, representing wind turbine generator set - main shaft forging - production process. For each field within this framework, specific data is populated from multiple sources, forming a four-dimensional data framework: component × direction × stage × dimension. Specifically, this is illustrated by the forging of the main shaft for a 16MW offshore wind turbine generator set (corresponding unit ID: P2-L3-E4, i.e., P2 wind turbine generator set). Taking the design and R&D, L3 sub-category, and E4 production process as an example, we focused on the specific part, the 16MW forged main shaft, and clarified its business links (P2) and structural levels (L3). According to the entire life cycle of offshore wind power equipment, it was broken down into three stages: design and R&D, production and manufacturing, and testing and inspection. This covered the complete process of the part from technical solution to qualified delivery. Nine types of core information elements, i.e., data attributes, were matched for each stage to form a two-dimensional data collection grid of stage-dimension. The business links and structural levels were combined with the two-dimensional data collection grid to generate corresponding structural data fragments. Using the three-dimensional decomposition method, 135 uniquely identified subdivided units were generated to achieve a refined and structured description of the supply chain.

[0029] Step 203: Combine the multi-source verification data and the structure data fragment to generate a content data fragment; The system assigns a confidence level label of 0 to 3 to each data point (official documents = 3, third-party reports = 2, expert interviews = 1, web crawling = 0) and attaches it to the structured data fragment to form an enhanced content fragment. When multiple sources describe the same fact, the system prioritizes the data with higher confidence. If there is a data conflict, it is marked as pending verification, and the manual verification process is automatically triggered when the confidence level is <1.

[0030] In an optional embodiment of this application, the step of generating a content data fragment by combining the multi-source verification data and the structured data fragment includes: Sub-step S2031: Determine the confidence label of the multi-source verification data; Sub-step S2032: Associate the confidence label with the structured data fragment to generate a content data fragment.

[0031] Data is collected from various sources, including official documents, third-party reports, expert interviews, and online information, forming multi-source verification data. The data is then scored for confidence based on its source, and corresponding confidence labels are determined. These labels are then used to populate structured data fragments, meaning each field is labeled with its corresponding confidence label. Fields with a confidence score <1 are subject to manual review. Data is corrected and labels are updated through methods such as vendor surveys and verification by authoritative institutions, ultimately generating content data fragments. Specifically, initial confidence labels can be assigned to data according to preset mapping rules, such as a confidence label of 3 for official documents and 2 for third-party reports, with a built-in rule engine continuously monitoring.

[0032] Specifically, refer to Figure 3 If any data is found to have an initial confidence label <1, a manual review process is immediately triggered. Reviewers are required to verify the data through methods such as telephone verification and cross-validation of authoritative information. They have the right to correct the data content and update the confidence label based on the verification results. By attaching a confidence label to each field, low-confidence data triggers manual review, ensuring data credibility.

[0033] Step 204: Identify the content data fragments based on preset rule indicators and determine the target rule indicators corresponding to the content data fragments; the preset rule indicators include capacity matching degree, yield indicators, and logistics indicators; After the system receives a content data fragment, the rule engine will scan the key data fields in the fragment like a detector and match them with the preset rule base to determine the target rule indicator triggered by the fragment. That is, the input content data fragment with confidence label is parsed and matched by the rule engine, and the triggered target rule indicator and its judgment result will be packaged into the diagnostic card.

[0034] In an optional embodiment of this application, the step of identifying the content data fragment based on a preset rule indicator and determining the target rule indicator corresponding to the content data fragment includes: Sub-step S2041: Identify the business dimensions of the content data fragment; Sub-step S2042: Based on the business dimension, match one or more corresponding target rule indicators from the preset rule indicators.

[0035] The system determines which business dimension the data segment to be analyzed belongs to within the supply chain, such as the supply of raw materials and components, localization status, spare parts inventory, lagging demonstration applications, construction and maintenance of marine machinery and equipment, and policy and market environment. Based on this business dimension, it automatically identifies and judges each link in the supply chain, accurately pinpoints bottlenecks and problems, and outputs standardized diagnostic results including the values ​​of indicators for each diagnostic dimension, judgment results, and problem levels. Specifically, for a data segment about a 16MW forging main shaft, the system parses the unit_id (unit ID) and the meaning of its fields. 2-L3-E4 clearly points to the production and manufacturing information of a specific component (spindle). According to the preset mapping relationship, the production and manufacturing capacity information of the specific component belongs to the assessment of supply capacity. Therefore, the system automatically classifies it into the business dimension of raw material and component supply. After determining the business dimension, it further selects the specific diagnostic rules applicable to that dimension. The raw material and component supply dimension corresponds to the capacity matching rule, and the localization dimension corresponds to the yield indicator rule. By executing the corresponding quantitative rules, the problems are accurately exposed, providing a solid and objective evidence chain starting point for the final risk warning.

[0036] Step 205: Call the target rule index to perform matching calculations on the content data fragments to generate initial detection data; For content data segments with confidence labels (such as production capacity and demand data for spindle forgings) and corresponding target rule indicators such as production capacity matching degree rules, the rule engine is invoked to execute preset algorithms for calculation and logical judgment, outputting initial detection data. The most typical form of this is a standardized diagnostic card. Specifically, the system extracts the specific field values ​​necessary for executing the target rule indicator from the content data segment. Taking spindle forgings as an example, the target rule indicator is the production capacity matching degree rule. From the data segment, the monthly production capacity is extracted as 120 pieces, and the project plan or market forecast data for the next 12 months' demand is extracted as 180 pieces. The system will simultaneously read the confidence labels of these fields. For example, if the demand forecast data comes from web scraping (confidence = 0), the system may mark or downgrade the reliability of the result in subsequent calculations or alerts. Then, it will call the calculation logic, formula, or decision tree bound to the target rule indicator, and use the extracted data to perform calculations. If the ratio of the demand forecast value for the next 12 months to the existing capacity is greater than a given threshold, it will be marked as a capacity bottleneck. Based on the calculation results and the decision results, the system will automatically generate a structured initial detection data containing the core information of the problem, i.e., a diagnostic card, in the following form: {"issue_id": "ISS-2025-0001","description": "Main shaft forgings demand shortfall of 33% in the next 12 months","impact_projects": ["A Offshore Wind Farm","B Offshore Wind Farm"],"risk_ids": ["R31 capacity shortage","R11 import restrictions"]} It can be seen that the data fragment matched and calculated the detection results of insufficient production capacity and import restrictions. A complex supply chain problem may be revealed by multiple such initial detection data from different perspectives, accurately describing where the problem occurred, how serious the problem is, who was affected, and what type of risk it belongs to. This provides the most basic and reliable data for the subsequent construction of a complete supply chain risk situation map.

[0037] Step 206: Based on the initial detection data, perform risk classification to determine the risk category; The initial detection data generated by the bottleneck diagnostic engine (such as the diagnostic card "33% demand gap for spindle forgings in the next 12 months") is automatically matched with the preset three-level risk decomposition structure (RBS / RiskBreakdown Structure) of risk category-risk subcategory-risk event. This allows the data point to be classified into a specific category (such as capacity risk) within the six major risk systems of policy, technology, capacity, quality, standards, and environment, providing a classification basis for subsequent quantitative assessment and early warning.

[0038] In an optional embodiment of this application, the step of classifying risks based on the initial detection data and determining the risk category includes: Sub-step S2061: Based on the initial detection data, identify the risk identification code corresponding to the initial detection data; Sub-step S2062: Based on the risk identification code, construct the association between the initial detection data and the preset risk decomposition structure; Sub-step S2063: Based on the aforementioned association, match risk categories.

[0039] When diagnosing problems, the system's built-in rule engine automatically matches the most relevant risk event codes from the RBS knowledge base to determine the risk category, based on the nature of the problem (e.g., supply shortage, technical failure, or policy change) and the elements involved (e.g., specific parts, technological directions). Specifically, it can pre-build a risk decomposition structure using the Delphi method, employing a three-level coding system of risk category - risk subcategory - risk event. This system clearly defines six major risk categories as primary classifications: policy (R1), technology (R2), capacity (R3), quality (R4), standards (R5), and environment (R6). Each major category is further subdivided into specific risk events, each with a unique code (e.g., R31 represents insufficient capacity, and R1101 represents a 25% tariff imposed by Europe and the US on wind power equipment).

[0040] The specific process of risk identification based on the Delphi method and RBS includes: 1. Expert selection Establish an expert think tank for the offshore wind power industry, comprising research institutes, design institutes, turbine manufacturers, component manufacturers, universities, construction companies, operation and maintenance companies, construction companies, and regulatory agencies. Adhering to the principles of comprehensive coverage, multi-level expertise, and experience matching, this ensures that the expert team covers all aspects of the offshore wind power supply chain while also incorporating both theoretical research and practical experience, guaranteeing the comprehensiveness and professionalism of risk identification.

[0041] Through industry association recommendations and targeted invitations from enterprises, 80 qualified candidate experts were selected and formal invitation letters were sent to them, including their research background, research objectives, and time commitments. Ultimately, 50 experts were confirmed to join the think tank (response rate ≥ 60%).

[0042] Expert Grouping: Experts are divided into 5 working groups (corresponding to technical directions P1-P5) according to their professional fields. Each group has a group leader who is responsible for coordinating and communicating opinions within the group.

[0043] 2. Three-wheeled Delphi Each round of the questionnaire uses a Likert 5-point scale for scoring. A coefficient of variation (CV) < 0.25 is considered convergent. The core logic is anonymous feedback, multiple iterations, and statistical convergence. The importance of risk points is quantified using the Likert 5-point scale, ultimately forming a unified risk list. The specific process is as follows: Questionnaire Design: Develop an open-ended questionnaire that includes two core questions based on your field of expertise: list the potential risks in the offshore wind power supply chain and the possible impact of each risk. Also include a diagnostic card for bottlenecks and preliminary test data as reference materials.

[0044] Anonymous distribution and collection: Questionnaires were distributed to 50 experts through an encrypted questionnaire platform, with a 7-day collection period. A total of 46 valid questionnaires were collected (a collection rate of 92%).

[0045] Summary of feedback: The collected questionnaires were coded and deduplicated, and a total of 128 initial risk points in 6 categories, namely policy, technology, production capacity, quality, standards, and environment, were identified, forming the first round of risk point list.

[0046] Scale Design: Based on the initial list, a Likert 5-point questionnaire was designed (1 point = minimal impact, 2 points = minor impact, 3 points = moderate impact, 4 points = significant impact, 5 points = extremely significant impact), requiring experts to rate the probability of occurrence and the degree of impact of each risk point.

[0047] Statistical Analysis: 42 valid questionnaires were collected. The mean and coefficient of variation (CV) of each risk point score were calculated (CV = standard deviation / mean, reflecting the dispersion of opinions). For example, the probability of the occurrence of the risk of the US and Europe imposing a 25% tariff on wind power equipment was 4.2 (CV = 0.32), and the impact was 4.5 (CV = 0.28). Since CV > 0.25, it needs to proceed to the final round of survey.

[0048] Feedback: The scoring results and dispersion of each risk point will be anonymously provided to the experts, who will be invited to re-examine their own scores in conjunction with the overall feedback, providing a reference for the final round of research.

[0049] Questionnaire optimization: For the 36 risk points with a CV > 0.25 in the second round, the latest industry data (such as the 2024 wind power equipment import tariff policy and spindle forging capacity data) were added, and the scoring questionnaire was distributed again.

[0050] Convergence Criteria: 40 valid questionnaires were collected, and the calculated CVs for all risk points were <0.25, meeting the convergence criteria. For example, the probability CV of the risk of Europe and the United States imposing a 25% tariff on wind power equipment decreased to 0.18, and the impact CV decreased to 0.20, with experts reaching a consensus.

[0051] A final list was compiled: 89 risk points with an average score of ≥3 were selected as the core risk pool for subsequent RBS construction.

[0052] 3. Risk Breakdown Structure (RBS) The risk is coded in three levels: major risk category - sub-risk category - risk event. For example: R1 policy risk, R11 trade barriers, R1101 25% tariff imposed by the US and Europe on wind power equipment. The hierarchy is defined as follows: Level 1 coding (R + number): represents the major risk category, corresponding to 6 major risk systems (e.g., R1 = policy risk, R2 = technology risk). Secondary code (major category code + two digits): represents the risk subclass, which is a sub-area of ​​major risk (e.g., R11 = policy risk - trade barriers). Level 3 coding (subclass code + three digits): represents a specific risk event and is a concrete representation of the risk that can be implemented (e.g., R1101 = 25% tariffs imposed by the US and Europe on trade barriers).

[0053] Step 207: Determine the early warning strategy corresponding to the risk category; The system uses the Drools rule engine to match real-time risk indicator values ​​with dynamic thresholds to trigger warnings. It automatically classifies warning levels I-IV based on the warning probability and impact level predicted by the LSTM-CNN model, and uses reinforcement learning (PPO algorithm) to iteratively optimize a complete warning strategy package that includes risk level, triggering rules, recommended handling measures, and push targets, thus realizing an intelligent closed loop from risk identification to handling suggestions.

[0054] In an optional embodiment of this application, the step of determining the early warning strategy corresponding to the risk category includes: Sub-step S2071: Classify the initial detection data based on the risk category to determine the warning level; Sub-step S2072: Match the corresponding early warning strategy based on the risk category and the early warning level.

[0055] Based on the risk identification results, a risk early warning indicator system is established, and reasonable thresholds are set for each early warning indicator, as follows:

[0056] Table 1 The Drools rule engine has a built-in dedicated mapping library for risk categories, dynamic thresholds, and triggering rules. Different risk categories correspond to independent threshold standards and triggering logic, enabling millisecond-level accurate early warnings and differentiated threshold matching. For example, for R3 capacity risk, the trigger threshold is bound to the capacity matching index (threshold ≤ 0.7); for R1 policy risk, the trigger threshold is bound to the policy stability index (threshold < 6 points); for R6 environmental risk (extreme sea state), the trigger threshold is bound to the meteorological warning level (threshold ≥ typhoon orange warning). Thresholds for different categories are not interchangeable. Dedicated rule matching: The rule engine calls customized triggering rules based on the risk category. For example, if the LSTM-CNN model predicts a 70% probability of R3 capacity risk (entering the orange warning range), and matches a rule that indicates a capacity gap > 30% and affects core projects, an early warning is triggered immediately. If the predicted probability of R1 policy risk is 70%, an early warning will only be triggered if the policy change involves tariffs and covers core suppliers, thus avoiding false alarms.

[0057] The system is based on two dimensions: risk category and warning level. It utilizes the Drools rule engine's built-in channel matching rule base for automatic decision-making and push notifications. Simultaneously, it pushes warning information by role (project manager, procurement manager, operations and maintenance dispatcher) through four methods: SMS, email, platform pop-ups, and API integration, ensuring zero omission of critical information. For example, when the spindle forging capacity shortage is determined to be a Level I (red) warning, a mandatory push is triggered across all channels. The system will send an SMS to the project manager: "[Level I Warning] Spindle forging capacity shortage 33%, affecting A / B wind farms. Immediately activate the joint reserve and backup plan." An email containing detailed data attachments and a supplier list will be sent to the procurement manager. A platform pop-up will be pushed to the operations and maintenance dispatcher, displaying a real-time risk dashboard. The system will automatically notify the supplier's ERP system via API integration, locking in emergency stock orders. For a Level II (orange) warning: email + platform pop-up + API integration will be used. The system combines various methods to avoid information overload, including canceling SMS alerts. Level III (yellow) alerts are only pushed to relevant responsible persons via platform pop-ups, while Level IV (blue) alerts are only connected to the backend monitoring system via API for data recording and are not actively pushed to individuals. By distributing alert instructions to various push service modules, accurate alerts are provided to ensure timeliness in high-concurrency scenarios and the effective execution of alert strategies.

[0058] Step 208: Execute the aforementioned early warning strategy.

[0059] The system pushes early warning information in parallel through an integrated communication interface based on the warning level and role permission matrix: Level I (red) triggers strong alerts via SMS, pop-ups, and API multi-channels; Level IV (blue) only sends a trend report via email. The push content includes a risk description and a list of response strategies generated by a reinforcement learning engine (such as activating joint reserve requirements or negotiating capacity transfer), while automatically searching the emergency plan knowledge base to match historical cases or standard operating procedures. The system creates a response tracking task for each early warning, continuously monitors changes in risk indicators and manual execution status, and feeds key nodes, execution results, and effectiveness data back to the data lake to update the risk assessment model, achieving a closed-loop evolution from static response to dynamic self-optimization.

[0060] This application's embodiment uses Cartesian product partitioning to generate 135 uniquely identifiable sub-units, establishing a data unit of part × direction × stage × dimension. Each field is labeled with a confidence level, and low-confidence data triggers manual review. Based on a preset business rule base, structured data fragments are automatically scanned and calculated, outputting standardized detection data, i.e., diagnostic cards, to accurately locate problems and associate them with potential risk codes. Combining the Delphi method and risk decomposition structure, the diagnostic cards are automatically mapped to RBS, achieving structured classification from specific problems to risk categories. A dynamically updated quantitative indicator system is established for each type of risk, with thresholds adaptively adjusted according to technological iterations and market changes. This enables risk analysis to penetrate from the equipment level to the part / process level, providing a high-quality, auditable data foundation for risk assessment and early warning, and reducing false alarm rates.

[0061] It should be noted that, for the sake of simplicity, the method embodiments are all described as a series of actions. However, those skilled in the art should understand that the embodiments of this application are not limited to the described order of actions, because according to the embodiments of this application, some steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also understand that the embodiments described in the specification are all preferred embodiments, and the actions involved are not necessarily required by the embodiments of this application.

[0062] Reference Figure 4 The diagram shows a structural block diagram of an embodiment of an offshore wind power supply chain anomaly detection device according to this application. The offshore wind power supply chain anomaly detection device may specifically include the following modules: The data acquisition module 401 is used to collect wind power supply chain data and multi-source verification data in response to the detection command for the offshore wind power supply chain. The generation module 402 is used to perform structural decomposition on the wind power supply chain data and generate structural data fragments; Combined module 403 is used to combine the multi-source verification data and the structure data fragment to generate content data fragment; The identification module 404 is used to identify the content data fragments based on preset rule indicators and generate initial detection data; Classification module 405 is used to classify risks based on the initial detection data and determine the risk category; The determination module 406 is used to determine the early warning strategy corresponding to the risk category; The execution module 407 is used to execute the warning strategy.

[0063] In an optional embodiment of this application, the generation module 402 includes: The first generation submodule is used to perform structural decomposition of the wind power supply chain data based on a preset supply chain decomposition model, and to determine the business links, structural levels and data attributes corresponding to the wind power supply chain data. The second generation submodule is used to combine the business process, the structural hierarchy, and the data attributes to generate a structured data fragment.

[0064] In an optional embodiment of this application, the bonding module 403 includes: The first combining submodule is used to determine the confidence label of the multi-source verification data; The second combining submodule is used to associate the confidence label with the structured data fragment to generate a content data fragment.

[0065] In an optional embodiment of this application, the preset rule indicators include capacity matching degree, yield indicator, and logistics indicator, and the identification module 404 includes: The first identification submodule is used to identify the content data segment based on a preset rule indicator and determine the target rule indicator corresponding to the content data segment. The second identification submodule is used to call the target rule index to perform matching calculations on the content data fragments and generate initial detection data.

[0066] In an optional embodiment of this application, the first identification submodule includes: The first identification unit is used to identify the business dimension of the content data fragment; The second identification unit is used to match one or more target rule indicators from the preset rule indicators based on the business dimension.

[0067] In an optional embodiment of this application, the classification module 405 includes: The first classification submodule is used to identify the risk identification code corresponding to the initial detection data based on the initial detection data; The second classification submodule is used to construct the association between the initial detection data and the preset risk decomposition structure based on the risk identification code; The third classification submodule is used to match risk categories based on the aforementioned association.

[0068] In an optional embodiment of this application, the determining module 406 includes: The first determining submodule is used to classify the initial detection data based on the risk category and determine the warning level; The second determining submodule is used to match the corresponding early warning strategy based on the risk category and the early warning level.

[0069] As the device embodiment is basically similar to the method embodiment, the description is relatively simple, and relevant parts can be found in the description of the method embodiment.

[0070] Reference Figure 5 This application also provides an electronic device, including: The processor 501 and the storage medium 502 store a computer program executable by the processor 501, which executes the computer program to implement the offshore wind power supply chain anomaly detection method as described in any of the embodiments of this application.

[0071] The memory may include random access memory (RAM) or non-volatile memory, such as at least one disk storage device. Optionally, the memory may also be at least one storage device located remotely from the aforementioned processor.

[0072] The processors mentioned above can be general-purpose processors, including central processing units (CPUs), network processors (NPs), etc.; they can also be digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.

[0073] Reference Figure 6 This application also provides a computer-readable storage medium 601, on which a computer program is stored. When the computer program is run by a processor, it executes the offshore wind power supply chain anomaly detection method as described in any one of the embodiments of this application.

[0074] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. The same or similar parts between the various embodiments can be referred to each other.

[0075] Those skilled in the art will understand that embodiments of this application can be provided as methods, apparatus, or computer program products. Therefore, embodiments of this application can take the form of entirely hardware embodiments, entirely software embodiments, or embodiments combining software and hardware aspects. Furthermore, embodiments of this application can take the form of computer program products implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0076] This application describes embodiments with reference to flowchart illustrations and / or block diagrams of methods, terminal devices (systems), and computer program products according to embodiments of this application. It should be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing terminal device to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing terminal device, generate instructions for implementing the flowchart illustrations. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0077] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing terminal device to operate in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0078] These computer program instructions can also be loaded onto a computer or other programmable data processing terminal equipment, causing a series of operational steps to be performed on the computer or other programmable terminal equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable terminal equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0079] Although preferred embodiments of the present application have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of the embodiments of the present application.

[0080] Finally, it should be noted that in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or terminal device that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or terminal device. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or terminal device that includes said element.

[0081] The above provides a detailed description of the offshore wind power supply chain anomaly detection method, the offshore wind power supply chain anomaly detection device, the electronic device, and the computer-readable storage medium provided in this application. Specific examples have been used to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the method and its core ideas. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of this application. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A method for detecting anomalies in an offshore wind power supply chain, characterized in that, include: In response to inspection directives targeting the offshore wind power supply chain, collect wind power supply chain data and multi-source verification data; The wind power supply chain data is structurally decomposed to generate structural data fragments; By combining the multi-source verification data and the structured data fragment, a content data fragment is generated; The content data fragments are identified based on preset rule indicators to generate initial detection data; Risk classification is performed based on the initial detection data to determine the risk category; Determine the early warning strategy corresponding to the risk category; Implement the aforementioned early warning strategy.

2. The method according to claim 1, characterized in that, The step of structurally decomposing the wind power supply chain data to generate structural data fragments includes: Based on a pre-defined supply chain decomposition model, the wind power supply chain data is structurally decomposed to determine the business links, structural levels, and data attributes corresponding to the wind power supply chain data. The business process, the structural hierarchy, and the data attributes are combined to generate a structured data fragment.

3. The method according to claim 1, characterized in that, The step of generating a content data fragment by combining the multi-source verification data and the structured data fragment includes: Determine the confidence labels of the multi-source verification data; The confidence labels are associated with the structured data fragments to generate content data fragments.

4. The method according to claim 1, characterized in that, The preset rule indicators include capacity matching degree, yield indicator, and logistics indicator. The step of identifying the content data fragments based on the preset rule indicators and generating initial detection data includes: The content data fragments are identified based on preset rule indicators, and the target rule indicators corresponding to the content data fragments are determined. The target rule index is invoked to perform matching calculations on the content data fragments to generate initial detection data.

5. The method according to claim 4, characterized in that, The step of identifying the content data fragment based on preset rule indicators and determining the target rule indicator corresponding to the content data fragment includes: Identify the business dimensions of the content data fragments; Based on the business dimension, one or more target rule indicators are matched from the preset rule indicators.

6. The method according to claim 1, characterized in that, The step of classifying risks based on the initial detection data and determining risk categories includes: Based on the initial detection data, identify the risk identification code corresponding to the initial detection data; Based on the risk identification code, the association between the initial detection data and the preset risk decomposition structure is constructed; Based on the aforementioned relationships, risk categories are matched.

7. The method according to claim 1, characterized in that, The steps for determining the early warning strategy corresponding to the risk category include: The initial detection data is classified according to the risk category to determine the warning level; A corresponding early warning strategy is matched based on the risk category and the early warning level.

8. An anomaly detection device for an offshore wind power supply chain, characterized in that, include: The data acquisition module is used to collect wind power supply chain data and multi-source verification data in response to testing commands for the offshore wind power supply chain. The generation module is used to perform structural decomposition on the wind power supply chain data and generate structural data fragments. The module is used to combine the multi-source verification data and the structured data fragments to generate content data fragments; The identification module is used to identify the content data fragments based on preset rule indicators and generate initial detection data; The classification module is used to classify risks based on the initial detection data and determine the risk category; The determination module is used to determine the early warning strategy corresponding to the risk category; The execution module is used to execute the warning strategy.

9. An electronic device, characterized in that, It includes a processor, a memory, and a program or instructions stored in the memory and executable on the processor, wherein the program or instructions, when executed by the processor, implement the steps of the offshore wind power supply chain risk anomaly detection method as described in claims 1-7.

10. A readable storage medium, characterized in that, The readable storage medium stores a program or instructions, which, when executed by a processor, implement the steps of the offshore wind power supply chain risk anomaly detection method as described in claims 1-7.

Images